REVIEWS in MINERALOGY & GEOCHEMISTRY Volume73
geochemical society
SULFUR IN MAGMAS AND MELTS: Its Importancefor Natural and Technical Processes EDITORS: Harald Behrens & James D. Webster
~
J
!
i Z460
z....o
.noo
nzo
U40
X..,.yEllcltatiOnE~y[e.V]
MINERALOGICAL SOCIETY OF AMERICA GEOCHEMICAL SOCIETY Series Editor: Jodi J. Rosso 2011
ISSN 1529-6466
RiMG Volume 73
Sulfur in Magmas and Melts: Its Importance for Natural and Technical Processes CONTENTS 1-8 Studies of Sulfur in Melts - Motivations and Overview
Behrens & Webster
Analytical and Spectroscopic Methods 9-39 41-78
AnaJytical Methods for Sulfur Determination...
Ripley et at.
Spectroscopic Studies on Sulfur Speciation in Glasses
Wilke et at.
Physical and Chemical Properties ofS-Bearing Silicate Melts 79-111
Diffusion and Redox Reactions of Sulfur in Silicate Melts
113-141
Role of Sulfur in Coloring & Melting Kinetics in Industrial Glass
143-165
Experimental Studies on Sulfur Solubility in Silicate Melts...
167-213
Modeling the Solubility of Sulfur in Magmas...
Behrens & Stelling Falcone et at. Backnaes & Deubener Baker & Moretti
Constraints from Natural and Experimental Systems 215-246
Sulfur Budget in Magmas...
247-283
Distribution of Sulfur Between Melt and Fluid...
285-314
Sulfur-bearing Magmatic Accessory Minerals
315-3 36
Sulfur in Extraterrestrial Bodies & the Deep Earth
337-361
Fining of Glass Melts
Wallace & Edmonds Webster & Botchamikov Parat et at. Ebel Muller-Simon
Natural and Technical Applications 363-421
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts & Magmas
423-492
Sulfur Degassing from Volcanoes...
493-551
Interactions Between Metal & Slag Melts: Steel Desulfurization
513-578
Role of Magmatic Sulfur in Ore Deposit Formation
Marini eta/. Oppenheimer et at. Lehmann & Nadif Simon & Ripley
9 780939 95 087 4
>
REVIEWS in MINERALOGY and GEOchEMIStRY Volume 73
2011
Sulfur in Magmas and Melts:
Its Importance for Natural and Technical Processes EDITORS Harald Behrens James D. Webster
University of Hannover Hannover, Germany American Museum of Natural History New York, U.S.A.
On the COver: Top Left: Image from Kawah Ijen volcano, Java, Indonesia in 1998. It shows sulfur being collected from a large fumarole within the crater for commerical use (supplied by J.D.Webster, New York USA). Top Right: View into a side fired glass melting furnace under operation (copyright by HVG-DGG, Offenbach am Main, Germany). Bottom Left: Microscope photo of an H2O-S bearing andesitic glass obtained by a decompression experiment at 1030 °C with pressure release from 4 kbar to 0.7 kbar (supplied by A. Fiege, Hannover, Germany). Bottom Right: S K-edge XANES spectrum of a fluid-saturated Etna basalt synthesized at 200 MPa, 1050 °C, log(fO2/bar) = –8.3 (supplied by M. Wilke, Potsdam, Germany).
Series Editor: Jodi J. Rosso MINERALOGIcAL SOcIEtY of AMERIcA GEOchEMIcAL SOcIEtY
Reviews in Mineralogy and Geochemistry, Volume 73 Sulfur in Magmas and Melts: Its Importance for Natural and technical Processes ISSN 1529-6466 ISBN 978-0-939950-87-4
Copyright 2011
The MINERALOGICAL SOCIETY of AMERICA 3635 Concorde Parkway, Suite 500 Chantilly, Virginia, 20151-1125, U.S.A. www.minsocam.org The appearance of the code at the bottom of the first page of each chapter in this volume indicates the copyright owner’s consent that copies of the article can be made for personal use or internal use or for the personal use or internal use of specific clients, provided the original publication is cited. The consent is given on the condition, however, that the copier pay the stated per-copy fee through the Copyright Clearance Center, Inc. for copying beyond that permitted by Sections 107 or 108 of the U.S. Copyright Law. This consent does not extend to other types of copying for general distribution, for advertising or promotional purposes, for creating new collective works, or for resale. For permission to reprint entire articles in these cases and the like, consult the Administrator of the Mineralogical Society of America as to the royalty due to the Society.
Sulfur in Magmas and Melts: Its Importance for Natural and Technical Processes
73
Reviews in Mineralogy and Geochemistry
73
FrOm the SerieS editOr The chapters in this volume represent a compilation of the material presented by the invited speakers at a short course on August 21-23, 2011 called “Sulfur in Magmas and Melts and its Importance for Natural and Technical Processes.” This Mineralogical Society of America and the Geochemical Society sponsored short course was held at the Hotel der Achtermann, in Goslar, Germany following the 2011 Goldschmidt Conference in Prague, Czech Republic. Following a nice overview in chapter 1 by the organizers Harald Behrens and James Webster, this volume is divided into 4 parts. (1) Analytical and Spectroscopic Methods -- chapters 2 and 3 (2) Physical and Chemical Properties of S-Bearing Silicate Melts -- chapters 4-7 (3) Constraints from Natural and Experimental Systems -- chapters 8-11 (4) Natural and Technical Applications -- chapters 12-16 Any supplemental materials associated with this volume, such as the supplemental tables compiled by Marini et al. (chapter 14), can be found at the MSA website, www. minsocam.org/MSA/RIM. Errata will also be posted there. The reader will also be able to find links to paper and electronic copies of this and other RiMG volumes. Jodi J. Rosso, Series Editor West Richland, Washington June 2011
1529-6466/11/0073-0000$05.00
DOI: 10.2138/rmg.2011.73.0
Sulfur in Magmas and Melts: Its Importance for Natural and Technical Processes
73
Reviews in Mineralogy and Geochemistry
73
TABLE OF CONTENTS
1
Studies of Sulfur in Melts – Motivations and Overview Harald Behrens, James D. Webster
INTRODUCTION ....................................................................................................................1 Background....................................................................................................................1 The behavior and importance of sulfur in melts ............................................................2 REVIEW CHAPTERS IN THIS VOLUME .............................................................................2 REMAINING ISSUES AND CONSIDERATIONS FOR FUTURE RESEARCH .................6 ACKNOWLEDGMENTS.........................................................................................................7 REFERENCES .........................................................................................................................8
2
Analytical Methods for Sulfur Determination in Glasses, Rocks, Minerals and Fluid Inclusions Edward M. Ripley, Chusi Li, Craig H. Moore Erika R. Elswick, J. Barry Maynard, Rick L. Paul Paul Sylvester, Jun Hun Seo, Nobomichi Shimizu
INTRODUCTION ....................................................................................................................9 ANALYTICAL TECHNIQUES .............................................................................................10 Total sulfur concentration in minerals, rocks and glasses using elemental analyzer – infrared absorption technology........................................10 Total sulfur concentration in minerals, rocks and glasses using elemental analyzer – mass spectrometer technology.........................................11 Total sulfur concentration of minerals, rocks and glasses using X-ray fluorescence .............................................................................................13 “Kiba” method for the determination of sulfur concentration in whole rocks and glasses ..............................................................................................15 Nuclear methods (activation analysis) for the determination of sulfur concentration in minerals, rocks and glasses ..........................................16 Sulfur analysis of minerals and glasses using the electron microprobe ......................19 Analyses of the sulfur concentration of minerals and glasses by secondary ion mass spectrometry (SIMS) .........................................................20 Analyses of the sulfur concentration of minerals and glasses by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) .........................22 iv
Sulfur in Magmas and Melts ‒ Table of Contents Determination of the sulfur concentration in fluid inclusions by LA-ICP-MS ..........31 SUMMARY ............................................................................................................................32 ACKNOWLEDGMENTS.......................................................................................................35 REFERENCES .......................................................................................................................35
3
Spectroscopic Studies on Sulfur Speciation in Synthetic and Natural Glasses Max Wilke, Kevin Klimm, Simon C. Kohn
INTRODUCTION ..................................................................................................................41 X-RAY EMISSION AND ABSORPTION SPECTROSCOPY ..............................................42 S Ka X-ray emission ...................................................................................................42 Determination of the S oxidation state using EPMA ..................................................45 X-ray excited high-resolution X-ray emission spectroscopy ......................................46 X-ray absorption spectroscopy ....................................................................................46 Determination of S oxidation state from XANES .......................................................54 Beam damage during Analysis by EPMA and XANES ..............................................55 Sulfur redox equilibrium determined in glasses by l(S Ka) and XANES ..................57 NUCLEAR MAGNETIC RESONANCE ...............................................................................60 33 S NMR of solid model compounds. ..........................................................................61 33 S NMR of glasses ......................................................................................................64 RAMAN AND IR SPECTROSCOPY ....................................................................................66 Raman spectroscopy of sulfur model compounds .......................................................66 Experimental details for Raman spectroscopy on glasses ...........................................67 Raman spectroscopy on sulfur in glasses ....................................................................69 Beam damage by Raman Spectroscopy ......................................................................71 Determination of the oxidation state using Raman spectroscopy ...............................71 SUMMARY AND OUTLOOK ..............................................................................................72 ACKNOWLEDGEMENTS ....................................................................................................73 REFERENCES .......................................................................................................................73
4
Diffusion and Redox Reactions of Sulfur in Silicate Melts Harald Behrens, Jan Stelling
INTRODUCTION ..................................................................................................................79 SULFUR DIFFUSION STUDIES ..........................................................................................80 Silica glass ...................................................................................................................82 Simple silicate glasses .................................................................................................82 Borosilicate glasses .....................................................................................................89 Aluminosilicate melts relevant to magmatic systems..................................................90 COMPARISON TO DIFFUSION OF OTHER VOLATILES ................................................92 EFFECT OF REDOX STATE ON SULFUR DIFFUSION ....................................................95 v
Sulfur in Magmas and Melts ‒ Table of Contents SULFIDE/SULFATE INTERDIFFUSION AND REDOX REACTIONS OF SULFUR .......96 Redox reactions in dry melts .......................................................................................96 Redox reactions in hydrous melts................................................................................98 SULFUR DIFFUSION VERSUS VISCOSITY .....................................................................99 SUMMARY AND OUTLOOK ............................................................................................102 ACKNOWLEDGMENTS.....................................................................................................103 RERERENCES .....................................................................................................................103
5
The Role of Sulfur Compounds in Coloring and Melting Kinetics of Industrial Glass Roberto Falcone, Stefano Ceola, Antonio Daneo, Stefano Maurina
INTRODUCTION ................................................................................................................113 SULFUR COMPOUNDS IN INDUSTRIAL GLASS PRODUCTION...............................114 Sodium sulfate ...........................................................................................................114 Slag ............................................................................................................................114 Glass cullet ................................................................................................................114 Filter dust ...................................................................................................................114 Pyrite .........................................................................................................................115 Others ........................................................................................................................115 INDUSTRIAL GLASS PRODUCTION ..............................................................................115 Batch preparation.......................................................................................................116 Melting ......................................................................................................................117 Batch reactions ..........................................................................................................117 Fining.........................................................................................................................121 Forming and post-forming .........................................................................................123 OPTICAL PROPERTIES AND COLORS OF INDUSTRIAL SLS CONTAINER GLASS ....................................................................................................124 COLOR GENERATION IN SLS GLASSES .......................................................................126 SULFUR SOLUBILITY, REDOX AND GLASS COLOR ..................................................128 The batch redox number ............................................................................................129 Experimental melting ................................................................................................131 CHEMICAL CHARACTERIzATION OF GLASS COLORS ............................................132 BUBBLES IN SLS GLASSES .............................................................................................134 Incomplete fining .......................................................................................................135 Redox.........................................................................................................................135 Reboil ........................................................................................................................135 Deposits .....................................................................................................................136 SUMMARY AND OUTLOOK ............................................................................................138 ACKNOWLEDGMENTS.....................................................................................................138 REFERENCES .....................................................................................................................139
vi
Sulfur in Magmas and Melts ‒ Table of Contents
6
Experimental Studies on Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure Linda Backnaes, Joachim Deubener
INTRODUCTION ................................................................................................................143 ANALYSIS METHODS FOR SULFUR CONTENT ..........................................................144 INDUSTRIAL MELTS .........................................................................................................144 Effect of oxygen fugacity ..........................................................................................144 Effect of melt temperature .........................................................................................149 Effect of melt composition ........................................................................................150 SOLUBILITY OF SULFUR-BEARING WASTE IN MELTS ............................................155 SOLUBILITY OF SULFUR IN NATURAL MELTS AND SYNTHETIC ANALOGS......157 SUMMARY AND OUTLOOK ............................................................................................160 AKNOWLEDGMENT .........................................................................................................161 REFERENCES .....................................................................................................................161
7
Modeling the Solubility of Sulfur in Magmas: A 50-Year Old Geochemical Challenge Don R. Baker, Roberto Moretti
INTRODUCTION ................................................................................................................167 THERMODYNAMICS AND MODELING SULFUR SOLUBILITY IN MAGMATIC SYSTEMS ...........................................................................................168 A few necessary definitions and concepts .................................................................168 Modeling philosophies ..............................................................................................169 Thermodynamic basis for dissolution of sulfur in silicate melts...............................171 A BRIEF HISTORY OF EXPERIMENTS AND EMPIRICAL MODELS FOR SULFUR SOLUBILITY IN SILICATE MELTS............................................................175 Sulfur behavior in systems with only sulfur-rich gas and silicate melts at 1 atm .....175 SULFIDE AND SULFATE SATURATION IN SILICATE MELTS .................................178 Sulfide saturation at 1 bar ..........................................................................................178 Sulfide and sulfate saturation at high-pressure with or without the presence of a hydrous fluid .............................................................................................182 THERMODYNAMIC MODELS FOR THE BEHAVIOR OF SULFUR IN MAGMATIC SYSTEMS ...........................................................................................191 Sulfur speciation in silicate melts ..............................................................................192 The thermodynamic model of Moretti and Ottonello................................................194 GEOLOGICAL APPLICATIONS .......................................................................................203 CONCLUSIONS...................................................................................................................208 ACKNOWLEDGMENTS ....................................................................................................209 REFERENCES .....................................................................................................................209
vii
Sulfur in Magmas and Melts ‒ Table of Contents
8
The Sulfur Budget in Magmas: Evidence from Melt Inclusions, Submarine Glasses, and Volcanic Gas Emissions Paul J. Wallace, Marie Edmonds
INTRODUCTION ................................................................................................................215 SULFUR CONCENTRATIONS IN MAGMAS ..................................................................216 MAGMATIC DEGASSING OF SULFUR ...........................................................................221 Degassing and vapor-melt partitioning......................................................................221 Degassing inferred from melt inclusions from mafic volcanoes ..............................223 Magmatic vapor phase and volcanic gases ................................................................226 “Excess sulfur” or more accurately, “excess volatiles” problem...............................229 Magmatic sulfur and ore deposits .............................................................................236 Recycling of sulfur in subduction zones ...................................................................237 ACKNOWLEDGMENTS.....................................................................................................239 REFERENCES .....................................................................................................................239
9
Distribution of Sulfur Between Melt and Fluid in S-O-H-C-Cl-Bearing Magmatic Systems at Shallow Crustal Pressures and Temperatures James D. Webster, Roman E. Botcharnikov
INTRODUCTION ................................................................................................................247 BACKGROUND ...................................................................................................................249 Experimental background .........................................................................................250 METHODS ...........................................................................................................................251 Experimentation: advantages and challenges ............................................................251 Analytical: issues and challenges .............................................................................256 EXPERIMENTAL RESULTS ON SULFUR PARTITIONING BETWEEN FLUID AND SILICATE MELT ..................................................................................................258 Felsic melts — S-H2O±CO2 ......................................................................................258 Rhyodacitic melts — S-H2O-Cl ...............................................................................261 Phonolitic melts — S-H2O-Cl ...................................................................................261 Andesitic melts — S-H2O±CO2±B............................................................................265 Basaltic melts — S-H2O±CO2-Cl ..............................................................................265 Summary on S partitioning between fluids and rhyolitic to basaltic melts at crustal conditions ...............................................................................269 APPLICATION OF THE EXPERIMENTAL DATA TO PROCESSES OF FLUID EXSOLUTION AND THE EVOLUTION OF MAGMA AND MAGMATIC FLUIDS ...........................................................................................271 Magmatic gas composition as an indicator of magma and volcanic degassing activity ...........................................................................................272
viii
Sulfur in Magmas and Melts ‒ Table of Contents Volatile mixing relationships and the influence of S on CO2, H2O, and Cl solubility in melt and new insights on vapor (fluid) saturation in felsic magmas. ............................................................................................274 SUGGESTIONS FOR FUTURE RESEARCH ....................................................................276 ACKNOWLEDGMENTS.....................................................................................................276 REFERENCES .....................................................................................................................276
10
Sulfur-bearing Magmatic Accessory Minerals Fleurice Parat, François Holtz, Martin J. Streck
INTRODUCTION: THE OCCURRENCE OF MAGMATIC SULFUR-BEARING MINERALS .....................................................................................................................285 Magmatic sulfides .....................................................................................................286 Magmatic sulfates and sulfate-bearing minerals .......................................................291 MINERAL STABILITY, PARAGENESES AND MINERAL/MELT PARTITIONING OF SULFUR .......................................................................................295 Transition from sulfide to sulfate stability fields in silicate melts ............................295 Stability range of sulfide phases in magmatic systems .............................................296 Sulfates and sulfate-bearing minerals ........................................................................302 THE IMPORTANCE OF S-BEARING ACCESSORY MINERALS FOR DECIPHERING MAGMA RESERVOIR PROCESSES .................................................306 In sItu SULFUR ISOTOPE IN S-BEARING MINERALS...............................................308 CONCLUDING REMARKS ................................................................................................309 ACKNOWLEDGMENTS.....................................................................................................309 REFERENCES .....................................................................................................................309
11
Sulfur in Extraterrestrial Bodies and the Deep Earth Denton S. Ebel
COSMOCHEMISTRY OF SULFUR ...................................................................................315 Silicate melts and sulfur in primitive source materials..............................................315 Sulfur content of the terrestrial planets .....................................................................318 EXPERIMENTAL CONSTRAINTS ....................................................................................318 Element partitioning ..................................................................................................318 Liquid silicate - liquid metal-sulfide .........................................................................320 Solid metal - liquid metal-sulfide ..............................................................................321 Sulfide saturation and immiscibility ..........................................................................323 Rheology: wetting and deformation ..........................................................................324 PLANETARY INTERIORS..................................................................................................324 Iron meteorites ...........................................................................................................324 Sulfur in core fractionation........................................................................................325 Earth core formation ..................................................................................................326 Sulfur and lithophile element partitioning.................................................................328 ix
Sulfur in Magmas and Melts ‒ Table of Contents MAGMAS OF OTHER SOLAR SYSTEM BODIES ..........................................................328 CONCLUSIONS...................................................................................................................330 ACKNOWLEDGMENTS.....................................................................................................330 REFERENCES .....................................................................................................................330
12
Fining of Glass Melts Hayo Müller-Simon
INTRODUCTION ................................................................................................................337 AGENTS USED FOR CHEMICAL FINING ......................................................................339 LABORATORY EXPERIMENTS ON SULFUR CHEMISTRY ........................................341 Equilibrium experiments ...........................................................................................341 Melting experiments ..................................................................................................344 MONITORING OF THE REACTION PARAMETERS ......................................................348 Oxidation state of iron ...............................................................................................348 Oxygen sensors..........................................................................................................349 Redox number concepts ............................................................................................350 Interdependence of redox related measurements ......................................................351 MODELS OF INDUSTRIAL SULFUR FINING ................................................................352 Solubility concept ......................................................................................................352 Equilibrium concepts .................................................................................................352 Dynamic equilibrium concepts ..................................................................................353 INVESTIGATIONS UNDER INDUSTRIAL CONDITIONS .............................................355 SUMMARY ..........................................................................................................................358 ACKNOWLEDGMENT .......................................................................................................359 REFERENCES .....................................................................................................................359
13
Sulfur Degassing From Volcanoes: Source Conditions, Surveillance, Plume Chemistry and Earth System Impacts Clive Oppenheimer, Bruno Scaillet, Robert S. Martin
INTRODUCTION ................................................................................................................363 Geodynamics and the geochemical behavior of sulfur..............................................363 Subduction zones .......................................................................................................364 Ocean ridge environments .........................................................................................369 Hot spots ....................................................................................................................371 Flood basalts and silicic parts of large igneous provinces.........................................372 MEASURING VOLCANIC SULFUR EMISSIONS ...........................................................373 Direct sampling .........................................................................................................377 In situ sensing ...........................................................................................................377 Ultraviolet spectroscopy ............................................................................................378 Broad-band infrared spectroscopy ............................................................................383 Laser spectroscopy ....................................................................................................385 x
Sulfur in Magmas and Melts ‒ Table of Contents Satellite remote sensing .............................................................................................385 INTERPRETATION OF SULFUR-EMISSION DATA........................................................386 Proportions of sulfur species .....................................................................................386 Sulfur fluxes...............................................................................................................387 Sulfur isotopes ...........................................................................................................388 VOLCANIC SULFUR EMISSION TO THE ATMOSPHERE ............................................389 Ice cores .....................................................................................................................392 ATMOSPHERIC AND CLIMATIC IMPACTS OF SULFUR DEGASSING......................394 Chemical schemes relevant to volcanic sulfur emissions .........................................394 Impacts of tropospheric sulfur emissions from volcanoes ........................................398 The atmospheric and climatic impact of the 1991 eruption of Mt. Pinatubo ............399 Requirements for a climate-forcing eruption ............................................................404 SUMMARY AND CONCLUSIONS ....................................................................................405 ACKNOWLEDGMENTS.....................................................................................................406 REFERENCES .....................................................................................................................406
14
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, and Magmas Luigi Marini, Roberto Moretti, Marina Accornero
INTRODUCTION ................................................................................................................423 BASIC PRINCIPLES ...........................................................................................................424 Terminology of sulfur isotope systematics ...............................................................424 δ33S and δ36S values ...................................................................................................425 Equilibrium fractionation factors ..............................................................................426 Isotope geothermometry ............................................................................................430 SULFUR ISOTOPIC COMPOSITION OF NATURAL SAMPLES ..................................430 Mantle-derived materials and igneous rocks .............................................................430 Sulfide and sulfate minerals from magmatic, magmatic-hydrothermal and related ore deposits ..................................................................................431 Sulfur isotopes in magmatic and magmatic-hydrothermal systems ..........................434 SULFUR STABLE ISOTOPES AND THE POTENTIAL FOR PROBING DEGASSING AND CRYSTALLIzATION PROCESSES AND SULFUR SOURCES IN MAGMAS ...............................................................................452 Magmatic degassing and sulfur isotope fractionation ...............................................453 Separation of S-bearing liquids and/or solid phases..................................................454 Parametric assessment ...............................................................................................454 Comparison of analytical S and δ34SΣS data from selected volcanic systems and theoretical models of degassing and separation of sulfides and anhydrite ......465 Conclusions and future research................................................................................481 ACKNOWLEDGMENTS.....................................................................................................482 REFERENCES .....................................................................................................................482
xi
Sulfur in Magmas and Melts ‒ Table of Contents
15
Interactions Between Metal and Slag Melts: Steel Desulfurization Jean Lehmann, Michèle Nadif
INTRODUCTION ................................................................................................................493 PARTITIONING OF SULFUR BETWEEN SLAG AND METAL MELTS........................494 Sulfide capacity .........................................................................................................494 Modeling ...................................................................................................................495 Desulfurization during secondary metallurgy operations .........................................496 METHODS OF DESULFURIzATION................................................................................499 Evolution of S-content during liquid steel refining ...................................................499 Desulfurization of steel by steel-slag stirring ...........................................................501 Slag formers additions ...............................................................................................503 Slag composition .......................................................................................................503 Stirring conditions .....................................................................................................503 Desulfurization of steel by lime powder injection ...................................................506 SUMMARY AND FUTURE WORK ...................................................................................507 LIST OF SYMBOLS ............................................................................................................509 REFERENCES .....................................................................................................................510
16
The Role of Magmatic Sulfur in the Formation of Ore Deposits Adam C. Simon, Edward M. Ripley
INTRODUCTION ................................................................................................................513 GEOCHEMISTRY OF SULFUR IN MAGMATIC-HYDROTHERMAL SYSTEMS........514 Sulfur basics ..............................................................................................................514 Behavior of sulfur in silicate melts ............................................................................516 THE PARTITIONING OF METALS AND SULFUR AMONG MAGMATIC PHASES ...519 The partitioning of sulfur between silicate melt and H-O-S-Cl fluid(s)....................519 Controls on the partitioning of ore metals among silicate melt and crystalline sulfides ...........................................................................................519 The partitioning of ore metals among silicate melt, sulfide liquid and sulfide crystals ..........................................................................................523 The partitioning of ore metals among silicate melt and S-bearing aqueous fluid(s) ...............................................................................................527 The transport of ore metals in magmatic-hydrothermal fluid(s) ...............................530 The partitioning of ore metals between vapor and brine at temperatures below the water-saturated granite solidus.........................................................539 PORPHYRY-TYPE ORE DEPOSITS ..................................................................................540 Porphyry basics .........................................................................................................540 Tectonic setting and associated magma composition of porphyry deposits ..............540 Source of sulfur in porphyry environments ...............................................................541 Source of ore fluids in porphyry environments ........................................................542 xii
Sulfur in Magmas and Melts ‒ Table of Contents Constraints on the composition of porphyry-ore forming fluids ...............................543 Harmonizing fluid transport data from nature and experiments ................................546 Oxidation state of causative magmas: the role of sulfide vs. sulfate .........................546 Causative magma sources: normal or enriched? .......................................................549 Deposition of metal-sulfides in the porphyry environment .......................................551 Ni-Cu-(PGE) DEPOSITS .....................................................................................................552 Characteristics and classification of magmatic Cu-Ni-(PGE) deposits .....................552 Resource and grade characteristics............................................................................553 A general model for magmatic Ni-Cu ore genesis ....................................................554 Source magmas for Ni-Cu deposits ...........................................................................556 Transport of sulfide melt............................................................................................556 PGE DEPOSITS IN LAYERED INTRUSIONS ..................................................................557 Characteristics and classification of PGE deposits....................................................557 Models for the genesis of PGE deposits in layered mafic intrusions ........................559 Source magmas for PGE deposits .............................................................................563 FUTURE RESEARCH: WHAT DO WE NEED? ................................................................563 ACKNOWLEDGMENTS.....................................................................................................564 REFERENCES .....................................................................................................................564
xiii
1
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 1-8, 2011 Copyright © Mineralogical Society of America
Studies of Sulfur in Melts – Motivations and Overview Harald Behrens Institut für Mineralogie and ZFM Center for Solid State Chemistry and New Materials Leibniz Universität Hannover Callinstr. 3 D-30167 Hannover, Germany
[email protected]
James D. Webster Dept. of Earth and Planetary Sciences, Division of Physical Sciences American Museum of Natural History Central Park West at 79th St. New York, New York 10024-5102, U.S.A.
[email protected]
INTRODUCTION Background For the past 37 years the Mineralogical Society of America, and in conjunction with the Geochemical Society (since 2000), have sponsored and published 72 review volumes that communicate the results of significant advances in research in the Earth sciences. Several of these have either directly or indirectly addressed the fundamental importance, role, and behavior of volatile components on processes influencing magma rheology, crystallization, evolution, eruption, and related metasomatism and mineralization. Volume 30—which was published in 1994—focused on this topic broadly, and this volume has provided a lasting summary on the geochemical and physical behaviors of a wide variety of magmatic volatiles (Carroll and Holloway 1994). Since that year, continued research has brought important and new knowledge about the role of the volatile component sulfur in natural magmas, and significant progress was made simultaneously in understanding the role of sulfur in industrial or technical processes such as glass or steel production. Here, in volume 73, we have assembled in 15 chapters the current state of research concerning sulfur in melts based on the extensive experience of various authors practically working on these topics. The behavior of sulfur in melts and its implications for natural and industrial processes are still insufficiently understood, and hence, are difficult to apply as a tool for interpreting problems of geological or industrial interest. In recent decades, various new investigations in the geosciences as well as in the engineering and material sciences have employed modern spectroscopic, analytical, theoretical, and experimental techniques to improve our understanding of the complex and volatile behavior of sulfur in a wide variety of molten systems. However, these different research initiatives (e.g., empirical vs. applied research and natural vs. technical applications) were rarely well integrated, and the scientific goals were usually approached with specific and relatively focused points of view. Consequently, bridging this longstanding gap between the earth sciences and material sciences is one of the major objectives of this published volume and its associated short course (presented August of 2011 in Goslar, Germany). Furthermore, it is currently an appropriate time to provide a comprehensive overview on the 1529-6466/11/0073-0001$05.00
DOI: 10.2138/rmg.2011.73.1
2
Behrens and Webster
state of the art concerning sulfur in silicate melts given the important advances made since 1994, and also to address important questions which remain to be resolved.
The behavior and importance of sulfur in melts Sulfur is one of the most abundant volatiles in terrestrial magmas. During volcanic eruptions, large amounts of sulfur are released to the atmosphere, mostly as SO2 and H2S, which have important climatic impacts. Sulfur also has significant effects on the partitioning of a wide variety of elements between silicate melts, liquid metals, gases, and solids, and consequently magmatic sulfur species exert major controls on the genesis of a large variety of ore deposits. Physicochemical processes involving sulfur also affected the evolution of the early solar system. Furthermore, sulfur plays important roles in technical processes involved in glass and steel production. In glass manufacturing, sulfate is often added for fining of glass melts, i.e., to remove bubbles which would otherwise cause significant quality degradation, and sulfur also influences the color of industrial glass. In steel production, molten metals are forced to interact with silicate slag melts for desulfurization which is required to produce high-performance steel. Although the specific problems related to sulfur in silicate melts vary between natural and industrial applications, the underlying physicochemical processes are nevertheless the same. The solubility of sulfur-bearing minerals and the partitioning of sulfur between coexisting phases (e.g., a melt and gas phase) determine how much sulfur can be dissolved in or extracted from melts. Moreover, the diffusivity of sulfur in melts exerts a major control on the formation of bubbles and on the kinetics of mineral dissolution and growth. It is also important to bear in mind that the behavior of sulfur in silicate melts is much more complex than that of other volatiles, such as water and carbon dioxide, because sulfur dissolves in silicate melts in at least two different oxidation states. At low oxygen fugacities below that of the nickel-nickel oxide oxygen buffer, sulfide (S2−) is the predominant sulfur species whereas at higher oxygen fugacities sulfate (SO42−) is dominant. Other species such as sulfite (S4+) may exist as well at specific conditions. As a consequence of the multi-species behavior of sulfur, it is often difficult to model and sometimes difficult to predict the behavior of sulfur in natural and industrial processes. Complex reactions which are coupled to other oxidation-reduction equilibria (e.g., those involving heterovalent cations such as iron) can occur due to changes in the oxidation state of sulfur. For instance, during the degassing of silicate melt at very high temperature and atmospheric pressure, sulfate becomes instable and dissociates forming sulfur dioxide and oxygen, and the latter species are able to oxidize ferrous to ferric iron.
REVIEW CHAPTERS IN THIS VOLUME In chapter 2, an overview on methods for quantifying the sulfur contents of glasses, minerals, and other solid phases is given by Ripley et al. (2011, this volume). The most common technique to gain information about the spatial distribution of sulfur in condensed matter is still electron microprobe analysis (EMPA), but other methods, i.e., laser-ablation combined with mass spectrometry (LA-ICPMS) or optical emission spectrometry and secondary ion mass spectrometry (SIMS) are also frequently used tools. X-ray fluorescence (XRF) is a widely employed method for analysis of rocks, owing to its ability to simultaneously analyze a wide range of elements in different matrices with a modest level of sample preparation. This chapter also describes nuclear reaction analysis, which involves bombardment of material with a high-energy beam and subsequent detection of gamma radiation, as another technique used to measure samples with very low sulfur concentrations. Issues bearing on proper standards, analytical detection limits, and analytical uncertainties are also discussed. In chapter 3, Wilke et al. (2011, this volume) report on the significant progress that has been made in the last few years in the characterization of sulfur speciation in glasses by spectroscopic
Sulfur in Melts – Motivations and Overview
3
techniques. X-ray absorption spectroscopy (XAS) is the most commonly used tool. This method can clearly distinguish oxidation states of sulfur in the glasses and yields information about the coordination of sulfur, but quantitative interpretation of XAS spectra in terms of relative proportions of sulfur species in glasses is still in its early stages and requires additional calibration work. Magic-angle spinning nuclear magnetic resonance spectroscopy (MAS NMR) also has the potential to yield insights into the coordination and the oxidation states of sulfur in glasses, but its application is limited by the low natural abundance of the sulfur isotope 33S used in NMR measurements. Recently, Raman spectroscopy has been intensively applied to study sulfur speciation, i.e., to analyze specific bands of sulfide, sulfate, and hydrogen sulfide in glasses. However, as described by Wilke et al., this technique is applicable only for specific melt compositions and at sulfur contents exceeding several hundreds of ppm by weight. A variety of processes in magmas and melts are controlled by sulfur diffusion and oxidation-reduction reactions of sulfur in the melt, and an overview on this topic is given in chapter 4 by Behrens and Stelling (2011, this volume). Close inspection of data in the published literature implies that differences in sulfur speciation, i.e., whether sulfide or sulfate is the dominant sulfur species in melts, have little influence on the mobility of sulfur in the melts. The evaluation of results of experimental studies in this chapter does, however, provide evidence that sulfur diffusion is strongly coupled to the dynamics of the silicate network and, hence, to the viscosity of the melt. It is shown that the Eyring relationship, which relates diffusivity to viscosity, is a good tool for predicting sulfur diffusivity in the melts. The kinetic issues of melting of raw materials in industrial glass manufacturing are described in chapter 5 by Falcone et al. (2011, this volume). During the production of sodalime-silicate glass, sulfur-containing raw materials (sulfates and sulfides) are added to improve the quality of the final product; sulfate is used for fining of glass melts, i.e., to remove bubbles from the glass batch. In addition, the oxidation-reduction conditions during glass melting and minor contents of iron exert major controls on the color of the glass products. In particular, the amber color of glasses is attributed to chemical associations, such as Fe3+-S2−, which are stable only in a narrow range of oxygen fugacity, the latter of which can be established by adding specific amounts of carbon sources to the raw materials prior to melting. It has long been of interest to predict the solubility of sulfur in silicate melts under the conditions relevant to glass and steel production as well as for natural processes. In chapter 6, Backnaes and Deubener (2011, this volume) give an overview of experimental results improving industrial glass or steel production. Particular attention is given to experiments on the equilibration of gas mixtures with silicate melts and on the effects of adding reducing agents such as carbon to the melt at ambient pressure. It is noted that the measured quantities of sulfur in the melt typically do not represent equilibrium conditions but instead are affected by kinetic factors such as diffusion and convection in the melt. Nevertheless, such experiments provide useful information on how different chemical components affect the ability of the melt to bind sulfur species. A variation in oxygen fugacity in the gas mixtures in contact with the melts at constant sulfur fugacity demonstrates that sulfur solubility is particularly high when sulfate is the stable sulfur species. High sulfur concentrations in the melts are also achieved when equilibrating silicate melts with gas mixtures under reducing conditions, and a distinctly V-shaped solubility curve for sulfur in the melt as a function of the logarithm of oxygen fugacity in the gas phase is often observed at ambient pressure. Useful approaches to estimate the solubility of sulfur in silicate melts relevant to natural systems are given in chapter 7 by Baker and Moretti (2011, this volume). Empirical models which reproduce experimental data accurately have been tested, but these models may have large uncertainties when extrapolated to conditions which are far from those covered by the experimental data. Thermodynamic models have been constructed and are more applicable and useful for extrapolation out of the range of the supporting experimental data. In particular,
4
Behrens and Webster
a new model is proposed by Baker and Moretti (2011, this volume) for prediction of sulfur capacity at anhydride saturation (SCAS) in silicate melts. As detailed in chapter 8 by Wallace and Edmonds (2011, this volume), silicate melt inclusions trapped in minerals provide the opportunity to investigate processes of magma evolution and subsequent eruptive processes. Silica-rich melts are characterized by relatively low sulfur solubilities, so it is clear that the recharge of mafic magma with more chemically evolved magmas (some of which may be saturated in vapor) contributes to the volatile budget of at least some intermediate and silicic volcanoes. Characteristic ranges in S content of pillowbasalt rim glasses and melt inclusions collected from mid-ocean ridge basalts (MORB) and basaltic arc magmas are summarized. This chapter addresses mafic magmas that are able to mingle and mix efficiently with resident magmas. It also describes how and why the mixed magmas associated with mafic recharge are usually highly oxidized and silicic in composition which promotes sulfur partitioning into the vapor phase. The sulfur data from melt inclusions are used to interpret volatile recycling in subduction zones, compositions of pre-eruptive volcanic gases, and the “excess sulfur” problem associated with some eruptive systems. The partitioning of sulfur between fluids and melts, as reviewed in chapter 9 by Webster and Botcharnikov (2011, this volume), controls the volatile budget of magmas and sulfur release on volcanic degassing. In general, sulfur is preferentially incorporated in a magmatic fluid phase, i.e., fluid/melt partition coefficients are larger than 1 and even may reach values of 1000 or more. Temperature, pressure, and oxygen and sulfur fugacities as well as fluid and melt composition affect the fluid/melt partitioning of sulfur. Typically, silica-rich melts (e.g., rhyolitic melts) display larger S partition coefficients than silica-poor (e.g., basaltic) melts. With knowledge of the partitioning behavior of sulfur, the measurements of magmatic gas compositions may be used as an indicator of magma degassing activity and style. Fluids generated in magmatic systems are usually composed of various components in the system S-O-H-C-Cl, but other halogens and N may also be involved, and as described in this chapter, experimental research on sulfur partitioning at elevated pressures between silicate melts and complex multi-component fluids is insufficient, at present, to model degassing in all magmas. An overview on the importance and role of sulfur-bearing minerals in magmas is given in chapter 10 by Parat et al. (2011, this volume). As described therein, S-bearing minerals represent only a negligible component of the mineral assemblage in magmatic rocks, and hence, are accessory minerals. In most cases, these S-bearing accessory phases are iron sulfides, typically pyrrhotite, but chalcopyrite, pentlandite, sphalerite or molybdenite also occur. Among sulfate minerals that are stable in magmatic systems, anhydrite is the most common phase. Other magmatic SO4-bearing minerals include S-rich apatite, haüyne/sodalite, and silvialite. Despite the small abundance of these S-bearing minerals, they are extremely useful for estimating the activity of various sulfur-bearing species in the magmas, to constrain oxygen fugacity and the S concentration in magmas prior to eruption and degassing (e.g., pre-eruptive sulfur concentration in melts). In chapter 11, Ebel (2011, this volume) addresses the current understanding of sulfur in extraterrestrial bodies and in the deep Earth. Our knowledge on the abundances of S, C, N, O, and other volatile components in the solar system is based mainly on meteorite data combined with spectroscopic measurements of the solar photosphere. However, sampling of comets in the NASA Stardust mission and experimental studies, i.e., on element partitioning between metals, metal liquids, silicate and silicate melts, at high pressures also provide useful information. The solar abundance of S expressed by the ratio of (S/(Fe+Ni+Co)) is approximately 0.5. It is preserved at least in the carbonaceous chondrites (CI) and perhaps also in comets, while the rocky planets and most meteorite parent bodies (asteroids) have either lost or never accreted a significant sulfur component. This chapter describes how silicate melts are involved in nearly every phase of planetary evolution, from accretionary impacts through core/mantle
Sulfur in Melts – Motivations and Overview
5
differentiation to volcanism and subduction. While the effect of dissolved sulfur on element partitioning between melt and solid metals has been investigated, it remains an open question about how sulfur affects the partitioning of other elements during core/mantle differentiation. The fining of molten glass, as described in chapter 12 by Müller-Simon (2011, this volume), is an important step in industrial glass manufacturing that is used to improve the quality of glass products. The primary melt may contain numerous bubbles which originate from air trapped in the cavities of the solid starting mixture before melting, and from CO2 that is generated through combustion of organic impurities and the decomposition of carbonates during melting. In the fining process, these bubbles expand by sequestration of other gas-soluble components in the melt facilitating the ascent of the bubbles to the melt-gas interface, so the number of bubbles decreases strongly in the product with heating time. Sulfate salts are often added to the raw material batch for fining of soda-lime silicate melts, which is the most commonly used base composition for container glass, flat glass, fiber glass and glass tableware products. Upon heating to temperatures above 1400 °C, sulfate ions dissociate to sulfur dioxide and oxygen which diffuse into the pre-existing bubbles. As detailed in this chapter, understanding the elementary processes of fining is crucial in optimizing the entire glass-production process. In Chapter 13, Oppenheimer et al. (2011, this volume) discuss sulfur degassing from volcanoes with respect to source conditions in magma, the surveillance of volcanic activities, the chemical composition of released gases, and global impacts of magmatic sulfur release. This review also addresses the speciation of sulfur in volcanic vapors, the causes of variability in sulfur abundance and speciation in different geodynamic contexts, techniques and resulting data in the measurement of sulfur emissions from volcanoes, the links between subsurface processes and surface observations, and the consequences of volcanic sulfur degassing for climate and the environment. The 1991 activity of Mt. Pinatubo is described as a well-documented example of the global impacts of a single major eruption (approximately 10 km3 of erupted material) that has been studied with significant instrumental detail. Although twenty years have already passed since that eruption, it is remarkable to note that new findings concerning its climatic, environmental, and ecological consequences are still emerging. Despite the tremendous insights afforded by this event, it represents only a very small sample of the broad range of volcanic eruption styles, geographic locations, and atmospheric states that combine to produce significant perturbations to atmospheric composition, radiation, and dynamics. Chapter 14, by Marini et al. (2011, this volume), reviews how the stable isotopes of sulfur serve as an important geochemical tool for the study of processes occurring in magmarelated, hydrothermal systems, melts, and magmas. Sulfur has four stable isotopes with natural abundances of 95.02% (32S), 0.75% (33S), 4.21% (34S), and 0.02% (36S). Importantly, small variations in the ratios of these isotopes can be induced by kinetic and thermodynamic effects, and these variations or fractionations are exacerbated given the multiple oxidation states of sulfur. The fractionation of sulfur isotopes between co-existing phases is particularly large when the oxidation state of sulfur differs in both phases and when the phases have different aggregate physical states (i.e., gas phases vs. condensed phases). This chapter summarizes how sulfur isotopes have been measured in volcanic gases as well as in samples from magmatic rocks and or magmatic ore deposits. These data are compared with theoretical models of degassing and the crystallization and physical separation of sulfides and anhydrite from melt to gain insights into the processes occurring in magma chambers and during the ascent of the magma to the surface. In particular, this chapter addresses sulfur isotope characteristics for magmas and eruptive products of Mt. Vesuvius, Mt. Mazama, and Mt. Etna. Desulfurization of molten metals by interaction with slags is widely applied in the production of high-quality steel as described in chapter 15 by Lehmann and Nadif (2011, this volume). Sulfur is mainly present in solid steel as manganese sulfide (MnS) inclusions which strongly affect the processing and properties of steel. Since such inclusions behave
6
Behrens and Webster
more plastically than steel during deformation, they act as crack-initiation sites and zones of weakness. Therefore, sulfur is detrimental to the malleability, ductility, toughness, formability, weldability, and corrosion resistance of steels, and hence, very low levels of sulfur (<0.003 to <0.001 wt% S) are required especially for high-quality, flat steel products. Such low-S levels are usually obtained by re-melting the raw metal in a converter with the addition of slags with a high sulfide-capturing capacity. An example slag composition consists of 50 wt% CaO, 18 wt% FeO, and 13 wt% SiO2 plus minor additional components. Understanding the interaction between liquid metal and slag with respect to thermodynamics (i.e., partitioning of sulfur) and kinetics (i.e., exchange reactions between slags and metal liquid and transport within the phases) is crucial for improving the manufacturing processes. A significant body of field, laboratory, and experimental evidence, as reviewed in the final chapter (i.e., 16) by Simon and Ripley (2011, this volume), suggests that sulfur plays a pivotal role in the generation of numerous magmatic and hydrothermal ore deposits. Sulfur is important in controlling the concentration of ore metals in the silicate melt via sequestration of metals in molten and crystallizing sulfides, in controlling the ability of a magmatic-hydrothermal fluid to scavenge ore metals from the silicate melt (by complexing of metals with various sulfur species), in moderating the ability of the hydrothermal fluid to retain and transport ore metals from magma to subsolidus deposition sites, and in controlling the ability of a particular ore metal to precipitate from hydrothermal fluids. Modeling these processes requires detailed knowledge on the speciation of sulfur in the melt, the speciation of metals in sulfur-bearing melts and hydrothermal fluids, and the behavior of sulfur in ore forming H-O-S-Cl fluid(s) at magmatic and sub-magmatic conditions. Furthermore as described in this chapter, redox conditions have a major impact on these properties because the oxidation state of metals and sulfur vary strongly with oxygen fugacity which affects the stability of ore metal complexes in fluids and melts. This physicochemical framework is used to discuss typical ore metal abundances in melts and fluids, the formation of different ore deposit types, their tectonic settings, and the nature of sulfide mineralization.
REMAINING ISSUES AND CONSIDERATIONS FOR FUTURE RESEARCH A recurrent observation in all chapters of this volume is that the speciation, and hence the resulting complex chemical and isotopic behaviors, of sulfur in these systems are strongly controlled by the capacity of this volatile to change valence states—potentially over a broad range—in vapor and condensed phases. Recent advances in analytical methods for determining compositional and structural data on sulfur in minerals, fluids, glasses, and melts, for example, have improved our understanding of the principal functions of this volatile component in natural and synthetic systems, but because of the associated chemical complexities there are still many open questions. It should be noted that some of these issues are addressed in the brief recent collection of scientific papers on sulfur in Elements by Métrich and Mandeville (2011). With regard to the variety of methods used for sulfur analysis and related observational data: (1) at present, there is no simple, readily available, and accurate means of directly measuring the concentration of S in quenched experimental fluids and in natural fluid inclusions (with potentially broad ranges in S concentrations in chemically complex fluids), but recent developments in the use of laser ablation ICP-MS are promising. (2) Given the problematic nature of quenching melts and fluids, in situ measurements of sulfur dissolution and speciation in fluids and melts of experiments at geologically relevant pressures and temperatures are needed. And, information on interactions
Sulfur in Melts – Motivations and Overview
7
between sulfur and other multivalent elements in silicate melts would be particularly beneficial from such experiments. (3) The use of high-spatial resolution, micro-analytical methods (e.g., SIMS) for measuring the stable isotopes of sulfur in silicate glasses has only recently been established, and consequently, there is significant potential and need to apply this new geochemical tool to interpret fluid exsolution and degassing processes through isotopic analyses of sulfur in matrix glasses and silicate melt inclusions. Similarly, we need increased data for sulfur isotopes in individual, and potentially zoned, sulfide and sulfate minerals to better understand magma storage conditions, magma evolution, and degassing. (4) Additional work is required to determine the temporal and spatial distribution of sulfur-bearing gases and particles emitted from volcanoes to the atmosphere. This is very important for sulfur because of its chemical reactivity in volcanic plumes and its key role in forming volcanic aerosols that modify the Earth’s albedo. (5) Development and improvement of in situ monitoring techniques are crucial for glass and steel production to reduce the time between observation of manufacturing errors and modifying the production process. Such issues include nickel sulfide-induced breakage of thermally toughened glasses and their prevention by heat-soak-tests, and the relevant feedback times for customers that can reach months (Kasper and Gelderie 2008). Furthermore, acquisition of real-time data on the concentrations of the different sulfur species in the glass melt and information on fining bubbles and the tank atmosphere would help to improve the efficiency of the fining process during glass production. The production of ultra-low-S content steel requires enhancement of the kinetics of desulfurization during steel-slag stirring at atmospheric pressure, and the implementation of in situ measurements is important to assess the gas stirring quality online during this process. With regard to the behavior of sulfur in silicate melts: (6) Experiments at conditions corresponding to those prevailing in the deep Earth (elevated pressure and temperature) are needed to determine the partitioning of sulfur (and, in particular, its isotopes) and associated chalcophile trace elements between relevant minerals, sulfide melts, and mafic and ultramafic melts. (7) The phase relations of geologically relevant melts saturated in S-bearing fluid and mineral phases require additional investigation, and this is particularly necessary for systems containing C-O-H-S-Cl-F-bearing, chemically complex, geologically relevant fluids. Current understanding of the partitioning of sulfur between these phases at controlled and known fugacities of oxygen and sulfur is particularly poor. Thermodynamic models predicting the dissolution of sulfur in minerals, melts, and fluid phases as well as phase relations in S-bearing systems are crucially needed. Improved thermodynamic models are necessary as a support and a guide for the strategies of future research on sulfur exsolution from melts, on sulfur isotope fractionation and on the role of sulfur-charged vapors and fluids in processes related to volcanic eruption, magma evolution, and mineralization.
ACKNOWLEDGMENTS We appreciate the support of the Mineralogical Society of America and the Geochemical Society. We acknowledge thoughtful reviews of this chapter by Renat Almeev, Joachim Deubener, Francois Holtz, and Max Wilke. The completion of this volume involved the kind
8
Behrens and Webster
and extensive reviews of referees too numerous to name. We appreciate editorial assistance by Nanette Nicholson, and final editorial handling by Jodi Rosso is gratefully acknowledged. We acknowledge and appreciate financial support from the Volkswagen Foundation (Germany) and the U.S. National Science Foundation.
REFERENCES Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Behrens H, Stelling J (2011) Diffusion and redox reactions of sulfur in silicate melts. Rev Mineral Geochem 73:79-111 Behrens H, Webster JD (2011) Studies of sulfur in melts – motivations and overview. Rev Mineral Geochem 73:1-8 Carroll MR, Holloway JR (eds) (1994) Volatiles in Magmas. Reviews in Mineralogy, Volume 30. Mineralogical Society of America. Ebel DS (2011) Sulfur in extraterrestrial bodies and the deep earth. Rev Mineral Geochem 73:315-336 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Gelderie U, Kasper A (2008) How to prevent spontaneous breakage of thermally toughened alkaline earth silicate glass. Eur J Glass Sci Technol Part A 49(3):126-132 Lehmann J, Nadif M (2011) Interactions between metal and slag melts: steel desulfurization. Rev Mineral Geochem 73:493-511 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Métrich N, Mandeville CW (2010) Sulfur in magmas. Elements 6(2):81-86 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285314 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Wallace PJ, Edmonds M (2011) The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Rev Mineral Geochem 73:215-246 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78
2
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 9-39, 2011 Copyright © Mineralogical Society of America
Analytical Methods for Sulfur Determination in Glasses, Rocks, Minerals and Fluid Inclusions Edward M. Ripley, Chusi Li, Craig H. Moore, Erika R. Elswick Department of Geological Sciences, Indiana University Bloomington, Indiana 47405 U.S.A.
[email protected]
J. Barry Maynard Department of Geology, University of Cincinnati Cincinnati, Ohio 45221-0013, U.S.A.
Rick L. Paul Analytical Chemistry Division, National Institute of Standards and Technology Gaithersburg, Maryland 20899, U.S.A.
Paul Sylvester Department of Earth Sciences and Inco Innovation Centre, Memorial University St John’s NL, A1B 3X5, Canada
Jun Hun Seo Institute of Geochemistry and Petrology, ETH Zurich 8092 Zurich, Switzerland
Nobomichi Shimizu Department of Geology and Geophysics, Woods Hole Oceanographic Institution Woods Hole, Massachusetts 02543, U.S.A. INTRODUCTION The analytical techniques that are normally utilized to determine the concentration of sulfur in geologic materials can be divided into bulk analytical methods that require sample powders and microanalytical methods done in situ. The most common method of bulk sulfur analysis is accomplished using combustion methods followed by detection of SO2 in an infrared cell. A similar method involves an elemental analyzer coupled to a mass spectrometer where the mass 64 ion beam is monitored and compared to a standard. X-ray fluorescence is another method of bulk powder analysis that is well established for the determination of sulfur in coal and plant materials, but is only rarely used for the analysis of rocks due to a relatively high detection limit and difficulties with sample fusion. Micro-analytical techniques normally involve individual minerals or glasses. The most commonly used method is electron microprobe analysis (EMPA), but ion microprobe (SIMS) and laser ablation-inductively coupled plasma-mass spectrometry (LA-ICP-MS) methods are useful for some samples. Nuclear methods of sulfur analyses are useful for some materials largely because of their very low detection limits. A relatively new technique is the application of LA-ICP-MS to determine the concentration of sulfur in fluid inclusions contained within glasses and minerals. 1529-6466/11/0073-0002$05.00
DOI: 10.2138/rmg.2011.73.2
10
Ripley et al.
In this chapter we review how these methods may be used to measure the concentration of sulfur in a variety of geologic materials. We emphasize the types of materials that can be analyzed, sample preparation, standardization techniques, detection limits and uncertainties that are associated with each of the methods. Elsewhere in this volume, the determination of sulfur species in glasses is discussed (Chapter 2 -- Wilke et al. 2011) and methods of isotopic analyses are treated (Chapter 15 -- Marini et al. 2011).
ANALYTICAL TECHNIQUES Total sulfur concentration in minerals, rocks and glasses using elemental analyzer – infrared absorption technology Automated, high-temperature combustion, carbon-sulfur analyzers utilize solid state infrared (IR) absorption for the detection of total sulfur (TS) and total carbon (TC). Samples are combusted in either an induction or resistance furnace in the presence of oxygen to form SO2 and CO2. LECO Corporation developed these instruments for the use in the steel industry in the 1930’s, but they have successfully made the transition to the analysis of natural materials, including glasses (e.g., Williams et al. 1957; Gibson and Andrawes 1978; Mathez 1980; Bach and Erzinger 1995; de Hoog et al. 2004). In the early 1980’s, ELTRA analyzers (ELTRA GmbH, Neuss, Germany) began producing elemental analyzers for the same market, with very comparable results. The signals emitted from the infrared reference cells are selective and correspond to the IR radiation absorbed from the SO2 and CO2 concentrations in the sample gas mixture. The instrument electronically linearizes and integrates the signal from the receiver cells with the initial sample mass to compute a concentration, usually in the wt% range. Most middle- to high-end instruments report a minimum sulfur detection limit between 100 and 200 ppm (e.g., ELTRA CS2000, LECO CS 600). For sulfur analysis, the combustion of the sample takes place in a temperature range of 1250 to 1450 °C (the temperature necessary for the decomposition of sulfate minerals) in an O2 stream. The oxygen gas also acts as a carrier gas to move the sample through a series of traps to remove water molecules, and halogens before reaching the IR detectors. The induction furnace is useful for samples with a minimal number of sulfur phases in a fairly uniform matrix, as heating and release of SO2 are rapid. With induction combustion a modifier, typically low-sulfur tungsten of iron powder is added to each sample prior to combustion. For more complex matrices, the resistance furnace holds the sample at temperature for a longer period of time (100-150 s) to allow for more recalcitrant forms of sulfur (e.g., barite) and carbon (e.g., graphite) to evolve into SO2 and CO2 gas. Most instruments require a range of 100 to 800 mg of ground sample material to make reproducible measurements. As expected, the leaner the sample the more sample material may be necessary to achieve reproducible results. Most middle- to high-end instruments report standard sulfur detection limits between 100 and 200 ppm (e.g., ELTRA CS2000, LECO CS 600), with a theoretical minimum of 0.6 μg of S in a homogeneous 1 g sample (0.6 ppm). The instruments are calibrated using standards containing similar sulfur species, concentrations and matrix to those of the samples. NIST or USGS standards such as GXR-4 copper mill-head (1.77% S) (e.g., Roy et al. 2009), or USGS Standard Reference Materials (SRMs) including SDO-1 Devonian Ohio Shale (9.95% C and 5.35% S) (e.g., Elswick et al. 2007), or CANMET CCU-1 copper concentrate (35.4% S), RTS-1 sulfide ore mill tailings (1.66% S) and SV-1a (10% S) nickel-copper-cobalt ore (Kerr 2001), are used for high sulfur samples. For lower sulfur samples, SRM USGS BCR-1 (464 ppm S) has been successfully demonstrated to be a useful standard (e.g., Gibson and Andrawes 1978), and it follows that standards such as Geological Survey of Japan GSJ JA-3 andesite (228 ppm S) would also be suitable. Alternatively, commercially available SRM’s of coal, soil and other geologic materials can be utilized for instrument calibration (e.g., Giblin et al. 1990; Bach and Erzinger 1995; de Hoog et al. 2004).
Analytical Methods
11
Total sulfur concentration in minerals, rocks and glasses using elemental analyzer – mass spectrometer technology A mass spectrometer with an attached elemental analyzer (EA, Fig. 1) can be used to determine the concentration of sulfur in mineral or whole-rock samples when the concentration is about 50 ppm or more (5 μg of sulfur in a 100 mg sample). The procedure is straightforward. Five or more samples of a sulfur-containing and chemically well-characterized mineral powder are prepared, spanning a range of weights, and are analyzed with the system. SO2 is produced and the mass 64 peak (32S16O16O) is utilized with the peak area proportional to the amount of SO2. The relationship between sulfur concentration and peak area is typically linear in a welltuned system. From the weights of a standard and knowledge of the chemical composition of the standard, the SO2 peak area can be converted to the weight of sulfur in each standard and then to the concentration of sulfur in each sample. Once this relationship is determined, samples of unknown composition can be weighed and analyzed and the resulting peak areas converted to sulfur weights and then concentrations. The procedure is described in more detail below (see also Studley et al. 2002). Samples that are to be combusted in an EA and analyzed for sulfur are weighed into small tin cups in an amount sufficient to produce the desired signal strength at the Faraday cup collector of the mass spectrometer. This amount depends on the setup and tuning of the particular mass spectrometer and the laboratory’s typical operating procedures, as well as the chemical composition of the sample. The weights needed for determining concentrations will range from about 0.02 mg to 0.2 mg for chalcopyrite (CuFeS2) and from about 0.06 mg to 0.6 mg of argentite (Ag2S) on systems that are run at a relatively low (approximately 1 volt)
Figure 1. Photograph showing the elemental analyzer that is coupled to a mass spectrometer for the determination of whole-rock sulfide contents. The reactor is housed in the furnace and contains tungsten oxide and native copper separated by quartz wool. The GC column is 80 cm in length and encased in Teflon.
12
Ripley et al.
peak height; a peak height of 2 volts would require approximately twice these amounts, a peak height of 5 volts approximately 5× these amounts, etc. To each tin cup an oxidizing agent, such as vanadium or niobium pentoxide, is also added. The weight of vanadium pentoxide added is approximately 5× the weight of sulfur in the sample, or more. The tin cups are then crimped shut and loaded into the auto-sampler of the EA. At a minimum, 5 samples of well-characterized standard powders whose weights span the weight (concentration) range of interest should be run with a set of unknowns. It is best if these samples are interspersed with the unknowns to alert the user to any problems that may develop during the course of the analyses. The (usually) automated analysis sequence is then begun (e.g., EA coupled to a ThermoFisher MAT 252 mass spectrometer, Fig. 1). Each sample drops into a hot reactor column. Uniform and complete combustion is accomplished, in part, by making sure that the sample is in the hot zone of the combustion furnace—typically the hot zone is at a temperature of approximately 1000 °C. A slotted quartz glass insert, which sits on top of the reactor column and is plugged at the bottom with quartz wool, is used to position the samples as they fall from the auto-sampler. As the sample enters the hot zone, oxygen is also introduced, and that, along with the vanadium/niobium pentoxide, assures complete combustion. Residue from the combusted samples eventually builds up in the insert and, when it becomes too full, the samples are no longer in the hot zone. Hence, the insert must be replaced periodically to insure good combustion. Sample combustion is complete within 60 s. The mixture of sulfur-bearing gas produced by the combustion flows through a heated column containing a reactive packing comprised of tungsten oxide and reduced copper so that both oxidation and reduction take place in the same reactor. Ideally, after undergoing reaction, all the sulfur-bearing gases have been converted to SO2—the species that is monitored in the mass spectrometer. Depending on the composition of the sample, other gases, may also be present, commonly these are CO2 and NO2. The packing of the reactor is optimized so that the SO2 peak is sharp. The peak can become broad, however, if the reactants in the reactor are nearly spent. Also, the peaks can broaden if there are certain minerals, such as pyroxene, amphibole, or biotite (Studley et al. 2002) included with the sulfur-bearing minerals in the sample. Also, broadening can occur if the insert contains too much material, especially if it is very fine grained. After the gas exits the column, it passes through a water trap (Fig. 1) and then into a short chromatographic column. In the column, the gases are separated into discrete pulses under normal conditions. However, if the abundance of an additional gas species, e.g., CO2 or NO2, is high it is possible that complete separation is not attained and theses peaks can overlap the SO2 peak resulting in erroneous measurement of the sulfur (SO2) peak area. To illustrate the process of data reduction and concentration determination we present an example. Following the procedures described above we prepared 5 samples of chalcopyrite (known composition) in weights of 0.0245, 0.0382, 0.0643, 0.1046 and 0.1773 mg. The resulting peak areas after processing the samples as described above were 2.310, 3.871, 6.593, 12.560 and 21.972 volt/s. The number of milligrams of sulfur in each sample was calculated by multiplying the weight of each sample by 0.349—the weight proportion of sulfur in a mole of chalcopyrite. The peak area for each sample was then plotted versus the weight of sulfur in the sample and the equation of a linear best-fit line calculated. This is shown in Figure 2. A sample of basalt with an unknown sulfur concentration was prepared. The sample weighed 35.6138 mg and produced a peak area of 21.396 volt/seconds. Using the equation for the best fit line, this peak area corresponds to 0.0610 mg of sulfur. This concentration of sulfur in this amount of sample is then 1712 ppm. The routine detection limit for whole-rock analyses is ~100 ppm sulfur using a 100 mg sample. Uncertainty is typically ± 2% of the amount of sulfur present in the sample. Dissolved sulfur in glasses has proven to be very difficult to extract and measure using this technique.
Analytical Methods
13
Figure 2. Plot showing peak area versus mg of sulfur in a chalcopyrite (CuFeS2) standard. SO2 was produced using an elemental analyzer and peak areas were determined by mass spectrometry.
Total sulfur concentration of minerals, rocks and glasses using X-ray fluorescence X-ray fluorescence (XRF) is a widely used method for analysis of rocks, owing to its ability to simultaneously analyze a wide range of elements in different matrices with a modest level of sample preparation. Newer machines will in fact take unprocessed samples such as archeological artifacts (e.g., Diana et al. 2007) and it is straightforward to do direct analyses of metals if a flat surface can be machined and polished. Despite these advantages, sulfur is an element that is commonly excluded from analytical routines. The use of XRF for sulfur in coals and plant materials is well established (e.g., Bower et al. 1986; Pearce et al. 1990) but difficulties with loss of sulfur during fusion with lithium metaborate (Giles et al. 1995; Hettipatharina et al. 2004) restrict the use of the XRF technique for sulfur during analyses for major elements in rocks. Pelikánová (1985) suggested the addition of LiNO3 or LiCO3 to the flux as oxidizing agents to convert sulfide to sulfate, which would then not be volatile, but this technique does not seem to have been widely used. A better approach may be to add excess Ba to the flux to convert all sulfur to BaSO4, which is not lost during the preparation of glass disks (Gazulla et al. 2008). X-ray fluorescence is also used for trace-element analysis of rock powders without fluxing. Instead the powders are compressed into disks, and these provide an opportunity to determine sulfur without the fluxing problems. In the laboratory at the University of Cincinnati, samples are routinely screened for sulfur when doing analyses of pressed powders and the protocol has been found to be a useful tool for the identification of high-sulfur samples that might otherwise have gone unsuspected (Lytle et al. 2005). However, some limitations to its usefulness have been discovered. The first limitation stems from the nature of standards typically available. Figure 3 is based on published sulfur values for 25 commonly used US Geological Survey and Japan Geological Survey rock standards and illustrates that there is a deficiency of available standards in the 0.21.0% range. Rocks seem to be generally low in sulfur with only black shales and mineralized rocks having values greater than 0.2%, and then the values are much greater, leaving this gap. A second limitation is that within these standards the correlation of X-ray intensity to reported % sulfur is poor (Fig. 3). Note that the standard SDO1 has been excluded here, because its much higher sulfur value tends to dominate any correlation. Despite the fair correlation coefficient, large errors would appear using these standards for calibration. The nature of the problem is not entirely clear, but is likely to stem from two issues. First it seems likely that
14
Ripley et al.
Figure 3. Plot of X-ray intensity versus reported % sulfur in available standards using XRF analysis.
some reported values are wrong. Leoni et al. (1982) evaluated many of these same standards and found that only 5 of 19 determinations were within 10% of the literature values for sulfur, which they attributed to incorrect reference values. Second is the possibility that the standards are heterogeneous in sulfur distribution among and within bottles. Such heterogeneity could arise during storage of ground powders, with heavier sulfide grains concentrating in the bottom of the bottles. Another problem is oxidation of pyrite sulfur to native sulfur during storage, which is then volatile under the electron beam in a vacuum (Chinchon et al. 1988). It is possible that JB2, JG2, and JG3 suffer from this problem based on much lower X-ray intensities than would be expected based on the certificate values for sulfur. Instead of relying on conventional rock standards, at the University of Cincinnati a calibration curve has been constructed by using a large database from shales analyzed by both XRF and by elemental analyzer (EA, Fig. 4). For the data in this plot, a Rigaku 3070 wave-length dispersive spectrometer with a rhodium source tube was used with a Ge crystal and a sulfur Kα position of 110.8° and backgrounds at 109.8° and 111.5°. The second background is potentially overlapped by a molybdenum L line at 111.86°, but a scan of SDO1, which is rich in both sulfur and molybdenum, did not show a significant molybdenum peak at this position. At higher molybdenum values (> 100 ppm) a interference may appear, however. Also note that the detection limit is relatively high. There is no correlation of X-ray intensity to EA sulfur values of less than about 0.1% or 1000 ppm. Based on sulfur analyses of black shales using a LECO elemental analyzer the accuracy of sulfur concentrations determined using XRF methods is within 3%. An obvious limitation with this approach is that each laboratory would need to establish an independent calibration. Another is that the samples used were all pyrite-bearing shales. If sulfate-rich samples are to be analyzed, a separate calibration is likely called for. Nevertheless, this approach provides a useful and quick screening for samples that are being run for trace
Analytical Methods
15
Figure 4. Plot of X-ray intensity versus sulfur concentration for shale samples measured at the University of Cincinnati.
elements on pressed powders. A selection can then be made for further measurement by methods such as those employing an elemental analyzer.
“Kiba” method for the determination of sulfur concentration in whole rocks and glasses Sasaki et al. (1979) used a mixture of tin (II)-chloride dihydrate and strong phosphoric acid, first described by Kiba et al. (1955), to extract various forms of sulfur from geologic materials. Kiba et al. (1955) used the mixture to reduce sulfate to hydrogen sulfide, but Sasaki et al. (1979) found that it was suitable to extract sulfides and sulfates from a variety of rock types. Sasaki and Ishihara (1979) and Ueda and Sakai (1984) used the method to extract sulfur from several types of rocks, and Krouse and Ueda (1987) applied the technique to other rock types as well. More recently, Mandeville et al. (1998, 2009) used the Kiba method to extract sulfur from fine-grained S-bearing minerals in volcanic and plutonic whole rocks. Sakai et al. (1978) and Ueda and Sakai (1983) described a vacuum-based method of Kiba extraction that allowed the collection of sulfur from sulfide and sulfate, as well as CO2 for the measurement of sulfur and carbon isotope ratios. Kiba reagent is prepared according to procedures described by Sasaki et al. (1979). Sample powders and Kiba reagent are placed in a flask and heated to ~280 °C. The mixture is continuously purged by N2 and the evolved H2S is carried through a Cl-scrubber trap and into a Cd- or Zn-acetate solution for conversion to CdS or ZnS. The CdS or ZnS is then mixed with Ag-nitrate to form Ag2S. The Ag2S-bearing solution is filtered and the Ag2S collected on glass wool or filter paper. For the most precise gravimetric analysis of the produced Ag2S, collection on filter paper that is dense enough to easily liberate the Ag2S is preferred. Only ~75% of the sulfur in chalcopyrite is liberated in the atmospheric heating method, and this must be considered in the analysis of whole-rock sulfur. Sulfur in pyrite and pyrrhotite is 100%
16
Ripley et al.
liberated, making the method a particularly useful one for the determination of sulfur in many types of fine-grained rocks. Ueda and Sakai (1983) found that in the vacuum method of Kiba analysis the sulfur in copper-bearing minerals was completely liberated. The vacuum-based method is laborious with only a few samples processed in an 8-hour period. Mandeville et al. (1998) reported that replicate analyses were within ±13% of accepted sulfur concentrations for a series of standards using the atmospheric heating method.
Nuclear methods (activation analysis) for the determination of sulfur concentration in minerals, rocks and glasses Nuclear methods comprise a group of chemical analysis techniques in which a material is bombarded with neutrons, charged particles, or gamma-ray photons, inducing a nuclear reaction, and emitted radiation (typically gamma rays or beta particles) is then measured. The analysis may be fully instrumental and nondestructive, or post-irradiation chemistry may be performed to isolate the elements of interest before counting in order to improve detection limits. Nuclear methods have distinct advantages over non-nuclear methods of chemical analysis. Because the measurement relies upon a nuclear rather than chemical reaction, such techniques have few sources of error in common with other analysis methods, and are therefore valuable for intercomparison. Chemical matrix effects are generally avoided since the analysis is independent of the chemical form of the element being measured. If neutrons are utilized, the entire sample volume is analyzed since both neutrons and emitted gammas penetrate the sample. With the exception of prompt gamma ray activation analysis to measure sulfur in coals, sulfur is not routinely measured by nuclear methods. Table 1 summarizes some nuclear methods that have been used to measure sulfur in materials and reported detection limits. There are no standard procedures or standard instruments, and detection limits and the magnitude of uncertainties are dependent upon the type, energy, and flux of the bombarding particles (characteristic of the irradiation facility used), and the radiation measured. The following sections focus on a general discussion of nuclear methods that have been used for sulfur measurement, all of which may be potentially applied to glasses and minerals. Thermal neutron activation analysis (NAA) utilizes thermal neutrons (i.e., neutrons with average kinetic energy ~0.025 eV) from a nuclear reactor. Some facilities in the United States where NAA is performed include the NIST Center for Neutron Research located at the National Institute of Standards and Technology (Gaithersburg, MD), as well as reactor facilities at the University of Missouri, and the University of Texas at Austin. Neutrons are captured by elemental nuclei, resulting in the formation of product nuclei of the same element but one mass unit higher than the capture nuclide; unstable (radioactive) product nuclei then decay, usually with the release of beta particles and gamma rays, and the resulting radiation is measured (see Fig. 5). Measurement of the energy of the characteristic radiation allows elemental identification, while measurement of its intensity and comparison with standards of known composition gives quantitative analysis. Samples may be measured intact after irradiation (instrumental neutron activation analysis or INAA), or chemical separations may be performed to isolate the elements of interest before counting to achieve better detection limits (radiochemical neutron activation Figure 5. Simplified schematic illustration showing the difference between thermal neutron NAA and PGAA. 59Co captures a neutron to form compound nucleus 60Co*, which immediately de-excites with the emission of prompt gamma rays (PGAA). 60Co then decays to 60 Ni with the emission of betas and decay gamma rays (NAA).
Abundance capture nuclide, reaction cross section or particle energy, product half-life
Interfering Reaction(s)
95.02%, > 2MeV, 14.28 d
4.21%, 0.29 b, 87.2 d
0.02%, 0.23 b, 5.1 min
35
P(n,γ)32P Cl(n,α)32P
31
Cl(n,γ)35S
35
Cl(n, p)37S
37
None after radiochemistry None after radiochemistry
β−, 0.167 β−, 1.7
95.02%, 0.53 b, stable
--------
γ, 0.841 2.380
4.21%, 85.2 mbf, 12.4 s
Cl(p,pn)34mCl
35
γ, 0.511
(0.6 ng)i
(0.3 to 1) μgh
γ, 2.129, 3.305
None after radiochemistry
(50 μg/cm2)g
(0.25%)f
β+, 9.9
γ, 2.13
(100 μg)e
(10 μg)c, (0.01 μg in absence of P and Cl)d
0.1 μg when Cl < 0.1 μg
(25 μg)a (70 μg)b
Reported Detection Limits
Dams et al. (1970); bYule (1965); cMcCandless (1964); dWayman (1964); eJurney et al. (1977); fKlie and Sharma (1982); gThomas and Schweiker (1972); hStrijckmans et al. (1985); iRosseau et al. (1984)
a
S(18O,t)47V
32
95.02%, 39 MeVi, 32.6 min
4.21%, > 6.5 MeVh, 32.2 min
S(p, n)34mCl
34
95.02%, 22.5 MeVg, 0.3 s
S(p, n)32Cl
32
Charged particle activation analysis; Radiation source: cyclotron
S(n, p)34P
34
14 MeV neutron activation analysis; Radiation source: D-T neutron generator
S(n,γ)33S
32
K (0.843), Ca (0.837)
none significant
Spectral Interference Element (Energy MeV)
γ, 3.104
Radiation Measured Energies (MeV)
Prompt gamma activation analysis; Radiation source: reactor or isotopic neutron source
S(n, p)32P
32
S(n,γ)35S
34
S n,γ)37S
36
Neutron activation analysis Radiation source: reactor
Reaction
Table 1. Some activation analysis methods used in the determination of sulfur. Data are from Walker et al. (1989) unless otherwise noted.
Analytical Methods 17
18
Ripley et al.
analysis or RNAA). Sample size is dependent upon the quantity of sulfur or other elements to be measured and the nature of the matrix; samples for NAA typically range from a few hundred milligrams to a few grams. Samples and standards (usually high-purity compounds or standardized solutions of high-purity compounds evaporated onto filter paper or aluminum foil) are typically irradiated together along with flux monitors (typically metal foils) to correct for errors arising from neutron flux gradients across the irradiation vessel. Because of their stability, sulfates (typically ammonium, sodium, potassium) are good standards for sulfur analysis. There is generally no need to match matrices of samples and standards. Sulfur may be determined by measurement of the characteristic 3.104 MeV gamma ray emitted by 37S, the neutron capture product of 36S (Yule 1965; Dams et al. 1970; Shah et al. 1970) after a short irradiation on the order of 5 to 10 min. Gamma rays are measured using a high-purity intrinsic germanium detector with associated counting electronics (Knoll 1989). Although detection limits below 100 μg have been reported, the sensitivity of this method suffers from the extremely low atomic abundance and natural variability of the capture isotope, 36 S (0.0136%), and the short half-life of 37S (5.1 min), which usually precludes post irradiation and limits the counting time. Precision is therefore limited by an insufficient number of counts (poor counting statistics). Shah et al. (1970) reported mass fractions of 1% to 2% of sulfur in petroleum with 1σ uncertainties of 5% to 10% largely due to poor counting statistics. It has also been shown that variability of 36S places a “worst case limit” of ±29% on measurement accuracy unless samples and standards of known isotopic composition are analyzed (Fleming and Lindstrom 1982). A detailed discussion of other uncertainties in instrumental neutron activation analysis may be found elsewhere (Greenberg et al. 2000). Sulfur may be determined by measurement of 35S (t1/2, 87 d) produced via neutron capture of 34S (Bouten and Hoste 1962; Li and Filby 1983; Paul 2008), and by measurement of 32P (t1/2, 14.28 d) formed via the 32S(n, p)32P reaction (McCandless 1964; Souliotis 1964; Wayman 1964). Both nuclides are pure beta emitters, with maximum beta energies of 1.7 MeV for 32P and 0.167 MeV for 35S. Radiochemical separation of these nuclides from the sample matrix is necessary to avoid spectral interferences and to improve detection limits. Correction for loss of sulfur during dissolution and separation may be accomplished through the addition of a carrier (mg amounts of non-irradiated sulfur added before sample dissolution) and specific procedures to measure the recovery yield. Measurement of betas is performed either by low-background gas proportional counting (if phosphorus or sulfur is precipitated and counted in the solid state) or by liquid scintillation counting (if absorbed in a liquid), the latter being preferred for 35S to minimize uncertainties arising from self-absorption of low energy betas. Both the 32P and 35S methods suffer from interference from competing nuclear reactions, as 32P is also produced by neutron capture of 31P, and 35S is produced from chlorine via the fast neutron reaction 35Cl (n, p) 35 S. These interferences are significant: in a typical reactor thermal neutron flux the ratio of 32P produced from phosphorus versus sulfur is > 50, while 35S produced from Cl is about 35× the activity produced from an equal amount of sulfur. Accurate measurement of sulfur therefore requires measurement of and correction for the interfering element. For measurement of sulfur via 35 S, a Cl/S ratio of 0.003 results in a 10% correction (Paul 2008). Li and Filby (1983) measured sulfur mass fractions from 0.7% to 4% in fuel oils with Cl/S ratios < 0.003 with 1σ uncertainties of < 3%, but for biological materials with < 0.3% sulfur and Cl/S > 0.3, uncertainties up to 50% were reported. Paul (2008) measured 5 μg/g sulfur in a 1 g sample of ultra-high-purity iron by the 35S method, with an irradiation time of 8 h. (Paul 2008) and an expanded uncertainty of about 10%, with a 30% correction for the chlorine (also measured by NAA). A detection limit of 0.1 μg was calculated for sulfur in a sample with 0.1 μg Cl. A general discussion of uncertainties arising in RNAA procedures is given elsewhere (Paul et al. 2003; Paul 2008). Prompt gamma-ray activation analysis (PGAA), also known as prompt neutron gammaray activation analysis (PNGAA), utilizes radiation emitted during neutron bombardment rather than radiation from delayed decay of radioactive product nuclei. This technique has
Analytical Methods
19
the advantage of being able to measure elements that do not form radioactive products upon neutron capture. Gamma rays emitted during de-excitation of compound product nuclei formed upon neutron capture (known as prompt or capture gamma rays) are measured using a high-purity intrinsic germanium detector. PGAA is typically performed using neutron beams extracted from reactors, although isotopic sources (e.g., 152Cf) are sometimes used. The use of cold neutrons with guided beams results in lower background and enhanced sensitivities (Paul 1997). PGAA is regularly performed at the NIST Center for Neutron Research, which is maintained as a user facility. Uncertainties in PGAA may arise from correction for background gamma rays from neutron capture in the shielding, detector, and the surrounding environment and from the effects of neutron scattering by hydrogen (Mackey et al. 2005). For this reason, samples are most often packaged and mounted for analysis using nonhydrogenous materials. Because neutron scattering can have a dramatic effect on element sensitivities, best results are obtained by matching the hydrogen content of samples and standards. Sulfur has been measured in coals, coal fly ash, cements, and biological materials using reactor-based thermal neutron PGAA at mass fractions down to 0.2% (Jurney et al. 1977; Failey et al. 1979; Germani et al. 1980), with 1σ uncertainties of 5% to 10% and reported detection limits of about 100 μg (Jurney et al. 1977). PGAA using a 152Cf neutron source has been used for on-line determination of sulfur in coals at mass fractions from 0.5% to 6% (Vourvopoulos and Womble 1989). The availability of portable neutron sources and reactor user facilities (Paul 1997) makes PGAA a viable technique for measurement of sulfur at mass fractions above 0.1%. Sulfur has been measured by 14 MeV neutron activation analysis using a D-T (deuteriumtritium) neutron generator as the radiation source (Shani and Cohen 1976; Klie and Sharma 1982) and quantified by measurement of 34P produced by the 34S(n, p)34P reaction. A lower limit of detection of 0.25% (mass fraction) has been reported for sulfur in coal (Klie and Sharma 1982). Charged-particle activation analysis has been used to measure sulfur, with a cyclotron as the radiation source, via the reaction 34S(p, n)34mCl in metals with and without radiochemical separation of chlorine (Dabney et al. 1973; Vanecasteele et al. 1980; Strijckmans et al. 1985), and in petroleum products using the 32S(p, n)32Cl reaction (Thomas and Schweiker 1972). Detection limits at lower ppb levels have been reported. Proton-induced prompt-gamma activation analysis has been used to measure sulfur in coals at % mass fractions (Olivier et al. 1986). Heavy ion activation analysis using the 32S(18O, t)47V reaction has been used to measure sulfur in metals, ceramics, and biological samples at mass fractions down to ppb (Rousseau et al. 1984).
Sulfur analysis of minerals and glasses using the electron microprobe Electron microprobe analysis is a nondestructive in situ technique used to determine the concentrations of elements in solid materials. When a sample is bombarded by an electron beam, each element emits a set of characteristic X-rays with specific energies and wavelengths. The concentration of an element in the sample is determined by comparing the X-rays of an unknown material with the intensity of a characteristic X-ray emitted from a standard in which the concentration of the element is known. Both the sample and standard must have a flat, polished surface to ensure precise comparison of X-ray intensity between them. All major elements in the sample must be analyzed so that the matrix effect can be corrected accurately using the correction procedure (ZAF or PAP) imbedded in the data acquisition software. Most instruments are equipped with wavelength-dispersive spectrometry (WDS) and energydispersive spectrometry (EDS). However, EDS is not suitable for quantitative analysis of many rock-forming minerals, volcanic and synthetic glasses, and melt inclusions because the energy peaks of the X-rays emitted from some of the constituent elements overlap (i.e., Na-Mg, Mg-Al, Al-Si, K-Ca). Most instruments can accommodate two types of samples: 25 mm (1″) circular and 26 × 46 mm rectangular polished sections or polished thin sections.
20
Ripley et al.
The smallest spatial resolution (beam diameter) of the electron beam in most instruments is 1 μm. Under the conditions of 15 kV and 20-nA beam current; the total excitement volume in the target is about 2× larger than the beam size. So, typically a polished area of >3 μm in diameter in the sample is required for routine analysis. During electron-beam bombardment, heat is generated on the surface of the target. For example, under the conditions of 15 kV, 20 nA and 1 μm beam size, the temperature in the target may reach 50oC. If the beam current increases to 100 nA, the temperature in the target may exceed 100oC at which de-volatilization may occur. A larger beam of 5-10 μm in diameter may reduce the bombardment heat by >50%. So, typically a 5-10 μm beam is used for the analysis of volcanic and synthetic glasses, and melt inclusions. Counting time has a minor effect on the detection limit for sulfur. For example, doubling the counting time only reduces the detection limit by <10%. In contrast, doubling the beam current may reduce the detection limit by about 50%. Under the conditions of 15 kV and 20-s peak counting time, the detection limit with 95% confidence (1σ) for sulfur is about 300 ppm if the beam current is set at 20 nA. The detection limit is reduced to <100 ppm if the beam current is increased to 100 nA. Barite (BaSO4), chalcopyrite (CuFeS2), pyrrhotite (Fe1−xS) and pyrite (FeS2) are commonly used as standards for sulfur. In most instruments, sulfur analysis is made using a PET crystal on a WDS spectrometer. The concentration of sulfur in the sample is determined by comparing the X-ray intensity of SKα between the sample and standard. The X-ray peak positions of SKα from the sample and the standard may differ due to variation in valence states (e.g., Carroll and Rutherford 1988; Wallace and Carmichael 1992). The relative peak shift can be determined using the WDS spectra of the sample and standard. It does not matter what standard is used for sulfur analysis so long as the relative peak shift of SKα is included in the analytical “label” which contains calibration information. Adjustment for SKα peak shift may be avoided if the appropriate calibration standard is used; i.e., BaSO4 or CaSO4 for oxidized glasses and troilite or pyrrhotite for reduced glasses. This approach is strongly recommended if the valence state of sulfur in the material to be analyzed is known. The most common monitor sample used in trace sulfur analysis of natural and synthetic glasses, and melt inclusions is a Smithsonian standard (USNM 111240, VG-2) which is a natural basaltic glass from the Juan de Fuca Ridge. A wet-chemical analysis of this glass yielded a concentration of 1320 ± 50 ppm (Wallace and Carmichael 1992). The sulfur contents of the VG-2 standard determined by electron microprobe analysis are 1420 ± 20 ppm (Wallace and Carmichael 1992), 1348 ± 62 ppm (Thordarson et al. 1996) , 1450 ± 30 ppm (Metrich et al. 2001),1416 ± 36 ppm (de Hoog et al. 2001), 1403 ± 30 ppm (O’Neill and Mavrogenes 2002) and 1414 ± 30 ppm (Liu et al. 2007). These results indicate that the uncertainty of trace sulfur analysis by electron microprobe at this concentration level is about 5% relative.
Analyses of the sulfur concentration of minerals and glasses by secondary ion mass spectrometry (SIMS) Applications of SIMS techniques to the determination of S concentrations in geologic materials have been developed as a part of an analytical platform for volatile elements (CO2, H2O, F, S, Cl) in volcanic glasses and glass inclusions in phenocrysts. Excellent reviews of the techniques using the small-footprint Cameca f-series instruments can be found in Ihinger et al. (1994) and Hauri et al. (2002). Techniques developed at the Woods Hole Oceanographic Institution use the Cameca IMS 1280 and are different from those described earlier in that high mass resolving power (MRP) is used to mass spectrometrically separate molecular ion interferences. Since this method of S analysis is a part of a comprehensive volatile analysis, and H2O contents are determined by measuring intensities of 16O1H−, 17O− must be separated mass spectrometrically, requiring an MRP > 4700, and analyses are conducted routinely using MRP = 5500-6000 (see Fig. 6a, mass spectra on mass 17). For the determination of S contents in silicate glasses, this range of MRP is completely sufficient for separating typical molecular
c
Figure 6. Plots of high-mass resolution mass spectra for masses 17 (for 16O1H vs. 17O), 32 (for 32S vs. 31P1H, 16O2) and 30 (for 30Si vs. 29Si1H). These spectra were taken for a basalt glass with a mass resolving power of ~ 6000.
a
b
Analytical Methods 21
22
Ripley et al.
ions such as 31P1H from 32S (MRP = 3360), 16O2 from 32S (MRP = 1801), and 29Si1H from 30Si (MRP = 2842). See the spectra involving mass 30 and 32 in Figures 6b and 6c. Typically, a beam of 133Cs+ ions with currents between 1.0 and 1.5 nA is focused into a spot of ~10 μm in diameter and is rastered over a square area of 30⋅30 μm, and a mechanical aperture (Field Aperture) is placed at the position of the secondary ion image focal plane, such that only the central area of 15⋅15 μm is being analyzed. The normal incidence electron gun is used to compensate the positive electrical charge deposited by the ion beam. For each analysis, after sputter-cleaning for a period of 4 min, secondary ion intensities of 12C−, 16O1H−, 19F−, 30Si−, 32 − S , and 35Cl− are measured in an ascending mass order and the cycle is repeated 10× with a total analysis time of approximately 20 min. Intensity ratios against 30Si− are then averaged for 10 cycles. With these instrumental conditions, the sulfur ion yield in glasses of basaltic composition is approximately 150 cps/ppm/nA, indicating that the detection limit is better than 7 ppb. In analysis of typical MORB glasses with ~900 ppm S, the 32S− intensity is greater than 1.5⋅105 cps, and counting errors for accumulated intensities for 10 cycles are <0.1%, and thus negligible. The 32S/30Si ratio is reproducible to within 0.5% (2σ) for each analysis and represents a minor contributor to overall analytical uncertainty. Sulfur concentrations in unknown glasses in a specific session are obtained by a calibration curve such as shown in Figure 7, derived for glasses of basaltic or basaltic-andesite compositions. Standard glasses are all natural glasses independently documented by electron probe and/or SIMS (Hauri et al. 2002; T. Plank personal communication). The double-error linear regression (Sohn and Menke 2002) of the standard data shown in the figure results in an equation for determination of S concentration: S (ppm) = 644(+20/−59) × (32S/30Si) The error of the slope is 3.1~9.1% (2σ), and represents by far the largest contributor to the random analytical error. The accuracy of the method can be judged through detailed crosscalibration studies, such as that of Rose-Koga et al. (2010), and is believed to be of the order of 5%. The average working curve for 10 separate sessions over a period of 3 years is: S (ppm) = 651±24 × (32S/30Si) In other words, the sulfur working curve is reproducible within 3.7% (2σ). Although the detection limit of this approach is extremely low (~6 ppb), instrumental background as determined by analysis of a synthetic forsterite with no sulfur is of the order of 300 ppb. As shown in Figure 7, NIST620 and 621 standard glasses with high-silica compositions (SiO2 = 71~72 wt%) plot on the same working curve with basalts/basaltic andesites (including Fe-Ti basalts). This demonstrates that “matrix effects” are insignificant for the formation of negatively charged secondary ions of sulfur from silicate glasses.
Analyses of the sulfur concentration of minerals and glasses by laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) Laser Ablation Inductively Coupled Plasma Mass Spectrometry (LA-ICP-MS) is the technique used perhaps most widely for in situ analyses of concentrations of trace elements in minerals and glasses. The technique has a long history of development dating back to Gray (1985) and an extensive bibliography that was summarized recently in Sylvester (2008a). Figure 8 shows a schematic of the experimental design of a LA-ICP-MS system. Analyses are carried out as follows: small amounts (a few micrograms to tens of nanograms) of target materials are ablated with a pulsed laser beam in a sealed chamber, and the aerosol (vapor and entrained particulates) from the ablation is transported in a helium gas stream to an argon plasma by way of tubing connecting the ablation chamber to the ICP-MS. Once in the argon plasma, the constituent chemical elements of the aerosol are atomized and ionized and
Analytical Methods
23
Figure 7. A representative working curve for sulfur in glasses of basaltic composition, showing the relationship between sulfur contents (ppm) and secondary ion intensity ratio 32S/30Si.
Figure 8. Schematic diagram of the experimental design of a LA-ICPMS system equipped with a sector field mass spectrometer. Three slits of different widths (LR = low, MR = medium, HR = high) are used in ICP-SFMS to control the level of mass resolution.
then drawn into a high-vacuum mass analyzer where they are separated and collected according to their mass to charge ratios. In LA-ICP-quadrupole (Q) MS, a quadrupole mass filter separates the sample ions in an oscillating electric field. In LA-ICP-sector field (SF) MS, shown in the schematic of Figure 8, magnetic and static electric fields are used as the mass analyzer. ICPSFMS instruments are equipped with mechanical slits placed in the ion path at both the entrance and exit of the mass analyzer. The narrower slits provide increased mass resolution, which can avoid spectral interferences, albeit with a loss of ion transmission and thus analyte sensitivity.
24
Ripley et al.
Quantification of trace-element concentrations is typically made by external calibration against standard reference materials ablated periodically along with the unknown samples. Variations in the mass of ablated material removed by the laser from sample to sample, referred to as the “ablation yields,” are corrected by internal standardization wherein the known concentration of an element is used to normalize the count rate data for all other analytes (Longerich et al. 1996). One strength of the technique, as in all microbeam methods, is the ability to determine the spatial distributions of elements throughout a sample by targeting particular domains for analysis. For LA-ICP-MS the spatial resolution is on the order of a micrometer using either laser analyses that drill down into the sample or across its surface (Mason and Mank 2001; Kosler 2008). Sample preparation is simple compared to many other techniques, requiring only the exposure of a flat, polished, clean surface of the mineral or glass. The sample surface does not need to be coated with a conductive material such as carbon or gold as for electron-beam analysis because electrostatic charging of the sample surface does not occur during laser ablation. The sample is commonly presented to the LA-ICP-MS as a standard petrographic thin section (typically 27 mm × 46 mm and 30 μm thick) or in a mount (typically a 25 mm diameter round) containing the mineral grains or glass chips embedded in epoxy resin. Larger samples such as slabs of rock, ore or coral can be accommodated where the ablation chamber has a large-volume design. Unlike in solution-based ICP-MS, no acid digestion of samples is required, which is time consuming and requires specialized expertise and apparatus. Although the concentrations of most elements in the periodic table can be measured by LAICP-MS, sulfur is one of the more challenging because of (1) spectral interferences, principally involving oxide compounds; (2) the lack of well-calibrated reference materials for use as calibration standards, particularly for silicates; and (3) problems with blank contamination, which can become particularly troublesome at low sulfur concentrations. Hence, sulfur is not commonly included in LA-ICP-MS analyses of silicate minerals and melt inclusions (e.g., Pettke et al. 2004) and only limited data for sulfur concentrations in silicates determined by LA-ICP-MS have been reported in the literature. Instead, analyses for sulfur in minerals have mainly involved sulfides, where sulfur signals are high compared to interfering oxide signals. But even in these cases the sulfur data are commonly collected only in conjunction with sulfur concentrations determined by another technique such as EPMA, to be used for internal standard element corrections for ablation yields. Nonetheless, the method has significant potential for sulfur analyses in silicates and other non-sulfur-rich matrices, and should be considered where it is useful to link sulfur measurements directly to the same domain in a mineral or glass analyzed for other minor and trace elements by LA-ICP-MS. Table 2 lists the principal spectral interferences that have potential to present complications for LA-ICP-MS measurements of sulfur concentration, and the mass resolution of the mass spectrometer needed to resolve them (calculated as the mass of the sulfur isotope divided by the mass difference between that sulfur isotope and the interfering peak). The interferences fall into three general types: oxide and hydroxide molecular ions; charged hydrides of sulfur; and doubly charged ions of Ni and Zn. Charged helide species such as 28Si4He+, potentially formed by the use of helium as the carrier gas for the ablated aerosol, have not been detected in previous studies (Guillong et al. 2008a). There is an argon interference (derived from the argon gas of the ICP) on the least abundant isotope of sulfur, 36S, but this isotope is so minor (0.015%) that for practical purposes it is hardly useful for sulfur measurements by LA-ICP-MS in any case. The analyst has three options for dealing with spectral interferences on sulfur in LAICPMS. First, if one is using a SFMS instrument operated in low-resolution mode or a QMS instrument, sulfur isotopes cannot be resolved from their spectral interferences. In these cases, one can assume that the large majority of the interferences are derived from the gas background of the LA-ICP-MS system and simply subtract off the background measurements prior to each
Analytical Methods
25
Table 2. Some potential spectral interferences on sulfur isotopes and the mass resolution needed to distinguish them in LA-ICP-MS. Isotope 64
Mass
Ni++ Zn++ 32 S+ 16 O–16O+ 14 N-18O+
31.963985 31.964574 31.972071 31.989830 32.002234
66
Zn++ 33S+ 32 1 S- H+ 16 O–17O+ 16 O-16O-1H+
32.963019 32.971458 32.979896 32.994047 32.997655
68
Zn++ S+ 33 1 S- H+ 16 O–18O+
33.962424 33.967867 33.979283 33.994075
36
35.967081 35.967546
64
34
S+ Ar+
36
S isotopic abundance (%)
Required resolution (M/Dm) 3957 4268
95.04 1802 1061 3910 0.75 3911 1461 1260 6247 4.20 2978 1297 0.015 77419
Notes: The isotopic mass data are from Audi et al. (2003). The percent natural abundance data for sulfur are from De Laeter et al. (2003). Mass resolution is calculated as the mass number (M) divided by the difference in mass between neighboring peaks (Dm) where the valley between the peaks is less than 10% of the peak height of the analyte to be measured.
ablation from the intensities measured for the sulfur isotopes during ablation. For analyses of silicate and other oxygen-rich minerals and glasses, there would be some oxide production attributable to the introduction of oxygen from the sample matrix itself, but this contribution appears to be minor (discussed below) and, in any case, is accounted for, if the external calibration sample has a similar bulk composition as the unknown minerals or glasses. Figure 9 shows mass scans made over m/z = 32, 33 or 34 prior to ablation, in the gas background alone, and then later, during ablation of the NIST SRM 610 synthetic soda-lime aluminosilicate glass, which contains only ~0.06 wt% sulfur (Jochum et al. 2006; Guillong et al. 2008b). The data were collected using an ICP-SFMS operated in low-resolution mode (=300). ICP-QMS instruments operate at similar mass resolutions. The expected positions of each of the sulfur isotopes and potential interfering peaks are indicated. One can see from the scans in Figure 9 that the gas background count rates at m/z = 32, 33 or 34 are very large in NIST SRM 610 relative to the count rates produced from actual sulfur in the glass. For instance, ~15,000 of 25,000 total counts (60%) on m/z = 34 may be attributed to the gas background. Thus, this approach has not been used widely for LA-ICP-MS analysis of sulfur-poor, silicate materials. Instead it has been employed for analyses of sulfur in sulfide minerals, where the S+/ O2+ ratio is very high, and in particular where sulfur was measured only as the internal standard element for ablation-yield corrections (e.g., Ballhaus and Sylvester 2000; Cabri et al. 2003; Wohlgemuth-Ueberwasser et al. 2007). Norman et al. (2003) used this method to measure sulfur concentrations of 1.7 to 37 wt% in sulfide ore reference materials prepared with lithium borate as fused glass discs, applying 57Fe as the internal standard isotope, on 100-μm diameter ablation spots. They reported detection limits for sulfur of 0.55 wt%, precisions of better than
26
Ripley et al.
Figure 9. Plots of mass scans over the 32S, 33S and 34S peaks in the gas background (A, B, C) and during laser ablation of the NIST SRM 610 glass (D, E, F) in low-resolution (300), EScan (electric) mode of a Thermo-Scientific Element-XR ICP-SFMS. The ablation was made using a Lambda Physik Compex 110 ArF excimer GeoLas system in helium gas with a 69 μm laser spot moved at 1 μm/s in a line raster, at 4 J/cm2 laser fluence and 10 Hz repetition rate, for 100 s. Note the scale change in counts per second (cps) intensity between the scans for 32S and those for 33S and 34S. The nominal masses of the ions are listed in Table 2.
10% and, for most samples, accuracies of better than 13%. Even though the 32S isotope is an order of magnitude more abundant than the 34S isotope, the 34S isotope has been preferred for these low-resolution measurements because the 16O18O+ interference at m/z = 34 is two orders of magnitude smaller than the 16O16O+ at m/z = 32 (see Fig. 10). A second method for LA-ICP-MS analyses of sulfur is to make the measurements with an SFMS instrument operating in a higher-resolution mode so that the peaks for the sulfur isotopes
Analytical Methods
27
Figure 10. Plots of mass scans over the 32S, 33S and 34S peaks in the gas background (A, B, C) and during laser ablation of the NIST SRM 610 glass (D, E, F) in medium-resolution (4000) LA-ICP-SFMS, using the same instrumentation and data acquisition parameters as in Figure 9. Note the logarithmic scale compared to the linear scale used in Figure 9. The peaks of the principal interferences are labeled.
can be separated from the peaks of the interfering ions. Figure 10 shows mass scans made for gas background and NIST SRM 610 in medium resolution, analogous to those made in low resolution in Figure 7. The scans indicate that, in NIST SRM 610, the largest interferences on 32 33 S, S and 34S are 16O16O+, 16O16OH+ and 16O18O+ respectively. The other potential interferences listed in Table 1 are minor to negligible. All three sulfur peaks can be distinguished from their interferences clearly in medium resolution. There is, however, a significant decrease in analyte sensitivity (about a factor of 15, comparing Figs. 9 and 10) between the low- and mediumresolution modes of SFMS because, as mentioned above, the narrower slits used for medium resolution transmit fewer ions to the mass analyzer. Thus it is preferable to use the largest isotope of sulfur, 32S, for concentration measurements in medium resolution. Axelsson and
Ripley et al.
28
Rodushkin (2001) used this method to measure sulfur concentrations in sphalerite by LAICP-SFMS, in line scans made with a 40-μm wide laser beam. Employing 66Zn as an internal standard isotope, they reported sulfur concentrations of 28-33 wt%, with precisions of better than 8% and accuracies better than 15%. This method has also been used extensively for sulfur isotope measurements in sulfides and sulfates by LA-ICP-SFMS using multicollector arrays, which allow 32S/34S and 33S/34S ratios to be measured very precisely (Bendall et al. 2006; Mason et al. 2006; Craddock et al. 2008). In order to compare the quality of data and detection limits that may be expected with these two approaches more fully, particularly for sulfur-poor minerals and glasses, which has not been well-documented in the literature, Table 3 shows previously unpublished results of sulfur analyses for the NIST SRM 612 and 614 synthetic soda-lime aluminosilicate glasses measured by LA-ICP-SFMS in low- (300) and medium- (4000) resolution modes. NIST SRM 610 glass was used as the external calibration standard and 44Ca was the internal standard isotope. The concentration of sulfur in NIST SRM 610 is not well established (discussed below) but the value Table 3. Determinations of sulfur concentrations and detection limits in NIST SRM 612 and 614 glasses in low and medium resolution by LA-ICP-SFMS. Isotope:
32
33
S
S (ppm) concentration
S (ppm) detection limit
34
S
S (ppm) concentration
S (ppm) detection limit
S
S (ppm) concentration
S (ppm) detection limit
NIST SRM 612/ Low Resolution Analysis 1 271 Analysis 2 356 Analysis 3 220 Mean ± SD 283 ± 69
132 158 137
315 291 268 291 ± 24
136 193 99
329 362 337 343 ± 18
34 21 15
NIST SRM 612/ Medium Resolution Analysis 1 364 Analysis 2 364 Analysis 3 354 Mean ± SD 361 ± 6
12 20 12
441 434 334 403 ± 60
97 175 125
380 384 372 379 ± 6
58 69 50
NIST SRM 614/ Low Resolution Analysis 1 461 Analysis 2 260 Analysis 3 331 Mean ± SD 351 ± 102
216 94 110
424 213 276 304 ± 108
142 80 162
339 308 328 325 ± 15
27 27 34
NIST SRM 614/ Medium Resolution Analysis 1 323 Analysis 2 334 Analysis 3 318 Mean ± SD 325 ± 8
14 13 13
333 297 375 335 ± 39
97 130 134
379 334 328 347 ± 28
51 61 66
Instrumentation: Thermo-Scientific Element-XR ICP-SFMS and Lambda Physik Compex 110 ArF (193 nm wavelength, 20 ns pulse width) excimer GeoLas system, located at Memorial University of Newfoundland. ICP-MS settings: Rf forward power=1150 W; <2 W reflected power; Pt guard electrode on; argon gas flows (coolant = 16 l/min, make-up = 0.704 l/min, auxiliary = 1.03 l/min); nickel sample (1.1 mm aperture) and skimmer (0.8 mm aperture) cones; low (300) and medium (4000) resolution. Ablation conditions: Helium ablation chamber and carrier gas (1.073 l/min); 69 μm laser spot moved at 1 μm/s in a line raster; 4 J/cm2 laser fluence; 10 Hz repetition rate. ICP-MS acquisition parameters: Single ion measuring; 1 point per peak; 5 ms per peak settling time; 5 ms per peak dwell time; 30 s gas background measurement; 60 s laser ablation measurement. External calibration standard: NIST SRM 610 = 693 ppm S. Internal standard element: Ca using the 44Ca isotope and 11.50 wt% CaO in NIST 610; 11.90 wt% CaO in NIST 612; 11.90 wt% CaO in NIST 614. Detection limits are calculated using the equation given by Longerich et al. (1996).
Analytical Methods
29
(693±69 ppm) of Jochum et al. (2006), determined at the Carnegie Institution of Washington using a CAMECA IMS6F ion probe (SIMS), was assumed for the concentration calculations because the same study also reported sulfur determinations by SIMS for both the NIST SRM 612 (350±35 ppm) and NIST SRM 614 (306±31 ppm) glasses, allowing for direct comparison here. The sulfur results for NIST SRM 612 and 614 by LA-ICP-SFMS are generally similar to those of Jochum et al. (2006) in both low-and medium-resolution modes (Table 3). As expected, measurement of the 34S isotope gives superior results in low resolution. Based on the lowresolution measurements of 34S, the detection limits for the sulfur analyses are calculated to be 15 to 34 ppm, the measurement reproducibility is ~5% (343±18 ppm for NIST SRM 612; 325±15 ppm for NIST SRM 614) and the accuracy compared to the Jochum et al. (2006) values is 2% (612) and 6% (614). The data also confirm that the 32S isotope gives the best results in medium resolution. Based on medium-resolution measurements of 32S, the detection limit for the sulfur analyses is calculated to be 12 to 20 ppm, the measurement reproducibility is ~2% (361±6 ppm for NIST SRM 612; 325±8 ppm for NIST SRM 614) and the accuracy compared to the Jochum et al. (2006) values is 3% (612) and 6% (614). The figures of merit suggest that in situ measurements for sulfur concentration in silicate glasses by LA-ICP-SFMS in medium-resolution mode are improved compared to those in low-resolution mode, but only marginally so. Detection limits are limited by the sulfur intensities of the gas background, with, for instance, some 13,000 cps of 32S and ~700 cps of 34S found in the gas background of the medium- resolution experiments run here. A third approach that could be used to reduce spectral interferences for sulfur measurements by laser ablation is collision and reaction cell (CRC) technology, which is now available on most commercial ICP-QMS instruments. The CRC is a device commonly placed behind the ion-extraction interface and in front of the mass analyzer of the QMS, into which a reactive gas is added, allowing collisions and reactions with analyte and polyatomic ions produced in the ICP torch. In the case of sulfur analyses, xenon gas (sometimes plus H2 gas) is added to the CRC, leading to the breakdown of O2+ and NO+ molecular ions through transfer reactions to produce Xe+ and XeH+ with a commensurate improvement in the S+/O2+ and S+/NO+ ratios (Rowan and Houk 1989; Mason et al. 1999). In solution mode, Mason et al. (1999) reported a loss in sulfur sensitivity of 10-20% with the addition of Xe, but background interferences were reduced to <1% of their original levels so that S+/O2+ was improved by a factor of 10. This allowed measurements of 50 mg sulfur per liter of dilute acid solution (~50 ppm) with 2% internal precision. Surprisingly, the method has not been well tested for laser analyses of sulfur concentration. Mason et al. (2006) however compared the CRC approach to the mediumresolution SFMS method for in situ measurement of sulfur isotope ratios in sulfides and sulfates using LA-multicollector-ICP-MS. They found that the two methods gave 34S/32S ratios of comparable precision and accuracy, suggesting that the CRC approach could provide in situ sulfur concentration data of similar quality to the medium-resolution SFMS method. LA-ICP-MS is a technique remarkably free of matrix effects so that an exact matrix match of the external calibration standard to the unknown is not usually necessary for trace-element measurements unless analytical uncertainties less than about 5% are required (Sylvester 2008b). For sulfur measurements in sulfide minerals and glasses, it is likely that almost any sulfide material with a known sulfur concentration would be an adequate calibration standard. The LA-ICP-MS community has not yet converged on a generally accepted sulfide material for this purpose, but several sulfides have been proposed as international reference materials for in situ trace-element analyses and in principle could be used for calibration of sulfur measurements in sulfides (e.g., Sylvester et al. 2005; Wohlgemuth-Ueberwasser et al. 2007). Probably the most widely distributed and currently available material is USGS MASS-1, a synthetic sulfide pressed powder pellet (Wilson et al. 2002).
30
Ripley et al.
It is not recommended to use sulfide calibration standards for analyses of silicate minerals and glasses. This is because ablation of the percent level concentrations of sulfur present in sulfide materials will lead to elevated sulfur backgrounds in the ICP-MS instrument due to contamination “memory effects” in the sample chamber, transfer tubing, ICP torch and cones. The sulfur concentrations of many natural silicates are expected to be at or below levels of 10s of ppm (e.g., Jochum et al. 2006) so that elevated background memory effects of sulfur can compromise detection limits severely. Also, there seems to be a substantial matrix effect between sulfides and silicates with regard to the relative sensitivities between sulfur and transition metals. Sylvester (2008b) found that the sensitivities of sulfur in NIST SRM 610 glass and USGS MASS-1 were very similar when ablated under identical conditions, but the sensitivities for Fe, Cu, Zn, Ag, and Pt were all enriched by some 50-60% in the sulfide relative to the silicate. The origin of this matrix effect is not understood. Sylvester (2008b) suggested that the sulfur signals in NIST SRM 610 were elevated anomalously by oxide interferences produced from oxygen present in the silicate itself; however, the medium-resolution experiments on NIST SRM 610 in Figure 9 suggest that this is not the case, and there must be another cause of the discrepancy between the intensities of sulfur and the other metals. The practical problem caused by the discrepancy comes in identifying an element to be used as the internal standard for sulfur measurements in silicates calibrated against a sulfide as the external standard. The transition metals are one group of elements present in substantial quantities in most sulfides and some silicates, but the matrix effects described above should discourage their use as internal standards for sulfur measurements in silicates calibrated against sulfides. So which reference material should be used for external calibration of sulfur measurements in silicates by LA-ICP-MS? The most obvious choice is the NIST SRM 610 glass because it is already used widely in LA-ICP-MS as a calibration standard for a large number of trace elements, and it contains somewhat more sulfur than the NIST SRM 612 and 614 glasses (Jochum et al. 2006), providing slightly improved measurement precision. Guillong et al. (2008a) showed that the sulfur concentrations in a scapolite specimen ([Na,Ca]4[Al3Si9O24]Cl with 2300 ppm sulfur) determined by EPMA could be reproduced to within 5% by LA-ICP-MS using NIST SRM 610 as the external standard. The principal drawback of NIST SRM 610 is that attempts to determine its sulfur concentration using various methods have given a range of rather imprecise values: 456±32 ppm by laser plasma ionization mass spectrometry (Rocholl et al. 1997); 550 ± 25 ppm (Ihinger et al. 1994), 560±60 ppm (Evans et al. 2008) and 693±69 ppm (Jochum et al. 2006) by SIMS; and 550±40 ppm (Guillong et al. 2008a) and 634±151 ppm (Fitzpatrick et al. 2008) by LA-ICP-SFMS. Mandeville (pers. comm.) reports that a value of 550±50 ppm sulfur for NIST SRM 610 is typically returned by electron microprobe analyses when anhydrite and barite are used as calibration standards and all of the sulfur is assumed to be present as the SO4 species. It is likely that the range of reported values for NIST SRM 610 largely represents calibration and measurement errors of the methods used, but it is also possible that some of the range reflects actual intrinsic variability in the sulfur concentrations of NIST SRM 610 material on the tens of micrometer scale. Further work is clearly needed to calibrate the sulfur concentration and micrometer-scale variability in NIST SRM 610 more precisely so that the glass can be used more effectively for LA-ICP-MS analyses of sulfur in silicates. NIST SRM 610 may also prove to be an appropriate calibration standard for analyses of sulfur in mineral matrices beyond silicates such as sulfates, phosphates and oxides but virtually no work has been done in this area. LA-ICP-MS measurements do not commonly include a correction for blank contamination of the sample other than for the contaminants present in the carrier and torch gases. The gasbackground data are collected before the laser is fired, and subtracted from the intensities measured while the sample is ablated. Contaminants present in the ablation chamber and
Analytical Methods
31
delivery system to the ICP-MS that are removed only during ablation are usually minor; also they are commonly added to both the ablated calibration standard and unknown materials in similar amounts so that they are accounted for satisfactorily in the concentration calculations. But this is not always the case and when the contamination of an element is large relative to amounts being measured, the lack of a proper, total procedural blank correction in LA-ICP-MS can result in significant errors. Guillong et al. (2008a) reported that, unfortunately, this can be the case for low-level sulfur measurements by LA-ICP-MS. They found that optical-grade fused silica and high-purity quartz containing less than 2 ppm sulfur gave apparent sulfur concentrations of approximately 200 ppm when measured by LA-ICP-MS. They showed that the sulfur contaminant was present somewhere in the ablation chamber, perhaps in the sample mount itself, and that the contaminant was released into the ablation stream by either mechanical or photochemical processes occurring during ablation. It remains to be determined whether their results are common to many LA-ICP-MS systems employed today. If so, it will be necessary to include a total procedural blank correction for low-level sulfur measurements by LA-ICP-MS. Ideally the blank would be a sulfur-free material, matrix-matched to the unknown samples, and ablated under similar conditions to the unknowns. The detection limits of the technique would depend largely on the size of the procedural blank for a given LA-ICP-MS system.
Determination of the sulfur concentration in fluid inclusions by LA-ICP-MS Inductively-coupled mass spectrometry is not normally a method of choice for determination of anion components like sulfur or halogens. The yield of positive S+ ions is poor, leading to modest detection limits in aqueous solutions (Bandura et al. 2002) or laser-ablation aerosols of solids. Two main exceptions include samples containing sulfur as a major element (Craddock et al. 2008) and the bulk analysis of fluid and melt inclusions. Quantitative analysis of sulfur in fluid inclusions has been a serious challenge for years, especially in studies of hydrothermal ore deposits, where sulfur is one of the most important elements (this volume, Simon and Ripley 2011). Sulfur not only controls chalcophile metal solubility and the precipitation of sulfide minerals (Richards and Larson 1998), but also is a major redox couple in hydrothermal fluids (Ohmoto and Lasaga 1982) and a key ligand for effective hydrothermal transport of Au and Cu in aqueous fluids and magmatic vapor (Stefansson and Seward 2004; Seo et al. 2009; Zajacz and Halter 2009). Sulfur and halogens trapped in fluid inclusions from Archaean sediments and hydrothermal systems imply secular variations in ocean composition and evolution of biological activities of early Earth (Foriel et al. 2004). Because of technical limitations, sulfur analyses of fluid inclusions have so far been restricted to bulk-extraction methods (Banks and Yardley 1992; Bray and Spooner 1992), but such results are commonly difficult to interpret in terms of geological processes. Quantitative analysis of sulfur in individually targeted, single fluid inclusions is possible by micro proton-induced X-ray emission (μ-PIXE) at high concentrations (Ryan et al. 1993) or the synchrotron XRF technique (Cauzid et al. 2007). However, these techniques are limited by partial overlap of the sulfur and chlorine peaks, which hampers sensitivities of sulfur detection. Absorption of low-energy X-ray photons in the host material requires fluid inclusions that are very shallow (less than 5 μm) below the sample surface and have well-defined geometry. Thus, typical detection limits of 0.1-1% sulfur in normal (20-50 μm) inclusions can be obtained. Laser ablation (LA-) ICP-MS is relatively cost-effective, fast and the most sensitive method suitable for multi-element analysis in fluid inclusions (Gunther et al. 1998; Heinrich et al. 2003) and has recently, after initial difficulties, been established for sulfur quantification in single fluid inclusions (Guillong et al. 2008a; Seo et al. 2009). LA-ICP-MS instrumentation (ETH-Geolas Excimer Laser Ablation propototype optics, coupled to a Perkin Elmer Elan 6100 DRC) and evaluation procedures follow those described
32
Ripley et al.
in Gunther et al. (1998) and Heinrich et al. (2003), with the possible addition of minor H2 to the ablation stream to enhance sensitivity for heavy metals during simultaneous multielement analysis of single inclusions (Guillong and Heinrich 2007). After subtraction of signal contributions from the host mineral, element ratios in the fluid are calibrated by reference materials (NIST SRM 610; which has a revised sulfur concentration of 550 ± 40 μg/g; Guillong et al. 2008a). Absolute concentrations of elements are calculated by internal standardization against Na, based on prior microthermometric determination of salinity corrected for other salts (Gunther et al. 1998; Heinrich et al. 2003). Two specific difficulties pertain to sulfur analysis by LA-ICPMS: the potential massinterference of all S+ isotopes with bi-atomic O2+ ions; and a previously unknown elementspecific contamination behaviour. To test the magnitude and significance of mass-interference for inclusions ablated from oxygen-rich host minerals, 30 cogenetic fluid inclusions in quartz were analyzed with two different mass-spectrometers. Constant microthermometric salinity (42.4±1.2 NaCl equiv. wt%) and textural occurrence along a single healed fracture indicates identical composition of all inclusions in this assemblage. Half of the inclusions were analyzed with a sector-field mass spectrometer (Thermo – Finnigan ELEMENT-2), which has sufficient mass resolution to resolve polyatomic interferences (16O16O+) from sulfur (32S+). The remaining inclusions were analyzed with the quadrupole ICP-MS. Analysis of transient signals and the consistency of resulting sulfur concentration between both data sets show that the precision and accuracy of sulfur quantification are not primarily limited by the mass interferences of O2, down to sulfur concentrations of ~100 μg/g (Guillong et al. 2008a). The scatter of 35% relative standard deviation (RSD) among all 30 sulfur analyses provides a measure of maximum uncertainty. Analytical accuracy has been confirmed by analyzing synthetic fluid inclusions with known sulfur concentration (Seo et al. 2011). Limits of detection for sulfur in fluid inclusions depend on inclusion size and fluid density, but decrease to 100 μg/g for large (60 μm) inclusions of liquid-like density (Guillong et al. 2008a). Greatest practical care must be devoted to an element-specific and ablation-dependent contamination effect, which requires special cleaning of the ablation chamber and careful monitoring of the baseline on 32S+. At ETH, we found elevated sulfur (and also chlorine) signals during the ablation of sulfur-free host minerals, which is evident from a rise in 32S+ intensity correlating with that on 26Si+ in the case of quartz. Laser-induced sulfur counts can also result from simple UV-illumination of the sample chamber alone. This signal comes from UV-induced ‘desorption’ of sulfur attached to the inner surfaces of the sample chamber, as tested in detail by Guillong et al. (2008a). Careful cleaning of the entire sample chamber using polishing paste and weak acids allows reduction of this contamination to less than an equivalent of 20 μg of sulfur per gram of ablated quartz. To subtract any remaining contamination during data evaluation with our program SILLS (Guillong et al. 2008b), we always set the sulfur baseline immediately before and after the transient sulfur peak from each fluid inclusion, as indicated in Figure 11. In summary, sulfur quantification in fluid inclusions by laser ablation ICP-MS is now possible with particular precautions, as part of the multi-element capability of this powerful method. Sulfur contents of fluid inclusions can be determined at geochemically useful accuracy and precision (<35% RSD in single assemblage), reaching detection limits (100 μg/g for 30-μm diameter brine inclusions) well below the sulfur concentration in many natural hydrothermal fluids.
SUMMARY The analytical methods described above are all in use for the detection of sulfur concentration in a range of materials (Table 4). Choice of the most appropriate method is a function of the type of material to be analyzed, sample availability, and to some extent the
Analytical Methods
33
Figure 11. Plots of gas blank-corrected transient signals (cps) of quartz-hosted single brine inclusions (30 μm size) in a molybdenite-quartz vein from the Bingham Canyon deposit, USA by LA-ICP-MS. (A) Good signals of 32S, 29Si, 23Na, and 32S/29Si ratios in fluid inclusion after cleaning of ablation chamber, measured in Feb. 2010. (B) High and unstable S/Si background ratios during quartz ablation (Feb. 2008) become lower and stable (Feb. 2010) after removal of S contamination by cleaning of the ablation chamber, leading to a much better detection limit (LOD).
expected sulfur concentration. A review of the literature shows that the method of choice for bulk sulfur analyses of rocks (including ores) involves an elemental analyzer coupled with an infrared detection cell (e.g., LECO, ELTRA systems). The “Kiba” method of sulfur extraction followed by gravimetric analysis allows the detection of sulfur in the range of 50-100 ppm, but care must be exercised in the collection of produced sulfide and final weighing. The method is particularly suitable for the collection of finely disseminated sulfides in rocks for isotopic analysis. For routine determination of sulfur in minerals and glasses the electron microprobe remains the most utilized technique, unless sulfur is present at concentrations less than 50 to 100 ppm. For the detection of low-levels of sulfur in minerals and glasses Secondary Ion Mass Spectrometry may become the norm. Only a few labs in the world at present are capable of exploring sulfur determination by SIMS; the future for routine sulfur analysis may be a function
Ripley et al.
34
Table 4. Comparison of methods for the determination of S concentration in minerals, rocks, glasses, and fluid inclusions Normal Detection Limit
Remarks
powdered minerals, rocks, glasses
100 ppm; minimum 0.6 ppm in a homogeneous 1-gram sample
standard choice for analyses of rocks and glasses
EA coupled to a mass spectrometer
powdered minerals, rocks, glasses
50 to 100 ppm
dissolved S in glasses may be difficult to extract
Kiba extraction
powdered minerals, rocks, glasses
50 to 100 ppm
accurate determinations require careful gravimetric procedures
X-ray fluorescence
powders (in form of fused glass disks or pressed disks); polished minerals
1000 ppm
high detection limit and poor precision has limited usefulness
Nuclear methods
powders or solid surfaces that can be irradiated
5 ppm in a 1 gram sample
uncertainty may be high
Electron microprobe
polished surfaces of minerals and glasses
100 ppm
match of sulfide and sulfate standards to unknowns may be critical
SIMS
polished surfaces of minerals and glasses
6 ppb, but variable due to instrument background
background may be as high as 300 ppb; accuracy of low-level standards may be a concern
LA-ICP-MS
polished surfaces of minerals or glasses; fluid inclusions
5 to 30 ppm
spectral interference, blank contamination and lack of appropriate standards for analysis of S in silicates have been limiting factors
Technique
Sample Types
EA with infrared detection cell
of instrument availability. Nuclear methods of sulfur analyses offer the potential for low-level sulfur determination, but standards and samples with known isotopic compositions must be run together to reduce large variations in accuracy. Laboratory availability will limit the routine use of nuclear methods of analyses for most geologic materials. Sulfur analyses using LA-ICP-MS technology is in its infancy, but the potential to determine low-levels of sulfur concurrently with other trace-elements ensures that research will continue to be directed at making sulfur analyses via LA-ICP-MS routine. Analyses of sulfur in fluid inclusions via LA-ICP-MS will advance research on ore genesis and volatile interaction with rocks. Microanalytical tools involving the ion probe and LA-ICP-MS offer an additional benefit of potential determination of sulfur isotope ratios (see the contributions in this volume by Marini et al. 2011 and Simon and Ripley 2011). Samples analyzed via mass spectrometry following high-temperature combustion offer a similar benefit, but the extraction of sulfur species dissolved in glass by this method has proven difficult with most EA systems.
Analytical Methods
35
ACKNOWLEDGMENTS We would like to thank Nicole Metrich and Charles Mandeville for reviews that improved the presentation of this chapter. The editorial assistance of Jim Webster and Harald Behrens is appreciated. Paul Sylvester expresses appreciation to Mike Tubrett and Kate Souders for the LA-ICP-MS measurements of sulfur in the NIST glasses.
REFERENCES Audi G, Wapstra AH, Thibault C (2003) The 2003 atomic mass evaluation. (II). Tables, graphs and references. Nucl Phys A 729:337-676 Axelsson MD, Rodushkin I (2001) Determination of major and trace elements in sphalerite using laser ablation double focusing sector field ICP-MS. J Geochem Explor 72:81-89 Bach W, Erzinger J (1995) Volatile components in basalts and basaltic glasses from the EPR at 9°30 N. Proc Ocean Drill Program Part B Sci Results 142:23-29 Ballhaus C, Sylvester P (2000) Noble metal enrichment processes in the Merensky reef, Bushveld complex. J Petrol 41:545-561 Bandura DR, Baranov VI, Tanner SD (2002) Detection of ultratrace phosphorus and sulfur by quadrupole ICPMS with dynamic reaction cell. Anal Chem 74: 1497-1502. Banks DA, Yardley BWD (1992) Crush-leach analysis of fluid inclusions in small natural and synthetic samples. Geochim Cosmochim Acta 56: 245-248. Bendall C, Lahaye Y, Fiebig J, Weyer S, Brey GP (2006) In situ sulfur isotope analysis by laser ablation MCICPMS. Appl Geochem 21:782-787 Bouten P, Hoste J (1962) The determination of sulfur and phosphorus in steel by neutron activation analysis. Anal Chim Acta 27:315-319 Bower NW, Gladney ES, Ferenbaugh RW (1986) Critical comparison of X-ray fluorescence and combustion infrared methods for the determination of sulphur in biological matrices. Analyst 111:105-106 Bray CJ, Spooner ETC (1992) Fluid inclusion volatile analysis by gas-chromatography with photoionization micro-thermal conductivity detectors - applications to magmatic MoS2 and other H2O-CO2 and H2O-CH4 fluids. Geochim Cosmochim Acta 56: 261-272 Cabri LJ, Sylvester PJ, Tubrett MN, Peregoedova A, Laflamme LHG (2003) Comparison of LAM-ICP-MS and Micro-PIXE results for palladium and rhodium in selected samples of Noril’sk and Talnakh sulfides. Can Mineral 41:321-329 Carroll MR, Rutherford M.J (1988) Sulfide speciation in hydrous experimental glasses of varying oxidation state: results from measured wavelength shifts of sulfur X-rays. Am Mineral 73: 845-849 Cauzid J, Philippot P, Martinez-Criado G, Menez B, Laboure S (2007) Contrasting Cu-complexing behaviour in vapour and liquid fluid inclusions from the Yankee Lode tin deposit, Mole Granite, Australia. Chem Geol 246: 39-54 Chinchon JS, Lopez-Soler A, Traveria A, Vaquer R (1988) X-ray fluorescence analysis of samples with elemental sulphur: Effect of sulphur sublimation. X-Ray Spectrom 17: 217-218 Craddock PR, Rouxel O, Ball L, Bach W (2008) Sulfur isotope measurement of sulfate and sulfide by highresolution MC-ICP-MS. Chem Geol 253:102-113 Dabney SA, Swindle DL, Beck JN, Francis G, Schweikert EA (1973) On the determination of sulfur by charged particle activation analysis. J Radioanal Nucl Chem 16:375-383 Dams R, Robbins JA, Rahn KA, Winchester JW (1970) Nondestructive neutron activation analysis of air pollution particulates. Anal Chem 42:861-867 de Hoog JM, Mason PD, van Bergen MJ (2001) Sulfur and chalcophile elements in subduction zones: constraints from a laser ablation ICP-MS study of melt inclusions from Galunggung Volcano, Indonesia. Geochim Cosmochim Acta 65:3147-3164 de Hoog JCM, Hattori HK, Hoblitt RP (2004) Oxidized sulfur-rich mafic magma at Moutn Pinatubo, Philippines. Contrib Mineral Petrol 146:750-761 De Laeter JR, Böhlke JK, De Bièvre P, Hidaka H, Peiser HS, Rosman KJR, Taylor PDP (2003) Atomic weights of the elements: Review 2000 (IUPAC Technical Report). Pure and Appl Chem 75:683-800 Diana M, Gabrielli N, Ridolfi S (2007) Sulfur determination on stone monuments with a field-transportable EDXRF system. X-Ray Spectrom 36:424-428, doi: 10.1002/xrs.1005 Elswick ER, Hower JC, Carmo AM, Sun T, Mardon SM (2007) Sulfur isotope geochemistry of coal and derived combustion products: an example from an Eastern Kentucky mine and power plant. Appl Geochem 22:2065-2077
36
Ripley et al.
Evans KA, O’Neill HStC, Mavrogenes JA (2008) Sulphur solubility and sulphide immiscibility in silicate melts as a function of the concentration of manganese, nickel, tungsten and copper at 1 atm and 1400 °C. Chem Geol 255:236-249 Failey MP, Anderson DL, Zoller WH, Gordon GE (1979) Neutron-capture prompt γ-ray activation analysis for multielement determination in complex samples. Anal Chem 51:2209-2221 Fitzpatrick AJ, Kyser K, Chipley D, Beauchemin D (2008) Fabrication of solid calibration standards by a sol-gel process and use in laser ablation ICPMS. J Anal At Spectrom 23:244-248 Fleming RF, Lindstrom RM (1982) Limitations on the accuracy of sulfur determination in NAA. Trans Am Nucl Soc 41:223 Foriel J, Philippot P, Rey P, Somogyi A, Banks D, Menez B (2004) Biological control of Cl/Br and low sulfate concentration in a 3.5-Gyr-old seawater from North Pole, Western Australia. Earth Planet Sci Lett 228:451463 Gazulla MP, Gómez MP, Orduña M, Rodrigo M (2008) New methodology for sulfur analysis in geological samples by WD-XRF spectrometry. X-ray Spectrom 38:3-8, doi 10.1002/xrs.1092 Germani MS, Gokmen I, Sigleo AC, Kowalczyk GS, Olmez I, Small AM, D.Anderson DL, Failey MP, Gulvali MC, Choquette CE, Lepel EA, Gordon GE, Zollar WH (1980) Concentrations of elements in the National Bureau of Standards’ bituminous and subbituminous coal Standard Reference Materials. Anal Chem 52:240-245 Giblin AE, Likens GE, White D, Howarth RW (1990) Sulfur storage and alkalinity generation in New England lake sediments. Limnol Oceanogr 35:852-869 Gibson EK Jr, Andrawes FF (1978) Sulfur abundances in the 74001/74002 drive tube core from Shorty Crater, Apollo 17. Lunar and Planetary Science Conference, 9th, Houston, TX, March 13-17, 1978, Proceedings. Volume 2. (A79-39176 16-91) New York, Pergamon Press, Inc., p 2011-2017 Giles HL, Hurley PW, Webster HWM (1995) Simple approach to the analysis of oxides, silicates and carbonates using X-ray fluorescence spectrometry. X-ray Spectrom 24:205-218 Gray AL (1985) Solid sample introduction by laser ablation for inductively coupled plasma source mass spectrometry. Analyst 110:551-556 Greenberg RR, Lindstrom RM, Simons DS (2000) Instrumental neutron activation analysis for certification of ion-implanted arsenic in silicon. J Radioanal Nucl Chem 245:57-63 Guillong M, Heinrich CA (2007) Sensitivity enhancement in laser ablation ICP-MS using small amounts of hydrogen in the carrier gas. J Anal At Spectrom 22:1488-1494 Guillong M, Latkoczy C, Seo JH, Gunther D, Heinrich CA (2008a) Determination of sulfur in fluid inclusions by laser ablation ICP-MS. J Anal At Spectrom 23:1581-1589 Guillong M, Meier DL, Allan MM, Heinrich CA, Yardley BWD (2008b) SILLS: A MATLAB-based program for the reduction of laser ablation ICP-MS data of homogeneous materials and inclusions. In: Laser Ablation ICP–MS in the Earth Sciences: Current Practices and Outstanding Issues. Sylvester P (ed) Mineralogical Association of Canada Short Course Series, 40:328-333 Gunther D, Audetat A, Frischknecht R, Heinrich CA (1998) Quantitative analysis of major, minor and trace elements in fluid inclusions using laser ablation inductively coupled plasma mass spectrometry. J Anal At Spectrom 13:263-270 Hauri E, Wang J, Dixon JE, King PL, Mandeville C, Newman S (2002) SIMS analysis of volatiles in silicate glasses. 1. Calibration, matrix effects and comparisons with FTIR. Chem Geol 183:99-114 Heinrich CA, Pettke T, Halter WE, Aigner-Torres M, Audetat A, Gunther D, Hattendorf B, Bleiner D, Guillong M, Horn I (2003) Quantitative multi-element analysis of minerals, fluid and melt inclusions by laserablation inductively-coupled-plasma mass-spectrometry. Geochim Cosmochim Acta 67:3473-3497 Hettipathirana TD, Grey NA, Naidu R (2004) Analysis of silicates using wavelength-dispersive X-ray fluorescence spectrometry for major elements: Effects of loss elimination and catch-weights. X-Ray Spectrom 33:117-123, doi: 10.1002/xrs.709. Ihinger PD, Hervig RL,McMillan PF (1994) Analytical methods for volatiles in glasses. Rev Mineral 30:67-121 Jochum KP, Stoll B, Herwig K, Willbold M, Hofmann AW (2006) MPI-DING reference glasses for in situ microanalysis: New reference values for element concentrations and isotope ratios. Geochem Geophys Geosyst 7:Q02008, doi:10.1029/2005GC001060 Jurney ET, Curtis DB, Gladney ES (1977) Determination of sulfur in environmental materials by thermal neutron capture prompt gamma-ray spectrometry. Anal Chem 49:1741-1743 Kerr 2001 The calculation and use of sulfide metal contents in the study of magmatic ore deposits: a methodological analysis. Explor Min Geol 10:289-301 Kiba T, Takagi T, Yoshimura Y, Kishi I (1955) Tin (II)-strong phosphoric acid. A new reagent for the determination of sulfate by reduction to hydrogen sulfide. Bull Chem Soc Japan 28: 641-644 Klie JH, Sharma HD (1982) Sulfur determination in coal by 14 MeV neutron activation analysis. J Radioanal Nucl Chem 71:299-309 Knoll GF (1989) Radiation Detection and Measurement. John Wiley and Sons, New York
Analytical Methods
37
Kosler J (2008) Laser ablation sampling strategies for concentration and isotope ratio analyses by ICP-MS. In: Laser Ablation ICP–MS in the Earth Sciences: Current Practices and Outstanding Issues. Sylvester P (ed) Mineralogical Association of Canada Short Course Series, 40:79-92 Krouse HR, Ueda A (1987) Contents and sulphur isotope composition of trace sulfate and sulphide in various rock types. Appl Geochem 2:127-131 Leoni L, Menichini M, Saitta M (1982) Determination of S, Cl and F in silicate rocks by X-ray fluorescence analyses. X-ray Spectrom 11:135-137 Li M, Filby RH (1983) Determination of sulfur in fly ash and fuel oil Standard Reference Materials by radiochemical neutron activation analysis and liquid scintillation counting. Anal Chem 55:2336-2340 Liu Y, Samaha N, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Longerich HP, Jackson SE, Gunther D (1996) Laser ablation inductively coupled plasma mass spectrometric transient signal data acquisition and analyte concentration calculation. J Anal At Spectrom 11:899-904 Lytle D, Gerke TL, Maynard JB (2005) Geochemistry of sulfur in iron corrosion scales found in drinking water distribution systems: Role of sulfate reducing bacteria. J Am Water Works Assoc 97:109-120 Mackey EA, Paul RL, Lindstrom RM, Anderson DL, Greenberg RR (2005) Evaluation of uncertainties in prompt gamma-ray activation analysis. J Radioanal Nucl Chem 265:273-281 Mandeville CW, Sasaki A, SaitoG, Faure K, King R, Hauri E (1998) open-system degassing of sulfur from Krakatau 1883 magma. Earth Planet Sci Lett 160:709-722 Mandeville CW, Webster J, Tappan C, Taylor B, Timbal A, Sasaki A, Hauri E, Bacon C (2009) Stable isotopic and petrologic evidence for open-system degassing during the climatic and pre-climactic eruptionsof Mt. Mazama, Crater Lake, Oregon. Geochim Cosmochim Acta 73:2978-3012 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Mason PRD, Kaspers K, van Bergen MJ (1999) Determination of sulfur isotope ratios and concentrations in water samples using ICP-MS incorporating hexapole ion optics. J Anal At Spectrom 14:1067-1074 Mason PRD, Kosler J, de Hoog, JCM, Sylvester PJ, Meffan-Main S (2006) In situ determination of sulfur isotopes in sulfur-rich materials by laser ablation-multiple collector-inductively coupled plasma mass spectrometry (LA-MC-ICP-MS). J Anal At Spectrom 21:177-186 Mason PRD, Mank AJG (2001) Depth-resolved analysis in multi-layered glass and metal materials using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). J Anal At Spectrom 16:1381-1388 Mathez EA (1980) Sulfide relations in Hole 418A flows and sulfur contents of glasses. In: Initial Reports of the Deep Sea Drilling Project, v. 51, 52, 53, Part 2. Donnelly T, Francheteau J, Bryan W, Robinson P, Flower M, Salisbury M, et al. (eds) Washington (U.S. Government Printing Office), p 1069-1085. http://www. deepseadrilling.org/51_52_53/volume/dsdp51_52_53pt2_31.pdf McCandless EL (1964) Determination of sulfur in polysaccharides by neutron activation analysis. Anal Biochem 7:357-365 Metrich N, Bertagnini A, Landi P, Rosi Mauro (2001) Crystallization driven by decompression and water loss at Stromboli Volcano (Aeolian Island, Italy. J Petrol 42:1471-1490 Norman M, Robinson P, Clark D (2003) Major- and trace-element analysis of sulfide ores by laser-ablation ICPMS, solution ICP-MS, and XRF: new data on international reference materials. Can Mineral 41:293-305 O’Neill HC, Mavrogenes JA (2002) The sulfide capacity and the sulfur content at sulfide saturation of silicate melts at 1400 °C and 1 bar. J Petrol 43:1049-1087 Ohmoto H, Lasaga AC (1982) Kinetics of reactions between aqueous sulfates and sulfides in hydrothermal systems. Geochim Cosmochim Acta 46:1727-1745 Olivier C, Peisach M, Morland HJ, De Wet BS (1986) Sulphur determination by proton-induced prompt gamma emission: The effect of the matrix and its importance in coal analysis. J Radioanal Nucl Chem 106:107-122 Paul RL (1997) Hydrogen measurement by prompt gamma-ray activation analysis. Analyst 122:35R-41R Paul RL (2008) Determination of sulfur in steels by radiochemical neutron activation analysis with liquid scintillation counting. J Radioanal Nucl Chem 276:243-249 Paul RL, Simons DS, Guthrie WF, Lu J (2003) Radiochemical neutron activation analysis for certification of ion-implanted phosphorus in silicon. Anal Chem 75:4028- 4033 Pearce WC, Hill JWF, Kerry I (1990) Use of X-ray fluorescence spectrometry for the direct multi-element analysis of coal powders. Analyst 115:1397-1403 Pelikánová M (1985) Determination of sulphur in silicate and carbonate rocks by wavelength dispersive XRF. Fresnius J Anal Chem 320:338-340 Pettke T, Halter WE, Webster JD, Aigner-Torres M, Heinrich CA (2004) Accurate quantification of melt inclusion chemistry by LA-ICPMS: a comparison with EMP and SIMS and advantages and possible limitations of these methods. Lithos 78:333-361 Richards JP, Larson PB (eds) (1998) Techniques in Hydrothermal Ore Deposits Geology. Reviews in Economic Geology, Volume 10. Society of Economic Geologists 264pp
38
Ripley et al.
Rocholl A, Simon K, Jochum KP, Bruhn F, Gehann R, Kramar U, Luecke W, Molzahn M, Pernicka E, Seufert HM, Spettel B, Stummeier J (1997) Chemical characterisation of NIST silicate glass certified reference material SRM 610 by ICP-MS,TIMS, LIMS, SSMS, INAA, AAS and PIXE. Geostand Geoanal Res 24:101-114 Rose-Koga EF, Albarede F (2010) A data brief on magnesium isotope compositions of marine calcareous sediments and ferromanganese nodules. Geochem Geophys Geosys 11, Q03006, doi: 1029/2009GC002899 Rousseau M, Friedli C, Lerch P (1984) Trace determination of sulfur by heavy ion activation analysis. Anal Chem 56:2854-2856 Rowan JT, Houk RS (1989) Attenuation of polyatomic ion interferences in inductively coupled plasma mass spectrometry by gas-phase collisions. Appl Spectrosc 43:976-980 Roy P, Balaram V, Singh RS, Krishna AK, Charan CD, Charan SN, Murthy NN (2009) A simplified and rapid method for the determination of sulphur in kimberlites and other geological samples by WD-XRF spectrometry. At Spectrosc 30:178-183 Ryan CG, Heinrich CA, Mernagh TP (1993) PIXE microanalysis of fluid inclusions and its application to study ore metal segregation between magmatic brine and vapor. Nucl Instrum Methods Phys Res Sect B 77:463471 Sakai H, Ueda A, Field CW (1978) δ 34S and concentration of sulphide and sulphate sulphurs in some ocean floor basalts and serpentinites. In: Short Papers of the Fourth International Conference, Geochronology, Cosmochronology, Isotope Geology. Zartman RE (ed), US Geol Surv Open-file Rept 78-701:372-374 Sasaki A, Arikawa Y, Folinsbee RE (1979) Kiba reagent method of sulphur extraction applied to isotope work. Bull Geol Survey Japan 30:241-245 Sasaki A, Ishihara S (1979) Sulfur isotopic composition of the magnetite-series and ilmenite-series granitoids in Japan. Contrib Mineral Petrol 68:107-115 Seo JH, Guillong M, Heinrich CA (2009) The role of sulfur in the formation of magmatic-hydrothermal coppergold deposits. Earth and Planet Sci Lett 282:323-328 Seo JH, Guilong M, Aerts M, Zajacz Z, Heinrich CA (2011) Microanalysis of S, Cl and Br in fluid inclusions by LA-ICP-MS. Chem Geol 284:35-44, doi:10.1016/j.chemgeo.2011.02.003 Shah KR, Filby RH, Haller WA (1970) Determination of trace elements in petroleum by neutron activation analysis. J Radioanal Chem 6:185-192 Shani G, Cohen D (1976) The lower detectable limit of sulfur by fast neutron activation analysis. Int J Appl Radiat Isot 27:349-350 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Sohn RA, Menke W (2002) Application of maximum likelihood and bootstrap methods to nonlinear curve-fit problems in geochemistry. Geochem Geophys Geosys 3, doi;10.1029/2001GC000253 Souliotis AG (1964) Combined radiochemical-neutron activation analysis method for the determination of sulfur and phosphorus in high-purity paper and beer. Anal Chem 36:811-814 Stefansson A, Seward TM (2004) Gold(I) complexing in aqueous sulphide solutions to 500 degrees C at 500 bar. Geochim Cosmochim Acta 68:4121-4143 Strijckmans K, De Brucker N, Vanecasteele C (1985) Determination of sulphur in fly ash by instrumental proton activation analysis. J. Radioanal Nucl Chem Lett 96:389-398 Studley SA, Ripley EM, Elswick ER, Dorais MJ, Fong J, Finkelstein D, Pratt LM (2002) Analysis of sulfides in whole rock matrices by elemental analyzer-continuous flow isotope ratio mass spectrometry. Chem Geol 192:141-148 Sylvester PJ (ed) (2008a) Laser ablation ICP-MS in the Earth Sciences: current practices and outstanding issues. Mineralogical Association of Canada Short Course Series, Volume 40 Sylvester PJ (2008b) Matrix effects in laser ablation ICP-MS. In: Laser ablation ICP-MS in the Earth sciences: current practices and outstanding issues. Sylvester P (ed), Mineralogical Association of Canada Short Course Series, 40:67-78 Sylvester PJ, Cabri LJ, Tubrett MN, McMahon G, Laflamme JHG, Peregoedova A (2005) Synthesis and evaluation of a fused pyrrhotite standard reference material for platinum group element and gold analysis by laser ablation-ICPMS. Abstract, 10th Intern Platinum Symp, p 16-20 Thomas JP, Schweiker EA (1972) A rapid method for assaying sulfur using proton activation analysis. Nucl Instrum Meth 99:461-467 Thordarson T, Self S, Oskarsson N, Hulsebosch T (1996) Sulfur, chlorine and fluorine degassing and atmospheric loading by the 1783-1784 AD Laki (Skaftai Fires) eruption in Iceland. Bull Volcanol 58:205-225 Ueda A, Sakai H (1983) Simultaneous determinations of the concentration and isotope ratio of sulfate- and sulfide-sulfur and carbonate-carbon in geological samples. Geochem J 17:185-196 Vanecasteele C, Dewaele J, Esprit M, Goethals P (1980) The determination of sulphur in copper, nickel, and aluminum alloys by proton activation analysis. Anal Chim Acta 119:121-127
Analytical Methods
39
Vourvopoulos G, Womble PC (1989) On-line sulfur determination in coal with prompt gamma neutron activation. Nucl Instrum Meth Phys Res B 36:200-205 Walker FW, Parrington JR, Feiner F (1989) Nuclides and Isotopes - Chart of the Nuclides, 14th edition. General Electric Wallace P, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874 Wayman CH (1964) Determination of total sulfur in water by neutron activation analysis. Anal Chem 36:665666 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Williams JP, Farncomb FJ, Magliocca TS (1957) Determination of sulfur in glass. J Am Ceram Soc 40:352-354 Wilson SA, Ridley WI, Koenig AE (2002) Development of sulfide calibration standards for the laser ablation inductively-coupled plasma mass spectrometry technique. J Anal At Spectrom 17:406-409 Wohlgemuth-Ueberwasser CC, Ballhaus C, Berndt J, Stotter nee Paliulionyte V, Meisel T (2007) Synthesis of PGE sulfide standards for laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS). Contrib Mineral Petrol 154:607-617 Yule HP (1965) Experimental reactor thermal-neutron activation analysis sensitivities. Anal Chem 37:29-132 Zajacz Z, Halter W (2009) Copper transport by high temperature, sulfur-rich magmatic vapor: Evidence from silicate melt and vapor inclusions in a basaltic andesite from the Villarrica volcano (Chile). Earth Planet Sci Lett 282:115-121
3
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 41-78, 2011 Copyright © Mineralogical Society of America
Spectroscopic Studies on Sulfur Speciation in Synthetic and Natural Glasses Max Wilke Helmholtzzentrum Potsdam Deutsches GeoForschungsZentrum GFZ Telegrafenberg, 14473 Potsdam, Germany
[email protected]
Kevin Klimm Institut für Geowissenschaften Johann Wolfgang Goethe-Universität Altenhöferallee 1, 60438 Frankfurt am Main, Germany
Simon C. Kohn Department of Earth Sciences University of Bristol Queens Rd., Bristol, BS8 1RJ, United Kingdom INtroduCtIoN Spectroscopic methods are powerful means to obtain information on the electronic or local structure of materials. By these methods, constraints can be provided on the chemical state, crystal chemistry or, in non-crystalline materials, the coordination environment or complexation of a given element. In this chapter, we focus on spectroscopic techniques that provide direct insight into the sulfur (S) species present in glasses and melts and which are applicable at the sulfur concentrations usually found in glasses (mostly below 1-2 wt%). Methods that are potentially suitable for this task are i) the wavelength analysis of X-ray emission spectra (mostly using the electron microprobe), ii) X-ray absorption spectroscopy, iii) 33S NMR and iv) Raman spectroscopy. For compounds such as sulfides or sulfates, containing S as a major component, further methods such as optical absorption spectroscopy in the UV-visible frequency range (UVVIS) or electron spin/paramagnetic resonance (ESR or EPR) spectroscopy (e.g., Ross 1974; Wincott and Vaughan 2006) are useful. However, these two techniques provide only an indirect view on sulfur because they actually probe the cations, which are often transition metals. In compounds where S is a major component, these data also provide information on the sulfur species, because the cations are linked to S as an anion. In glasses, where S is only a minor component and cations are mostly coordinated by oxygen, it is difficult to link the observations made on the cation to the S species. Nevertheless, there are some studies on glasses where the additional information by UV-VIS or ESR spectroscopy was used to constrain possible species of sulfur in the glass (e.g., Beerkens 2003; Bingham et al. 2010). Here, only very brief introductions to the spectroscopic techniques will be provided. For a more detailed introduction to the spectroscopic techniques introduced here the reader is referred to other literature on this topic (e.g., Hawthorne 1988; Beran and Libowitzky 2004). Information on other analytical techniques related to sulfur is found in the chapter by Ripley et al. (2011, this volume). We will first discuss X-ray based techniques such as X-ray 1529-6466/11/0073-0003$05.00
DOI: 10.2138/rmg.2011.73.3
Wilke, Klimm, Kohn
42
emission and X-ray absorption spectroscopy that have been most widely used to characterize the sulfur oxidation state in glass. From X-ray absorption spectroscopy additional information on the type and arrangement of neighboring atoms may be derived. Next, the current state of 33S NMR spectroscopy on sulfur in glasses will be presented. Finally, the strength and limitations of Raman spectroscopy for evaluating sulfur species in glasses will be discussed. For all methods we will discuss the potential for constraining the speciation of sulfur in glasses and, as far as possible, how they can be used to quantify relative proportions of the sulfur species. Information on coordination and nearest neighbors of sulfur is only qualitative, so far. Quantitative determinations of the sulfur oxidation state in quenched melts are used to constrain the thermodynamics of sulfur redox reactions in silicate melts.
X-ray eMISSIoN aNd abSorptIoN SpeCtroSCopy S Ka X-ray emission Measuring the wavelength of S Ka radiation [l(S Ka)] provides information on the valence state and the local electronic environment of sulfur in the matter of interest. Generally, S Ka X-ray radiation is emitted if the K-shell 1s electrons have been excited to higher energy levels either by X-rays with an energy of at least 2.47 keV or by electrons with an excitation voltage of at least 7-8 kV. Ka radiation is emitted upon an electronic transition from the L to the K-shell, or in other notation from the S 2p electron orbital to the 1s orbital level. For the sake of integrity it has to be noted that the S Ka radiation can be distinguished in two discrete energies S Ka1 and S Ka2 due to the splitting of the 2p electron orbital into two energy levels, 2p1/2 and 2p3/2. In the following, we will simply use the term S Ka, that refers to the doublet of S Ka1 and S Ka2 wavelengths as the two emission lines are hard to distinguish from each other due to the experimental energy resolution commonly used. Variations in the wavelength of the emitted S Ka is related to shifts in the energy levels of the electron shells due to changes in the valence electron population, which influences the screening of the nuclear charge and thus the nuclear potential experienced by the 2p electrons (Alonso Mori et al. 2009). The wavelength of the emitted Ka radiation varies in a similar fashion as the ionic radii of S2− (184 pm) and S6+ (29 pm) (Shannon 1976). The energy of the Ka radiation for S2− is approximately 0.2 eV lower and for S6+ 1.2 eV higher compared to the Ka radiation of elemental sulfur (Faessler and Goehring 1952; Matthews et al. 1999). This corresponds to a wavelength shift Dl(S Ka) of ~0.003 Å between sulfide and sulfate (Carroll and Rutherford 1988). The energy difference between S4+ and S0 is about half that of S6+ (Alonso Mori et al. 2009). Mixtures of sulfur in various valence states display intermediate wavelengths for l(S Ka) between that of pure S6+ and pure S2−. Therefore, in contrast to XANES or Raman spectroscopy, measuring the S Ka wavelength does not provide direct constraints on the sulfur species, at least with the wavelength resolution and counting statistics usually available—for example, a mixture of S6+ and S2− in a sulfur-bearing compound can display identical wavelengths to pure S4+ (see also below). Measurements of l(S Ka) are performed by scanning the wavelength of emitted X-rays during exposure of solid material to a high-energy electron beam. Typically such measurements are performed using electron probe microanalysers (EPMA). With this technique, the X-rays are excited by electrons and the wavelength is determined by scanning the angle of a spectrometer crystal with known crystal lattice (usually a crystal made of Pentaerythritol, PET) such that Bragg’s Law (nl = 2d sin q) is fulfilled for l(S Ka). Usually a full scan in the range of 0.612 to 0.616 sin q is performed resulting in a spectrum of counts versus sin q (Fig. 1). The peak maximum of the spectrum corresponds to the l(S Ka) of interest. It has to be noted that so far all studies measuring the l(S Ka) for determining the S6+/S2− ratios performed scans over the full range of sin q. However, it is also possible to determine the S6+/S2− by comparing two intensities at fixed sin q relatively in a similar fashion as it has been
Spectroscopic Studies of Sulfur Speciation in Glasses
43
Figure 1. Scans of the l(S Ka) peak position for FeS (black circles) and BaSO4 (open circles) crystalline compounds (Klimm unpublished data). Operating conditions of the EPMA were 20 kV and 30 nA with a spot diameter of 15 mm. Each spectrometer was moved 0.00004 sinq units for 100 steps over the range of 0.612 to 0.616 sinq during a single spot analysis. Counting time for each step varied was 1600 ms which results in total counting time of each spot of 160 s. Peak positions (indicated by grey lines) were obtained by fitting the wave scan spectra with a Gaussian function (solid and dotted lines). The calculated Dl(S Ka) wavelength shift between the FeS and BaSO4 reference compound is in this case 3.07×103 Å.
!
demonstrated for Fe La and Fe Lb intensities in order to determine the Fe3+/Fe2+ ratio in garnets (Hoefer and Brey 2007). l(S Ka) shift of sulfur model compounds. The relationship of S Ka wavelength shift and the sulfur valence state was established 40 to 50 years ago by analyzing numerous solid sulfur reference compounds with known sulfur valence state by X-ray excited emission spectroscopy (e.g., Faessler and Goehring 1952; Chappell and White 1968; Connolly and Haughton 1972). Nowadays, l(S Ka) measurements are usually performed by EPMA and are calibrated with sulfate and sulfide compounds to determine the maximum wavelength shift [Dl(S Ka)]. The wavelength shift measured on unknown samples is compared to this calibration for determining an apparent sulfur valence state (e.g., Carroll and Rutherford 1988; Wallace and Carmichael 1994; Jugo et al. 2005). Sulfide (S2−), the most reduced variety of sulfur, yields highest values of l(S Ka) whereas the oxidized variety sulfate (S6+) yields lowest values. In addition to the valence state of sulfur (S6+ or S2−), the charge balancing ion of the sulfide or sulfate compound also affects the exact wavelength of l(S Ka). For instance, the l(S Ka) for sulfides is higher by ~0.0001 Å for FeS2 than for MnS2 (Faessler and Goehring 1952). For sulfates it is higher by Dl(S Ka) ~ 0.0002 Å for CaSO4 (anhydrite) than for BaSO4 (barite), which is in the same order of magnitude as the precision of the EPMA analysis reported by Wallace and Carmichael (1994). The highest Dl(S Ka) between sulfate and sulfide are obtained for sulfides and barite and are in the range of Dl(S Ka) = 0.292 to 0.313 Å (Table 1) depending on the diffracting crystal, spectrometer geometry, etc. Because of the high sulfur concentration in the sulfide and sulfate reference compounds, high quality scans of wavelength spectra can be obtained at “normal” electron beam conditions such as 15 kV, 20-40 nA with acquisition times of less than a minute (Carroll and Rutherford 1988). Typical scans of the wavelength for FeS and barite are shown in Figure 1. Because of the relatively small difference between Ka1 and Ka2 wavelength and the experimental resolution the l(S Ka) scans follow a Gaussian function, which is used for fitting in order to determine the peak position and thus l(S Ka). It should be noted that FeS can be present as pyrrhotite or troilite, which is not always specified by the authors. Rowe et al. (2007) used both and do not report any difference in terms of l(S Ka). Therefore, the more general term FeS is used here. Experimental details for EPMA of S in glasses. The sample preparation for determining the l(S Ka) of S in glasses using an electron microprobe follows the common procedure of sample preparation for any conventional analysis using EPMA (see also Ripley et al. 2011, this volume). Glass chips are embedded in epoxy resin and polished using a combination of
Wilke, Klimm, Kohn
44
table 1. S Ka wavelength shift between sulfide and sulfate model compounds. Compound l(S Ka) (Å×103)
reference Sulfide
barite
anhydrite
Scapolite
pyrite
CaS
-
2.00
2.00 1.84
-
Chappell & White 1968
FeS2
-
2.1
-
-
Connolly & Haughton 1972
FeS
-
3.06±0.08
2.86±0.18
-
Carroll & Rutherford 1988
FeS
3.13±0.11
3.00±0.09
-
0.39±0.12
Wallace & Carmichael 1994
FeS
3.11±0.10
-
-
-
Métrich & Clocchiatti 1996
ZnS
3.07±0.18
-
-
-
Gurenko & Schmincke 1998; Gurenko & Schmincke 2000
ZnS
3.04±0.20
-
-
-
Jugo et al. 2005
FeS
2.92±0.11
-
-
-
Klimm et al. submitted
Source
Note: Dl(S Ka) wavelength shift calculated relative to given reference sulfide in units of Ångström times 103.
abrasive paper and diamond-bearing polishing paste or solutions, respectively. Although it has been argued that some surface oxidation may occur during sample preparation generating S6+ in nominally sulfate free samples (Fleet et al. 2005b), thus affecting the measured sulfur valence state, there is no definitive evidence that such a process commonly occurs in all samples. XANES and Raman spectroscopy of only sulfide-bearing alkali silicate glasses that are usually very sensitive to corrosion and oxidation showed no evidence for any oxidation during sample preparation (Klimm and Botcharnikov 2010; Klimm et al. submitted). The polished sample should be perfectly plane and the surface roughness should be less than 1 mm in order to guarantee optimal conditions for analysis. In order to perform l(S Ka) determination on natural melt inclusions care has to be taken to consider the size of the inclusions. Because of the excitation depth of a few mm, depending on the analytical conditions such as acceleration voltage, beam current and beam focus, a sufficiently large volume of glass is required. It also has to be noted that melt inclusions or synthetic glasses may contain additional crystalline sulfate/ sulfide phases or droplets of quenched immiscible sulfide liquid indicating sulfur saturation, which can distort the S6+/S2− determination of sulfur dissolved in the glass. A general difficulty of analyzing sulfur in silicate glasses using EPMA is the relatively low abundance of sulfur, which is related to the relatively low solubility of sulfur in silicate melts. Sulfur solubility may vary considerably as a function of various parameters such as glass composition, oxygen fugacity, temperature and pressure and sulfur concentrations in coexisting phases. Typical sulfur levels in natural glass compositions are low and usually below ~3000 ppm total S (e.g., Wallace and Carmichael 1992; Clemente et al. 2004; Backnaes and Deubener 2011,this volume; Ebel 2011, this volume; Baker and Moretti 2011, this volume). This in return requires relatively high beam currents and counting times to lower the detection limit and increase counting statistics during EPMA analysis. Typical conditions for the determination of l(S Ka) by EPMA in natural glasses are acceleration voltages of 15-30 kV and beam currents of 20-40 nA. Counting times of ≥1 min are used in order to achieve sufficient counting statistics in spectra acquired at low S concentrations. Usually, reliable l(S Ka) wavelength measurements can be performed for concentrations as low as 100 ppm S with >10 min total counting time (e.g., Wallace and Carmichael 1992; Nilsson and Peach 1993). The uncertainty in peak position at S concentrations <100 ppm exceeds the total sulfide-sulfate peak shift (Matthews et al. 1999). Simple glass compositions such as soda lime silicate glass may dissolve sulfur levels up to a
Spectroscopic Studies of Sulfur Speciation in Glasses
45
few weight percent (see Backnaes and Deubener 2011, this volume). For such glasses counting times may be reduced (Klimm et al. submitted). l(S Ka) shift of sulfur in glasses. Measurements of the l(S Ka) have mostly been performed on glasses of geological relevant compositions in order to elucidate the sulfur dissolution mechanism. The first determination of l(S Ka) on glasses was performed by Connolly and Haughton (1972). They used a synthetic sulfur saturated glass of basaltic composition synthesized at 1200 °C and log fO2 = −9.08. They compared the l(S Ka) of the synthetic basalt to that measured on natural scapolite, a Na-Ca framework aluminosilicate that contains SO42− groups. The observed wavelength shift between basalt and scapolite was similar to that between pyrite and anhydrite, and they concluded that the sulfur in this basaltic glass was present as an anion coordinated by cations. Carroll and Rutherford (1988) performed l(S Ka) measurements on a series of experimental glasses of basaltic, andesitic, trachyandesitic and dacitic composition. The glasses were synthesized under a range of fO2 from reducing to oxidizing conditions (log fO2 = QFM-2 to QFM+6 where QFM corresponds to the log fO2 of the solid oxygen buffer assemblage quartz-fayalite-magnetite). The determined l(S Ka) were compared to l(S Ka) of synthetic FeS as a reference compound for S2− in order to determine the wavelength shift, Dl(S Ka), of the samples relative to this sulfide standard in the form of: Dl(S K a )sample = l(S K a )sulfide standard − l(S K a )sample
(1)
The experimental glasses yield wavelength shifts of Dl(S Ka) = 0 at reducing fO2 or Dl(S Ka) ~ 0.0029 Å at oxidizing fO2, which is similar to Dl(S Ka) observed for scapolite. At intermediate fO2, Dl(S Ka) was within the above range and systematically increased with increasing fO2 indicating that the sulfur valence state in the silicate glasses is a function of fO2 and changes from sulfide to sulfate (a full discussion on the relationship of fO2 and sulfur valence state is given below). Jugo et al. (2005) determined Dl(S Ka) wavelength shifts of experimental basaltic glasses synthesized at either oxidizing or reducing conditions. The determined Dl(S Ka) values of the glasses were either similar to that of the sulfide reference (sphalerite, ZnS) or similar to the sulfate reference (barite). However, because of the lack of experiments at intermediate fO2 Jugo et al. (2005) did not obtain intermediate values of Dl(S Ka). The work of Carroll and Rutherford (1988) is the only study to date that reports intermediate l(S Ka) in experimentally synthesized glasses of geological relevant composition equilibrated at controlled fO2. A number of studies performed wavelength shift measurements using EPMA on natural glasses of volcanic origin in order to gather information on the sulfur speciation in magmas of various tectonic settings. The compositions studied include submarine glasses of basalts and back-arc magmas (Wallace and Carmichael 1992, 1994; Nilsson and Peach 1993; de Hoog et al. 2004; Rowe et al. 2007), intra-oceanic basaltic and alkaline compositions (Métrich and Clocchiatti 1996; Gurenko and Schmincke 1998, 2000; Rowe et al. 2007) or glasses of subduction-zone related magmas (Matthews et al. 1999; Rowe et al. 2007, 2009). Values of Dl(S Ka) determined in all studies were similar to FeS in the case of primitive basaltic compositions or fell in the range between those of FeS and anhydrite, as seen for the experimental glasses (Carroll and Rutherford 1988). Only a few attempts have been made to apply S Ka wavelength shift determination on more simple glass compositions such as albite or alkali silicate glass. Reported Dl(S Ka) wavelength shifts for alkali silicate glasses also fall in the range between FeS and Anhydrite depending on the valence state of the starting sulfur compound or on the fO2 during glass synthesis (Winther et al. 1998; Tsujimura et al. 2004; Klimm et al. submitted).
determination of the S oxidation state using epMa Carroll and Rutherford (1988) concluded that sulfur in the glass is present as sulfate (S6+), sulfide (S2−) or a mixture of both. This was based on the fact that glasses saturated with either
46
Wilke, Klimm, Kohn
anhydrite or FeS show similar Dl(S Ka) as the coexisting S-bearing crystals. Assuming that the shift of Dl(S Ka) is a linear function of the relative proportions of S6+ and S2− in the glass and that the Dl(S Ka) for the sulfate standard represents 100% sulfur as sulfate S6+/SS, the following proportionality is obtained: S6 + Dl(S K a )sample = SS Dl(S K a )sulfate
(2)
where Dl(S Ka)sample is the wavelength shift of the measured sample relative to a reference (FeS in most cases; see Eqn. 1) and Dl(S Ka)sulfate is the wavelength shift of a sulfate standard (anhydrite in most cases) relative to the same reference. It has to be noted that the determined S6+/SS ratio is not a direct indicator of the relative proportions of S6+ and S2− and should be taken as “sulfate mole fraction equivalent” X(S6+)eq. (Winther et al. 1998; Jugo et al. 2005). For example, sulfides with bi-anions in their structure [e.g., (S-S)2− in FeS2 and MnS2] show positive deviations in l(S Ka) resulting in apparent X(S6+) ranging between 0.11 and 0.18 (Winther et al. 1998).
X-ray excited high-resolution X-ray emission spectroscopy It has been known for decades that X-ray emission can be excited using an X-ray beam (e.g., Faessler and Goehring 1952; Chappell and White 1968; Connolly and Haughton 1972). Alonso Mori et al. (2009) have shown recently, that the combination of a very intense X-ray beam from a synchrotron radiation source and a highly resolving wavelength dispersive spectrometer with sub-eV resolution allows detection of the complete fine structure of the S Ka doublet and even the fine structure of the S Kb emission. These authors only studied model compounds but provided the basis for further applications. This technique should probably be able to differentiate between mixtures of S2− and S6+ and, for example, S4+. Bingham et al. (2010) provide first emission spectra collected on simple glass compositions of industrial relevance synthesized at various redox conditions. All spectra could be fitted assuming only mixtures of S2− and S6+. In such spectra, S4+ should appear as an extra peak at a position in-between those of S2− and S6+. If S4+ was present at all in these samples it was only to a very minor extent.
X-ray absorption spectroscopy X-ray absorption spectroscopy or better X-ray absorption fine-structure spectroscopy (XAFS) is able to provide information on the chemical state and on the local structural environment of the studied element. X-ray absorption spectroscopy is performed by scanning the energy of the incident X-ray beam across a certain energy threshold needed for excitation of a core-level electron, typically at the K or L-absorption edge of a given element. If the energy bandwidth of the incoming beam is small enough, the fine structure present in the otherwise smoothly varying X-ray absorption coefficient is observed, which is dependent on the chemical state and structural environment of the absorbing element. The observed fine structure arises from transitions of the photoelectron to higher bound-state energy levels (before or at the edge) or from backscattering of the photoelectron from the atoms surrounding the absorbing element (at the edge and after the edge). X-ray absorption spectroscopy is basically grouped into two methods: i) X-ray absorption near-edge structure (XANES), using the fine structure occurring just below and up to ca. 100-150 eV above the edge. This region in the spectrum is sensitive to the electronic structure of the absorbing element, which changes with valence state as well as with the structural arrangement and type of surrounding atoms; ii) Extended X-ray absorption fine structure (EXAFS), the fine structure occurring at energies far above the edge (starting from ca. 100 up to 1000 eV or further), that provides quantitative information on the distances between the absorbing atom and its surrounding atoms as well as the number of surrounding atoms. As XAFS works as a local probe, the structural information is accessible even for systems lacking long-range order, such as silicate glasses and melts or aqueous solutions and hydrothermal or even supercritical fluids.
Spectroscopic Studies of Sulfur Speciation in Glasses
47
A large variety of sulfur-bearing compounds have been studied using XANES. Compounds studied include those where S is a major constituent such as natural and synthetic sulfides, elemental sulfur, sulfites, sulfosalts, and sulfates (e.g., Hitchcock et al. 1987; Li et al. 1994, 1995; Mosselmans et al. 1995; Farrell and Fleet 1999, 2001; Farrell et al. 2002; Fleet 2005; Fleet et al. 2005a; Wincott and Vaughan 2006; Figueiredo and da Silva 2009). As S is an important component in organic geochemistry and soil systems many studies have been performed on related compounds, i.e., sulfides, organic compounds such as asphaltenes, bitumens or humic substances (e.g., George and Gorbaty 1989; Kasrai et al. 1994; Vairavamurthy et al. 1997; Jokic et al. 2003; Prietzel et al. 2003; see also references in Fleet 2005). Most studies used the XANES at the S K-edge (2472 eV) where the S 1s electron is excited. Also the S L2,3-edge (L3: 162.5 eV, L2: 163.6 eV) may be used, where the 2p electrons are excited (e.g., Li et al. 1994, Farrell et al. 2002). However, application to glasses has never been attempted, probably due to the low concentrations of sulfur and the interference with the Si L1-edge in this energy region. The wide range of oxidation states covered by sulfur, and its ability to form bonds with atoms covering a large range of values in electronegativity, result in strong variations in the local electronic structure of sulfur. Spectroscopic methods provide powerful means to obtain experimental information because the XANES arises from transitions close to the Fermi level. At the S K-edge, the S 1s electron is excited to the 3p-like lowest unoccupied level. With increasing oxidation state the position of the edge shifts to higher energy due to differences in the screening of the charge of the nucleus. The shift between S2− and S6+ may reach 12 eV, and thus the XANES is a useful means for determining the oxidation state. However, the relationship of the K-edge shift and oxidation state is usually further complicated by the nature of the coordinating elements as well as by effects of the local geometry. In the following, basic observations for the S K-edge XANES and their interpretations made on crystalline sulfur compounds are introduced briefly, followed by a detailed discussion of XANES application on S in silicate glasses. S K-edge XANES of model compounds. Figure 2 shows a compilation of XANES spectra measured on crystalline model compounds that we consider most important for the analysis of sulfur in silicate glass. The highest energy position of the XANES is found for S6+, where sulfur forms the oxy-anion sulfate with tetrahedral symmetry. Spectra of all crystalline sulfate compounds show a very strong peak at 2482 eV that has a shoulder at higher energies varying in intensity and position among the different sulfates. At higher energies (> 2490 eV) further features occur that also vary among the various sulfates. The strong peak can be directly related to electron excitations to molecular orbitals formed by S 3p and O 2p orbitals, whereas all features at higher energies can be related to effects by the next-nearest neighbors or more distant neighbors (Alonso Mori et al. 2009). The edge of the S4+ in the sulfite oxy-anion, which has trigonal pyramidal symmetry, is found at lower energies. The main peak is found at 2478 eV with a high-energy shoulder and a further peak at 2481 eV. While the main peak is assigned as for sulfate, the high-energy shoulder indicates hybridization of the S 3s and 3p orbitals, which is possible due to the trigonal symmetry of the sulfite molecule (Alonso Mori et al. 2009). In sulfides, the first coordination shell is formed by metal cations. The different local structural environments as well as the different valence shell configurations of the metal cations have a strong influence on the XANES. In particular the edge position shows considerable variations (Li et al. 1995; Womes et al. 1997; Farrell and Fleet 1999, 2001; Farrell et al. 2002; Fleet 2005; Wincott and Vaughan 2006; Alonso Mori et al. 2009). Fe-bearing sulfides such as pyrrhotite show a prominent peak at energies between 2470 and 2472 eV followed by a broad feature at higher energies (2475-2480 eV). The exact structure depends on the composition and structure as shown for a series of Fe-Cu sulfides by Alonso Mori et al. (2009) or for a suite of synthetic Fe-Mn sulfides by Farrell et al. (2002). In the case of sulfides, the 3p orbitals are fully occupied. Hybridization with metal 3d orbitals will transfer some 3p electron density to the cation and
48
Wilke, Klimm, Kohn
Figure 2. Normalized S K-edge XANES spectra of model compounds. Spectra are background-subtracted and normalized to edge-jump. The plot was compiled from data provided through the public ID21 database or directly by authors as indicated (measurement mode as indicated): Gypsum: ID21 database, A. Scheinost, transmission; Hauyn: ID21 database, fluorescence, M. Wilke; Na2SO4, Na2SO3, CaS: R. Alonso Mori, fluorescence selfabsorption corrected; CaS peak for S4+ from beam damage; Pyrite: ID21 database, M. Sandström, transmission; Pyrrhotite: M. Wilke, fluorescence; Sphalerite: ID21 database, A. Scheinost, transmission, (for further examples of model compounds see also Fleet et al. 2005a,b, Fleet 2005, Jugo et al. 2010., Alonso Mori et al. 2009, or ID21 database: http://www.esrf.eu/ UsersAndScience/Experiments/Imaging/ ID21/php/Database%20SCompounds).
thus creates unoccupied states of 3p character that can be assigned to the spectral feature at the lowest energy found for Fe sulfides. The degree of electron density transfer to the 3d orbitals strongly depends on the metal cation. The prominent feature at 2470-2472 eV is absent in MgS and CaS (c.f. Farrell et al. 2002; Fleet 2005) because no 3d states are available for hybridization. The feature is likewise absent if the 3d orbital of the cation is filled as in the case of ZnS. The broad feature at 2475-2480 eV in the spectrum of pyrrhotite was assigned to S 3p states hybridized with Fe 4p states based on electronic structure calculations (Womes et al. 1997). Compared to pyrrhotite, the edge of pyrite is shifted to higher energies resulting in a more intense maximum at 2472 eV in accordance with the formal S oxidation state of -1. The peak at 2472 eV is attributed to excitation to S 3p s* states because S 3p and Fe 3d hybridization is less favored in this compound (Womes et al. 1997). As mentioned above, the energy position of the edge can be used as a first proxy for the oxidation state. The position of the edge is usually measured by determining the first maximum in the derivative of the spectrum. Figure 3 illustrates the variation of the XANES edge position with the formal oxidation state of sulfur model compounds, which is adapted after Fleet et al. (2005b) using the data of Alonso Mori et al. (2009). Compounds with negatively charged sulfur display a wide range of edge positions reflecting the strong influence of the electronic structure on the energy position as discussed above. Therefore, the edge position is not a wellsuited parameter for precise measurements of the oxidation state. More reliable procedures to determine the S oxidation state are discussed below.
Spectroscopic Studies of Sulfur Speciation in Glasses
49
Figure 3. Plot of edge position vs. formal oxidation state of sulfur after Fleet et al. (2005b) for various sulfur compounds. The edge position was determined as first maximum in the first derivative of the spectrum. Data for plot are taken from Alonso Mori et al. (2009) and shifted by −0.86 eV (Alonso Mori pers. comm., mistake in the reported energy position of the reference compound, elementary sulfur). For more details see Alonso Mori et al. (2009) or Fleet et al. 2005b.
Experimental details for XANES on glasses. The sample preparation for XANES measurements at the S K-edge depends on the way the spectra are going to be measured. For samples containing sulfur as a major component transmission measurements are ideal because these would be least affected by experimental artifacts. However, the ideal sample thickness at the S K-edge is very small due to the strong absorption of X-rays at theses energies. A thickness of less than 5 microns is necessary for one absorption length (sample thickness where absorption is 1/e) for silicate glasses. On the other hand, silicate glasses usually contain S only in trace amounts. In this case, the indirect measurement of the absorption spectrum by use of the fluorescence yield is adequate, because sulfur itself can be neglected for the total absorption of the sample. In concentrated samples, sulfur dominates the total absorption and the strong increase across the edge counterbalances the detected fluorescence yield by decreasing the effective escape depth of the fluorescence. This effect is called self-absorption and will distort the spectra by damping maxima and minima in the fine structure (see also Bunker 2010). Indirect measurement using the total electron yield (photo electrons and Auger electrons) would be another option. However, the escape depth of these electrons is very shallow so that only the surface and a few nanometers below are actually probed (see e.g., Fleet et al. 2005b). In case of fluorescence, the escape depth of radiation is in the order of several microns, and thus a true measurement of the bulk if the surface is pristine and clean. The Si in silicate glass further complicates XRF measurements of S due to the strong fluorescence signal. Therefore an energy dispersive detector is required. The high Si fluorescence signal usually will lead to detector saturation and thus an inefficient use of the detector. Filtering of the fluorescence signal with an 8 mm Kapton turned out to be an efficient way to improve the fluorescence detection (e.g., Wilke et al. 2008). Spectra on glasses are taken either on powdered samples or on polished sections. In case of powdered samples care must be taken that the speciation of the sulfur in the glass is not changed at the glass surface. Evans et al. (2009) reported that oxidation significantly affected the resulting spectra for some of their reduced samples measured as powders. If bulk samples or powders are used care must be also taken not to include any sulfur-rich inclusions (e.g., crystals). An ideal combination is the use of a polished glass sample with a thickness of 100-200 mm and a small beam with diameter in the range of a few hundred microns, so that one is able to check optically that a homogeneous portion of the glass is analyzed. This is of particular interest
50
Wilke, Klimm, Kohn
for samples that were equilibrated with an aqueous fluid phase, because these may contain bubbles with S-bearing fluid (Wilke unpublished). Depending on the stability of the sample in the beam (see also below), smaller beam sizes may be usable. In terms of spectral resolution the energy bandwidth delivered from a Si (111) doublecrystal monochromator is adequate. As X-ray sources, synchrotron radiation from bending magnets as well as undulators has been used. Undulators are periodic magnetic structures used to produce very intense radiation (e.g., see Bunker 2010 for more details regarding synchrotron radiation). The high intensity delivered from an undulator seems ideal due to the low S contents in the glass but care must be taken to check for potential beam damage during the analysis (see below). If a focused beam is needed an undulator source is probably the only way to acquire spectra in reasonable time. In any case, it is important to adopt the analytical protocol as well as the sample preparation to the needs of the specific problem and considerable amounts of time should be dedicated to check all relevant experimental parameters. The lowest S concentration that can be analyzed by XANES is difficult to define as it depends on many parameters, such as beam intensity, spot size, sample and fluorescence detector used. For example, a rhyodacite glass containing ca. 200 ppm S was successfully analyzed using a polished 200 mm thick glass section, a beam delivered from an undulator collimated to 200 mm and a single element Si-drift detector. The spectrum was collected with 8 s/point, so that the total acquisition time was ca. 80 min (Wilke unpubl.). S K-edge XANES of glasses – implications for S species. The first measurements of the S K-edge XANES on quenched melts were reported by Paris et al. (2001), who compared spectra of glasses of natural and simple composition (Na-Ca silicate glass) synthesized at various redox conditions. Bonnin-Mosbah et al. (2002) and Métrich et al. (2002) presented the first XANES data on natural melt inclusions in samples from mid-ocean ridges, ocean islands and arc-related volcanism. As found earlier (e.g., Nilsson and Peach 1993; Wallace and Carmichael 1994) their data show systematic variations of the oxidation state of sulfur as a function of the geotectonic setting, i.e., more reduced sulfur for oceanic settings and more oxidized sulfur for arc-related settings. McKeown et al. (2004) investigated the incorporation of S in synthetic borosilicate nuclear waste glass covering a large variety of compositions and synthesis conditions. XANES data of glasses from the system Na-Al-Si-O, to which CaS or CaSO4 was added, and synthetic Galapagos basaltic glasses were studied by Fleet et al. (2005b). Backnaes et al. (2008) studied simple glass compositions of technical relevance that were bubbled with gaseous SO2 or doped with sulfur compounds at high temperature to investigate the dissolution of tetravalent sulfur at ambient pressure. Wilke et al. (2008) reported data from synthetic oxidized and reduced glasses of anhydrous and hydrous basaltic, hydrous andesitic and hydrous Na-Ca silicate composition in order to elucidate the effect of beam-sample interaction on the resulting spectra. XANES data of natural samples were compared to synthetic samples by Métrich et al. (2009). Evans et al. (2009) studied glasses from the Ca-Mg-Al-Si-O system synthesized at varying redox conditions. This study focused on the effect of doping by Mn, Ni and W on the S speciation. Bingham et al. (2010) presented data from several spectroscopic techniques, including XANES, applied to technical soda-lime glasses and slag synthesized at various redox conditions. Jugo et al. (2010) and Klimm et al. (submitted) determined the sulfur oxidation state in glasses of natural basaltic or simpler potassium silicate to albitic compositions, respectively that were synthesized under various controlled fO2 and contained various S species. Finally, Stelling et al. (2011) investigated interactions between sulfur and H2O in soda-silica and soda-lime melts by looking at inter-diffusion profiles of S and H2O quenched from high T and P. Here, XANES was used to detect changes in the S valence state or coordination along the profile. Examples of S K-edge XANES for various glass compositions are shown in Figure 4. For oxidized glasses containing only S6+, a very intense peak at 2482 eV with a small shoulder on the high-energy side is observed throughout, which is very similar compared to the sulfate
Spectroscopic Studies of Sulfur Speciation in Glasses
51
Figure 4. Examples of normalized S K-edge spectra of oxidized and reduced glasses. From top to bottom: dry syn. basalt, 1 GPa 1300 °C, oxidized; hydrous syn. Etna basalt, 0.2 GPa, 1050 °C, oxidized; hydrous Na-Ca-Si glass, 0.2 GPa, 1000 °C, oxidized; hydrous NaCa-Si glass, 0.2 GPa, 1000 °C, reduced; dry syn. basalt, 1 GPa 1300 °C, reduced; natural basalt LauBasin: Ocean floor sample; natural basalt Loihi Seamount: Kilauea ocean floor sample (spectra from Jugo et al. 2010, Stelling 2009, Klimm et al. submitted). Spectra are backgroundsubtracted and normalized to edge-jump.
model compounds. Also the region at higher energies shows quite strong similarity to sulfate model compounds and almost no variability depending on glass composition (cf. Paris et al. 2001; McKeown et al. 2004; Fleet et al. 2005b; Métrich et al. 2009). It can thus be inferred, that the S in the sulfate oxy-anion has a very similar structural environment in these glasses compared to crystalline references. Brendebach et al. (2009) further constrained this structural similarity by obtaining structural data from analysis of EXAFS data for S6+ in borosilicate waste glass. According to their analysis they conclude that the sulfate oxy-anion is present in their glass as isolated groups surrounded by cations and do not bond to the borosilicate network via bridging oxygens. Spectra of reduced glasses show a broad hump with a maximum at 2476-2477 eV. At 2470-2471 eV, a shoulder at the beginning of the edge jump is observed in the natural and synthetic samples shown. The spectra measured on simple glasses—particularly Fe-free glass compositions—do not show the shoulder at 2470-2471 eV. As shown by one example of hydrous soda-lime silicate glass in Figure 4, the edge position is usually shifted to slightly higher values in comparison to Fe-bearing glasses (see also Métrich et al. 2009 for further examples). The example of the hydrous soda-lime silicate glass also shows a sharp feature at 2466 eV, which was so far only observed in simple alkali and earth-alkali silicate glasses and is most intense for hydrous glasses (Métrich et al. 2009; Stelling et al. 2011; Klimm et al. submitted). Fleet et al. (2005b) assigned the S XANES found for reduced Fe-bearing glasses to the pre-dominance of Fe mono-sulfide units in the glass. The difference to the spectrum measured
52
Wilke, Klimm, Kohn
on pyrrhotite implies that these units differ in the local structure compared to sulfides, which results in weaker hybridization of S 3p and Fe 3d orbitals—the main source for the peak found at 2470-2471 eV (see above). In Fe-free glasses monosulfide units with other cations are formed (e.g., CaS or NaS), which is constrained by the resemblance of the spectra to crystalline mono-sulfides and numerical simulations (Fleet et al. 2005b; Kravtsova et al. 2004). There is considerable difference between spectra obtained on complex (Fe-bearing) glass compositions to those taken on simple compositions. Even among simple compositions, spectra show differences, for example, sulfide in Na-silicate and Na-Ca-silicate glasses (e.g., Backnaes et al. 2008; Stelling 2009; Stelling et al. 2011). These composition-dependent differences may suggest that sulfide is associated with network modifiers and does not replace bridging oxygens in substantial amounts. This is consistent with the first ab initio molecular dynamics simulations of soda-lime-silica melts demonstrating that sulfide is primarily replacing non-bridging oxygens (Machacek et al. 2010). Free S2− is not a relevant species according to these simulations. Further quantitative assignment of nearest neighbors for sulfide in glasses based on S K-edge XANES is difficult due to the many potential possibilities (Fe, Mg, Ca, Na etc.) and due to the unknown stereochemistry of each possible unit and therefore remains a field for future research. The exact assignment of the sharp feature at 2466 eV is still pending. However, it can be clearly correlated with S2− present in the glass and it appears to be particularly important in hydrous glasses. Stelling et al. (2011) observed this peak within H2O-S2− inter-diffusion concentration profiles of Na silicate and Na-Ca silicate glasses and observed a strong spatial correlation of the peak at 2466 eV to changes in the H2O speciation (detected by IR and Raman spectroscopy). Hence, correct assignment of this feature might bring valuable information for interactions between S and H2O in volatile bearing melts and glasses. The proposed pre-dominance of Fe-S units in silicate glasses is indirectly supported by the strong correlation of sulfide solubility and Fe content (e.g., Carroll and Webster 1994; see also Baker and Moretti 2011, this volume). Further evidence for Fe-S units in the glasses is provided by Raman spectroscopy, i.e., in sulfide-bearing glasses, bands are observed in the range of vibrations typical for metal-sulfide compounds (Klimm and Botcharnikov 2010; see below). Finally, the chromophore responsible for the amber color of container glasses is assigned to Fe-S units in the glass (see Falcone et al. 2011, this volume). In a few XANES spectra of Fe-bearing glasses, a separate peak is observed with the maximum at 2470 eV, so that the spectra closely resemble that of pyrrhotite. Examples with this feature, as shown in Figure 4, can be found for natural samples where the cooling history is largely unknown. Although the samples look microscopically glassy and show no evidence for devitrification or exsolution of a sulfide liquid, the XANES indicates a structural re-organization during cooling. XANES is a very sensitive tool for such changes as shown earlier for Ni and Fe in hydrous quenched melts (Farges et al. 2001; Wilke et al. 2006). Métrich et al. (2009) report XANES spectra for a few synthetic glasses that feature this separate peak at 2470 eV. By comparing measurements with a large (200 mm) and a small beam (1 mm) they could show that the separate peak stems from exsolved Fe-S globules present in their glasses. Evidence for co-existence of sulfur species by XANES. A particular advantage of XANES spectroscopy is that it also provides very helpful experimental evidence for determining coexisting sulfur species in the redox region where the transition from the reduced to the oxidized species occurs. In their pioneering work, Carroll and Rutherford (1988) and Wallace and Carmichael (1994) who used the fluorescence wavelength shift measured by EPMA to determine the average S oxidation state, assigned the sulfur redox equilibrium to the reaction: S2−(m) + 2 O2(g) = SO42−(m)
(3)
This equation assumes that only two species, S2− and S6+, are present in silicate melts and glasses. A final proof of this assumption, however, could not be provided with wavelength analysis using an electron microprobe alone (see also Jugo et al. 2005; Jugo 2009). The first
Spectroscopic Studies of Sulfur Speciation in Glasses
53
XANES data of Paris et al. (2001) that were actually taken (among others) on the samples of Carroll and Rutherford (1988), do show only evidence for S2− and S6+ and these species were found to coexist. This finding supported the interpretations that were based on the fluorescence wavelength shift data and on the assumption that no other S species was present at intermediate redox conditions. Similar evidence for the coexistence of S2− and S6+ was found in subsequent studies (McKeown et al. 2004; Fleet et al. 2005b; Backnaes et al. 2008; Wilke et al. 2008; Métrich et al. 2009; Evans et al. 2009). Métrich et al. (2002) observed the presence of S4+ in a few of their natural glass inclusion samples. The existence of another S species would have major consequences for the thermodynamic treatment of sulfur redox equilibria in silicate melts. However, it was shown later that the presence of S4+ can be related to interactions between sulfur and the intense X-ray beam during the measurement (Wilke et al. 2008; Métrich et al. 2009, see also below). Furthermore, the results by Backnaes et al. (2008) show that S4+ is not present in silicate melts even if tetravalent sulfur compounds are dissolved in the melt at high temperature. Finally, the co-existence of S2− and S6+ in the range of oxygen fugacities of the transition from sulfide to sulfate has been demonstrated by Jugo et al. (2010) (see Fig. 5) for a suite of hydrous basaltic samples synthesized at 200 MPa and 1050°C. The relative decrease of the S6+ features or the relative increase of the S2− features in the spectra correlates well with the decrease of the oxygen fugacity during the synthesis (see also below). Bingham et al. (2010) still argue for the existence of intermediate sulfur species, such as S4+. However, in our view the very good XANES data they provide does not support their interpretations. Furthermore, they
Figure 5. Normalized S K-edge spectra of hydrous synthetic Etna Basalt synthesized at 0.2 GPa, 1050 °C and oxygen fugacities indicated. Oxygen fugacity in log units relative to quartzfayalite-magnetite buffer (QFM). Spectra from Jugo et al. (2010). Spectra are background-subtracted and normalized to edge-jump.
54
Wilke, Klimm, Kohn
show that their X-ray emission spectra (the first reported spectra of this type for S in glass) can be fitted assuming only S2− and S6+.
determination of S oxidation state from XaNeS XANES spectra may be used to quantitatively determine the average oxidation state as shown in the case of Fe or V (Wilke et al. 2001, 2004; Sutton et al. 2005; Cottrell et al. 2009). Any procedure has to involve spectral features of the XANES whose intensity variation can be related unambiguously to the average formal oxidation state. As shown above, the S K-edge position does not provide a good basis for this. Paris et al. (2001) suggested fitting the XANES peaks using lorentzian-gaussian functions and an arctangent function as background. However, they could not confirm all the results that were previously obtained on their samples by the wavelength shift method. McKeown et al. (2004) fitted the spectra taken on glasses with linear combinations of spectra taken on crystalline model compounds. This approach assumes that the XANES of the models can be transferred to the glass, i.e., that the local atomic structures as well as the electronic structures in glass and model are comparable. While it might be acceptable for S6+ it is certainly problematic for S2− since coordinating cations are unknown and, moreover, orbital hybridization for a given cation-sulfide pair is likely to be different between sulfide crystal and a silicate glass (e.g., Fleet et al. 2005b, see also above). Another option for determining the S oxidation state in glasses is to use linear combinations of spectra taken on completely oxidized or reduced glasses. One could use the whole range of the XANES and fit the spectrum of the unknown using the two end-members. This type of fit, however, is strongly influenced by regions of the spectra that are not sensitive to the oxidation state. It is sufficient and more precise to focus on features in the spectrum that are particularly sensitive to S2− or S6+ as suggested by Jugo et al. (2010), where they take the signals of the normalized spectra integrated in two energy windows for sulfide and sulfate, respectively. For sulfide the spectra were integrated between 2475.7 eV and 2480 eV yielding I(S2−); for sulfate they were integrated between 2481.5 eV and 2484 eV yielding I(S6+) (see also Fig. 6a). To define a relationship between the variation of the two signal windows and S6+/SS, a sequence of spectra was calculated by linear combination of spectra of a completely oxidized glass and a completely reduced glass (Fig. 6a). The I(S2−) and I(S6+) parameters were determined for this calculated sequence and the trend of the intensity parameter defined as I(S6+) over the sum of the two signals (SI = I(S2−) + I(S6+)) was plotted against S6+/SS (Fig. 6b). In this protocol, it is assumed that mixtures of the two end-members match the spectra of samples with mixed oxidation state. Within the energy windows used for integration this assumption is valid as was checked by comparison. The trend was best fitted by an exponential function:
( )
I S6 + / SI − A S6 + = −C ln B SS
(4)
where the fit parameters A, B, and C for the synthetic basaltic composition used are 1.24, −0.95, and 0.81, respectively. This relationship can be used to estimate S6+/SS in samples showing S2− and S6+ features in XANES spectra. It is very likely that this relationship varies depending on the composition of the glass, i.e., because of the variations of sulfide features with local coordination. In contrast to iron (e.g., Wilke et al. 2004), verification of this procedure by an independent method is very difficult if not impossible due to the low S concentrations in the glass and the lack of suitable methods. Even the comparison with measurements by the S Ka wavelength shift will give uncertain results due to the likeliness of beam damage affecting the results by EPMA. A possible crosscheck could be done by X-ray excited high-resolution X-ray emission analysis. However this has not been attempted, so far.
parameter determined on the calculated spectra shown in a) as a0function of S6+/ΣS
Oxidized Glass
2470 2480 (Markers). The line represents a fit of Eq. (4) to the data with the parameters indicated.
Both graphs are modifiedSpectroscopic after Jugo et al. Studies (2010). sulfide
(a)
Intensity(sulfate) / ∑ Intensity (sulfate+sulfide)
Normalized Absorption
Energy (eV) a) of Sulfur Speciation in Glasses
sulfate
8
6
4
2
Reduced Glass
Oxidized Glass
a)
2480
2490
Intensity(sulfate) / ∑ Intensity (sulfate+sulfide)
Energy (eV) 1525
55
0.8
0.7
0.6
0.5
Fit of Eq. (4): Coefficient values ± one standard deviation A = 1.24 ± 0.01 B = -0.95 ± 0.01 C = 0.81 ± 0.02
0.4
0.0
2500
2500
(b)
0.9
0.3
0 2470
2490
b)
0.2
0.4
0.6
0.8
1.0
6+
S / ∑S
Figure 6. a) Sequence of spectra calculated by linear combinations of the two spectra indicated as reduced and oxidized glasses. Bars at the top indicate energy windows used for integrating the signals for sulfide and sulfate, which are used to calculate the intensity parameter, which is used to determine S6+/SS (see 0.9 text). b) Variation of intensity parameter determined on the calculated spectra shown in a) as a function of 6+ S /SS (Markers). The line represents a fit of Equation (4) to the data with the parameters indicated. Both 0.8 graphs are modified after Jugo et al. (2010). 0.7
beam damage during analysis by epMa and XaNeS
1525
0.6
EPMA. The strong interaction of the high-energetic electron beam with the sample very often0.5induces changes in the properties of the sample itself that affect either the intensity or the wavelength of the emitted radiation of the element under consideration. Métrich and Clocchiatti Fit of Eq. (4): (1996) the first time-dependent measurements of l(S Ka) using EPMA on natural 0.4 published Coefficient values ± one standard deviation 1.24 ± to 0.01evaluate if beam-sample interaction has an effect on the resulting melt inclusionsAB in ==order -0.95 ± 0.01 C = 0.81 ± 0.02 data.0.3The acquisition time was varied between ~3 to 10 min with spot sizes of 10 to 30 mm. 0.0 0.2 0.4 The results show that the0.6total 0.8sulfur 1.0concentration is not significantly affected. However, 6+ S /of ∑S l(S Ka) showed a systematic shift towards lower wavelength—or b) the determined values increasing Dl(S Ka) when compared to the l(S Ka) of the FeS reference—with increasing acquisition time or decreasing spot size of the electron beam. This suggests that during electron 71 of the S2− was beam exposure, a modification in the oxidation state of sulfur occurs and some 6+ 4+ oxidized to S or S , which cannot be distinguished by this method. A similar oxidation process was observed by Rowe et al. (2007) who performed l(S Ka) wavelength shift determination on olivine-hosted melt inclusions with acquisition times from 1 min up to 1 h. Their results also show a shift towards more oxidizing S6+/SS ratios with increasing acquisition time that follows a linear relationship (Fig. 7a). The increase is stronger at low S concentrations or if the initial S6+/ SS ratio is low (observed increase of S6+/SS from 0.05 to 0.30 for 1600 ppm S compared to an increase of S6+/SS from 0.8 to 0.9 for 2700 ppm S after 1h hour of acquisition). To circumvent the problem of oxidation or at least minimize the effect of electron beam-sample interaction both Métrich and Clocchiatti (1996) and Rowe et al. (2007), suggested that the sample should be moved under the electron beam during beam exposure, for example, at a rate of 1 mm/min. Klimm et al. (submitted) performed l(S Ka) measurements in addition to XANES measurements on a series of synthetic water-bearing soda-lime silicate glasses synthesized at oxidizing and reducing fO2. Because the sulfur concentration in these glasses was up to an order of
Wilke, Klimm, Kohn
56 1.0
a) Basalt
1.0
y = 0.14x + 81.10 R2 = 0.94
0.8
S6+/ΣS
0.4
incremental
S6+/ΣS
0.8
0.6
b) Soda lime glass
2700 ppm S 1600 ppm S
0.6
0.4
y = 0.45x + 6.28 R2 = 0.81
S 6+/ ΣS, XANES
0.2
0.2
0.57 0.00
0.0
0
10
20
30
time (min)
40
50
60
0.0
0
40
80
time (s)
120
160
Figure 7. a) Increase of S6+/SS with electron beam exposure time for basaltic glasses, variable beam currents and a beam size of 5-7 mm measured by EPMA. One-minute analyses are based on 50 incremental acquisitions with the beam scanning the sample at a rate of 1mm/min (redrawn after Rowe et al. 2007). b) S6+/SS of soda-lime glass using l (SKa) wavelength shifts by EPMA as a function of acquisition time for repeated wavelength scans for two different glasses (redrawn after Klimm et al., submitted): black dots: soda-lime glass that initially only contain S2−; grey dots: soda-lime glass that initially contains a mixture of S6+ and S2−. Initial S6+/SS of both glasses have been determined by XANES and are given in the box. Curves (solid and dotted lines) are exponential fits of the data. Extrapolation to zero time (the intersection of the exponential fits with the y-axis) yields S6+/SS values of -0.01 (black curve) and 0.66 (dotted curve).
magnitude higher (2000 to 20000 ppm S) than in natural glass compositions acquisition times of only 10 to 160 s yielded high quality spectra for determining l(S Ka). The results demonstrate that for such glass compositions oxidation occurred after less than 10 s of beam exposure. Glass that initially only contains S2− (as confirmed by XANES spectroscopy) showed 20% S6+ after 10 s and 90% S6+ after 80 s total acquisition time by l(S Ka). Comparison to glasses with higher initial S6+/SS showed that the oxidation rate is higher if the initial S6+/SS ratio is low. However, the increase in S6+/SS with time did not follow a linear trend as suggested by Rowe et al. (2007). Instead, it followed an exponential trend up to values between 80 and 100% S6+ as shown in Figure 7b. Klimm et al. (submitted) showed that back-extrapolation of a series of S6+/ SS determinations for various acquisitions to zero-time, before any oxidation has occurred, are consistent with those determined using XANES spectroscopy. The difference in the chemical composition could provide a possible explanation for the differences in the kinetics of beaminduced oxidation, so that Fe-bearing glasses are less sensitive to beam damage. It also has to be noted that glasses that contain only S6+ (from XANES) show wavelength shifts indicating S6+/SS values below 1, similar to observations pointed out by Jugo et al. (2005, see also below). This suggests that also reduction may occur during electron beam exposure. However, the observed wavelength-shift values in these oxidized samples were immediate and no correction could be applied. The presence of S4+ cannot be detected by measuring the l(S Ka) by EPMA. However, Wilke et al. (2008) demonstrated that S4+ is very likely generated in initially S4+-free glasses during EPMA measurements. XANES spectra show a signal of S4+ if the particular area of analysis had previously experienced electron beam exposure. “Fresh” glass surfaces, which had not been affected by any previous beam-sample interaction, did not provide any signal corresponding to S4+. Wilke et al. (2008) suggested that the generation of S4+ could explain why most literature data do not report wavelength shifts equal to the sulfate reference compounds
Spectroscopic Studies of Sulfur Speciation in Glasses
57
for oxidized or sulfate-saturated glasses, as documented in Matthews et al. (1999), Jugo et al. (2005) and Klimm et al. (submitted). XaNeS. Although the interaction between X-ray photons and the sample are much weaker compared to electrons, even for irradiation by photons the resulting photo-ionization may induce changes in element species. Wilke et al. (2008) showed that the occurrence of S4+ in oxidized glasses is closely related to beam damage, which occurred during the XANES acquisition using an intense focused beam (spot size < 1 mm). In a time-series of XANES spectra that were acquired in steps of one minute the appearance and increase of a peak at the energy for S4+ and the decrease of the peak for S6+ was observed. Métrich et al. (2009) confirmed these findings. Klimm et al. (submitted) observed a dramatic oxidation of S2− to S6+ during XANES acquisition in simple silicate glasses that contain significant amount of water (few wt%). For instance, hydrous potassium silicate glasses that only contain S2− showed a strong signal for S6+ (corresponding to 40% S6+) if the sample was not moved under the focused beam (1 mm) during acquisition. Since the beam damage is not immediate experimental procedures can be designed to avoid this effect. The exact procedure has to be adapted to the beam parameters and sample composition. This may include use of a less intense beam, reduction of acquisition time on a single spot with averaging over several spots or movement of the sample during the acquisition.
Sulfur redox equilibrium determined in glasses by l(S Ka) and XaNeS The relationship between S6+/SS in quenched melts and fO2 provides the possibility to constrain the redox conditions of magmatic processes. Therefore, several studies have attempted to determine this relationship. In their pioneering study, Carroll and Rutherford (1988) were able to show the change of S6+/SS determined by the Dl(S Ka) as a function of the oxygen fugacity (fO2) during the synthesis (Fig. 8a). At log fO2 ≤ QFM-2, S6+/SS equals to 0 indicating that only S2− is present in the glass. In contrast, S6+/SS yield values slightly below 1 (~0.9) at log fO2 ≥ QFM+2 indicating that sulfur is predominantly dissolved as S6+ in the glass. In the range of log fO2 = QFM-2 to QFM+2, Dl(S Ka) data imply that sulfur is dissolved as a mixture of S2− and S6+, if only these two species are assumed to be present. S6+/SS ratios increased nearly sigmoidally with the steepest increase of S6+/SS from ~0.2 to 0.9 at fO2 in the range of QFM+1 to QFM+2 (Carroll and Rutherford 1988). A similar relation between S6+/SS and fO2 is also observed in studies on natural glasses using l(S Ka) (Wallace and Carmichael 1992; Nilsson and Peach 1993; Wallace and Carmichael 1994; de Hoog et al. 2004; Gurenko and Schmincke 1998, 2000; Matthews et al. 1999; Rowe et al. 2007). The fO2 during the formation of natural samples in these studies is estimated by determining the Fe3+/Fe2+ ratio of the glasses and applying the compositional model by Kress and Carmichael (1991) that describes the relationship of Fe3+/SFe as a function of fO2. Jugo et al. (2010) provided the first quantitative determinations of the sulfur oxidation state by XANES. These data are shown in Figure 8b, which were determined on the XANES spectra of the series of basaltic glasses shown in Figure 5. The data points indicate a slightly sharper turnover from S6+ to S2− at slightly more reducing conditions than the data by Carroll and Rutherford (1988), i.e., between QFM and QFM+2. The sulfur speciation derived by these spectroscopic methods is consistent with that derived by thermodynamic analysis of sulfur solubility measurements. These data imply the coexistence of S6+ and S2− over a similar range of oxygen fugacity (e.g., Nagashima and Katsura 1973, see Backnaes and Deubener 2011, Fig. 2, this volume). The oxidation reaction of S2− to S6+ in melts can be expressed by the equilibrium given in Equation (3), for which the equilibrium constant follows as: K=
a(SO24 − ) a(S2 − ) ⋅ f 2 O2
(5)
with a for activity and f for fugacity. Assuming that the activities of the two sulfur species can
log (S /S )
2-
-3
-2
-1
0
1
-1
(b)
0
∆QFM
1
2
Carroll & Rutherford 1988 Wallace & Carmichael 1994 XANES Jugo et al. 2010 Fit Jugo et al. 2010
EPMA
Figure 8. a) S6+/SS vs. log fO2 (expressed relative to the log fO2 of the quartz-fayalite-magnetite equilibrium) for silicate glasses. S6+/SS was determined by S Ka wavelength shift and XANES as indicated. Data points by Carroll and Rutherford (1988) represent individual Dl(S Ka) measurements on experimental glasses of basaltic, andesitic, trachyandesitic and dacitic composition (see also their Table 1). Trend of Wallace and Carmichael (1994) corresponds to Equation (7) with given parameters (see text), which is also based on Dl(S Ka). Trend of Jugo et al. (2005) results form a compilation of literature data (see text). Data points of Jugo et al. (2010) were determined on hydrous basaltic glasses by XANES using the spectra shown in Figure 5. The trend corresponds to a fit of Equation (8) to the data assuming a stoichiometric factor of 2 for fO2 as outlined in the text. b) S6+/S2− vs. log fO2 for datasets shown in a) as indicated. Trend of Wallace and Carmichael (1994) corresponds to Equation (7) with given parameters.
6+
(a)
58 Wilke, Klimm, Kohn
Spectroscopic Studies of Sulfur Speciation in Glasses
59
be described by their concentrations (in square brackets), we can rearrange to log
[SO24 − ] = log K + 2 log fO2 [S2 − ]
(6)
Thus, plotting the logarithm of the ratio of S6+/S2− as a function of log fO2 at constant P and T should result in a straight line with a slope of 2 and an intercept that is a function of the equilibrium constant. The data by Carroll and Rutherford (1988) and Jugo et al. (2010) plotted in such a way are shown in Figure 8b. Both datasets define a linear correlation with oxygen fugacity on this graph, but with different slopes. Based on Equation (6), Wallace and Carmichael (1994) derived an empirical equation by combining the S6+/SS of natural glasses (Wallace and Carmichael 1992) with the S6+/SS of experimental glasses (Carroll and Rutherford 1988), all derived by the wavelength shift method, as follows: log
[SO24 − ] b a log fO2 + + c = 2− [S ] T
(7)
where a = 1.02, b = 25410 K, c = −10. As shown in Figure 8b, this equation describes well the data measured by Carroll and Rutherford (1988). The parameter a, which describes the slope, differs from the ideal value 2 and indicates significant deviation from the equilibrium constant derived in Equations (5) and (6). In contrast, the data by Jugo et al. (2010) can be fitted using the “ideal” slope of 2 and even releasing the parameter for the slope yields a value of 2.1±0.1, which is, within error, identical to the ideal slope (Jugo et al 2010). The consequence of the different slopes for the sulfur redox equilibrium is best illustrated by going back to the sigmoidal change of S6+/SS with fO2 shown in Figure 8a. Equation (6) can be re-arranged to plot S6+/SS as a function of log fO2: S6 + 1 = A − BDQFM SS 1 + 10
(8)
where A controls the position along the ∆QFM-axis and B corresponds to the slope in Equation (6) (see also Jugo et al. 2005, 2010). The variation of S6+/SS with fO2 as derived from the fit parameters of Equation (7) by Wallace and Carmichael (1994) is shown in Figure 8a (with A = 1.19 and B = 1.02). Matthews et al. (1999) and Jugo et al. (2005) derived similar empirical equations that included newer data of S6+/SS by EPMA (Nilsson and Peach 1993; Matthews et al. 1999; de Hoog et al. 2004; Jugo et al. 2005). The trend shown in Figure 8a is the one derived by Jugo et al. (2005) with A = 1.22 and B = 0.95. The XANES-derived data can be described with Equation (8) by setting B = 2 and a fitted value of A = 2.1 (Jugo et al. 2010). The comparison of the three curves illustrates that the transition from sulfide to sulfate occurs over a much shorter interval of fO2 for the XANES-derived S6+/SS than for those by the S Ka wavelength shift. A possible reason for this discrepancy may be that during l(S Ka) measurements beam damage occurred, so that S6+ is reduced to S4+ or that some S2− is oxidized for reduced samples. Both processes would lead to an extended fO2 range for the redox transition and could explain the discrepancy, so that the shallow slope for data determined by l(S Ka) with EPMA is simply an artifact of the method (Jugo et al. 2010). Matthews et al. (1999) and Jugo et al. (2005) addressed another problem of l(S Ka) data that was previously neglected, i.e., the fact that the Dl(S Ka) of oxidized glasses always show values that are lower than those determined on the crystalline sulfate references. There are three ways to interpret this observation: i) either the presence of sulfide in the glass; ii) a fundamental difference of Dl(S Ka) for crystalline sulfates and sulfate dissolved in the glass; iii) the presence of another sulfur species such as S4+. In applying Equation (8), Jugo et al. (2005) included another fit parameter that replaces the numerator of 1 in Equation (8) and derived a value of 0.89. This result implies that apparently only a maximum of ca. 90% sulfate
60
Wilke, Klimm, Kohn
(or minimum of 10% sulfide) is present in glasses even at very high fO2 of QFM+6, where only S6+ should be stable. Jugo et al. (2005) suggested that presence of S4+ could be one explanation for the reduced Dl(S Ka). They suggested S4+ as an additional sulfur species being stable at intermediate redox conditions. Later, Wilke et al. (2008) re-analyzed the samples of Jugo et al. (2005) by XANES and could show in fact that S4+ is present. After repolishing the samples, however, no evidence for S4+ was left, so that this species likely formed during the EPMA measurements (see also section above on beam damage). Two studies, Métrich and Clocchiatti (1996) and Rowe et al. (2009), have used the fit of S6+/SS as a function of fO2 provided by Wallace and Carmichael (1994) to determine the fO2 for the formation of melt inclusions. However, it has to be noted that this oxygen barometer is only applicable for a narrow range of redox conditions, given the narrow fO2 range where sulfide and sulfate coexist. Thus, in many cases either sulfide or sulfate will be found indicating conditions more oxidized or more reduced than the sulfur redox reaction in melts. Furthermore, post-entrapment processes, such as loss of volatiles or reactions during cooling, may alter the S6+/SS found in the melt inclusion. Finally, S6+/SS data and calibrations derived by l(S Ka) measurements with EPMA should always be considered within the limitations defined by the analytical problems discussed earlier.
NuClear MaGNetIC reSoNaNCe Nuclear magnetic resonance (NMR) is an enormously powerful technique that can be applied to problems as diverse as: cation coordination numbers in solutions; porosity and permeability in oil well logging; protein structure and dynamics; inorganic materials chemistry; and medical imaging. Many thousands of papers using the family of NMR techniques are published each year, and those dealing with inorganic solids are a small proportion of the whole. Nonetheless, over the last 25 years, NMR has played a crucial role in improving our understanding of the structure of silicate glasses and the structural and dynamic relationship between silicate melts and quenched glasses. A good review of solid state NMR of inorganic materials is provided by MacKenzie and Smith (2002), reviews of the use of NMR in mineralogy and geochemistry include Kirkpatrick (1988), Stebbins (1998) and Fechtelkord (2004) and more specific reviews of NMR studies of silicate melts and glasses are provided by Zwanziger (1998) and Kohn (2004). None of these reviews are very recent, and the reader is directed to series such as Progress in Nuclear Magnetic Resonance Spectroscopy and Annual Reports on NMR Spectroscopy for up-to-date reviews of specific applications of NMR. A detailed introduction to NMR is not warranted here as there are so many excellent books and review articles on NMR, but in essence, NMR is a technique that probes the difference in energy of nuclear spin energy levels. The separation of the energy level is given by mB DE = I
(9)
where DE is the difference in energy between energy levels (and hence the energy of the transition), m is the nuclear magnetic moment, Bo is the applied magnetic field and I is the nuclear spin. m is given by m=
ghI 2π
(10)
where g is the magnetogyric ratio and h is Planck’s constant. The difference in energy of the nuclear spin energy levels, DE, is always small, therefore the NMR frequency, n, is small. Furthermore the population difference between the energy levels is usually small, so NMR is intrinsically not a very sensitive technique. Much of the useful information comes from
Spectroscopic Studies of Sulfur Speciation in Glasses
61
the chemical shift. This is the small difference in resonance frequency between a site in a sample and a reference compound. It arises because different chemical environments have different distributions of electron density and the magnetic fields induced by the motions of these electrons slightly modify the external magnetic field experienced by the nucleus. Samples containing a single element in several different chemical environments can thus give multiple peaks in NMR spectra. Glasses, where there is a distribution of chemical environments, give peaks broadened by chemical shift distribution (csd). In solids there are several interactions between nuclei, which broaden the NMR resonance. These interactions are averaged by rapid tumbling of molecules in liquids, but in solids other strategies have to be used to obtain high resolution spectra. The most important for inorganic materials is magic angle spinning (MAS) a technique whereby the sample is physically spun around a specific axis (54.74° from the direction of the magnetic field) at frequencies up to 50 Hz. MAS removes nuclear dipole-dipole interactions, and broadening from chemical shift anisotropy (CSA). In cases where I > 1/2 the nucleus has a nuclear electric quadrupole moment and additional line broadening mechanisms exist that are not completely narrowed by MAS. In these cases, the width and shape of the MAS spectrum of the central (+1/2 to −1/2) transition depend additionally on i) the quadrupole moment of the nucleus, ii) the electric field gradient at the site of the nucleus of interest, and iii) a parameter that is characteristic of the geometry of the site known as the asymmetry parameter (h). For a given nucleus the quadrupole coupling constant (Cq) and asymmetry parameter can often be unambiguously derived from experimental spectra, and the two parameters reflect the nature and magnitude of distortion of the local environment of the nucleus. An overview of the issues in solid state NMR of quadrupolar nuclei is given by Ashbrook and Duer (2006). 33
S is the only suitable isotope of sulfur for NMR measurements, but its nuclear spin properties are far from ideal. It has a low natural abundance (0.76%), with a low magnetogyric ratio and sensitivity, and it has a spin of 3/2 giving it a quadrupole moment. At natural abundance, the sensitivity is 1.72×10−5 compared with 1H. 33S is not a commonly studied nucleus, even in solutions, but a range of different compounds especially those related to organic and biological chemistry have been studied. This work is reviewed by Barbarella (1993), Jackowski (2001), Chesnut and Quin (2004) and Musio (2009) and provides a useful framework for the chemical shifts expected in solids. 33
S NMr of solid model compounds
The best way to ascertain the utility of NMR on any particular nucleus is to perform measurements on model compounds of known structure. Early surveys of this nature for 33S in solids were published by Eckert and Yesinowski (1986) and Daunch and Rinaldi (1996). Eckert and Yesinowski (1986) established that it was feasible to collect static 33S NMR spectra of sulfides and sulfates, but that even with an external magnetic field of 11.7 T, the low Larmor frequencies mean that a rolling baseline artifact obscures the broad resonances from samples. Eckert and Yesinowski (1986) used a special pulse sequence to overcome this problem and compiled isotropic chemical shifts and quadrupole coupling constants for a range of materials. Daunch and Rinaldi (1996) reassessed the potential of 33S NMR by using a higher fields of 14.1 T and fast MAS. Because the dipole moment of 33S is relatively small, magic angle spinning of solid samples can completely remove line-broadening arising from dipole-dipole interaction and spinning faster than the static line width will eliminate broadening from chemical shift anisotropy. Daunch and Rinaldi (1996) found that the width of resonances is reduced by a factor of 10 compared with the techniques of Eckert and Yesinowski (1986) because of averaging of dipolar coupling, CSA and quadrupolar broadening. More comprehensive studies were provided by Wagler et al. (2003, 2004) who obtained MAS spectra for a range of different inorganic sulfides and sulfates respectively. Figure 9 shows 33S MAS NMR spectra of sulfides obtained at a field of 17.6 T with magic angle spinning up to 6 kHz (Wagler et al. 2003). The lines are all narrow and show no observable quadrupolar lineshape, suggesting quadrupole
62
Wilke, Klimm, Kohn
Figure 9. 33S MAS NMR spectra of sulfide model compounds (Wagler et al. 2003). The line widths are generally small because of small quadrupole coupling constants for sulfide in these compounds where the S site has a high degree of symmetry, but note the large chemical shift range of around 600 ppm. [Used by permission of Elsevier Limited, from Wagler et al. (2003) J Magn Reson 161, Fig. 1, p. 193]
coupling constants less than 0.5 MHz. There is however a large range in isotropic chemical shifts, related to the electronic structure (e.g., ionicity) of the crystals. Figure 10 shows spectra for a range of crystalline sulfates (Wagler et al. 2004). In this case there is very little variation in isotropic chemical shift (approximated by the shift of the left hand edge of the quadrupolar lineshape), but the width and shape of the quadrupolar resonance varies substantially. Fitting the line shapes gives Cq varying from less than 0.5 MHz for NH4Al(SO4)2·12H2O to 1.7 MHz for BaSO4, and h varying from 0.1 to 1.0 over the whole set of model compounds. The biggest control on Cq is the distribution of oxygen nearest neighbors around the sulfur. Figure 11 shows the correlation between Cq for sulfate model compounds vs. D(SO)m, the difference between the longest and shortest S-O bond length in the sulfate group. The potential of 33S NMR in the study of cement chemistry was investigated by de Lacaillerie et al (2006). They showed that the natural abundance spectra of gypsum and ettringite were quite different, and they proposed that 33 S has potential for quantitative analysis of sulfate mineralogy in cement-related systems. They performed their measurements at 19.6 T and again they emphasized the importance of using the highest possible magnetic field. Several studies have recently developed improved multiple pulse techniques for detection of difficult nuclei such as 33S. Jakobsen et al (2006, 2007), observed the spinning sidebands of the satellite transitions (e.g., +3/2 to +1/2 rather than the normal +1/2 to −1/2 transition) of two different alums [KAl(SO4)2, NH4Al(SO4)2] and two tetrathiometallates [(NH4)2WS4, (NH4)2MoS4], and obtained accurate values of Cq, h and isotropic chemical shift. In the case of the alums they observed interesting structural behavior as a function of temperature, and in the tetrathiometallates they resolved several different S sites. Further technological improvements were reported by Hansen et al. (2008), who applied multiple-pulse sequences to induce population transfer between the different spin energy levels in order to enhance the intensity of the central transition. The signal enhancement was in the range 1.74-2.25, which equates to a reduction in spectrum acquisition time by a factor of nearly five for a given signal-to-noise ratio. They also provided improved data on a range of model compounds. Further applications of 33S
Spectroscopic Studies of Sulfur Speciation in Glasses
63
!
Figure 10. 33S MAS NMR spectra for sulfate model compounds (Wagler et al. 2004). Most of the spectra show a quadrupolar doublet line shape because of residual 2nd order quadrupole interaction. [Used by permission of Elsevier Limited, from Wagler et al. (2004) J Magn Reson 170, Fig. 2, p. 340]
Figure 11. Plot of quadrupole constant (Cq) for sulphate model compounds vs ∆(SO)m, the difference between the longest and shortest S-O bond length in the sulfate group (Wagler et al. 2004). The circles are Cq determined from experimental NMR spectra, the squares are Cq from ab initio calculation. [Used by permission of Elsevier Limited, from Wagler et al. (2004) J Magn Reson 170, Fig. 3, p. 342]
64
Wilke, Klimm, Kohn
at natural abundance with signal enhancement were reported by Jakobsen et al (2009, 2010) and Sutrisno et al (2009) who used both high fields and quadrupolar Carr-Purcell-MeiboomGill (QCPMG) to obtain high quality spectra from ever more complex materials with a wide range of Cq. Most recently Moudrakovski et al (2010) used fields up to 21.1 T and the QCPMG technique to acquire 33S NMR spectra (with and without MAS) of K2SO4, KHSO4, K2S2O7, and K2S2O8 at natural abundance. They fitted their spectra to obtain improved values of isotropic chemical shift, chemical shift anisotropy, quadrupole coupling constant and asymmetry parameter. They also performed density-functional-theory (DFT) calculations to complement their experimental data. Quadrupolar constants, up to 16 MHz were observed. 33
S NMr of glasses
There are very few published 33S NMR spectra of glasses. However, the data published so far suggest that the best results can be obtained using 33S enrichment, the highest available magnetic fields and magic angle spinning. Typically at least a few tens of mg of sample is required, so the spatial resolution of XANES, EPMA or Raman is not attainable. Couch et al. (2004) presented spectra for 6 (mostly hydrous) silicate glasses as well as some model compounds obtained at a frequency of 46.05 MHz using a field of 14.1 T. The spectra were relatively good with acquisition times of 15-60 hours and 99% enriched 33S in the starting materials for the glasses, even though the sulfur concentration was only 0.5 to 2 wt% S. Peaks at 324-336 ppm, attributed to sulfate were observed in all glasses. No quadrupolar lineshape was observed, but that is typical of spectra of quadrupolar nuclei in glasses, where a range of environments usually leads to an asymmetric broadened line (Kohn et al. 1998). In addition one sample showed an additional peak at 360 ppm that was assigned to thiosulfate. However it was suggested that the thiosulfate was present in the sample as particles of nanocrystalline thiosulfate at the surface of the glass rather than as a species dissolved in the glass. O’Dell et al. (2008) applied some of the latest techniques for enhancing 33S NMR signals to a hydrous glass containing 1.15 wt% of 99%-enriched S. They showed that the MAS signal could be improved by a factor of up to 1.85 (see Fig. 12) with no line-shape distortion. The data suggest that natural abundance studies of 33S in glasses are not yet feasible, but with continuing technological improvements they may be possible in the future. Recently Klimm et al. (submitted) reported 33S MAS NMR spectra for a much wider range of glass compositions, and used improved experimental conditions compared with Couch et al. (2004), most notably a magnetic field strength of 18.8 T. The hydrous glasses include samples synthesized at both oxidizing and reducing conditions, where S is expected to be present as sulfate and sulfide, respectively. XANES measurements were made on the same samples to provide a wellcharacterized value for the S6+/SS ratio for each sample. Representative spectra from the Klimm et al. (submitted) study are shown in Figure 13. Two samples show peaks at about −200
Figure 12. Three separate, normalized, 33S MAS NMR spectra of the same sulphate-bearing silicate glass, showing the signal enhancement that is possible using rotor assisted population transfer (RAPT) and hyperbolic secant adiabatic pulse sequences (redrawn after O’Dell et al. 2008)
Spectroscopic Studies of Sulfur Speciation in Glasses
65
!
Figure 13. S MAS NMR spectra of a range of S-containing silicate glasses equilibrated at different oxygen fugacities (after Klimm et al. submitted). 33
ppm that are attributed to sulfide. The peaks are of different widths and slightly different positions in the two samples suggesting some compositional control (RS50 has a trondhjemitic composition whereas RS11 is albite). No crystalline sulfides have peaks at −200 ppm, and it is tentatively suggested that this resonance can be attributed to dissolved molecular H2S in the glasses. Potentially, it could be also related to the same sulfide species that gives rise to the Raman band that was assigned to HS− (Klimm and Botcharnikov 2010, Klimm et al. submitted, see also below). All the other samples have compositions of either sodium calcium silicate (RS55, RS19), rubidium silicate (RS54), sodium silicate (RS16) or potassium silicate (RS53, RS41, RS18) and show sulfate resonances at about 330 ppm. There are a few small additional features in some spectra, but none are completely reproducible, and the evidence for other structural species in the glass is not convincing. It is interesting to note that some of the sulfate resonances are broader than others, and a range of average Cq from 1.1-1.4 MHz is derived from fitting. It was also observed that some samples where both sulfide and sulfate are known to be present (from XANES) only showed sulfate resonances, implying that dissolved sulfide is not observable in some compositions. This could be because of a large dispersion of the chemical shift (remember the wide range in shifts observed previously for model compounds) or a large Cq. The latter is not consistent with the model compounds, but perhaps sulfide dissolved in silicate melts has a much more asymmetric environment than simple crystalline sulfides. The large dispersion of chemical shift would be consistent with the results by XANES as well as by simulations (Machacek et al. 2010). These indicate that S2− is probably predominantly associated to network-modifying cations, which very likely would result in a relatively large dispersion of sulfur sites in the glass structure. In conclusion, NMR has not yet fulfilled its promise as a method to obtained detailed structural information on S dissolution mechanisms in glasses. The low frequency and natural abundance of 33S make the collection of spectra with adequate signal-to-noise ratio challenging. The key to collecting useful spectra is use of the highest possible magnetic fields. A decade or so ago 14.1 T was the highest commercially available magnetic field for an NMR spectrometer,
66
Wilke, Klimm, Kohn
but now 18.8 T, 21.1 T and even 23.5 T instruments are becoming available. The higher fields dramatically increase signal-to-noise ratio and decrease linewidths. Although difficult, NMR should continue to be developed as a complement to XANES and Raman spectroscopy.
raMaN aNd Ir SpeCtroSCopy Vibrational spectroscopy such as Raman or Infrared spectroscopy provides direct information on structural properties of materials by measuring vibrational transitions of bonded atoms in crystalline and non-crystalline structures or molecular units. Raman spectroscopy arises from the Raman effect or Raman scattering that is the inelastic scattering of photons, which results in a frequency shift of the scattered light in comparison to the excitation frequency. A requirement for the observation of the Raman effect is the presence of Raman active vibrational modes within the matter of interest. Several normal vibrational modes can be distinguished: n1, n2, n3 and n4, which corresponds to symmetric stretching, symmetric bending, asymmetric stretching and bending of a bond, respectively (often described as A1 for n1, E for n2 and F for n3 and n4. The theoretical and experimental aspects of Raman spectroscopy and its application have been extensively reviewed in the past (e.g., Karr 1975; McMillan 1985; Smith and Dent 2005) including its application on silicate glasses (McMillan and Wolf 1995) and on sulfide and sulfate compounds (Myneni 2000; Wincott and Vaughan 2006). As far as experimental details are concerned, we will only discuss a few important issues for application to sulfur in silicate glass. Raman spectroscopy has already been applied successfully to investigate sulfur-bearing fluid inclusions in natural minerals (e.g., Rosasco et al. 1975; Rosasco and Roedder 1979; Beny et al. 1982; Burke 2001). Bands related to sulfur species have been assigned by a comparison of band positions to bands occurring in Raman spectra of sulfur reference compounds. However, the number of studies applying Raman spectroscopy to study sulfur in glasses is still rather limited as will be shown below. Infrared spectroscopy is also frequently applied to crystalline sulfides and sulfates to study the vibrational modes of these compounds (e.g., Wincott and Vaughan 2006; Ross 1974). For glasses, only a few studies can be found that report infrared spectra focused on sulfur. These are mostly restricted to chalcogenide-based glasses with very high sulfur contents or other compositions of technical interest (e.g., Mei et al. 2003; Kim et al. 2005; Santagneli et al. 2008). To our knowledge, no studies report infrared spectra focused on sulfur in silicate glasses. This is probably because sulfur concentrations are low and it is rather difficult to separate weak but broad bands from the background. In addition, bonds that are visible with Raman spectroscopy are usually weaker in IR spectra and vice versa because of the selection rules for vibrational transitions. This has been shown experimentally for instance for H2O (e.g., Karr 1975; Smith and Dent 2005).
raman spectroscopy of sulfur model compounds Raman spectroscopy has been applied routinely as a fingerprinting method in order to quickly and easily identify minerals of geological interest (e.g., Mernagh and Trudu 1993; Sarma et al. 1998; Hope et al. 2001; White 2009). Raman spectra of minerals have been collected over the years and have become available online as databases or libraries, such as the RRUFF database at www.RRUFF.info (Downs 2006), and can be used as reference spectra. Most common natural sulfur-bearing minerals incorporate sulfur as sulfate (S6+) or sulfide (S2−). Because the structural environments of S6+ and S2− in solid sulfur compounds are fundamentally different, with S6+ being connected to oxygen as a SO42− group and S2− mostly being connected to metallic cations (e.g., FeS, FeS2, NiS), the wavenumbers at which Raman bands due to S are observed are fundamentally different and thus allow S2− and S6+ to be distinguished in Raman spectra. For S6+ the most intense features observed in Raman spectra are related to the n1 S-O
Spectroscopic Studies of Sulfur Speciation in Glasses
67
[n1(SO4)] stretching mode of sulfate and occur at wavenumbers of 1000±25 cm−1 depending on the exact composition and structure of the sulfate compound (e.g., Rosasco and Roedder 1979; Beny et al. 1982; Burke 2001; McKeown et al. 2001; Tsujimura et al. 2004; Lenoir et al. 2009; White 2009). For instance, the Raman spectra of anhydrite (CaSO4) display the highest peak at 1016 cm−1 corresponding to n1(SO4) bands whereas at 497 cm−1 a much smaller band is characteristic for the symmetric bending mode of the SO42− compound [n2(SO4)] (Fig. 14, Sarma et al. 1998). For gypsum, which is the hydrated form of CaSO4 (·2H2O), n1(SO4) and n2(SO4) are observed at slightly lower wavenumbers of 1008 and 495 cm−1, respectively. Other sulfates such as potassium-bearing sulfates such as mercallite (KHSO4) have n1(SO4) shifted to 977 cm−1. For sulfides, the most intense bands occur at much lower wavenumbers in the range of 250 to 450 cm−1 (Hope et al. 2001; Mernagh and Trudu 1993; White 2009). For instance, pyrite and marcasite (both as FeS2) display the most intense bands at 375 and 325 cm−1, respectively, and these are also related to n1 stretching modes (Fig. 14). Sulfides with cations of a less metallic character show the most intense features at lower wavenumbers (e.g., MgS with n1 at ~300 cm−1, Siebert et al. 2004).
experimental details for raman spectroscopy on glasses Modern experimental setups for Raman spectroscopy are often based on a confocal microscope, which allows a spatially resolved analysis in the micrometer range. A major advantage of this setup is that no special sample preparation is usually required and materials can be exposed directly to the laser beam. With a confocal micro Raman setup, the laser beam can be focused easily on the surface of the sample. Therefore, glass chips can be mounted directly under the laser beam. However, care has to be taken that the surface is carefully cleaned because a thin film of organic solvents or microcrystalline dust particles may completely overly the signal of the glass. If band intensities are to be used for quantification more efforts should be dedicated to produce a polished sample surface in order to minimize any effects of the surface to the measured signal. Since a confocal setup restricts the acquisition of signal to a well-defined volume, micro-Raman spectroscopic studies on fluid inclusion trapped within natural minerals are possible (e.g., Rosasco et al. 1975; Rosasco and Roedder 1979; Beny et al. 1982; Burke 2001). This allows the determination of sulfur speciation below the sample surface and may be of interest when determining the sulfur speciation of melt inclusions, which are not exposed to the surface. One problem for the identification of Raman bands related to sulfur in glasses is the relatively low signal of the bands due to the low sulfur concentration in the glass. In addition, the bands are relatively broad due to disorder within the glass structure in contrast to very sharp bands observed in solid crystalline sulfur compounds. This is especially problematic in the case of natural glasses where sulfur abundance barely exceeds 1000 ppm and sulfur might be present in various valence states (S2− and S6+). In this case, the counting statistics of the spectra should be optimal in order to have a good signal to noise ratio. This can be achieved by increasing the laser intensity and acquisition time or by performing multiple scans that sum up to one spectrum. Increasing the laser intensity and acquisition time may lead to the saturation of the detector and is thus intrinsically limited, but it varies for each specific Raman spectrometer setup (Klimm and Botcharnikov 2010). Very high laser intensities may also damage the glass. Typical acquisition times for Raman spectra of sulfur bearing glasses are in the range of 20-1000 s for the 514.5 nm wavelength Ar+ laser and the 532 nm wavelength ND:YAG laser (e.g., Hapanowicz and Condrate 1996; Tsujimura et al. 2004; Lenoir et al. 2009; Klimm and Botcharnikov 2010). With these conditions sulfur species have been identified down to concentrations of ~0.01 wt% S for the glass compositions studied (Klimm and Botcharnikov 2010). It also has to be noted that background fluorescence is a common phenomenon in Raman spectroscopy, which can be induced by various parameters of the sample (e.g., transition metals, organics of epoxy glue). For instance, Klimm and Botcharnikov (2010) reported fluorescence in the Raman spectrum of potassium silicate glasses. The spectrum showed a sudden overall
Wilke, Klimm, Kohn
68
model compounds
1017 cm-1
Anhydrite 1008 cm-1
Gypsum
Normalized Intensity
1016 cm-1
Arcanite 977 cm-1
Mercallite 374 cm-1
Pyrite 324 cm-1
Marcasite 200
400
600
800
-1
Wavenumber [cm ]
1000
1200
Figure 14. Raman spectra of selected sulfate and sulfide reference compounds from the public RRUFF database at www.rruff.info (Downs 2006): Anhydrite, CaSO4 (RRUFFID: R061102); Gypsum CaSO4·2H2O (RRUFFID: X050096); Arcanite K2SO4 (RRUFFID: R070040); Mercallite, KHSO4 (RRUFFID: R070072); Pyrite, FeS2 (RRUFFID: ID060882); Marcasite, FeS2 (RRUFFID: R070692). Numbers in italics indicate wavenumbers of the n1-mode.
Spectroscopic Studies of Sulfur Speciation in Glasses
69
increase in intensities over a broad range of wavenumbers. However, in this case the information of the spectrum was still visible. Fluorescence is especially problematic for determining peak areas to quantifying the abundance of a certain species in the glass because of the difficulty in performing a reliable background subtraction. The only way to circumvent fluorescence is probably to use a laser with a different energy that does not excite fluorescence.
raman spectroscopy on sulfur in glasses Determining sulfur species in glasses by Raman spectroscopy is associated with difficulties due to the, sometimes, uncertain interpretation of the spectra. A Raman spectrum reflects all Raman active modes of the measured sample. Therefore, Raman spectra of sulfur-bearing glasses are a summary of vibrational modes related to the glass structure—for example, Si-O bonds, those related to sulfur bonds (e.g., S-O), and if present those related to O-H or C-O bonds. Due to the relatively low solubility of sulfur in glasses, especially in those of geological interest (see the contributions in this volume by Backnaes and Deubener 2011 Baker and Moretti 2011, and Ebel 2011) the resulting bands related to sulfur (especially for sulfide) are relatively weak and often make a correct identification very difficult. In addition, it is difficult to synthesize appropriate glasses containing sulfur in various valence states because the synthesis has to be performed under controlled redox conditions. Most information has been gathered on the presence of sulfate because syntheses of oxidized glasses are generally easier to perform, so that little information has been obtained on reduced sulfur species using Raman spectroscopy. Usually Raman spectra of sulfur containing glasses are compared to Raman spectra of solid reference compounds or liquids. The similarity of the position of bands to sulfur-bearing reference compounds is commonly used as a confirmation of the presence of a specific sulfur species in the measured glass (“fingerprint method”). However, because sulfur is dissolved in the glass and because of differences in coordination between glasses and crystals, bands are usually shifted by a few wavenumbers compared to reference compounds. Only a limited number of studies are available that specifically investigated sulfur incorporated in silicate glasses because of the low band intensity and the difficulty in assigning and interpreting Raman bands for S in glasses, Raman spectroscopic investigations on sulfur in glasses have therefore been performed mainly on glasses of technical interest. The compositions studied include alkali silicate glasses (Konijnendijk and Buster 1977; Hapanowicz and Condrate 1996; Ooura and Hanada 1998; Tsujimura et al. 2004; Klimm and Botcharnikov 2010; Fig. 15), borosilicate glasses (McKeown et al. 2001; Manara et al. 2007; Lenoir et al. 2009) and borate glasses (Ahmed et al. 1997). Recently, the results of Raman spectroscopic investigations of sulfur in hydrous natural glass compositions such as basalt, andesite and rhyodacite have been published (Klimm and Botcharnikov 2010). As shown in Figure 15, most oxidized glasses show a significant band in the range of ~970 to 1010 cm−1 which corresponds to the n1(SO4) band. The range of wavenumbers observed for the n1(SO4) band in glasses is related to differences in the corresponding glass structure, i.e., it is a function of the content and type of network-modifying cations. The band is shifted to high wavenumbers for cations with high ionic field strength (e.g., Li). The largest variations of the n1(SO4) band are observed for borate glasses with various cations (McKeown et al. 2001). For other simple glass compositions, such as potassium silicate, sodium silicate, soda lime silicate or borosilicate glasses the n1(SO4) band occurs at 990 cm−1 (Hapanowicz and Condrate 1996; McKeown et al. 2001; Tsujimura et al. 2004; Manara et al. 2007; Lenoir et al. 2009; Klimm and Botcharnikov 2010). Although bands related to S6+ have not been assigned in Raman spectra of borate glasses by Ahmed et al. (1997), some spectra also display a band at 990 cm−1, indicating that S6+ is present in those glasses. In more complex natural glasses such as basalt, andesite and rhyodacite the n1(SO4) band is observed at 1000 cm−1 (Klimm and Botcharnikov 2010). It should be noted that the n1(SO4) band in glasses is often superimposed by the intense bands of the Si-O vibrations in this spectral range, which may complicate correct identification and separation of
Wilke, Klimm, Kohn
70
HS-
SO42-
Fe-S
a) Soda lime glass
ce
u ed
Figure 15. Raman spectra of hydrous S-bearing glass. a) Soda-lime silicate glass (0.36 wt% S2− and 0.54 wt% S6+, respectively), synthesized at 200 MPa, 1000 °C: In reduced samples bands can be assigned to S as sulfide. In oxidized glass only bands related to sulfate are observed. b) Basaltic glass, synthesized at 200 MPa. 1050 °C (0.13 wt% S2− and 0.60 wt% S6+, respectively; data from Klimm and Botcharnikov 2010).
oxidized
HS-
SO42-
b) Basalt Fe-S
Counts
r
d
reduced
oxidized 500
1000
1500
2000
-1
Wavenumber [cm ]
2500
3000
the sulfur band. In critical cases, where the n1(SO4) band is completely superimposed by bands related to the silicate network as demonstrated for relatively polymerized geological relevant glass compositions with low S solubility, the detection of this band is limited to sulfur contents above 150 ppm S6+ (Klimm and Botcharnikov 2010; Klimm et al. submitted). Assignment of bands in Raman spectra related to reduced sulfur incorporated in glasses has been not so straightforward. Ahmed et al. (1997) suggested that S3−, S2−, S22−, S32−, S52− and S62− can be identified in Raman spectra of borate glasses. However, the bands related to these polysulfides are rather weak and the presented spectra are quite noisy, so that the band assignment might be an over-interpretation. McKeown et al. (2001) and Tsujimura et al. (2004) reported bands in the range of 350 to 380 cm−1, at ~ 300 cm−1 and at ~460 cm−1 in spectra taken on sodium silicate and borosilicate glasses. The band at 460 cm−1 is close to the frequency of the n2(SO4) band. Bands between 300 and 380 cm−1 could indicate S-S bonds or intermediate S species. Therefore, the authors speculated on the presence of polysulfide (e.g., S62−), dithionate (S5+) or thiosulfite (S4+), because l(S Ka) analysis and S K-edge XANES of these glasses suggest the presence of reduced sulfur species. These observations are, at least partially, in contrast to those made by Klimm and Botcharnikov (2010), who investigated S in either fully oxidized or reduced hydrous glasses. In simple glass compositions, such as potassium silicate and soda lime glass with a few wt. % of water (below water saturation) a band at 2574 cm−1 appears at reducing conditions, which was assigned to the presence of HS− in these glasses (Rosasco and Roedder 1979, see Fig. 15). This band was also observed by Stelling et al. (2011) for hydrous sulfide-bearing Na-Si and Na-Ca-Si glasses. The slightly asymmetric feature of the band at 2574 cm−1 can be explained by the presence of additional H2S in the glass as it has also been demonstrated for natural fluid inclusions (Rosasco and Roedder 1979). No bands
Spectroscopic Studies of Sulfur Speciation in Glasses
71
were detected in the 300-400 cm−1 region that could have been assigned to any sulfide species. Klimm and Botcharnikov (2010) did not observe the HS− band in reduced glasses of hydrous basaltic, andesitic and rhyodacitic compositions. In contrast to hydrous S-free basaltic, andesitic and rhyodacitic glasses, the reduced S-bearing ones show a band at 400 cm−1, which can be related to iron-sulfide bonding within the silicate glass structure when compared to sulfide reference compounds. Based on this observation, Klimm and Botcharnikov (2010) suggested that the dissolution mechanism of sulfide in silicate glasses is different in iron bearing and iron-free systems, which is in agreement with observations by XANES (Fleet et al. 2005b, see also above) and is consistent with the correlation of sulfur solubility with the iron content of the melt (e.g., Wallace and Carmichael 1992, Carroll and Webster 1994; also see in this volume Backnaes and Deubener 2011 and Baker and Moretti 2011).
beam damage by raman Spectroscopy Unlike other spectroscopic studies using intense X-ray or electron beams (S K-edge XANES and l(S Ka) wavelength shift), no beam damage has so far been reported in studies investigating sulfur speciation by Raman spectroscopy. However, it is known that during laser irradiation the temperature of glasses increases. By using a high-intensity laser, such a temperature increase may lead to degassing of the volatiles (e.g., H2O, CO2, O2) that are dissolved in the glass. The exsolved gases form bubbles within the glass, which can be observed microscopically. This process is often described as “boiling” and is mostly observed in dark iron-bearing glass compositions. Hapanowicz et al. (1995) described such a degassing process for lead silicate glasses after the irradiation of the glass with a high intensity 4-kW CO2 laser and subsequent analysis with Raman spectroscopy. The bubbles formed contained oxygen, which had exsolved from the glass although no change in the silicate glass structure was observed. Advances in detector sensitivity allow the use of lasers with shorter wavelength and less power as well as reduced acquisition times, minimizing the effect of heating during laser beam exposure. For instance, Klimm and Botcharnikov (2010) did not observe any beam damage such as “boiling” during laser beam exposure or changes in the spectra during acquisition using the 532 nm line of a Nd:YAG laser on hydrous silicate glasses. Increasing the acquisition time or laser intensity during spectrum acquisition resulted in saturation of the detector rather than damaging the silicate glasses. However, it has to be noted that laser beam damage may occur during Raman spectroscopic investigations and glass samples should be carefully investigated after laser beam exposure. Apart from “boiling” the S species present in the glass may change during irradiation, which should be tested by performing time-dependent measurements.
determination of the oxidation state using raman spectroscopy So far, it has been demonstrated that Raman spectroscopy on silicate glasses allows sulfate (S ) and sulfide (S2−) species in silicate glasses to be identified. However, in order to determine the S6+/SS ratio of sulfur dissolved in glasses, in a comparable way to (S Ka) wavelength shift measurement and S K-edge XANES spectroscopy, a quantification of each species present in a Raman spectrum is required. Quantification has only been successfully performed for the sulfate content in oxidized borosilicate glasses (Fig. 16, Lenoir et al. 2009). The intensity of the sulfate band is directly related to the concentration of sulfate in the glass. Because the sulfate band is often superimposed by bands related to the silicate glass structure, a deconvolution of the wavelength region where the sulfate and silicate bands occur is necessary in order to determine the intensity of the band related to sulfate. In addition to the sulfate concentration, the intensity of this band also depends on the bulk glass compositions. Therefore, a calibration of the band intensity as a function of sulfur content for each glass composition is required in order to quantify the S6+ abundance in glasses. Furthermore, sulfur bands should be scaled to bands of the glass network in order to eliminate effects of the Raman measurement conditions. The same calibration and scaling has to be applied to other sulfur species such as S2−. However, 6+
Wilke, Klimm, Kohn
72
Borosilicate
Intensity [a.u.]
990 cm-1
S6+
900
1000
1100
1200 -1
Wavenumber [cm ] Figure 16. Deconvolution of baseline-corrected Raman spectrum of Na borosilicate glass in the region 850-1250 cm−1 in order to extract the intensity of the n1 band of sulfate (Redrawn after Lenoir et al. 2009). Spectrum was fitted using four Gaussians to describe bands related to the silicate stretching vibrations and one Gaussian for the sulfate band (shaded).
considering the small band intensity for S2− compared to S6+, especially in natural glasses (Klimm and Botcharnikov 2010) an accurate quantification of S6+/SS is difficult if possible at all. Thus, the determination of this parameter by Raman spectroscopy remains the object of future Raman spectroscopic investigations and might only be possible in specific systems. Nonetheless, the qualitative detection of different sulfur species allows a rough estimation of the fO2 during formation by Raman spectroscopy; if both, S6+ and S2− coexist in a glass a rough estimate of the corresponding fO2 is possible. Assuming that the S6+/SS ratio is simply a function of fO2 and other parameters such as composition, pressure and temperature have minor effects on the S6+/SS equilibrium, such conditions are in the range of QFM = 0 to QFM+2 according to the latest calibration of S6+/SS vs. fO2 (Jugo et al. 2010). If only S2− is determined by Raman spectroscopy then fO2 ≤ QFM and, accordingly, if only S6+ is determined then fO2 ≥ QFM+2. Thus, when applied to natural glass samples (e.g., melt inclusions) a preliminary estimation of the redox conditions is possible, which can be further constrained by more robust methods.
SuMMary aNd outlooK The various spectroscopic techniques proposed here provide valuable experimental insight into the incorporation of sulfur in silicate glasses and melts and substantially contribute to a better understanding of sulfur speciation in these compounds. Each method shown here has its advantages and disadvantages and the choice of which one to use depends on the problem and the specific questions to be answered. The method easiest to apply is certainly Raman spectroscopy due to its widespread availability, which should give at least qualitative data. S Ka wavelength analysis can be used as an add-on of EPMA. However, the S oxidation state determined might be strongly affected by beam damage. X-ray induced high-resolution X-ray emission analysis might be a possible way to go in the future. Satisfactory 33S NMR data are difficult to obtain particularly due to low abundance of the relevant isotope and the potential of this method strongly depends on the availability of strong magnetic fields. If data can be obtained they provide very detailed insight to the species present in the glass at least for oxidized samples, but more methodological developments are required. XANES is the most elaborate approach since access to beamtime at synchrotron radiation facilities is limited. However, this
Spectroscopic Studies of Sulfur Speciation in Glasses
73
method probably provides the deepest and most robust insight to the sulfur species. On one hand, the S oxidation state can be determined quantitatively with only small artifacts from the analysis. On the other, these spectra can be used to constrain models for the S incorporation in melts and glass by the use of ab initio methods to simulate XANES spectra, which has only been applied to S in limited cases (e.g., Kravtsova et al. 2004) and not yet to S in glasses. All techniques provide evidence that S2− and S6+ are the dominant species for S in silicate glasses. Most of the earlier evidence for the presence of S4+ in glass can now be explained in terms of experimental artifacts. Spectroscopic evidence points out that S6+ is present in glasses within the sulfate anion. The sulfate anion forms an isolated unit that is not connected to the network of SiO4 tetrahedra via bridging oxygens. The sulfide anion appears to predominantly replace non-bridging oxygens, based on spectroscopic evidence and the very first molecular simulation study (Machacek et al. 2010). The latter also indicates that free S2− is not relevant in silicate melts. Spectroscopic as well sulfur solubility data suggest association of Fe and S2− in silicate melts. Apart from this observation, there is no systematic understanding of interactions or association of S2− with other components of the melt. Further insight could come from quantitative understanding of experimental spectra with ab initio simulation of spectra, particularly if they are combined with the outcome of computational molecular dynamics studies. The structure of glasses represents that of the melt frozen-in at the temperature of the glass transition, so that the S species found in the glass might be different from those present at high temperature in the melt. Experiments looking at sulfur speciation in situ at high temperature and/ or high pressure remain a field for future research. Such studies might resolve the question of the stability of S4+ in melts. Although measurements at ambient pressure are feasible using both X-ray and Raman techniques, the low energy of S Ka and S K-edge means that measurements at high pressure are rather unlikely using X-ray spectroscopy, but might be possible by Raman spectroscopy.
aCKNoWledGeMeNtS We thank R. Alonso Mori for providing data of XANES spectra. We appreciate reviews by M. Fleet, D. Neuville as well as comments from H. Behrens that were very helpful for improving the manuscript.
reFereNCeS Ahmed AA, Sharaf, NA, Condrate RA (1997) Raman microprobe investigation of sulfur-doped alkali borate glasses. J Non-Cryst Solids 210:59-69 Alonso Mori R, Paris E, Giuli G, Eeckhout SG, Kavcic M, Zitnik M, Bucar K, Pettersson LGM, Glatzel P (2009) Electronic structure of sulfur studied by X-ray absorption and emission spectroscopy. Anal Chem 81:6516-6525 Ashbrook SE, Duer MJ (2006) Structural Information from quadrupolar nuclei in Solid State NMR. Concepts in Magnetic Resonance Part A, Vol. 28A(3) 183–248 Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Backnaes L, Stelling J, Behrens H, Goettlicher J, Mangold S, Verheijen O, Beerkens RGC, Deubener J (2008) Dissolution mechanisms of tetravalent sulfur in silicate melts: evidences from sulfur K edge XANES studies on glass. J Am Ceram Soc 91:721-727 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Barbarella G (1993) Sulfur-33 NMR. Progress in NMR spectroscopy 25:317-343 Beerkens RGC (2003) Amber chromophore formation in sulfur- and iron-containing soda-lime-silica glasses. Glass Sci Technol 76:166-175
74
Wilke, Klimm, Kohn
Beny C, Guilhaumou N, Touray JC (1982) Native-sulfur-bearing fluid inclusions in the CO2-H2O-H2O-S-system - micro-thermometry and Raman micro-probe (Mole) analysis - thermochemical interpretations. Chem Geol 37:113-127 Beran A, Libowitzky E (eds) (2004) Spectroscopic methods in mineralogy. EMU Notes in Mineralogy, 6, Eötvös University Press Budapest Bingham PA, Connelly AJ, Hand RJ, Hyatt NC, Northrup PA, Alonso Mori R, Glatzel P, Kavčič M, Žitnik M, Bučar K, Edge R (2010) A multi-spectroscopic investigation of sulphur speciation in silicate glasses and slags. Eur J Glass Sci Technol A 51:63-80 Bonnin-Mosbah M, Métrich N, Susini J, Salome M, Massare D, Menez B (2002) Micro X-ray absorption near edge structure at the sulfur and iron K-edges in natural silicate glasses. Spectrochim Acta B 57:711-725 Brendebach B, Denecke MA, Roth G, Weisenburger S (2009) Sulfur incorporation in high level nuclear waste glass: a S K-edge XAFS investigation. J Phys Conf Ser 190:012186 Bunker G (2010) Introduction to XAFS: A Practical Guide to X-ray Absorption Fine Structure Spectroscopy. Cambridge University Press Burke EAJ (2001) Raman microspectrometry of fluid inclusions. Lithos 55:139-158 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation states: results from measured wavelength shifts of sulfur X-rays. Am Mineral 73:845-849 Carroll MR, Webster JD (1994) Solubilities of sulfur, noble gases, nitrogen, chlorine, and fluorine in magmas. Rev Mineral 30: 231-279 Chappell BW, White AJR (1968) X-ray spectrographic determination of sulfur coordination in scapolite. Am Mineral 53:1735-1738 Chesnut DB, Quin LD (2004) 33S NMR shieldings and chemical bonding in compounds of sulfur. Heteroat Chem 15:216-224 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulfur in hydrous rhyolitic melts. J Petrol 45:21712196 Connolly JW, Haughton DR (1972) Valence of sulfur in glass of basaltic composition formed under conditions of low oxidation potential. Am Mineral 57:1515-1517 Cottrell E, Kelley KA, Lanzirotti A, Fischer RA (2009) High-precision determination of iron oxidation state in silicate glasses using XANES. Chem Geol 268:167-179 Couch S, Howes AP, Kohn SC, Smith ME (2004) 33S solid state NMR of sulfur speciation in silicate glasses. Solid State NMR 26:203-208 Daunch WA, Rinaldi PL (1996) Natural-abundance solid-state 33S NMR with high-speed magic angle spinning. J Magn Reson Ser A 123:219-221 de Hoog JCM, Hattori KH, Hoblitt RP (2004) Oxidized sulfur-rich mafic magma at Mount Pinatubo, Philippines. Contrib Mineral Petrol 146:750-761 de Lacaillerie JBD, Barberon F, Bresson B, Fonollosa P, Zann H, Fedorov VE, Naumov, Gan NGZ (2006) Applicability of natural abundance 33S solid-state NMR to cement chemistry. Cement Concrete Res 36:1781-1783 Downs RT (2006) The RRUFF project: an integrated study of the chemistry, crystallography, Raman and infrared spectroscopy of minerals. Program and Abstracts of the 19th General Meeting of the International Mineralogical Association in Kobe, Japan, O03-13 Ebel DS (2011) Sulfur in extraterrestrial bodies and the deep earth. Rev Mineral Geochem 73:315-336 Eckert H, Yesinowski JP (1986) 33S NMR at natural abundance in solids. J Am Chem Soc 108:2140-2146 Evans KA, O’Neill HStC, Mavrogenes JA, Keller NS, Jang LY, Lee JF (2009) XANES evidence for sulfur speciation in Mn-, Ni- and W-bearing silicate melts. Geochim Cosmochim Acta 73: 6847-6867 Faessler A, Goehring M (1952) Rontgenspektrum und Bindungszustand - Die K-Alpha-Fluoreszenzstrahlung des Schwefels. Naturwissenschaften 39:169-177 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Farges F, Munoz M, Siewert R, Malavergne V, Brown GE Jr, Behrens H, Nowak M, Petit PE (2001) Transition elements in water-bearing silicate glasses/melts. Part II. Ni in water-bearing glasses. Geochim Cosmochim Acta 65:1679-1693 Farrell SP, Fleet ME (1999) Evolution of local electronic structure in cubic Mg1-xFexS by S K-edge XANES spectroscopy. Solid State Commun 113:69-72 Farrell SP, Fleet ME (2001) Sulfur K-edge XANES study of local electronic structure in ternary monosulfide solid solution [(Fe,Co,Ni]0.923S]. Phys Chem Miner 28:17-27 Farrell SP, Fleet ME, Stekhin IE, Kravtsova A, Soldatov AV, Liu XY (2002) Evolution of local electronic structure in alabandite and niningerirte solid solutions [(Mn,Fe)S, (Mg,Mn)S, (Mg,Fe)S] using sulfur Kand L-edge XANES spectroscopy. Am Mineral 87:1321-1332 Fechtelkord M (2004) Solid state NMR spectroscopy as supporting method in Rietveld refinements of rockforming minerals: New developments and examples. EMU Notes Mineral 6:421-463
Spectroscopic Studies of Sulfur Speciation in Glasses
75
Figueiredo MO, da Silva TP (2009) Effect of oxygen sharing on the white line of S K-edge XANES spectra of sulfate minerals. Eur J Mineral 21:79-83 Fleet ME (2005) XANES spectroscopy of sulfur in Earth materials. Can Mineral 43:1811-1838 Fleet ME, Liu X, Harmer SL, King, PL (2005b) Sulfur K-edge XANES spectroscopy: Chemical state and content of sulfur in silicate glasses. Can Mineral 43:1605-1618 Fleet ME, Liu XY, Harmer SL, Nesbitt HW (2005a) Chemical state of sulfur in natural and synthetic lazurite by S K-edge XANES and X-ray photoelectron spectroscopy. Can Mineral 43:1589-1603 George GN, Gorbaty ML (1989) Sulfur K-edge X-ray absorption spectroscopy of petroleum asphaltenes and model compounds. J Am Chem Soc 111:3182-3186 Gurenko AA, Schmincke HU (1998) Petrology, geochemistry, S, Cl and F abundances, and S oxidation state of sideromelane glass shards from Pleistocene ash layers north and south of Gran Canaria (ODP Leg 157). Contrib Mineral Petrol 131:95-110 Gurenko AA, Schmincke HU (2000) S concentrations and its speciation in Miocene basaltic magmas north and south of Gran Canaria (Canary Islands): Constraints from glass inclusions in olivine and clinopyroxene. Geochim Cosmochim Acta 64:2321-2337 Hansen MR, Brorson M, Bildsøe H, Skibsted J, Jakobsen HJ (2008) Sensitivity enhancement in naturalabundance solid-state 33S MAS NMR spectroscopy employing adiabatic inversion pulses to the satellite transitions. J Magn Reson 190:316-326 Hapanowicz RP, Condrate RA (1996) Raman spectral investigation of sulfate inclusions in sodium calcium silicate glasses. J Solid State Chem 123:183-185 Hapanowicz RP, Condrate RA, Buerhop C (1995) Raman microprobe study of bubbles formed because of laser damage in glasses. Applied Spectroscopy 49:1369-1370 Hawthorne FC (ed) (1988) Spectroscopic Methods in Mineralogy and Geology. Reviews in Mineralology, Volume 18. Mineralogical Society of America Hitchcock AP, Bodeur S, Tronc M (1987) Sulfur and chlorine K-shell X-ray absorption spectra of SCl2, S2Cl2, SOCl2, and SO2Cl2. Chem Phys 115:93-101 Hoefer HE, Brey GP (2007) The iron oxidation state of garnet by electron microprobe: Its determination with the flank method combined with major-element analysis. Am Mineral 92:873-885 Hope GA, Woods, R, Munce CG (2001) Raman microprobe mineral identification. Miner Eng 14:1565-1577 Jackowski K (2001) Gas-phase 17O and 33S NMR spectroscopy. J Mol Struct 563-564:159-162 Jakobsen HJ, Bildsøe H, Skibsted J, Brorson M, Gor’kov P, Gan Z (2010) A strategy for acquisition and analysis of complex natural abundance 33S solid-state NMR spectra of a disordered tetrathio transition-metal anion. J Magn Reson 202:173-179 Jakobsen HJ, Bildsøe H, Skibsted J, Brorson M, Srinivasan BR, Nätherd C, Bensch W (2009) New opportunities in acquisition and analysis of natural abundance complex solid-state 33S MAS NMR spectra: (CH3NH3)2WS4, Phys Chem Chem Phys 11:6981-6986 Jakobsen HJ, Hove AR, Bildsøe H, Skibsted J (2006) Satellite transitions in natural abundance solid-state S-33 MAS NMR of alums - Sign change with zero-crossing of C-Q in a variable temperature study. J Magn Reson 180:170-177 Jakobsen HJ, Hove AR, Bildsøe H, Skibsted J, Brorson M (2007) Advancements in natural abundance solidstate 33S MAS NMR: characterization of transition-metal MLS bonds in ammonium tetrathiometallates. Chem Commun 1629-1631 Jokic A, Cutler JN, Ponomarenko E, van der Kamp G, Anderson DW (2003) Organic carbon and sulfur compounds in wetland soils: insights on structure and transformation processes using K-edge XANES and NMR spectroscopy. Geochim Cosmochim Acta 67:2585-2597 Jugo PJ (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37:415-418 Jugo PJ, Luth RW, Richards JP (2005) Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim Cosmochim Acta 69:497-503 Jugo PJ, Wilke M, Botcharnikov RE (2010) Sulfur K-edge XANES analysis of natural and synthetic basaltic glasses: Implications for S speciation and S content as function of oxygen fugacity. Geochim Cosmochim Acta 74:5926-5938 Karr C (1975) Infrared and Raman Spectroscopy of Lunar and Terrestrial Minerals. Academic Press. Kasrai M, Bancroft GM, Brunner RW, Jonasson RG, Brown JR, Tan KH, Feng X (1994) Sulfur speciation in bitumens and asphaltenes by X-ray absorption fine structure spectroscopy. Geochim Cosmochim Acta 58:2865-2872 Kim Y, Saienga J, Martin SW (2005) Glass formation in and structural investigation of Li2S + GeS2 + GeO2 composition using Raman and IR spectroscopy. J Non-Cryst Solids 351:3716-3724 Kirkpatrick RJ (1988) MAS NMR spectroscopy of minerals and glasses. Rev Mineral 18:341-403 Klimm K, Botcharnikov RE (2010) The determination of sulfate and sulfide species in hydrous silicate glasses using Raman spectroscopy. Am Mineral 95:1574-1579
76
Wilke, Klimm, Kohn
Klimm K, Kohn SC, O’Dell LA, Botcharnikov RE, Smith ME (submitted) The dissolution mechanism of sulfur in silicate melts; I. Assessment of analytical techniques in determining sulfur speciation. Geochim Cosmochim Acta Kohn SC (2004) NMR studies of silicate glasses. EMU Notes Mineral 6:399-419 Kohn SC, Smith ME, Dirken PJ, van Eck ERH, Kentgens APM, Dupree R (1998) Sodium environments in dry and hydrous albite glasses. Improved 23Na solid state NMR data and their implications for water dissolution mechanisms. Geochim Cosmochim Acta 62:79-87 Konijnendijk WL, Buster J (1977) Raman-scattering measurements of silicate-glasses containing sulfate. J NonCryst Solids 23:401-418 Kravtsova AN, Stekhin IE, Soldatov AV, Liu X, Fleet ME (2004) Electronic structure of MS (M=Ca,Mg,Fe,Mn): X-ray absorption analysis. Phys Rev B 69:134109/1-134109/12 Kress VC, Carmichael ISE (1991) The compressibility of silicate liquids containing Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on their redox states. Contrib Mineral Petrol 108:82-92 Lenoir M, Grandjean A, Poissonnet S, Neuville DR (2009) Quantitation of sulfate solubility in borosilicate glasses using Raman spectroscopy. J Non-Cryst Solids 355:1468-1473 Li D, Bancroft GM, Kasrai M, Fleet ME, Feng XH, Tan K (1995) S K- and L-edge X-ray absorption spectroscopy of metal sulfides and sulfates: applications in mineralogy and geochemistry. Can Mineral 33:949-960 Li D, Bancroft GM, Kasrai M, Fleet ME, Yang BX, Feng XH, Tan K, Peng MS (1994) S K-edge and L-edge X-ray-absorption spectroscopy of sphalerite, chalcopyrite and stannite. Phys Chem Miner 20:489-499 Machacek J, Gedeon O, Liska M, Marhoul F (2010) Molecular simulations of silicate melts doped with sulfur and nitrogen. J Non-Cryst Solids 356:2458-2464 MacKenzie KJD, Smith ME (2002) Multinuclear Solid State Nuclear Magnetic Resonance of Inorganic Materials. Pergamon Press Manara D, Grandjean A, Pinet O, Dussossoy JL, Neuville DR (2007) Sulfur behavior in silicate glasses and melts: Implications for sulfate incorporation in nuclear waste glasses as a function of alkali cation and V2O5 content. J Non-Cryst Solids 353:12-23 Matthews SJ, Moncrieff DHS, Carroll MR (1999) Empirical calibration of the sulfur valence oxygen barometer from natural and experimental glasses: method and applications. Mineral Mag 63:421-431 McKeown DA, Muller IS, Gan H, Pegg IL, Kendziora CA (2001) Raman studies of sulfur in borosilicate waste glasses: sulfate environments. J Non-Cryst Solids 288:191-199 McKeown DA, Muller IS, Gan H, Pegg IL, Stolte WC (2004) Determination of sulfur environments in borosilicate waste glasses using X-ray absorption near-edge spectroscopy. J Non-Cryst Solids 333:74-84 McMillan PF, Wolf GH (1995) Vibrational spectroscopy of silicate liquids. Rev Mineral 32:247-315 McMillan PF (1985) Vibrational spectroscopy in the mineral sciences. Rev Mineral 14:9-63 Mei Q, Saienga J, Schrooten J, Meyer B, Martin SW (2003) Preparation and characterization of glasses in the Ag2S + B2S3 + GeS2 system. J Non-Cryst Solids 324:264-276 Mernagh TP, Trudu AG (1993) A laser Raman microprobe study of some geologically important sulfide minerals. Chem Geol 103: 113-127 Métrich N, Berry AJ, O’Neill HStC., Susini J (2009) The oxidation state of sulfur in synthetic and natural glasses determined by X-ray absorption spectroscopy. Geochim Cosmochim Acta 73:2382-2399 Métrich N, Bonnin-Mosbah M, Susini J, Menez B, Galoisy L (2002) Presence of sulfite (SIV) in magmas: implications for volcanic sulfur emissions. Geophys Res Lett 29:1538 Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim Cosmochim Acta 60:4151-4160 Mosselmans JFW, Pattrick RAD, Vanderlaan G, Charnock JM, Vaughan DJ, Henderson CMB., Garner CD (1995) X-ray absorption near-edge spectra of transition metal disulfides FeS2 (pyrite and marcasite), CoS2, NiS2 and CuS2, and their isomorphs FeAsS and CoAsS. Phys Chem Miner 22:311-317 Moudrakovski I, Lang S, Patchkovskii S, Ripmeester J (2010) High field 33S solid state NMR and first-principles calculations in potassium sulfates. J Phys Chem A 114:309-316 Musio R (2009) Applications of 33S NMR spectroscopy. Annu Rep NMR Spectrosc 68:1-88 Myneni, SCB (2000) X-ray and vibrational spectroscopy of sulfate in earth materials. Rev Mineral Geochem 40:113-172 Nagashima S, Katsura T (1973) The solubility of sulfur in Na2O-SiO2 melts under various oxygen partial pressures at 1100 °C, 1250 °C and 1300 °C. Bull Chem Soc Jpn 46:3099-3103 Nilsson K, Peach CL (1993) Sulfur speciation, oxidation state, and sulfur concentrations in backarc magmas. Geochim Cosmochim Acta 57:3807-3813 O’Dell LA, Klimm K, Freitas JCC, Kohn SC, Smith ME (2008) 33S MAS NMR of a disordered sulfur-doped silicate: Signal enhancement via RAPT, QCPMG and adiabatic pulses. Appl Magn Reson 35:247-259 Ooura M, Hanada T (1998) Compositional dependence of solubility of sulfate in silicate glasses. Glass Technology 39:68-73
Spectroscopic Studies of Sulfur Speciation in Glasses
77
Paris E, Giuli G, Carroll MR, Davoli I (2001) The valence and speciation of sulfur in glasses by X-ray absorption spectroscopy. Can Mineral 39:331-339 Prietzel J, Thieme J, Neuhäusler U, Susini J, Kögel-Knabner I (2003) Speciation of sulfur in soils and soil particles by X-ray spectromicroscopy. Eur J Soil Science 54:423-443 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Rosasco GJ, Roedder E (1979) Application of a new Raman microprobe spectrometer to nondestructive analysis of sulfate and other ions in individual phases in fluid inclusions in minerals. Geochim Cosmochim Acta 43:1907-1915 Rosasco GJ, Roedder E, Simmons JH (1975) Laser-excited Raman-spectroscopy for nondestructive partial analysis of individual phases in fluid inclusions in minerals. Science 190:557-560 Ross SD (1974) Sulfates and other oxy-anions of Group VI. In: The Infrared Spectra of Minerals. Farmer VC (ed) Mineralogical Society Monograph 4, London, p 423-444 Rowe MC, Kent AJR, Nielsen RL (2007) Determination of sulfur speciation and oxidation state of olivine hosted melt inclusions. Chem Geol 236:303-322 Rowe MC, Kent AJR, Nielsen RL (2009) Subduction influence on oxygen fugacity and trace and volatile elements in basalts across the cascade volcanic arc. J Petrol 50:61-91 Santagneli SH, Schneider J, Skripachev I, Ribeiro SJL, Messaddeq Y (2008) Preparation and characterization of new glassy system As2P2S8-Ga2S3. J Phys Chem B 112:4943-4947 Sarma LP, Prasad PSR, Ravikumar N (1998) Raman spectroscopic study of phase transitions in natural gypsum. J Raman Spectr 29:851-856 Shannon RD (1976) Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Crystallogr A32:751-767 Siebert J, Malavergne V, Guyot F, Combes R, Martinez I (2004) The behaviour of sulfur in metal-silicate core segregation experiments under reducing conditions. Phys Earth Planet Inter 143:433-443 Smith E, Dent G (2005) Modern Raman Spectroscopy: A Practical Approach. Wiley Stebbins JF (1988) NMR spectroscopy and dynamic processes in mineralogy and geochemistry. Rev Mineral 18:405-429 Stelling J (2009) Diffusion, Speziation und Löslichkeit von Schwefel in Silikatschmelzen. PhD Thesis Leibniz Universität Hannover, Germany Stelling J, Behrens H, Wilke M, Göttlicher J, Chalmin E (2011) Interaction between sulfide and H2O in silicate melts. Geochim Cosmochim Acta 75:3542-3557 Sutrisno A, Terskikhb VV, Huang Y (2009) A natural abundance 33S solid-state NMR study of layered transition metal disulfides at ultrahigh magnetic field. Chem Commun 186-188 Sutton SR, Karner J, Papike J, Delaney JS, Shearer C, Newville M, Eng P, Rivers M, Dyar MD (2005) Vanadium K edge XANES of synthetic and natural basaltic glasses and application to microscale oxygen barometry. Geochim Cosmochim Acta 69:2333-2348 Tsujimura T, Xue XY, Kanzaki M, Walter MJ (2004) Sulfur speciation and network structural changes in sodium silicate glasses: Constraints from NMR and Raman spectroscopy. Geochim Cosmochim Acta 68:50815101 Vairavamurthy A, Maletic D, Wang S, Manowitz B, Eglington T, Lyons T (1997) Characterization of sulfurcontaining functional groups in sedimentary humic substances by X-ray Absorption near-edge structure spectroscopy. Energy Fuels 11:546-553 Wagler TA, Daunch WA, Panzner M, Youngs WJ, Rinaldi PL (2004) Solid-state 33S MAS NMR of inorganic sulfates, J Magn Reson 170:336-344 Wagler TA, Daunch WA, Rinaldi PL, Palmer AR (2003) Solid state 33S NMR of inorganic sulfides. J Magn Reson 161:191-197 Wallace PJ, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874 Wallace PJ, Carmichael ISE. (1994) S speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts. Am Mineral 79:161-167 White SN (2009) Laser Raman spectroscopy as a technique for identification of seafloor hydrothermal and cold seep minerals. Chem Geol 259:240-252 Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov RE, Kohn SC, Janousch M (2008) The origin of S4+ detected in silicate glasses by XANES. Am Mineral 93:235-240 Wilke M, Partzsch GM, Bernhardt R, Lattard D (2004) Determination of the iron oxidation state in basaltic glasses using XANES at the K-edge. Chem Geol 213:71-87 Wilke M, Schmidt C, Farges F, Malavergne V, Gautron L, Simionovici A, Hahn M, Petit PE (2006) Structural environment of Fe in water-bearing silicate glass and melt – evidence from X-ray absorption spectroscopy. Chem Geol 229:144-161 Wincott PL, Vaughan DJ (2006) Spectroscopic studies of sulfides. Rev Mineral Geochem 61:181-229
78
Wilke, Klimm, Kohn
Winther KT, Watson EB, Korenowski GM (1998) Magmatic sulfur compounds and sulfur diffusion in albite melt at 1 GPa and 1300-1500 degrees C. Am Mineral 83:1141-1151 Womes M, Karnatak RC, Esteva JM, Lefebvre I, Allan G, Olibier-Fourcade J, Jumas JC (1997) Electronic structures of FeS and FeS2: X-ray absorption spectroscopy and band structure calculations. J Phys Chem Solids 58:345-352 Zwanziger JW (1998) Structure and chemical modification in oxide glasses. Int Rev Phys Chem 17:65-90
4
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 79-111, 2011 Copyright © Mineralogical Society of America
Diffusion and Redox Reactions of Sulfur in Silicate Melts Harald Behrens and Jan Stelling Institut für Mineralogie and ZFM Center for Solid State Chemistry and New Materials Leibniz Universität Hannover, Callinstr. 3 D-30167 Hannover, Germany
[email protected]
INTRODUCTION Although large-scale transport in magmas is usually controlled by convection, diffusion of sulfur in silicate melts, nevertheless, plays a crucial role in the kinetics of various magmatic processes. The most important process is probably the degassing of magmas and melts initiated by an oversaturation of the melt with respect to dissolved volatiles. Such oversaturation can be caused in nature, i.e., by decompression when a magma ascends to the Earth surface. But the degassing of melts is also of major interest for technical applications, e.g., during fining of melts in industrial glass production. The kinetics of volcanic eruptions and the fining of glass melts are basically controlled by the dynamics of bubble nucleation and growth, and these properties are strongly affected by the diffusion of volatiles from the melt into bubbles (Nemec 1980a,b; Sparks et al. 1994; Müller-Simon 2011, this volume). Hence, understanding the diffusion of volatiles in melts provides a necessary tool for modeling bubble nucleation and growth. However, the diffusivities of volatiles such as H2O, CO2, SO2 and H2S in the melt do not only affect the degassing rate, they may also cause a fractionation of volatiles between gas phase and melt as well, due to different diffusivities. For instance, in hydrous magmas the diffusion of H2O is usually much faster than CO2 and sulfur diffusion (Baker et al. 2005; Zhang and Ni 2010), so rapid, disequilibrium removal of bubbles from the melt may cause an artificial enrichment of some components in the bubbles. Other processes which may be governed by sulfur diffusion in the melts are the dissolution and precipitation of minerals. In glass manufacturing, crucial problems are the kinetics of dissolution of the raw materials in the melt batch and of the homogenization of the melt (see Falcone et al. 2011, this volume). In nature, the formation of sulfur-bearing accessory minerals requires extraction of sulfur from a large melt volume, due to the low sulfur solubility in the melts (see Parat et al. 2011, this volume). Hence, rates of diffusion of sulfur towards or away from the minerals are important for the kinetics of crystal growth and crystal dissolution. But diffusion may affect isotope partitioning between melts and minerals as well if the mass of the isotopes and, hence, their diffusivities in the melt differ strongly. The mechanisms of sulfur diffusion in silicate melts are much more complex than those of other reactive volatiles such as H2O and CO2. Water and carbon dioxide both dissolve as unreacted species (molecules) and reacted species (hydroxyl groups and carbonate groups). For these two components the molecular species diffuses much faster than the reacted species, which often can be considered in a first approximation as immobile (Zhang and Ni 2010). Hence, diffusional transport is controlled by the fraction of molecular species relative to the reacted species and their mobilities in the melt. The dominant sulfur species in silicate melts, assuming that speciation in glasses represents the melt at least near the glass transition, are 1529-6466/11/0073-0004$05.00
DOI: 10.2138/rmg.2011.73.4
80
Behrens & Stelling
sulfide under reducing conditions and sulfate under oxidizing conditions (see Wilke et al. 2011, this volume). In both cases there is not a simple transfer to a molecular form. No molecular form exists for sulfide except possibly in hydrous melts where H2S could be formed, and sulfate can decompose to molecular SO2 only by a change in oxidation state (e.g., SO42−(m) ↔ SO2(g) + 0.5O2(g) + O2−(m), where m refers to melt and g to gas). Hence, the most likely process of diffusion of sulfur in the melt occurs via ionic sulfur species, and fluxes of other melt constituents, e.g. alkali ions, are required for local charge neutralization. As a consequence, the diffusion of sulfur in the melt will be associated with a large structural perturbation, i.e. local re-arrangement of the structure is needed because the diffusing unit is too large to fit into existing holes of the silicate network. Hence, one might expect a large influence of the kinetics of structural relaxation (i.e., viscosity) on sulfur diffusion. Sulfur diffusion studies on silicate glasses and melts are rare for natural melts as well as for melts of interest for technical application. Previous reviews on melts relevant to nature can be found in Watson (1994), Baker et al. (2005) and Zhang et al. (2007). In part, the lack of diffusion data is due to technical difficulties in doing well-controlled diffusion experiments. These problems are related to the volatile character of sulfur, to the speciation of sulfur which changes with redox conditions, and also to the overprinting of diffusive transport by convection in diffusion sample experiments. In this chapter, we will first give an overview of sulfur diffusion studies on silicate melts. Several of these studies show large variations of the diffusion coefficients at constant pressuretemperature (P-T) conditions, and an important question is whether other transport mechanisms such as convection may have affected the experiments. Next, we will compare the diffusion of sulfur with that of other volatiles such as H2O, CO2 and noble gases. For understanding the rate-controlling processes of sulfur diffusion in silicate melts, the effect of oxygen fugacity will be discussed, and findings on diffusion of individual sulfur species will be presented. Then we will have a look at diffusion-controlled redox processes in which sulfur is involved. This is interesting in view of the interaction of magmas with fluids which may induce changes in sulfur speciation (Webster and Botcharnikov 2011, this volume). Finally, considering that sulfur diffusion is usually very slow in silicate melts, one needs to address how strongly the mobility of sulfur in silicate melts is coupled to the dynamics of the silicate network, i.e., to melt viscosity.
SULFUR DIFFUSION STUDIES Experimental techniques to determine diffusion coefficients of ions, atoms, and molecules in silicate melts are described in Ryerson (1987), Watson and Dohmen (2010) and Behrens (2010). Information about analytical techniques to measure diffusion profiles can be found in Cherniak et al. (2010). In most of the sulfur diffusion studies, chemical diffusion coefficients were measured, meaning that a gradient in sulfur concentration was present in the experimental diffusion samples. In a few studies (Watson 1994; Szurman et al. 2007; Backnaes 2011), the radioactive isotope 35S was used to monitor sulfur diffusion and the samples were virtually free of chemical potential gradients. Conditions and results of the sulfur diffusion studies are listed in Appendix Table 1. Compositions of the studied melts can be found in the Appendix Table 2. The temperature dependence of diffusivity is usually expressed by an Arrhenius relationship −E D = Di ,o ⋅ exp a i RT
(1)
where i refers to the diffusing species. Di,o is the pre-exponential factor, Ea is the activation energy for diffusion, R is the universal gas constant and T is the absolute temperature. Figure 1 presents a compilation of sulfur diffusion data for melts with compositions relevant for industrial processes or for magmatic systems.
Diffusion & Redox Reactions of Sulfur in Silicate Melts
81
Figure 1. Compilation of sulfur diffusion data for (a) simple melts and melts of interest for glass sciences: B10 - NCS, NS3 (Backnaes 2010; Backnaes et al. 2011); B62 - NCS (Brückner 1962); E05 - SiO2 glass (Espiau de Lamaestre et al. 2005); K&N95 - NCS (Klouzek and Nemec 1995), N&M81 - float glass (Nemec and Mühlbauer 1981); Sch87 - alkali borosilicate SRL-131 (Schreiber et al. 1987); Sch89 - E-glass (Schreiber et al. 1989); S09 – NCS, NS3 (Stelling 2009); Sz07_ox – barium aluminoborosilicate AF45, oxidizing cond. (Szurman et al. 2007); Sz07_red - AF45, reducing conditions (Szurman et al. 2007) and (b) melts of interest for geosciences: B&R96 - rhyolite (Baker and Rutherford 1996); F05 - E bas = Etna basalt, S bas = Stromboli basalt (Freda et al. 2005); W94 - andesite & dacite (Watson 1994), W98 - albite (Winther et al. 1998). Lines are Arrhenius relationships listed in Appendix Table 1; compositions are given in Appendix Table 2.
82
Behrens & Stelling
In past, the terminology of diffusion coefficients has been ambiguous. In some older papers (Nemec and Mühlbauer 1981; Schreiber et al. 1987, 1989; Klouzek and Nemec 1995), the diffusing species was often not clearly identified and labels such as SO2 or SO3 were used for diffusion coefficients; these labels refer to the gas phase used in the experiments rather than to the mobile species in the silicate melt. In this review the term sulfur diffusion (S-diffusion) refers to experiments in which the diffusing sulfur species could not be clearly identified or in which different sulfur species coexisted. Diffusion coefficients for species such as sulfide or sulfate are given only when the data are based on measurements of profiles of discrete species, i.e., identified by using micro XANES (X-ray absorption near-edge structure) spectroscopy. The terms tracer diffusion and 35S-diffusion refer to experiments in which the radioactive isotope 35S was used. In the following, the studies are discussed in order of an increasing complexity of melt composition. Some recent work of one of the authors on sodium silicate and soda lime silicate melts is presented in more detail than other work, because this work has been so far published only as a dissertation (Stelling 2009).
Silica glass Espiau de Lamaestre et al. (2005) studied S-diffusion profiles in silica glass by secondary ion mass spectrometry (SIMS) after implanting sulfur close to the glass surface using a highly energetic particle beam and subsequent diffusion annealing at temperatures of 1073-1173 K in a silica tube furnace under dry N2. The obtained diffusivities between 2.5×10−18 m2/s and 1.6×10−17 m2/s are the slowest ones measured for sulfur diffusion in glasses to date (Fig. 1). These data may be compared to noble gas diffusion and are intermediate between diffusivities for krypton and xenon in silica glasses (Behrens 2010). The authors (Espiau de Lamaestre et al. 2005) assume that species equilibrium is achieved during experiment by the reaction S2− + 2O2 ↔ SO42−; however, sulfur speciation was not measured. For the implantation process one might expect different sulfur species than those found in glasses produced by quenching of silicate melts. Large local stress is induced in the glass around the implanted atoms and local relaxation of the silica network is very slow at the temperatures of the diffusion experiments which are far below the glass-transition temperatures of silica glasses (Agrarwal et al. 1995). Hence, the diffusion data may not be transferable to a relaxed, stress-free silica glass.
Simple silicate glasses Studies on S-diffusion in simple Al- and B-free silicate compositions are restricted to soda lime silicate and sodium trisilicate melts. The first systematic study on transport of sulfur in silicate melts was performed by Brückner (1961a,b, 1962) on soda lime silicate glass melts with a composition often used as a simplified window glass composition (16 Na2O - 10 CaO - 74 SiO2 in mol%, denoted hereafter as NCS). He studied interface reactions and material exchange between silicate melts and sulfur-bearing gases or molten salts. Brückner (1961b) tried to model the results of experiments in platinum crucibles involving uptake of sulfur by the NCS melt overlain by Na2SO4 melt. He assumed a simple diffusion-controlled process and found that the derived sulfur diffusivity of ~3×10−7 m2/s was relatively fast compared to sodium diffusion. Furthermore, the derived diffusivity varied very little with temperature in the range of 1373-1873 K. The author argued that convection had a major influence on the transport of sulfur in the melt, induced probably by reactions at the salt/silicate interface and by density fluctuations in the melt near the interface. Subsequently, Brückner (1962) performed two experiments in which a sulfate-bearing melt was placed above a sulfate-free melt in a platinum capsule, one was heated for 30 minutes and the other for 65 minutes at 1573 K. Sulfur profiles were measured using bulk analyses of glass sections parallel to the diffusion interface. The data suffer from the imprecision of the analytical technique available at that time, and also
Diffusion & Redox Reactions of Sulfur in Silicate Melts
83
from the degassing of sulfur to the outer atmosphere, which strongly affected the profiles in the upper part of the melt batch. Brückner (1962) argued that the measured values of D = 8×10−10 m2/s after 30 min and D = 2×10−10 m2/s after 65 min represent upper limits for the diffusion coefficient. Even higher apparent diffusivities were found in experiments when a SO2-O2 gas mixture in ratio 2:1 was blown over the glass melt. The author noted that the uptake of sulfur was accompanied by a loss of sodium. Sodium diffuses through the melt to the gas/melt interface and forms a layer of sodium sulfate liquid by reaction with sulfur dioxide and oxygen (Na2O + ½ O2 + SO2 = Na2SO4), but sodium is evaporated as well to the atmosphere under the chosen experimental conditions. The studies of Brückner (1961a,b, 1962) point out the need to avoid the contribution of convection to the overall transport of sulfur in melts and give the first example of the coupling of sulfur fluxes to fluxes of other melt components. Nemec and Mühlbauer (1981) reported a diffusion equation, which they assigned to SO3 diffusion (the term S-diffusion is more appropriate), in a float glass melt compositionally similar to the NCS melt. According to Nemec (pers. communication) the data are based on photographic measurement of the growth rate of ascending bubbles in the sulfate-bearing melt in the T-range of 1698-1773 K. Using gas chromatographic analysis of volatiles released from re-heated quenched glasses, it could be demonstrated that the content of the fining gases SO2 and O2 was about 98% in the bubbles. Diffusion coefficients of SO3 were derived by modeling the growth kinetics assuming a simple diffusion-controlled mechanism. An inverse approach was used by Klouzek and Nemec (1995) who measured the rate of dissolution of SO2 gas in a NCS glass melt. The gas was loaded via a tube in a silica container which was inserted upside down in the melt. Then the position of the gas/melt interface was recorded with a video camera, and diffusion data were obtained by fitting the change in the height of the gas column to the solution of Fick’s 2nd law for spherical coordinates to account for the curved gas/melt interface. In the case of SO2 and SO2-O2 gases, the experimental data show noticeable discrepancies to the trend expected by the theoretical equation implying that the dissolution process is not exclusively controlled by diffusion of sulfur species in the melt. The authors suggest that chemical reaction and convection have an effect as well on the absorption kinetics. Fitting only the initial period of the experiments, estimates of sulfur diffusivities in soda lime silicate melts were obtained which range from 9.0×10−12 m2/s at 1523 K to 5.1×10−11 m2/s at 1723 K. Using the same experimental approach, in a subsequent study Nemec and Klouzek (1998) found evidence for the formation of a sulfate layer at the interface between soda lime silicate glass melts and SO2 ± O2 in particular at lower temperatures (1473 K). This reaction includes extraction of Na2O and/or CaO from the melt and, hence, induces a flux of these components to the melt surface, as observed already by Brückner (1962). A model to describe the interaction between sulfate-bearing melts and SO2-bearing bubbles was proposed by Nemec and Ullrich (1998). In the PhD thesis of Stelling (2009) diffusion-couple experiments were used to study chemical diffusion of sulfur in NCS melts and sodium trisilicate (NS3) melts. In the experiments, sulfur-free and sulfur-bearing glass cylinders (diameter 4 mm, length 3-4 mm) with polished base planes in mutual contact were loaded in platinum capsules (Fig. 2). Sulfur contents between 0.10 and 0.25 wt% were adjusted in the sulfur-bearing halves of the couple; these sulfur contents are much lower than the expected sulfur solubility for the conditions of the diffusion experiments. Sulfur-bearing NS3 glasses were produced by melting mixtures of glass powder and appropriate amounts of Na2SO4, FeS or Na2S in platinum capsules in an internally heated gas pressure vessel (IHPV) at 1273 K, 100 MPa for ca. 15 h. The syntheses conditions were chosen to be similar to those used in the diffusion experiments, in order to prevent complications during the diffusion runs induced by disequilibrium of hydrogen between the pressure medium and the experimental charge (the capsule walls are permeable to hydrogen at temperature). However, XANES spectra (recorded on glass powders or glass sections at the SUL-X or XAS beam line at
84
Behrens & Stelling
Figure 2. Scheme of a diffusion couple sample, before (left) and after (right) the experiment.
ANKA, the synchrotron facility at the Karlsruhe Institute of Technology (KIT), Germany, or the ID21 beam line at the European Synchrotron Radiation Facility (ESRF), Grenoble, France) demonstrate that only sulfide remains stable in the nominally anhydrous melt while sulfate is partially reduced under these conditions in the IHPV (Stelling 2009). Therefore, a different synthesis strategy was chosen for NCS glasses in order to obtain samples containing exclusively sulfate or sulfide. To produce sulfate-bearing glasses, about 1 wt% sulfur as Na2SO4 was mixed with glass powder and melted in a platinum crucible at temperatures between 1640 K and 1920 K in air for 1.5 to 3 h. The remaining sulfur content in the glasses after this fining experiment was typically 0.10-0.15 wt%. Sulfide-bearing glasses were prepared by melting up to 7 wt% of sulfur as Na2S and FeS with glass powder in a graphite crucible for 60 to 90 min at 1700 to 1800 K. To protect the graphite crucible against oxidation in air, it was inserted in an alumina crucible and covered with a platinum lid. Sulfur distribution in both sulfate- and sulfide-bearing NCS glasses was homogeneous and glasses were free of salt-relicts over several millimeters as evidenced by analyses of glass fragments using a carbon/sulfur analyzer ELTRA CS800 or by investigation of thin sections of the glasses using a CAMECA SX-100 electron microprobe. Results of S K-edge XANES spectroscopy demonstrate that either sulfate or sulfide was the only sulfur species in the respective glasses. The NS3 glasses have typically higher water contents (0.1-0.6 wt%) than NCS glasses (<0.1 wt%) as measured by IR microspectroscopy on thin sections of starting glasses and diffusion samples (see Appendix Tables 3 and 4). This is a consequence of using different techniques for syntheses: the partial pressure of H2O was very low during melting of NCS in crucibles at ambient pressure. On the other hand, the platinum capsules used in the IHPV syntheses of NS3 glasses are closed to H2O loss, and thus H2O adsorbed on the starting glass powder remained in the final glass product. In addition, H2O can be generated inside the capsule through reduction of sulfate by hydrogen which diffuses from the pressure medium into the capsule (Backnaes et al. 2008). After sealing the capsules using arc-welding, diffusion experiments were carried out in an IHPV at a confining argon pressure of 100 MPa. Pressurizing has the advantage of fixing the Pt-tube tightly to the glass cylinders, and thus minimizing the risk of creeping of the melt at high temperature. On the other hand, considering that typical activation volumes for diffusion correspond to the size of the diffusing species, no significant change of diffusivity is expected for such a low pressure compared to ambient pressure.
Diffusion & Redox Reactions of Sulfur in Silicate Melts
85
Redox conditions were imposed by the intrinsic hydrogen fugacity of the vessel which is typically in the range of 0.02-0.10 MPa for the IHPVs used in Hannover (Schuessler et al. 2008). Under water-saturated conditions the corresponding oxygen fugacity is 2.6 to 3.5 orders of magnitude higher than the oxygen fugacity adjusted by the nickel/nickel oxide (NNO) buffer. Due to the low water contents of a few hundred ppm in weight (measured by IR spectroscopy) in the glass melts, the oxygen fugacity, which is controlled by the reaction H2 + ½ O2 = H2O, is below that of the NNO buffer. Under these conditions, sulfide is expected to be the stable sulfur species in the melts (see Wilke et al. 2011, this volume). However, in experiments with water-poor glasses containing sulfate, it could be demonstrated by S K-edge micro-XANES spectroscopy that only in a layer of ~100 mm thickness is sulfate reduced to sulfide within 8 h at 1373 K and 100 MPa in the sulfur diffusion experiments (Stelling 2009). The reduced rim is small compared to the radius of the cylinders, and in the cylinder cores, where the concentration profile is measured, the speciation is still unchanged as shown for some samples by XANES spectroscopy. Diffusion profiles were routinely measured by electron microprobe (CAMECA SX-100, acceleration voltage of 15 kV, beam current of 40 nA for sulfur, beam defocused to 20 mm, wavelength for sulfur analyses was always set to the peak maximum). An example is shown in Figure 3a. For selected samples, S K-edge XANES spectra were recorded along the diffusion profile to determine the speciation of sulfur after the experiments (see Fig. 3b). In order to analyze profiles of individual species, X-ray fluorescence scans were performed for specific energies at the SUL-X beamline at ANKA Karlsruhe or the ID21 beamline at ERSF Grenoble in steps of 10-50 mm using accumulation times of 2-10 s and focus diameters on the sample of 1-50 mm. Intensities were recorded at 2474.0 eV and 2477.0 eV (sulfide) and 2482.5 eV (sulfate). This measurement procedure allows monitoring changes in speciation with high local resolution without producing beam damage (see Wilke et al. 2011, this volume). Stelling (2009) showed that no change of the oxidation state occurs along the diffusion path of sulfate. As shown in Figure 3a, profiles recorded by XRF scans and by EMPA agreed very well. A crucial test as to whether diffusion experiments may be affected by melt convection is to measure diffusivities as a function of run duration. As shown in Figure 4 the apparent sulfur diffusivity in NCS melts at 1100 °C increases systematically with time for both sulfideand sulfate-bearing diffusion couples. Furthermore, in some samples the contact between the two halves of the diffusion couples appears to have been distorted from a plane; although this was often difficult to recognize by microscope. In other experiments, sulfur profiles deviate considerably from the trend predicted by the solution of Fick’s 2nd law for the given boundary conditions (initial concentration step at the interface, two semi-infinite media, see Crank (1975): x C − Ci ,min 1 − erf Ci = Ci ,min + i ,max 4D t 2 i
(2)
where Ci,min and Ci,max are the initial concentrations of the diffusing particle i in both halves of the diffusion couple, x is the distance to the inflection point of the profile, Di is the diffusion coefficient and t is the run duration. These observations indicate that convection can be a severe problem in sulfur diffusion experiments with silicate melts and small temperature gradients or chemical gradients which induce density gradients in the melt that can easily initiate convective fluxes. It is interesting to note that trace-element diffusion in sulfur-free NCS melts (Behrens and Haack 2007) was not noticeably affected by convection. This may imply that the sulfur gradient itself promotes local convective mixing in silicate melts. In Appendix Table 3 and Figure 5a,b a compilation of diffusion data is given for selected sulfur diffusion experiments with NCS and NS3 melts performed by Stelling (2009). The only data plotted are those not obviously affected by convection. Experiments that yielded anomalously high diffusion coefficients or that showed distorted diffusion couple interfaces or
86
Behrens & Stelling
S (wt %)
Figure 3. (a) Diffusion profile measured by XRF scans in comparison to the EMPA profile for soda lime silicate melt (#G28, 100 MPa, 1373 K, 8 h). Spatial resolution of measurement spots was 10 mm for EMPA and ~ 30 mm for the XRF scans, the intensity of the XRF energy was normalized to the beam current. Dashed lines represent fits by Equation (2). Data from Stelling (2009). (b) Selected S K-edge XANES spectra, recorded along the profile shown above, give evidence that sulfate is the dominant species in the diffusion sample. Spectra were normalized to the pre- and post-edge region of the Sulfur K edge, respectively. Features below 2480 eV in the spectrum recorded at the interface are due mainly to low sulfur content and, hence, more scatter in spectrum. Note that a more defocused beam of 100 mm diameter was used to collect the spectra in order to avoid beam damage. Spectra are plotted with an offset for clarity, redrawn after Stelling (2009).
Diffusion & Redox Reactions of Sulfur in Silicate Melts
87
Figure 4. Variation of apparent diffusivities for sulfur in soda lime silicate melts at 100 MPa and 1373 K with run duration. Increased diffusivity at long durations is attributed to contributions of local melt convection. Data from Stelling (2009).
anomalous profiles were excluded. However, the scatter of the data for NCS melt is obvious and even these data may present only upper limits for the diffusion coefficients. No significant difference of the diffusivity of sulfide and sulfate in NCS melt was found. Tentatively, the selected data (13 runs with S2−-bearing samples + 15 runs with S6+-bearing samples, see Appendix Table 3) were fitted by a single Arrhenius relationship kJ 215 ± 31 2 m mol 5.4 × 10 −5 exp − DS = s RT
(3)
which reproduces the experimental data in the T-range of 1273-1573 K with a 1s standard deviation of 0.35 log units. For NS3 melts containing a mixture of sulfate and sulfide, the lowest diffusion data at given temperatures were used to define an Arrhenius relationship in the range of 1273-1523 K: kJ 177 ± 23 2 m mol 7.7 × 10 −7 exp − DS = s RT
(4)
which reproduces the experimental data (7 runs) with a 1s standard deviation of 0.17 log units. It is noteworthy that the lowest diffusion coefficients were obtained for NS3 samples with the lowest water contents of 0.11-0.26 wt%, as measured by IR spectroscopy using the two-band method described in Behrens and Stuke (2003). The new data for NCS melts indicate that sulfur diffusion is significantly slower than previous estimates for sulfur diffusivity reported by Brückner (1961b), Nemec and Mühlbauer (1981) and Klouzek and Nemec (1995). This supports the hypothesis that the experiments in the earlier studies were not controlled by sulfur diffusion only (Fig. 5a). Further evidence for this hypothesis is given by the recent study of Backnaes et al. (2011) who performed 35S-tracer-
88
Behrens & Stelling
Figure 5. Arrhenius plot for sulfur diffusion in simple silicate melts at 100 MPa. (a) Soda lime silicate (NCS). (b) Sodium trisilicate glass (NS3). Data sources: S09 - NCS, NS3 (Stelling 2009); B10 - NCS, NS3 (Backnaes 2010; Backnaes et al. 2011); K&N95 - NCS (Klouzek and Nemec 1995); N&M81 - Float glass (Nemec and Mühlbauer 1981), B62 - NCS (Brückner 1962). EMPA or SR-XRF refers to profile measurements by electron microprobe and by m XANES spectroscopy, respectively. Lines represent fits by Equation (1), Arrhenius parameters are listed in Appendix Table 1. Eyring diffusivities were calculated for NCS and NS3 melts by Equation (13) assuming a jump distance of 0.3 nm using viscosity data of Bornhöft and Brückner (1999).
Diffusion & Redox Reactions of Sulfur in Silicate Melts
89
diffusion experiments with NCS melts and NS3 melts following the protocol of Szurman et al. (2007). Some NCS melts were nominally sulfur-free melts and other melts contained 1380 ± 45 ppm in weight S as sulfate. The initial experiments at ambient pressure failed because of pronounced advective fluxes in the diffusion samples, so subsequent experiments were performed in an IHPV at 100-MPa confining pressure. Experimental conditions are identical to those described above for the diffusion-couple experiments of Stelling (2009) carried out in the same lab. In the case of NCS melts, 35S-diffusivities for short-term runs with sulfur-free and sulfatebearing base melts agree within error, but experiments with longer durations systematically yielded slightly higher diffusivities for the sulfate-bearing melt, implying that the data might be affected to a certain extent by advective fluxes. As shown in Figure 5a, 35S-diffusion data for NCS melts from Backnaes et al. (2011) are 1-2 log units lower than the estimates from earlier studies (Brückner 1961b; Nemec and Mühlbauer 1981; Klouzek and Nemec 1995), but they are in good agreement with the diffusion-couple experiments presented in this review. Backnaes et al. (2011) found that 35S-diffusion coefficients in NS3 melts do not differ much from those in NCS melts in the range of 1273-1473 K (Fig. 5b). The data for NS3 melts are not as robust as those for NCS melts since only four successful experiments were performed, but a duplicate at 1273 K with longer duration (12 h compared to 6 h) yielded a slightly lower diffusivity (7.6×10−14 m2/s compared to 10.5×10−14 m2/s), implying that at least for this temperature, convection has not been a severe problem. At 1473 K, tracer-diffusion and chemical-diffusion data agree within error while at 1273 K the chemical-diffusion experiments yield diffusivities about half order of magnitude lower than the tracer-diffusion experiments. Whether the difference is due to differences of sulfur speciation or concentration, or it reflects experimental difficulties remains an open question, i.e., because the speciation is unknown for the tracer experiments with nominally sulfur-free melts.
Borosilicate glasses Borosilicate glass is widely used in households, laboratories and industrial plants because of its low coefficient of thermal expansion and because of its high chemical durability. Major components are usually SiO2, B2O3 and alkali oxides, but for specific applications aluminarich and alkali-poor compositions are chosen. Sulfur diffusion in such melts was studied in three papers. Schreiber et al. (1987, 1989) measured the uptake of SO2-bearing gasses by melt spheres, which were fixed in a metallic ring, as a function of time using bulk-chemical analyses of the glasses after quench. An alkali borosilicate melt was studied in the range of 1313-1463 K (Schreiber et al. 1987) and a commercial E-glass melt (basically an alkali-poor aluminoborosilicate, see Appendix Table 2) in the range of 1373-1773 K (Schreiber et al. 1989). 35
S-tracer diffusion in a technical alkali-poor barium aluminoborosilicate glass melt (AF45, Schott AG) was studied by Szurman et al. (2007) at temperatures 1173-1673 K and ambient pressure. To avoid convection during the experiments, a sandwich arrangement was chosen with the radioactive tracer 35S (b-emitter with a half-life of 87.4 days) placed as a dried droplet of sodium sulfate solution between two glass cylinders which were surrounded by a platinum ring. The whole assemblage was fixed in a platinum tube by squeezing the ends of the tube, and experiments were performed either under oxidizing conditions in air or under reducing conditions in a nitrogen flow which was passed over Ni/NiO powder. After the diffusion experiments, thin layers (≈10 mm) of glass were ground off successively, and the residual b activity was measured with a methane gas flow counter (using 10% CH4; 90% Ar). 35
S-tracer diffusion in the barium aluminoborosilicate melt (Szurman et al. 2007) was more than three orders of magnitude slower than chemical diffusion of sulfur in the alkali borosilicate melts (SRL-131) or the E-glass melts studied by Schreiber et al. (1987, 1989). Furthermore, the activation energy for sulfur diffusion in the barium aluminoborosilicate melt is much higher (~258 kJ/mol) than the activation energies for the alkali borosilicate (~45 kJ/
90
Behrens & Stelling
mol) and E-glass (~162 kJ/mol) estimated from the graphical representations in Schreiber et al. (1987, 1989) (Fig. 1a). These findings can not be fully explained by different melt viscosities, and it is suspected that the kinetics of SO2 uptake in the experiments of Schreiber et al. (1987, 1989) was not controlled by sulfur diffusion only.
Aluminosilicate melts relevant to magmatic systems The first measurements on sulfur diffusion in melts relevant for nature, including obsidian, dacite as well as synthetic Fe-free dacite, synthetic low-K andesite and a lunar ultramafic (very rich in Fe and Mg) composition, were performed by Watson et al. (1993) and are described in the review of Watson (1994). The compositions of the melts are listed in Appendix Table 3. Sulfur loss from melt droplets of the lunar composition containing about 0.3 wt% S was analyzed by electron microprobe after heating at 1573 K at near-ambient pressure under reducing conditions (0.7 log units below the iron/wustite buffer). Two experiments with different run durations (40-180 min) gave the same diffusion coefficient of 6.6×10−12 m2/s, implying that the transport of sulfur was controlled by diffusion and that convection had no significant effect. For other melt compositions, diffusion couples with sulfur-bearing and sulfur-free melts or sandwiches of sulfur-free cylinders with 35S-tracer in between were used in the temperature range of 1300-1500 °C. Experiments were carried out at 1 GPa in a piston cylinder apparatus using Mo capsules in most cases to fix the oxygen fugacity at reducing conditions. According to Watson (1994), the oxygen fugacity was near the iron/wustite buffer and, hence, in the stability field of sulfide. Profiles for sulfur after chemical diffusion experiments were measured by electron microprobe along the cylindrical axis while tracer experiments were analyzed by making b-tracks of the sectioned specimens and scanning these with a photodensitometer. Arrhenius relationships derived for dry dacite and andesite and hydrous andesite containing ~5.5 wt% H2O are listed in Appendix Table 1 and are compared to other melt compositions in Figure 1b. It is noteworthy that the equations are based only on two or three samples, and the derived activation energies should be considered as estimates. A systematic increase of sulfur diffusivity with water content was observed for andesite melts which can be attributed to the change in melt viscosity as discussed later. Baker and Rutherford (1996) determined diffusion rates for sulfur in rhyolite melt at temperatures of 1073-1373 K, pressures of 0.1-200 MPa, water contents of 0-7.3 wt%, and oxygen fugacities from the quartz-fayalite-magnetite (QFM) buffer to air. The experiments involved dissolution of anhydrite (under oxidizing conditions) or pyrrhotite (under reducing conditions) into rhyolite melt over time scales of hours to days. Sulfur-concentration profiles in the quenched glasses were measured by electron microprobe, and the diffusion of sulfur in the melt was strongly enhanced by dissolved water. Adding 7 wt% of H2O to the dry melt increases the sulfur diffusivity by 1.5 to 2 orders of magnitude; Baker and Rutherford only report the water content added to the capsules before the experiment. According to the empirical model of Liu et al. (2005), the water solubility in rhyolitic melts at 200 MPa is expected to vary between 5.6 wt% at 1273 K and 6.0 wt% at 1073 K. The scatter of the data is very large, i.e., by one order of magnitude at 1173 K for water-saturated melts at 200 MPa. Discrepancies in the data are particularly large for nominally dry melts, i.e. much lower diffusivities were measured in melts of experiments conducted in air versus those conducted in a cold seal pressure vessel (CSPV). A possible explanation may be differences in water content, because transport properties of “dry” rhyolitic melts dramatically change with addition of small amounts of H2O (Neuville et al. 1993; Dingwell et al. 1996; Schulze et al. 1996; Zhang et al. 2003). Assuming a negligible pressure effect, Baker and Rutherford (1996) combined their data at 200 MPa with data from Watson (1994) at 1 GPa to constrain Arrhenius relationships for dry and hydrous rhyolite melts, which are listed in Appendix Table 1 and shown in Figure 1b. More likely, the equation for hydrous melts reflects diffusivity for melts containing 5-6 wt% H2O, rather than 7 wt% H2O, as labeled by the authors.
Diffusion & Redox Reactions of Sulfur in Silicate Melts
91
Diffusion of sulfur in albitic (NaAlSi3O8) melts was studied by Winther et al. (1998) in the temperature range of 1573-1773 K using the diffusion-couple technique with one side of the couple being doped with 0.11 wt% sulfur as Na2SO4. Diffusion profiles were measured by electron microprobe, and various analytical and spectroscopic techniques were applied to identify sulfur species in the experimental glasses. The experiments were carried out in a piston cylinder apparatus at 1 GPa, and Fe2O3 was added to the sample assemblage to adjust the fO2 to oxidizing conditions. The sulfate ion was found to be very stable in the albite melt even at low fO2, and it was the dominant species in all glasses, as evidenced by the wavelength position of the S Ka peak in electron microprobe measurements. In the presence of graphite, the glass acquired a characteristic strong violet color, which was attributed to the conversion of some of the sulfate to S2− and S3− radical anions. In experiments at 1673 K, the effects of run duration, sulfur contents and addition of components such as graphite (as reducing agent) or NaNO3 (as oxidizing agent) were systematically investigated. The overall data set shows variations in diffusivity by more than one order of magnitude at this temperature. The authors attributed higher sulfur diffusivities in some of the runs to higher water contents of the melts. Although no IR spectroscopic data were given to support this hypothesis, the relative changes of diffusivity are consistent with the dramatic decrease in viscosity of stoichiometric NaAlSi3O8 melts when small amounts of H2O or Na2O are added (e.g., Whittington et al. 2004), the latter were possibly produced in the experiments of Winther et al. (1998) by dissociation of NaNO3. Through a comparison of total sulfur profiles with relative abundances of S2− and S3−, Winther et al. (1998) concluded that the radical anions S2− and S3− have higher mobility than sulfate ions in albitic melts and, hence, will govern bulk-sulfur diffusion when present. However, the question remains whether these sulfur radicals have, indeed, a general importance or are more a peculiarity of the conditions and the melt studied by Winther et al. (1998), i.e., no evidence for such species was found in other studies on sulfur speciation in silicate melts (see Wilke et al. 2011, this volume). The activation energy for sulfur diffusion in albite melts (458 ± 51 kJ/ mol) is much higher than that for other melt compositions (Appendix Table 1), and resembles the activation energy for viscous flow (402 kJ/mol) derived from the high-temperature viscosity data in Toplis et al. (1997) using an Arrhenius relationship for viscosity. Sulfur diffusion in basaltic melts with water concentrations of ~0 and ~3.5 wt% was studied by Freda et al. (2005) in the temperature range from 1498 K to 1723 K in a piston cylinder apparatus using the diffusion-couple technique. Reducing conditions in the stability field of sulfide were adjusted by inserting the diffusion couple into a graphite capsule. Concentration profiles of oxide components and sulfur were measured by electron microprobe, and apparent water contents were estimated along the profiles using the by-difference method (Devine et al. 1995). Duplicate runs with different durations (varied by a factor of 2) indicate that the reproducibility is within 0.3 log units for the diffusion data. Variation of pressure from 0.5 to 1 GPa had no significant effect on sulfur diffusivity in basaltic melts. Sulfur diffusion was slightly faster in alkali basalt from Etna compared to a high-K calc-alkaline basalt from Stromboli (see Appendix Table 1 for the Arrhenius parameters and Appendix Table 2 for the compositions). Combining the data sets for both melts, the authors’ proposed as general equations for sulfur diffusion in dry basaltic melt: 4.0 × 10 −4 DS =
m2 226.3 ± 58.3 kJ/mol exp − s RT
(5)
and in basaltic melts containing 3.5 wt% H2O 5.91 × 10 −7 DS =
m2 130.8 ± 82.6 kJ/mol exp − s RT
(6)
Although the water contents are not well established (i.e., the “dry” melt probably contains
92
Behrens & Stelling
several tenths to hundreds of ppm in weight H2O, or even more), it is clear that dissolved water has a much smaller effect on sulfur diffusion in basaltic melts than in rhyolitic melts. This is consistent with the larger effect of water on the viscosity in rhyolite melts. Using the data of Freda et al. (2005) and ignoring the differences between Etna and Stromboli basalts, Zhang et al. (2007) proposed an equation to predict sulfur diffusivity in basalt at 1498-1723 K and 0.5-1 GPa in the range of 0-4 wt % H2O: DS = 2.72 × 10 −4
m2 27692 − 651.6 × CW exp − s T
(7)
where CW is the H2O content in weight percent. This equation reproduces the experimental data within ±0.35 log units. The underlying idea of the equation is that sulfur diffusion varies exponentially with water content as observed for diffusion of other volatiles, i.e. molecular H2O, CO2 , and Ar (see Zhang and Ni 2010; Behrens 2010). Sulfur diffusion data for andesitic melts reported by Watson (1994) also are consistent with an exponential variation of sulfur diffusion with melt water content.
COMPARISON TO DIFFUSION OF OTHER VOLATILES In Figure 6a,b, the diffusion of sulfur in natural melts is compared to that of the most abundant volatiles H2O and CO2. In water-poor (~ 0.2 wt% H2O) basaltic melts, sulfur- diffusion is slower by more than one order of magnitude than bulk H2O diffusion at temperatures around 1500 K (Zhang and Stolper 1991; Freda et al. 2005). In dacitic and andesitic melts, the differences in diffusivities are of the same magnitude for these two volatiles for the same temperatures (Watson 1994; Behrens et al. 2004). On the other hand, in rhyolitic melts sulfur diffusion is slower than H2O diffusion by even 3 orders of magnitude at 1273 K and 2 orders of magnitude at 1773 K. Although these data are for different pressures (0.2 GPa for rhyolite, 1 GPa for the other melts), it appears that diffusive fractionation of volatiles upon decompression in silica-rich melts can be more pronounced because of slow sulfur diffusion. When bubbles nucleate and grow they are supplied with dissolved volatiles by diffusion in the melt. Faster diffusing species will fill the bubbles preferentially if the system does not come to equilibrium. CO2 diffusion does not vary much with anhydrous melt composition (Watson 1994; Nowak et al. 2004; Spickenbom et al. 2010; Zhang and Ni 2010). In water-poor rhyolitic to andesitic melts CO2 diffusivity is intermediate between those of sulfur and H2O, while in basaltic melts S and CO2 diffusivities are very similar. That might be due to the dominance of carbonate as the C species in basaltic melts while the more polymerized melts contain higher fractions of molecular CO2 (Nowak et al. 2004). Carbonate as an anionic species may have similar mobility to sulfate and sulfide because diffusion of all these species are strongly affected by the melt viscosity. It is interesting to compare the diffusion of sulfur with those of noble gases. Noble gases interact with the silicate structure through van der Waals forces only, and the diffusivities are mainly controlled by the size of noble gases, which determines the strain required for jumps
Figure 6 (on facing page). Comparison of diffusion of sulfur with other volatiles. All data are for a pressure of 1 GPa except for rhyolite, for which data refer to a pressure of 0.2 GPa. (a) Dry melts. Sulfur diffusion: basalt (F05, Freda et al. 2005), andesite, dacite (W94, Watson 1994), rhyolite (B&R96, Baker and Rutherford 1996); H2O diffusion at 0.2 wt% H2O: basalt (Z&S91, Zhang and Stolper 1991), rhyolite (N&Z08, Ni and Zhang 2008), andesite, dacite (B04, Behrens et al. 2004); CO2 diffusion: nat. melts (N&Z10, Ni and Zhang 2010); Rn diffusion: andesite (G99, Gauthier et al. 1999). (b) Hydrous melts. Sulfur diffusion: basalt (3.5 wt% H2O, F05), andesite (5.5 wt% H2O, W94), rhyolite (5.8 wt% H2O, B&R96); H2O: andesite (5.5 wt% H2O, B04), rhyolite (5.8 wt% H2O, N&Z08); CO2: rhyolite (8 wt% H2O, W94).
Diffusion & Redox Reactions of Sulfur in Silicate Melts
93
94
Behrens & Stelling
of the atoms from one cavity in the structure to an adjacent one. Argon diffusivity is similar to molecular CO2 diffusivity at least in silicic melts (Behrens 2010) in which the molecular form of CO2 is predominant. Considering the CO2 diffusion data shown in Figure 7a, it is evident that argon diffusion is much faster than sulfur diffusion in silicic melts. Even radon diffusion appears to be faster than sulfur diffusion in andesite melts (Fig. 6a). However, in this comparison one needs to consider the different experimental pressures, i.e. radon diffusion was measured at ambient pressure (Gauthier et al. 1999, 2000) while sulfur diffusion was measured at 1 GPa (Watson 1994). A more comprehensive comparison between noble gas diffusion and sulfur diffusion is possible for melts with albitic compositions since the whole range of noble gases from He to Xe has been studied (Shelby and Eagan 1976; Roselieb et al. 1992, 1995; Carroll 1991; Carroll et al. 1993; Spickenbom et al. 2010). A clear trend of decreasing diffusivity with increasing size of the noble gases has been established, but diffusion of all of these elements is much faster than sulfur diffusion measured by Winther et al. (1998). This again provides evidence that sulfur belongs to the group of slowest diffusing components in silicate melts. The addition of water to silicate melts strongly enhances the diffusion of all volatiles in the melts (Fig. 6a,b). For hydrous rhyolitic and andesitic melts, the difference in diffusivity between sulfur and the other volatiles is larger than in water-poor melts. In rhyolitic melts containing 6-8 wt% of H2O, both CO2 and H2O are faster by about three log units than sulfur diffusion (Watson 1994; Baker and Rutherford 1996; Ni and Zhang 2008), and in andesitic melts with 5.5 wt% dissolved water the difference in diffusion rates between sulfur and H2O is
Figure 7. Comparison of 35S-tracer diffusion data for barium aluminoborosilicate, AF45 under reducing and oxidizing conditions compared to Eyring diffusivity (data from Szurman 2005 and Szurman et al. 2007). For details about redox conditions see text.
Diffusion & Redox Reactions of Sulfur in Silicate Melts
95
two orders of magnitude (Watson 1994; Behrens et al. 2004). Considering the relatively weak effect of water on sulfur diffusion in basaltic melts (Fig. 1b), no large differences in volatile diffusivity are expected for hydrous basaltic melts at magmatic temperatures.
EFFECT OF REDOX STATE ON SULFUR DIFFUSION The effect of sulfur speciation on sulfur diffusion has been debated for a long time. From a geometrical point of view, one might expect that sulfide can diffuse faster than sulfate, simply because of the smaller ionic radius. But both sulfide and sulfate are anionic species, and one needs to take into account that cations are required for local charge compensation. Hence, the effective size of the diffusing species may be much larger than the size of the anion itself. The question, however, is whether this difference has any influence on the specific anion mobility. In contrast to noble gases, sulfide and sulfate interact with their chemical environment, which means that movement of sulfur species will not be a simple jumping from one cavity in the silicate network to another one. The movement of the sulfur ions surrounded by cations requires breaking and re-forming of multiple T-O bonds where T stands for the tetrahedrally coordinated cations such as Si and Al. Hence, the mobility of sulfur species is strongly determined by the dynamic character of the melt and local relaxations around the anions during movement through the melt. An enhancement of sulfur diffusion in the melt is possible if a rapidly moving sulfur species acts as a transporter for sulfur. Polysulfide radical ions might be such vehicles under certain reducing conditions, as suggested by Winther et al. (1998), but such conditions are not common in nature or in industry plants for glass manufacturing. An idea of the importance of sulfur speciation for sulfur diffusion can be obtained through studies in which experiments were performed at various oxygen fugacities or in which profiles of sulfur species were directly measured using, for instance, m-XANES. Most of the diffusion studies in glass sciences were performed under oxidizing conditions, i.e. in air. Szurman et al. (2007) also carried out 35S-tracer diffusion under reducing conditions which were established by passing a nitrogen flux over a mixture of nickel and nickel oxide. According to the authors, the oxygen partial pressure is about 10−5 bar under these conditions, which is still in the stability field of sulfate identified by spectroscopic data (see Fig. 3 Wilke et al. 2011, this volume). Hence, the enhancement of sulfur diffusion by about 0.3 log units under reducing conditions can not be clearly assigned to a change in speciation. Furthermore, it has to be noted that the plotted data in Szurman et al. (2007) represent only average values for the given temperatures and the data reported by Szurman (2005) show large variation by up to one order of magnitude at constant temperature (Fig. 7). Thus, the redox effect is within experimental uncertainty. Sulfur diffusion studies relevant to the Earth sciences have been conducted under either reducing or oxidizing conditions. Exceptions are the work of Baker and Rutherford (1996) on rhyolitic melts where various solid oxygen buffers were used to adjust the oxygen fugacity and the study of Winther et al. (1998) where different additives to the melt were used to vary redox conditions. No significant effect of oxygen fugacity on sulfur diffusion in rhyolitic melts was found by Baker and Rutherford (1996) in experiments controlled by the MnO/Mn3O4 (MNO) buffer and the QFM buffer. Sulfur speciation was not measured, but it can be expected that sulfate was the dominant species at the MNO buffer while sulfide is the stable species at the QFM buffer. Baker and Rutherford (1996) argued that the diffusion data represent diffusion of a single species, most likely sulfide. However, another explanation might be that the transport properties of the molten medium, i.e., the melt viscosity, controls the diffusion, and the properties of the different sulfur species have little effect. The experiments of Winther et al. (1998) on albitic melts have been already discussed in a previous section. Numerous interesting findings were made, but the interpretation of the diffusion data is not simple. It cannot be excluded that some of the variations in diffusivity
Behrens & Stelling
96
originates from different (unmeasured) water contents of the samples which have a strong effect on the melt viscosity in particular for albitic composition. Diffusion data for sulfide and sulfate in soda lime silicate melts show no significant difference (Fig. 5a). No species other than sulfide and sulfate were observed in XANES spectra of post-experimental glasses (see Fig. 3b for an example of a sulfate-bearing glass). However, it cannot be excluded that such species are present at the percent level. Hence, an unambiguous proof for the absence of a rapidly diffusing sulfur species cannot be given by spectroscopic observations. Chemical transport of sulfur through the melt requires the movement of neutral units such as Na2SO4 or Na2S which can be only achieved by coupling the fluxes of different particles (atoms, ions or molecules). On the other hand, tracer diffusion, i.e., the exchange of one sulfur isotope by another one, requires no coupling of particle fluxes and could be achieved by isotope exchange with a rapid sub-species. The good agreement of chemical diffusion data and tracer diffusion data for NCS and NS3 melts implies that such mobile species do not affect sulfur diffusion at least in soda lime silicate melts.
SULFIDE/SULFATE INTERDIFFUSION AND REDOX REACTIONS OF SULFUR It is an interesting question how the redox state of sulfur in a melt can adapt to changes in oxygen fugacity imposed, for instance, by an adjacent melt or fluid. Such a scenario is important in particular for magma mixing and magma/fluid interaction but also for fining of industrial glass melts. Natural melts often contain high concentration of iron, and the redox equilibrium of iron which can be expressed by: 2 FeO(m) + 0.5 O2 ( m) ↔ Fe 2O3 ( m)
(8)
where m refers to the melt phase. This reaction will strongly interfere with the redox equilibrium of sulfur which may be described by: SO24 − ( m) ↔ S2 − ( m) + 2 O2 ( m)
(9)
Redox reactions in dry melts In dry iron-free melts, a local change in redox state can be achieved either by diffusion of sulfide and/or sulfate, or by transport of oxygen through the melt. In the latter case, the concentration and the diffusivity of dissolved molecular oxygen in the melt are expected to be controlling parameters. To shed light on the mechanisms of this process, Stelling (2009) performed diffusion-couple experiments with NCS melts doped on one side with sodium sulfide or pyrrhotite and on the other with sodium sulfate (Appendix Table 4). The concentration of sulfate and sulfide, respectively, in the parts of the diffusion couple was about 0.15 wt%, with almost constant concentrations in the melt. Formally, this setup may be considered as an interdiffusion experiment in which a diffusion flux of one divalent anion (S2−) is compensated by a counter flux of another divalent anion (SO42−) while the molten matrix remains virtually unchanged. Experiments were performed in an IHPV at temperatures of 1373 and 1473 K and 100 MPa for 1 to 8 h. Some of the starting glasses contained traces of dissolved iron when FeS was used as a dopant, others were iron-free (Na2S as dopant). Concentration profiles of both species were measured after the experiment using X-ray fluorescence (XRF) scans as described in a previous section. An example of the measured profiles is given in Figure 8. The profiles of both species are highly symmetric and well fitted by Equation (2). Experimental conditions and the obtained diffusion coefficients for sulfide (DS2−) and sulfate (DSO42−) are listed in Appendix Table 4. In each experiment, the diffusion coefficients of both species agree within error (maximum difference of 0.25 log units) which is consistent with an interdiffusion mechanism. However,
97
Normalized intensity at 2477.0 eV (x10-3)
Diffusion & Redox Reactions of Sulfur in Silicate Melts
Figure 8. Example of a sulfide/sulfate interdiffusion experiment (G14, 100 MPa, 1473 K, 4 h, 30 min) with soda lime silicate melt. Dashed curves are fits by Equation (2). The intensity of the XRF energy was normalized to the beam current. Redrawn after Stelling (2009).
the whole data set at constant temperature shows variations by 0.4 log units at 1373 K (2 runs) and 1.2 log units at 1473 K (6 runs) (Fig. 9). These variations are, in part, due to analytical problems resulting from the unexpected shortness of the profiles but may be affected as well by contributions of non-diffusive transport. Despite these experimental and analytical problems, it is clear that the sulfide/sulfate interdiffusion experiments yield diffusion coefficients which are much smaller than those measured in diffusion couples only containing one sulfur species and measured in 35S tracer experiments. The symmetry of the profiles and the identical diffusion coefficients for sulfide and sulfate found in the interdiffusion experiments give further evidence that the diffusion coefficients for sulfide and sulfate do not differ much in silicate melts. Moreover, it can be excluded that rapid diffusion of an oxygen species is responsible for local transformation of sulfide into sulfate and vice versa. A possible explanation for the slow sulfide/sulfate interdiffusion is given by the theory developed by Darken (1948) to describe interdiffusion in binary alloys. In an alloy composed by two metals A and B, the interdiffusion coefficient D can be expressed by: = D
( x A ⋅ DA + xB ⋅ DB ) ⋅ f
(10)
where xi and Di are the mole fraction and the self-diffusion coefficient of the respective metal, and f is the thermodynamic factor, which is calculated as f = ∂lnaA/∂lnxA where aA is the activity of component A. As a consequence of the Gibbs-Duhem relation, there is only one thermodynamic factor for a binary alloy (∂lnaA/∂lnxA = ∂lnaB/∂lnxB). As discussed by Mehrer (2007), the thermodynamic factor can be larger or smaller than unity, depending on the change of the Gibbs free energy upon mixing the two components. It is clear that this concept cannot be directly applied to the interdiffusion of sulfide and sulfate in silicate melts, i.e., because the diffusing species are minor components only, embedded in a dynamic silicate network. The fluxes of sulfide and sulfide are not necessarily coupled, but diffusion of mobile ions (i.e., alkalis) and relaxation of the silicate network can serve for local charge balance. Nevertheless, the Darken concept demonstrates that variations
Behrens & Stelling
98
Figure 9. Arrhenius plot for sulfide/sulfate interdiffusion at 100 MPa in soda lime silicate melts (data from Stelling 2009), compared to chemical sulfur diffusion [S09] and 35S tracer diffusion [B10]. The dashed line represents the Eyring diffusivity calculated using data of Bornhöft and Brückner (1999).
of the chemical potentials of the diffusing components along a diffusion profile can strongly affect the interdiffusion rate. According to Equation (9), the local oxygen fugacity changes as a function of the sulfide/sulfate ratio. Our hypothesis is that this variation in oxygen fugacity controls the interdiffusion rates of both sulfur species, but further research is required to approve this hypothesis. An interesting possibility is the combination of the 35S-tracer with a chemical diffusion couple. This is basically the same type of experiment as described above, only that one half of the couple (either the sulfide-bearing or the sulfate-bearing half) is doped with 35S to monitor the self diffusion of sulfur. Such an experiment can clarify whether self diffusion of sulfur is decoupled from sulfide/sulfate interdiffusion.
Redox reactions in hydrous melts In hydrous iron-free melts, an additional possibility for generation of oxygen or removal of oxygen is given by: H 2O(m) H 2 (m) + 0.5 O2 (m)
(11)
and this implies that water diffusing into a melt may initiate oxidation of sulfide while hydrogen diffusing into a melt may initiate reduction of sulfate. The solubility and the diffusivity of water in silicate melts are very high and, hence, one might expect that H2O diffusing into an initially dry sulfide-bearing melt may initiate an oxidation of sulfide to sulfate. This process was experimentally studied by Stelling et al. (2011) in NCS and NS3 melts using two different setups. In type (i) experiment, a dry sulfide-bearing NCS glass cylinder was loaded with liquid water into a platinum capsule so that the melt was water-saturated at the contact with the fluid. In type (ii) experiments, a water-undersaturated sulfur-free glass (~ 3.0 wt% H2O) was placed
Diffusion & Redox Reactions of Sulfur in Silicate Melts
99
in contact with a dry sulfide-bearing glass (0.15-0.20 wt% S, added as FeS or Na2S before synthesis) to form a diffusion couple Hence, according to Equation (11) the conditions are more oxidized in the former case. Experiments were run at 200 MPa (i) or 100 MPa (ii) in IHPVs at 1273 to 1523 K for 4 to 20 min. Run duration was corrected for heating and cooling (Stelling et al. 2011) with an estimated uncertainty of ±1 min. Profiling by electron microprobe (sulfur) and infrared microscopy (H2O) demonstrated that H2O diffusion in the melts is faster by 1.5-2.3 orders of magnitude than sulfur diffusion and, hence, H2O can be considered as a rapidly diffusing oxidant while sulfur behaves in a quasi-immobile fashion in these experiments (Fig. 10a). In Raman spectra, a band at 2576 cm−1 appears in the sulfide-H2O transition zone which is attributed to fundamental S-H stretching vibrations (Klimm and Botcharnikov 2010). The formation of new IR absorption bands at 5025 cm−1 (at the expense of the combination band of molecular H2O at 5225 cm−1) and at 3400 cm−1 was observed at the front of the in-diffusing water in the sulfide-bearing melt. The appearance and intensity of these two IR bands are correlated with systematic changes in S K-edge XANES spectra. A pre-edge excitation at 2466.5 eV occurs with increasing H2O concentration while the sulfide peak at 2474.0 eV decreases in intensity relative to the peak at 2477.0 eV, and a feature grows at 2472.3 eV (Fig 10b). The observations by Raman, IR and XANES spectroscopy indicate a well-coordinated S2−-H2O complex which was probably formed in the glasses during cooling at the glass transition. No oxidation of sulfide was observed in any of the diffusion-couple experiments (type (ii) experiments) while XANES spectra of the experiments with a free H2O fluid (type (i) experiments) show almost complete transformation of sulfide to sulfate near the melt surface and coexistence of sulfate in sulfide in the center of the melt. This can be explained by a lower H2O activity in the diffusion-couple experiments or by the need of a sink for hydrogen (e.g., a fluid which can dissolve high concentration of hydrogen) to promote oxidation of sulfide by H2O via the reaction S2− + 4 H2O = SO42− + 4 H2. Sulfite could not be detected in any of the XANES spectra implying that this species, if it exists in the melt, is subordinate or transient only. The observation of H-S species in the glasses may raise the question whether such a species could be a transport vehicle for sulfur, similar to water molecules for hydrogen and oxygen transport in melts (Behrens et al. 2007). It is worth noting that H-S species were observed only in iron-free melts (Klimm and Botcharnikov 2010; Stelling et al. 2011), i.e., if iron is present the formation of Fe-S bonds is preferred to H-S bonds (note also that the sulfide capacity of melts increases strongly with Fe-content (see Baker and Moretti 2011, this volume). That means for natural (iron-bearing) melts, H-S species have much lower stability than H2O molecules and are not likely candidates as fast sulfur transporters. In (iron-poor) melts of technical interest, H-S species could play a role for sulfur transport; however, the water contents are usually too small under technical glass production conditions (typically <0.1 wt%) to expect high concentrations of H-S species.
SULFUR DIFFUSION VERSUS VISCOSITY As pointed out in previous sections, there are several pieces of evidence that sulfur diffusion is strongly coupled to the melt viscosity, i.e., the large variation of DS with melt composition shown in Figure 1 mainly originates from variations in melt viscosity. Two empirical equations have been proposed to interconnect diffusivity and viscosity, i.e., the Stokes law and the Eyring relationship. Both approaches postulate a reciprocal dependence between diffusivity and viscosity and differ only in the adjustable parameter. Stokes law was derived for molecular liquids. The characteristic parameter to link the viscosity h and the diffusivity Dh is the radius r of the diffusing molecule:
100
Behrens & Stelling
S (wt %)
Figure 10. Effects of H2O diffusion into sulfide-bearing NCS melts (200 MPa, 1323 K, 20 min). (a) Comparison of diffusion profiles measured by IR spectroscopy (H2O) and EMPA (S) (b) S K edges XANES spectra at various distances to the initial contact plane of the diffusion couple. Spectra were normalized to the pre- and post-edge region of the S K edge, respectively. See text for details. Redrawn after Stelling et al. (2011).
Diffusion & Redox Reactions of Sulfur in Silicate Melts Dh =
kT 6 πhr
101 (12)
where k is the Boltzmann constant. In the Eyring relationship, it is assumed that the same transition state is passed during viscous flow and diffusion (Chakraborty 1995). The characteristic value to relate the transport properties is the jumping distance l: Dh =
kT lh
(13)
The Eyring relationship was found to give a good estimate for the diffusivity of network formers (e.g., Si4+, Al3+, in dry melts also oxygen) and high-field strength elements (i.e., Zr4+) in silicate melts above the liquidus (Chakraborty 1995; Behrens and Hahn 2009). Comparison of experimental diffusion and viscosity data yields l values close to 0.3 nm which roughly corresponds to the size of SiO4-tetrahedra (Chakraborty 1995; Koepke and Behrens 2001; Tinker and Lesher 2001; Mungall 2002; Reid et al. 2003; Behrens and Haack 2007; Behrens and Hahn 2009). On the other hand, the radius calculated by Equation (13) is in the order of 0.02 nm and has no physical meaning. To test the relationship between melt viscosity and sulfur diffusion, we have plotted both variables against each other in Figure 11. Viscosities for nominally dry melts were calculated for the experimental temperatures using the Vogel-Fulcher-Tamman (VFT) equation (Fulcher 1925; Tamman and Hesse 1926) log h= A +
B T − T0
(14)
with parameters A, B and T0 listed in Appendix Table 5. In this equation T is in °C, as usual for VFT equations. For hydrous rhyolite, andesite and basalt equations from the literature were used to calculate the viscosity as a function of water content and temperature (Zhang et al. 2003; Vetere et al. 2008; Misiti et al. 2009). In the comparison of the data, one has to be aware that viscosity and sulfur diffusion were not always determined for exactly the same melt compositions. The following trends are visible in Figure 11. In general, diffusion data are close to the Eyring diffusivities (± one log unit) for viscosities smaller than 104 Pa·s, but at higher viscosities the sulfur diffusivity progressively departs from the Eyring diffusivity towards higher values. This trend is also visible within individual data sets if a sufficiently large T range was experimentally covered, e.g., for the barium aluminoborosilicate AF45 (Szurman et al. 2007) or for hydrous rhyolite (Watson 1994; Baker and Rutherford 1996). A smaller slope of log DS compared to log Dh in function of log h is evident for basaltic melts (only the Stromboli basalt is plotted here from Freda et al. (2005) since viscosity data were measured for exactly that melt composition by Misiti et al. (2009)) and andesitic melts (Watson 1994), in agreement with progressive decoupling of sulfur diffusivity from melt relaxation with lower temperature (high viscosity). There are some data sets which do not fit this trend well. As mentioned above, sulfur diffusivities reported by Brückner (1962) suffer from experimental and analytical problems and have to be considered as maximum values only. Data from Schreiber et al. (1987, 1989) for boron-bearing melts based on gas sorption in melt spheres are much higher than expected by viscosity. An uncertainty here is that the exact composition of the E-glass studied by Schreiber et al. (1989) is not known and the alkali borosilicate of Schreiber et al. (1987) has a slightly different composition than the melt used for viscosity determination. Thus, some of the discrepancies may originate from non-matching melt viscosities. However, data from Nemec
102
Behrens & Stelling
Figure 11. Comparison of sulfur diffusion with melt viscosities. Empty Circles – technical compositions, Filled squares – natural compositions.
and Mühlbauer (1981) and Klouzek and Nemec (1995) based on sulfur exchange between a gas phase and silicate melts are also one order of magnitude higher than the Eyring diffusivity. Hence, the kinetics of exchange of sulfur components between gases and silicate melts might be affected by other processes, i.e., the interface reaction (Brückner 1962; Nemec 1980a,b: Nemec and Ullrich 1998) and/or micro-convection near the melt surface.
SUMMARY AND OUTLOOK A variation of sulfur diffusivity with oxygen fugacity is not evident from experimental studies, and no significant difference was found for diffusivity of sulfide and sulfate in silicate melts. However, experimental difficulties (i.e., to avoid any convection) limit the accuracy of the diffusion data and a difference in diffusivity of both species up to a factor of 2 can not be ruled out. The experimental data indicate that diffusion of sulfur in silicate melts is strongly coupled to melt viscosity and that the Eyring relationship can be used for an estimation of sulfur diffusivity for melts with viscosity <104 Pa·s, when viscosity data are available. At higher viscosity, the Eyring relationship will systematically underestimate sulfur diffusivity. In natural melts with rhyolitic to basaltic compositions, sulfur diffusion is one or more orders of magnitude slower than H2O diffusion which may induce a kinetic fractionation during melt degassing when fluid and melt are rapidly separated. Considering the coupling of sulfur diffusion to melt viscosity, significant isotopic fractionation of sulfur by diffusion is not expected.
Diffusion & Redox Reactions of Sulfur in Silicate Melts
103
An open question remains whether certain conditions promote the existence of a rapidly moving sulfur species in silicate melts which may considerably enhance transport and isotopic exchange of sulfur. Strong bonding of sulfide to ferrous iron under reducing conditions and the stability of the large sulfate-cation association under oxidizing conditions will limit such a possibility, i.e., in iron-bearing melts. H2S might be a vehicle for sulfur transport in hydrous iron-free melts. However, this has probably little application for silicate melts in nature and industrial glass formation. Major tasks for future work will be the study of the kinetics of melt degassing and its relationship to the diffusion of the involved volatiles.
ACKNOWLEDGMENTS Part of the research was funded by the German Science Foundation (DFG). We thank Minoru Tomozawa, Bruce Watson, Sumit Chakraborty and Jim Webster for their constructive reviews.
RERERENCES Agarwal A, Davis KM, Tomozawa M (1995) A simple IR spectroscopic method for determining fictive temperature of silica glasses. J Non-Cryst Solids 185:191-198 Backnaes L (2010) Coupled processes in sulphur-bearing silicate melts - a study of speciation, diffusion and viscosity. PhD dissertation, TU Clausthal, Germany, 165 pp Backnaes L, Deubener J, Behrens H, Stelling J, Cichy SB, Bartels A (2011) Diffusion of the 35S isotope in sodalime-silica and sodium trisilicate glass melts. J Non-Cryst Solids, doi:10.1016/j.jnoncrysol.2011.03.037, in press Backnaes L, Stelling J, Behrens H, Goettlicher J, Mangold S, Verheijen O, Beerkens R, Deubener J (2008) Dissolution mechanisms of sulphur in silicate melts - implications from XANES sulphur K-edge studies on glasses. J Am Ceram Soc 91:721-727 Baker DR, Freda C, Brooker RA, Scarlato P (2005) Volatile diffusion in silicate melts and its effects on melt inclusions. Ann Geophys 48:699-717 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Baker LL, Rutherford MJ (1996) Sulfur diffusion in rhyolite melts. Contrib Mineral Petrol 123:335-344 Behrens H (2010) Noble gas diffusion in silicate glasses and melts. Rev Mineral Geochim 72:227-267 Behrens H, Haack M (2007) Cation diffusion in soda-lime-silica glass melts. J Non-Cryst Solids 353:4743-4752 Behrens H, Hahn M (2009) Trace element diffusion and viscous flow in potassium-rich trachytic and phonolitic melts. Chem Geol 259:63-77 Behrens H, Stuke A (2003) Quantification of H2O contents in silicate glasses using IR spectroscopy - a calibration based on hydrous glasses analyzed by Karl-Fischer titration. Glass Sci Tech 76:176-189 Behrens H, Zhang Y, Leschik M, Wiedenbeck M, Heide G, Frischat GH (2007) Molecular H2O as carrier for oxygen diffusion in hydrous silicate melts. Earth Planet Sci Lett 254:69-76 Behrens H, Zhang Y, Xu ZG (2004) H2O diffusion in dacitic and andesitic melts. Geochim Cosmochim Acta 68:5139-5150 Bornhöft H, Brückner R (1999) Elastic and inelastic properties of soda lime silicate glass melts. Glass Sci Tech 72:315-328 Brückner R (1961a) Zur Kinetik des Stoffaustausches an den Grenzflächen zwischen Silikatglas und Salzschmelzen und des Stofftransportes in Silikatglasschmelzen unter besondere Berücksichtigung des Verhaltens von Na2SO4 und seinen Zersetzungsprodukten. Teil I. Grenzflächenenergetische Ausgleichsprozesse bei Stoffaustauschvorgänge. Glastechn Ber. Glass Sci Tech 34:438-56 Brückner R (1961b) Zur Kinetik des Stoffaustausches an den Grenzflächen zwischen Silikatglas und Salzschmelzen und des Stofftransportes in Silikatglasschmelzen unter besondere Berücksichtigung des Verhaltens von Na2SO4 und seinen Zersetzungsprodukten. Teil II. Der Substanzaustausch zwischen Sililcatglas- und Salzschmelzen. Glastechn Ber Glass Sci Tech 34:515-28 Brückner R (1962) Zur Kinetik des Stoffaustausches an den Grenzflächen zwischen Silikatglas- und Salzschmelzen und des Stofftransportes in Silikatglasschmelzen unter besonderer Berücksichtigung des Verhaltens von Na2SO4 und seinen Zersetzungsprodukten. Teil III. Glastechn Ber Glass Sci Tech 35:93105 Carroll MR (1991) Diffusion of Ar in rhyolite, orthoclase and albite composition glass. Earth Planet Sci Lett 103:156-168
104
Behrens & Stelling
Carroll MR, Sutton SR, Rivers ML, Woolum D (1993) An experimental study of krypton diffusion and solubility in silicic glasses. Chem Geol 109:9-28 Chakraborty S (1995) Diffusion in silicate melts. Rev Mineral 32:411-503 Cherniak DJ, Hervig R, Koepke J, Zhang Y, Zhao D (2010) Analytical methods in diffusion studies. Rev Mineral Geochem 72:107-170 Crank J (1975) The Mathematics of Diffusion. Oxford University Press, 2nd ed Darken LS (1948) Diffusion, mobility and their interrelation through free energy in binary metallic systems. Trans AIME 175:184-201 Devine JD, Gardner JE, Brack HP, Layne GD, Rutherford MJ (1995) Comparison of microanalytical methods for estimating H2O content in silicic glasses. Am Mineral 80:319-328 Dingwell DB, Romano C, Hess KU (1996) The effect of water on the viscosity of a haplogranitic melt under P-T-X conditions relevant to silicic volcanism. Contrib Mineral Petrol 124:19-28 Espiau de Lamaestre R, Jomard F, Majimel J, Bernas H (2005) Diffusion properties of chalcogens (S, Se, Te) into pure silica. J Non-Cryst Solids 351:3031-3036 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Freda C, Baker DR, Scarlato P (2005) Sulfur diffusion in basaltic melts. Geochim Cosmochim Acta 69:50615069 Fulcher GS (1925) Analysis of recent measurements of the viscosity of glasses. J Am Ceram Soc 8:339-355,789794 Gauthier PJ, Condomines M, Hammouda T (1999) An experimental investigation of radon diffusion in an anhydrous andesitic melt at atmospheric pressure: Implications for radon degassing from erupting magmas. Geochim Cosmochim Acta 63:645-656 Gauthier PJ, Condomines M, Hammouda T (2000) Erratum to “An experimental investigation of radon diffusion in an anhydrous andesitic melt at atmospheric pressure: Implications for radon degassing from erupting magmas”. Geochim Cosmochim Acta 64:959-960 Klimm K, Botcharnikov R (2010) The determination of sulfate and sulfide species in hydrous silicate glasses using Raman spectroscopy. Am Mineral 95:1574-1579 Klouzek J, Nemec L (1995) A study of gas absorption by a glass melt using image recording and analysis. Glastechn Ber Glass Sci Tech 86:128-33 Koepke J, Behrens H (2001) Trace element diffusion in andesitic melts: An application of synchrotron X-ray fluorescence analysis. Geochim Cosmochim Acta 65:1481-1498 Liu Y, Zhang Y, Behrens H (2005) Solubility of H2O in rhyolitic melts at low pressures and a new empirical model to predict H2O and CO2 solubility in rhyolitic melts. J Volcanol Geotherm Res 143:219-235 Mehrer H (2007) Diffusion in Solids - Fundamentals, Methods, Materials, Diffusion-Controlled Processes. Volume 155 in Springer Series in Solid-State Sciences. Springer Misiti V, Vetere F, Mangiacapra A, Behrens H, Cavallo A, Scarlato P, Dingwell DB (2009) Viscosity of high-K basalt from the 5th April 2003 Stromboli paroxysmal explosion. Chem Geol 260:278-285 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Mungall JE (2002) Empirical models relating viscosity and tracer diffusion in magmatic silicate melts. Geochim Cosmochim Acta 66:125-143 Nemec L (1980a) The behavior of bubbles in glass melts. Part 1. Bubble size controlled by diffusion. Glass Technol 21:134-138 Nemec L (1980b) The behavior of bubbles in glass melts. Part 2. Bubble size controlled by diffusion and chemical reaction. Glass Technol 21:139-144 Nemec L, Klouzek J (1998) Interaction of gas mixtures containing SO2 and O2 with glass liquids. J Non-Cryst Solids 231:152-160 Nemec L, Mühlbauer M (1981) Verhalten von Gasblasen in der Glasschmelze bei konstanter Temperatur. Glass Sci Tech 54:99-108 Nemec L, Ullrich J (1998) Calculations of interactions of gas bubbles with glass liquids containing sulphates. J Non-Cryst Solids 238:98-114 Neuville DR, Courtial P, Dingwell DB, Richet P (1993) Thermodynamic and rheological properties of rhyolite and andesite melts. Contrib Mineral Petrol 113:572-581 Ni H, Zhang Y (2008) H2O diffusion models in rhyolitic melt with new high pressure data. Chem Geol 250:6878 Nowak M, Schreen D, Spickenbom K (2004) Argon and CO2 on the race track in silicate melts: A tool for the development of a CO2 speciation and diffusion model. Geochim Cosmochim 68:5127-5138 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285314 Prado MO, Fredericci C, Zanotto ED (2003a) Isothermal sintering with concurrent crystallization of polydispersed soda-lime-silica glass beads. J Non-Cryst Solids 331:145-156
Diffusion & Redox Reactions of Sulfur in Silicate Melts
105
Prado MO, Fredericci C, Zanotto ED (2003b) Non-isothermal sintering with concurrent crystallization of polydispersed soda-lime-silica glass beads J Non-Cryst Solids 331:157-167 Reid JE, Suzuki A, Funakoshi K-I, Terasaki H, Poe BT, Rubie DC, Ohtani E (2003) The viscosity of CaMgSi2O6 liquid at pressures up to 13 GPa. Phys Earth Planet Int 139:45-54 Roselieb K, Rammensee W, Büttner H, Rosenhauer M (1992) Solubility and diffusion of noble gases in vitreous albite. Chem Geol 96:241-266 Roselieb K, Rammensee W, Büttner H, Rosenhauer M (1995) Diffusion of noble gases in melts of the system SiO2-NaAlSi2O6. Chem Geol 120:1-13 Ryerson FJ (1987) Diffusion measurements: experimental methods. Methods Exp Phys 24A:89-130 Schreiber HD, Kozak SJ, Balazs GB, Fritchman AL, Schreiber CW (1989) Equilibrium and transport-properties of gases in E-glass melts. J Am Ceram Soc 72:1680-1691 Schreiber HD, Schreiber CW, Leonhard PG, Mc Manus KK, Trandel BD (1987) Solubility and diffusion of gases in a reference borosilicate melt. Diff Defect Data 53-54:345-350 Schuessler JA, Botcharnikov RE, Behrens H, Misiti V, Freda C (2008) Oxidation state of iron in hydrous phonotephritic melts. Am Mineral 93:1493-1504 Schulze F, Behrens H, Holtz F, Roux J, Johannes W (1996) The influence of water on the viscosity of a haplogranitic melt. Am Mineral 81:1155-1165 Shelby JE, Eagan RJ (1976) Helium migration in sodium aluminosilicate glasses. J Am Ceram Soc 59:420-425 Sipp A, Neuville DR, Richet P (1997) Viscosity, configurational entropy and relaxation kinetics of borosilicate melts J Non-Cryst Solids 211:281-293 Sparks RSJ, Barclay J, Jaupart C, Mader JM, Phillips JC (1994) Physical aspects of magmatic degassing. 1. experimental and theoretical contraints on vesiculation. Rev Mineral 30:413-445 Spickenbom K, Sierralta M, Nowak M (2010) Carbon dioxide and argon diffusion in silicate melts: insights into the CO2 speciation in magmas. Geochim Cosmochim Acta 74:6541-6564 Stelling J (2009) Diffusion, Speziation und Löslichkeit von Schwefel in Silikatschmelzen. PhD dissertation, Leibniz University of Hannover, Germany 165 pp Stelling J, Behrens H, Wilke M, Göttlicher J, Chalmin E (2011) Interaction between sulphide and H2O in silicate melts. Geochim Cosmochim Acta 75:3542–3557 Szurman M (2005) Schwefeldiffusion in Alumoborosilicatschmelzen. PhD dissertation, TU Clausthal, Germany, 118 pp Szurman M, Heide G, Frischat GH (2007) Sulphur diffusion in silicate glass melts with particular reference to Schott AF45 glass. Glass Technol: Eur J Glass Sci Tech A 48:242-246 Tamman G, Hesse W (1926) Die Abhängigkeit der Viskosität von der Temperatur bei unterkühlten Flüssigkeiten. Z Anorg Allg Chemie 156:245-257 Tinker D, Lesher CE (2001) Self diffusion of Si and O in dacitic liquid at high pressures. Am Mineral 86:1-13 Toplis M, Dingwell DB, Hess K-U, Lenci T (1997) Viscosity, fragility, and configurational entropy of melts along the join SiO2-NaAlSiO4. Am Mineral 82:979-990 Vetere F, Behrens H, Schuessler JA, Holtz F, Misiti V, Borchers L (2008) Viscosity of andesite melts – implication for magma mixing prior to Unzen 1991-1995 eruption. J Volc Geotherm Res (Unzen special issue) 175:208-217 Watson EB (1994) Diffusion in volatile-bearing magmas. Rev Mineral 30:371-411 Watson EB, Dohmen R (2010) Non-traditional and emerging methods for characterizing diffusion in minerals and mineral aggregates. Rev Mineral Geochem 72:61-105 Watson EB, Wark DA, Delano JW (1993) Initial report on sulfur diffusion in magmas. EOS Tran Am Geophys Union 74:620 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Whittington A, Hellwig BM, Behrens H, Joachim B, Stechern A (2009) The viscosity of hydrous dacitic liquids: Implications for the rheology of evolving silicic magmas. Bull Volc 71:185-199 Whittington AG, Richet P, Behrens H, Holtz F, Scaillet B (2004) Experimental temperature-X(H2O)-viscosity relationship for leucogranites, and comparison with synthetic silicic liquids. Trans Royal Soc Edinburgh: Earth Sci 95:59-72 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Winther KT, Watson EB, Korenowski GM (1998) Magmatic sulfur compounds and sulfur diffusion in albite melt at 1 GPa and 1300-1500 °C. Am Mineral 83:1141-1151 Zhang Y, Ni H (2010) Diffusion of H, C, and O components in silicate melts. Rev Mineral Geochem 72:171-225 Zhang Y, Stolper EM (1991) Water diffusion in a basaltic melt. Nature 351:306-309 Zhang Y, Xu Z, Liu Y (2003) Viscosity of hydrous rhyolitic melts inferred from kinetic experiments, and a new viscosity model. Am Mineral 88:1741-1752 Zhang Y, Xu Z, Zhu M, Wang H (2007) Silicate melt properties and volcanic eruptions. Rev Geophys 45:1-27
Method
diffusion couple
sodium trisilicate, NS3
bubble growth sorption of gas sorption of gas 35
S tracer, in air S tracer, in air
35 35 35 35
float glass
alkali borosilicate, SRL-131
lime aluminoborosilicate, PPG E-glass
barium aluminoborosilicate, AF45
barium aluminoborosilicate, AF45
AF45 + 5 wt% Na2O
AF45 + 10 wt% Na2O
AF45 + 15 wt% Na2O
cryst. diss; diff. couple cryst. diss; diff. couple diffusion couple
rhyolite, 5.8 ± 0.3 wt.% H2O
Fe-free dacite, nominal dry
diffusion couple
rhyolite, nominal dry
albite melt, nominal dry
Sulfur diffusion in melts of natural relevance
S tracer, in air
S tracer, in air
S tracer, red.
diffusion couple
soda lime silicate, NCS
S tracer
sorption of gas
soda lime silicate, NCS 35
diff. couple in air
soda lime silicate, NCS
soda lime silicate, NCS ± sulfate
S tracer
sodium trisilicate, NS3
35
S implantation + anneal in dry N2
silica glass
Sulfur diffusion in simple and industrial melts
Glass/melt
0.1 0.1
S6+
0.1 0.1 0.1
S6+ S6+ S6+
200 1000
S2−
200
1000
0.1
0.1
0.1
0.1
variable
variable
S6+
S6+
S
6+
S6+
S6+
0.1
100
S6+ or S2−
unknown
100
unknown
S
100
6+
100
S6+ or S2− unknown
0.1
P (MPa)
unknown
dom. S species*
1573-1773
1123-1773
1273-1773
1573-1773
1173
1173
1173
1173-1673
1173-1673
1373-1773
1312-1462
1698-1773
1273-1573
1273-1473
1523-1723
1573
1273-1473
1273-1523
1073-1173
T range (K)
3
6+1
3+1
7
1
1
1
8
23
3
unknown
unknown
12(S6+),13(S2−)
6+3
3
2
4
7
3
runs
257±16
258±10
167
60
128
215±31
216±7
188
m /s
2
m /s
2
−3.85
−7.09
−6.00
1.17±1.22
263
142±29
205±24
458±51
D = (3.88 ± 0.30)×10−13 m2/s
D = (2.26 ± 0.30)×10
−14
D = (1.32 ± 0.08)×10−15 m2/s
−3.44±0.69
−3.77±0.49
−3.79
−6.96
−6.35
−4.27±1.15
−4.6±0.3
−4.59
−10
167±17
177±22
193
Ea (kJ/mol)
D = (2↔8)×10
−6.3±0.6
−6.11±0.96
−8.22
log D0 (D0 in m2/s)
Appendix Table 1. Compilation of sulfur diffusion data for silicate melts.
Watson (1994)
Baker and Rutherford (1996) f
Baker and Rutherford (1996) e
Winther et al. (1998)
Szurman (2005)
Szurman (2005)
Szurman (2005)
Szurman et al. (2007) d
Szurman et al. (2007) d
Schreiber et al. (1989) c
Schreiber et al. (1987) c
Nemec and Mühlbauer (1981)b
Stelling (2009)
Backnaes et al. (2011)
Klouzek and Nemec (1995) a
Brückner (1962)
Backnaes et al. (2011)
Stelling (2009)
Espiau de Lamaestre et al. (2005)
Reference
106 Behrens & Stelling
diffusion couple diffusion couple diffusion couple diffusion couple
Etna basalt, nominal dry
Stromboli basalt, nominal dry
Etna basalt, ~4 wt% H2O
Stromboli basalt, ~3 wt% H2O
1000 500-1000 500-1000 500-1000 500-1000 0.1
S2− S2− S2− 2− 2− 2−
S
S
S
1000
S2−
1573
1498-1723
1498-1723
1523-1723
1523-1723
1573-1773
1573-1773
2
4
4
7
6
2
3
−12
D = 6.6×10
−5.57
−5.79
−5.52
−3.52
−7.49
−6.00
m /s
2
155±69
139±69
176±48
224±38
115
191
Watson (1994)
Freda et al. (2005)
Freda et al. (2005)
Freda et al. (2005)
Freda et al. (2005)
Watson (1994)
Watson (1994)
*
Notes: The predominant sulfur species was often not determined in the studies, and estimates are given here based on the experimental conditions (reducing/oxidizing) in the experiments. For details about redox conditions see text. a Arrhenius parameters were calculated by regression of diffusivities given in the paper. b Experimental details given by Nemec (pers. communication). c Arrhenius parameters and T range were estimated from the graphical representation in the paper. d Error of log D0 determined by regression of diffusion data in Szurman (2005). e In this Arrhenius relation a datum from Watson at 1773 K, 1 GPa was included to better constrain the activation energy. f Arrhenius parameters were obtained by refitting all data from Baker and Rutherford (1996) at 200 MPa with >6 wt% H2O plus one datum from Watson (1994) for a melt with 5.5 wt% H2O added at 1 GPa. The given water content refers to the water solubility in the melt at 200 MPa (Liu et al. 2005), see text for details.
desorption
diffusion couple
low-K andesite, 5.5 wt% H2O
lunar basalt
diffusion couple
low-K andesite, nominal dry
Diffusion & Redox Reactions of Sulfur in Silicate Melts 107
57.9 56.51
alkali borosilicate, SRL-131
E-glass
76.6 68.3 65.1 61.2 50.78 47.25 42.9
obsidian (Los Possos obsidian)
Fe-free dacite
Natural dacite
low-K andesite
Stromboli basalt
Etna basalt
lunar basalt
3.5
1.67
0.94
0.8
0.5
0.6
0.08
0.1
0.12
1.0 6.4
14.7
9.82
10.77
11.56
13.58
8.3
16.73
18.47
17.6
16.5
16.7
12.3
13.1
19.3
14.3
9.82
8.42
7.88
8.45
1.30
22.1
9.86
6.38
4.8
5.2
-
1.05
0.7
0.55
0.10
13.5
6.31
6.35
2.5
2.1
3.5
0.04
0.1
2.65
2.0
3.80
8.5
10.58
12.2
7.5
5.3
5.5
0.37
0.5
18.36
0.13
0.15
0.15
0.15
8.40
9.63
10
10
CaO
0.5
3.74
2.43
4
3.8
4
4.41
3.7
11.1
0.45
17.7
13.14
8.42
4.67
0.14
13.00
16.68
16
16
22.97
<0.1
1.7
1.89
1.6
1.4
1.5
4.64
4.8
0.35
5.7
0.50
Na2O K2O
Winther et al. (1998)
Freda et al. (2005) Freda et al. (2005) Watson (1994)
0.18 MnO, 0.54 P2O5, 160 ppm S 0.6 ZrO2, 0.3 MnO
Watson (1994)
Watson (1994)
Watson (1994)
Baker and Rutherford (1996)
Watson (1994) d
0.15 MnO, 0.38 P2O5, 160 ppm S
0.05 MnO
0.12 ZrO2, 0.18 GeO2, 0.14 Ga2O3
Schreiber et al. (1987)
0.5 ZrO2, 0.5 La2O3
Sipp et al. (1997) c
Szurman (2005)
Szurman (2005)
Szurman (2005)
Szurman (2005)
Nemec and Mühlbauer (1981) b
Stelling (2009), Backnaes (2010)
Klouzek and Nemec (1995) a
Brückner (1962)
Stelling (2009) a, Backnaes (2010)
Reference
10.23 BaO, 0.22 As2O3
11.27 BaO, 0.23 As2O3
11.63 BaO, 0.24 As2O3
12.10 BaO, 0.26 As2O3
0.3 SO3
others
a
Notes: Nominal composition in mol% b Composition given by Nemec (pers. communication) c Composition of the E-glass used by Schreiber et al. (1989) was not given. Here the composition of the E-glass used for viscosity measurements by Sipp et al. (1997) is reported. d Composition given by Watson (pers. communication)
68.2 76.8
albite
obsidian
Melts of natural relevance
60.4
63.5
AF45 + 5 wt% Na2O 58.39
65.33
barium aluminoborosilicate, AF45
AF45 + 15 wt% Na2O
72.60
float glass
AF45 + 10 wt% Na2O
75.21
74
soda lime silicate, NCS
soda lime silicate, NCS (Schott)
74
76.23
SiO2 TiO2 B2O3 Al2O3 FeOtotal MgO
soda lime silicate, NCS
sodium trisilicate, NS3 (Schott)
Simple and industrial melts
Glass/melt
Appendix Table 2. Compositions of melts in wt% used in sulfur diffusion studies.
108 Behrens & Stelling
Diffusion & Redox Reactions of Sulfur in Silicate Melts Appendix Table 3. Results of diffusion couple experiments with soda lime silicate and sodium trisilicate glass melts at 100 MPa (Stelling 2009). diffusing species
T (°C)
Dwell time (s)
H2O (wt.%)
log D (D in m2/s)
dom. S6+
1273
75600
n.a.
−13.13±0.08
NS3-D02-1000
dom. S
6+
1273
86400
0.257
−13.53±0.05
NS3-D09-1100
dom. S6+
1373
25200
0.440
−12.59±0.06
NS3-D16-1150
dom. S6+
1423
18000
0.229
−12.77±0.06
NS3-G31
dom. S6+
1473
7200
0.113
−12.40±0.06
NS3-G31 (XRF)
dom. S6+
1473
7200
0.113
−12.53±0.04
NS3-D11-1200
dom. S
6+
1473
10800
0.515
−10.89±0.03
NS3-D13-1250
dom. S6+
1523
7200
0.275
−11.97±0.05
NS3-D04-1000
S2
1273
19080
0.247
−13.31±0.07
NS3-D05-1000
S2-
1273
65100
0.235
−13.22±0.06
Soda lime silicate NCS-Dox-1000
S6+
1273
172800
0.076
−12.75±0.04
NCS-Dox-1050
S6+
1323
86400
0.025
−13.12±0.05
NCS-Dox-1050 (XRF)
S6+
1323
86400
0.025
−12.87±0.03
NCS-DC-1100
S
6+
1373
64800
n.a.
−13.75±0.05
NCS-Dox-1100I
S6+
1373
3600
n.a.
−13.13±0.09
NCS-Dox-1100II
S6+
1373
10800
n.a.
−12.95±0.06
NCS-Dox-1100III
S6+
1373
32400
n.a.
−12.26±0.06
NCS-G28
S6+
1373
28800
0.021
−13.03±0.04
NCS-G28 (XRF)
S6+
1373
28800
0.021
−12.98±0.03
NCS-Dox-1150
S6+
1423
64800
0.046
−12.17±0.04
NCS-Dox-1150 (XRF)
S6+
1423
64800
0.046
−12.16±0.04
NCS-DC-1200
S6+
1473
14400
n.a.
−13.00±0.05
NCS-Dox-1200II
S
6+
1473
3600
n.a.
−11.87±0.07
NCS-Dox-1250
S6+
1523
14400
0.012
−11.52±0.03
NCS-Dox-1250 (XRF)
S6+
1523
14400
0.012
−11.53±0.12
NCS-Dred-1000
S2-
1273
172800
n.a.
−12.66±0.03
NCS-Dred-1050
S2-
1323
86400
0.009
−12.25±0.03
NCS-Dred-1050 (XRF)
S2-
1323
86400
0.009
−12.22±0.03
NCS-Dred-1100II
S2-
1373
3600
n.a.
−12.82±0.10
NCS-Dred-1100IV
S2-
1373
10800
n.a.
−12.78±0.07
NCS-Dred-1150
S2-
1423
64800
0.009
−11.84±0.04
NCS-Dred-1150 (XRF)
S
2-
1423
64800
0.009
−11.73±0.03
NCS-Dred-1200
S2-
1473
14400
n.a.
−11.53±0.06
NCS-Dred-1200II
S2-
1473
3600
n.a.
−12.02±0.04
NCS-Dred-1250
S2-
1523
14400
n.a.
−11.61±0.06
NCS-Dred-1250 (XRF)
S2-
1523
14400
n.a.
−11.71±0.03
NCS-Dred-1300
S
2-
1573
10800
n.a.
−11.18±0.06
NCS-Dred-1400
S2-
1573
420
n.a.
−11.49±0.09
Run # Sodium trisilicate NS3-D01-1000
Notes: The sodium trisilicate starting glasses were synthesized in an IHPV adding FeS (source of S2−) or Na2SO4 (source of S6+). Soda lime silicate starting glasses were synthesized at ambient pressure adding Na2S (source of S2−) or Na2SO4 (source of S6+). Samples marked with XRF were analyzed using XRF line scans at fixed absorption energies, other profiles were measured by EMPA. All runs were conducted in NQ-IHPVs at PH2 ~ 0.2 bar with heating rates of 30 K/min and initial cooling rates of ~150 K/min. Effective run durations used in the calculation of the diffusion coefficients were calculated by adding an increment of ~150 s to the dwell time, to account for heating and cooling. H2O contents of some post-experimental glasses were measured by IR microspectroscopy (Behrens and Stuke 2003).
109
Behrens & Stelling
110
Appendix Table 4. Results of sulfide/sulfate interdiffusion experiments with soda lime silicate glass melts at 100 MPa (Stelling 2009). Run # G15 G16 G11 G12 G13 G14 G23 G24
Diffusing species
Sulfur source
T (K)
Dwell time (s)
max. H2O (wt%)
log DS (D in m2/s)
S2− vs. S6+ S2− vs. S6+ 2− S vs. S6+ 2− S vs. S6+ 2− S vs. S6+ S2− vs. S6+ S2− vs. S6+ 2− S vs. S6+
Na2S Na2SO4 FeS Na2SO4 Na2S Na2SO4 FeS Na2SO4 Na2S Na2SO4 FeS Na2SO4 Na2S Na2SO4 FeS Na2SO4
1373
21600
1373
21600
1473
3600
1473
3600
1473
16200
1473
16200
1473
16200
1473
16200
n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a. 0.011 0.011 0.012 0.012
−13.68±0.08 −13.70±0.06 −13.85±0.08 −14.10±0.09 −12.89±0.07 −13.01±0.07 −13.38±0.11 −13.46±0.10 −13.46±0.11 −13.45±0.07 −13.58±0.09 −13.52±0.06 −13.88±0.10 −13.67±0.07 −14.06±0.12 −13.85±0.08
Notes: Sulfur sources were added to glass powder in the syntheses of the starting glasses. All runs conducted were conducted in NQ-IHPVs at PH2 ~ 0.2 bar with heating rates of 30 K/min and initial cooling rates of 150 K/min. Effective run durations used in the calculation of the diffusion coefficients were obtained by adding typically an increment of ~150 s to the dwell time, to account for heating and cooling. Diffusion coefficients were determined using XRF line scans at fixed absorption energies for each species. H2O contents of some glasses were measured by IR microspectroscopy (Behrens and Stuke 2003).
2507.06
2
723-1673
773-1573
T
w
1122-1894 1290-2139 1324-2189
5515.26 1907.23 218.66 × exp −469.1 w log h = −4.26 + + × exp 2+ T Fe (T − 275.43) (T − 626.24 ) Fe T tot
× exp −483.99
257.5 559.4 22.3
1023-1623 1023-1623 673-1873 1423-1773 1223-1573 1273-1623 1323-1773 673-1873 873-1273
T range (K)
573-1923
Notes: w refers to the water content in wt%, x to the mole fraction of water on a single oxygen basis.
Andesite
+
(T − 218.14 ) (T − 565.34 )
8605.42
Equations for H2O-bearing melts
14900.0 8132.0 19181.0
503.5 547.8 533.0 511.2 609.0 744.0 715.0 510.5 551.9
T0 (°C)
log h = −7.36 +
Stromboli basalt
Rhyolite
−6.385 −4.760 −5.880
Dry melts with natural relevance albite, NAS 75:50 Fe-free dacite, 0.01 wt% H2O “dry” rhyolite, 6 ppm H2O
3705.4 4223.7 4358.4 6333.0 3842.0 2236.0 2097.0 4983.2 6606.0
VFT parameters B
1812.2 1.9448 1+ T 49584 1795.5 + − x log h = − log exp 18.5611 − exp 1.47517 T T
−2.037 −2.462 −2.700 −3.400 −2.200 −1.400 −1.600 −2.086 −4.820
A
Simple and industrial melts sodium trisilicate, NS3 soda lime silicate, NCS float glass barium aluminoborosilicate, AF45 AF45 + 5 wt% Na2O AF45 + 10 wt% Na2O AF45 + 15 wt% Na2O Aluminoborosilicate, ABS E-glass
Glass/melt
Appendix Table 5. Relations used to calculate melt viscosities.
Vetere et al. (2008)
Zhang et al. (2003)
Misiti et al. (2009)
Toplis et al. (1997) Whittington et al. (2009) Neuville et al. (1993)
Bornhöft & Brückner (1999) Bornhöft & Brückner (1999) Prado et al. (2003b) Szurman (2005) Szurman (2005) Szurman (2005) Szurman (2005) Prado et al. (2003a) Sipp et al. (1997)
Reference
Diffusion & Redox Reactions of Sulfur in Silicate Melts 111
5
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 113-141, 2011 Copyright © Mineralogical Society of America
The Role of Sulfur Compounds in Coloring and Melting Kinetics of Industrial Glass Roberto Falcone, Stefano Ceola, Antonio Daneo, Stefano Maurina Stazione Sperimentale del Vetro Via Briati 10 I-30141 Venice-Murano, Italy
[email protected]
InTRoDuCTIon Soda-lime-silica (SLS) glass is the most widely used of all commercial types of glass. This type of glass is mainly used for manufacturing windowpanes, household glassware and glass containers (e.g. bottles, jars) for foodstuffs and beverages. These types of endproducts differ in their application, and production method (e.g., blowing and pressing for containers and glassware, float process for windows) as well as in their chemical composition. Nevertheless, they are all produced by melting mixed raw materials (batch) in a glass furnace at maximum temperatures ranging between 1500-1600 °C. The batch consists mainly of silica sand, sodium carbonate (soda), lime, dolomite and variable amounts of glass cullet. Small quantities of alumina-bearing raw materials, fining agents (e.g., sodium sulfate), coloring and reducing/oxidizing agents are also added to the batch. For the production of SLS glass, sulfur containing raw materials (e.g. sulfates and sulfides) play an important role in determining the final product’s quality. These compounds are involved in the final part of the fusion process, known as the fining process. In this process, the decomposition of raw materials generates a large amount of gas. The evolution of those gases from the glass melt is enhanced by the presence of the sulfur compound (Kloužek et al. 2007). Moreover, sulfur compounds act as oxidizing (sulfates) or reducing (sulfides) agents, playing a decisive role in the coloring mechanism of the final glass. Sulfates also enhance the kinetics of the sand’s dissolution by wetting the sand grains at relatively low temperatures thereby, accelerating the melting process (Albayrak and Sengel 2008; Müller-Simon and Gitzhofer 2008; Daneo et al. 2006, 2009). In Table 1, typical composition ranges for SLS containers glasses produced in Italy are reported in wt% of oxides (data from Stazione Sperimentale del Vetro, SSV). The total sulfur concentration in the final glass, conventionally expressed as total sulfates (SO3,tot = 2.497Stot), varies from less than 0.02 wt% up to approx. 0.3 wt% depending mainly on the melt’s redox state and, secondarily, on the maximum melting temperature. Moreover, sulfur may be present in the final glass exclusively as sulfates in oxidized glass and both as sulfide and sulfate in variable ratios depending on reduced glass types. The total sulfur amount present in glass and its oxidation state(s) are determined by its desired composition and by the process parameters 1529-6466/11/0073-0005$05.00
Table 1. Typical composition ranges (wt%) of industrial SLS container glasses produced in Italy (data from SSV). oxide
min
max
SiO2 Al2O3 Na2O K2O CaO MgO BaO SO3,tot Fe2O3tot Cr2O3
70.0 1.0 11.0 < 0.05 8.5 < 0.05 < 0.01 < 0.02 0.015 < 0.0003
73.0 3.0 14.0 2.5 12.0 4.0 1.5 0.30 0.90 0.25
DOI: 10.2138/rmg.2011.73.5
Falcone, Ceola, Daneo, Maurina
114
decided with regards to its manufacturing. In the production of SLS glass, a lack of sulfur means a lack of fining agents and, consequently, the final glass object will contain many small bubbles. On the other hand, if too much sulfur is introduced into the batch, the sulfur exceeding the solubility limits established by the redox state the and melting temperature forms SO2 and H2S bubbles that leaving the furnace render air emissions more pollutant.
SulFuR CoMpounDS In InDuSTRIAl GlASS pRoDuCTIon In SLS containers and flat glass manufacturing sulfur is introduced into the batch with different raw materials.
Sodium sulfate Sodium sulfate, Na2SO4, commonly referred to as “salt cake”, is used as a fining and oxidizing agent. It helps remove small bubbles from molten glass, fluxes the melt and prevents scum formation during refining. This raw material may be recovered from naturally occurring brines; however, glass industry commonly uses sodium sulfate produced as a by-product of other processes in the chemical industry. Its main industrial processes include the production of hydrochloric acid (a reaction between sulfuric acid and common salt) and synthetic textile fibers such as rayon (the neutralization of surplus sulfuric acid with sodium hydroxide).
Slag Processed blast furnace slag (commercially known as calumite or vitrite) is a secondary raw material obtained as a by-product in the metallurgy industry (blast furnace slag). The use of this material started in the early 1950s in America; it was introduced into Europe in the 1970s. Processed blast furnace slag contains about 1 wt% of sulfides and it is therefore classified as a reducing agent by glass technologists. Nevertheless, this raw material is also an inexpensive source of alumina. The introduction of small amounts of slag into a batch (up to 4-5 wt%) increases melting rates and improves glass refining, resulting in higher quality glass and lower energy consumption (Simpson 1976, 1979; Hreglich et al. 1979). Details of usage of slags in steel production for desulfurization are described by Lehmann and Nadif (2011, this volume)
Glass cullet In the production of hollow glass (glass containers), variable amounts of recycled glass cullet are introduced into the batch. This important secondary raw material contains about 0.1-0.3 wt% SO3,tot, depending on the various types of glass fractions contributing to cullet composition.
Filter dust In order to comply with the environmental limitations imposed by strict national and European rules for gas and solid emissions, all glass plants are equipped with fume and waste gas treatment systems for the removal of dusts and gaseous pollutants (e.g., SOx, HCl, HF). Filtration systems are primarily made up of electrostatic precipitators where a basic reagent (e.g., calcium hydroxide, sodium carbonate, sodium bicarbonate, sodium hydroxide) is used to neutralize acid gases. The solid product of this reaction contains a high percentage of sulfates (total sulfur expressed as SO3 usually ranges between 40 and 55 wt%) and minor amounts of sulfides (from <1 wt% to about 15 wt%). These dusts are continuously recycled in the original source tank; nevertheless, due to compositional variations and the presence of undesired components (mainly in furnaces where high amounts of recycled cullet are used), the amount of filter dust in the glass batch usually accounts for less than 1 wt% (Kraub et al. 1995; Scalet 1996).
Sulfur Compounds in Industrial Glass
115
pyrite Naturally occurring iron sulfide (FeS2) is used as a sulfur source and reducing agent in amber-colored glass production.
others Moreover, small but not negligible amounts of sulfur can be involuntarily added to the batch as undesired impurities from other natural raw materials (e.g. sands, limestone and dolostone) and heavy oil fuel (Daneo et al. 2006; Müller-Simon and Gitzhofer 2008).
InDuSTRIAl GlASS pRoDuCTIon Today, glasses can be produced using a wide variety of alternative methods. Nevertheless, the majority is still produced by melting a well-mixed “batch” of appropriately weighed quantities of raw material. Regardless of the final end-product, all industrial glass manufacturing processes share the same main steps (Shelby 1997): batch preparation, melting, fining, forming, and post forming. A schematic overview of the main steps involved in the industrial production process of glass containers is reported in Figure 1.
Silica sand
Soda ash
SiO2
Na2CO3
Limestone CaCO3 Dolostone CaMg(CO3)2
Other raw materials
Al2O3–bearing fining, coloring oxidising, reducing etc. …
Recycled glass cullet
Weighing and batch mixing
Melting (1500°C)
Bridgewall
Fining and homogenizing (1550°C)
Throat Conditioning (1300°C) Delivery Forming (950–1250°C) Hot-end coating Annealing lehr (500-550°C) Cold-end coating Inspection and product testing
Cullet
Packing, warehousing and shipping
Figure 1. Schematic overview of glass containers production.
116
Falcone, Ceola, Daneo, Maurina
Batch preparation During the batch preparation phase, raw materials are weighed and blended to achieve the desired final glass composition. Batch materials can be divided into five categories based on their role in the process; glass-formers (network-formers), fluxes, stabilizers, colorant and fining agents. Glass-formers are the most essential components of any glass batch as they provide the glass with network-forming oxide. The primary network-formers in commercial glasses are silica (SiO2) and boric oxide (B2O3). The vast majority of commercial glasses use silica as the glass-former. In the production of SLS glass products (mainly containers and flat glass), the main glass-former batch component is silica sand (quartz, SiO2). Moreover, silica is also introduced into glass by other natural silicate raw materials, such as feldspar, nephelinesyenite, and by slag or glass cullet. Boron oxide is used in the production of borosilicate glasses, the most important of which are neutral pharmaceutical glasses and thermally and chemically resistant glasses (e.g., Pyrex® or Duran® type). B2O3 is introduced as synthetic borax (sodium borate hydrate, Na2B4O7·10H2O) and/or boric acid (H3BO3) while colemanite, a naturally occurring mineral (calcium borate with nominal composition Ca2B6O11·5H2O), is rarely used (Sinton 2006). Due to the very low, temperature dependent solubility of sulfur in borosilicate melts, sulfates are not used as fining agents in the production of these glasses, whereas sodium chloride is used as an alternative to arsenic or antimony oxide (Volf 1961, 1990). Therefore, these types of glasses will not be discussed further in this chapter. Due to the high melting temperature (>2000 °C) required to produce vitreous silica, the production of silicate glasses requires the addition of a flux to reduce the processing temperature to within practical limits. The most commonly used fluxes are alkali oxides, mainly Na2O and then K2O. In SLS glass, sodium carbonate produced with the Solvay process is added to the batch as a flux. A specific type of soda with appropriate grain size and bulk density, also known as dense soda ash, has been developed to meet the requirements of the glassmakers (Sinton 2006). Nevertheless, the addition of alkali oxides results in serious degradation of the chemical durability of the glass, which usually refers to the behavior of glass with respect to water and aqueous solutions (Paul 1990; Scholze 1991). This degradation is countered by the addition of stabilizers (property modifiers). Common stabilizers are alkaline earth oxides (mainly CaO and MgO) which are introduced into the batch in the form of carbonates (limestone and dolostone). Nevertheless, SLS hollow glasses may undergo weathering phenomena on the internal surface of the container during storage in pallets in adverse conditions before filling (Verità et al. 2007; Chopinet et al. 2008; Falcone et al. 2011). Since alumina (Al2O3) strongly improves the chemical resistance of the final glass (Paul 1990; Scholze 1991), aluminabearing raw materials such as calcined or hydrated alumina, alkali feldspar, nepheline syenite, and processed slag are introduced into the batch in order to reach Al2O3 concentrations up to 2.5-3 wt% in the final glass. As previously mentioned, another important raw material in glass manufacturing is cullet, i.e., recycled glass obtained from within the plant (rejected products) and recycled glass from other recycling companies. Cullet is less expensive than virgin raw materials and it reduces the energy required for melting. In the production of some types of colored glass containers, cullet can constitute up to more than 80 % of the batch. Other minor components of the batch include coloring and the previously mentioned fining agents (sulfates). Furthermore, reducing and oxidizing raw materials are added to define the redox state of the melt and the glass end product as discussed below.
Sulfur Compounds in Industrial Glass
117
Melting In the melting furnace, the batch goes through four phases: melting, fining homogenizing and conditioning. Melting starts when the batch enters the furnace and is completed when all crystalline components are completely dissolved. A clearly defined space separation between melting, homogenizing and fining zones is not possible, since some homogenizing and refining take place even in the melting zone. However, melting should normally be completed in the first half of the melting chamber. Typical maximum melting temperatures, for industrial SLS glasses, range between 1500 and 1600 °C. In the melting process the formation of glass currents in the tank is essential for the satisfactory operation in the glass furnace (good final quality of the glass and optimum furnace energy consumption) (Trier 1987). Glass currents promote diffusion and contribute significantly to the homogenization of the inhomogeneous primary melt. The most important glass currents are withdrawal currents and convection currents. The former are originated by the continuous removal of glass at one end of the furnace and the continuous charging of batch at the opposite end. The convection currents are generated by density differences in the bath of the molten glass, due to thermal effects (e.g., batch charging, heating from above, heat losses from the tank, etc.). The temperature (and thereby density) distribution in the glass tank largely depends on the design of the furnace and on several operative factors. Since part of the glass surface is shielded from flame radiation by being covered with batch and foam, the heat losses in these zones are compensated for by glass currents. Another very important factor for current generation is the depth temperature gradient, which result from the melt being heated from above. Temperature differences between the hot glass surface and tank bottom depend on the glass color; Trier (1987) reports that clear glasses give gradients of 1.3-1.5 °C/cm, whereas glasses with lower transparency such as amber and green glass give gradients ranging from 2.5 to 3.5 °C/cm. Moreover glass currents are also influenced by bubblers and electric boosting if these devices are used to improve melt homogenization. In addition to improving glass melting and homogenization, glass currents ensure that the individual stages of glass melting (i.e., batch melting, refining and conditioning of the glass) take place at the correct time and in the correct location in the tank, as will be discussed in the next paragraph. Glass currents make a great contribution to batch melting of and to the homogenization of molten glass by promoting chemical diffusion which leads to a leveling out of concentrations. Diffusion-controlled processes in multi-component systems such as SLS batches and melts are complex and can be differentiated according to the type of diffusing species: network formers, network modifiers, and gases (Shaeffer 1984). The description of such processes goes beyond the aim of this paper. For a detailed review on sulfur diffusion in silicate melts see Behrens and Stelling (2011, this volume).
Batch reactions Batches for SLS glass production are composed of particles of different raw materials, mainly quartz, sodium carbonate and calcium carbonate. At increasing temperature the batch particles start to react at their contact surface producing primary intermediate phases. The melting process is strongly influenced by formation of these intermediate reaction products and by their chemical and physical properties. Knowledge of SLS glass batch reactions (phase evolution and kinetics analysis of batch components) is therefore essential to developing predictive models for the glass industry. These models enable glassmakers to improve the efficacy of the melting and fining processes, by defining parameters such as raw materials grain size, heating rate, etc.. The first stage of the melting phase includes all reactions that occur before the first melt appears (solid state reactions). These reactions are mainly dependent on solid particle-particle interactions and are diffusion-controlled processes. The degree to which solid state reactions convert batch materials into solid products depend on several parameters.
118
Falcone, Ceola, Daneo, Maurina
Packing density, grain size and degree of homogeneity determine the contact area between reactants and the diffusion distances. Moreover reactions are strongly influenced by the rate of heating. Kinetic measurements showed that equilibrium for ternary SLS batches require times in the order of tens of min to be reached at temperature lower than 900 °C (Dolan and Misture 2004). Since the heating rates in industrial melting are generally in the order of hundreds of degrees per minute, batch reactions in the glass tanks occur in a non-equilibrium regime. However, according to Hrma (1990), if the batch particles are fine and well mixed and there is sufficient time for the processes to progress, solid state reactions may consume large portions of silica and decompose most of the carbonates. The second stage of melting is characterized by the presence of original batch constituents, intermediate crystalline reaction products, and melt. Both low viscosity molten salts and high viscosity glass forming (silicates) melts can be produced in the batch. The liquid phase assists reactions of solids by dissolving them and transferring ions more rapidly than processes involving solid state diffusion. Therefore the reaction rates increase significantly when the first melt occurs (Hrma 1990). If we consider typical basic industrial SLS glass compositions (e.g., 70-75 wt% SiO2, 1216 wt% Na2O and 10-14 wt% CaO) in the ternary Na2O-CaO-SiO2 diagram in Figure 2, we can see that this composition lies inside the “eutectic valley” running between a binary eutectic in the Na2O-SiO2 diagram and another eutectic in the CaO-SiO2 diagram. This area corresponds to a liquidus temperature of about 1000 °C (Barton 2001; Barton and Guillemet 2005). In the ternary Na2O-CaO-SiO2 phase diagram, SLS glass compositions fall into a region where the most probable solid phases are tridymite (SiO2), b-wollastonite (CaSiO3) and devitrite (Na2Ca3Si6O16) (see Fig. 2). In fact these phases are the most common crystalline inclusions (defects called stones by glass technologists) arising from devitrification in SLS melts. Molten glass may devitrify as a result of a nucleation process and crystal growth which may occur if the melt temperature drops below the liquidus temperature (Clark-Monks and Parker 1980).
Figure 2. Soda-lime-silica glass composition in the Na2O-CaO-SiO2 phase diagram; T = tridymite, W = wollastonite, D = devitrite (redrawn after Levin et al. 1964).
Sulfur Compounds in Industrial Glass
119
During the very first stage of batch reaction only a small proportion of quartz particles is in contact with carbonates that initiate the batch reactions, whereas a large proportion of quartz particles is in contact with other quartz particles and thereby does not participate in initial batch reactions. At this stage, silica and carbonates are converted into new crystalline phases. For example sodium carbonate and silica produce sodium metasilicate (Na2Si2O3) which melts at a temperature as high as 1089 °C. Nevertheless this compound can further react with silica to form sodium disilicate (Na2Si2O5), or with sodium carbonate to form sodium orthosilicate. Moreover sodium carbonates can react with calcium carbonate to form the double carbonate Na2Ca(CO3)2 (Hrma 1990; Novotný and Lošot 2008).
Temperature (°C)
Based on the binary SiO2-Na2O phase diagram in Figure 3, formation of Na2SiO3 by diffusion of Na+ to the SiO2 boundary is expected to range from 670 °C to 780 °C (Hlavac 1983; Kim and Sanders 1991). Metasilicate reacts with quartz to form a disilicate Na2Si2O5 which melts at 874 °C but forms an eutectic with silica at 790 °C (Hrma 1985; Novotný and Lošot 2008). Studies on the reaction kinetics of binary SiO2 - Na2CO3 mixture report that, at temperatures close to the formation of the first eutectic melt (790 °C), kinetics of sodium disilicate formation are initially much slower than those of sodium metasilicate. The latter is the thermodynamically stable equilibrium phase whereas Na2Si2O3 is probably an intermediate phase in the disilicate formation (Hrma 1990; Dolan and Misture 2004). According to Novotný and Lošot (2008) the initial melting pathway is strongly influenced by the grain size of the raw materials, which also affects the fining process. In case of relatively larger grains sodium metasilicate would be the predominant intermediate product. By contrast, the use of very fine silica sand would results in a pathway with sodium disilicate as the paramount intermediate
SiO2 (wt%) Figure 3. Na2O-SiO2 phase diagram (after Levin et al. 1964).
120
Falcone, Ceola, Daneo, Maurina
product. In the metasilicate pathway the first melt would correspond to molten Na2SiO3 which has a relatively low viscosity which would promote melt degasssing. In the disilicate pathway a more viscous liquid phase, probably consisting Na2Si2O5-SiO2 eutectic melt, would form initially which would hamper CO2 from escaping from the melt. Based on the Na2CO3-CaCO3 phase diagram Na2Ca(CO3)2 is thermally stable above 335 °C. The double carbonate melts between 725 °C and 813 °C (eutectic temperatures) depending on the soda-lime batch ratio, at temperatures that are lower than the melting points of sodium carbonate (850 °C) and sodium disilicate (874 °C). Results obtained by Dolan and Misture (2004) using in situ high-temperature X-ray diffraction (HTXRD) measurements on ternary quartz, sodium carbonate and calcium carbonate laboratory and SLS industrial batches at different heating rates, suggest that initially, the reaction path mainly follows the binary systems (mainly Na2O-SiO2) of the ternary phase diagram Na2O-CaO-SiO2. Carbonate decomposition was observed only for CaCO3 with formation of the relatively stable CaO which precludes reaction with quartz until about 1000 °C. By contrast sodium carbonate does not decompose to a stable solid oxide maintaining the reactive carbonate form with much faster reaction kinetics with quartz compared to calcite. Results on a Na2CO3CaCO3 laboratory batch indicate that only small amounts of Na2Ca(CO3)2 is formed at a slow heating rate (10 °C/min), whereas no detectable amount of this phase was found at high heating rate (200 °C/min), before calcite decomposition occurred (750 °C). These results suggest that in SLS batch melting the lowest temperature of liquid formation is not a result of the predicted low temperature eutectic (725-813 °C). Due to the slow kinetics of double carbonate formation and the reaction kinetics of sodium silicates, the reaction path most likely initially follows the Na2O-SiO2 phase diagram (to the Na2Si2O5), followed by further reactions with carbonates and oxides to form ternary phases (Na4CaSi3O9 and Na2Ca2Si3O9). These results are in agreement with previous works where sodium metasilicate and sodium disilicate were identified as the intermediate reaction products, while ternary compounds would form at later stages (Hrma 1990). Based on this data and HTXRD measurements, Dolan and Misture (2004) proposed a possible reaction sequence for binary and ternary phases in a ternary SLS batch, before complete batch melting occurs. The reaction path for the formation of binary phases mainly considers the reaction between silica and sodium carbonate to form sodium metasilicate which further reacts with silica to form sodium disilicate; reaction of quartz with lime to form calcium silicate is also reported: Na2CO3 + SiO2 → Na2SiO3 + CO2 (g)
(1)
Na2SiO3 + SiO2 → Na2Si2O5
(2)
CaO + SiO2 → CaSiO3
(3)
The reaction path for formation of ternary phases considers formation of Na4CaSi3O9 and Na2Ca2Si3O9 are formed by further reaction of sodium disilicate with carbonates and oxides: Na2Si2O5 + 2CaO + SiO2 → Na2Ca2Si3O9
(4)
2CaSiO3 +Na2CO3 + SiO2 → Na2Ca2Si3O9 + CO2 (g)
(5)
It should be noted that high volumes of CO2 are released from the carbonates decompositions. The same intermediate phases were detected in a range of ternary compositions, thus confirming that phase formation is mainly dependent on particle-particle interactions, rather than on the overall compositions. Dolan and Misture (2004) also tested the effect of cullet on batch reactions. The “solvent” effect on quartz was illustrated by results on binary quartz-cullet mixture where the complete melting of quartz was detected from 900 °C to 1200 °C. Moreover the addition of cullet to
Sulfur Compounds in Industrial Glass
121
the same SLS batch did not alter the reaction sequence, (i.e., the same intermediate phases observed on cullet-free were detected). However reaction of cullet with other batch components was confirmed, as well as the influence of cullet on the amounts of rates of phase formation. When the stage of melting reactions is over, the batch has been transformed in a fluid with suspended residual refractory silica grains and bubbles. In this stage, evolution of oxygen from redox reactions, and evolution of fining gases occur resulting in the dissolution of residual solid grains and in the removal gaseous inclusions.
Fining The term fining refers to the physical and chemical processes occurring in the melting chamber during which bubbles (mainly CO2 inclusions produced by the decomposition of carbonates) are removed from the melt (i.e., the complete degassing and homogenization of the primary melt). This topic is extensively described by Müller-Simon (2011, this volume). Here, only a brief overview is reported so to provide a complete description of the industrial process. For a good final glass quality it is essential that only fully refined glass reaches the working end. As previously mentioned, glass currents play a decisive role and, although the discussion of this complex matter goes beyond the aim of this paper, a few general information are reported here. In flame heated tanks, firing along the length of the furnace is arranged so that a temperature maximum occurs in the refining zone (hot-spot, i.e., the area with the highest temperature). Convection currents in conjunction with withdrawal currents, result in a counterclockwise rotating current which prevents the batch blanket and foam from advancing too far into the refining zone (thermal barrier) and which forces melted glass upwards into the zone of high temperature (Trier 1987). Moreover, in glass container furnaces, a weir (also called bridgewall, height is typically 0.3 × glass depth) is placed in the fining section. The weir forces the molten glass upwards into the hot-spot, improving the fining process and preventing the ingress of contaminated glass from the floor into the throat. The upward glass current at the hot-spot can be reinforced by the use of bubblers, which produce controlled formation of bubbles at the tank bottom, and also improve melt homogenization. Electrical boosting can be also used to improve glass refining. During the first stages of glass melting, the decomposition of the raw materials (especially carbonates) gives rise to the formation of many gaseous inclusions, constituted mainly of CO2. The rate at which these bubbles leave the melt due to buoyancy is directly proportional to the bubbles’ cross section (bigger bubbles tend to reach the surface of the glass melt more rapidly) and inversely proportional to the viscosity of the melt. This can be summarized by the following relationship (Levich 1962): ν=
2r 2 ( r m − r g ) g 9h
(6)
where h is the viscosity, r is the bubble radius, g is the gravity acceleration and rm is the density of the melt and rg is the density of the bubble content. For a common SLS container glass, we obtain a rate of approximately 1 cm per hour for a bubble with a 0.1 mm radius at 1450 °C (melt viscosity about 10 Pa·s). In order to enhance the buoyancy of the smaller bubbles towards the surface, fining agents are added to the batch mixture (chemical fining). During the final stage of the melting process, the fining compounds decompose, developing large bubbles within the melt. These large bubbles tend to reach the top level of the glass melt, remixing the molten glass as they rise, thus homogenizing the melt. As they move, they capture and drag smaller bubbles that have not yet burst. Chemical fining is currently the most effective way to remove bubbles from glass melt.
122
Falcone, Ceola, Daneo, Maurina
In SLS glass production, sodium sulfate is widely used as a fining agent; this compound decomposes above 1550 °C, developing SO2 and O2 bubbles (Backnaes and Deubener 2011, this volume). During the fusion process, at a temperature just above the formation of the first permanent silicate liquid phase (about 1100 °C), sodium sulfate is a liquid almost completely insoluble in the glass and tends to occupy interstices within solid aggregate, between silicate melts and solids or at melt/gas interfaces. Liquid sulfate is highly fluid and promotes wettability and melting of solid batch particles at an increased rate and gas bubbles are expelled more rapidly from the melt; in other words, liquid sulfate acts as a detergent and fining agent. Nevertheless, at low or medium time-temperature regimes, the reaction of liquid sulfate with sand grains is very slow and it may inhibit the dissolution of the solid batch particles. Therefore, the unbalanced use of sulfate may result in an excess of foaming, undissolved sand grains and redox instabilities For this reason, in order to avoid such drawbacks during glass melting, the time-temperature regime needs to be kept under control and balanced with respect to the glass pull out. The thermal decomposition of the sodium sulfate starts above 1450 °C: 2Na2SO4 (m) → 2Na2O (m) + O2 (g) + 2SO2 (g)
(7)
The decomposition products—Na2O, O2 and SO2—are transferred from an un-decomposed, liquid phase (liquid Na2SO4) to the glass melt, causing a release of energy and, as a consequence, convection. This convection further accelerates dissolution of the residual solid particles that are still present in the melt. At 1550 °C, the SO2 and O2 formed by decomposition of dissolved Na2SO4 diffuse into small pre-existing bubbles in the melt and enlarge these bubbles, thus enhancing their buoyancy towards the surface. The partial pressure of the gases already present in the bubble (mainly Ar, CO2, N2 and H2O) decreases. The partial pressure decrease of these gases into the bubble disturbs the equilibrium previously established with the same gases dissolved in the melt. This causes diffusion of the gases from the melt into the bubble, as described by Henry’s Law equilibrium (Beerkens 2005, 2007). If a reducing agent (i.e., carbon or slag) is added to the glass batch the sulfate decomposition temperature is lowered. In presence of carbon sulfates are reduced to sulfides at lower temperature (1100-1300 °C), with production of SO2. Sulfides react with sulfates to produce more SO2. Due to this supersaturation of SO2 in the melt there is a first generation of bubbles that accelerates sand grain dissolution by foaming and bubble nucleation, resulting in more intense melt convection (Kloužek et al. 2007). At high carbon content (i.e., C/SO42− molar ratios > 1) all sulfates are decomposed and consumed by the reaction with carbon. For lower carbon content (i.e., C/SO42− molar ratios < 1) a second generation of bubbles occurs at temperatures higher than 1450 °C due to the thermal decomposition of the residual sulfate present in the melt giving SO2 and O2 , according to reaction (7) (Arkosiovà et al. 2008). Therefore, the introduction of proper amounts of carbon or other reducing raw materials in the batch results in a more effective melt fining. After the fining stage, the temperature is lowered to working conditions (generally 10001100 °C); during cooling, small bubbles collapse and the gases dissolve into the melt. Afterwards, while the glass melt cools, some bubbles made of SO2 and O2 are re-absorbed with the formation of sulfate groups within the silica network. If the temperature drops quickly the bubble does not vanish and will form a solid sulfate nano-layer on its surface in the case of oxidized glass. In reduced glass, oxygen is promptly dissolved by reducing agents (such as ferrous iron) in the glass melt, leaving bubbles made solely of SO2.
Sulfur Compounds in Industrial Glass
123
Forming and post-forming After refining, molten glass undergoes forming operations, and this is the step in which the final product takes its physical shape. Before being fed to forming machines the glass must be reduced to a suitable working temperature (conditioning in the working end or distributing head). The melting end is connected to the working end by a narrow passage called throat which form a physical separation between these two units in container tanks. Due to temperature gradients of the melt, the temperature of the glass flowing from the melting end to the working end depends on the depth and position of the throat. Moreover, throats are designed in order to minimize the formation of return glass currents from the working end to the melting end (Trier 1987). The forming process varies widely, depending on the type of glass being manufactured. In the next sections the main forming processes of SLS glass for the production of glass containers and flat glass are described. Container glass. The molten glass flows through the feeder for conditioning at about 1300 °C and is then forced through small holes, cut into gobs using mechanical shears and then transferred into container-shaped molds at about 1100 °C (gob feeding). Today, nearly all container manufacturers use the so-called individual section (IS) machine to automatically form containers from gobs. This machine is capable of handling a variety of mold types and sizes and it can produce more than 100 containers per minute. Two forming modes are commonly used in IS machines: the blow-and-blow and the press-and-blow. In the first mode, the gob is first transferred to a blank mold (parison mold) and settled using compressed air (settle blow). A counter blow is used to create a parison. The parison is then inverted and transferred to a second blow mold where its final shape is formed (finishing mold). The pressand-blow method uses a plunger to form the initial parison into a blank mold. The parison is then inverted and transferred into the blow mold where it undergoes final shaping in a vacuum or by air pressure. After forming the containers are transferred to post-forming processes such as hot-end coating, annealing and cold-end coating. The annealing process is designed to eliminate or limit internal stresses due to rapid cooling (the surfaces of molten glass cool more rapidly than in the center) which may cause glass objects to crack, shatter or even explode some later time. Containers are subject to strictly controlled cooling in a continuous-belt lehr (a special type of oven used specifically for glass annealing). Inside the lehr, the glass is allowed to cool to a temperature known as the annealing point (500-550 °C). When the glass reaches this point, the lehr temperature is stabilized for a specific length of time (30-90 min) to allow stresses present in the glass to relax. This phase is followed by a period of cooling with a pre-defined temperature gradient. A hot-end coating (thin tin oxide layer) and a cold-end coating (organic lubricants) are applied on the container surface shortly after forming and upon exiting the annealing lehr respectively; the former acts as an adhesive for the latter. These coatings are very effective in protecting the glass surface from mechanical damage due to sliding contacts on production, inspection, and high speed filling lines as well as during transport (Geotti-Bianchini et al. 1994). Finally, the end-products are sent to quality control and packing. Flat glass. The FLOAT glass process developed in the 1950s by Pilkington Brothers is used worldwide to produce almost all flat glass. In this process, molten glass flows horizontally from the forehearth onto a pool of molten tin, at a temperature of about 1050-1100 °C. The tin bath is kept in a molten state by electrical heating under a blanket of inert gas (i.e., nitrogen and hydrogen) to prevent the oxidation of tin. The molten glass passes over the molten tin, floating over it and conforming to the perfect flatness of the tin surface. As it undergoes this process, it develops a uniform thickness without distortion and perfectly fire-polished surfaces. At the
Falcone, Ceola, Daneo, Maurina
124
end of the pool, the continuous glass ribbon (about 600 °C) is transferred to the annealing lehr and to other finishing processes (cutting, metallic coating deposition, strengthening, etc.). In some cases a pyrolitic coating (oxide/s coating) is deposited before the annealing process.
opTICAl pRopeRTIeS AnD ColoRS oF InDuSTRIAl SlS ConTAIneR GlASS The color of a glass is that of the light either transmitted or reflected by the glass. The former is the more important. In this paper we refer to the term glass color as to the transmitted light color. The most common colors for SLS glass containers are: colorless, half white, emerald green, yellow-green (commercially known as UVAG, ultra-violet-absorbing-glass) and amber. The choice of the container’s color by food and beverage producers mainly depends on two requirements, food preservation (i.e., protection from light radiation for photosensitive foods or drinks) and marketing (i.e., visibility of the contained product and association of a typical container color to a specific product). Colorless glass containers (white and half white) promote the visibility of the content but provide poor protection for the foodstuff that it contains, whereas colored glass containers (mainly UVAG and amber) provide the highest level of protection for photosensitive foods and beverages (Jacobsson and Högberg 1947; Mastrobattista 1990; Locardi 1992). The color of the light transmitted by a glass depends upon three elements, i.e., the spectral energy distribution of the incident light, the interaction of the light with glass and the interaction of the transmitted light with the eye of the observer. Therefore, any method for the quantitative description of glass color must take into account all these three elements. In order to standardize color quantification, the Commission Internationale de l’Eclairage (CIE) defined some standard illuminants and colorimetric observers. One of the most used standard illuminant (illuminant C) is a close approximation to average daylight whereas the standard observers are based on the standard curve of human eye sensitivity (see Commission Internationale de l’Eclairage 2004 technical report CIE 15:2004 “Colorimetry”). Glass colors can be represented in terms of color coordinates, similar to colors of other materials, i.e., by points plotted in the colorimetric space introduced in 1931 by the (CIE). These coordinates are based on X, Y and Z tristimulus values. The tristimulus values are a mathematical tool created to represent the primary stimuli red, green and blue of a given light color. In fact, the light of any color can be matched to give its equivalent by a proper mixture of red, green and blue lights. In order to represent colors in the CIE colorimetric space, the color coordinates (or chromatricity coordinates) are defined as follows: = x
X Y Z = , y = , z X+Y+Z X+Y+Z X+Y+Z
(8)
These quantities provide the relative amounts of red, green and blue light, respectively, that would simulate the glass color and are designed so that x + y + z = 1. Therefore the chromaticity (color quality) of a sample can be represented by only two coordinates (x and y) in the CIE space (Bamford 1977; Paul 1990; Polato and Daneo 1991). Colorimetric coordinates in CIE space are determined by means of spectrophotometric measurements using standard illuminants and observers (e.g., illuminant C and observer CIE 1931). Nevertheless the Colorimetric CIE space is not isometric, that is: equal distances between two points (colors) in different parts of the diagram do not correspond to equal color differences as perceived by the human eye. Therefore, maximum acceptable color difference cannot be established exactly and, over the past few years, this has led to controversies among producers and customers. On account of this situation, the CIE recommends the CIE L*a*b* 1976 space. In this color space the cylindrical coordinates L*, a* and b* are derived from X, Y and Z tristimulus values of the color of the sample, obtained by applying weighing functions to the range of visible spectral transmittance curves measured by spectrophotometry. The a* axis varies from green
Sulfur Compounds in Industrial Glass
125
to red and the b* axis from blue to yellow. The L* axis is perpendicular to the (a*, b*) plane and its coordinates represent the lightness, i.e., the level of grey between black (complete absorption of incident light) and white (total colorlessness). The CIE L*a*b* 1976 space is an almost uniform color space where distances between points in the diagram are proportional to corresponding color differences, thus enabling a quantitative evaluation of color differences that correlates well with visual judgments (Polato et al. 1989). A detailed description of this approach can be found in Wiszecki and Stiles (1982). In Figure 4 the most common industrial glass container colors are plotted in the CIE a*b* plane (data from SSV).
Figure 4. Commercial container glass colors plotted in the CIE a*b* 1976 diagram (data from SSV).
Spectrophotometric measurements can also be used to characterize the optical properties of the color of glass containers: luminance, dominant wavelength and colorimetric purity. These quantities are necessary for the complete and unambiguous quantitative evaluation of a glass color. Luminance is the transparency of glass and it corresponds to the fraction of light that is transmitted through the glass and perceived by the human eye. It is obtained by correcting the spectral Transmittance T(l), defined as the ratio of the light radiation transmitted through the glass (It) to the total incident light radiation I0, with the emission curve of the CIE standard illuminant C and with the CIE standard colorimetric observer CIE 1931. Based on this, the quantity of luminance represents, as much as possible, the real physiological human perception of transmitted light. The Dominant wavelength describes a polychromatic light in terms of monochromatic radiation that evokes an identical perception of hue. Colorimetric purity is the percentage of the dominant wavelength on the color; it describes the color’s degree of paleness and corresponds to perceptual quantity “saturation” (CIE 15:2004). Due to the lack of an international reference for the quantification of the degree of protection needed for foodstuff and beverages, Filtering capacity (FC) has been defined by SSV as practical way to evaluate the ability of a type of glass to block dangerous radiation in a selected spectral range (usually 350-450 nm). FC is the difference from 100 the average value of transmittance measured in the range 350-450 with steps of 10 nm each. In the range 350-450 nm, it is calculated as follows (Polato et al. 1988):
Falcone, Ceola, Daneo, Maurina
126
λ = 450
FC = 100 −
∑ T (λ )
λ = 350
( 9)
11
where l = 350, 360, … 450 nm and T is the spectral transmittance at wavelength l. Table 2 shows typical optical property values and belonging to several types of soda-lime glass. The filtering capacity values show the high degree of guaranteed protection provided by amber and yellow-green glasses. On the other hand, the luminance values show that transparency (and therefore visibility of the contained product) is reduced to one third from colorless to amber glasses. Table 2. Typical optical properties values of SLS container glasses (from Daneo et al. 2006). Filtering capacity (%)
Color
Dominant wavelength (nm)
luminance (%)
(290-350 nm)
(350-450 nm)
Colorless Half white Emerald green Yellow-green - UVAG Amber
576 505 553 566 582
90 86 56 43 30
60 79 82 95 100
12 16 53 87 100
ColoR GeneRATIon In SlS GlASSeS In industrial SLS glasses, color is produced by the selective absorption of radiation in the visible range by transition metal ions (e.g., iron, nickel, cobalt, chromium, copper, manganese, other) dissolved in the glass during the melting stage (Bamford 1977). These coloring oxides may be introduced as undesired impurities within the raw materials (as it’s the case of iron in colorless glasses) or may be added to the batch as oxides in small quantities in order to produce the desired glass color. Figure 5 shows typical spectral transmittance curves in the visible range of commercial glass samples obtained through spectrophotometry of 3-mm thick samples. The absorption spectra due to transition metal ions in glasses have been interpreted by the ligand field theory (Cotton et al. 1995). This theory provides a comprehensive explanation of transition metal complex spectra in glasses as well as crystal structures and other chemical compounds. The absorption bands correspond to electronic transitions between energy levels in the partly filled 3d shell. The intensities of bands in glass are governed by the selection rules for the electron transitions and vary significantly from one transition metal ion to another. Low or high absorption intensities usually derive from formally spin-forbidden or spinallowed transitions, respectively (Bamford 1977). The electronic transitions responsible for the spectral absorption can be identified based on the valence state of the transition metal, the field symmetry and coordination (tetrahedral or octahedral) to the surrounding negatively charged ions (the ligands, oxygens in glasses) (Guloyan 2007). A detailed description of the theory applied to colored glass is reported in Paul (1990). The specific character of the various coloring agents depends on the glass composition. For a given glass composition the intensity of absorption by a coloring agent at a given wavelength is described by the Lambert-Beer’s law (Paul 1990; Scholze 1991): I t = I 0 ⋅ 10 − ( e⋅c ⋅d )
(10)
Sulfur Compounds in Industrial Glass
127 Colorless Half wHite emerald green
Transmission (%)
uvag amber
Wavelength (nm) Figure 5. Spectral transmission curves of different glass types, thickness = 3 mm (from Daneo et al. 2006).
where It and I0 are the transmitted and incident light intensities, respectively, c is the concentration of the coloring agent, and e is a wavelength dependent proportionality constant called extinction coefficient (usually expressed in L·mol−1·cm−1). At equal concentration and glass thickness, a coloring ion giving a higher extinction coefficient produces a stronger light absorption and consequently, a stronger coloring effect if the absorption band has effects in the range of visible light radiation. The value of the absorption coefficient mainly depends on the probability of the electronic transition. The unitless and wavelength dependent quantity absorbance D=
( e·c·d )
(11)
is called optical density in glass technology and can be measured experimentally by means of absorption spectrophotometry by the following: I D = log 0 It
(12)
where I0 and It are the intensities of incident and transmitted light respectively. By measuring D on a glass sample of a given base composition with known thickness and coloring ion concentration, it is possible to calculate the absorption coefficient e from Equation (11). Table 3 combines experimental data and assignments from the ligand field theory (Bamford 1977) and lists the most common SLS glass transition metal coloring ions, their coordination, absorption wavelengths, the corresponding optical densities (per cm path length per percentage by weight oxide), and the colors produced in glasses. The significant differences in the optical densities of ferric and ferrous iron should be noted as well as the high values of D for Cr3+ absorption bands. In Table 3 the absorption at 290 nm and 410 nm is also reported, the latter providing the amber color in SLS glass. The amber color in SLS glasses is produced by the simultaneous presence of ferric iron and sulfides. Based on spectroscopic and analytical results, Douglas and Zaman (1969) proposed a reasonable model for the amber chromophore made up of a special complex in which a ferric iron is tetrahedrally coordinated by three oxygen molecules and one
Falcone, Ceola, Daneo, Maurina
128
Table 3. Most common SLS glass coloring ions (from Bamford 1977). Wavelength (nm)
optical density
Color in SlS glass
tetrahedral
380 420 435
1.27 0.35 0.34
yellow
Fe2+
octahedral
1050
9.1
green-blue
3+
Cr
octahedral
450 650
6.4 5.9
emerald green
Cu2+
octahedral
780
3.0
turquoise
Mn3+
octahedral
490
4.0
violet
Co2+
tetrahedral
530 590 645
31.8 47.7 45.3
blue
Fe3+/S2-
tetrahedral
290 410
Ion
Coordination
Fe3+
amber
sulfide, along with the appropriate number of alkali to neutralize the charge on the complex group. The optical absorption produced by this chromophore originates in electron transfer between the sulfide and ferric ions, rather than in electronic transitions of a transition metal and the extinction coefficients are a 100× higher than those for transition metal absorption (Harding and Ryder 1970; Bamford 1977; Paul 1990). Since polyvalent coloring ions produce different glass colors at different oxidation states, the redox state of the melt/glass has a great significance for the final color of the glass and its constancy is important for glass color stability. Colorless glass requires high-purity raw materials (low iron and very low chromium content) and strongly oxidizing melting conditions to avoid the strong coloring effect of Fe2+. The ferrous iron absorption band is centered at 1050 nm, well outside the range of visible light but, due to the breadth of the band and its high extinction coefficient, it also produces absorption in part of the visible range. This effect results in an intense green-blue coloration, which is perceived as about three time stronger than the yellow color produced by an equivalent amount of Fe3+ (Bamford 1977). By contrast, amber and yellow-green (UVAG) glasses require reducing conditions for the formation and development of the amber chromophore. The glass redox state in the final product is characterized by measuring the Fe2+/Fetot and S2−/Stot ratios with appropriate techniques (see below). Interaction of sulfur species with heterovalent cations, i.e., ferrous and ferric iron, has a strong influence on the color of glasses and therefore, information on sulfur solubility and speciation is needed, which is summarized in the following section.
SulFuR SoluBIlITy, ReDox AnD GlASS ColoR The solubility of a gas in the glass melt can be driven by two processes: physical solubility and chemical solubility. In the first case, gas molecules occupy the free space available in the interstice of the silicon oxide network. The second process arises when the gas molecule chemically reacts with the network components, becoming part of the glass structure (Kramer 2005). The topic of this paragraph is discussed extensively in Backnaes and Deubener (2011, this volume) and only a short, but necessary, description is reported here in order to link sulfur solubility to industrial glass color.
Sulfur Compounds in Industrial Glass
129
When sulfur dissolves chemically in the glass melt, its solubility depends on composition, and redox state of the melt, temperature, and melting process duration. The redox state depends on partial oxygen pressure and can be defined as the ratio between the ferrous iron (Fe2+) and the total iron (Fetot) in the glass. For reduced glass, the ratio between sulfides and the total amount of sulfur is also used as an indicator of the redox state of the glass.
The batch redox number To define the final redox (and, consequently the color) of a glass, proper amounts of oxidizing and reducing raw materials are added to the batch. The solubility of sulfur in the glass melt and the Fe2+/Fetot ratio are strongly dependent on the relative amounts of these raw materials as well as on the redox state of the glass cullet used in the batch. Glass technologists use an empirical approach to predict the effectiveness of these raw materials in the glassmelting process and to define the theoretical oxidation-reduction state of a glass batch. This method was first proposed by Manring and Hopkins (1958) and then revised by Simpson and Myers (1978). Based on both practical experience and theoretical equations, redox values have been assigned to commonly used oxidizing and reducing raw materials. These values are positive for oxidizing agents and negative for reducing agents and are calculated based on 1 kg of raw material per 2000 kg of sand in a glass batch (Simpson and Myers 1978; Daneo et al. 2006). In Table 4 redox values of the most common oxidizing and reducing raw materials are listed. Using these factors, a redox number is calculated for a glass batch by multiplying the redox value of each raw material by its weight in kg in reference to 2000 kg of silica sand. Afterwards, the values are algebraically summed in order to obtain the batch’s redox number. The main natural raw materials such as silica sand, limestone, dolostone etc., usually contain impurities such as carbon, iron oxides, sulfides and sulfates that can influence the final batch redox number, due to the high amounts of these materials in the batch. Redox values pertaining to these raw materials are therefore calculated by determining their chemical oxygen demand, Table 4. Redox values of the most common oxidizing and reducing raw materials (from Daneo et al. 2006). oxidizing Raw material Manganese oxide (IV) (MnO2) Arsenic oxide (V) (As2O5) Calcium sulphate (CaSO4) Sodium sulfate (salt cake) (Na2SO4) Potassium bichromate K2Cr2O7 Gypsum (CaSO4.2H2O) Baryte (BaSO4) Sodiun nitrate (NaNO3) Haematite (III) (Fe2O3) Magnetite (Fe3O4)
Reducing Redox value +1.09 +0.93 +0.70 +0.67 +0.65 +0.56 +0.40 +0.32 +0.25 +0.19
Raw material Carbon (100% C) Coke (85% C) Carbocite (65% C) Pyrite (FeS2) Pyrrothite (FeS) Ferrochromite (FeCrO3) Arsenic oxide (III) (As2O3) Slag
Redox value −6.70 −5.70 −4.36 −1.20 −1.60 −1.00 −0.93 −0.14
Falcone, Ceola, Daneo, Maurina
130
COD (Peters 1993) and transforming them into equivalent carbon by the simple stoichiometry: C + O2 → CO2
(13)
The equivalent carbon is then used in the batch redox number calculation with the redox value −6.7. The same procedure is applied to determine the redox value of glass cullet. Table 5 shows typical ranges of batch redox numbers for different industrial SLS glass container colors. Table 6 illustrates an example of batch redox number calculation for an amber glass batch composition. It is important to stress that optimum redox numbers may vary with the type of furnace, the fuel used and the general melting conditions. In fact, the batch redox number cannot account for either all of the possible melting condition variations or for the glass production rate (pull rate) in industrial plants that can lead to different redox states in the end product. Nevertheless, once established, it can be used as a reference value for that particular melting unit. In order to monitor the real redox state of the glass melt during the melting process, inline oxygen sensors have been developed. These devices enable the detection of uncontrolled redox variations at an early stage and a faster correction of the batch redox number, particularly when high percentages of recycled cullet are used in the glass batch (Laimböck 2008; MüllerSimon 2011, this volume). Table 5. Typical batch redox number ranges for different types of container glass (from Daneo et al. 2006) Glass color
Batch redox number range
Colorless Half white Emerald green Yellow green UVAG Amber
+15 / +7 +5 / −3 −3 / −10 −16 / −20 −20 / −30
Table 6. Example of a batch redox number calculation for a cullet free amber glass batch. Raw material
Weight (kg)
CoD (mg oxygen / g raw material)
equivalent C per kg of raw material (kg)
Silica sand
2000
0.15
0.11
Soda-ash
600
Limestone
330
0.25
0.03
Dolostone
300
0.25
0.03
Equivalent C total
0.17
Redox value
Redox number
−6.7
−1.1
Pyrite
6
−1.2
− 7.2
Graphite (85% C)
0.7
−5.7
−4.0
Slag
80
−0.14
−11.2
Batch redox number
−23.5
Sulfur Compounds in Industrial Glass
131
experimental melting In order to investigate sulfur behavior in SLS melting experiments at different redox conditions have been performed (Barbon et al. 1991). In order to define real sulfur solubility in a glass melt, an equilibrium state between the melt and the gas phase is required. In melting experiments, an equilibrium state is rarely reached and therefore, sulfur behavior in the glass melt would be more correctly described by the amount of sulfur retained by the glass, usually expressed as weight % of SO3,tot. The amount of retained sulfur depends largely on the experimental parameters. Nevertheless, at constant experimental conditions, it can be used as an approximate indicator of sulfur solubility in glass. Figure 6 shows the variation of retained sulfur concentration (expressed as wt% SO3,tot) and the % of Fe2+ of Fetot versus the batch redox number, for a SLS glass composition. The curve was obtained experimentally by adding carbon to the same SLS batch glass melted at 1350 °C. The corresponding glass contained 0.33 wt% of total iron (expressed as Fe2O3), without any other coloring element (Barbon et al. 1991; Maurina et al. 2008). In the Figure, the color of the glass produced by iron at different redox values is also reported. Under oxidizing conditions, (redox number +15 to +10 and Fe2+/Fetot < 30%), sulfur is soluble largely as sulfate (S6+) (Wilke et al. 2011, this volume). Iron is present mostly as Fe3+ which gives the glass a pale yellowish color. At more reduced conditions, i.e., higher contents of ferrous iron, the retained sulfur decreases progressively. Initially Fe2+ increases moderately; after entering the region of negative redox numbers the abundance of ferrous iron increases dramatically. Glass color changes from yellow to green (30-50% Fe2+) to blue (approx. 60% Fe2+) due to the progressive increase of Fe2+. In correspondence of a batch redox number −15 (Fe2+ = 65-70%) sulfur begins to dissolve as sulfide (S2−). The corresponding glass color suddenly turns from blue to amber-yellow, due to the formation of the amber chromophore Fe3+/S2-, turning dark brown at extremely reduced conditions. At a batch redox number around −20, Fe2+correspond to about 80% of total iron, whereas retained sulfur reaches its minimum concentration (minimum of solubility). For redox numbers lower than −20 (industrial amber color region) we observe a significant increase of retained sulfur in the glass in the form of sulfides and a moderate increase of Fe2+ (80-85 %).
Figure 6. Total retained sulfur solubility versus Fe2+/Fetot ratio (in %) for a soda-lime-silica glass batch after experiments at 1350 °C in which various amounts of carbon were added to soda-lime silicate melts (data from Barbon et al. 1991).
132
Falcone, Ceola, Daneo, Maurina
The effect of temperature on the glass redox and sulfur solubility (Backnaes and Deubener 2011, this volume) also has to be considered. Dubois et al. (1990) report that if the melting temperature for oxidized glasses increases, glasses with a higher ferrous content and lower retained sulfur are obtained. They concluded that a temperature increase lowers the sulfate solubility and enhances the thermal decomposition of ferric iron. Barbon et al. (1991) investigated the effect of temperature on the redox state of the resulting glasses for batch redox numbers ranging from +30 to −70. By melting the same batch at different temperatures (1350 °C and 1500 °C), an increase of the retained sulfur (in the form of sulfide) was observed in the glass obtained with batch redox numbers lower than −20. By contrast, in the oxidized range a higher melting temperature resulted in lower retained sulfur content due to reduced sulfate solubility. Moreover, the minimum of sulfur solubility was attained at lower redox number. A shift of the Fe2+ to Fe3+ curve to higher redox numbers was also observed, in agreement with Dubois et al. (1990).
CheMICAl ChARACTeRIzATIon oF GlASS ColoRS The concentrations and redox state of coloring polyvalent elements in the end product are routinely analyzed in industrial and research glass laboratories in order to monitor the quality and stability (constancy) of glass color. The concentration of elements influencing glass color can be determined by traditional wet chemistry methods and instrumental analytical techniques (Guadagnino and Corumluoglu 1997; Corumluoglu and Guadagnino 1999; Ripley et al. 2011, this volume). Nevertheless most of these techniques are expensive and time consuming. In the field of industrial glass rapid and frequent analyses are required to strictly monitor the production process. Glass compositions are therefore usually determined by quantitative Wavelength Dispersive X-ray Fluorescence Spectrometry (WDXRF) which provides reliable, accurate, fast and low-cost analysis. Severe limitations occur on sulfur analysis in glass samples by WDXRF due to sample preparation as glass beads (loss of sulfur during fusion with lithium tetra- or meta-borate) or pressed pellets (poor correlation of X-ray intensities to standard sulfur concentrations), as described by Ripley et al. (2011, this volume). Consequently, WDXRF analysis of glass samples at SSV are routinely performed on the flat glass surface polished with cerium oxide fine powder (without any other type of preparation such as grinding, fluxing, etc.) using a sequential X-ray spectrometer calibrated according to international glass standards (NBS, SGT, etc.). This procedure enables the quantitative determination of all glass components including total iron, total sulfur and chromium (conventionally expressed as SO3,tot, Fe2O3,tot and Cr2O3, respectively) even at very low concentrations on the order of tens of ppm. Ferrous oxide (FeO) and sulfide (S2−) concentrations in SLS glass samples can be rapidly determined by means of visible-near infrared range (Vis-NIR) spectrophotometry FeO can be determined by measuring the absorption band centered at 1050 nm. Sulfides can be determined by measuring the intensity of the amber chromophore optical absorption (Bamford and Hudson 1965; Cable and Hulme 1985; Boheme et al. 1991). A good correlation between FeO concentrations determined by wet chemistry and the absorption at 1050 nm was obtained by using a base line joining two points of the spectral curve at 770 and 1500 nm, where the absorption coefficient of Fe2+ is practically constant and the interference of amber and chromium chromophores is negligible (Daneo et al. 1992). For sulfides determination the amber absorption is measured at 550 nm, where the interference due to the absorption band of Cr3+ is low and can be corrected according to the equation: [Fe3 + ][S2 − ] = a × A550 − b × [Cr2 O3 ] − c
(14)
where the coefficients a, b and c are > 0 and have been determined by the statistical treatment of
Sulfur Compounds in Industrial Glass
133
a large number of reference samples analyzed by wet chemistry or other analytical techniques to determine the concentrations of iron and chromium. These coefficients are not absolute but differ depending on the instrument type and on the analytical methods for the chemical analysis. The spectrophotometric determination of sulfides is an indirect measurement since only the sulfides bond in the amber complex are determined. Nevertheless, Daneo et al. (1992) demonstrated that the corrected 550 nm absorption is proportional to the total sulfides content in the range of industrial hollow glass compositions, by comparing wet chemistry and spectrophotometric data. Table 7 lists typical wt% concentration ranges of Fe2O3,tot, SO3,tot and Cr2O3 together with the Fe2+/Fetot and S2−/Stot ratios (expressed in %) of different containers glass (extra-colorless, colorless, half white, emerald green, yellow-green UVAG and amber glass) produced by Italian manufacturers (Daneo et al. 2009). In white glass, Fe2O3tot, present as an undesired impurity from raw materials, ranges between 0.015 wt% (extra white glass) and 0.03-0.07 wt%. Fe2+ is lower than 30% of total iron; total retained sulfur (expressed as SO3,tot) is higher than 0.20 wt%; Cr2O3 must be lower than 0.0003 wt% (3 ppm). In half-white glass, the Fe2O3,tot and SO3,tot concentrations range between 0.1 wt% and 0.2 wt% and Fe2+ ranges between 30% and 40% of total iron. In emerald green glass, the color is mainly provided by about 0.2 wt% of Cr2O3, that is added to the batch as chromite ((Fe,Mg)Cr2O4). In this type of glass, Fe2O3,tot concentration ranges between 0.4 and 0.6 wt% (iron is introduced in the batch mainly as hematite) and the ferrous iron usually ranges between 45% and 55% of total iron. The yellow-green (UVAG) and amber glasses are characterized by high iron contents (Fe2O3,tot > 0.5 wt% and > 0.3 wt% respectively); SO3,tot is typically about 0.02 wt% for yellow-green glass (minimum sulfur retention) and lower than 0.06 wt% for amber glass. These types of glass have high Fe2+ content (>70% - 80% of Fetot) but still contain sufficient amounts of Fe3+ that ensure the formation of the amber chromophore. Yellow-green color of UVAG glass is produced by the simultaneous presence of amber chromophore and trivalent chromium (Cr2O3 about 0.1 wt%). The latter provides the green hue. Table 7. Ranges of coloring ions concentrations (in wt% of oxide) and redox pairs ratios (in %) in SLS container glasses produced in Italy (data from SSV). Glass type
Fe2o3,tot
So3,tot
Cr2o3
Fe2+/Fetot
S2–/Stot
Colorless Half white Emerald green Yellow green UVAG Amber
0.015-0.07 0.1-0.2 0.4-0.6 0.4-1.0 0.3-0.5
0.20-0.30 0.08-0.20 0.04-0.05 0.01-0.03 0.04-0.06
— traces 0.15 0.25 0.06 0.13 0.03 0.06
1 -30 30 40 43 56 70 80 74 82
— — — 75 90 85 95
Figure 7 shows typical ranges of total retained sulfur concentration (expressed as SO3,tot) of different types of container glass versus their Fe2+/Fetot. The general trend of the data mirrors the retained sulfur curve in Figure 6. The total amount of retained sulfur decreases progressively from oxidized colorless glasses to reduced yellow-green and then increases going toward highly reduced amber glass. Sulfur speciation in industrial SLS glass melt has recently been investigated by XANES spectroscopy. Bingham et al. (2010) reported the presence of intermediate S4+ and S5+, in addition to main sulfur species S6+ and S2−, primarily in yellow-green (UVAG) glasses produced under moderately reduced conditions that coincide with minimum sulfur solubility. However, the existence of intermediate redox states of sulfur between sulfide and sulfate is still
134
Falcone, Ceola, Daneo, Maurina
Figure 7. Typical ranges of total retained sulfur and Fe2+/Fetot ratio for different container glass types produced in Italy (data from SSV).
controversial. Other studies show evidence that such XANES features can be formed as beam damage (Wilke et al. 2008; 2011, this volume).
BuBBleS In SlS GlASSeS Bubbles in soda-lime container glass are caused by different phenomena linked to the different stages of the production process. In particular, bubbles are the main defects of SLS glass products that may derive from uncontrolled variations of retained sulfur. These defects can be caused by a local event that changes the uniformity of temperature or composition in the glass. Gaseous inclusions represent a serious aesthetic problem, especially for high quality glass production. The presence of visible bubbles in high quality cosmetic and perfumery containers causes the rejection of the product by the customer. Bubbles on the bore (mouth) of a container might cause mechanical problems in the cork process and large bubbles on the container bodies might lead to holes or even glass fragmentation inside the container and contamination of the contained product. Therefore, knowledge of the composition of gas bubbles is fundamental for determining the origin of the defect and undertaking necessary corrective operations in the glass production line (Backhausen et al. 2004). Gas compositions in bubbles can be analyzed using a mass spectrometer with a quadrupole analyzer (Greim et al. 1981). A small fragment of glass with a bubble is placed in the breaking chamber. Inside this chamber, the air is pumped out by means of turbo-molecular pumps, until reaching an internal pressure of 10−12 to 10−13 bar. At this vacuum level, the bubble is broken by a plunger and the gases inside reach the ionization chamber through the transfer line. The ionization chamber creates the ions which, after being accelerated by an appropriate device, simultaneously reach the quadrupole analyzer (dynamic analysis). High vacuum levels are necessary because of the small size of the bubbles usually analyzed. Since gas pressure in the bubbles at room temperature range from 0.2 to 0.3 bar, it is necessary to have as much gas as possible reaching the analyzer. All the surfaces in contact with the gas are gold plated in order to limit the absorption of the gas species on the internal walls and obtain low detection limit. As an example the detection limit of GIA522 Mass spectrometer produced by IPI is approximately 5×10−12 bar·l (which corresponds to a bubble with a diameter of 0.1 mm with an internal pressure of 10−2 bar). In order to promote bubble breakage by means of a plunger, two opposite converging cuts near the bubble are made on
Sulfur Compounds in Industrial Glass
135
the sample. This preparation also reduces the fresh fractured surface that could absorb part of the gas after breaking. Gaseous inclusions containing sulfur are mainly caused by three different issues: incomplete fining, redox instability and reboil.
Incomplete fining The fining process depends on the maximum temperature reached by the melt in the hotspot, and the time that the melt stays at this temperature. The fining process is insufficient or incomplete when, due to an uneven balance of the time-temperature choice, the molten glass still contains medium-sized bubbles (up to 1 mm) after leaving the fining stage. The time the melt stays in the furnace is influenced by the pull rate (amount of glass extracted from the furnace); time is shorter as the pull rate is increased. This is a common issue at plants whose economic budget is strictly related to final product quantity generated per day, and where the presence of some defects is not that critical (containers for low quality products). The incomplete decomposition of the fining agents lead to the incomplete reaction and release of SO2 and O2 gases and eventually leads to the presence of SO2 bubbles in which the main gas, in oxidized glasses, may be condensated and other gases dissolved in the melt, such as CO2, may diffuse into the cavity (Golob and Lowell Swarts 1984). The temperature in the working end is strongly influenced by the quantity and type of raw materials in the furnaces. For instance, the higher the fraction of glass cullet, the lower the heat that is required to melt the batch material, or at given heat, the temperature in the glass throat needs to be higher (Conradt et al. 1997). If the amount of cullet is not kept strictly under control, temperature fluctuations may arise during initial melting stages. These fluctuations can cause a displacement of the hot-spot of the furnace, and eventually lead to incomplete fining. As a consequence sulfates may evolve completely into SO2 and O2 before the complete elimination of CO2 from the melt has occurred. Thus CO2 bubbles in the molten glass persist after it has passed the throat and cannot be eliminated.
Redox During the fusion process, the redox state must be kept under control; otherwise, uncontrolled variations may lead to the formation of SO2-bearing bubbles. This phenomenon can derive result from drastic redox gradient, where oxidized glass undergoes a sudden reduction, or when reduced glass suffers fast oxidation. In both cases, sulfur oversaturation can occur in the glass melt, followed by the formation of SO2-bearing bubbles. Redox variations arise where two glass melts with different redox states come into contact. The two most common causes are the use of different-colored cullets or residual glass from a former production run, after color change. Moreover, in a particular production (tableware, perfumery, etc.), glass is colored by adding a frit (i.e., a colored low temperature melting glass) to the colorless glass in the feeder. The mixing of these two different glasses may lead to the formation of sulfur-containing bubbles. Sometimes, a burner with too much or too little oxidant flame might change the redox state of the glass melt surface giving rise to sulfur bubbles.
Reboil The term “reboil” refers to the sudden formation of bubbles after the fining phase as a consequence of a local temperature increase, due to an ill-regulated burner over the glass melt, or to malfunctioning of the thermo-regulation system. This temperature increase causes a local reduction of sulfur solubility and, in the case of oversaturation, the formation of SO2-bearing bubbles as main gas component. If this phenomenon happens in the proximity of the working end, the glass melt will not have enough time to re-dissolve the SO2 gas. In this case, the bubbles are limited to specific production intervals and a single production line. Table 8 reports typical compositional ranges of bubbles in SLS container glasses due to incomplete fining, redox and reboil.
Falcone, Ceola, Daneo, Maurina
136
Table 8. Typical composition ranges (in vol%) of bubbles in SLS container glasses due to incomplete fining, redox and reboil (from Maurina et al. 2008). Gas
Fining
Redox
Reboil
SO2
traces - 78.2%
87.1-95.3%
96.9-97.9%
N2
0.4-7.1%
2.8-7.9%
1.1-1.5%
CO2
21.2-91.7%
1.7-4.4%
1.0-1.5%
H2S
0 - traces
traces - 0.2%
—
COS
0-1.1%
—
—
Ar
—
0 - traces
—
Pressure
10-100 mbar
80-120 mbar
80-120 mbar
Deposits
S0 - sulfate
S0 - sulfate
S0 - sulfate
Deposits The gases contained in sulfur-bearing bubbles, may react with the surrounding glass forming sulfur and sulfate deposits before reaching the working end. SO2 in the bubble decomposes into SO3 and S: 3SO2 → 2SO3 + S
(15)
SO3 reacts with Na2O in the glass melt, forming Na2SO4 deposits on the internal bubble surface: SO3 + Na2O → Na2SO4
(16)
The gas pressure inside the bubble is reduced because of this mechanism (Golob and Lowell Swarts 1984) and sometimes gas bubbles contain only residual gases and deposits. The lowering of the pressure inside the bubble drives the diffusion of the gases from melt to the bubble, in order to re-establish the equilibrium between gas and melt. Elemental sulfur and sulfates solid deposits can be analyzed by Optical Microscope (OM) and Scanning Electron Microscope and Energy Dispersive X-ray microanalysis (SEM-EDX). Observed under the optical microscope, sulfate deposits often display an irregular shape, which, in some cases, has a peculiar dendritic structure (snowflake structure). Figure 8 shows sulfate deposits and small spheres of elemental sulfur condensation deposited on the internal surface of a bubble in a SLS container glass. The identification of sulfur condensates can be confirmed by heating the bubble and observing the effect of increasing temperature on the deposits. Sulfur condensates melt at about 120 °C and changes color at about 145 °C. These changes are reversible upon cooling (Fowkes and Parton 1969). Deposit morphologies can be observed at high magnification using SEM-EDX analysis, which also provides information about their chemical composition. However, there are some technical difficulties. Sample preparation is quite difficult, because the bubble must be opened without spoiling or losing the deposit inside the internal bubble surface. Moreover, the concavity of the internal surface is not suitable for SEM-EDX analysis, which is best performed on flat surfaces. Figure 9 shows the SEM secondary electron images of the internal surface of a bubble with sodium sulfate deposits are shown at different magnifications. Due to the small quantity of deposits present and the concavity of the sample, only qualitative elemental microanalysis could be performed on this sample. Nevertheless, the analyses gave clear evidence for the chemical nature of the deposit (Maurina et al. 2008).
Sulfur Compounds in Industrial Glass
137
Figure 8. Detail of bubble with sulfur and sulfates deposits (optical stereo microscope).
Figure 9. SEM secondary electron images of the internal surface of a bubble with sulfates deposit at low (above) and high (below) magnification.
138
Falcone, Ceola, Daneo, Maurina SuMMARy AnD ouTlooK
In the production of industrial SLS glass, sulfur compounds play an important role in determining the quality of the final product. These compounds actively participate in the fining process. Moreover they take a decisive role in the coloring mechanism of reduced glasses (yellow-green and amber) by producing the amber color. The absorption bands at 290 and 425 nm of this chromophore provide the highest level of protection to photosensitive foodstuffs and beverages. Glass container color quality is monitored by measuring optical properties and CIE L*a*b* 1976 coordinates. Moreover it is essential to monitor redox equilibria in order to ensure the right, stable and reproducible color quality. The batch redox number concept alone is not sufficient for ensuring the optimal quality of the end product. Therefore the analytical determination of the redox pairs ratios Fe2+/Fe3+, S2-/S6+ and Cr2O3 concentration are of critical importance in order to ensure stable and homogeneous final coloration, in particular for reduced glasses such as yellow-green UVAG and amber. For these types of glass, small optical property variations can be evident to the observer and, if they fall outside acceptance limits, the products may be contested or rejected by the customer. Furthermore, uncontrolled variations of retained sulfur in glass melts may cause sulfur bubbles which spoil the final product quality and can cause product rejection by customers. Therefore, the analytical characterization of the gases inside the bubbles is fundamental for determining their origin and undertaking necessary corrective operations Today, due to commercial reasons, glassmakers are increasingly requested to perform frequent glass color changes in the furnace. In industrial glass tank furnaces color changes are carried out by the so called changeover on the run, i.e., the progressive change of the batch composition while production operations continue (draining and refilling is theoretically possible only for small tanks for special glasses). During the transitions periods (which vary from several hours to several days, depending upon starting and target colors) usually no saleable glass is produced. Therefore, for economic reasons, the change must be performed as quickly as possible. In recent years, the amount of glass cullet in the batch has increased up to 80-90 wt% for particular glass types such as emerald green and UVAG yellow-green. This environmentally positive trend has several drawbacks mainly due variability of cullet compositions, and also in terms of redox affecting contaminants. Based on these assumptions future tasks in glass technology should include the development of procedures and analytical methods for the fast glass quality specifications, with particular concern on consistency, uniformity and reproducibility of glass color, in order to assist glassmakers particularly during the critical phase of color change. Moreover, the development of rapid methods for the characterization of glass cullet and the effect of high percentages of glass cullet in the batch on sulfur solubility, redox and glass color and bubbles formation should be studied, as well as considering the inevitable variations of industrial cullet compositions.
ACKnoWleDGMenTS The authors wish to thank prof. Harald Behrens from the University of Hannover for his careful revision work and assistance during the manuscript preparation and all the reviewers for their constructive comments and suggestions. Special thanks to Linda Falcone, Robert Linnen, Sharon Webb and Jim Webster for their English revision work and to Marta Vallotto for the drawing of Figure 1.
Sulfur Compounds in Industrial Glass
139
ReFeRenCeS Albayrak G, Sengel H (2008) Review of sulphate chemistry and its impact on glass production. Glass Technol Eur J Glass Sci Technol A 49:289-296 Arkosiovà M, Klouzek J, Nemec (2008) The role of sulfur in glass melting processes. Ceram Silik 52:155-159 Backhausen J, Muysenberg E, Ullrich J (2004) The importance of understanding the basic quality issues for improving the technology in glass melting. Int Glass J 129:31-36 Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Bamford CR (1977) Color Generation and Control in Glass. Elsevier, Amsterdam Bamford CR, Hudson EJ (1965) A spectrophotometric method for the determination of the ferrous/ferric ratio in soda-lime-silica glass. Proc of the 7th ICG Congress 1:1-11 Barbon F, Geotti-Bianchini F, Hreglich S, Scandellari L, Verità M (1991) Effect of the batch redox number and melting temperature on the redox equilibria in soda-lime industrial glass. Proc of the 1st ESG Conference Fundamentals of the glass manufacturing process, p 252-256 Barton J (2001) The evolution of the composition of industrial glasses: a historical perspective. Riv Staz Sper Vetro 2:3-10 Barton J, Guillemet C (2005) Le Verre – Science et Technologie. EDP Sciences, France Beerkens R (2005) Sulphate decomposition and sulphur chemistry in glass melting processes. Glass Technol Eur J Glass Sci Technol A 46:39-46 Beerkens R (2007) Sulphur chemistry and sulphate fining and foaming of glass melts. Glass Technol Eur J Glass Sci Technol A 48:41-52 Behrens H, Stelling J (2011) Diffusion and redox reactions of sulfur in silicate melts. Rev Mineral Geochem 73:79-111 Bingham PA, Connelly AJ, Hand RJ, Hyatt NC, Northrup PA, Alonso Mori R, Glatzel P, Kavcic M, Zitnik M, Bucar K, Edge R (2010) A multi-spectroscopic investigation of sulfur speciation in silicate glasses and slags. Glass Technol Eur J Glass Sci Technol A 51:63-80 Boheme W, Mueller HW, Liekmeier W (1991) A spectrometer concept for dedicated applications in UV-VIS spectroscopy. Int Lab Jan-Feb:18-24 Cable M, Hulme R (1985) Measurement of the ferrous:ferric ratio in flint glasses and its uses. Glass Technol 26:170-175 Chopinet M-H, Verità M, Falcone R, Lehuédé P, Vallotto M, Nardone M, Sodo A (2008) Soda-lime-silica glass containers: chemical durability and weathering products. Adv Mater Res 39-40:305-310 Clark-Monks C, Parker JM (1980) Stones and Cords in Glass. Society of Glass Tecnology, Sheffield Commission Internationale de l’Eclairage (2004) Colorimetry, 3rd Edition. Technical Report CIE 015:2004 Conradt R, Kham NS, Eiumnoh C, Pimkhaokham P (1997) Melting behavior of batches containing ground cullets. In: Proceedings of Fundamentals of Glass Science and Technology. Glafo, Växjö, Sweden. p 290-296 Corumluoglu O, Guadagnino E (1999) Determination of ferrous iron and total iron in glass by a colorimetric method. Glass Technol 40:24-28 Cotton FA, Wilkinson G, Gaus P (1995) Basic Inorganic Chemistry, 3rd ed. Wiley, New York. Chapter 23. Daneo A, Falcone R, Hreglich S (2006) Vetri per contenitori: ruolo dello stato redox nella formazione e stabilità del colore. Riv Staz Sper Vetro 5:5-16 Daneo A, Falcone R, Hreglich S (2009) Effect of the redox state on container glass color stability. Glass Technol Eur J Glass Sci Technol A 50:147-150 Daneo A, Polato P, Scandellari M L, Verità M (1992) Spectrophotometric determination of ferrous iron and sulphides in industrial hollow soda-lime glass. Bol Soc Esp Ceram Vidrio 31:509-514 Dolan MD, Misture ST (2004) Analysis of glass batch reactions using in situ x-ray diffraction. Part 2. Sodalime-silica glass batches. Glass Technol 45:167-174 Douglas R W, Zaman M S (1969) The chromophore in iron-sulphur amber glass. Phys Chem Glasses 10:125132 Dubois B, Rey S, Petit-Marie D, Barton JL (1990) Effect of melting temperature on redox state of glass refined with sulfate. Glastech Ber 63K:253-260 Falcone R, Licenziati F, Orsega EF, Verità M (2011) The dependence of the weathering of soda-lime-silica glass on environmental parameters: a preliminary investigation. Glass Technol Eur J Glass Sci Technol A 52:23-29 Fowkes AJ, Parton C (1969) Sulphur and iodine condensates in bubble in glass. Glass Technol 5:147-150 Geotti-Bianchini F, Verità M, De Riu L, Stella A (1994) Evaluation of tin oxide coatings on glass containers using HECM and EPMA. Glass Technol 35:216-223 Golob H R, Lowell Swarts E (1984) Diproportion of SO2 in bubbles within soda-containing glasses. J Am Ceram Soc 67:564-567
140
Falcone, Ceola, Daneo, Maurina
Greim J, Knodler H, Schaeffer H A (1981) Mass spectrometric analysis of the SO2 gas reaction with soda-limesilica glass surfaces. Verres Réfract 35:315-318 Guadagnino E, Corumluoglu O (1997) Indirect determination of sulphide sulphur in glass by flame atomic absorption spectrometry. Glass Technol 38:179-182 Guloyan YA (2007) Complete analysis of ionic staining of glasses by transition-metal compounds. Glass Ceram 64:153-158 Harding FL, Ryder RJ (1970) Amber colour in commercial silicate glasses. J Can Ceram Soc 39:59-63 Hlavac J (1983) The Technology of Glass and Ceramics. An Introduction. Elsevier, Amsterdam Hreglic S, Scandellari M, Verità M (1979) Impiego delle scorie d’altoforno come materia prima nella produzione del vetro. Riv Staz Sper Vetro 9:205-214 Hrma P (1985) Reaction between sodium carbonate and silica sand at 874 °C < T < 1022 °C. J Am Ceram Soc 68:337-341 Hrma P (1990) Batch melting reactions. In: Chemistry of Glasses - 2nd Edition. Paul A (ed) Chapman and Hall, London New York, p 157-178 Jacobsson B, Högberg B (1947) The sensitivity of beer to light. Protection afforded by glass bottles. Wallerstein Lab Comm 29:5-16 Kim S, Sanders TH (1991) Phase diagram for ceramists. J Am Ceram Soc 74:1833-1840 Kloužek J, Arkosiová M, Lubomír N, Cincibusová P (2007) The role of sulphur compounds in glass melting. Glass Technol Eur J Glass Sci Technol A 48:176-182 Kramer F W (2005) Solubility of gases in glass melts. Ber Bunsenges Phys Chem 100:1512-1514 Kraub M, Lenhart A, Ratka H, Kircher U (1995) Characterization of filter dust in the glass container industry and definition of its influence on green glass melts. Glastech Ber Glass Sci Technol 68:278-284 Laimböck P (2008) In-line oxygen sensors for the glass melt and the float bath. Adv Mater Res 39-40:443-446 Lehmann J, Nadif M (2011) Interactions between metal and slag melts: steel desulfurization. Rev Mineral Geochem 73:493-511 Levich,VG (1962) Physiochemical Hydrodynamics. Prentice-Hall, Englewood Cliffs, NJ USA Levin EM, Robbins CR, McMurdie HF (1964) Phase Diagrams for Ceramists. The American Ceramic Society, Columbus Locardi B (1992) Qualità del vetro e conservazione degli oli vegetali. Riv Staz Sper Vetro 2:67-72 Manring WH, Hopkins RW (1958) Use of sulphates in glass. The Glass Industry 39:139-142 Mastrobattista G (1990) Effect of light on extra virgin olive oils in different types of glass bottles. Ital J Food Sci 3:191-195 Maurina S, Falcone R, Vallotto M (2008) Origine e caratterizzazione delle inclusion gassose contenenti zolfo nei vetri sodico-calcici per contenitori. Riv Staz Sper Vetro 4:17-15 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Müller-Simon H, Gitzhofer K (2008) Sulphur mass flow balances in industrial glass melting furnaces. Glass Technol Eur J Glass Sci Technol A 49:289-296 Novotný F, Lošot R (2008) Chemical reactions in a soda-lime silicate batch. Advanced Materials Research 39-40:459-464 Paul A (1990) Chemistry of Glasses - 2nd Edition. Chapman and Hall, London New York Peters A (1993) Determination of reducing components in glassmaking raw materials. Glastech Ber 66:19-24 Polato P, Daneo A (1991) Specificazione del colore: dall’atlante di Munsell al diagramma colorimetrico CIE. Riv Staz Sper Vetro 6:303-314 Polato P, Daneo A, Segato P (1988) Caratterizzazione ottica del vetro cavo. Riv Staz Sper Vetro 1:5-18 Polato P, Segato P, Daneo A (1989) Determination of color difference or “colorlessness” for glasses using the spectrophotometric method. Riv Staz Sper Vetro 1:81-90 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Scalet B (1996) Problems related to the use of cullet and the reuse of dusts from fumes treatment plants. Intern Glass J 87:65-69 Scholze H (1991) Glass, Nature, Structure and Properties. Springer Verlag, New York Shaeffer HA (1984) Diffusion-controlled processes in glass forming melts. J Non-Cryst Solids 67:19-33 Shelby JE (1997) Introduction to Glass Science and Technology. Royal Society of Chemistry Simpson W (1976) Calumite slag as a glassmaking raw material for the increase of furnace productivity. Glass Technol 17:35-40 Simpson W (1979) Calumite slag – its benefits to the glassmakers. Riv Staz Sper Vetro 9:364-368 Simpson W, Myers DD (1978) The redox number concept and its use by the glass technologist. Glass Technol 19:82-85 Sinton C W (2006) Raw Material for Glass and Ceramics. John Wiley and Sons, New Jersey Trier W (1987) Glass Furnaces: Design Construction and Operation. Society of Glass Technology, Sheffield
Sulfur Compounds in Industrial Glass
141
Verità M, Falcone R, Sommariva G, Chopinet M-H, Lehuédé P (2009) Weathering of the inners surface of soda-lime-silica glass containers exposed to the atmosphere. Glass Technol Eur J Glass Sci Technol A 50:65-70 Volf MB (1961) Technical Glasses. Sir Issac Pitman and Sons, Eds, London Volf MB (1990) Technical Approach to Glass. Elsevier, Amsterdam and New York Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov RE, Kohn SC, Janousch M (2008) The origin of S4+ detected in silicate glasses by XANES. Am Mineral 93:235-240 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Wiszecki G, Stiles WS (1982) Color Science – Concepts and Methods, Quantitative Data and Formulae – 2nd Edition. John Wiley and sons, New York
6
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 143-165, 2011 Copyright © Mineralogical Society of America
Experimental Studies on Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure Linda Backnaes* and Joachim Deubener Institute of Non-Metallic Materials Clausthal University of Technology, Zehntnerstr. 2a D-38678 Clausthal-Zellerfeld, Germany
[email protected] *current affiliation: SCHOTT Electronic Packaging, Christoph-Dorner-Str. 29, D-84028 Landshut, Germany
INtroDuctIoN Various experimental studies in silicate melts have been performed to understand the dependence of sulfur solubility on melt composition and experimental conditions. These experiments were motivated by glass technologists due to the role swulfur compounds play in fining (Müller-Simon 2011, this volume) and coloring (Falcone et al. 2011, this volume) melts. In particular, the risk of foaming in the glass tank and rejects in the glass production due to discoloration and bubbles demand a systematic approach for sulfur solubility in silicate melts. Also sulfur solubility experiments have been conducted by metallurgists, whose interest is centered on the interaction of metal and slag melts to desulfurize steel products (Lehmann and Nadif 2011, this volume). Besides technical applications, sulfur solubility experiments at atmospheric pressure are of importance to geoscientists in modelling near-surface conditions such as sulfur degassing from volcanoes (Oppenheimer et al. 2011, this volume) and the role of sulfur in the formation of ore deposits (Simon and Ripley 2011, this volume). In order to study the effect of polymerization of the silicate network on sulfur solubility, melt compositions were varied in the experiments across broad limits. It has been shown that the solubility generally increases with increasing network modifier to network former ratio (Baker and Moretti 2011, this volume). This supports the idea that the presence of free volume and cationic charge compensators in the network structure promote incorporation and mobility of anionic sulfur species (Behrens and Stelling 2011, this volume). Sulfur speciation is also responsible for the strong dependence of sulfur solubility on oxygen fugacity as reported by Baker and Moretti (2011, this volume) and Müller-Simon (2011, this volume) for natural and technical melts, respectively. Sulfur was found to be stable in these melts as sulfide S2− under reducing conditions and as sulfate (SO4)2− under oxidizing conditions with a sharp transition at oxygen fugacities close to that of the Ni-NiO buffer (Wilke et al. 2011, this volume). In most of the experimental studies on sulfur solubility, sulfur was introduced by either mixing solid sulfur sources such as sodium sulfate to the glass batch before melting or by passing sulfur-bearing gases through/over the hot silicate melt. For any sulfur source the sulfur retention of the melt will depend on dwell time of the sulfurization experiment while the sulfur solubility in the melt at saturation with the present sulfur source, is independent of dwelling. This means that in order to be able to talk about sulfur solubility, the experiments need to be performed with consideration of the equilibrium between at least two phases, for example melt + fluid, melt + salt or melt + gas (Baker and Moretti 2011, this volume). As a consequence of the slow mobility of the sulfur species in silicate melts (Behrens and Stelling 2011, this volume) 1529-6466/11/0073-0006$05.00
06_Backnaes_Deubener.indd 143
DOI: 10.2138/rmg.2011.73.6
6/22/2011 5:15:05 PM
Backnaes & Deubener
144
these specific equilibria may be reached in large sample volumes and at low temperatures only after days or even longer dwell times. The lack of detailed description of experimental conditions in several reports makes it difficult to identify whether the sulfur content of the glasses reflects equilibrium conditions or a kinetically controlled state specific to the applied procedures. With regard to gas reactions, a strong control on the fugacity of sulfur and oxygen needs to be undertaken. This has often not been done because the scope of the experiments was mainly to improve manufacturing procedures. For example, in an experiment designed to investigate the fining of melts, an equilibrium state is usually not reached, and therefore the sulfur contents of the glass product does not represent the true solubility of sulfur in the melt. In the course of this paper the term solubility is used only for experiments in which equilibrium or at least near-equilibrium conditions were achieved while the term retention is used for all other experiments.
ANALySIS MEthoDS for SuLfur coNtENt There are several ways of measuring the sulfur content in a glass. Since the sulfur contents are usually quite low however, the analysis method needs to have a high sensitivity for element detection. The analysis methods for sulfur determination in glasses are thoroughly discussed in Ripley et al. (2011, this volume), and hence only the methods employed in the studies presented in this chapter are briefly mentioned below. Two main analysis methods were employed: Electron microprobe analysis (EMPA) (Haughton et al. 1974; Buchanan and Nolan 1979; Kaushik et al. 2006; Bingham et al. 2007, 2008; Mishra et al. 2008; Lenoir et al. 2009) and X-ray fluorescence (XRF) analysis (Papadopoulos 1973; Barbon et al. 1991; Ooura and Hanada 1998; McKeown et al. 2001; Manara et al. 2007). In some of the older papers, a method based on acid extraction was employed (Pearce and Beisler 1965; Nagashima and Katsura 1973; Katsura and Nagashima 1974), in which the glass was dissolved in acid (hydrochloric acid, strong phosphoric acid, or mixtures of stannous chloride dehydrate with phosphoric acid). When using the phosphor-based acids (Nagashima and Katsura 1973; Katsura and Nagashima 1974), the sulfur was dissolved as hydrogen sulfide, whereupon it was fixed as zinc sulfide and photometrically determined. When using hydrochloric acid (Pearce and Beisler 1965), the dissolved material was filtered, and the filtrate containing the sulfur fraction was analyzed gravimetrically (as BaSO4). Combustion analysis was used for sulfur detection by Fincham and Richardson (1954). In this method, the glass is rapidly combusted and the amount of sulfur dioxide gas (SO2) is quantified through infra-red spectroscopy. A similar method, namely evolved gas analysis (EGA) combined with the use of a mass spectrometer, was used by Klouzek et al. (2006, 2007) for the determination of sulfur gases released from a melt in the furnace. Therefore mixtures of glass, coke and sodium sulfate were filled in a silica glass tube and placed in a laboratory furnace that was heated at a rate of 5 K min−1. Gases evolved from the sample were flushed out to sampling loops by a stream of air. The loops were repeatedly connected to the stream of carrier gas of a chromatograph. Furthermore, combustion analysis was used by Beerkens and Kahl (2002), which dissolved the evolved SO2 gas in NaOH peroxide and determined the sulfur content of the resulting solution by titration.
INDuStrIAL MELtS Effect of oxygen fugacity Since sulfur is a polyvalent element, one of the most influential parameters with regard to sulfur solubility and speciation in silicate melts is the oxygen fugacity (fO2) in the melt. The
06_Backnaes_Deubener.indd 144
6/22/2011 5:15:06 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
145
effect of fO2 on sulfur solubility has been investigated by several research groups (Fincham and Richardson 1954; Holmquist 1966; Nagashima and Katsura 1973; Papadopoulos 1973; Schreiber et al. 1987; Beerkens et al. 2002, 2003a, 2003b; Müller-Simon et al. 2008). In the pioneering work of Fincham and Richardson (1954) sulfur solubility in the binary systems CaO-SiO2, MgO-SiO2, FeO-SiO2 and CaO-Al2O3 as well as the ternary system CaO-SiO2-Al2O3 was investigated. The melts were equilibrated with a furnace atmosphere containing a H2-CO2-SO2 gas mixture of constant sulfur dioxide fugacity (fSO2) but different fO2. The reaction time in the temperature range from 1350 to 1650 °C was varied to reach constant sulfur content in the melt. Thus, their experiments determined the sulfur solubility at saturation with the sulfur-bearing gases. The sulfur content of the melts was analyzed by combustion after quench. At low oxygen fugacities (fO2 < 10−7 bar), Fincham and Richardson (1954) found that the solubility of sulfur is enhanced with decreasing fO2, but declined with decreasing fO2 under oxidizing conditions (fO2 > 10−2 bar). The following equilibrium reactions have been proposed to explain the partitioning between the sulfur in gas (g) and melt (m) and the solubility of sulfur as reduced sulfide (S2−) under low and as oxidized sulfate (SO42−) under high oxygen fugacities: 3 SO2 ( g ) + O2 − (m) ↔ O2 ( g ) + S2 − ( m) 2
(1)
1 SO2 ( g ) + O2 ( g ) + O2 − ( m) ↔ SO 4 2 − ( m) 2
(2)
with the corresponding equilibrium constants: K sulfide = K sulfate =
fO2 3/ 2 aS2−
(3)
fSO2 aO2− aSO 2−
(4)
4
fSO2 fO2 1/ 2 aO2−
where aS2−, aSO42− and aO2− are the activities of sulfide, sulfate and free oxygen in the melt, respectively. Rearranging Equations (3) and (4) in logarithmic scales and assuming ideal behavior of the mixed compounds the solubility of sulfur as sulfide (Ssulfide) and as sulfate (Ssulfate) is: 3 logSSulfide = − log PO2 + log PSO2 + log K sulfide × cO2− 2
(
logSSulfate =
1 log PO2 + log PSO2 + log ( K sulfate × cO2− ) 2
)
(5) (6)
where PO2, PSO2 are the partial pressures of oxygen and sulfur dioxide and cO2− is the concentration of free oxygen in the melt (O2−) Nagashima and Katsura (1973) investigated the sulfur solubility in binary Na2O-SiO2 melts of molar ratios of 1:3, 1:2 and 1:1 at 1100 °C, 1250 °C and 1300 °C by keeping the SO2 partial pressure constant but varying the oxygen partial pressure. They passed a pre-mixed gas flow of SO2 with addition of CO2 and H2 over the melt to control the oxygen and sulfur dioxide fugacity. The sulfur content of the glasses was measured through an acid extraction method. Gas-saturated melts were obtained at 1300 °C and 1250 °C as shown from the lack of variation in sulfur content with run duration. For the mixture of 1:1 Na2O to SiO2 at 1100 °C however, a significant deviation from equilibrium with the gas phase was reported, since the sulfur concentration levels were reported to be close to 8% (it was however not stated if the value referred to mol% or wt%). Furthermore, the experimental dwell times at the given temperatures was not stated. The equilibrium compositions of the gas phase were calculated numerically by minimizing free energy of the gas mixture using the method of White
06_Backnaes_Deubener.indd 145
6/22/2011 5:15:06 PM
146
Backnaes & Deubener
et al. (1958) and comparing the calculated PO2 with the actual oxygen partial pressure measured by a ZrO2 sensor. At constant temperature, the sulfur solubility for all three glass compositions showed a minimum for intermediate oxygen partial pressures in the range PO2 = 10−7-10−9 bar, with the minimum solubility being shifted to lower oxygen partial pressures as the sodium oxide content increases (Fig. 1). The maximum sulfur solubility was found at either high or low oxygen partial pressures, i.e., during either oxidizing or reducing conditions, when sulfur is present in the form of sulfate and sulfide respectively. In Figure 1 the logarithm of the oxygen fugacity was specified relative to the nickel-nickel oxide (NNO) buffer since in the relevant temperature range NNO equals QFM + 1. In this diagram it is assumed that PO2 equals fO2. Experimental conditions for the data presented in Figure 1 are listed in Table 1 of the Appendix. A different approach to investigate the dependence of sulfur solubility on oxygen fugacity was used by Beerkens and Kahl (2002) for soda-lime-silicate melts (74 SiO2-16 Na2O-10 CaO). The glasses were produced by adding sodium sulfate (1 wt%) and iron (up to 1.5 wt%), and carbon powder as reducing agent (in various amounts) to the batch. Batches of 700 g (with a 7 cm initial batch height) were pre-melted in silica and alumina crucibles at 1200 °C for 16 h and then heated up to 1400 °C in air and subsequently cooled down to room temperature. The oxygen
figure 1. Total sulfur in the melt as a function of oxygen fugacity (relative to the log fO2 of the nickel-nickel oxide equilibrium DNNO) in different silicate melts. Straight lines indicate sulfate and sulfide solubility trends according to (d(logS)/d(logfO2)) = 1/2 and (d(logS)/d(logfO2)) = −3/2, respectively. Data: 39.4 SiO243.3 CaO-17.3 Al2O3-melts: Black full symbols, square = 1425 °C, circle = 1500 °C, triangle = 1550 °C, Fincham and Richardson (1954); SiO2-Na2O-melts at 1200 °C: Half-filled symbols, circle = trisilicate, square = disilicate, Holmquist (1966); SiO2-Na2O-melts at 1250 °C: Open symbols, circle = trisilicate, square = disilicate, star = metasilicate, Nagashima and Katsura (1973); 74 SiO2 – 16 Na2O – 10 CaO melts at 1400 °C: Gray halftone symbols, square = 0.12 wt% FeO , pyramid = 0.3 wt % FeO, diamond = 1.5 wt% FeO, circle = lower Na2O content, Beerkens and Kahl (2002). It needs to be emphasised that the data set from Beerkens and Kahl (2002) does not represent equilibrium between the sulphur and melt phases. Lines connecting data are intended as a visual guide.
06_Backnaes_Deubener.indd 146
6/22/2011 5:15:07 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
147
fugacity in the melt was measured during heating from 1200 °C to 1400 °C and during cooling with a ZrO2 oxygen sensor (immersed 3-5 cm below the glass surface), and was within the PO2-range of 10−0.2-10−7.5 bar. The sulfur content of the glass was analyzed by a hot extraction method. A minimum of sulfur retention was detected in the PO2-range of 10−5-10−6 bar (Fig 1). A major focus of the study of Beerkens and Kahl (2002) was on the variation of the intensity of amber coloration with redox state of the melt using the intensity of the absorption band at a wavelength of 410 nm as the controlling parameter. The increase of PO2 is associated with a change from amber-colored to colorless glasses. In contrary to the studies of Nagashima and Katsura (1973) and Fincham and Richardson (1954), equilibrium was not reached in the experiments of Beerkens and Kahl (2002) and, hence, the concentrations of sulfur in the postexperimental glasses represent only sulfur retention. However, the trends of S-content of melts as a function of PO2 are quite similar in these studies and agree with the so called “Budd-curve” introduced in glass technology through the pioneering work on sulfur retention of Budd (1965). Beerkens and Kahl (2002) interpreted their results with reference to the stability of the different sulfur species, i.e., sulfate being stable at high oxygen fugacity, and sulfide at reducing oxygen fugacities. As the sulfur retention was at its minimum in the sulfite stability field at intermediate oxygen fugacity they concluded that sulfite is not a stable species in the melt. Spectroscopic data summarized by Wilke et al. (2011, this volume) support this assumption and showed that sulfide and sulfate occurred in its own specific oxygen partial pressure range while both species can coexist in the melt only in a narrow range of oxygen fugacity centered about one order of magnitude above the equilibrium fO2 of the quartz-fayalite-magnetite (QFM+1). Furthermore, Backnaes et al. (2008) detected solely sulfate and sulfide species in technical soda-lime-silica glasses using XANES spectroscopy. However, this is not an unambiguous proof that sulfite does not exist in the high temperature melts. As noted by Müller-Simon (2011, this volume) sulfite may convert to other sulfur species during cooling. According to Equations (5) and (6) slopes d(log S)/d(log fO2) of −1.5 and +0.5 are expected when sulfide and sulfate, respectively, are the sole sulfur species. From inspecting the data collected in Figure 1 it can be seen that only the data from Fincham and Richardson (1954) and Nagashima and Katsura (1973) reveal similar slopes, but there too, the slight deviations from models of sulfur solubility (Baker and Moretti 2011, this volume) are evident. The data from the other reports show noticeable discrepancy from the expected trend that indicates either variable sulfur dioxide fugacity or contributions of other sulfur dissolution reactions within these series of experiments. Whether sulfur dissolves exclusively as sulfate at high oxygen fugacities and sulfide at low oxygen fugacities can be tested by inspecting the ranges with respect to the oxygen fugacity of constant values of K′sulfide and K′sulfate—i.e., plotting log K′ vs. log PO2 with logK′ = log(K × cO2−) in Figure 2. Fincham and Richardson (1954) defined an upper limit for exclusive sulfide and a lower limit for exclusive sulfate solubility at oxygen fugacities of 10−6 bar and 10−4 bar respectively for a calcium alumosilicate melt at 1500 °C. Nagashima and Katsura (1973) concluded that for sodium silicate compositions at 1250 °C sulfur solubility is governed by the sulfate equilibrium (Eqn. 4) at oxygen fugacities above 10−7 bar and by the sulfide equilibrium (Eqn. 3) at oxygen fugacities below 10−9 bar. The close correlation of sulfur solubility and sulfur speciation is in agreement with spectroscopy data (XANES) of Jugo et al. (2010), where a sharp step in the S6+/SS ratio was detected at oxygen fugacities close to NNO buffer (≈ 10−4.2 bar at 1500 °C and ≈ 10−7 bar at 1250 °C). Therefore in Figure 2 the logarithm of the oxygen fugacity was specified relative to the nickel-nickel oxide (NNO) buffer and it was assumed for simplicity that PO2 equals fO2. The oxidation of sulfur under high oxygen fugacities (Eqn. 2) has also been studied by Holmquist (1966) by equilibrating binary sodium silicate melts with mixtures of nitrogen,
06_Backnaes_Deubener.indd 147
6/22/2011 5:15:07 PM
148
Backnaes & Deubener
figure 2. Equilibrium constants for sulfide reaction Equation (5) (dashed lines) and sulfate reaction Equation (6) (solid lines) as a function of oxygen fugacity (relative to the log fO2 of the nickel-nickel oxide equilibrium DNNO). Data: 39.4 SiO2-43.3 CaO-17.3 Al2O3-melts at 1500 °C: Symbols: Black full circles, Fincham and Richardson (1954); SiO2-Na2O-melts at 1250 °C: Open symbols, circle = trisilicate, square = disilicate, star = metasilicate, Nagashima and Katsura (1973); Lines are intended as a visual guide.
hydrogen and sulfur dioxide gases. At relative high oxygen partial pressures (10−2-10−4 bar), with sulfur dioxide gas as the sulfur source and sulfate being the exclusive sulfur species in the melt, the dissolution mechanism was described. Using different Na2O/SiO2 ratios in the range from 1.5 to 4, Holmquist (1966) demonstrated that the slope of logS vs. log (PSO2 × PO21/2) was close to unity at sulfur fugacities near to saturation, which confirmed the assumption that sulfur dissolves exclusively as single SO42− groups in the glass network under highly oxidized conditions (Fig. 3). The dependence of sulfur retention on oxygen fugacity is addressed in glass technology commonly by using redox numbers based on the refining reaction of carbon and sodium sulfate with the silicate melt and the chemical oxygen demand to oxidize carbon entirely to carbon dioxide. Empirical factors were introduced to evaluate the various raw materials according to their ability to influence the redox state of the batch as a whole. Two different redox number concepts are frequently used: Glass redox number (GRN) by Simpson and Myers (1978) and chemical oxygen demand (COD) by Manring and Diken (1980). Details on the calculation of different redox numbers can be found in Müller-Simon (2011, this volume) and consequences for coloring glasses are shown in Falcone et al. (2011, this volume). Since industrial redox numbers are calculated on the basis of relative proportions of constituent raw material, they are not equivalent to the oxygen fugacity in the melt. However, empirical relations between redox numbers and PO2 have been introduced (Müller-Simon 1999) using in-situ measurements of oxygen partial pressures during tank operation. Simpson and Myers (1978) added calumite slags (sulfide-bearing earth alkaline alumosilicate glass) as a reducing agent to highly oxidized multi-component soda-lime silicate batches
06_Backnaes_Deubener.indd 148
6/22/2011 5:15:08 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
149
figure 3. Plot of the logarithm of total sulfur vs. logarithm of (PSO2 × PO20.5/bar1.5) to test the validity of the sulfate reaction (Eqn. 6) for Na2O-SiO2 melts at 1200 °C (gray halftone symbols: Holmquist 1966) and 1250 °C (open symbols: Nagashima and Katsura 1973). Dotted line indicates sulfur saturation in the melt in equilibrium with liquid sodium sulfate, dash-dotted line indicates sulfate solubility according to d(logS)/ d(log(PSO2 × PO20.5)) = 1 and lines connecting data points are intended as a visual guide.
used to manufacture container glassware. Sulfur was analyzed in final glass products after melting batches in a glass tank at 1500-1530 °C. Manring and Diken (1980) studied melting and fining of sulfur-bearing soda-lime-silica glass batches in laboratory (hot stage microscopy) and industrial practice (water-cooled periscope). They recalculated Budd’s sulfur retention data (Budd 1965) on the basis of COD redox number. Barbon et al. (1991) investigated redox equilibrium and coloration of multi-component soda-lime-silicate glass batches by observing changes in iron and sulfur concentration as well as melting temperature and redox number. The batches were melted under a controlled atmosphere of PO2 = 10−2 bar. Since the objective was to imitate processes in industrial glass production however, it is possible that equilibrium between the added sodium sulfate and the silicate melt was not reached. The total sulfur and iron contents in the glass were analyzed through X-ray fluorescence, while the ferrous iron and sulfide contents were obtained through chemical analysis. The sulfate and ferric iron contents were determined as the difference between the total contents and sulfide or ferrous iron, respectively. Finally, Müller-Simon and Glitzhofer (2008) balanced sulfur in the production process from measurements of sulfur in raw materials, fuel, combustion air, waste gas and final soda-limesilica glasses. In this study ICP-AES was used to quantify sulfur retention in flint (white) and in green and amber colored glasses. Sulfur retention was correlated to the reducing ability of the different batches used to produce the glasses. Figure 4 compiles the available sulfur retention data of industrially produced soda-limesilica glasses as a function of their batch redox number. Based on sulfur speciation (see Fig. 1), typical V-shaped curves are evident in Figure 4, with the minimum sulfur solubility located between redox number −60 and −20. Sulfur fining at production temperatures close to 1500 °C is limiting the sulfur retention in the final glass products at a level below 0.15 wt% S.
06_Backnaes_Deubener.indd 149
6/22/2011 5:15:08 PM
150
Backnaes & Deubener
figure 4. Sulfur retention in industrial soda-lime-silica glasses as a function of GRN and COD batch redox numbers. Symbols and lines: Full circle, Barbon et al. (1991); Open cirlce, Simpson and Myers (1978); Dashed line, “Budd”-curve after Manring and Diken (1980); Dashed-dotted line, Klouzek et al. (2006) calculated from oxygen partial pressures at 1500 °C using the empirical relation: COD = −1.76 + 9.7 × logPO2 (Müller-Simon 1999); Stars, Müller-Simon and Glitzhofer (2008).
Effect of melt temperature The degassing of sulfur from the melt during heating is the premise for using sulfate as a fining agent during industrial melting, and the process consequently leads to a sulfur depleted melt at high temperatures (sulfur degassing reactions of glass carbon mixtures are at 1100-1300 °C and at ca. 1550 °C, see e.g. Müller-Simon 2011, this volume). Consequently, at oxygen fugacities above NNO an increasing melting temperature has a negative effect on the retention of sulfur in silicate melts (Fincham and Richardson 1954; Holmquist 1966; Papadopoulos 1973). Figure 5 shows an Arrhenian decrease in sulfur solubility with increasing dwell temperature for relative high oxygen fugacities where sulfur is dissolved as sulfate in the melt, but due to the change in the dissolution mechanism of sulfur with oxygen fugacity, at low oxygen fugacities (sulfide solubility) the opposite trend is evident (Fincham and Richardson 1954). From inspecting the slopes in Figure 5 it can be seen that sulfide reaction (Eqn. 1) is endothermic, while the sulfate reaction (Eqn. 2) is exothermic. The latter enthalpy decreases as the Na2O content increases, which can be attributed to the increasing depolymerization of the melt. As in situ high temperature Raman measurements experiments have shown (e.g., Mysen and Frantz 1994), increasing temperature favors disproportionation and therefore non-bridging oxygen (NBO) in the silicate network, particularly for silicate melts of low NBO/T ratio.
Effect of melt composition The following section will be divided into two separate parts; one focusing on studies where sulfur was added as a salt to the melt, and the other where sulfur was added in gaseous form. Generally the composition of a melt strongly affects sulfur solubility due to the changing activity of free oxygen ions (= melt basicity) and the NBO/BO ratio (ratio of non-bridging to bridging oxygen in the melt) is changed (Nagashima and Katsura 1973; Papadopoulos 1973; Ooura and Hanada 1998; Beerkens and Kahl 2002; Manara et al. 2007; Chopinet 2007).
06_Backnaes_Deubener.indd 150
6/22/2011 5:15:08 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
151
figure 5. Temperature dependence of total sulfur in silicate melts equilibrated with gas mixtures for different oxygen partial pressures. At log(PO2(bar)) = 0 and −4.33 sulfur dissolves as sulfate while at log(PO2(bar)) = −9.16 sulfide is the stable sulfur species. Data: 39.4 SiO2-43.3 CaO-17.3 Al2O3-melts at log(PO2(bar)) = 9.16 and 0: Black full symbols, Fincham and Richardson (1954); SiO2-Na2O-melts at log(PO2(bar)) = -4.33: Open symbols, circle = trisilicate, square = disilicate, star = metasilicate, Nagashima and Katsura (1973). Lines are intended as a visual guide.
According to Equations (1) and (2), free oxygen O2−(m) is needed for incorporation of sulfur in the melt while bridging oxygen O0(m) do not support sulfur dissolution. Acid-base reactions following Lewis’ (1923) concept of acids and bases being acceptors and donators of an electron pair, respectively, were assumed to be active in molten oxides via transfer of “free” oxygen ions O2− from metal oxides to silica (Lux 1939; Flood and Förland 1947). This is explained in detail in by Baker and Moretti (2011, this volume). According to Fincham and Richardson (1954) and Toop and Sammis (1962) the equilibrium of the three different oxygen species can be expressed as: O0 ( m) + O2 − ( m) ↔ 2O − ( m)
(7)
where O0(m) is the bridging oxygen (BO), O2−(m) is the free oxygen and O−(m) is the nonbridging oxygen (NBO). This reaction can be used to describe the activity of free oxygen in the melt as a measure of the melt basicity, which is also a measure of the depolymerization, i.e., the ratio between non-bridging and bridging oxygens via: aO2− =
aO−2 K aO0
≈ K oxygen −1
(NBO)2 (BO)
(8)
where Koxygen is the composition-dependent constant of the oxygen reaction. In Equation (8) the activities were replaced by molar fractions of NBO and BO assuming ideal mixing of oxygen species in the melt. According to Equation (8), free oxygen ions are increasingly formed as the composition changes from pure silica to metal oxides. The real distribution of BO and NBO, i.e., the degree of polymerization, can be derived from the Qn distribution measured by
06_Backnaes_Deubener.indd 151
6/22/2011 5:15:09 PM
152
Backnaes & Deubener
spectroscopic experiments such as NMR and Raman (Qn denotes a SiO4 tetrahedron in which n oxygens are bridging (BO) to other silicon tetrahedra and 4−n are nonbridging (NBO)). Sulfur added as salt. In a 1998 study, Ooura and Hanada doped binary melts of SiO2-R2O composition (R = Na, K, Li) as well as ternary melts SiO2-MO-Na2O (M = Mg, Ca, Ba) with alkali sulfates to investigate the sulfate solubility at 1350 °C in air atmosphere. The sulfate contents of the glasses were analyzed using energy dispersive X-ray fluorescence spectrometry. All samples were synthesized for 20 minutes in a Pt-crucible as batches calculated to produce 5 g of glass product. The dwell time was estimated based on an experiment with a 72.5 SiO212.5 BaCO3-12.5 Na2CO3-2.5 Na2SO4 batch. For this particular melt the sulfate concentration decreased from an initial concentration of 2.5 mol% to 2 mol% within the first 20 minutes of synthesis, and between 20 and 60 minutes dwell time no change in sulfate concentration was measured. However, the sulfur retention was higher than the equilibrium solubility expected at 1350 °C, as can be seen from comparison with solubility data in sodium silicate melts at 1200 and 1250 °C in Figure 3. This implies that equilibrium distribution of sulfur between melt and gas was not reached. Besides increasing sulfur retention with increasing depolymerization of the melt, expressed as the number of non-bridging oxygen per tetrahedron (NBO/T), Ooura and Hanada (1998) established a trend of increasing sulfur retention with increasing alkaline earth fraction in the melt in the order Mg-Ca-Ba, with Mg showing the lowest retention and Ba the highest (Fig. 6). The sulfur retention scales negatively with the field strength of the alkaline earth metals M which is defined as F = Z / (MOdistance)2 where Z is the charge of the cation (Fig. 6). This may indicate that the Qn groups in the different silicate melts are not energetically equivalent since, from their abundance (cations with higher field strength show an increase in the abundance of Q species with small n values; Maekawa et al. 1991), the opposite trend is expected. One may also argue that the abundance of free oxygen is not the only structural parameter determining sulfur solubility. Sulfate needs cations as next neighbours, and large divalent cations may be better modulators between the silicate network and sulfate groups than small divalent cations. Sulfur retention in borosilicate glasses for nuclear waste storage has been investigated by Manara et al. (2007). In this study the base glass was a Na2O-B2O3-SiO2 melt, which was doped with various amounts of alkali oxides in order to determine their specific effects on sulfur solubility and sulfurization dynamics. Sulfur was added as sodium sulfate or cesium sulfate to the melt. The synthesis was performed at a viscosity of 10 Pa·s with dwell times of up to 1000 minutes (temperature was therefore adjusted between 1073 and 1473 K with a heating rate of ≈ 4 K min−1), and the resulting sulfur content was analyzed by XRF and EDS analysis (Fig. 7 ). Sodium borosilicate glasses consist of two network former cations: silicon and boron. Depending on the molar ratio of sodium oxide to boron trioxide R = (Na2O)/(B2O3) the added sodium oxide can associate in the glass structure either with silicon tetrahedra and creating NBOs or with BO3-units transforming them to boron tetrahedra (BO4) creating no NBOs in this process. According to the revised Yun, Bray and Dell (YBD) model which incorporates besides BO3 and BO4 complex borosilicate units such as reedmergnerite (Na2O-B2O3-6 SiO2) and danburite (Na2O-B2O3-2 SiO2) creation of NBOs in sodium borosilicate glasses can be described as follows (Manara et al. 2009): For R < 0.5 sodium cations can act as chargecompensators for the formation of four-coordinated boron atoms (BO3 units are transformed to BO4), and in the range from 0.5 to R* (R* = upper limit for charge compensation) they can compensate charge of borosilicate groups (reedmergnerite and danburite units). For R > R* the additional alkali cations cause depolymerization of the glassy network, starting to form NBOs in the silica tetrahedrons. NBOs are then formed also in borate units at higher sodium contents. The study pointed out that the sulfur retention in fully polymerized borosilicate melts with R < R* was practically zero, in the range from 0.5R to R* very small, but increased sharply at R*, which confirmed that basicity of the melt, i.e., the presence and concentration of free oxygen in
06_Backnaes_Deubener.indd 152
6/22/2011 5:15:09 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
153
figure 6. Total sulfur as a function of NBO/T for sodium silicate and sodium-alkaline-earth (CaO, MgO, BaO)-silicate melts at 1350 °C. Data: Black full symbols - Ooura and Hanada (1998); open and grey halftone - Holmquist (1966) and Nagashima et al. (1973). The insert shows the dependence of sulfur retention on field strength of alkaline earth ions for sodium-alkaline-earth-silicate melts with NBO/T = 0.67 at 1350 °C. Lines are intended as a visual guide.
figure 7. Effect of Na2O to B2O3 ratio on total sulfur content in sodium borosilicate glasses. Data at R ≈ 1.7 was affected by evaporation effects. Line is intended as a visual guide. (Modified after Manara et al. 2007)
06_Backnaes_Deubener.indd 153
6/22/2011 5:15:10 PM
Backnaes & Deubener
154
the melt as determined by Koxygen in Equation (7) is the determining factor of sulfur solubility in borosilicate melts. Manara et al. (2007) reported that the sulfur retention decreases when sulfur is added as Cs2SO4 as opposed to Na2SO4, due to slower kinetics of sulfate incorporation when Cs is present (see Behrens and Stelling 2011, this volume). In a 1965 paper, Pearce and Beisler investigated the unmixing in Na2O-SiO2-Na2SO4 melts at 1200 °C. The dwell time was set at 72 h, and it was assumed that equilibrium between the melt and sulfate phase was achieved. The sulfate content of the glass was analyzed using the acid extraction method, as described in the beginning of this chapter. The validity of the method was checked against the combustion technique presented by Fincham and Richardson (1954). Evidence was found for a miscibility gap of two liquids at Na2O-SiO2 ratios lower than 1. One liquid was almost pure sodium sulfate and the other a silica-rich composition 16 Na2O-81 SiO23 Na2SO4 (wt%). In highly SiO2-rich melts, tridymite was also observed. In melts with a Na2OSiO2 ratio larger than one no liquid-liquid phase separation was found (Kordes et al. 1951). Sulfur added in gaseous form. In a 1973 study, Papadopoulos carried out experiments on soda-lime-silica glasses with Na2O- and CaO-concentrations ranging from 8.5-13.5 mol% and 9-14 mol%, respectively. A SO2-O2-gas mixture was bubbled through the melt in the temperature range 1340-1480 °C. PSO2 at the dwell temperature was determined by the partial pressures of O2 and SO2 in the gas mixtures. On the basis of variations in dwell time, it was assumed that bubbling times of 7 h at 1340 °C and 5 h at 1480 °C were sufficient to achieve equilibrium between the melt and the gas. The content of the sulfur in the glass was determined by X-ray fluorescence analysis. Papadopoulos (1973) proposed the J-parameter as a measure of the activity of free oxygen ions in the melt. It was defined as follows: J=
(Na + )2 (NBO)2 (BO)
(9)
where (Na+), (NBO) and (BO) are the molar fractions of sodium, non-bridging and bridging oxygens respectively. Assuming that the activity of free oxygen in melts is very small as compared to the total oxygen, i.e., (NBO) + (BO) ≈ (Ototal), the presence of NBO and BO can be readily calculated from composition. As an example for a 74 SiO2-16 Na2O-10 CaO melt with (Na+) = 0.32, (NBO) = 0.52, BO = (NBO) − (Ototal) = 1.22, Equation (9) yields J = 0.0227. The results showed that J and the sulfur solubility are positive correlated and that even small changes in the melt composition can cause a strong change in the J-parameter, and hence strong variation in sulfur solubility (Fig. 8).
SoLuBILIty of SuLfur-BEArINg wAStE IN MELtS Dealing with the immobilization of toxic compounds and nuclear waste in glasses, sulfur is often the limiting factor due to its relatively low solubility. Residuals of incineration or of ionexchanging resins or effluents of radioactive solutions are typically waste products with high sulfate contents. Thus, the low sulfur solubility may constitute an important technological issue in the development of a vitrification process, as it can dictate the radioactive waste-load-limiting factor (Manara et al. 2007). The solubility of different species of sulfur has been investigated with the emphasis on trying to optimize the conditions for higher sulfur retention (McKeown et al. 2001; Kaushik et al. 2006; Bingham et al. 2007, 2008; Mishra et al. 2008; Lenoir et al. 2009). Many of these studies concentrate on the speciation of sulfur in silicate melts using X-ray absorption spectroscopy (XAS) or Raman studies, since the valence state gives information on the incorporation mode of the sulfur atoms in the melt, and hence the possibility of improving the solubility. As already described in this chapter, the solubility of sulfur in silicate melts is at its highest level at very oxidizing and very reducing atmospheres, when sulfur is incorporated as S6+ and S2− respectively (Wilke et al. 2011, this volume).
06_Backnaes_Deubener.indd 154
6/22/2011 5:15:11 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
155
figure 8. Total sulfur as a function of the J-parameter. Data: Filled and open circles, soda-lime-silica melts at 1370 °C, Papadopoulus (1973); Gray halftone squares, sodium silicate melts at 1200 °C, Holmquist (1966). Straight lines are best fit through data. PSO3 calculated from partial pressures in the input mixture.
Studies on incorporation of sulfur-bearing toxic and nuclear waste into glass matrices have mostly focused on chemically resistant borosilicate or phosphate-bearing melts that can incorporate higher amounts of sulfur than pure silicate melts. Therefore, in the papers from Bingham et al. (2007, 2008) the systems P2O5-Al2O3-Na2O-Fe2O3 + Na2SO4 and P2O5-Fe2O3RySO4 (R = Li, Na, K, Mg, Ca, Ba, Pb and y = 1, 2) were studied. The melt was stirred at a maximum melting temperature of 1100-1150 °C for 2 h. The composition of the glasses including its sulfur content was analyzed by EDS (energy dispersive X-ray spectroscopy), and the sulfur content of the glass was correlated to the oxide activity of the melt by comparing it to the cation field strength, optical basicity, oxygen to phosphorous molar ratio as well as P2O5content. The optical basicity has been used as a measure of the activity of free oxygen ions in the melt by probing the UV absorption bands (charge transfer bands) for electron transfer from oxygen atoms to cations (Duffy and Ingram 1971). The optical basicity L is defined as the relative shift of the absorption band frequency n of the probe cation (Pb2+, Tl+) in a glass relative to the uncomplexed “free ion” and crystalline “ionic oxide” reference state (CaO) as: L = (nfree ion − nglass) / (nfree ion − nCaO). Duffy (1993) proposed an empirial relationship to calculate the effective optical basicity for a multi oxide composition:
L= X AOa /2 L(AO a / 2 ) + X BOb/2 L(BOb / 2 ) + ... th
(10)
where L(AOa/2) and L(BOb/2) are the optical basicities of the oxides AOa/2 and BOb/2, respectively, and XAOa/2 and XBOb/2 are the mole fractions. The optical basicity gives information on the proportion of oxygen atoms the oxides bring to the melt. In Figure 9 a linear decline of the log sulfur retention as a function of P2O5 in the melt is observed, whereas the log sulfur retention scales positively with the calculated optical basicity Lth (Eqn. 10) but negatively with the mean field strength F. Bingham et al. (2008) assumed that all the iron was present as ferric iron. Additionally, analysis of sulfur retention as a function of both F and O/P atomic ratio were undertaken. An increase in the former was found to reduce the sulfur content, while an increase in the latter causes an increase. The authors however clearly state that the sulfur retention reported in their study do not represent the equilibrium between melt and sulfur salt, since this is rarely the state strived for during waste glass immobilization
06_Backnaes_Deubener.indd 155
6/22/2011 5:15:11 PM
156
Backnaes & Deubener
figure 9. Sulfur retention in phosphate melts at 1150 ± 25 °C as a function of a) P2O5-concentration and mean field strength F (insert), b) theoretical optical basicity calculated by Equation (12). Dashed lines are best fit through the data (Modified after Bingham et al. 2008).
experiments. The study was only concerned with how much sulfur is present in the melt after a controlled melting and cooling protocol, i.e., results which could be extrapolated on a larger scale of waste glass immobilization. Therefore, the results should not be taken as solubility values of sulfur as such. For borosilicate melts, one of the first studies on the vitrification of sulfur-bearing waste was published by McKeown et al (2001). The study investigated the structural incorporation of sulfur in the melt, and the sulfate content was given for Na2O-Fe2O3-Al2O3-B2O3-SiO2 melts containing various amounts of additional network modifiers such as CaO, Li2O, Cs2O, BaO, K2O and/or MgO. The sulfur was added as a mixture of SO2-O2-SO3 for 7 h at 1208 °C to the pre-melted base glass. The resulting glasses contained sulfur amounts (established through X-ray fluorescence measurements) in the range 0.028-0.6 wt%, with the highest sulfur content occurring in the melt with a CaO-content of nearly 25 wt%, and the lowest in a melt containing no additional network modifiers, i.e., the pure Na2O-Fe2O3-Al2O3-B2O3-SiO2 melt. Raman spectroscopy indicated that sulfur was in the sulfate form and no evidence of reduced sulfur species was reported. Sodium alkaline earth borosilicate glasses (SiO2-B2O3-Na2O-BaO-CaO) were investigated for the same purpose by Kaushik et al. (2006) and Mishra et al. (2008). The latter group showed that as much as 3 mol% SO42− (added as sodium sulfate and analyzed by microprobe analysis) can be incorporated in a melt of composition 40.42 SiO2-20.21 B2O3-21.22 Na2O-15.15 BaO (mol%) without phase separation. Above this concentration level BaSO4 crystallized in the melt. The batch was melted at 1000 °C for 4 h and no information was given if equilibrium was reached. With a sulfate content of 3 mol% in the melt it was also found that the glass transition temperature decreases by about 40 K compared to the sulfate-free melts and keeping the Si/ Na, Ba/Na, B/Na ratios constant. Structural changes in the glass matrix were studied through 29 Si and 11B MAS NMR spectroscopy. Though Mishra et al. (2008) concluded that at low to moderate sulfate concentration the SO42−-ion was acting as a network modifying ion, XANES spectra indicate that more likely (SO4)2− is an anion which is not directly bond to the silicate network, i.e., S-O-Si bonds are not very likely, see Wilke et al. (2011, this volume). Lenoir et al. (2009) studied the solubility of sulfur in borosilicate glasses using electron microprobe analysis and Raman spectroscopy. Two types of borosilicate glass compositions were investigated: a SiO2-B2O3-Na2O and a SiO2-B2O3-BaO system. Sulfur was added as Na2SO4 for the Na2O-containing melts and as BaSO4 for the BaO-bearing melts. In both
06_Backnaes_Deubener.indd 156
6/22/2011 5:15:12 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
157
experiments sulfate was added in large excess (10 wt%) to the expected sulfur solubility in the melt in equilibrium with either Na2SO4 or BaSO4. The base glasses were melted in Pt-Rh(10%)crucibles at 1100 °C (Na-bearing melt) or 1200 °C (Ba-bearing melt) for 2 h. A slow, albeit unspecified, heating regime was used in order to allow for decomposition of the borate, nitrate and carbonate components. In order to assure homogeneity, the melting procedure was repeated and subsequently the glass was ground to a powder and mixed with the various sulfur source salts and remelted at 1200 °C (Na-melt) or 1300 °C (Ba-melt) for 2 h and quenched. After quenching, the glasses were crushed and washed after experiment to remove crystalline phases, and to ensure that only matrix-bond sulfur was analyzed. Sulfate was identified and quantified by the vibration band at 990 cm−1 in the Raman spectrum of the glass. Quantitative electron microprobe analyses of selected samples were used for calibration of the integrated Raman intensity, and the results were verified using electron microprobe analysis. It was found that Na2SO4 was more easily incorporated into the sodium borosilicate than BaSO4 into a barium borosilicate. An explanation of this finding was not given. Furthermore, it was shown that the sulfate content of the glass (i.e., the intensity of the Raman band at 990 cm−1) decreases with increasing dwell time at 980 °C. This is in partial agreement with previous work by Ooura and Hanada (1998), where a decrease in sulfur content with increasing dwell time at 1350 °C was found. However in Ooura and Hanada (1998) no significant change in the sulfur content after a dwell time of 20 minutes was reported, while the sulfate content in similar experiments by other workers (Lenoir et al. 2009) showed a continuous decrease until a dwell time of 800 minutes. However, it needs to be emphasized that the experimental parameters were quite different, since Lenoir et al. (2009) used a pre-melted base glass, while Ooura and Hanada (1998) added the sulfur-bearing component directly to the batch.
SoLuBILIty of SuLfur IN NAturAL MELtS AND SyNthEtIc ANALogS In most solubility studies of sulfur in natural melts, the experiments are performed at high pressures to simulate the conditions in the Earth’s interior. A few papers however are concerned with the sulfur solubility at near-to 1 atm. These will be reviewed in this section to compare the data with those of technical melts as described above. Details on modeling the solubility of sulfur in magmas can be found in Baker and Moretti (2011, this volume). In 1974, Haughton et al. published a paper on the solubility of sulfur in anhydrous mafic magmas at 1200 °C. Fugacities of oxygen and sulfur were controlled by a mixture of SO2-CO2CO. Partial pressures were calculated from the input mixture using the method of White et al. (1958) and Heald et al. (1963). Equilibrium between gas and the silicate melt as well as between a sulfide-rich phase formed in the melt and the melt itself was reached, thus, sulfur solubility at sulfide saturation was reported. The sulfur content of the glass was investigated using electron microprobe analysis. Since the experiments were performed at low oxygen fugacity, the authors assumed that all sulfur was present as sulfide. One of the investigated parameters was the sulfur solubility as a function of FeO-content of the melt, and the results are presented in Figure 10. As shown, the effect of FeO-content on the sulfur solubility in mafic magmas is minor for FeOcontents lower than 10 wt%. However as the FeO-content increases above 10 wt%, the effect on the sulfur solubility becomes substantial. The result, is however, a consequence of the formation of iron sulfide in the melt, which buffers the sulfide solubility in the silicate melt (see Baker and Moretti 2011, this volume). Furthermore, it was stated by Haughton et al. (1974) that a decrease in temperature causes a substantial decrease in the melts capacity to retain sulfide. This is in agreement with the results of Fincham and Richardson (1954) (Fig. 5). At high sulfide contents, Haughton et al. found evidence for a phase separation in iron-rich melts with coexisting sulfide-saturated silicate melt and pure sulfide melt (as also shown in Fig. 3 of Baker and Moretti 2011, this volume). The
06_Backnaes_Deubener.indd 157
6/22/2011 5:15:12 PM
158
Backnaes & Deubener
figure 10. Total sulfur as a function of FeO-content in anhydrous mafic magmas at ambient pressure and 1200 °C (Modified after Haughton et al. 1974).
formation of the additional sulfide melt in melts of different FeO-content depends on fO2, and fS2. As the oxygen fugacity decreases, the phase separation occurs at increasingly lower sulfur concentrations and FeO-content. The sulfur fugacity shows the opposite pattern as a decrease in sulfur fugacity with a simultaneous decrease in FeO hinders a phase separation. In 1974 Katsura and Nagashima published a paper on solubility studies of tholeiitic basalt, hawaiite and rhyodacite at 1250-1300 °C at 1 atm. The oxygen and sulfur fugacities during the experimental run were established through a gas mixture of CO2-H2-SO2, and equilibrium between the gas and melt phases was assumed to be reached by comparing the thermodynamically calculated values for PO2 (White et al. 1958) with actual measurements using an ZrO2 sensor. Equilibrium within the gas phase was reached at 1250 °C and 1300 °C whereas at 1100 °C noticeable differences between measured and calculated PO2 were reported. Equilibrium between the melt and the gas phase was established after 3 to 6 h of dwelling at 1300 °C and 1250 °C respectively, i.e., the sulfur content in the melt as determined by the acid extraction method (Nagashima and Katsura 1973) after quenching, reached a constant value. Experiments were conducted with respect to oxygen fugacity, and a fairly similar picture as in Figure 1 was obtained (Fig. 11). Consistent with other studies it was found that the sulfur solubility of the melts is highly dependent on the oxygen partial pressure, and for all three melts a solubility minimum is located close to NNO, i.e., in the PO2 range = 10−7-10−5 bar at 1250 °C. Beyond the solubility minimum, i.e., at PO2 < 10−7 and PO2 > 10−5, the tholeiite and hawaiite show generally higher sulfur solubility than the rhyodacite, at both oxidized and reducing conditions. This finding can be understood by a higher degree of polymerization of the rhyodacite melt (see Fig. 13). The solubility of sulfur in synthetic tholeiitic melts as well as the immiscibility of sulfide was studied by Buchanan and Nolan (1979). The authors controlled the oxygen and sulfur fugacity through a CO2-CO-SO2-mixture. The sulfur content was analyzed by electron microprobe. In Figure 12 the sulfur retention in the melt is shown as a function of fS2 for different values of fO2. The general trend is that high sulfur retention is promoted by low fO2 and high fS2, as also stated by e.g., Fincham and Richardson (1954), Nagashima and Katsura (1973), Barbon et al. (1991), Beerkens and Kahl (2003).
06_Backnaes_Deubener.indd 158
6/22/2011 5:15:12 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
159
figure 11. Total sulfur as a function of oxygen partial pressure at 1250 °C and 2.1 vol% SO2 in the gas phase. Lines are intended as a visual guide. (Modified after Katsura and Nagashima 1974).
figure 12. Sulfur retention in synthetic tholeiitic melts at ambient pressure as a function of log(fS2(bar)) for different fO2. Lines are intended as a visual guide. (Modified after Buchanan and Nolan 1979).
06_Backnaes_Deubener.indd 159
6/22/2011 5:15:13 PM
160
Backnaes & Deubener
figure 13. Measured sulfur contents in glass inclusions trapped in minerals as a function of SiO2 content in comparison to sulfur solubility in synthetic binary sodium silicate melts (1200 °C) Data: Natural melts, Ducea et al. (1994); Synthetic melts, Holmquist (1966). Line connecting data is intended as visual guide.
Ducea et al. (1994) investigated the solubility of sulfur in calc-alkaline, alkaline and tholeiitic melts by analyzing the sulfur content in lava, pumice and melt inclusions as well as through analysis of volcanic gases. The analyses of melt inclusions to determine sulfur solubility are only partially useful since the conditions at which the inclusions were formed are often not well known, and inclusions might be affected by post-entrapment events (loss of volatiles, partial crystallization of melts). In order to investigate glasses that were close to complete saturation, volcanic glasses associated with a crystallized sulfur mineral and showing no evidence of degassing were preferred. The authors thus compiled data from the literature on sulfur solubility (Baldridge et al. 1981; Luhr et al. 1984; Byers et al. 1985; Capaccioni et al. 1987; Conticelli et al. 1987; Johnson et al. 1987; Drexler et al. 1990; Fournelle et al. 1990; Alt et al. 1993; Metrich et al. 1993; Allard et al. 1994; Herzig et al. 1994; Matthews et al. 1994; Gerlach et al. 1996). Part of the data analyzed by Ducea et al. (1994) can be seen in Figure 13. It is shown that the basaltic samples collected at Etna volcano in Italy can contain up to 7× more sulfur in comparison to the highly polymerized dacites and rhyolites from Julcani volcano in Peru. Despite the limitations with respect to equilibrium solubility when studying melt inclusions the compositional dependence of the natural melts shown Figure 13 is in good agreement with the solubility data of synthetic soda-silicate melts at saturation with liquid sulfate (Fig. 3). However, in the binary melts the sulfur solubility is approximately one order of magnitude higher than in the natural melts compiled by Ducea et al. (1994).
SuMMAry AND outLook Sulfur solubility is an important issue in glass technology. The demand for optimized sulfate fining and reduced SOx-emissions has motivated systematic studies on the impact of process parameters (temperature, atmosphere, raw materials) on sulfur solubility. In order to
06_Backnaes_Deubener.indd 160
6/22/2011 5:15:14 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
161
achieve an increase in load of sulfur-bearing waste by a vitrification process, understanding what parameters contribute to high sulfur solubility is of crucial importance. Furthermore, in the geosciences, sulfur-bearing lavas have motivated studies on sulfur solubility and speciation at near-ambient pressures. These studies used simple model glass systems, binary and ternary compositions that varied structural parameters over a wide range. Together with data of natural silicate glasses (multi-component glasses) and their analog compositions synthesized in the laboratory two general trends have been established: First, the sulfur solubility depends on sulfur speciation, i.e., high solubility is found under oxidizing condition (fO2 > NNO), when sulfur is dissolved in the melt structure as sulfate, and under reducing condition (fO2 < NNO) when sulfide is the dominant sulfur species in the melt. This behavior forms a typical V-shaped equilibrium solubility curve between a sulfur-bearing gas or gas mixture and a melt. However, for most data reported in the literature, deviations from the expected trend for sulfide and sulfate solubility, i.e., slopes d(log S)/d(log fO2) of −1.5 and +0.5, were evident, indicating either variable sulfur dioxide fugacity or contributions of other sulfur dissolution reactions. At moderate oxygen fugacity (fO2 ≈ NNO), however, the sulfur retention in all glass systems studied was extremely low, confirming the low stability of the sulfite species upon quenching the glass. Secondly, adopting Lewis’ type acids and base concepts, free oxygen ions were assumed to act as electron donors that equilibrate with bridging and non-bridging oxygen of the silicate network. Thus, basicity parameters such as NBO/T, J, L, and others can serve as a measure to quantify the activity of free oxygen ions that correlate positively with the sulfur solubility in the melt. These parameters account for both structural aspects the dimensionality of the silicate network and the type of bonds involved in the near range structure. Systematic studies that keep the former constant and vary the latter, or vice versa, have not been developed fully however, but are desirable to gain deeper insights into the compositional dependence of sulfur solubility with respect to the mechanisms of incorporation of sulfur species in the glass structure. Finally, since the sulfur solubility in most silicate melts is relative small (< 1 wt%), the consequences for melt dynamics (e.g. viscosity), the glass transition temperature and the physical properties of quenched glasses (e.g. refractive index), is poorly understood due to lack of experimental evidence. Following the compositional trends established in this chapter, silicate melts of high basicity, i.e., orthosilicate compositions composed of isolated silica tetrahedra (termed in glass technology termed as “invert glasses”) will provide the basis of structure-property relation studies, particularly with respect to the optical, chemical and mechanical properties, for sulfur dissolved in silicate melts, that is of considerable importance for future developments.
AkNowLEDgMENt The authors would like to acknowledge Deutsche Forschungsgemeinschaft (DFG) for the financial support for a project on sulphur properties in silicate melts under grant DE598/9-1 in the scope of which this paper was produced. We are especially grateful to the anonymous reviewers for their valuable comments and to H. Behrens for fruitful discussions and helping to improve the manuscript in the revision phase, as well as to Gordon Moore and Marie Edmonds for proof reading the text.
rEfErENcES Allard P, Carbonelle J, Metrich N, Loyer H, Zetwog P (1994) Sulphur output and magma degassing budget at Stromboli volcano. Nature 368:326-329 Alt JC, Shanks WC, Jackson MC (1993) Cycling of sulfur in subduction zones: The geochemistry of sulfur in the Mariana island arc and back-arc through. Earth Planet Sci Lett 119:477-494
06_Backnaes_Deubener.indd 161
6/22/2011 5:15:14 PM
162
Backnaes & Deubener
Backnaes L, Stelling J, Behrens H, Goettlicher J, Mangold S, Verheijen O, Beerkens RGC, Deubener J (2008) Dissolution mechanisms of tetravalent sulfur in silicate melts: Evidences from sulfur K edge XANES studies on glasses. J Am Ceram Soc 91:721-727 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Baldridge WS, Carmichael ISE, Albee AL (1981) Crystallization paths of leucite-bearing lavas: Examples from Italy. Contrib Mineral Petrol 76:321-335 Barbon F, Geotti-Bianchini F, Hreglich S, Scandellari S, Verita M (editors) (1991) Effect of the Batch Redox Number and Melting Temperature on the Redox Equilibria in Soda-Lime Industrial Glass. Proceedings of the First European Society of Glass Science and Technology Conference held in Sheffield, UK on 9-12 September 1991. Society of Glass Technology, 264 p. Beerkens RGC, Kahl K (2002) Chemistry of sulfur in soda-lime-silica glass melts. Phys Chem Glasses 43:189198 Beerkens RGC (2003a) Sulfate decomposition and sodium oxide activity in soda-lime-silica glass melts. J Am Ceram Soc 86:1893-1899 Beerkens RGC (2003b) Amber chromophore formation in sulfur- and iron-containing soda-lime silica glass. Glass Sci Technol 76:166-175 Behrens H, Stelling J (2011) Diffusion and redox reactions of sulfur in silicate melts. Rev Mineral Geochem 73:79-111 Bingham PA, Connelly AJ, Hand RJ, Hyatt NC, Northrup PA (2007) Incorporation and speciation of sulfur in glasses for waste immobilisation. Glass Technol Eur J Glass Sci Technol A 50:135-138 Bingham PA, Hand RJ (2008) Sulfate incorporation and glass formation in phosphate systems for nuclear and toxic waste immobilization. Mater Res Bull 43:1679-1696 Buchanan DL, Nolan J (1979) Solubility of sulfur and sulfide immiscibility in synthetic tholeiitic melts and their relevance to Bushveld-complex rocks. Can Mineral 17:483-484 Budd SM (1965) Oxidation-reduction equilibrium in glass a special reference to sulfur. ACS Symposium: Gases in Glass. 67th Ann Meeting of Am Ceram Soc, May 1965 Byers CD, Garcia MO, Muenow DW (1985) Volatiles in pillow rim glasses from Loihi and Kilauea volcanos, Hawaii. Geochim Cosmochim Acta 49:1887-1896 Capaccioni B, Nappi G, Renzulli A, Santi P (1987) The eruptive history of Vepe Caldera (Latera Volcano): A model inferred from structural and geochemical data. Periodico di Mineralogia 56:269-284 Chopinet M-H (2007) Influence of Na2O activity on the behaviour of sulfur in glass. Glass Technol Eur J Glass Sci Technol 50:117-120 Conticelli S, Francalanci L, Manetti P, Peccerillo A (1987) Evolution of the Latera Volcano, Vulsinian District (Central Italy): Stratigraphic and petrological data. Periodica di Mineralogia 56:175-200 Ducea MN, McInnes BIA, Wyllie PJ (1994) Sulfur variations in glasses from volcanic rocks: Effect of melt composition on sulfur solubility. Inter Geol Rev 36:703-714 Duffy JA, Ingram MD (1971) Establishment of an optical scale for Lewis basicity in inorganic oxyacids, molten salts, and glasses. J Am Ceram Soc 93:6448-6454 Duffy JA (1993) A review of optical basicity and its applications to oxide systems. Geochim Cosmochim Acta 57:3961-3970 Drexler JD, Munoz JL (1990) Recent advances in geology of granite-related mineral deposits. In: Granitic Magmatism and Related Mineralization. Ishihara S, Takenouchi S (eds) Society of Mining Geologists of Japan, Mining Geology, spec issue 8:72-79 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Fincham CJB, Richardson FD (1954) The behaviour of sulfur in silicate and aluminate melts. Proc Royal Soc London A 223:40-62 Flood H, Förland T (1947) The acids and basic properties of oxides. Acta Chem Scand 1:592-604 Fournelle J (1990) Anhydrite in Nevado del Ruiz November 1985 pumice: relevance to the sulfur problem. J Volcan Geotherm Res 42:189-201 Gerlach TM, Westrich HR, Symonds RB (1996) Pre-eruption vapor in magma of the climactic Mount Pinatubo eruption: Source of the giant stratosphere sulfur dioxide cloud. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CT, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, Quezon City and University of Washington Press, p 415-433 Heald EF, Naughton J., Barnes IL (1963) The chemistry of volcanic gases. Use of equilibrium calibrations in the interpretationof volcanic gas samples. J Geophys Res 68:545-557 Haughton DR, Roeder PL, Skinner BJ (1974) Solubility of sulfur in mafic magmas. Econ Geol 69:451-467 Herzig P, Hannington M, McInnes B, Stoffers P, Vilinger H, Seifert R, Binns R, Liebe T (1994) Submarine volcanism and hydrothermal venting studied in Papua, New Guinea. EOS (Trans Am Geophys Union) 75:513-516
06_Backnaes_Deubener.indd 162
6/22/2011 5:15:14 PM
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure
163
Holmquist S (1966) Oxygen activity and the solubility of sulfur trioxide in sodium silicate melts. J Am Ceram Soc 49:467-473 Johnson RW, Jaques AL, Langmuir CH, Perfit MR, McColloch MT, Staudigel H, Chapell BW, Taylor SR (1987) Ridge subduction and forearc volcanism: Petrology and geochemistry of rocks dredged from the western Solomon arc and Woodlark basin. In: Marine Geology, Geophysics, and Geochemistry of the Woodlark Basin, Solomon Islands. Taylor B, Exon N (eds) Circum-Pacific Council for Energy and Mineral Resources Earth Science series 38:517-531 Jugo PJ, Wilke M, Botcharnikov RE (2010) Sulfur K-edge XANES analysis of natural and synthetic basaltic glasses: Implications for S speciation and S content as function of oxygen fugacity. Geochim Cosmochim Acta 74:5926-5938 Katsura T, Nagashima S (1974) Solubility of sulfur in some magmas at 1 atmosphere. Geochim Cosmochim Acta 38:517-531 Kaushik CP, Mishra RK, Sengupta P, Kumar A, Das D, Kale GB, Raj K (2006) Barium borosilicate glass - a potential matrix for immobilization of sulfate bearing high-level radioactive liquid waste. J Nucl Mater 358:129-138 Klouzek J, Arkosiova M, Nemec L (2006) Redox equilibria of sulfur in glass melts, Ceramics-Silikaty 50:134139 Klouzek J, Arkosiova M, Nemec L, Cincibusova P (2007) The role of sulfur compounds in glass melting. Glass Technol Eur J Glass Technol A 48:28-30 Kordes E, Zöfelt B, Pröger H (1951) Die Mischungslücke im flüssigen Zustand zwischen Na-Ca Silicaten und Na2SO4. Z Anorg Allg Chem 264:255-271 Lehmann J, Nadif M (2011) Interactions between metal and slag melts: steel desulfurization. Rev Mineral Geochem 73:493-511 Lenoir M, Grandjean A, Poissonnet S, Neuville DR (2009) Quantification of sulfate solubility in bororsilicate glasses using Raman spectroscopy. J Non-Cryst Solids 355:1468-1473 Lewis JN (1923) Valence and the Structure of Atoms and Molecules. Chemical Catalog Company, New York Luhr JF, Carmichael ISE, Varenkamp JC (1984) The 1982 eruptions of El Chichon volcano, Chiapas, Mexico: Mineralogy and petrology of the anhydrite-bearing pumices. J Volcanol Geotherm Res 23:69-108 Lux H (1939) “Säure” und “Basen” im Schmelzfluss: Bestimmung der Sauerstoffionen-Konzentration. Z Elektrochem 45:305-309 Maekawa H, Maekawa T, Kawamura K, Yokokawa T (1991) The structural groupsodf alkali silicate glasses determined from 29Si MAS-NMR. J Non-Cryst Solids 127:53-67 Manring WH, Diken GM (1980) A practical approach to evaluating redox phenomena involved in the meltingfining of soda-lime glasses. J Non-Cryst Solids 38&39:813-815 Manara D, Grandjean A, Pinet O, Dussossoy JL, Neuville DR (2007) Sulfur behaviour in silicate glasses and melts: Implications for sulfate incorporation in nuclear waste glasses as a function of alkali cation and V2O5-content. J Non-Cryst Solids 353:12-23 Manara D, Grandjean A, Neuville DR (2009) Structure of borosilicate glasses and melts: A revision of the Yun, Bray and Dell model. J Non-Cryst Solids 355:2528-2531 Matthews SJ, Jones AP, Gardeweg MC (1994) Lascar volcano, Northern Chile: Evidence for steady-state disequilibrium. J Petrol 35:401-432 McKeown DA, Muller IS, Gan H, Pegg IL, Stolte WC, Schlachter AS, Shuh DK (2001) Raman studies of sulfur in borosilicate waste glasses: sulfate environments. J Non-Cryst Solids 288:191-199 Metrich N, Clocchiatti R, Mosbah M, Chaussidon M (1993) The 1989-1990 activity of Etna magma mingling and ascent of a H2O-Cl-S rich basaltic magma: Evidence from melt inclusions. J Volcanol Geotherm Res 59:131-144 Mishra RK, Sudarsan KV, Sengupta P, Vatsa RK, Tyagi AK, Kaushik CP, Das D, Raj K (2008) Role of sulfate in structural modification of sodium barium borosilicate glasses developed for nuclear waste immobilization. J Am Ceram Soc 91:3903-3907 Mysen BO, Frantz JD (1994) Alkali silicate glass and melt structure in the temperature range 25-1651 °C at atmospheric pressure and implications for mixing behaviour of structural units. Geochim Cosmochim Acta 65:2413-2431 Müller-Simon H (1999) Sulfatläuterung in Kalk-Natron-Silicatgläsern. In: Grundlagen des Industriellen Glasschmelzprozesses. Condrat R (ed) Verlag der Deutschen Glastechnischen Gesellschaft, Frankfurt/M., p 45-72 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Müller-Simon H, Glitzhofer K (2008) Sulphur mass flow balance in industrial melting furnaces. Glass Technol Eur J Glass Sci Technol A 49:83-90 Nagashima S, Katsura T (1973) The solubility of sulfur in Na2O-SiO2 melts under various oxygen partial pressures at 1100 °C, 1250 °C and 1300 °C. Bull Chem Soc Japan 46:3099-3103 Ooura M, Hanada T (1998) Compositional dependence of solubility of sulfate in silicate glasses. Glass Technol 39:68-73
06_Backnaes_Deubener.indd 163
6/22/2011 5:15:14 PM
164
Backnaes & Deubener
Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Papadopoulos K (1973) The solubility of SO3 in soda-lime silica melts. Phys Chem Glasses 14:60-65 Pearce ML, Beisler JF (1965) Miscibility gap in the system sodium oxide-silica-sodium sulfate at 1200 °C. J Am Ceram Soc 48:40-42 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Schreiber HD, Kozak SJ, Leonard PG, McManus KK (1987) Sulfur chemistry in a borosilicate melts. Part 1: Redox equilibria and solubility. Glasstech Ber 60:389-398 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Simpson W, Myers DD (1978) The redox number concept and its use by the glass technologist. Glass Technol 19:82-85 Toop GW, Sammis CS (1962) Activities of ions in silicate melts. Trans Metall Soc AIME 224:878-887 White WB, Johnson SM, Dantzig GB (1958) Chemical equilibrium in complex mixtures. J Chem Phys 28:751755 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78
06_Backnaes_Deubener.indd 164
6/22/2011 5:15:14 PM
06_Backnaes_Deubener.indd 165
addition of sodium sulfate to glass batch addition of sodium sulfate to glass batch
10 CaO 16 Na2O 74SiO2
72.7 SiO2 1.5 Al2O3 9.7 CaO 2.6 MgO 13 Na2O 0.3 K2O
reached / gasd
5.3 4.1 5.7
0.066 0.066
3.4
3.4
0.34
0.009
0.26
3.4
5.3
5.2
5.0
4.7
Notes: a Sulfur solubility at saturation with SO2-bearing gas was reached for dwell times > 4.5 h at 1500 °C. Run duration of all samples was 6 h. b Sulfur solubility at saturation with SO2-bearing gas was reached for dwell times > 5 h at 1250 °C. Run duration of all samples > 5 h. c Information on change in sulfur content with variations in dwell time not provided. d The saturation of the melt by SO2 was assumed as the specific amount of released SO2 exceeded its solubility in the melt.
25 Na2O 75 SiO2
1500 1300
1400
treatment in SO2bearing gas mixtures
33.3 Na2O 66.7 SiO2
50 Na2O 50 SiO2 4.4
0.22 not reached reached / gasb not reached reached / gasb not reached reached / gasb probably not reachedc
20 Na2O 80 SiO2 1100 1250 1100 1250 1100 1250
0.38
0.73
25 Na2O 75 SiO2
28.6 Na2O 71.4 SiO2
reached / liqiud
1.62
1250
4.6
2.38
36.4 Na2O 63.6 SiO2
33.3 Na2O 66.7 SiO2 treatment in SO2bearing gas mixtures
4.5
3.30
39.4 SiO2 43.3 CaO 17.3 Al2O3
40 Na2O 60 SiO2
3.8 4.2 5.1
0.003 0.005 0.015
reached / gasa reached / gasa reached / gasa
1425 1500 1550
treatment in SO2bearing gas mixtures
DNNo
max. total sulfur (wt%) oxidized
Equilibrium / saturation
T ( °c)
Method
glass/melt
0.059 0.059
0.18
0.014
0.051
0.463
0.120 0.283 0.113
max total sulfur (wt%) reduced
-0.6 -0.8
-1.9
-5.4
-5.4
-5.4
-5.3 -6.5 -5
DNNo
Klouzek et al. (2006)
Beerkens and Kahl (2002)
Nagashima and Katsura (1973)
Holmquist (1966)
Fincham and Richardson (1954)
reference
Appendix table 1. Compilation of experimental data on solubility and retention of sulfur in silicate melts at ambient pressure with respect to sulfur speciation. “Oxidized” and “reduced” refers to sulfate and sulfide solubility respectively. Oxygen fugacity is specified relative to the log fO2 of the nickelnickel oxide equilibrium DNNO.
Sulfur Solubility in Silicate Melts at Near-Atmospheric Pressure 165
6/22/2011 5:15:14 PM
06_Backnaes_Deubener.indd 166
6/22/2011 5:15:14 PM
7
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 167-213, 2011 Copyright © Mineralogical Society of America
Modeling the Solubility of Sulfur in Magmas: A 50-Year Old Geochemical Challenge Don R. Baker Earth and Planetary Sciences, McGill University 3450 rue University Montréal, Quebec, H3A 2A7, Canada and Sincrotrone Trieste S.C.p.A. di interesse nazionale Strada Statale 14 - km 163,5 in AREA Science Park 34149 Basovizza, Trieste, Italy
[email protected]
Roberto Moretti Centro Interdipartimentale di Ricerca in Ingegneria Ambientale (CIRIAM) & Dipartimento di Ingegneria Civile Seconda Università degli Studi di Napoli and Istituto Nazionale di Geofisica e Vulcanologia, sezione Osservatorio Vesuviano Via Diocleziano 328 80124 Napoli, Italy
[email protected]
IntRODuCtIOn There are myriad reasons why we wish to understand the behavior of sulfur in magmatic systems, reasons that vary from pure intellectual curiosity to possible impacts on society and its resources. Since ancient times sulfur has been associated with volcanic activity, and the role of sulfur in the formation of ore deposits has long been recognized because of the necessity of metal ores for our modern life-style (e.g., Barnes 1979; Naldrett 1989; Simon and Ripley 2011, this volume). Recently the mechanisms and quantities of sulfur freed from natural magmas have become an important environmental issue due to their potential effects on global climate change. For example, the average annual volcanic SO2 emission rate of 7.5 to 10.5 teragrams (Tg) per year (Halmer et al. 2002) may contribute 10% of the global atmospheric sulfur input (Halmer et al. 2002; Smith et al. 2004), and individual eruptive episodes can rapidly contribute gigantic sulfur loads to the atmosphere, 100’s to 1000’s of Tg, depending on the scale of the eruption (Self 2006). Such sulfur emissions can produce potentially catastrophic local and global changes (e.g., Fedele et al. 2003; Ward 2009); Courtillot and Rennes (2003) correlated the timing of flood basalts with extinction events in Earth’s history and hypothesized a causal relation. Part of the kill mechanism responsible for extinction may be volcanically derived sulfur creating anoxic oceans and another part of the mechanism may be climatic changes brought about by sulfur injection into the atmosphere (Ward 2009). Indeed, Erwin (2006) advocates that sulfur released from the eruption of the Siberian Flood basalts played a role in the end-Permian extinction. In light of the evidence that volcanic degassing is a significant source of sulfur to the atmosphere (Stoiber et al. 1987; Symonds et al. 1994; Andres and Kasgnoc 1998; Graf et al. 1529-6466/11/0073-0007$05.00
DOI: 10.2138/rmg.2011.73.7
168
Baker & Moretti
1998; Halmer et al. 2002; Stevenson et al. 2003; Wallace 2005), there is a need to understand the mechanisms and quantities of sulfur released into the atmosphere from volcanic activity. It is essential that we understand whether the sum of anthropogenic and magma-derived sulfur can partially compensate the effect of CO2 on global climate change. For example, Crutzen (2006) suggested injecting one to two teragrams of sulfur per year into the stratosphere to mitigate global warming, an amount that could be naturally emitted by a small eruption of 1-3 km3 of dense rock. To completely understand the magmatic portion of the global sulfur cycle we must know the mechanisms by which sulfur is transported in magmas (in fluid or gas, silicate melt, sulfide melt, crystal) and the quantities of sulfur that can be stored in magmatic reservoirs. Experimental studies of sulfur’s behavior in silicate melts were begun by metallurgists in the first half of the 20th century. Investigations in geologically relevant systems have been performed since the 1960’s; however, experimentalists cannot study all possible melt and rock compositions at all conditions. Thus, to bridge the gap between the studied compositions and conditions, and to hopefully extrapolate beyond them, quantitative models of sulfur’s behavior in magmatic systems have been constructed by many authors. Such models, although originally designed for application in geological environments can also be used to investigate the behavior of sulfur in industrial silicate melts, such as slags associated with production of steel or container glasses (for more details about application of thermodynamic modeling in technical applications see chapters in this volume by Falcone et al. 2011, Lehmann and Nadif 2011, and Müller-Simon 2011). The ideal model allows a scientist to calculate the fugacity of sulfur in a magmatic system or a furnace and the saturation state of the melt with respect to sulfide and sulfate. We begin with a discussion of modeling philosophies and thermodynamics. This section is followed by a history of experimental studies and empirical models that provided important constraints on the saturation of silicate melts with either a sulfide or sulfate phase; then we demonstrate their limitations and discuss the development of more rigorous thermodynamic models. Example applications of both types of models to natural systems are presented; however, each model has its benefits and its drawbacks. We conclude that at the present time even the most rigorous models for sulfur’s behavior in silicate melts and sulfide or sulfate saturation have limitations that must be recognized by users if they are to obtain reliable constraints on the storage and transport of sulfur in magmatic systems.
tHERMODYnAMICS AnD MODELInG SuLFuR SOLuBILItY In MAGMAtIC SYStEMS A few necessary definitions and concepts Before we begin our review of models for sulfur dissolution in silicate melts it is necessary to clarify certain terms and conditions that have been repeatedly misused and misunderstood in the geological literature. We apologize for this rather pedantic beginning, but the question of terminology is critical and many of terms discussed below will appear again and again in this contribution. The first term to define is saturation. Saturation refers to the condition at which two phases coexist with each other at equilibrium, e.g., a liquid saturated with a gas or fluid, where the term gas refers to a low density vapor phase and fluid refers to volatile-dominated phase whose density approaches that of silicate melts, which occurs at pressures above the liquid-vapor critical end point. The strict thermodynamic definition of a silicate melt saturated with sulfur requires the equilibrium coexistence of a silicate melt and a phase (solid, liquid or fluid) composed of pure sulfur. This occurrence is so rarely achieved that use of the term “sulfur saturated” should be avoided, except when it accurately describes equilibrium with a
Modeling the Solubility of Sulfur in Magmas
169
phase composed of pure sulfur. Explicitly providing the names of the two phases coexisting at equilibrium such as a gas-saturated melt, a fluid-saturated melt, a pyrrhotite-saturated silicate melt, or a sulfide melt-saturated silicate melt is the correct usage. Importantly, an element does not saturate another phase, unless that element constitutes a pure phase. However, it is common practice in Earth sciences to identify sulfur in magmatic systems as some kind of undistinguished component and to discuss “sulfur saturation.” This comes mainly from the tendency, for example in volcanology, to talk about, and work on, volatile components such water, carbon dioxide, chlorine and, of course, sulfur. When dealing with equilibrium between a sulfur-bearing fluid and a silicate melt (i.e., a fluid-saturated melt) scientists are often interested in the solubility of sulfur, which can be defined as the equilibrium concentration of sulfur in a phase (in the case of this chapter a silicate melt) under a given set of thermodynamic parameters and coexisting phases. Alternatively, scientists wish to know the partitioning of sulfur between the fluid and the melt, because as the sulfur concentration in the fluid changes, so must the sulfur concentration in the melt coexisting with that fluid at equilibrium. These concentration changes are the direct reflection of the equilibrium phase j requirement that the chemical potential, m component i , of a component (e.g., sulfur) must be equal in all coexisting phases. Partitioning can be expressed simply as the concentration (often in weight fraction or percentage) of sulfur in the fluid divided by the concentration of sulfur in the melt, which is commonly abbreviated as Dfl/m. The application of this type of partition coefficient is founded upon Henrian behavior (i.e., a linear relationship between component activity and concentration) often seen for trace elements in silicate crystals and melts. However, the concentration range over which sulfur behaves in a Henrian manner in silicate melts and geological fluids remains to be adequately defined, so such simple partition coefficients must be applied with caution. Alternatively, mole fractions of sulfur in the melt and the fluid can be used to define a molar, or Nernstian, partition coefficient based upon a sulfur exchange reaction between the melt and the fluid. One advantage of Nernstian partition coefficients is that in some cases they may be valid at concentrations higher than the Henrian limit. Additionally, Nernstian partition coefficients are a natural outgrowth of any thermodynamic modeling of elemental partitioning. In many circumstances we are interested in the concentration of sulfur in a silicate melt that is saturated with a sulfide phase, the sulfur concentration at sulfide saturation, abbreviated as the SCSS. The sulfide with which the silicate melt is saturated can be either crystalline, commonly pyrrhotite, or a liquid, commonly dominated by sulfur and iron in approximately a 1-to-1 molar ratio, but either of these phases will also contain other metals and oxygen. As discussed below, at high oxygen fugacity, fO2, the melt can become saturated with sulfate (commonly anhydrite) and in this case the sulfur concentration in the melt at anhydrite saturation can be defined, the SCAS. Although the concentrations of sulfur in melts saturated with either a sulfide or sulfate phase have been referred to as the sulfur solubility, use of the terms SCSS or SCAS are more thermodynamically correct because the reaction involved is not one simply of elemental sulfur dissolution into a silicate melt.
Modeling philosophies Ultimately all models for the saturation of silicate melts with a sulfur-bearing phase and the partitioning of sulfur between two coexisting phases are based upon experimental data. The data can vary from thermodynamic measurements of enthalpies, entropies and heat capacities to direct measurements of the partitioning of sulfur between two phases. In order to take the observed data and use it to create a general model for computing the partitioning of sulfur among the phases of interest (melt(s), fluid(s), crystals), we must establish some quantitative laws to model how sulfur dissolves in each phase depending on the system parameters (e.g., composition, temperature, pressure and related variables). But, before creating a model from the measured data the researcher must explicitly or implicitly define a modeling philosophy.
Baker & Moretti
170
We consider two philosophical end-members for modeling partitioning and solubility for any component in any phase based upon experimental measurements used for calibration. The first end-member is predominantly empirical and can contain any variables the modeler wishes to include to fit the experimental observations. Although models approaching this endmember often contain thermodynamic variables, such as pressure, P, and temperature, T (all temperatures in this contribution are in K unless otherwise specified), many of the variables are chosen based only upon the quality of the fit they provide to the data. In principle this endmember requires no constraints and variables are chosen solely based upon the statistical fit to the data. In this case predictions will be strictly valid only within the compositional domain used for calibrating the empirical function, but the model cannot be safely extrapolated to pressure-temperature-compositions conditions outside its calibration space. The second end-member models the chemical behavior based upon a more-or-less rigorous thermodynamic treatment defined by chemical reactions and constrained by as many thermodynamic quantities (e.g., enthalpies, entropies, etc.) as possible. In principle, if all thermodynamic data (volume, enthalpy, entropy, heat capacity) for all components in all phases involved in the reactions were available, no solubility and saturation experiments would be necessary. The application of the thermodynamic principles of Gibbs free-energy minimization and equality of chemical potentials of components between phases at equilibrium would be strictly sufficient to model phase diagrams and, coupled with mass balance, the partitioning of the element of interest, such as sulfur, between phases within the system. Currently, we are far from having the necessary data for this direct approach in geological systems and thermodynamic models for sulfur solubility rely upon the same data sets as empirical models. However, even thermodynamic end-member models retain some aspects of the empirical philosophy because the choice of the chemical components and reactions used in the models are dictated by experimental results. The large number of variables sometimes found in empirical models often results in better fits (statistically lower residuals) than thermodynamically based models. However, even though empirical models may provide a better fit to the measured data, we stress that they may not accurately interpolate between measured data points or extrapolate beyond the data. The advantage of thermodynamic models is that they should be easily and correctly interpolated and extrapolated, but possibly at the expense of more exact fits to the data. Most models discussed below fall between these two end-members and display the advantages, and disadvantages, of both end-members. Because of the absence of well-established calorimetric or volumetric studies, the use of experimental measurements of mass partitioning (e.g., from solubility/saturation or other phase equilibria experiments) is mandatory for calibration. The philosophy for constructing a model from experimental results is summarized in the flowchart of Figure 1, which demonstrates the two-way link between theory and experiments. Much caution is needed in the construction of models, since reliable and accurate models must be provided along with: 1)
equations from theories of general validity, avoiding purely empirical fitting parameters as much as possible;
2)
accurate procedures of numerical analysis, in order to find the stable and physicallysound mathematical conditions that minimize uncertainties;
3)
reliable data should be uniformly spread all over the P-T-X domain of interest (where X represents composition) in order to maximize the quality of the model and minimize any interpolations or extrapolations. The best practice would be to use compositional data within their Euclidean sample space (simplex), in order to
and
Modeling the Solubility of Sulfur in Magmas Experiments
Model verification and implementation
171
Check of consistency
Analysis of the simplex structure
Applications to natural magmas
Figure 1. Flowchart for any modeling procedure based on the calibration of unknown thermodynamic parameters using the information contained in solubility data.
better evaluate their statistics out of the cloud of spurious effects due, for example, to closure (e.g., Pawlowsky-Glahn and Egozcue 2006). This can be achieved by working with chemical ratios.
thermodynamic basis for dissolution of sulfur in silicate melts Although a totally empirical end-member model requires nothing more than experimental data and a computer program to perform a fit, the foundation of a thermodynamic model is much more detailed. To begin the creation of a model, one, or more, chemical reactions must be constructed. As an example of this consider the classical reaction for sulfur dissolution into a silicate melt at fO2’s below the nickel-nickel oxide, NNO, buffer (for a summary of geological oxygen fugacity buffers see Frost 1991) and 1 bar pressure: 1 gas 1 S2 + O2 − , silicate melt ↔ O2gas + S2 − , silicate melt 2 2
(1)
O2− and S2− are shorthand notations for the ionic species containing all of the dissolved components of the type M2/vv+O and M2/vv+S in the silicate melt, where Mv+ is a generic cation of valence v+. Following basic thermodynamics, at equilibrium the change in Gibbs free energy of the reaction must equal zero, DG = 0, and DG equals 1 2
( ) (a 0 = ∆G = ∆H − T ∆S + ∫ P∆V + RT ln (a ) (a aOgas2−
1 gas 2 S2
silicate melt S2 −
silicate melt O2 −
) )
(2)
where DH is the enthalpy of the reaction, DS is the entropy of the reaction, DV is the volume change of the reaction, R is the gas constant and the ai j’s in the natural logarithm term are the activities of the different components (subscript i) in the different phases (superscript j). The natural logarithm term in Equation (2) is, of course, the equilibrium constant of Equation (1), K(1): 1 gas 2 O2
silicate melt S2 −
1 gas 2 S2
silicate melt O2 −
(a ) (a K (1) = (a ) (a
) )
(3)
Baker & Moretti
172
The enthalpic, entropic and volume terms in Equation (2) are those for the pure end-member components, S2gas, O2gas, M2/vv+O and M2/vv+S, in the reaction. The activities take into account any excess Gibbs free energy of mixing (excess heats and volumes of mixing). In the limiting case of ideal mixing, where the only contribution of mixing to the energy of the reaction is the entropy of mixing, the activities can be replaced by the mole fraction of the components in the phase raised to the site multiplicity, n, e.g., aij = (Xij)n. In a binary silicate melt composed of forsterite (Mg2SiO4) and fayalite (Fe2SiO4) components the site multiplicity for Mg or Fe could be modeled with a value of 2, however because the site multiplicity cannot be unambiguously identified in silicate melts (unlike crystals) it is considered equal to 1 and n is dropped from this equation relating activity and composition. Unfortunately, most mixing in silicate melts is not ideal and aij = (giXij), where gi is the activity coefficient, which may be compositionally independent at low concentrations (Henrian behavior), but typically is a complex function of concentration. The thermodynamic approach demands the identification of activity-composition relationships and the reactions between the components in the system. These conditions can be expressed by considering the example of SO2 and H2S partitioned between two phases, gas and melt. Considering first SO2, we write the equation for dissolution into the melt SO2gas ↔ SO2silicate melt
( 4a )
from which we can write the thermodynamic equation: gas gas gas o , gas µSO + RT ln aSO = µSO = RT ln fSO 2 2 2 2 silicate melt silicate melt = µSO = RT ln fSO 2 2
( 4b)
o , silicate melt silicate melt = µSO + RT ln aSO 2 2
where fSO2gas is the fugacity of SO2. It should be noted that fugacity corresponds to activity, which is by definition the ratio of fugacity to the standard state fugacity (f/f°), when the 1 bar, T of interest standard state is chosen such that fSO2° = 1. We can rearrange Equation (4b) to calculate the fugacity: gas gas o , silicate melt o , gas silicate melt silicate melt fSO = ϕSO2 XSO P = µSO − µSO + RT ln γ SO XSO2 2 2 2 2 2
( 4c )
where jSO2 and XSO2gas are the fugacity coefficient and mole fraction of SO2 in the gas phase, and XSO2silicate melt is its mole fraction in the melt phase. Treating H2S in a similar manner yields H 2Sgas ↔ H 2Ssilicate melt
(5a )
melt silicate melt silicate melt fHgas = ϕH2 S X Hgas2 S P = µ oH,2silicate − µ oH,2gas X H2 S S S + RT ln γ H 2 S 2S
( 5b )
and
These equations must be constrained by mass balance, e.g.:
∑S = n S 1
silicate melt
+ n2SO2gas + n3H 2Sgas
(6)
where SS is the total moles of S in the system and the ni’s in Equation (6) are the number of moles of each species or component. The challenge in solving the equations is to determine the activity coefficients and fugacity coefficients of each species in the system. The term fugacity applied here has a general thermodynamic meaning, representing a chemical potential. This does not necessarily imply that a free gas phase must exist in order to define fSO2 and fH2S, by analogy with oxygen fugacity, fO2, which can be defined and used in fluid-undersaturated systems as well. The misconception that the use of fugacity requires a free gaseous or fluid phase has often led to serious misunderstandings about the applicability of the
Modeling the Solubility of Sulfur in Magmas
173
fugacity of a component in fluid-undersaturated systems. Indeed the fugacity of any thermodynamic species, including those normally considered as solids, can be rigorously defined in the absence of a free gaseous or fluid phase and used in thermodynamic calculations. Another way to visualize the fugacity concept is by considering a virtual fluid that coexists with a melt; the fugacity corresponds to the activity of the component of interest in the virtual fluid. The thermodynamic treatment is much more complicated if additional sulfur-bearing phases such as pyrrhotite, anhydrite, or even sodalite, are present in the system (cf. Parat et al. 2011, this volume). The sulfur-bearing components are no longer limited only to SO2 and H2S, but increase in the presence of solid phases, especially if these are solid-solutions (e.g., FeS-NiSCuS), and additional equilibria must be considered, for example: FeSsulfide ↔ FeSsilicate melt
(7a )
and, for that matter, one could define reactions between any FeS in the gas phase with either the silicate melt or the sulfide: FeSgas ↔ FeSsilicate melt and
FeSgas ↔ FeSsulfide
(7b)
And, of course, the mass balance Equation (6) must be modified to account for the presence of sulfur in the additional phases. However, two major problems arise when using the classical thermodynamic approach based on macroscopic components. The first problem is the difficulty of expressing the mixing properties between components, especially in the melt phase, in such a way that classical mixing models (e.g., Margules parameters) involving oxides can reproduce the shape of the Gibbs free energy of the mixture, typically characterized by a cuspate minimum (e.g., Ottonello 2001). The second problem is that we lack many of the thermodynamic properties (enthalpy, entropy, heat capacity, volume) at the conditions of interest, which span a wide range of T, P and composition. For example, knowledge of volume data, particularly the partial molar volumes in mixtures, is fundamental for determining the effect of pressure, particularly when volatile components are involved and a fluid phase is present. But such data are currently unavailable for sulfur species in silicate melts and coexisting fluids. The thermodynamic approach does not necessarily contain any structural implications concerning the components and phases under investigation. An example of such an approach is the non-ideal approach, based on a series expansion of the interaction parameters (e.g., Margules parameters, Wij), between the hypothesized species in the melt. If we truncate the series expansion after the first term, the free energy of mixing is expressed as: n −1
G mixing excess = N ∑
n
∑XXW i
j
ij
(8)
i=1 j=i+1
where N is the sum of the number of moles of each component, ni, and taking the derivative of this excess Gibbs free energy of mixing with respect to the moles of the component of interest yields the activity coefficient: ∂G mixing excess = RTlnγ i ∂ni P,T,n j ≠i
( 9)
This very general approach was introduced by Ghiorso et al. (1983) to model melt-mineral phase equilibria using species defined by end-member mineral compositions. For some volatile species, the adoption of this classical thermodynamic procedure works well for calibration based upon experimentally determined element partitioning. This type of model accurately predicts water and carbon dioxide solubility in silicate melts (Papale et al. 2006), where water in the melt can be treated as a single thermodynamic component (irrespective of how it
Baker & Moretti
174
speciates between OH− groups and molecular H2O in the melt), and all carbon in the melt can be attributed to CO2 (irrespective of the ratio of molecular CO2 to CO32− groups dissolved in the melt). Other H- and C-bearing components can be neglected because their concentrations in silicate melts at shallow depths and typical igneous fO2’s are very small and their abundances in the fluid minimal. The calibration of thermodynamic models involving silicate melts implicitly demands assumptions or approximations about the nature of the melt components necessary to describe equilibria with solid or gas phases. Consider sulfur solubility when a H-, O-, C- and S-bearing fluid phase is in contact with a silicate melt. We can to a first approximation neglect the solubility of S2, SO, and SO3, in the melt and limit our attention to the most important gaseous species, H2S and SO2. However, we cannot establish the same correspondence we did for H2O and CO2 components and easily attribute a proportion of dissolved sulfur to H2Ssilicate melt and SO2silicate melt in order to retrieve thermodynamic data and calibrate models on the basis of fSO2 and fH2S (Eqns. 4 and 5). The correspondence fails because with increasing sulfur in the system at fO2’s near the NNO oxygen fugacity buffer the eventual saturation of the silicate melt with a liquid or solid sulfide solid solution (e.g., monosulfide solution, MSS) will complicate the matter, even if we oversimplify the situation and assume a very simple scheme in which only a FeS component is considered. Additionally this approach is complicated by the dissolution of SO2 as sulfide and/ or sulfate ions in the melt, which is poorly treated by classical mixing models. A simple example of the thermodynamic approach is to take the equilibrium constant from a reaction of interest and fit it to measurements. For example, let us consider the exchange reaction FeO silicate melt +
1 fluid 1 S2 ↔ FeSsulfide melt or po + O2fluid 2 2
(10)
where the superscript po stands for pyrrhotite. The equilibrium constant of Equation (10), at the T and P of interest is: 1 fluid 2 O2
(a ) logK (10) = log (a )
1 fluid 2 S2
sulfide melt or po aFeS 1 ⋅ silicate melt = logK (10)1,T − aFeO 2.303RT
∫ ( ∆V
melt
)
+ ∆V fluid dP
(111)
and relates the activities at the P,T of interest with the 1 bar, T of interest standard state (denoted by the superscript 1,T); DVmelt and DVfluid are the volume changes of melt and fluid respectively in Equation (10), which must be considered for high-pressure equilibria. By simplifying the expression for FeO activity in melts, only considering 1-bar conditions, and assuming that the activity of FeS in pyrrhotite is unity, Bockrath et al. (2004) quantified Equation (11) and found: log fS2 = 6.7 −
12800 melt − 2 log X FeO + ∆FMQ T
(12)
where DFMQ refers to the log of the measured fO2 minus the log of the fO2 of the fayalite-quartzmagnetite buffer (FMQ) at the same temperature. The first two terms in Equation (12) are related to the entropic and enthalpic contributions to the Gibbs free energy of Equation (10); the third term expresses the effect of melt composition on the fS2, and the last term demonstrates the influence of fO2. With knowledge of the appropriate variables (T, fO2, and XFeO), this expression from Bockrath et al. (2004) allows computation of the fS2 at sulfide saturation of a silicate melt, but strictly only holds at 1 bar in the temperature range of 1473 to 1673 K. Liu et al. (2007) incorporated a DVreaction term into Equation (12) to extend it to high pressure. Simple expressions
Modeling the Solubility of Sulfur in Magmas
175
like these are extremely useful, preserve a thermodynamic flavor, and formally respect the reaction constraints imposed by stoichiometric coefficients (i.e., the factor of 2 before the log term in Eqn. 12). But, such relationships are most often calibrated for specific compositions. Therefore, they have limited applicability because they disregard the effect of compositional variables in determining activity-composition relationships of oxide and sulfide components. In particular, these approaches cannot apply to both simple and complex melt systems (e.g., a CaOMgO-Al2O3-SiO2 system and a natural melt). In order to accurately describe the partitioning of sulfur between the various phases (melt, fluid and coexisting sulfur-bearing phases) we must change the way we describe the thermodynamic properties of the silicate melt components, and this is the most difficult challenge in the construction of thermodynamic models for the behavior of sulfur in silicate melts. Such modifications will be discussed in one example below, after a review of experimental studies and empirical models of the interactions between sulfurbearing phases and silicate melts.
A BRIEF HIStORY OF EXPERIMEntS AnD EMPIRICAL MODELS FOR SuLFuR SOLuBILItY In SILICAtE MELtS Sulfur behavior in systems with only sulfur-rich gas and silicate melts at 1 atm The earliest investigations on the behavior of sulfur in silicate melts were performed by metallurgists anxious to control the amount of sulfur in slags. These researchers’ experimental studies laid the groundwork for both empirical and thermodynamic models of the interaction of sulfur species with silicate melts. The contribution by Backnaes and Deubener (2011, this volume) provides a summary of many experiments involving sulfur and silicate melts at 1 atm. In this section we briefly review the experiments that were specifically used to create models for the dissolution of sulfur into compositionally complex silicate melts at 1 atm pressure. The classic reference for sulfur’s behavior in silicate melts coexisting with a gas phase at 1 atm is Fincham and Richardson (1954), who reported a compilation of their experimental studies and those of previous researchers on the interactions between H2-CO2-SO2 gas mixtures and simple silicate melts in the systems CaO-SiO2, MgO-SiO2, FeO-SiO2, CaO-Al2O3 and CaO-SiO2-Al2O3. They provided compelling evidence that sulfur dissolves into the melt as sulfide at relatively low fO2 which can be described by the reaction 1 gas 1 S2 + O2 −, silicate melt ↔ O2gas + S2 −, silicate melt 2 2
(13a )
which leads to the following thermodynamic equation 1 1 melt log[S2 − , silicate melt ] = logK (13a) + log fSgas − log fOga2 s + log aOsilicate 2− 2 2 2
(
)
(
)
(
)
(13b)
where [S2−, silicate melt] is the concentration of sulfide dissolved in the silicate melt, K(13a) is the equilibrium constant of Equation (13a), and aO2−silicate melt is the activity of “free oxygen” in the melt, which is typically considered constant for a given melt composition. At higher fO2’s sulfur was found to dissolve as sulfate, whose dissolution reaction can be written as 1 gas 3 gas S2 + O2 + O2 − ,silicate melt ↔ SO24 − ,silicate melt 2 2
(14a )
and 1 3 melt log SO24 −, silicate melt = log K (14a ) + log fSgas + log fOgas + log aOsilicate 2− 2 2 2 2
(
)
(
)
(
)
(14 b)
Baker & Moretti
176
These reactions were based upon the slopes of sulfur concentrations in melts at varying oxygen fugacity, but constant sulfur fugacity. Fincham and Richardson (1954) experimentally determined that the solubility reaction in a CaO-SiO2-Al2O3 melt at 1550 °C changed from sulfide to sulfate species at an fO2 of ~10−5.5 bars, where the sulfur concentration in the melt was at a minimum (Fig. 2). However, the reactions Fincham and Richardson (1954) applied to describe sulfur dissolution in their experiments are different from Equations (13) and (14) because of their use of SO2 gas in experiments (Fig. 2), producing the following equation at low oxygen fugacity 3 SO2gas + O2 − ,silicate melt ↔ O2gas + S2 − ,silicate melt 2
(15)
and at high oxygen fugacity, 1 SO2gas + O2gas + O2 −, silicate melt ↔ SO24 −, silicate melt 2
(16)
whose slopes defined by the dissolved sulfur in the melt and the oxygen fugacity support the presence of S2− species dominating in the melt at lower fO2 and SO42− species at higher fO2. Importantly, the experiments of Fincham and Richardson were only saturated with a sulfurbearing gas; no crystalline or molten sulfide or sulfate was stable, and as will be demonstrated later, the saturation of the system with another sulfur-bearing phase has a measurable effect on the depth of the solubility minimum. Because sulfur contents of slags are a complex function of T, fO2 and fS2, metallurgists first faced the problem of defining a parameter able to embody all the dependencies upon these terms. In order to create a general expression to relate fO2 and fS2 to the sulfur concentration in the melt, Fincham and Richardson (1954) redefined the sulfide capacity of Rosenqvist (1951) as CS = [S]
fO12/ 2 fS12/ 2
(17)
and the sulfate capacity as CSO4 = [S]
1 f f
3 / 2 1/ 2 O2 S2
(18)
where [S] is the concentration of sulfur in the silicate melt, which Fincham and Richardson (1954) measured in weight percent, although other authors define CS with sulfur measured in ppm. Equations (17) and (18) are easily seen to be related to the equilibrium constant expressions for Equations (13) and (14), respectively. With these definitions, the sulfide and sulfate capacities of different melt compositions can be easily compared, and CS became the commonly accepted method of expressing the amount of sulfur in a silicate melt. It must be stressed that CS is a pseudo-equilibrium constant and does not, in any way, determine the sulfur solubility in the silicate melt at saturation with another liquid or solid phase. Its nature is very similar to that of an equilibrium constant, inasmuch as CS is a constant for each composition (at a given T and P) as long as the sulfur concentration in the silicate melt is relatively low (i.e. the activities of oxide components in S-bearing melts are nearly identical to those observed in the sulfur-free melt). Although the thermodynamic analyses of Fincham and Richardson (1954) have been superseded, the canonical equations they wrote for sulfur dissolution into silicate melts and the concept of sulfide and sulfate capacities remain cornerstones for investigations of sulfur behavior in silicate melts. The most important realization of Richardson and coworkers is that the high reactivity of sulfur in gas-silicate melt reactions needs an ionic notation to be described, involving a redox
Modeling the Solubility of Sulfur in Magmas
177
log S (ppm) in silicate melt
5
4
1/2 S 2 + 3/2 O 2 + O
3
1/2 S 2 + O
melt
= 1/2 O 2 + S
melt
melt
= SO4
melt
2
1
o
CaO-Al2O3-SiO2 melt, 1500 C after Fincham and Richardson (1954)
0
-10
-9
-8
-7
log f O
-6
-5
-4
-3
2
Figure 2. Solubility of sulfur in silicate melts at conditions of a fixed mole fraction of SO2 in the input gas phase and varying fO2 measured by Fincham and Richardson (1954) in a CaO-Al2O3-SiO2 melt at 1 bar pressure. Please see text for further discussion of the reactions portrayed in these figures.
exchange coupling of sulfur and oxygen species. Richardson and coworkers’ research in the 1950’s made significant progress not only in the study of sulfur dissolution in melts, but also in the study of silicate melts themselves. Indeed, they advanced our structural understanding of silicate melts by recognizing that the reactivity of sulfur affects the melt structure. They also introduced a notation that defines the three types of oxygens in silicate melts: those bonded to two silicon atoms (-Si-O-Si-, or O0), or bridging oxygen, those bonded to one silicon atom and associated with metal atoms (e.g., -Si-O−·Ca2+·−O-Si-, or simply O−), or non-bridging oxygen, and those associated with metal atoms only (e.g., O2−·Ca2+, or O2−), or free oxygen. They clearly stated that the proportions of these three types of oxygen species will depend upon the proportions of silica and metal oxide composing a binary system. These proportions should be controlled by an equilibrium of the type: O2 − + Si-O-Si ↔ 2Si-O −
(19)
with an equilibrium constant of: K (19) =
2 aSi-O −
aO2− aSi-O-Si
(20)
Equations (19) and (20) can be rewritten in a more general fashion by considering the quasichemical species O0 and O−. In fact, Equation (20) does not imply a unique, T-P dependent, equilibrium constant: depending on the chemistry of the melt system, other network formers can intervene in place of Si (e.g., Al), and many networks modifiers can be present, shifting the melt’s acid-base equilibria (see below), and hence energetic, properties. Equation (20) is the basis for mixing models of silicate melts, as shown by Toop and Samis (1962) and discussed below. The validity of this reaction between oxygen species is independent of the presence of sulfur in silicate melts. However, it was possible to discern this basic reaction in silicate melts only when investigating sulfur because of its high reactivity, which results in its dissolution to a completely dissociated species. Indeed, the measurements allowed Richardson and
Baker & Moretti
178
Fincham (1956) to recognize that sulfur solubility was inversely correlated with the amount of silica, and that oxygens associated with metal atoms are the only oxygen species entering into Equations (13) and (14). These insights into the nature of silicate melts were a major milestone, but raised important questions concerning how to generalize results from binary systems to complex multicomponent (e.g., natural) systems and the applicability of the reaction mechanisms discovered. Calculating the activity of O2− anions in compositionally complex melts is a daunting task, which has led to the development of polymeric models in order to assess the melt reactivity (discussed below). Many other metallurgists studied the sulfide capacity of different slag compositions from the 1950’s, providing a useful database (summarized in: Moretti and Ottonello 2003a,b, 2005; Backnaes and Deubener 2011, this volume). Although these numerous studies of sulfur dissolution into slag melts (e.g., Abraham and Richardson 1960; Young et al. 1992) and studies of other simplified sulfide systems of geological interest, such as Fe-S-O (Naldrett 1969) and FeS-FeO-Fe2O3-SiO2 (MacLean 1969) added to our knowledge, the next great advance in our understanding of sulfur behavior in magmatic systems appeared with the paper of Katsura and Nagashima (1974). Katsura and Nagashima (1974) used similar techniques to those of Fincham and Richardson (1954), but for the first time a series of natural melt compositions (tholeiite, hawaiite, and rhyolite) were exposed at high temperatures to H2-CO2-SO2 gas mixtures at 1 atm. and the sulfur concentrations and speciation in the melts determined. The melts were not saturated with crystalline or molten sulfides or sulfates, and the functional relationships between dissolved sulfur in the natural melts and fO2 were consistent with the Fincham and Richardson (1954) hypothesis of the dominance of sulfide in the melt at lower fO2 and sulfate at higher fO2. Using wet-chemical, analytical techniques that could determine sulfide and sulfate in the quenched glass, and at constant XSO2gas values, Katsura and Nagashima (1974) demonstrated that sulfur dissolved as sulfide at lower fO2 and as sulfate at higher fO2, confirming Fincham and Richardson (1954). The fO2 at the minimum sulfur concentration (< 0.3 wt%) was approximately 10−8 bar. Shima and Naldrett (1975) studied sulfur dissolution in an ultramafic melt; they demonstrated that a plot of sulfur concentration versus log fS2 at constant fO2 displayed a slope of 1/2, consistent with the predictions of Fincham and Richardson (1954). Buchanan and Nolan (1979) and Danckwerth et al. (1979) performed further investigations of the behavior of sulfur in geologically relevant melts in the CaO-MgO-Al2O3-SiO2-FeO system (Buchanan and Nolan 1979; Buchanan et al. 1983) and in high FeO and TiO2 lunar melts (Danckwerth et al. 1979). These data expanded the compositional base for melts saturated with an S-bearing gas, but Buchanan and Nolan (1979) and Buchanan et al. (1983) disagreed with the results of Fincham and Richardson (1954). Buchanan and Nolan’s (1979) and Buchanan et al.’s (1983) findings have been questioned by O’Neill and Mavrogenes (2002) who performed a detailed study of sulfur partitioning between sulfur-bearing gases and geologically relevant melts in the system CaO-MgO-Al2O3-SiO2 ± FeO ± TiO2 and other melts of differing compositions at 1 atm. O’Neill and Mavrogenes (2002) demonstrated that these systems followed the behavior originally seen by Fincham and Richardson (1954) for simple systems and were able to use the data to define an empirical relationship between the CS and melt composition at 1400 °C.
SuLFIDE AnD SuLFAtE SAtuRAtIOn In SILICAtE MELtS Sulfide saturation at 1 bar The earliest studies of sulfide saturation of silicate melts were performed by metallurgists who measured the partitioning of sulfur between melts (or complex slags) and coexisting sulfide liquids and metals. These studies, although not investigating the gas phase, established some basic principles of general validity. More details of some of these studies, as well as other
Modeling the Solubility of Sulfur in Magmas
179
studies on compositionally simple systems are found in the contribution in this volume by Backnaes and Deubener (2011, this volume). A common feature of this earlier époque is that the nature of the equilibrium interaction between phases and their components was not deduced from the behavior observed for exchanged components, e.g., sulfur. Rather, step-wise assumptions were made in order to reproduce the observed behavior. For example, Darken and Larsen (1942) studied the partitioning of sulfur between slags and metals by assuming that all sulfur exists in melts as FeS, CaS and MnS. Therefore, they investigated exchange reactions between slag components given by corresponding couples of metal oxides and sulfides, for example: FeSsulfide + CaO silicate melt ↔ CaSsilicate melt + FeO silicate melt
(21)
FeSsulfide + MnO silicate melt ↔ MnSsilicate melt + FeO silicate melt
(22)
and In their treatment they assumed that the free concentration of a silica molecule is negligible, e.g., that SiO2 is associated with other oxides. Although, Darken and Larsen (1942) recognized that their knowledge of melt constitution was far from complete, they were able to construct a model that successfully described the experimental results available to them at that time. The success of this kind of model was limited to restricted compositional domains: the wider the compositional domain, the larger the number of possible complexes needed to explain the observed partitioning, hence the greater the difficulties in establishing a general model for the behavior of sulfur in silicate melts. Thirty-two years later Haughton et al. (1974) studied sulfur solubility in natural mafic melts by equilibrating them at 1200 °C with CO-CO2-SO2 gas mixtures at low oxygen fugacities where sulfide is the stable sulfur species in the melt. Haughton et al. (1974) used the electron microprobe as their analytical tool; this allowed them, unlike previous studies, to analyze silicate melts that were saturated with a sulfide phase, liquid FeS. Haughton et al. (1974) demonstrated that the sulfur concentration in melts saturated with sulfide, the SCSS, was positively correlated with fO2 at constant fS2 and negatively correlated with fS2 at constant fO2 (Fig. 3). This behavior is due to the presence of sulfide in the experiments of Haughton et al. (1974) which buffers the melt composition. A schematic reaction explaining the behavior observed by Haughton et al. (1974) was given by Liu (2005) who combined Equation (13a) with the reaction 1 1 FeO sulfide melt + S2gas ↔ FeSsulfide melt + O2gas 2 2
(23)
4 FeSsulfide melt + 2O2 −, silicate melt + O2gas ↔ 2S2 −, silicate melt + S2gas + 4 FeO sulfide melt
( 24 )
to create
which replicates the positive effect of fO2 and negative effect of fS2 on the SCSS seen by Haughton et al. (1974). Because Haughton et al. (1974) investigated a variety of mafic melt compositions they were able to create the first model (of which we are aware) to calculate the sulfide capacity for natural, mafic melts of different compositions. Haughton et al. (1974) remark that “the transposition of sulfide capacities derived from binary compositions to multicomponent melts is complex.” In fact, Fincham and Richardson (1954) demonstrated that in ternary and more complicated melts, the activity coefficient of FeO reaches a maximum at intermediate compositions because it is inversely proportional to the solubility of FeO in these melts. Haughton et al. (1974) argued that since the sulfide capacity is so strongly controlled by the amount of FeO in a melt, and because the activity coefficient of FeO reflects its solubility, the sulfide capacity must be a function of the activity coefficient of FeO. Haughton et al. (1974) also pointed out that silica and alumina display negative correlations with the SCSS; CaO and MgO show small positive correlations,
Baker & Moretti
180
Figure 3. Schematic drawing of the effects of fS2 and fO2 on the saturation surfaces for silicate and sulfide melts. The points represent two melts coexisting with each other and connected by the tie-line seen in the figure. Drawn after Haughton et al. (1974).
while Na2O and K2O show small negative correlations (but too small to be definitive). TiO2, however, exhibits a strong positive correlation, almost matching FeO in its influence. Haughton et al. (1974) rearranged Equation (13b) to calculate the concentration of dissolved sulfur in the melt, [S2−, silicate melt]: 1 1 melt log S2 −, silicate melt − log fSgas + log fOgas = log K((13a ) + log aOsilicate 2− 2 2 2 2
(
)
(
)
(
)
(25)
The left-hand side of Equation (25) is the log (CS) (see Eqn. 17) and the right-hand side contains the equilibrium constant, which is fixed at a given temperature and pressure, plus the term for the activity of O2−, silicate melt, which is a function of melt composition. By fitting all of their data Haughton et al. (1974) were able to predict the log(CS) for basaltic melt compositions at 1200 °C and one atmosphere with an empirical equation: log(CS ) = −5.704 + 3.15 X FeO + 2.65 X CaO + 0.12 X MgO + 0.77 X TiO2 + 0.75 ( X Na2 O + X K 2 O ) (26)
where the concentration of sulfur in the CS term is expressed in weight percent, −5.704 can be associated with the logK(13a) term and the X’s are the mole fractions of the various oxides in the melt whose weighted sum reflects the log(aO2−silicate melt) term. Haughton et al. (1974) made no claims of thermodynamic rigor, but promoted Equation (26) as a method to calculate the CS. As already demonstrated by Richardson and Fincham (1956), log(CS) is linearly related to the reciprocal of temperature, with constant slope but intercept values depending in a complex fashion on melt chemistry. According to Haughton et al. (1974), a slope of −0.819×104/T is a reasonable approximation for most mafic magmas. Based on this assumption, if the ratio of fO2 to fS2 and all other variables are kept constant upon cooling, the solubility of sulfur would decrease by a factor of ten from 1200 °C to 1040 °C. As discussed by Haughton et al. (1974), and as remains true today for more recent models, the challenge in applying Equations such as (25) and (26) to calculate the SCSS requires knowledge of both fO2 and fS2 in natural systems. Nevertheless, Haughton et al. (1974) demonstrated that the concentrations of sulfur calculated by their model were consistent with measurements of natural mafic rocks available at that time. Shima and Naldrett (1974) performed 1-atm experiments at 1450 °C and fO2’s of 10−10.4 and 10−9.2 bars on an ultramafic melt and in some experiments saturated it with molten FeS. They found that at fS2’s between 10−2.4 and 10−2.0 bars the silicate melt became saturated with FeS and contained between 1600 and 2700 ppm sulfur at the SCSS. Incidentally, it was Shima and Naldrett (1975) who introduced the definition of the SCSS.
Modeling the Solubility of Sulfur in Magmas
181
Although Wallace and Carmichael (1992) performed no experiments, they compiled data from previous studies and created a model for the prediction of the mole fraction of sulfur, XSsilicate melt in basaltic melts at 1 atm pressure: −1.75 × 10 4 + 0.744 T + 8.07 X CaO + 6.37 ( X Na2 O + X K 2 O )
ln XSsilicate melt = 0.300 ln fS2 − 0.179 ln fO2 + 0.388 ln X FeO + +1.56 XSiO2 + 8.69 X FeOtotal
(27)
where the X’s are the mole fractions of oxides in the melt and the coefficients were calibrated from experimental data in the literature. This equation can be combined with the thermodynamic values for the Equation (23) to calculate the SCSS. Their model works well for basalts and similar compositions, for which the model was designed, but the predictions become less accurate for more silicic melts where the SCSS is at much lower sulfur concentrations. O’Neill and Mavrogenes (2002) combined their experimental results (discussed above) with thermodynamic data and constructed a model to predict the SCSS at 1 bar. Following a methodology similar to Haughton et al. (1974), their predicted SCSS values were calculated through the sulfide capacity and the Gibbs free energy change of the iron, sulfur, and oxygen exchange between silicate and sulfide melts at 1 atm: 1 1 FeO silicate melt + S2silicate melt ↔ FeSsulfide melt + O2sulfide melt 2 2
(28a )
and an expression for the Gibbs free energy change of Equation (13a) ∆G o (13a ) = 122175 − 80.280T + 8.474T ln T
(in J mol −1 )
(28b)
The lnCS was calculated by fitting their experiments in a manner similar to Haughton et al. (1974) ln(CS ) = −5.018 + 7.56 X Ca + 4.48 X Mg + 4.24( X Na + X K )
(29)
+5.20 X Ti + 26.31X Fe + 1.06 X Al + 48.48 X Fe X Ti
where the CS term in this case is calculated using ppm as the unit of sulfur concentration in the silicate melt, and the mole fractions of each cation are calculated by dividing the concentration of a cation by the sum of the concentrations all cations dissolved in the silicate melt e.g.,: XSi =
moles Si moles Si + moles Ti + moles Al + moles Fe + ...
(30)
The concentration of sulfur (in ppm) at the SCSS, [S]SCSS, can be calculated from: ln [S]SCSS =
∆G o (13a ) sulfide silicate melt + ln CS + ln aFeS − ln aFeO RT
(31)
where O’Neill and Mavrogenes (2002) considered that aFeSsulfide is typically close to 1 and can therefore be approximated as 1, and that aFeOsilicate melt can be considered equal to XFeOsilicate melt. O’Neill and Mavrogenes (2002) demonstrated the success of their model for many mafic melt compositions, except a few high-FeO, high-TiO2 melts of lunar composition studied by Danckwerth et al. (1979). As discussed previously, O’Neill and Mavrogenes (2002) cast doubt on the results of Buchanan and Nolan (1979) and Buchanan et al. (1983) because some of their experimental results do not display the Fincham and Richardson (1954) theoretical slope of 1/2 when sulfur concentration is plotted as a function of fS2 at constant fO2.
Baker & Moretti
182
Sulfide and sulfate saturation at high-pressure with or without the presence of a hydrous fluid High-pressure studies on the behavior of sulfur in magmatic melts began in the late 1970’s with studies by Helz on the sulfide melt saturation of basaltic melts (Helz 1977) and Helz and Wyllie’s (1979) study of the system CaCO3-Ca(OH)2-CaS. Mysen and Popp (1980) studied sulfur solubility in albite and diopside melts at high pressure (discussed below), and Huang and Williams (1980) studied the system Fe-S-Si-O to 3.2 GPa. The first comprehensive, high-pressure study of sulfur behavior in magmatic melts of different compositions was Wendlandt (1982), who studied sulfide saturation of two basalts and an andesite at pressures from 1.25 to 3.0 GPa and 1300 to 1460 °C. Wendlandt (1982) placed mixtures of rock + sulfide into graphite capsules and then inserted this capsule into a platinum capsule that was welded closed; the oxygen fugacity in this type of assembly is near the carboncarbon monoxide-carbon dioxide-oxygen buffer, below the FMQ buffer (Jacobsson and Oskarsson 1991). The fS2 in such a capsule cannot be explicitly controlled and the fugacities of volatiles change with temperature and pressure. Wendlandt (1982) convincingly demonstrated that the SCSS decreased with increasing pressure at constant temperature and increased with increasing temperature at constant pressure (Fig. 4). These results on anhydrous melts have been confirmed by subsequent studies of the SCSS as a function of temperature and pressure. The observed influences of pressure and temperature were approximately linear, indicating that even though fugacities were not precisely controlled, they did not change significantly in the experiments. Wendlandt (1982) confirmed the effect of FeO on the SCSS seen by Haughton et al. (1974) and presented evidence for the dissolution of FeO into the sulfide melt (cf. Eqn. 24). The negative pressure effect found by Wendlandt (1982) was in direct contrast to the positive effect of pressure found by Mysen and Popp (1980) in their study of sulfide saturation in diopside and albite melts. Wendlandt (1982) argued that the differences were due to different capsule design; Mysen and Popp (1980) used boron nitride capsules which impose an extremely low fO2 on the samples, and used sulfide buffers that imposed high fS2’s. An additional factor that was recognized more recently is that the behavior of sulfur in very low-FeO* (total iron
2000
S in silicate melt (ppm)
1800
Grande Ronde basalt Mt. Hood andesite
1600 1400 1200 1000
2000 1800 1600 1400 1200 1000 800 600
800 600 400 200 0
2.0 GPa
o
1420 C
1300 1350 1400 1450 o
after Wendlandt (1982)
Temp.( C) 1
1.5
2
Pressure (GPa)
2.5
3
Figure 4. Wendlandt’s (1982) measurements of the effects of pressure and temperature (see inset) on the sulfur concentration at sulfide saturation, the SCSS, for anhydrous basaltic and andesitic melts at oxygen fugacities controlled by graphite in platinum capsules.
Modeling the Solubility of Sulfur in Magmas
183
as FeO) melts is quite different than in higher FeO* melts (e.g., Bradbury 1983; Poulson and Ohmoto 1990). The results of Mysen and Popp (1980) have not been investigated in more recent studies because of their apparent inapplicability to sulfide saturation of natural magmas on terrestrial planets. Bradbury (1983) performed a detailed study of pyrrhotite solubility in water-saturated albitic melts. By using different compositions of pyrrhotite he was able to vary the FeO* concentrations in the melts between approximately 0.1 and 1.95 wt% FeO*. He performed experiments from 900 to 1000 °C and 100 to 600 MPa at the FMQ buffer. Using the iron concentrations of the pyrrhotites in the experiments, Bradbury was able to calculate the fS2 of his experiments from the calibration of Froese and Gunter (1976). Bradbury demonstrated that the SCSS was highest in low iron melts, with less than 0.2 wt% FeO*, and decreased with increasing iron in the melt (Fig. 5). Based upon his experimental results he proposed that pyrrhotite dissolves by producing a SH− species in the silicate melt FeSsulfide + O2 −, silicate melt + H 2 O gas ↔ FeO silicate melt + SH −, silicate melt + OH −, silicate melt
(32)
Bradbury’s study is also unique because he found no effect of temperature, unlike other highpressure, but anhydrous, experiments performed over a similar temperature range (cf., Fig. 4 inset). Carroll and Rutherford (1985, 1987, 1988) investigated the solubility of both sulfide and sulfate phases in hydrous andesitic and dacitic melts. Carroll and Rutherford (1985) demonstrated that at fluid-saturated conditions of 1025 °C, pressures from 100 to 225 MPa, and the graphite-methane and FMQ oxygen fugacity buffers, the SCSS increased with increasing pressure and with increasing FeO* (up to 29.6 wt% FeO*) in the melts. The effect of pressure on the SCSS in these fluid-saturated melts is opposite to the effect observed by Wendlandt (1982) for anhydrous melts. At the same temperatures and pressures, but at higher oxygen fugacities, defined by the MnO-Mn3O4 and Fe3O4-Fe2O3 buffers, the stable sulfur-bearing mineral was anhydrite and the sulfur concentration in the melt at saturation, the SCAS, was much higher and increased with pressure. Carroll and Rutherford (1987) investigated the conditions at which an andesitic magma would saturate with sulfide or anhydrite; they demonstrated that at the nickel-
S in silicate melt (ppm)
4000 3500
SCSS of an albitic melt
3000
400 MPa, 950 C (Bradbury, 1983)
o
2500 2000 1500 1000 500 0 0
0.25
0.5
0.75
FeO* (wt%)
1
1.25
1.5
Figure 5. The SCSS for a water-saturated albitic melt as a function of FeO* (total iron as FeO) concentration at 950 °C and 400 MPa from Bradbury (1983). Note the dramatic increase in the SCSS at low FeO*. For further details please see the text.
184
Baker & Moretti
nickel oxide buffer, pyrrhotite is expected to crystallize whereas at the MnO-Mn3O4 buffer and higher oxygen fugacities anhydrite is expected. They provided further evidence for the positive relationship between the FeO* concentration in the melt and the SCSS and demonstrated that anhydrite saturation was more sensitive to pressure and temperature (both positively correlated) than pyrrhotite saturation. Because of the limited compositional range investigated, Carroll and Rutherford (1985, 1987) did not create a general model to calculate the CS, the SCSS or the SCAS. Carroll and Rutherford (1988) found that near the FMQ buffer the amount of sulfate was near zero, but at 2 to 3 log units of fO2 higher only sulfate was stable in the melt. Because the sulfide to sulfate ratio is a function of the oxygen fugacity, as shown in the equations of Fincham and Richardson (1954), the position of the S-Ka X-ray emission peak provides a technique to estimate the oxygen fugacity of experimental and natural glasses, discussed below and elsewhere in this volume (Wilke et al. 2011, this volume). Markus and Baker (1989) published an abstract on their study of the SCSS in an andesitic melt at 1200 to 1400 °C and at fO2’s defined by their capsules of graphite enclosed in platinum, i.e., below the FMQ buffer where all sulfur is dissolved in the melt as sulfide. They demonstrated that the SCSS in andesitic melts was less than in basaltic melts and increased with temperature. Curiously, they found an apparent increase in the SCSS with increasing sulfur in the bulk composition studied, an observation inconsistent with the thermodynamically expected behavior. This effect was particularly noticeable at total sulfur concentrations above 1 wt% in the system. Although not reported in the abstract, this increase was attributed to the presence of small, less than 1-mm in size, sulfide melts present in the quenched glass at spatial scales below the resolution of the backscattered imaging used locate areas for analysis. Such quenched sulfide melts became more abundant with increasing total sulfur in the system and created the apparent increase in the SCSS with increasing sulfur. Such experimental artifacts must be carefully avoided for accurate measurements of the true SCSS. Luhr (1990) studied sulfide and anhydrite solubilities in basaltic and trachyandesitic bulk compositions in fluid-saturated experiments at pressures from 100 to 400 MPa, temperatures from 800 to 1000 °C, and oxygen fugacities defined by the FMQ, NNO, MnO-Mn3O4 and Fe3O4Fe2O3 buffers. The results of Luhr (1990) are in accord with those of Carroll and Rutherford (1987), partially because they studied similar rock compositions from the 1982 eruption of El Chichon volcano, Mexico. Luhr (1990) found that pyrrhotite saturated the silicate melts at the FMQ and NNO buffers, but, as seen by Carroll and Rutherford (1987), anhydrite was stable at higher fO2’s. Luhr also found a positive dependence of the sulfur concentrations at sulfide saturation with pressure, in contrast to Wendlandt’s (1982) findings in anhydrous systems. Poulson and Ohmoto (1990) compiled all of the experimental data available at that time and proposed the existence of three different equilibrium reactions between sulfide and silicate melts. For silicate melts with FeO* concentrations less than 1 wt%: FeSsulfide melt + O2 −, silicate melt ↔ FeO silicate melt + S2 −, silicate melt
(33a )
for melts with FeO* concentrations between 1 and 10 wt%: FeSsulfide melt ↔ FeSsilicate melt
(33b)
and for melts containing greater than 10 wt% FeO*: FeSsulfide melt + 2 FeO silicate melt ↔ Fe3SO2silicate melt
(33c)
They based their reactions upon the slopes of data in plots of the mole fraction of sulfur versus the mole fraction of iron at conditions where S2− is the sulfur species in the melt. Although their proposal of differing dissolution mechanisms at sulfide saturation of silicate melts has largely been discarded, they were instrumental in highlighting the strange behavior of the SCSS at low
Modeling the Solubility of Sulfur in Magmas
185
FeO concentrations as seen by Bradbury (1983). This behavior, however, was later explained by the model of O’Neill and Mavrogenes (2002) discussed above. Mavrogenes and O’Neill (1999) performed sulfide melt saturation experiments on anhydrous basaltic and picritic melts at pressures from 0.5 to 9.0 GPa and temperatures from 1400 to 1800 °C in capsules of either iron or iron-iridium alloys. Mavrogenes and O’Neill (1999) combined Equation (13a), the definition of Cs, and the reaction describing the distribution of Fe between silicate and sulfide melts, Equation (23), to produce: ln[Ssilicate melt ] =
∆G o (23) sulfide silicate melt + ln CS − ln aFeS + ln aFeO RT
(34)
where ln[Ssilicate melt] is the natural logarithm of the concentration of sulfur in the silicate melt saturated with sulfide and expressed in ppm, ∆G°(23) is the standard state Gibbs free energy of Equation (23), aFeSsulfide is the activity of FeS in the sulfide and aFeOsilicate melt is the activity of FeO in the silicate melt. Because ln CS is fundamentally an equilibrium constant, Mavrogenes and O’Neill (1999) realized that both CS and ∆G° could be expressed by the sum of entropic, enthalpic and pressure terms; on this basis they proposed that experiments could be used to calibrate an equation that expresses the SCSS for their basaltic melt compositions, but yet does not explicitly rely upon knowledge of fO2 and fS2, which are often poorly constrained, especially for natural samples: ln[Ssilicate melt ] =
−6684 P sulfide + 11.52 − 0.047 + ln aFeS T T
(35)
where pressure is measured in bars and aFeSsufide is the activity of FeS in the sulfide phase; its value is a reflection of the fO2 and fS2. However, because aFeSsulfide is often close to 1, as discussed by O’Neill and Mavrogenes (2002), this value can be adopted for many natural rocks and the last term disappears from Equation (35). The values for the constants in Equation (35) were extracted from their experiments on basaltic melts; they also produced similar equations to describe their results on picritic melts and the results of Wendlandt (1982). The model of Mavrogenes and O’Neill (1999) works well for the prediction of the SCSS in basaltic melts, however this model does not allow the calculation of the effect of melt composition on sulfur solubility and inaccurately predicts the SCSS in more silicic melts (see Liu et al. 2007). Holzheid and Grove (2002) extended the model of Mavrogenes and O’Neill (1999) by adding the melt compositional dependence of the SCSS. To express the effect of composition they used the non-bridging oxygen-to-tetrahedral cation ratio, NBO/T (written as nbo/t in the equation below to avoid confusion between tetrahedral cations, t, and temperature, T), of the melt (Mysen and Richet 2005). Using their own experiments at high pressure on melt compositions from komatiitic to andesitic and data from Mavrogenes and O’Neill (1999), they created a new model for the SCSS that also did not depend upon fO2 and fS2: ln[Ssilicate melt ] =
−10129 nbo P sulfide + 12.84 − 0.060 + 0.793 + ln aFeS T t T
(36)
Scaillet and Pichavant (2003) provided a model for calculating the sulfur solubility in basaltic melts by extending the model of Wallace and Carmichael (1992) to more oxidizing conditions, in which fS2 is the only variable in the model: log[Ssilicate melt ] = 3.2211 + 2.0928(log fS2 ) + 9.5397 × 10 −2 (log fS2 )2
(37)
+3.5864 × 10 −2 (log fS2 )3
where log [Ssilicate melt] is the base 10 logarithm of the sulfur concentration in the silicate melt measured in ppm.
186
Baker & Moretti
Clemente et al. (2004) investigated the solubility of sulfide and sulfate in hydrous rhyolitic melts at 200 MPa, temperatures from 800 to 1000 °C, and fO2’s from 2.3 log units below to 2.9 log units above the NNO buffer. In some cases Clemente et al. (2004) stabilized both sulfate and sulfide in their experiments, thus providing reliable constraints on the fS2. Oxygen fugacities were calculated from hydrogen fugacities in the experiments and the Burnham (1979) model was applied to determine water solubility in the silicate melts. Sulfur fugacites were calculated from the compositions of the pyrrhotite in the run products or by the modified Redlich-Kwong equation of state for experiments where a fluid phase exsolved at high fS2. At low fO2, below NNO+1 log unit, Clemente et al. (2004) considered a reaction proposed by Burnham (1979) that was very similar to that provided by Bradbury (1983), but without the involvement of iron, because they found, as did Bradbury (1983), that melts with the lowest FeO had the highest SCSS: H 2O gas + SH −, silicate melt ↔ H 2Sgas + OH −, silicate melt
(38)
and they proposed that at higher oxygen fugacities the reaction might be SO2gas + 2OH −, silicate melt ↔ SO24 −, silicate melt + H 2gas
(39)
From their experiments Clemente et al. (2004) created a model for rhyolitic melts at 200 MPa that relates the sulfur concentration (in ppm) of silicate melts saturated with sulfide liquid, pyrrhotite, or anhydrite to the oxygen and sulfur fugacities: log[Ssilicate melt ] = 0.001T − 0.2567∆NNO + 0.1713∆FFS + 0.0034 ∆NNO × ∆FFS
( 40)
where T is in °C, the fO2 is referenced to the difference between the log of the fO2 of the sample and the log of the fO2 at the NNO buffer, DNNO, and fS2 is referenced to the Fe-FeS buffer (FFS) in a similar manner, DFFS. This model is applicable to fO2’s between DNNO = 1.5 and −2. Although this model reproduces most of their data, extension of their model to compositions other than rhyolitic is impossible. Botcharnikov et al. (2004) studied the combined solubilities of sulfur and chlorine in a volatile-saturated melt of the Mt. Unzen (Japan) rhyodacite composition for which almost all experiments were saturated with pyrrhotite. They found that chlorine could increase the solubility of sulfur in the melts up to approximately 200 ppm, or twice the value measured in the chlorine-free system. Although no solubility models were presented in this study, the authors strongly argue that sulfur solubilities in melts are controlled by the fluid phase and the concentration of FeO in the melt. Costa et al. (2004) investigated sulfide and sulfate saturation of a dacitic melt and constrained the fO2 at which the sulfur-bearing phase changed from sulfide to sulfate between 1.4 and 2.5 log units above the NNO buffer. Li and Ripley (2005) compiled all of the experimental data available at that time to create a general model for the SCSS expressed as the mole fraction of sulfur in the melt as a function of T, P and melt composition. They updated their model with a new equation a few years later (Li and Ripley 2009) that calculates the mole fraction of sulfur in the melt at the SCSS, XSSCSS: 10 4 ln XSSCSS = −1.76 − 0.474 − 0.021P + 5.559 X FeO + 2.565 X TiO2 T +2.709 X CaO − 3.192 XSiO2 −3.049 X H2 O
(41a )
where P is in kbar and T in Kelvins in Equations (41a) and (41b) (below). Li and Ripley (2009) also compiled published measurements on the saturation of melts with sulfate and fit the available data to yield the equation for the mole fraction of sulfur in a melt saturated with anhydrite, XSSCAS:
Modeling the Solubility of Sulfur in Magmas lnX
SCAS S
10 4 = 10.07 − 1.151 + 0.104P − 7.1XSiO2 − 14.02X MgO − 14.164X Al2 O3 T
187 ( 41b)
Li and Ripley (2005, 2009) used these equations to model the general formation of igneous ore deposits by sulfide saturation and separation. Jugo et al. (2005a,b) investigated sulfide and sulfate saturation at 1.0 GPa in an anhydrous basaltic melt and demonstrated that melts became saturated with sulfate at fO2’s 2 log units above the FMQ buffer. These studies demonstrated that at approximately one log unit below the same buffer, the saturating sulfur-bearing phase was sulfide and the SCSS was on the order of 1000 ppm rather than ~ 1 wt% at the higher oxygen fugacities investigated. Jugo (2009) made a comparison of his results on basaltic melts with previous results on more felsic melts and demonstrated the similarity of sulfur’s behavior for both melt compositions and the enhanced sulfur concentrations in silicate melts saturated with sulfate. Scaillet and Pichavant (2005) updated their model from 2003 based upon the results of Luhr (1990), O’Neill and Mavrogenes (2002) and Clemente et al. (2004) and created an empirical formula to calculate the amount of sulfur dissolved in a melt given the melt composition, fO2 and fS2: log[Ssilicate melt ] = 7.28 × 10 −6 P + 0.00084107T + 0.00488128∆NNO3 + 0.0818873∆NNO2 −0.0224068∆NNO × ∆FFS + 0.22801636 ∆FFS −0.012467 wt% SiO2 − 0.0015766 wt%Al 2O3 +0.37362348 wt% Fe 2 O3 + 0.0674383 wt% FeO
(42)
+0.01121929 wt% MgO + 0.02000831 wt% CaO +0.05644745 wt%Na 2 O − 0.0248037 wt% K 2 O +0.00672403 wt%TiO2 + 0.06868295 wt% H 2 O + 0.05778453 wt% OH
where P is in bar and T is in °C. They then used the same techniques and thermodynamic data as O’Neill and Mavrogenes (2002) to calculate the value of the SCSS. Scaillet and Pichavant (2005) demonstrated that their model successfully reproduced the calibrating experiments and applied the results to the investigation of fluid compositions at different Italian volcanoes. Scaillet and Macdonald (2006) measured the sulfide and sulfate saturation in peralkaline rhyolitic melts and the partitioning of sulfur between the hydrous fluid phase and the melt. Although they did not provide a model for sulfide or sulfate saturation, their study is notable as one of the few that produced melts that were simultaneously saturated with a hydrous fluid, sulfide and sulfate phases. Scaillet and Macdonald (2006) demonstrated that peralkaline rhyolitic melts displayed measurably higher values of the SCSS and SCAS than peraluminous melts. Liu et al. (2007) performed sulfide saturation experiments on melt compositions varying from basaltic to rhyolitic and temperatures from 1050 to 1450 °C at 500 MPa and 1 GPa. Consistent with previous studies, the SCSS increased with temperature and decreased with either pressure or increasing silica concentration in the melt. Intriguingly, Liu et al. (2007) found that the addition of water, but not at quantities necessary to saturate the melt with a volatile phase, increased the SCSS for silicic melts, but decreased it for mafic ones. Based upon their observations and those of Wendlandt (1982), they modeled the sulfide dissolution reaction as: FeSsulfide + FeO silicate melt + O2 −, silicate melt ↔ S2 −, silicate melt + 2 FeO sulfide
(43)
Because the amount of FeOsulfide was small, they set its activity to a constant and aFeSsulfide to 1, as had Mavrogenes and O’Neill (1999) and Holzheid and Grove (2002). They used the concentrations of iron and sulfur in the silicate melt as proxies for their activities. As a proxy
Baker & Moretti
188
for the activity of O2−, silicate melt they defined a parameter that accounted for the effects of melt composition on the SCSS: the MFM parameter (modified from the FM parameter of Ryerson and Watson 1987) that describes the melt composition using cation mole fractions: MFM =
Na + K + 2(Ca + Mg + Fe 2 + ) Si × ( Al + Fe3 + )
(44)
where the mole fractions of Fe2+ and Fe3+ can be obtained either from the analysis of the rock itself of by application of the equation of Kress and Carmichael (1991). Liu et al. (2007) combined their experimental results with data from many previous experimental studies on natural compositions to create a model for the SCSS that included additional terms for the concentration of water in the melt and interactions between water and the anhydrous melt (Fig. 6): 4454.6 P − 0.03190 + 0.71006 ln(MFM) T T (45) −1.98063 (MFM) X H2 Omelt + 0.21867 ln X H2 Omelt + 0.36192 ln X FeOmelt
ln[Ssilicate melt ]SCSS = 11.35251 −
(
)
(
)
Where ln[Ssilicate melt]SCSS is the natural logarithm of the concentration of sulfur (in ppm) at the SCSS, P is in bar, and XH2Omelt is the total mole fraction of water in the melt. Liu et al. performed an independent test of this model by comparison of the predicted SCSS to the SCSS measured in experiments that were not used to construct the model and demonstrated the improvement provided by this model compared to previous ones. With this model the SCSS can be calculated for common igneous melt compositions to a high degree of precision (~10%) at oxygen fugacities between approximately 2 log units above and below the FMQ buffer (based upon the Jugo et al. 2005b, calibration of sulfur species versus fO2) at pressures from 1 atm to 9.0 GPa
Modeled ln SCSS (ppm)
9
Liu et al. (2007) model for the SCSS in komatiitic to rhyolitic melts 5% 10%
8 7
20%
6
Liu et al. (2007) Haughton et. al. (1974) Other anhydrous high pressur e
5
Luhr (1990) Clemente et al. (2004) Holzheid & Grove (2002)
4
Mavrogenes & O’Neill (1999) High f O (Liu et al. 2007) 2
Clemente et al. (2004) (Fe/S)mol< 2
3 2
} measurement uncertainties 4
5
6
7
Measured ln SCSS (ppm)
8
9
Figure 6. Modeled versus measured values of the SCSS using the model of Liu et al. (2007) for melt compositions varying from komatiitic to rhyolitic and pressures from 1 bar to 9.0 GPa, 800 to 1800 °C. The low FeO* melts, [Fe/S]mol <2, of Clemente et al. (2004) display dramatically higher values of the SCSS than melts with higher FeO* concentrations, similar to what is seen in Bradbury’s results displayed in Figure 5. Typical 1-sigma uncertainties in the measured concentrations of sulfur in silicate glasses (quenched melts) are shown at the bottom of the figure. Relative percentage differences between the measured and modeled SCSS are given by the dashed, dotted, and dash-dot lines. (Figure modified after Liu et al. 2007.)
Modeling the Solubility of Sulfur in Magmas
189
and temperatures from 800 to 1800 °C. However, Liu et al. (2007) demonstrated that the SCSS of the low FeO* melts (< ~ 0.2 wt% FeO*, or [Fe/S]mol < 2) of Clemente et al. (2004) could not be reproduced by their model (Fig. 6). A program running in Scilab (http://www.scilab.org) for calculating the SCSS with the model of Liu et al. (2007) is included as an electronic appendix to this contribution; this program also calculates the SCAS (discussed below). Ariskin et al. (2008) published an abstract with a new model for the calculation of the SCSS. Few details of the model are published in the abstract, but the model is based upon the hypothesized presence of multiple iron sulfide complexes in the melt, FenS2(n−1), where n is ≥ 2. The model contains 14 fitted parameters, reproduces the calibration data base to within 10% and is applicable over the temperature range of 1100 to 1400 °C. More details of this model must await its presentation in a lengthier publication. Moune et al. (2009) studied the saturation of hydrous basaltic and basaltic-andesitic melts with pyrrhotite at 1050 °C, pressures of 200 and 300 MPa, and oxygen fugacities varying from one log unit below the FMQ to one log unit above the FMQ buffer. At water-undersaturated conditions Moune et al.’s measured values for the SCSS were broadly consistent with those predicted by the Liu et al. (2007) model, but at water-saturated conditions the measured values for the SCSS were up to 10× greater than predicted by the Liu et al. (2007) model. Moune et al. (2009) argued that the higher SCSS values in the water-saturated experiments may have been due to their higher fO2’s, compared to the water-undersaturated experiments. Webster et al. (2009) investigated the partitioning of sulfur between anhydrite-saturated phonolitic melts and a Cl-bearing, hydrous fluid at 200 MPa and ~ 900 to 1000 °C. The concentration of sulfur in these melts varied from 200 to 1100 ppm and fluid/melt partition coefficients for most of the experiments were between 50 and 300, although one was 2 and another exceeded 1000. These authors found no effect of chlorine on the sulfur partition coefficient. More information on the partitioning of sulfur between silicate melts and fluids is found elsewhere in this volume (Webster and Botcharnikov 2011, this volume). The success of the Liu et al. (2007) empirical model for the SCSS encouraged the authors of this contribution to apply the same melt compositional parameter, the MFM value, to model the sulfur concentration at anhydrite saturation, the SCAS, based upon the reaction: S6 + , melt + 4O2 −, melt + Ca 2 + , melt ↔ CaSO4anhydrite
( 46 )
Following the same logic as Liu et al. (2007) and using the experimental results of Carroll and Rutherford (1987), Luhr (1990), Costa et al. (2004), Jugo et al. (2005a) and Scaillet and MacDonald (2006) to calibrate the model, we created an expression for the SCAS, which reproduces most of the data to within 5% (Fig. 7): ln[Ssilicate melt ]SCAS = 23.53502 −
19073.8 P + 4055.8 + 0.82637 ln (MFM) T T
(
)
(
)
−0.79932(MFM) X H2 Omelt + 0.81241 ln X H2 Omelt − 0.21087 ln ( X CaOmelt )
( 47)
where ln[Ssilicate melt]SCAS is the natural logarithm of the concentration of sulfur (in ppm) at the SCAS, P is in GPa, and XH2Omelt is the total mole fraction of water in the melt. The digits presented in Equation (47) exceed those that are significant, but are kept to avoid round-off errors in the calculations. When the calculated and measured values of the SCAS are compared, the χ2 value (see definition in Liu et al. 2007) is 1.17 and the average squared deviation between the measured and modeled values of the SCAS is 0.09 natural log units (in ppm). Of the 89 data points used for the calibration, only two of them display modeled values of the SCAS that deviate more than 10 relative percent from the measured value (Fig. 7).
Baker & Moretti
190
Modeled ln SCAS (ppm)
10 9 5%
%
10
8 7 6 5 4 4
5
6
7
8
Measured ln SCAS (ppm)
9
10
Figure 7. Modeled versus measured values of the sulfur concentration in silicate melts at anhydrite saturation, SCAS, using the model presented in this chapter. Modeled and measured values typically agree to within 5% relative. For uncertainties in the measured values see Figure 6. Relative percentage differences between the measured and modeled SCAS are given by the dashed and dotted lines. Please see the text for further details.
This model for the SCAS is based primarily upon a small number of experiments using water-rich, dacitic to rhyolitic melts, which is where magmatic sulfates have been found in nature. Some of the papers providing the calibrating experiments only publish a range of oxygen fugacities (used to calculate the MFM value) or water concentrations; in these cases the average of the values provided in the papers were used to calibrate the model. Therefore application of this model, particularly to more basic and to water-poor melts, should be done cautiously. Furthermore, this equation is only applicable for oxygen fugacities higher than approximately 1.5 log units above the NNO buffer, because experimental results demonstrate that sulfide is the sulfur-bearing phase that saturates silicate melts at lower oxygen fugacities. The Scilab program included as an electronic appendix to this contribution computes the SCAS using Equation (47) at oxygen fugacites of NNO+1.5 and above and the SCSS using Equation (45) at oxygen fugacities of NNO+1.5 and lower. As discussed below, this fO2 boundary is an approximation and the exact value depends upon pressure, temperature, and composition, including the concentration of water in the melt. This program does not accurately predict the sulfur concentrations at either sulfide or sulfate saturation in the region where the ratio of dissolved sulfide to sulfate changes rapidly (Fig. 8), but it can be used to constrain the possible concentrations of sulfur in a melt saturated with either sulfide or sulfate at these conditions. Both the SCSS model of Liu et al. (2007) and the SCAS model proposed in this contribution require some water in the analysis. If a totally anhydrous melt is used as input data, the Scilab program adds 0.001 moles, (~200 ppm) of H2O to the analysis. The accuracy of the model for the SCSS is tempered by the results of Moune et al. (2009), which imply that the SCSS predicted by the Liu et al. (2007) model for water-saturated basaltic melts may be considered minima, and that the true SCSS can be higher. This short history of experiments and the development of empirical models to describe the behavior of sulfur in silicate melts during the past 50 years demonstrate that we have made substantial progress. With each study a new set of results became available to expand the data base that could then be fit with some combination of compositional variables, temperature and
Modeling the Solubility of Sulfur in Magmas
191
1.0 0.9 0.8
Nilsson and Peach (1993) Wallace and Carmichael (1994) Metrich and Clocchiatti (1996)
/Total S
0.7 0.6 Moretti and Ottonello (2003)
0.4
X-S
6+
0.5
Wallace and Carmichael (1994)
0.3
0.2 0.1 0.0 -4
-3
-2
-1
0
∆ NNO
1
2
3
4
Figure 8. The mole fraction of total sulfur as S6+ in melts (X-S6+/Total S) as a function of the difference between the log of the fO2 and the fO2 at the nickel-nickel oxide buffer, DNNO, measured by Nilsson and Peach (1993), Wallace and Carmichael (1994) and Métrich and Clocchatti (1994) using the energy shift of the sulfur Ka emission line. Also shown are the expected variations in X-S6+/Total S with DNNO calculated by Wallace and Carmichael (1994), the solid line, and by Moretti and Ottonello (2003a), the dashed line. Only data for geological systems are reported. See text for details.
pressure. The models evolved from being limited to a restricted set of compositions and temperatures to a broad range that encompasses most common igneous melts. We now have the ability to predict the SCSS and the SCAS for many different natural melt compositions, as well as different fO2’s, water concentrations and fS2’s in the system. However much we may extol these models, each of them has its own limitations, the most significant one being the potential problems that may arise when the model is extrapolated beyond its compositional base and range of fO2’s over which it was calibrated (which for studies of natural compositions are often limited). Notable questions still remain, however. One of the most intriguing is the clear negative effect of pressure on the SCSS for fluid-undersaturated melts and the positive effect for fluidsaturated melts. Another question is how the composition of the fluid affects the sulfur solubility in the melt. The high SCSS values at low FeO concentrations needs further investigation, although the compositional range where this behavior occurs is not relevant to most geological processes. Future experimental studies investigating the SCSS and SCAS in hydrous basic to silicic melts will provide some answers to these questions, but to truly understand sulfur in magmatic processes we must construct a rigorous thermodynamic model of sulfur’s behavior in silicate melts.
tHERMODYnAMIC MODELS FOR tHE BEHAvIOR OF SuLFuR In MAGMAtIC SYStEMS Few true thermodynamic models exist that describe the behavior of sulfur in systems containing silicate melt +/− fluid (or gas at low pressure) and +/− sulfides or sulfates. This scarcity of models is due to the complex nature of sulfur’s behavior in silicate melts, in particular its different redox states. Although in most silicate magmas sulfur is dominated by S2−, even small amounts of S6+ affect sulfur solubility. Thus before considering possible thermodynamic models it is first necessary to review the effect of oxygen fugacity on sulfur speciation in magmatic systems.
Baker & Moretti
192 Sulfur speciation in silicate melts
Empirical models for sulfur’s behavior in magmatic systems do not require knowledge of the sulfur oxidation state and speciation in the melt, however thermodynamic models demand an accurate “accounting” of sulfur in the different species present in the melt. Wilke et al. (2011, this volume) provide a detailed review of sulfur speciation in silicate melts and glasses, but briefly the dominant sulfur species currently recognized in magmatic silicate melts are the ones discovered by Fincham and Richardson (1954): S2− and S6+, or sulfide and sulfate species. The stability of these species and any reactions between them must be considered and quantified in any thermodynamic model of sulfur’s behavior in silicate melts. Based upon the work of Carroll and Rutherford (1985, 1987, 1988), Wallace and Carmichael (1992, 1994) considered the sulfide–sulfate speciation equilibrium S2 −, silicate melt + 2O2fluid ↔ SO24 −, silicate melt
(48a )
which is obtained by subtracting Equation (14a) from Equation (13a), and yields the following thermodynamic relationship: 0 = ∆H − T ∆S + RT ln
silicate melt aSO 2− 4
(
melt aSsilicate fOfluid 2− 2
)
2
( 48b)
Wallace and Carmichael (1994) calibrated Equation (48a) with the measurements from Carroll and Rutherford (1985, 1987, 1988) and found log
silicate melt XSO 2− 4
X
silicate melt S2 −
= 1.02 log fO2 +
25410 − 10 T
( 49)
Note that the coefficient for logfO2 in Equation (49) is significantly smaller than that expected on the basis of the stoichiometry of Equation (48a): 1.02 instead of 2. This low value may be suggesting the presence of a non-stoichiometric reaction. Using this relationship Wallace and Carmichael (1994) demonstrated that submarine glassy lavas varied in oxygen fugacity from a little above the NNO buffer to 3 log units below it (Fig. 8). Additional studies by many research teams followed soon thereafter, and using this correlation between the ratio of sulfur species to the oxygen fugacities researchers investigated backarc magmas (e.g., Nilsson and Peach 1993), island arc magmas, and within-plate magmas (e.g., Métrich and Clocchiatti 1996) and confirmed that the common range of oxygen fugacities for igneous rocks is similar to that found by Wallace and Carmichael (1994, Fig. 8). Both Matthews et al. (1999) and Jugo et al. (2005b) provided updated, empirical calibrations of this oxygen fugacity determination. Moretti and Ottonello (2003a) made a preliminary thermodynamic assessment of this relationship between sulfur species and oxygen fugacity, ignoring any compositional dependencies (which below are shown to be significant), by combining data from the geological and the materials science literature to find that: log
silicate melt XSO 2− 4
X
silicate melt S2 −
= 2 log fO2 +
35078 − 7.6021 T
( 50 )
Note that the prefactor of the fO2 term in the Equation (50) is the correct stoichiometric coefficient of Equation (48a), 2, however this calibration yields fO2’s where the dominant sulfur species in the melt changes from sulfide to sulfate approximately 1 log unit lower than that of Wallace and Carmichael (1994) (Fig. 8; data from glass science and technology and are not reported in this figure, for those data see Moretti and Ottonello 2003a). Moretti and Ottonello (2003a) realized that the curve described by Equation (50) is only one of the many possible curves generated by shifting the constant term and the 1/T coefficient for any composition.
Modeling the Solubility of Sulfur in Magmas
193
Some of the difference in the locations of the curves in Figure 8 may be in part due to Moretti and Ottonello’s (2003a) use of results published in Clemente et al. (2004) on experimentally produced rhyolitic melts that on a molecular basis contained more sulfur than iron, a rare, if ever achieved, condition on Earth. All these studies, designed to find the oxidation state in the various magmatic and geodynamic settings, are affected by the simple approach applied to the speciation state of sulfur. However, for this approach to be accurate it requires that the activity coefficients of the sulfide and sulfate species be either the same in every melt composition or to cancel each other. Neither of these assumptions is likely to be true. Furthermore, there are large uncertainties associated with values at both high and low oxygen fugacities, i.e., sulfate fractions close to 1 and 0, respectively. Figure 8 demonstrates that the region in which the speciation of sulfur changes most rapidly with oxygen fugacity is between NNO−1 and NNO+1, the oxygen fugacity range of many terrestrial magmas. This dramatic change in speciation over a narrow range of oxygen fugacities reflects the transition that is seen near the so-called sulfur-solubility minimum (Fig. 2). Thus, this is the most strategic region to study if we want to understand the behavior of sulfur in natural silicate melts. However, before continuing with the construction of a complete thermodynamic model, the presence of other sulfur species in silicate melts must be considered. Glass science literature contains reports of the occurrence of other sulfur species, such as sulfite or dissolved SO2, and elemental sulfur, both inferred through voltammetric studies (Lafage and Taxil 1993; Tilquin et al. 1997) and Raman measurements (Konijnendijk and Buster 1977). However, their occurrence is always minor compared to sulfate. Sulfite has been also detected by gravimetry and titrimetry (Close and Tillman 1969) as well as by spectrophotometry (Pyare and Nath 1986). The occurrence of polysulfides has been reported by Ahmed et al. (1997), who detected S5−, S3− and S2− in alkali-borate glasses by means of Raman scattering and by Winther et al. (1998) who detected S3− and S2− in albite glass melted in the presence of graphite at 1 GPa. However, as Wilke et al. (2008, 2011 this volume) demonstrate, sulfide and sulfate species dominate in magmatic silicate melts and therfore thermodynamic models for sulfur equilibria need not be concerned with other sulfur species in most silicate melts. Consideration of any possible speciation equation based on coexistence of S2− and S6+ leads to an inevitable paradox (Moretti and Ottonello 2003a) as illustrated in Figure 9: an equilibrium constant which equals the activity of one of the reactants, raised to the power of its stoichiometric coefficient and therefore unaffected by the concentration of all other components in the reaction. From this analysis, some important features emerge: i)
Equilibrium species different than sulfate and sulfide may occur, possibly as anions not connected to the silicate structure. But they cannot modify the sulfate/sulfide ratio enough to appreciably eliminate the thermodynamic “paradox” involving Equation (48a). Moreover, their stoichiometry is such that they should appear around the “minimum” of sulfur solubility (Fig. 2), so that the relative amounts of any other intermediate species are expected to be negligible with respect to sulfide and sulfate anions.
ii) If we accept that Equations (13a), (14a) and (48a) relate “quasi-chemical” species of sulfide, S2−, and sulfate, SO42−, which may not actually correspond to discrete molecular entities, we implicitly recognize that each molten sulfide and sulfate present in the silicate melt contributes to the bulk solubility mechanism through its own chemical reaction. Thus the solubility of sulfide in the melt involves the sum of all reactions between sulfur in the fluid, in the sulfide melt or crystalline phase and each cation dissolved in the silicate melt (e.g., 2S2−, silicate melt + Si4+, silicate melt ↔ SiS2sulfide melt or crystal, 2S2-, silicate melt + Ti4+, silicate melt ↔ TiS2sulfide melt or crystal, 3S2−, silicate melt + 2Al3+, silicate melt ↔ Al2S3sulfide melt or crystal, …), not simply a single reaction involving
Baker & Moretti
194 40
2-
35
S + 2O2 = SO4
2-
30
log K
25 20
log K = -1.97 log fO
15
r = 0.98
10 5 0 -20
2
2
log K = log f
2-
SO4
- log f
S
2-
- 2 log f O
2
(after Moretti and Ottonello, 2003a) -15
-10
log f O
-5
0
2
Figure 9. Apparent sulfur redox equilibrium constant logK(49a) plotted against logfO2 (drawn after Moretti and Ottonello 2003a). See discussion in the text. Data are from both material science studies and geological systems (see Moretti and Ottonello 2003a, for the original data sources).
Fe2+ and sulfur, as has often been assumed in the past. In a similar manner the solubility of sulfate involves all reactions between SO42- and cations in the melt (e.g., 2SO42−, silicate melt + Si4+, silicate melt ↔ Si(SO4)2anhydrite, …), and the ratio of sulfate to total sulfide dissolved in the melt is a function of the interaction between all of the possible reactions. Therefore, as the concentration of the cations in the melt changes so must the position of the curve in Figure 8. This concept implies that there are many mechanisms, i.e., reactions, embodied in the equilibria presented in Equations (13a), (14a) and (48a) that must be considered. Although such a sulfur solubility model based upon a full thermodynamic treatment of the system is much more complicated than the empirical composition-weighted expressions discussed above, it is superior, as will be explained in the following paragraphs.
the thermodynamic model of Moretti and Ottonello As the reader will by now appreciate, dissolution of sulfur into natural silicate melts is complex and requires careful consideration of the oxidation state. Thus, constructing a rigorous thermodynamic model to describe the behavior is challenging. Indeed, it is so challenging that currently only one model for sulfur’s behavior in magmatic silicate melts at pressures of 1 atm and higher can be considered truly a thermodynamic model, that of Moretti and Ottonello (2003a,b, 2005). Although other thermodynamic models for silicate slag systems at 1 atm have been constructed (e.g., Reddy and Blander 1987; Pelton et al. 1993), we are unaware of any comprehensive model in the materials science literature that can model the interaction of sulfur with melts of equal compositional complexity to those found in nature. The construction of a thermodynamic model for the behavior of sulfur in magmatic systems requires defining the thermodynamic species present in silicate melts and determining their activities based upon the components normally found in a silicate melt and typically reported as oxides (SiO2, TiO2, Al2O3, Fe2O3, FeO, MnO, MgO, CaO, Na2O, K2O, P2O5, H2O, and CO2) as well as possible components for sulfur: H2Sfluid, SO2fluid, S2−, melt, SO42−, melt. The thermodynamic approach does not necessarily imply neglecting the need for structural models for silicate melts and knowledge of the actual microscopic species dissolved in them. However, thermodynamic
Modeling the Solubility of Sulfur in Magmas
195
approaches are macroscopic and in principle independent of structural considerations, although the choice of species and components reflect inferences about the microscopic nature of the investigated system and the reactions occurring therein. The non-ideal contributions to the activities of species in geological solutions are frequently expressed by Margules parameters, however the most-commonly used set of Margules parameters for silicate species in geological melts are those of Ghiorso and co-workers (e.g., Ghiorso et al. 1983), and their compilations are not suitable representations of the complex thermodynamics of the multiple oxidation state sulfur species dissolved in silicate melts and the exsolution or crystallization of sulfide and sulfate phases from silicate melts. Instead of using Margules parameters, Moretti and Ottonello (2003a,b, 2005) chose to consider the activities of individual ions in silicate melts for their thermodynamic model and apply thermodynamic formalisms commonly used for molten salts and polymer mixtures. A major advantage of this approach is its immediate application to acid-base and redox exchanges occurring in silicate melts, such as those involving divalent and trivalent iron (Ottonello and Moretti 2004; Moretti 2005) and, of course, sulfur-bearing species. In order to model the activity of a cation-anion complex, such as Mav+Ob2− (where a is the stoichiometric coefficient for cation M of charge v+ and b is the stoichiometric coefficient of oxygen of charge 2−), Moretti and Ottonello applied Temkin (1945) notation involving ion mixing over two “sub-lattices” (or matrices): aMav+ Ob2− = aM v+ aO2−
[M v+ ] = ∑[cations ]
a
[O2 − ] ∑[ anions ]
b
(51)
where the square brackets indicate the molar concentrations of ions in the melt. Calculation of the cation activity, aMv+, is readily accomplished given an analysis of the melt, but direct application of the Temkin model to compositionally complex silicate melts is impossible because the anion matrix (composition, number and size of anions) varies in a complicated fashion with composition, as exemplified by Equation (19) that describes the reaction between O2− (free oxygens) and Si-O-Si (O0, bridging oxygens) to produce 2 Si-O− (O−, non-bridging oxygens). This complexity is the origin of the difficulty in the application of even non-ideal, Margules-based, thermodynamic approaches. This complexity is due to the chemical reactivity in silicate melts that affects the anion matrix (Toop and Samis 1962; Ottonello et al. 2001), but which may be described by concepts developed in polymer chemistry. Recent developments in polymeric models (Ottonello 2001, 2005; Ottonello and Moretti 2004; Moretti 2005; Ariskin and Polyakov 2008) demonstrated that the mixing properties of silicate melts can be well reproduced by the distribution of anionic units, including oxide ions or “free oxygens.” The basis of these new polymeric models is still the Temkin model for ionic component speciation over anionic and cationic matrices. These ionic-polymeric models allow us to deal with the chemical complexity of silicate systems through application of a simplified description of the acid-base properties of components and of the basic thermodynamic quantities involved in chemical equilibria, including the activities of highly-reactive ionic species, such as those of sulfur. Therefore, ionic-polymeric models implement the Temkin mixing model to investigate the influence of melt composition on chemical reactions occurring in the system, particularly changes in the anion matrix due to polyanion speciation over it. The chemical reactions in silicate melts are acid-base and redox exchanges on which the well-known chemistry of aqueous solutions is based. But, with a difference: the characteristic process of an acid-base reaction in oxide systems, as defined by Flood and Forland (1947), is “the transfer of an oxygen ion from a state of polarization to another.” This concept equally applies to thermodynamic species in silicate melts, which can be considered oxygen-based solvents.
Baker & Moretti
196
The link between redox and acid-base exchanges in the Lux-Flood formalism is represented by the “normal oxygen electrode” redox equilibrium: 1 O 2 + 2e − ↔ O 2 − 2
( 52 )
A basic oxide is defined as being capable of dissociating to release oxygen ions and an acidic oxide is one that associates with oxygen ions to produce a base: Base ↔ Acid + O2−
(53a)
Notice the similarity with the reaction involving the breakdown of carbonic acid to produce bicarbonate and a hydrogen ion in an aqueous solution: H2CO3 ↔ HCO3 + H+
(53b)
except that in silicate melts the base dissociates to release the oxygen anion, whereas in aqueous solutions the acid dissociates to release the hydrogen cation (or proton). It is well established that the Lux-Flood acid-base property of dissolved oxides markedly affects the extent of polymerization by producing or consuming free oxygen ions in the melt (O2−). Thus, for a generic oxide M2/vO (Fraser 1975, 1977) two reactions are possible, one with acidic behavior: M 2 / v O silicate melt + O2 −, silicate melt ↔ M 2 / v O22 −, silicate melt
(54a)
and one with basic behavior: M 2 / v O silicate melt ↔
2 v + , silicate melt M + O2 −, silicate melt v
(54 b)
where, to reiterate, the type of behavior depends upon whether the O2− ion is produced or consumed. The Moretti and Ottonello thermodynamic approach rests on the principle that for each composition, at a given set of P-T values, the melt is characterized by an equilibrium distribution of several ionic species of oxygen (O0, O−, and O2−), metal cations, and ionic polymers composed of linked SiO44− units (or AlO45−, PO43−, etc.), and that the reactivity of individual polymeric units is the same, independent of the extent of bulk polymerization (Toop and Samis 1962). In order to calculate the activity of each species it is only necessary to compute the relative proportions of the oxygen quasi-species (O2−, O−, O0) that contribute to the anion matrix together with other anions (such as the sulfur anions S2− and SO42−) that enter into Equation (51). To solve the polymerization equation and calculate the fractions of oxygen quasi-species Moretti and Ottonello utilized the optical basicity of the melt. The optical basicity is a measure of the power of oxygen atoms in an oxide, or of anions (e.g., S), in a melt to donate electrons; alkali earth oxides have the highest optical basicity values, alkaline earth oxides have intermediate values, and the values of SiO2 and Al2O3 are lower still (see Duffy 1996 for a very readable introduction to optical basicity and Ottonello et al. 2001 for its application to silicate melt equilibria calculations). The equilibrium constant for Equation (19) is calculated using the reciprocal of the optical basicity, the “basicity moderating parameter” of the oxide-forming cation (gMv+ for modifiers, and gTh+ for formers), which represents the ability of the cation in an oxide to reduce the localized donor properties of the oxide, for each cation in the melt weighed by it abundance. Ottonello et al (2001) produced a parameterization of Equation (19) based on optical basicity through the following equation: K (19) = exp 4.662
(∑ X
Miv +
)
γ Miv+ − ∑ XT η+ γ T η+ − 1.1445 j j
(55)
Modeling the Solubility of Sulfur in Magmas
197
where XMv+ and XTh+ are respectively the atom fractions of network modifiers and network formers in one mole of the multicomponent melt or slag. Equation (55) establishes a formal link between acid-base properties of the medium (expressed as a balance between formers and modifiers’ basicities) and the polymerization constant. Although Equation (55) disregards temperature and pressure, it works properly when applied to the extensive database of sulfur solubility in silicate melts (Moretti and Ottonello 2003a,b, 2005). It is worth noting that adopting the difference in the basicity moderating parameters of network formers and network modifiers to quantify the bulk extent of polymerisation reactions in multicomponent melts has the same significance as adopting the difference in electronegativity between constituent atoms in predicting the heat of formation of simple compounds. The calibration of Ottonello et al. (2001) demonstrates that the equilibrium constant, K(19), is positively correlated with the concentration of network modifying cations (e.g., Na, K, Ca, Mg) and negatively correlated with the concentrations of network forming cations (e.g., Si, Al), as expected from inspection of Equation (19). The equilibrium constant for Equation (19) combined with the bulk composition of the melt allows calculation of the proportions of each type of oxygen quasi-species present (details in Ottonello et al. 2001 and Moretti 2005). The S[anions] term in Equation (51) is composed of the sum of O2−, OH−, S2−, SO42−, and S[structons], where structons are polymeric Si-O units in the melt. (Moretti and Ottonello 2005). The summation of all structons in the melt is calculated from the sum of the network forming cation concentration in the melt together with those of O0 and O− (see details in Ottonello et al. 2001; Moretti and Ottonello 2003b; Moretti 2005). Moretti and Ottonello (2003a,b, 2005) considered the following types of reactions involving sulfur-bearing components (fictive components) in silicate melts: M 2 / vSsilicate melt ↔ melt M 2 / vSO silicate ↔ 4
2 v + , silicate melt M + S2 −, silicate melt v
(56a )
2 v + , silicate melt M + SO24 −, silicate melt v
(56b)
Reactions based upon Equations (56a) and (56b) can be constructed for each cation present in the system and combined with reactions describing the solution of sulfur into the melt from the gas or fluid phase: 1 fluid 1 S2 + O2 − , silicate melt ↔ O2fluid + S2 − , silicate melt 2 2
(57a )
1 3 O2 −, silicate melt + S2fluid + O2fluid ↔ SO24 −, silicate meltt 2 2
(57b)
to yield equations such as 1 1 M 2 / v O silicate melt + S2fluid ↔ M 2 / vSsilicate melt + O2fluid 2 2
(58aa )
3 1 melt M 2/v O silicate melt + S2fluid + O2fluid ↔ M 2 / vSO silicate 4 2 2
(5 58b)
The following equilibrium constants are derived for Equations (58a) and (58b): melt aOfluid aMsilicate 2/v S 2 K O-S, M (58a ) = silicate fluid melt a M2 / v O aS2
1
2
(59a )
Baker & Moretti
198
melt aMsilicate 2 / v SO 4
K O-SO4 , M (58b) =
melt aMsilicate 2/v O
−3 2
(a ) (a ) fluid O2
fluid S2
−1 2
(59b)
Although equilibrium constants K(59a) and K(59b) refer to a specific oxide-forming cation M, the oxide-sulfide and oxide-sulfate disproportionations for all cations in multicomponent slags or silicate melts may be readily generalised in terms of the Flood-Grjotheim thermochemical cycle (see details in Flood and Grjotheim 1952, and the generalization provided by Moretti and Ottonello 2005) to create composite equilibrium constants for oxide-sulfide and oxidesulfate exchanges (Ka′O-S and Ka′O-SO4, respectively) constructed from the weighted sum of the individual equilibrium constants (such as in Eqn. 59a,b) for each cation: ′ = N ′Av+ ln C A K O-S, A + N B′ v + ln C B K O-S, B + . . . ln K O-S
(60a )
′ 4 = N ′Av+ ln C A KO - SO4 , A + N B′ v + ln C B K O-SO4 , B + . . . ln K O-SO
(60 b)
or in a general form (for instance for sulfide): n
′ = ∑ N M′ v+ ln CO-S, Mi K O-S, Mi ln K O-S i
(61)
i =1
where N′ are the “electrically equivalent fractions,” i.e.: vi+ nMiv+
N M′ iv+ =
(62)
∑v n
+ i Miv +
and CA, CB, etc. are the energetic differences relating the different standard states of pure components (oxide, sulfide, sulfate) in the melt phase at the P-T of interest and pure components (oxide, sulfide, sulfate) in the pure liquid phase at the same conditions. Because the polymeric approach assumes a Temkin model solution of constituent salts (Eqn. 51), the electrically equivalent fractions are calculated for the appropriate matrix, either cationic or anionic. Let us now consider the oxide-sulfide disproportionation equilibria in the context of the polymeric approach. The approach does not involve any modification of the Toop-Samis conceptual framework. The reference state condition for M2/vS consistent with the Toop-Samis development is one of completely dissociated sulfide in solution in which the (Temkin model) activity is one, the same as for the oxide components. We have thus, for instance, that for a particular cation Mv+, the dissolution (“diss.” in the equations below) equilibria in the Temkin model demand: melt silicate melt aMsilicate aO2− v+
a
silicate melt M2/ v O
melt silicate melt aMsilicate aS2− v+ melt aMsilicate 2/ v S
(
melt aMsilicate v+
)
2 v
silicate melt aSO 2− 4
melt aMsilicate 2 / v SO 4
1 K (54a )
(63a )
= K diss. M2 / v S = K (56a )
(63b)
= K diss. M2 / v O =
= K diss. M2/ v SO4 = K (56 b)
(63c)
The reference state is such that equilibrium constants K(54a), K(56a) and K(56b) each reduce to one. We may ascribe this apparent inconsistency as due to the energy difference between the standard state of completely dissociated, Temkin model oxide, sulfide and sulfate components, for which
Modeling the Solubility of Sulfur in Magmas
( + RT ln ( a
199
silicate melt silicate melt S2 − M v+
) )
(64 b)
silicate melt melt melt µ silicate = µ*,M2silicate + RT ln aMsilicv+ate melt aSO 2− M2 / v SO 4 / v SO 4 4
)
(64c)
melt melt melt silicate melt µ silicate = µ*,M2silicate + RT ln aMsilicate aO2− v+ M2 / v O /v O melt melt µ silicate = µ*,M2silicate M2 / v S /vS
a
(
(64a )
and the true melt components, whose structural condition in the polymeric network demands a standard state of “pure component in the melt phase” that is different than that of “pure component in pure liquid phase,” and for which we have: e melt melt melt µ silicate = µ oM, 2silicate + RT ln aMsilicat M2 / v O /v O 2/v O
(65a )
melt melt melt = µ oM, 2silicate + RT ln aMsil2icate µ silicate M2 / v S / v S, melt /vS
(65b)
melt melt melt = µ oM, 2silicate + RT ln aMsilicate µ silicate M2 / v SO 4 / v SO 4 , melt 2 / v SO 4
(65c)
Therefore, the oxide-sulfide and oxide-sulfate equilibria (58a) and (58b) require: 2−
S O2 −
melt melt melt melt µ oM, silicate − µ oM, 2silicate + µ*,M2silicate − µ*,M2silicate S /v O /v O /vS = K O-S, M exp 2 / v RT
a fluid 2 ′ M Sfluid = K O-S, aO 2
aSfluid 2 fluid aO2
1
2
(66a )
1 2
and ate melt melt melt melt µ oM, 2silicate + µ*,M2silicate − µ*,M2silicate − µ oM, 2silic SO24 − / v SO 4 /v O /v O / v SO 4 = K exp O-SO4 , M RT O2 −
(
1 fluid 2 S2
′ 4 ,M a = K O-SO
fluid aS2
(
1 2
)(
aOfluid 2
)
3 2
(66 b)
3 fluid 2 O2
) (a )
Equations (66a) and (66b) contain the energy difference between the two standard states, which is the key with which solubility experiments based on heterogeneous equilibria (58a) and (58b) may be compared to a solubility model adopting ionic fractions on ‘structural’ sites. The lefthand terms of Equations (66a) and (66b) refer to concentration ratios and equal the activity ratios for these species, whose activity coefficients may be assumed to be identical within a reasonable approximation. By using the definition of sulfide capacity it follows that
(
′ , M O2 − CS,c = K O-S
)
fO°2 ° fS2
1
2 MS ∑ mol oxides
(67)
with: Smoloxides = molar summation of all oxides in the system per 100 g of material, MS = atomic weight of sulfur, and f°O2 and f°S2 are the fugacities of gaseous components at their 1-bar standard state. An equation similar to Equation (67) can be written for the sulfate capacity. The equilibrium constant for Equation (58a), i.e., Equation (59a), can be expressed with an Arrhenian dependence (e.g., for sulfide dissolution): ′ , M = AO-S ′ ,M + ln K O-S
′ ,M BO-S T
(68)
200
Baker & Moretti
Moretti and Ottonello (2003a,b; 2005) started with the tabulated thermodynamic data for liquid sulfide-oxide and sulfate-oxide exchanges in the presence of S2 and O2 in order to define such Arrhenian dependencies. They then refined the data using experimentally measured sulfide and sulfate capacities to calibrate the constant C terms appearing in Equations (60a,b) and (61), which sum to the original entropic terms of the liquid sulfide-oxide and sulfate-oxide exchange. These terms can be considered as annealing entropies, needed to change from the reference state condition of “pure liquid component” to that of “pure melt component” or (e.g, for sulfide): annealing ′ Mi − AO-S, Mi ) ∆SO-S, = RCO-S, M = R ( AO-S, Mi
(69)
In those cases in which thermodynamic data for pure liquid sulfides and sulfates were not available (e.g., TiS2; FeIII2S3 and many sulfate components), the enthalpic ‘B’ parameters in Equation (68) have been constrained by the sufide and sulfate capacity database (Moretti and Ottonello 2005). The model extension to compute sulfur dissolution at P > 1 bar is based on the computation of the equilibrium constant shift with P for any specific oxide-sulfide and oxide-sulfate exchange. In the case of exchanges of the type of Equation (58a), we have for example: ′P -TM = AO-S, ′ M+ ln K O-S,
′ M BO-S, 1 − T RT
P
∫ ( ∆V
m
+ ∆V fl ) dP
(70)
1
where subscripts m and fl refer melt and fluid (or gas) components, respectively. Details on the data used and the strategies applied to reduce the number of unconstrained parameters can be found in Moretti and Ottonello (2005) and in Moretti and Baker (2008). The success of this thermodynamic model is displayed through the comparison of the modeled and measured sulfide capacity, CS, and sulfate capacity, CSO4, as defined by Equations (17) and (18) (Fig. 10), and its capability to assess sulfur solubility in both relatively simple systems, such as metallurgical slags, and complex systems, such as natural silicate melts (Fig. 11). For these reasons, the model can be used to define the compositional dependencies of the sulfide and sulfate capacities of different hydrous magmatic compositions (Fig. 12). From the CS and CSO4 information provided by such diagrams, it is possible to relate sulfur mass partitioning with gas fugacities and provide a full characterization of the system by adopting the following relationship: melt silicate melt melt Stotal = Ssilicate + Ssilicate = CS fO−21/ 2 fS12/ 2 + CSO4 fO32/ 2 fS12/ 2 sulfide sulfate
(71)
where in this case the sulfur concentrations and the CS and CSO4 are expressed in weight percent. This equation can be used to estimate the total sulfur concentration in melts, or as oxybarometer, or as a sulfur barometer, depending on the needs and information available. Moretti and Baker (2008) built upon the model of Moretti and Ottonello (2003a,b, 2005) to calculate the saturation of silicate melts with sulfide (either molten or crystalline). They calibrated their model for sulfide saturation using all available data in the literature and demonstrated possible complexities of sulfur’s behavior in magmatic systems (discussed below). In addition, they stressed that some of the complexities to be expected in magmatic systems may not be discerned when using simple, empirical models for the saturation of silicate melts with sulfides. They also updated the equation of Bockrath et al. (2004) that calculates the sulfur fugacity at sulfide-saturated conditions. Unfortunately, no similar thermodynamic model for the saturation of magmatic-composition melts with sulfate is currently available. Software for the application of the models of Moretti and Ottonello (2005) and Moretti and Baker (2008) are available from R. Moretti.
Modeling the Solubility of Sulfur in Magmas
(a)
201
0
Modeled log C S
-2
Sulfur solubility model of Moretti and Ottonello (2003, 2005)
-4
-6
-8
Anhydrous melts Hydrous melts
10% -10 -10
(b)
-8
-6
-4
-2
Measured log CS
0
20
Modeled log C SO
4
18
Sulfur solubility model of Moretti and Ottonello (2003, 2005)
16
10%
14 12 10 8 Anhydrous melts Hydrous melts
6 4 4
6
8
10
12
14
Measured logC SO
16
18
20
4
Figure 10. Comparison of the calculated CS, in weight percent, defined by Equation (17), for sulfide (Fig. 10a), and the calculated CSO4, in weight percent, defined by Equation (18), for sulfate (Fig. 10b), with the measured values using the conjugated Toop-Samis-Grjotheim model of Moretti and Ottonello (2003a,b; 2005). Note that the agreement between measurements and model calculations is almost always within 10%. (Drawn after Moretti and Ottonello 2005.)
Although the polymeric model used by Moretti and Ottonello (2003a,b, 2005) is but one of many ways to model silicate melt thermodynamics, it provides a useful formalism to constrain reactions and interaction parameters. Margules-like approaches rarely provide the correct interaction terms for the same component in both compositionally simple (slags) and complex (natural melts) systems. But, applying the polymerization model allowed Moretti and Ottonello (2003a,b, 2005) to find the best solutions for the dissolution of sulfur (as well as other elements) in limiting binaries (e.g., SiO2-CaO) and extrapolate them to more complex systems.
Baker & Moretti
202
5000
DB3 synthetic basalt
4000
S (ppm)
a)
T=1400°C P=1 bar
3000 2000 1000
d)
0 -10
-8
-6
-4
logfO2 - logfS2 5000
S (ppm)
4000
b)
MORB T=1400°C P=1 bar
3000 2000 1000
e)
0 -10
-8
-6
-4
logfO2 - logfS2 2500 Cape Vogel andesite
S (ppm)
2000
T=1400°C P=1 bar
1500 1000 500
f)
0 -10
-8
-6
-4
logfO2 - logfS2 2000
c)
hydrous dacite
S (ppm)
1500 T=992°C P=2083 bar
T=798°C P=2034 bar
1000 500
g) 0 -17
-16
-15
-14
-13
-12
-11
-10
-9
logfO2 - logfS2
Figure 11. Solubility of sulfur in silicate melts as sulfide: predictions from the model presented by Moretti and Ottonello (2003a,b, 2005) vs. experiments for both simple slags of metallurgical interest (panels a to c) and natural melts or their analogs (panels d to g). Experimental data (gray circles) are from Abraham et al. (1960) in panel a, Kalyanram et al. (1960) in panel b, Brown et al. (1982) in panel c, O’Neill and Mavrogenes (2002) for panels d to f (gray triangles) and from Clemente et al. (2004) in panel g (gray squares and circles). Black diamonds in panels d to g are computations from the Moretti and Ottonello (2005) model; symbols have been connected by a heavy line as a guide for the eye in each panel. Asterisks in panels d to f are computations based on the model of O’Neill and Mavrogenes (2002); the thin line connecting them has been drawn as a guide for the eye in each panel. Panels e to g are modified from Moretti and Ottonello (2005).
Modeling the Solubility of Sulfur in Magmas
(a)
-4 -5
log C S
-8
o
1050 C
-7
50 MPa
1200 C
1100 C
o
-6
o
o
1150 C
Kilauea basalt melt
200 MPa
100 MPa
203
500 MPa 400 MPa o 900 C 300 MPa o
850 C o 800 C o 750 C o 700 C
10 MPa
-9 -10 -11 -12 -13
Rhyolitic melt
-14 0
(b)
2
4
8
6
H2O (wt%)
22
10
Rhyolitic melt o
700 C
20
o
750 C
18
o
log C SO
4
800 C
16 14
o
850 C 10 MPa
50 MPa 100 MPa
12 10 8 0
200 MPa
300 MPa
o
1050 C o
1100 C
Kilauea basalt melt 2
o
400 MPa 900 C 500 MPa
o
1150 C o
1200 C
4
6
H2O (wt%)
8
10
Figure 12. (a) The sulfide capacity, CS (in wt%) as defined by Equation (17), vs. water concentration in the melt for a water-saturated Kilauea basaltic melt composition at 1050 to 1200 °C (upper 4 curves) and a water-saturated rhyolitic melt at 700 to 900 °C (lower 5 curves) and pressures from 1 atm (0.0 wt% H2O) to 500 MPa as calculated by the model of Moretti and Ottonello (2005). (b) Sulfate capacity, CSO4 (in wt%) as defined by Equation (18) vs. water concentration in the melt for a water-saturated Kilauea basaltic melt at 1050 to 1200 °C (lower 4 curves) and a water-saturated rhyolitic melt at 700 to 900 °C (upper 5 curves) and pressures from 1 atm (0.0 wt% H2O) to 500 MPa, as calculated by the model of Moretti and Ottonello (2005). The subvertical dashed lines in both figures indicate the pressures, in MPa. Water contents calculated with the model of Papale et al. (2006). Please see text for further discussion. (Drawn after Moretti and Ottonello 2005.)
GEOLOGICAL APPLICAtIOnS The relative solubility (or partitioning) of sulfur between a fluid or gas phase and a silicate melt, as well as the saturation of the system with either a sulfide (liquid or crystalline) or a sulfate, depend upon pressure, temperature, fO2, fS2 and composition. Despite the lack of critical data needed for modeling the behavior of sulfur (e.g., hydrous, basic and intermediate melt
Baker & Moretti
204
compositions under conditions of known oxygen and sulfur fugacities), after half a century of research we are to the point where quantitative, or at least semi-quantitative, models can be constructed that constrain the evolution of sulfur in magmatic systems. When complete compositional and fugacity information is available for a magmatic system the behavior of sulfur can be more-or-less easily modeled. However, it is rare that there is enough data for a backward model of sulfur in the magmatic system. Instead, forward models are typically created and matched with the rocks in hand. A recent example of forward modeling is presented in Moretti and Baker (2008) who investigated the effects of intensive variables on sulfide saturation in a generic rhyolitic melt with 0.5% H2O at 800 °C, a rhyolitic melt at water-saturated conditions and 800 °C and a basaltic melt with 0.5 wt% H2O at 1200 °C, all at crustal conditions. The oxygen fugacity width of the transition region between S2− and S6+ is similar for all combinations of melts and water concentrations modeled by Moretti and Baker (2008). However, the oxygen fugacity at which the calculated sulfate species becomes the dominant species in the melt changes by 2 orders of magnitude as composition, pressure, temperature and water concentration change (Fig. 13). This point occurs at oxygen fugacities between approximately DNNO-0.5 at 100 MPa and DNNO+0.25 at 500 MPa for both the basalt at 1200 °C and water-poor rhyolite at 800 °C. Whereas for the water-saturated rhyolite at 800 °C the change occurs at DNNO+0.5 at 100 MPa and DNNO+2 at 500 MPa. These differences are due to compositional effects on the activities of the sulfide and sulfate species dissolved in the melt. The compositional, pressure, and temperature effects demonstrated in Figure 13 are a caution on the extent to which we can rely on any general correlation between the ratio of sulfate to sulfide in the melt as a measure of oxygen fugacity during magmatic evolution (e.g., Fig. 8). The SCSS for silicate melts changes significantly near the NNO buffer (Fig. 14). Below this oxygen fugacity the saturating concentration of sulfur in the melt is constant and increases with increasing water concentration at constant temperature, as well as with decreasing silica concentration and increasing temperature, as discussed previously in the review of experimental 1.0 0.9
o
Basalt, 0.5% H 2O, 100 MPa, 1200 C o
Basalt, 0.5% H 2O, 500 MPa, 1200 C o
Basalt, 0.5% H 2O, 2 GPa, 1200 C
0.7
Rhyolite, H 2O saturated, 500 MPa, 800 C
0.6
Rhyolite, 0.5% H 2O, 500 MPa, 800 C
/Total S
0.8
o
Rhyolite, H 2O saturated, 100 MPa, 800 C o
o
Rhyolite, 0.5% H 2O, 100 MPa, 800 C o
0.4
X-S
6+
0.5
0.3
0.2 0.1 0.0 -4
-3
-2
-1
0
∆ NNO
1
2
3
Figure 13. Comparison of the calculated molar ratio of S6+/Total S (X-S6+/Total S) in melts as a function of DNNO for basaltic and rhyolitic melts at different temperatures, pressures and water concentrations in the melt. Drawn after Moretti and Baker (2008). Note how unlike in Figure 8, the curves are functions of the intensive thermodynamic variables describing the melt, e.g., pressure, temperature and composition. Please see the text for further discussion.
Modeling the Solubility of Sulfur in Magmas
205
o
Basalt, 0.5% H2O, 100 MPa, 1200 C o
Basalt, 0.5% H2O, 500 MPa, 1200 C o
Basalt, 0.5% H2O, 2 GPa, 1200 C
6
o
Rhyolite, H2O saturated, 100 MPa, 800 C
log SCSS or SCAS (in ppm)
o
Rhyolite, H2O saturated, 500 MPa, 800 C o
5
Rhyolite, 0.5% H 2O, 100 MPa, 800 C o
Rhyolite, 0.5% H 2O, 500 MPa, 800 C
4 3
Basalt
2 1
Rhyolite, water-saturated
0 -1 -2
Rhyolite, 0.5 wt% H 2O
-3
-2
sulfide
-1
0
∆ NNO
1
sulfate
2
3
Figure 14. Comparison of the models of sulfide saturation by Moretti and Baker (2008) and Liu et al. (2007) for the SCSS, and the new sulfate saturation model of the SCAS as functions of fO2, displayed as DNNO. A basaltic melt and a rhyolitic melt with differing water concentrations are modeled. The lines correspond to the calculations of Moretti and Baker (2008) and the open and filled symbols represent the SCSS (using Liu et al. 2007) and SCAS (using this model in this paper), respectively. The circles correspond to the basaltic melt at 1200 °C with 0.5 wt% H2O at 100 MPa; the squares correspond to the rhyolitic melt at 800 °C with 0.5 wt% H2O at 100 MPa, and the diamonds correspond to the water saturated (11 wt% H2O) rhyolitic melt at 800 °C, 500 MPa. Other calculations using empirical models were omitted for clarity but produce similar results to those displayed. No solubility minimum is seen in the SCSS or SCAS of these sulfide- or sulfatesaturated melts, unlike what is observed in systems where only a gas phase is in equilibrium with a melt and oxygen and sulfur fugacities are controlled (Fig. 2). Please see the text for further discussion.
studies. Above the NNO buffer the SCSS increases rapidly with a slope defined by the stoichiometry of the reactions precipitating FeS, which is crystalline pyrrhotite for the rhyolitic melts and liquid FeS for the basaltic melt. Indeed, the calculations indicate SCSS values that exceed 100%. This impossibility is prevented by the limit of FeS (crystalline or liquid) stability with a silicate melt shown in Figure 14, as estimated from the data of Jugo et al. (2005a,b) and consistent with the stability of an FeS melt formed at NNO+1.4 in the study of Liu et al. (2007). Thus, above this limit the calculated values of the SCSS represent metastable equilibria. This limitation is not considered prohibitive because the vast majority of terrestrial silicate magmas (and possibly magmas in the solar system) appear to have oxygen fugacities below NNO+1.4 (e.g., Carmichael 1991). The trend of the SCSS as a function of oxygen fugacity demonstrates that there is no solubility minimum when the melt is saturated with a sulfide, unlike the sulfideundersaturated case where the minimum is seen when the sulfur fugacity is fixed and the oxygen fugacity varies at constant temperature (Fig. 2). Thus, we do not expect to see such a solubility minimum in natural rocks saturated with a sulfide or sulfate. However, the modeling reported in Figure 14 demonstrates that when a high-temperature, sulfide-saturated magma is mixed with a lower temperature magma the SCSS of the mixed melt may be low enough to exsolve significant amounts of sulfur-bearing gas and act as an eruption trigger, as proposed for the 1991 eruption of Mt. Pinatubo (Kress 1997). The effect of pressure on the SCSS is complicated by the presence, or absence, of fluid. At fluid-absent conditions the effect of pressure is to lower the SCSS, whereas the presence of fluid in the case of the
206
Baker & Moretti
water-saturated rhyolite results in an increase in the SCSS with increasing pressure (Fig. 14), as seen experimentally. The differing behavior is a result of the subtle effects of maintaining thermodynamic equilibrium between the fluid, melt and sulfide phases and the increasing water concentration in the water-saturated melt with increasing pressure. In many cases the details of the intrinsic parameters, particularly fS2 and possibly fO2 may not be well known. In those cases a more simple empirical model, such as those by Liu et al. (2007) and Li and Ripley (2009), may be more useful. These models are very good at reproducing the data set upon which they are based and have sometimes been tested with independent measurements of the SCSS. However, as exemplified by the comparison of calculations and experiments in Moune et al. (2009), these models still have their shortcomings. Comparing the model of Moretti and Baker (2008) with the empirical model of Liu et al. (2007) demonstrates that both models are only interim solutions for the calculation of the SCSS (Fig. 14). The two models are in good agreement for the SCSS for basaltic melts with 0.5 wt% H2O at 1200 °C, 100 MPa and fO2’s at and below the NNO buffer. At these oxygen fugacity conditions the SCSS is not sensitive to changes in fO2. However, at higher fO2’s the model of Liu et al. (2007) does not predict the orders of magnitude increase in the SCSS seen in the Moretti and Baker model. Although the Liu et al. (2007) model shows little sensitivity to fO2, the two models are in general agreement for the rhyoltic melts, both water-saturated and with low water concentrations above the NNO buffer at 800 °C (Fig. 14). However, below the NNO buffer the Liu et al. (2007) model predicts values of the SCSS that are one order of magnitude higher than the Moretti and Baker (2008) model for water-saturated rhyolitic melts and about 3 orders of magnitude higher than for rhyolitic melts with low water concentrations. Comparison of experimental results in the studies discussed above with the SCSS calculated by the two models indicates that most probably the true values for the SCSS of water-poor rhyolitic melts fall between the predictions made by the two models. The SCAS calculated with the empirical model introduced in this contribution is plotted at fO2’s above NNO+1.5 (Fig. 14). Over the fO2 range considered, the effect of fO2 on the SCAS is negligible, providing, of course, that the fO2 is in the sulfate stability region. The calculations demonstrate that as the sulfur-rich, saturating phase changes from sulfide to sulfate the concentrations of sulfur in the basaltic melt and in the rhyolitic melt with 11 wt% water increase by factors of approximately 2× to 8×. Interestingly, the predicted SCAS of the low water concentration rhyolitic melt is lower than its SCSS. Although the SCAS model reproduces the calibration experiments accurately (within less than 5% in most cases, Fig. 7), the surprising behavior seen at small concentrations of water in the rhyolitic melt may be an artifact of the few data available for calibration in this compositional region. As discussed above, empirical models should be considered as “works in progress” and none accurately predict the SCSS or SCAS in the oxygen fugacity region where the sulfate/sulfide ratio changes rapidly (Fig. 8), but they can provide constraints on the sulfur concentrations at saturation with either sulfide or sulfate in this fO2 region. Investigating the dissolution and exsolution of sulfur from multicomponent gases into silicate melts can produce apparently complicated patterns of sulfur concentrations in melts. Interpreting data from natural magmatic systems can be misleading if we do not consider nonlinear effects that apparently modify the simple stoichiometric relationships as exemplified by the sulfur solubility minimum (Fig. 2), which can be derived directly from Equations (13a) and (14a). Inspecting data plotted on a log S (in either wt% or ppm) vs. log fO2 diagram could lead to the unwarranted conclusion that the solubility minimum does not exist (e.g., Scaillet and Pichavant 2003), because it is not observed in this kind of plot. However, it should be recalled that the solubility minimum is seen when plotting sulfur concentration versus fO2 at constant SO2 fugacity (Fig. 2). Inclusion of the Moretti and Ottonello (2003a,b, 2005) model into a general model for CO2-H2O-SO2-H2S gas-silicate melt equilibria (Moretti et al. 2003) demonstrates how differently the behavior of sulfur can appear. Applying this model to the behavior of sulfur
Modeling the Solubility of Sulfur in Magmas
207
in natural, fluid-saturated magmas does not produce a minimum in the dissolved sulfur concentration in the melt as the fO2 changes (Fig. 15a). However this model nevertheless respects the stoichiometric constraints embodied in the sulfur dissolution reactions, if the dissolved sulfur concentration in the melt is normalized by the fugacity of SO2 (Fig. 15b). This example demonstrates some of the pitfalls that empirical models with simple mass partitioning can produce and reminds us of the utility of constructing and applying the more-complex and rigorous thermodynamic models.
(a)
5000
S (ppm)
4000
3000
S
2000
Total
4+
S 1000
2-
Basaltic melt in equilibrium with fluid 10 MPa, 1400 K
S
after Moretti et al. (2003)
0
(b)
-2
-1
0
1
∆ NNO
2
-1 SO2 + O
2
log [S (wt)/f SO ]
-1.5
2-,melt
-2
= 3/2 O 2 + S
2-,melt
SO2 + 1/2 O 2 + O
2-,melt
2-,melt
= SO4
-2.5 -3 Basaltic melt in equilibrium with fluid 10 MPa , 1400 K
-3.5
after Moretti et al. (2003)
-4
-2
-1
0
∆ NNO
1
2
Figure 15. Example of multicomponent gas-melt saturation (Moretti et al. 2003) of a tholeitiic melt at 10 MPa, 1400 K, under progressively oxidizing conditions and for total (exsolved + dissolved) volatile contents of 3 wt% H2Ototal, 1 wt% CO2total and 0.5 wt% Stotal. (a) The concentration of sulfide, sulfate and total dissolved sulfur in the melt are plotted as a function of fO2, expressed as DNNO. Note the lack of any observable sulfur concentration minimum, unlike what is seen in Figure 2, because the fSO2 in this case varies with fO2. (b) The concentration of sulfur dissolved in the melt normalized to the fSO2 as a function of fO2. When the data are plotted in this manner the sulfur concentration in the melt calculated by the model of Moretti et al. (2003) respects the theoretical slopes of the hypothesized reactions for sulfur dissolution in silicate melts (Fig. 2). (Drawn after Moretti et al. 2003.)
208
Baker & Moretti
Although we must acknowledge the current limitations of all the models investigated, they are still useful tools for petrogenetic and volcanic investigations, as many of the papers referenced in this contribution demonstrate. In the latter role, the model of Moretti and Ottonello (2003a,b; 2005) was one of the tools used by Aiuppa et al. (2007) to forecast volcanic eruptions at Mt. Etna, Italy. The interpretation of sulfur isotope data in magmatic systems (Marini et al. 2011, this volume) is another good example of the importance that composition-dependent models have in studying the saturation and speciation properties of sulfur in melts. The models for the SCSS and SCAS provide constraints on the maximum amount of sulfur that can be dissolved in magmatic melts. Any additional sulfur (the “excess sulfur”), often necessary to balance the observed sulfur degassing of volcanoes with the measured sulfur concentrations in quenched glasses, must therefore have been stored in another phase (usually considered to be the gas or fluid phase) of the volcanic system. The “excess sulfur”, above that which can be stored in the melt, can be used to estimate the mass of the gas phase present at depth. Thus, constraints on the possible masses of gases trapped in volcanic systems and the potential mass of sulfur degassed from volcanoes into the atmosphere can be assessed using models for the SCSS and the SCAS. Understanding the conditions at which a magma becomes saturated with a sulfide phase is directly applicable to models for the formation of ore deposits and therefore to our search for these resources (Barnes 1979; Naldrett 1989; Simon and Ripley 2011, this volume). Knowledge of the SCSS for magmas of differing composition can be applied to the search for high-temperature, sulfide-bearing mineral deposits, particularly for magmatic sulfide deposits such as those associated with mafic magmatic systems (e.g., Thakurta et al. 2008). In particular, models for the SCSS may play a key role in the search for Ni-Cu ore deposits, which produce not only these two metals but also more-valuable platinum group metals (e.g., Koek et al. 2010). Combining models for the SCSS with those for the exsolution of volatiles (e.g., Papale et al. 2006) can predict whether ore metals would be found associated with sulfides trapped within the magmatic system or if they would have been transported away in the fluid phase (Barnes 1979). Such knowledge can provide a valuable guide in the search for ore deposits.
COnCLuSIOnS We characterize our current understanding of the behavior of sulfur in silicate melts of magmatic compositions as “good.” We have a basic understanding of the mechanisms of sulfur dissolution in compositionally complex silicate melts and a choice of both empirical and thermodynamic models for the calculation of the SCSS and the SCAS. However, none of the models are perfect, and, in general, their accuracies are approximately ± 10% (relative). Nevertheless, this level of accuracy is sufficient for many petrological investigations and can provide constraints on the evolution of sulfur in magmatic systems. There is no doubt the models and their ability to predict the saturation of silicate melts with sulfur-rich phases will improve in the future. However these improvements require additional experimental data, particularly in the critical fO2 range near the NNO buffer where the speciation of dissolved sulfur changes from S2− to S6+ with increasing fO2. In particular, systematic experiments need to be performed over a wide range of melt and fluid compositions (including variations in water and halogen concentrations), temperatures and pressures, with controlled fugacities of oxygen, sulfur and water. Such experiments are technologically challenging and few laboratories have the resources to perform such experiments, but they are critical in the advancement of our understanding of sulfur in magmatic systems. Other technologically challenging experiments involve direct determination of basic thermodynamic properties (calorimetric and volumetric properties) of sulfur species in melts. However, the low concentrations of the saturation limits imposed by sulfide (either crystalline or liquid) and sulfate (anhydrite) precipitation represent a major obstacle in this endeavor.
Modeling the Solubility of Sulfur in Magmas
209
Sulfur is a key element whose behavior in magmatic systems must be understood because of its importance in ore deposit formation, volcanic eruption prediction, and global climate change. The models discussed in this contribution, as well as future models, will contribute to this understanding and allow us to better constrain sulfur’s role as we search for the increasing resources our society needs, seek better techniques to predict hazardous volcanic eruptions, and improve our understanding of the impact of volcanic degassing on global climate change.
ACknOwLEDGMEntS Many thanks go to H. Behrens and J.D. Webster for their excellent reviews of an earlier version of this manuscript that improved it significantly. RM acknowledges financial support from MIUR (Italian Government) in the frame of the PRIN 2007 project (INGV-Naples research Unit). Support for D.R.B. research has come from his NSERC Discovery grant; this paper was written while D.R.B. on sabbatical at Sincrotrone Trieste who graciously provided both hospitality and a stimulating environment for work.
REFEREnCES Abraham KP, Davies MW, Richardson FD (1960) Sulphide capacities of silicate melts. J Iron Steel Inst 196:309-312 Abraham KP, Richardson FD (1960) Sulphide capacities of silicate melts, Part II. J Iron Steel Inst 196:313-317 Ahmed AA, Sharaf NA, Condrate RA Sr (1997) Raman microprobe investigation of sulphur-doped alkali borate glasses. J Non-Cryst Solids 210:59-69 Aiuppa A, Moretti R, Federico C, Guidice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007) Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology 35:1115-1118 Andres RJ, Kasgnoc AD (1998) A time-averaged inventory of subaerial volcanic sulfur emissions. J Geophys Res 103:25251-25261 Ariskin AA, Bychkov KA, Danyushevsky LV, Barmina GS (2008) A model of S solubility in basaltic melts at 1 atm. Geochim Cosmochim Acta 72: Supplement 1, A31 Ariskin AA, Polyakov VB (2008) Chemical interactions and configurational disorder in silicate melts. Geochem Inter 46:429-447 Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Barnes HL (ed) (1979) Geochemistry of Hydrothermal Ore Deposits. 2nd ed. Wiley, New York Bockrath C, Ballhaus C, Holzhied A (2004) Stabilities of laurite RuS2 and monosulfide liquid solution at magmatic temperature. Chem Geol 208:265-271 Botcharnikov RE, Behrens H, Holtz F, Koepke J, Sato H (2004) Sulfur and chlorine solubility in Mt. Unzen rhyodacitic melt at 850 °C and 200 MPa. Chem Geol 213:207-225 Bradbury JW (1983) Pyrrhotite Solubility in Hydrous Albite Melts. Ph. D. dissertation, Penn. State Univ. Brown SD, Roxburgh RJ, Ghita I, Bell HB (1982) Sulphide capacity of titania-containing slags. Ironmaking and Steelmaking 9:163-167 Buchanan DL, Nolan J (1979) Solubility of sulfur and sulfide immiscibility in synthetic tholeiitic melts and the relevance to Bushveld-complex rocks. Can Mineral 17:483-494 Buchanan DL, Nolan J, Wilkinson N, de Villiers JR (1983) An experimental investigation of sulphur solubility as a function of temperature in synthetic silicate melts. Geol Soc South Africa Special Pub 7:383-391 Burnham CW (1979). The importance of volatiles constituents. In: The Evolution of the Igneous Rocks: Fiftieth Anniversary Perspectives. Yoder HS (ed) Princeton University Press, Princeton, p 439-482 Carmichael ISE (1991) The redox states of basic and silicic magmas: a reflection of their source regions? Contrib Mineral Petrol 106: 29-141 Carroll MR, Rutherford MJ (1985) Sulfide and sulfate saturation in hydrous silicate melts. J Geophys Res 90:601-612 Carroll MR, Rutherford MJ (1987) The stability of igneous anhydrite: Experimental results and implications for sulfur behavior in the 1982 El Chichon trachyandesite and other evolved magmas. J Petrol 28:781-801 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation state: results from measured wavelength shifts of sulfur X-rays. Am Mineral 73:845-849 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulfur in hydrous rhyolitic melts. J Petrol 45:21712196
210
Baker & Moretti
Close WP, Tillman JF (1969) Chemical analysis of some elements in oxidation-reduction equilibria in silicate glasses. Glass Technol 10:134-136 Costa F, Scaillet B, Pichavant M (2004) Petrological and experimental constraints on the pre-eruption conditions of Holocene dacite from Valcán San Pedro (36°S, Chilean Andes) and the importance of sulphur in silicic subduction-related magmas. J Petrol 45:855-881 Courtillot VE, Rennes PR (2003) On the ages of flood basalt events. C R Geosci 335:113-140 Crutzen PJ (2006) Albedo enhancement by stratospheric sulfur injections: A contribution to resolve a policy dilemma? Climate Change 77:211-220, doi:10.1007/s10584-006-9101-y Danckwerth PA, Hess PC, Rutherford MJ (1979) The solubility of sulfur in high-TiO2 mare basalts. Proc. 10th Lunar Planet. Sci. Conf:517-530 Darken LS, Larsen BM (1942) Distribution of manganese and of sulphur between slag and metal in the openhearth furnace. Trans Am Inst Mining Metall Eng 150:87-109 Duffy JA (1996) Optical basicity: A practical acid–base theory for oxides and oxyanions. J Chem Education 73:1138-1142 Erwin DH (2006) Extinction: How Life on Earth Nearly Ended 250 Million Years Ago. Princeton University Press, Princeton Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Fedele L, Orsi G, Giaccio B, Isaia R, Scaillet B (2003) The Campanian Ignimbrite eruption, Heinrich Event 4, and Palaeolithic change in Europe: A high-resolution investigation. In: Volcanism and the Earth’s Atmosphere. Robock A, Oppenheimer C (eds) AGU Monograph 139, p 301-325 Fincham CJ, Richardson FD (1954) The behavior of sulfur in silicate and aluminate melts. Proc R Soc Lond 223A:40-61 Flood H, Forland HT (1947) The acidic and basic properties of oxides. Acta Chem Scand 1:592-604 Flood H, Grjotheim T (1952) Thermodynamic calculation of slag equilibria. J Iron Steel Inst 171:64-80 Fraser DG (1975) An investigation of some long-chain oxi-acid systems. Ph.D. thesis. University of Oxford. Fraser DG (1977) Thermodynamic properties of silicate melts. In: Thermodynamics in Geology. Fraser DG (ed) Reidel, Berlin, p 300-325 Froese E, Gunter AE (1976) A note on the pyrrhotite-sulfur vapor equilibrium. Econ Geol 71:1589-1594 Frost BR (1991) Introduction to oxygen fugacity and its petrologic importance. Rev Mineral 25:1-9 Ghiorso MS, Carmichael ISE, Rivers ML, Sack RO (1983) The Gibbs free energy of mixing of natural silicate liquids; an expanded regular solution approximation for the calculation of magmatic intensive variables. Contrib Mineral Petrol 84:107-145 Graf HF, Langmann B, Feichter J (1998) The contribution of Earth degassing to atmosphere sulfur budget. Chem Geol 147:131-145 Halmer HH, Schmincke H-U, Graf H-F (2002) The annual volcanic gas input into the atmosphere, in particular into the stratosphere: a global data set for the past 100 years. J Volcan Geotherm Res 115:511-528 Haughton DR, Roeder PL, Skinner BJ (1974) Solubility of sulfur in mafic magmas. Econ Geol 69:451-467 Helz GR, Wyllie PJ (1979) Liquidus relationships in the system CaCO3-Ca(OH)2-CaS and the solubility of sulfur in carbonatite magmas. Geochim Cosmochim Acta 43:259-265 Helz RT (1977) Determination of the P-T dependence of the first appearance of FeS-rich liquid in natural basalts to 20 kb. EOS Trans, Am Geophys Union 58:523 Holzheid A, Grove TL (2002) Sulfur saturation limits in silicate melts and their implications for core formation scenarios for terrestrial planets. Am Mineral 87:227-237 Huang W-L, Williams RJ (1980) Melting relations of portions of the system Fe-S-Si-O to 32 kb with implications to the nature of the mantle-core boundary. Lunar Planet Sci XI: 486-488 Jacobsson S, Oskarsson N (1991) The system C-O in equilibrium with graphite at high pressure and temperature, An experimental study. Geochim Cosmochim Acta 58:9-17 Jugo P J, Luth RW, Richards JP (2005b) Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim Cosmochim Acta 69:497-503 Jugo PJ (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37:415-418, doi: 10.1130/ G25527A. Jugo PJ, Luth RW, Richards JP (2005a) An experimental study of the sulfur content in basaltic melts saturated with immiscible sulfide or sulfate liquids at 1300 °C and 1.0 GPa. J Petrol 46:783-798 Kalyanram MR, Macfarlane TG, Bell HB (1960) The activity of calcium oxide in slags in the systems CaOMgO-Al2O3-SiO2 at 1500 °C. J Iron Steel Inst 195:58-64 Katsura T, Nagashima S (1974) Solubility of sulfur in some magmas at 1 atmosphere. Geochim Cosmochim Acta 38:517-531 Koek M, Kreuzer OP, Maier WD, Porwal AK, Thompson M, Guj P (2010) A review of the PGM industry, deposit models and exploration practices: Implications for Australia’s PGM potential. Resources Policy 35:20-35
Modeling the Solubility of Sulfur in Magmas
211
Konijnendijk WL, Buster JHJM (1977) Raman-scattering measurements of silicate glasses containing sulphate. J Non-Cryst Solids 23:401-418 Kress VC (1997) Magma mixing as a source for Pinatubo sulfur. Nature 389:591-593 Kress VC, Carmichael ISE (1991) The compressibility of silicate liquids containing Fe2O3 and the effect of composition, temperature, oxygen fugacity and pressure on their redox states. Contrib Mineral Petrol 108:82-92 Lafage B, Taxil P (1993) Titration of molten soda lime silicate-glasses by square-wave voltammetry. J Electrochem Soc 140:3089-3093 Lehmann J, Nadif M (2011) Interactions between metal and slag melts: steel desulfurization. Rev Mineral Geochem 73:493-511 Li C, Ripley EM (2005) Empirical equations to predict the sulfur content of mafic magmas at sulfide saturation and applications to magmatic sulfide deposits. Mineral Deposit 40:218-230 Li C, Ripley EM (2009) Sulfur contents at sulfide-liquid or anhydrite saturation in silicate melts: Empirical equations and example applications. Econ Geol 104:405-412 Liu Y (2005) Sulfur Concentration at Sulfide Saturation in Anhydrous Silicate Melts at Crustal Conditions. M.Sc. thesis, McGill Univ. Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Luhr JF (1990) Experimental phase relations of water- and sulfur-saturated arc magmas and the 1982 eruptions of El Chichón volcano. J Petrol 31:1071-1114 MacLean WH (1969) Liquidus phase relations in the FeS-FeO-Fe2O3-SiO2 system and their application in geology. Econ Geol 64:865-884 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Markus RS, Baker DR (1989) Sulfur solubility in anhydrous andesitic melts. EOS Trans, Am Geophys Union 70:1402 Matthews SJ, Moncrieff DHS, Carroll MR (1999) Empirical calibration of the sulphur valence oxygen barometer from natural and experimental glasses: Methods and applications. Mineral Mag 63:421-431 Mavrogenes JA, O’Neill HS (1999) The relative effects of pressure, temperature and oxygen fugacity on the solubility of sulfide in mafic magmas. Geochim Cosmochim Acta 63:1173-1180 Métrich N, Clocchiatti R (1996). Sulfur abundance and its speciation in oxidized alkaline melt. Geochim Cosmochim Acta 60:4151-4160 Moretti R (2005) Polymerisation, basicity, oxidation state and their role in ionic modelling of silicate melts, Ann Geophys 48:583-608 Moretti R, Baker DR (2008) Modeling of the interplay of fO2 and fS2 along the FeS-Silicate Melt equilibrium. Chem Geol 256:286-298. doi: 10.1016/j.chemgeo.2008.06.055 Moretti R, Ottonello G (2003a) Polymerization and disproportionation of iron and sulfur in silicate melts: Insights from an optical basicity-based approach. J Non-Cryst Sol 323:111-119 Moretti R, Ottonello G (2003b). A polymeric approach to the sulfide capacity of silicate slags and melts. Metall Mater Trans B 34B:399-410 Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts, the conjugated Toop-SamisGrjotheim (CTSFG) model. Geochim Cosmochim Acta 69:801-823 Moretti R, Papale P, Ottonello G (2003) A model for the saturation of C-H-O-S fluids in silicate melts. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc London Special Pub 213, London, p 81-101 Moune S, Holtz F, Botcharnikov RE (2009) Sulphur solubility in andesitic to basaltic melts: implications for Hekla volcano. Contrib Mineral Petrol 157:691-707 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Mysen BO, Popp RK (1980) Solubility of sulfur in CaMgSiO6 and NaAlSi3O8 melts at high pressure and temperature with controlled fO2 and fS2. Am J Sci 280:78-92 Mysen BO, Richet P (2005) Developments in Geochemistry 10: Silicate Glasses and Melts (Properties and Structure). Elsevier, New York Naldrett AJ (1969) A portion of the system Fe-S-O between 900° and 1,080 °C and its application to sulfide ore magmas. J Petrol 10:171-201 Naldrett AJ (1989) Magmatic sulfide deposits. Clarendon Press, Oxford University Press, Oxford Nilsson K, Peach CL (1993) Sulfur speciation, oxidation state and sulfur concentration in backarc magmas. Geochim Cosmochim Acta 57:3807-3813 O’Neill HS, Mavrogenes JA (2002) The sulfide capacity and the sulfur content at sulfide saturation of silicate melts at 1400 °C and 1 bar. J Petrol 43:1049-1087 Ottonello G (2001) Thermodynamic constraints arising from the polymeric approach to silicate slags: the system CaO-FeO-SiO2 as an example. J Non-Cryst Solids 282:72-85
212
Baker & Moretti
Ottonello G (2005) Chemical interactions and configurational disorder in silicate melts. Ann Geophys 48:561581 Ottonello G, Moretti R (2004) Lux-Flood basicity of binary silicate melts. J Phys Chem Solids 65:1609-1614 Ottonello G, Moretti R, Marini L, Vetuschi Zuccolini M (2001). On the oxidation state of iron in silicate melts and glasses: A thermochemical model. Chem Geol 174:157-179 Papale P, Moretti R, Barbato R (2006) The compositional dependence of the saturation surface of H2O + CO2 fluids in silicate melts. Chem Geol 229:78-95 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285314 Pawlowsky-Glahn V, Egozcue JJ (2006) Compositional data and their analysis: an introduction. In: Compositional Data Analysis in the Geosciences: From Theory to Practice. Buccianti A, Mateu-Figueras G, Pawlowsky-Glahn V (eds) Geol Soc London Spec Pub, 264, London, p 203-206 Pelton AD, Eriksson G, Romero-Serrano A (1993) Calculation of sulfide capacities of multicomponent slags. Metall Trans B 24B:817-825 Poulson SR, Ohmoto H (1990) An evaluation of the solubility of sulfide sulfur in silicate melts from experimental data and natural samples. Chem Geol 85:57-75 Pyare R, Nath P (1986) A simple and rapid spectrophotometric method for determination of sulfite and sulfate in binary sodium-silicate glasses. Glass Tech 1:21-23 Reddy RG, Blander M (1987) Modeling of sulfide capacities of silicate melts. Metall Trans B 18B:591-596 Richardson FD, Fincham CJB (1956) Sulphur in silicate and aluminate slags. J Iron Steel Inst 178:4-15 Rosenqvist T (1951) A thermodynamic study of the reaction CaS + H2O = CaO + H2S and the desulphurization of liquid metals with lime. Trans Am Inst Mining Metall Eng 191:535-540 Ryerson FJ, Watson EB (1987) Rutile saturation in magmas, implications for Ti-Nb-Ta depletion in island-arc basalts. Earth Planet Sci Lett 86:225-239 Scaillet B, MacDonald R (2006) Experimental and thermodynamic constraints on the sulphur yield of peralkaline and peraluminous silicic flood eruptions. J Petrol 47:1413-1437. Scaillet B, Pichavant M (2003) Experimental constraints on volatile abundances in arc magmas and their implications for degassing processes. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc London Special Pub 213, London, p 23-52 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671-698 Self S (2006) The effects and consequences of very large explosive volcanic eruptions. Phil Trans Roy Soc A 364:2073-2097 doi:10.1098/rsta.2006.1814 Shima J, Naldrett AJ (1975) Solubility of sulfur in an ultramafic melt and the relevance of the system Fe-S-O. Econ Geol 70:960-967 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Smith SJ, Andres R, Conception E, Lurz J (2004) Sulfur Dioxide Emissions 1850-2000: Methods and Results. Pacific Northwest National Laboratory Research Report 14537 Stevenson DS, Johnson CE, Collins WJ, Derwent RG (2003) The tropospheric sulphur cycle and the role of volcanic SO2. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc London Special Pub 213, London, p 295-305 Stoiber R, Williams SN, Huebert B (1987) Annual contribution of sulfur dioxide to the atmosphere by volcanoes. J Volc Geotherm Res 33:1-8 Symonds RB, Rose WI, Bluth GS, Gerlach TM (1994) Volcanic gas studies, methods, results and applications. Rev Mineralogy 30:1-66 Temkin M (1945) Mixtures of fused salts as ionic solutions. Acta Phys Chim USSR 20:411-420 Thakurta J, Ripley EM, Li C (2008) Geochemical constraints on the origin of sulfide mineralization in the Duke Island Complex, Southeastern Alaska. Geochem Geophys Geosys 9:Q07003 Tilquin JY, Duveiller P, Gilbert J, Claes P (1997) Electrochemical behaviour of sulfate in sodium silicates at 1000 degrees C. Electrochim Acta 42:2339-2346 Toop GW, Samis CS (1962) Activities of ions in silicate melts. Trans Metall Soc AIME 224:878-887 Wallace P, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874 Wallace PJ (2005) Volatiles in subduction zone magmas, concentrations and fluxes based on melt inclusion and volcanic gas data. J Volcanol Geotherm Res 140:217-240 Wallace PJ, Carmichael ISE (1994) S speciation in submarine basaltic glasses as determined by measurements SKa X-ray wavelength shifts. Am Mineral 79:161-167 Ward PL (2009) Sulfur dioxide initiates global change in four ways. Thin Solid Films 517:3188-3203 Webster J D, Sintoni M F, De Vivo B (2009). The partitioning behavior of Cl, S, and H2O in aqueous vapor– ±saline–liquid saturated phonolitic and trachytic melts at 200 MPa. Chem Geol 263:19-36. Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283
Modeling the Solubility of Sulfur in Magmas
213
Wendlandt RF (1982) Sulfide saturation of basalt and andesite melts at high pressures and temperatures. Am Mineral 67:877-885 Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov R, Kohn SC, Janousch M (2008) The origin of S4+ detected in silicate glasses by XANES. Am Mineral 93:235-240 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Winther KT, Watson EB, Korenowski GM (1998) Magmatic sulfur compounds and sulfur diffusion in albite melt at 1 GPa and 1300-1500 °C. Am Mineral 83:1141-1151 Young RW, Duffy JA, Hassall GJ, Zu Z (1992) Use of optical basicity concept for determining phosphorus and sulphur slag-metal partitions. Ironmaking Steelmaking 19:201-219
8
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 215-246, 2011 Copyright © Mineralogical Society of America
The Sulfur Budget in Magmas: Evidence from Melt Inclusions, Submarine Glasses, and Volcanic Gas Emissions Paul J. Wallace Department of Geological Sciences, University of Oregon Eugene, Oregon 97403-1272, U.S.A.
[email protected]
Marie Edmonds Department of Earth Sciences University of Cambridge Cambridge, CB2 3EQ, United Kingdom
[email protected]
INTRODUCTION The major magmatic volatile components—H2O, CO2, S, Cl, and F— play an important role in the formation, evolution, and eruption of magma. Knowledge of magmatic concentrations and fluxes of these volatiles is thus important for understanding explosive eruptive behavior of volcanoes, recycling of volatiles in subduction zones, formation of magmatic-hydrothermal ore deposits, fluxes of volcanic gases to Earth’s atmosphere, and potential climatic impacts of large volcanic eruptions. Over the past 30 years, new analytical techniques for measuring volatiles in melt inclusions and glasses from volcanic rocks and new developments in remote sensing technology used for quantifying volcanic emissions have led to major advances in our understanding of volatiles in magmatic systems and their fluxes from Earth’s mantle to the crust and hydrosphere. Sulfur plays a particularly important role in many of the processes noted above because it affects partitioning of metals into sulfide phases or vapor in magmas during crustal storage, and when released to the atmosphere, it forms sulfuric acid aerosol droplets that catalyze ozone destruction, influences other aspects of atmospheric chemistry, and blocks incoming solar radiation. In addition, S may play a role in causing oxidation of the mantle wedge above subduction zones (Kelley and Cottrell 2009). In silicate melts, the solubility behavior, activitycomposition relations, and vapor-melt partitioning of S are complex due to multiple valence states and species (S2−, S6+ in melt; H2S, S2, SO2, SO3 in vapor) and the occurrence of nonvolatile S-rich phases (immiscible Fe-S-O liquid, pyrrhotite, monosulfide and intermediate solid solutions, anhydrite). Sulfur dioxide (SO2) is the easiest of the main magmatic volatiles to measure in volcanic plumes using ground- and satellite-based remote sensing techniques because of its relatively high concentration in volcanic plumes relative to background values. More recently, techniques for measuring H2S in volcanic plumes have been developed (e.g., McGee et al. 2001). There is now a wealth of data on SO2 fluxes from active volcanoes based on remote sensing techniques (see Oppenheimer et al. 2011, this volume). These data have led to a long-standing puzzle, 1529-6466/11/0073-0008$05.00
DOI: 10.2138/rmg.2011.73.8
216
Wallace & Edmonds
known as the excess sulfur problem, concerning S in magmatic systems: concentrations of dissolved S in magmas before eruption, as estimated from melt inclusion data, are typically far too low to account for the total mass of SO2 released during the eruption as measured by remote sensing techniques (Stoiber and Jepson 1973; Rose et al. 1982; Andres et al. 1991; Wallace 2001; Shinohara 2008). The likely cause of this excess S problem is the presence of an exsolved C-O-H-S vapor phase in the magma before the time of eruption (Anderson 1975; Luhr et al. 1984; Andres et al. 1991; Lowenstern 1993; Gerlach et al. 1994, 1996; Gerlach and McGee 1994; Wallace and Gerlach 1994; Giggenbach 1997; Wallace 2001; Scaillet and Pichavant 2003; Scaillet et al. 2003; Shinohara 2008). As a result, eruptions of silicic magma can release large amounts of SO2 derived from pre-eruptive exsolved vapor, despite the fact that such magmas generally have low concentrations of dissolved S (see also Baker and Moretti 2011, this volume). In this chapter, we review the concentration, solubility controls, and degassing behavior of S in magmas from different tectonic environments, with an emphasis on subduction-zone magmatism. We focus on the following major questions: 1.
What are the concentrations of S in magmas across the spectrum from basalt to rhyolite, and what do the variations tell us about S losses to sulfide or sulfate immiscible melts or minerals or to a vapor phase during magma differentiation?
2.
Given the common occurrence of excess S emissions from erupting volcanoes, how do the large mass fractions of vapor form in magma and, specifically, what is the role of mafic magma recharge in the underlying magmatic system?
3.
How can we best use volcanic gas composition to better understand the partitioning behavior of sulfur between melt and vapor?
4.
How do S and magma redox state affect the partitioning of metals during partial melting in the mantle, sulfide fractionation during magma transit through the crust, and exsolution of a vapor phase, and how do these relate to the formation of magmatichydrothermal ore deposits?
5.
What do the higher S contents in arc basaltic magmas compared to magmas in other tectonic environments tell us about S recycling in subduction zones?
SULFUR CONCENTRATIONS IN MAGMAS Information on the S and other volatile contents of magmas comes primarily from two types of samples: submarine glassy pillow rims and melt inclusions trapped in phenocrysts. Matrix glasses from subaerially erupted samples can also be analyzed, but these are always strongly degassed and provide no real information on magmatic concentrations except, perhaps, for highly soluble volatiles like F. For example, subaerial matrix glasses contain <50 to ~150 ppm S (e.g., Luhr 2001; Cervantes and Wallace 2003a; Métrich et al. 2001; Witter et al. 2004, 2005; Spilliaert et al. 2006a) compared to magmatic concentrations of >800 ppm S (see below). Exceptions to this might occur when sulfur diffusion into bubbles is inhibited due to kinetic factors such as during explosive eruptions with very high magma ascent rates, for example (data on the diffusivity of S in silicate melts can be found in Baker et al. 2005, Freda et al. 2005, and Behrens and Stelling 2011, this volume). Submarine pillow basalts, by contrast, are commonly quenched at a high enough hydrostatic pressure to inhibit degassing of S from the glassy, rapidly quenched pillow rims (Moore and Fabbi 1971; Dixon et al. 1991). However, during shallow-water eruptions and even during deep-water eruption of H2O-rich magma, S can be partially to strongly degassed (e.g., Davis et al. 1991). For H2O-poor MORB and Hawaiian tholeiite, degassing of S only occurs at water depths ≤ ~1000 m.
The Sulfur Budget in Magmas
217
Melt inclusions represent small volumes of silicate melt trapped at pressure inside of growing phenocrysts. In principle, providing they remain sealed during eruption, the host crystal acts like a tiny pressure vessel to keep the inclusion at relatively high pressure so that exsolution of volatiles does not occur. However, melt inclusion data can be complicated in some cases by either post-entrapment effects or the fact that inclusions may be trapped at middle- to upper-crustal pressures after some degassing of melt has already occurred (Lowenstern 1995; Métrich and Wallace 2008). The main post-entrapment effect of concern for S is that diffusive exchange between melt and olivine often causes a decrease in the FeOT content of the melt inclusion (Danyushevsky et al. 2000), and this can potentially cause exsolution of a sulfide globule in the inclusion because of decreased solubility of S in the silicate melt (Danyushevsky et al. 2002). Post-entrapment changes in oxygen fugacity may also occur (Gaetani et al. 2010), and this could potentially change S speciation in a melt inclusion or cause a sulfide globule to either exsolve or dissolve. Recognizing the effects of partial degassing before inclusion trapping will be discussed below. Complementing the analytical data on pillow rim glasses and melt inclusions are numerous experimental studies on the solubility and speciation of S under a wide range of conditions, and these experiments are valuable for understanding variations of S in natural samples. Analytical techniques for measuring S concentration and speciation in melt inclusions and glassy pillow rinds are described in Métrich and Wallace (2008), Ripley et al. (2011, this volume) and Wilke et al. (2011, this volume). MORB pillow rim glasses and melt inclusions contain ~800-1300 ppm S at 8-10 wt% FeOT, and the S contents increase with increasing FeOT up to ~2000-2400 ppm S at 15-16 wt% FeOT (Fig. 1). This correlation is caused by the effect of melt Fe in increasing the solubility of S in silicate melts (Mathez 1976; Wallace and Carmichael 1992; O’Neill and Mavrogenes 2002; Liu et al. 2007; Baker and Moretti 2011, this volume; Ebel 2011, this volume). At the relatively low oxygen fugacities of MORB magmas (
Figure 1. S contents vs. MgO and total FeO concentrations of submarine MORB glasses from the Galapagos spreading center (GSC) (Perfit et al. 1983) and the East Pacific Rise (EPR) (Le Roux et al. 2006). Line in B shows the S concentration limit in FeS-O-liquid-saturated MORB melts calculated using the S solubility model of Wallace and Carmichael (1992) updated with the temperature dependence from Mavrogenes and O’Neill (1999). The S concentration for Fepoor Galapagos basalt is calculated at 1170 °C and an fO2 that is −1.5 log units more reduced than the NNO buffer. For Fe-rich Galapagos basalt, the S concentration is calculated at 1070 °C and fO2 of −0.5 log units more reduced than NNO. These values are based on data in Wallace and Carmichael (1992).
D
Sulfide saturation
Sulfide saturation
B
D
D
D
B
C
C
A
A
D
D
C
C
218 Wallace & Edmonds
The Sulfur Budget in Magmas
219
of magnetite found in quenched sulfide globules in MORB glasses are consistent with relatively low fO2 for MORB magmas (see discussion below). Olivine-hosted melt inclusions from basaltic arc magmas typically have 900-2500 ppm S, but values extend as high as 4000-7000 ppm (Fig. 2). The highest values are found in phlogopite-bearing arc shoshonites and leucitites (Vigouroux et al. 2008). The S contents of arc magmas are commonly much higher than for MORB magmas of the same FeOT though there is considerable overlap in the datasets. Sulfur solubility in silicate melts increases dramatically with increasing fO2 as the speciation of S dissolved in the melt changes from predominantly S2− to S6+ (Carroll and Rutherford 1987, 1988; Luhr 1990; Wallace and Carmichael 1994; Métrich and Clocchiatti 1996; Jugo et al. 2005b; Jugo 2009). It has been shown that the presence of S4+ species in natural oxidized melt inclusions and low-pressure silicate glasses is an analytical artifact caused by electron microprobe beam damage (Métrich et al. 2003) or by irradiation with the synchrotron X-ray beam during analysis (Wilke et al. 2008; see also Wilke et al. 2011, this volume). In more oxidized melts, the maximum dissolved S is controlled by saturation with either anhydrite (Carroll and Rutherford 1987; Luhr 1990) or immiscible molten sulfate (Jugo et al. 2005a). The maximum dissolved S in basaltic melt under oxidized conditions is ~1.5 wt% S at 1300 °C and 1 GPa pressure. The mafic arc melts shown in Figure 2 are more likely to be at maximum temperatures of 1150-1200 °C given their compositions and H2O contents. Given the strong temperature dependence of sulfate solubility, likely maximum S contents at these temperatures are ~5000-7000 ppm (using a temperature dependence based on fits of Luhr 1990, experimental data to the expression lnS = a/T(°C) + b, where a and b are constants and lnS expresses the S concentration in the melt). Experiments on a hydrous, mafic magma composition at 1150 °C and 413 MPa under oxidizing conditions yielded anhydrite+sulfide liquid-saturated melts with 6500 ppm S and 1.7 wt% H2O (Pichavant et al. 2006). At intermediate fO2, both sulfide and sulfate may coexist (~NNO+1.5), and the S content of sulfide-saturated melts likely increases exponentially over the range from FMQ to NNO+1.5 (Fig. 3; see also Wilke et al. 2011, this volume). The high S contents of melt inclusions from many arc basaltic magmas require fO2 greater than FMQ, which is higher than values estimated for MORB magmas. This conclusion is supported by Fe3+/Fe2+ data for volcanic rocks (Carmichael 1991; Wallace and Carmichael 1999), measurements of Fe3+/Fe2+ (Kelley and Cottrell 2009) and S Ka wavelengths in basaltic melt inclusions (Métrich and Clocchiatti 1996; Luhr 2001; Vigouroux et al. 2008), and analysis of mantle xenoliths (Brandon and Draper 1996; Parkinson and Arculus 1999; McInnes et al. 2001). We note, however, that the relative fO2 of MORB vs. arc magmas has been somewhat controversial, in part because recent data have been used to suggest a higher fO2 for MORB, averaging around FMQ-0.4 (Bezos and Humler 2005), and in part because V/Sc systematics in the different types of basalts have been used to argue that both MORB and arc basalts have fO2 in the range of FMQ±1 (Lee et al. 2005; Mallman and O’Neill 2009). With increasing SiO2 content, the S contents of arc magmas decrease by an order of magnitude or more (Fig. 2). This decrease is probably caused by compositional and temperature effects on S solubility. For both sulfide- and sulfate-saturated melts, the solubility of S strongly decreases with decreasing temperature. At temperatures of ~800 °C, maximum solubilities are around 50-200 ppm (Luhr 1990). Thus most or all silicic arc magmas are likely sulfide ± sulfate saturated (e.g., Larocque et al. 2000; Luhr 2008; Parat et al. 2011, this volume) whereas many mafic arc magmas may not be saturated with either phase but instead have less S than the maximum solubility values. For example, in the central Oregon Cascades, mafic arc magmas with fO2 around NNO contain sulfide inclusions in olivine phenocrysts, but magmas with higher S contents and higher inferred fO2 do not contain either sulfide or a sulfate phase (Ruscitto et al. 2010). However, there is evidence for primary sulfide globules in at least some more oxidized basaltic magmas, such as basaltic enclaves from the 1991 eruption of Mount Pinatubo (NNO+1;
Figure 2. Variations in S content, S/K2O ratio, and K2O contents vs. SiO2 and total FeO concentrations for melt inclusions from convergent margin volcanic rocks. All data are from the GeoRoc database. For each sample suite, we have removed analyses that were the most obviously affected by degassing before inclusion entrapment based on low S/K2O compared to compositionally similar inclusions. Curves show the solubility of S in experimental anhydrite-saturated melts (starting materials were basalt (B) and trachyandesite (TA)) at the MNO buffer (solid curves) and sulfide-saturated melts at the FMQ buffer and 800-1000 °C and 100-400 MPa (Luhr 1990). Curves in D show the compositional variations in the experimental melts (Luhr 1990).
B B
A A
B B
A A
D
C
D
D
D
C
C
220 Wallace & Edmonds
The Sulfur Budget in Magmas
221
Arc
MORB
OIB BAB
Figure 3. Sulfur solubility vs. log fO2 for basaltic melts based on experimental data compared to log fO2 and the range of measured S contents for basalts from different tectonic environments. The black curve shows the exponential increase in S concentration for sulfide-saturated melts with increasing fO2. At the peak S content, melts are saturated with both sulfide+anhydrite (or sulfate melt). At higher fO2, the melts are saturated with sulfate only. Oxygen fugacity ranges for different types of basaltic magma are shown: Arc=subduction zone basalts; BAB=back arc basin basalts; OIB=oceanic island basalts; MORB=mid-ocean ridge basalts. For each basalt type, the top and bottom of the box indicate the range of measured S contents in basaltic glasses and melt inclusions based on data in Figures 1 and 2 and Wallace and Carmichael (1992, 1994). The range of S values shown for MORB are for glasses with 8-12 wt% FeOT because glasses with higher FeOT have substantially higher S solubility than is depicted by the experimental sulfide-saturation curve (see Fig. 1b). Figure is modified from Jugo (2009).
Figure 3 arc magmas DiMuro et al. 2008). An interesting question then is to what extent intermediate have their S contents limited by sulfide or sulfate saturation or whether the decreases in melt S with increasing SiO2 could be caused by partial degassing of S before inclusion entrapment at middle to upper crustal pressures. Many intermediate magmas show evidence of having been saturated with immiscible Fe-S-O liquid, but this sulfide phase is commonly converted to Feoxide globules, probably as a result of pre-eruptive shallow degassing of S (Larocque et al. 2000). Arc-like magmas ranging from basalt to rhyolite in the Manus back-arc basin show a trend of increasing Cu and Au with increasing SiO2 up to ~60 wt% SiO2, after which the concentrations decrease (Moss and Scott 2001; Sun et al. 2004). Jenner et al. (2010) have shown that the decreases in chalcophile element concentrations are caused by the onset of sulfide saturation rather than loss to a vapor or fluid phase. The sulfide saturation may occur due to the the onset of magnetite crystallization, which in turn causes a decrease in fO2. This evidence suggests that the decreases in S content seen in arc magmas as SiO2 increases may be primarily caused by sulfide saturation. MAGMATIC DEGASSING OF SULFUR Degassing and vapor-melt partitioning Arc magmas commonly contain enough dissolved H2O and CO2 to be vapor saturated in the middle to upper crust (Scaillet and Pichavant 2003; Wallace 2005; Métrich and Wallace
Wallace & Edmonds
222
2008). In addition, because H2O content has a strong effect on liquidus temperature, loss of H2O by degassing provides a strong drive for crystallization (e.g., Eichelberger 1995; Johnson et al. 2008). As a result, most of the observed phenocrysts in basaltic arc magmas crystallize during ascent, degassing, and shallow storage in the crust. Melt inclusions in crystals preserve a record of these processes, but the inclusions represent melts that had variably degassed prior to entrapment. This variable degassing hinders our ability to determine the primary volatile contents of basaltic melts. The extent to which this variable degassing hides the primary volatile contents is inversely proportional to the solubility of the component in question. Thus for low solubility components like CO2, melt inclusions formed in the middle to upper crust may never record values that are close to primary values (Wallace 2005). The major volatile components that degas the least are F and Cl, and they are least affected by this problem. H2O and S lie between the two extremes and show variable degrees of degassing in melt inclusion data sets.
CO2
0
200
e (MPa)
H2O
1.0
S
0.8 0.6
0.4 0.2 250 0.0
S
300
e (MPa)
Fraction of initial amount remaining (C / Cinitial)
olitic magma
CO2
1.0
0.8 0.6
0.4 0.2 4000.0
Low-H2O basaltic magma (Kilauea) H2O Low-H 2O basaltic magma (Kilauea) 1.0
S
A
0
0.6
Summit Reservoir
Summit Reservoir
0.4 0.2
CO2
0.0 300 50
1.0
50
CO2
100 150 200 150 Pressure 200 (MPa) 250
100
Pressure (MPa)
250
300
300
H2O-rich rhyolitic magma H2O-rich rhyolitic magma
B
B
0.8 0.6
A
A
Cl
Cl 0.8
Fraction of initial amount remaining (C / Cinitial)
Fraction of initial amount remaining (C / Cinitial)
magma (Kilauea)
Fraction of initial amount remaining (C / Cinitial)
Insight into the relative effects of partial degassing before inclusion entrapment on different volatiles can be gained using the simplified vapor-melt partitioning and solubility expressions described by Giggenbach (1997). Degassing of a real multi-component vapor phase will certainly be more complex (see Webster and Botcharnikov 2011, this volume). Both examples shown in Figure 4 assume a case of closed-system ascent and degassing. For high-temperature, low-fO2, and H2O-poor basaltic magma like that at Kilauea (Dixon et al. 1991; Wallace and Anderson 1998), olivines crystallizing in the summit reservoir (~2-6 km depth) trap melts that
H2O
0.4
Cl
H2O Cl
Crystal Growth
Crystal Growth S
CO2
0.2
0.0 0 500 100
100
200
FigurePressure 4
S
200
300
300 400 Pressure (MPa)
(MPa)
Figure 4. Fraction of initial dissolved volatile contents remaining in melt as a function of pressure for (A) lowH2O Kilauea basaltic magma (initial volatile concentrations are 0.5 wt% H2O, 7000 ppm CO2, 1000 ppm S, 100 ppm Cl) and (B) H2O-rich rhyolitic magma (initial volatile concentrations are 5.5 wt% H2O, 1000 ppm CO2, 1000 ppm S, 100 ppm Cl). See text for details. Curves were calculated using B equations and solubility parameters from Giggenbach (1987) and DSvapormelt values for rhyolitic melt from Wallace (2003). Note difference in scale on x-axes.
CO2
400
500
500
Figure 4
Figure 4
The Sulfur Budget in Magmas
223
have already lost 70-90% of the initial CO2 content of ~7000 ppm (Fig. 4a). In contrast, the melts have lost, at most, a trivial amount of the more soluble H2O, Cl, and S, making melt inclusions reliable indicators of the original magmatic concentrations of these volatiles. The main difference between this case and the case of low-temperature, H2O-rich rhyolitic magma (Fig. 4b), is that the lower temperature of the rhyolite causes S solubility to be quite low, comparable to CO2. As a result, at the ~200-MPa pressure that is common for phenocryst crystallization in such magmas, most of the initial S has degassed from the melt and is contained in a coexisting vapor phase. The ramifications of this behavior for understanding SO2 emissions from volcanoes are profound and will be discussed later in this chapter. Because of the complex behavior of S in melts and gases, accurate models for predicting vapor-melt partitioning of S must include the effects of temperature, pressure, melt composition, and the fugacities of O2, S2, and H2O. This has been done recently using both a simple empirical relationship (Scaillet and Pichavant 2005) and a more rigorous thermodynamic approach (Moretti and Baker 2008). These models make it possible to calculate the mole fractions of H2S and SO2 (the two dominant species) in the vapor phase in equilibrium with melt if the fO2, temperature, pressure and H2O content are known. It should be noted that these models are calibrated using experimental data, but there is little data for hydrous mafic melts in which the total S content of the quenched vapor phase is measured or the S-species fugacities can be constrained. Such data do exist for intermediate to silicic compositions (Luhr 1990; Keppler 1996, 2010; Clemente et al. 2004; see also Webster and Botcharnikov 2011, this volume), so the vapor-melt partitioning models are much better calibrated in this range. As an example of the potential effects of other volatiles on S solubility, Liu et al. (2007) report experimental data on an Etna basalt at 0.5-1 GPa in which increasing H2O from ~0 to 4 wt% (all experiments were vapor-undersaturated) caused a decrease in melt S content by about 50%. Additional experimental data for basaltic melts are presented by Webster and Botcharnikov (2011, this volume). Calculated DSvapor-melt values for S using the empirical model of Scaillet and Pichavant (2005) are shown in Figure 5 for a primitive calc-alkaline basaltic andesite (Jorullo volcano) from central Mexico (Johnson et al. 2010). DSvapor-melt values greater than one indicate preferential partitioning of S into the vapor phase, so the results show that S partitions slightly to strongly into the vapor phase over the range of conditions shown. Higher pressure increases the solubility of S in hydrous melts (Luhr 1990) and thus decreases the DSvapor-melt value. Higher fO2 also increases the solubility, but this effect is more than offset by the higher fugacities of SO2 in the vapor phase with increasing fO2, such that DSvapor-melt increases with fO2. Sulfur concentration also affects DSvapor-melt because it causes a non-linear increase in SO2 fugacity in the vapor phase. This is shown by the calculated curve for the Irazù 1723 eruption, which has an initial S content of ~4000 ppm S (Benjamin et al. 2007) compared to 1400 ppm used in the calculated Jorullo value. We compare the calculated pre-eruptive vapor compositions arising from these models with volcanic gas data in a subsequent section.
Degassing inferred from melt inclusions from mafic volcanoes Melt inclusion data provide evidence for progressive exsolution of S from melt into the gas phase during decompression. In some volcanoes, the S contents of melt inclusions show a clear decrease with decreasing H2O (Fig. 6), whereas in other cases (e.g., Johnson et al. 2010), data are more scattered, with loss of S occurring mainly at low pressure. At Arenal and Irazù volcanoes in Costa Rica, S degassing starts at pressures of ~150-200 MPa (Wade et al. 2006; Benjamin et al. 2007). A pressure of ~140 MPa is reported for the onset of efficient S degassing in the H2O-rich Etna basalt (Spilliaert et al. 2006a). In that case, ~80% of the initial S is lost in a pressure range between 140 and 10 MPa. Data for high-Mg basaltic andesite melt inclusions from cinder cones in central Mexico also show evidence that S degassing begins at ~100 MPa (Johnson et al. 2010). These pressures of S degassing deduced from melt inclusion data are consistent with recent experimental results (Lesne et al. in review; see Webster and
Wallace & Edmonds
224
DSvapor-melt Figure 5. Pressure dependence of S vapor-melt partitioning calculated using the empirical model of Scaillet and Pichavant (2005) and the Redlich-Kwong equation of state for fugacity coefficients for pure SO2. Calculations were made for a primitive basaltic melt from Jorullo volcano (Johnson et al. 2010) at log fO2 of NNO to NNO +1.5 at 1140 °C. Increases in fO2 (at a constant pressure) result in higher values of DSvapor-melt. The curve for Irazù is based on data from Benjamin et al. (2007) and is calculated at 1050°C for melt with higher initial S than Jorullo. The higher initial S and lower temperature for Irazù result in higher DSvapor-melt values at a given pressure. 5000
Irazu-MI-Ol Irazu-MI-Cpx
4000
A
~ 150 MPa
140 MPa
4000
Arenal-MI-Ol
B
Etna
Arenal model parent
3000
3000
ed
2000
2000
sa tur at
S (ppm) in melt
5000
Sulfide saturation
1000 0 0
1
2
3
H2 O (wt%) in melt
4
5
Su lfi d e
1000 0 0
1
2
H 2 O (wt%) in melt
3
4
Figure 6. Variation of H2O and S concentrations in melt inclusions from (A) Irazù (Benjamin et al. 2007) and Arenal (Wade et al. 2006) and (B) Etna (Spilliaert et al. 2006a). Sulfur-degassing paths were calculated based on the approach in Sisson and Layne (1993), which uses the relationship between K2O and H2O to calculate the proportion of H2O in the separating assemblage of crystals and bubbles and assumes no fractionation of K2O. The S-degassing paths were calculated using DSvapor-melt = 110 for Irazù and 70 for Arenal. Etna data include olivine-hosted melt inclusions from samples of the 2002 flank eruption (filled circles) and three lava fountains in 2000 (open squares). The pressure at which S degassing begins is constrained from the CO2 and H2O contents of the melt inclusions. For closed-system degassing, the S decrease is consistent with DSvapor-melt values ranging from 0 at higher pressures to 60 at lower pressures (Spilliaert et al. 2006a). Figure is modified from Métrich and Wallace (2008).
The Sulfur Budget in Magmas
225
Botcharnikov 2011, this volume). At all of these volcanoes, S is dissolved in melt primarily as S6+ based on measurements of SKa wavelengths of the glasses. The pressures for S degassing are far higher than the low pressure of S exsolution (≤ 10 MPa) reported for submarine basalts (e.g., Dixon et al. 1991), which contain S mainly as S2− because of their lower fO2. The evolution of S and H2O during decompression depends on magma ascent rate and degassing conditions (open vs. closed system). Changes in redox state towards more reduced conditions can drive the melt to sulfide saturation during late stage crystallization at shallow depth, and this affects the S degassing path. Faster magma ascent inhibits these processes until low pressure, causing S degassing to be controlled by vapor-melt partitioning. In cogenetic basaltic and trachybasaltic magmas at Etna, the S content of sulfide-saturated melt inclusions trapped in Fo < 75 olivine representative of the shallow, partly degassed magma are only half that of sulfide-free inclusions in Fo ~ 80 olivine that crystallized during fast ascent of basaltic magma under dominantly closed-system conditions (Spilliaert et al. 2006a). In contrast to the differences in S content, the H2O contents of the different inclusions are comparable (~1.5 wt%). Relations between S, K2O, and H2O can be used to infer and quantify S loss by degassing because S should behave like an incompatible element in the absence of any degassing or formation of a S-rich phase (Fig. 7). Variations in S/K2O in cogenetic suites of melt inclusions show evidence for S loss by degassing simultaneous with fractional crystallization of melt. The two processes are linked by cause and effect because loss of H2O by degassing causes S in the melt to partition into the vapor phase and simultaneously drives melt crystallization by increasing the liquidus temperature (Sisson and Layne 1993; Roggensack 2001; Wade et al. 2006; Johnson et al. 2008, 2010). This does not mean that the actual magma temperature increases, though in some cases it may (Blundy et al. 2006), but is caused by the fact that the wet liquidus temperature of a silicate melt is lower than the dry liquidus temperature. Assuming that degassing can be described by a Rayleigh fractionation process, Sisson and Layne (1993) Jorullo MI value of 34 for calc-alkaline basalt of Fuego used melt inclusion data to calculate a DSvapor-melt Paricutin MI for calc-alkaline basalt to basaltic volcano in Guatemala. Other calculated valuesHoya of Alvarez DSvapor-melt MI Jorullo andesite using this same method are 70 for Arenalmatrix volcano in Costa Rica (Wade et al. 2006), Jorullo MI Paricutin MI
Crystallization
0.3
0.1
0 0.5
Hoya Alvarez MI Jorullo matrix Paricutin matrix Hoya Alvarez matrix
Degassing + crystallization
Degassing
0.2
S / K2O (by wt) in melt
S / K2O (by wt) in melt
Paricutin matrix Hoya Alvarez matrix
Crystallization
0.3 1
0.2
1.5
2
2.5 Degassing
Degassing + crystallization
3
K2O (wt%) in melt
Figure 7. S/K2O vs. K2O in melt inclusions (MI) and matrix glasses (matrix) from Jorullo and Parícutin volcanoes (calc-alkaline basalt to basaltic andesite) and Hoya Alvarez maar (alkali basalt) in central Mexico. Vectors indicate 0.1 general paths for melts undergoing crystallization, degassing, and degassing + crystallization. Decreases in S/K2O with increasing K2O indicate S degassing during differentiation of the magmas at Jorullo and Parícutin. Hoya Alvarez melts appear not to have degassed S during differentiation, though decreases in S/K2O in matrix glasses indicate eruptive degassing of S. Figure is from modified from Johnson et al. (2010).
0 0.5
1
1.5
2
2.5
3
226
Wallace & Edmonds
75 for Irazù, also in Costa Rica (T. Plank, written commun., based on data in Benjamin et al. 2007), and 2-21 for Jorullo volcano (Johnson et al. 2010). Values for more K-rich arc magmas are 40 for Stromboli volcano (K-rich basalt to shoshonite; Métrich et al. 2001) and 5-20 for cinder cones in the western Trans-Mexican Volcanic Belt (shoshonite; Vigouroux et al. 2008). The Irazù value is higher than the values calculated at high pressure using Scaillet and Pichavant (2005), but we consider the agreement to be reasonably good given that the latter model does not have any experimental hydrous mafic melts in its calibration dataset. Some values of DSvapor-melt from cinder cones in Mexico are lower than the calculated values (Johnson et al. 2010). This could perhaps be caused by slow diffusion of S (Freda et al. 2005) during rapid ascent and degassing, but we note that the predicted effects of slow S diffusion have not been observed for other melt inclusion data sets (see discussion in Métrich and Wallace 2008). Comparisons of open-system and closed-system degassing have been made using methods whereby a value of DSvapor-melt can be calculated for each inclusion based on its difference from the initial concentrations of S and H2O (Spilliaert et al. 2006a; Johnson et al. 2010). Closedsystem models typically yield higher DSvapor-melt values than open-system models because Rayleigh fractionation is a more efficient mechanism of degassing; thus for a given initial and final dissolved S content, an open-system model requires a lower DSvapor-melt to explain the data. Melt inclusion data from Etna yield closed-system degassing DSvapor-melt values from 0-60 compared to open-sytem values of 0.4-25, and the values increase with decreasing pressure (Spilliaert et al. 2006a). The melt inclusion data show significant increases in DSvapor-melt at pressures < 140 MPa, suggesting extensive degassing of S occurs shallowly, and is consistent with the model predictions (Fig. 5) based on Scaillet and Pichavant (2005). In the case of Etna, the compositions of volcanic gases were shown to be incompatible with an open-system degassing model (Spilliaert et al. 2006a), and thus the comparison of melt inclusions and volcanic gas compositions can be used to discriminate between degassing processes.
Magmatic vapor phase and volcanic gases Magmatic vapor and volcanic gases contain sulfur predominantly in the form of sulfur dioxide (SO2) and hydrogen sulfide (H2S). Measurements of volcanic gas fluxes (by spectroscopy) and composition (by FTIR and electrochemical sensors) are widely used as a volcano-monitoring tool (Christopher et al. 2010; Oppenheimer 2010; Oppenheimer et al. 2011, this volume). It has been widely recognized that the flux and composition of gases varies from volcano to volcano and during a single eruption (e.g., Allard et al. 2005; Aiuppa et al. 2007; Shinohara 2008), and these data, when combined with petrological studies, have the potential to tell us about the controls on sulfur behavior in magmatic systems. Measurement of the flux of S-bearing gases is often the only way to quantify the total S output of the volcanic system. There are caveats of course; the S output might not represent the total S content of the system owing to the effects of hydrothermal scrubbing (e.g., Symonds et al. 1994). It is also necessary to be able to measure both dominant forms of S in gases (H2S and SO2) as the molar ratio between them is often close to unity, particularly for silicic systems. Thus volcanic gas data complement petrological studies of S behavior in melts; because S partitions into vapor at high pressures, information about the total S content of the system is difficult to extract from the melt inclusion record alone. The rapid increase in the amount of published volcanic gas data (owing largely to the development of new methodologies, see Oppenheimer et al. 2011, this volume) and the high temporal resolution of gas data records, allow us to extract a considerable amount of information regarding S partitioning behavior, the source of S in volcanic systems, and dynamic degassing mechanisms in the magma storage region and conduit. Vapor-saturation models allow calculation of vapor compositions in equilibrium with magmas over a range of crustal pressures (Newman and Lowenstern 2002; Moretti et al. 2003; Scaillet and Pichavant 2003; Moretti and Ottonello 2005; Papale et al. 2006; Burgisser and Scaillet 2007; Burgisser et
The Sulfur Budget in Magmas
227
al. 2008). These models can be compared to natural volcanic gas data in order to infer vapormelt partitioning, the pressure of vapor-melt separation and to assess whether closed- versus open-system degassing is operating (Moretti et al. 2003; Scaillet and Pichavant 2003; Moretti and Papale 2004; Burgisser and Scaillet 2007). Inevitably, ambiguities and uncertainties arise when there are non-unique outcomes to chemical models and also when trying to compare models to gas data, which are collected using a variety of methods that vary in their accuracy and precision. Furthermore, the chemical models do not cover the entire range of complexity of multi-component fluids, because scaling by data from well-constrained experiments (as reported in Webster and Botcharnikov 2011, this volume) are missing. Models of sulfur degassing and comparison to volcanic gas data. Vapor saturation models for silicate melts involving equilibria between a H2O-CO2-SO2-H2S gas phase and a silicate melt with dissolved volatiles predict how melt and vapor evolve during both second boiling (crystallization-induced vapor exsolution) and decompression (Moretti et al. 2003; Scaillet and Pichavant 2003; Moretti and Papale 2004; Moretti and Ottonello 2005; Burgisser and Scaillet 2007; Burgisser et al. 2008). An important first-order result of the modeling is that the simple behavior predicted by one-component solubility models is replaced by highly non-linear and complex behavior upon adding additional volatile components to the system. The behavior of S in melts is influenced dramatically by the fugacity of oxygen and sulfur, and temperature and melt composition (e.g., Carroll and Rutherford 1987; Luhr 1990; Wallace and Carmichael 1992; Scaillet et al. 1998; Clemente et al. 2004; Jugo et al. 2005a,b; Keppler 2010; see earlier section). Redox conditions during closed-system decompression degassing strongly modulate vapor SO2/H2S ratios, the most important redox equilibria being FeO-Fe2O3 in basaltic systems (Wallace and Carmichael 1992) and SO2-H2S in more silicic melt-vapor systems. Some general observations may be made by comparing computed pre-eruptive vapor compositions to volcanic gas data; for intermediate to silicic systems, volcanic gases are generally more enriched in H2O than calculated pre-eruptive vapor, and the C/S ratio is higher than predicted for simple closedsystem decompression degassing of a vapor-saturated melt. For mafic systems, the discrepancy is the other way around; there is generally less CO2 in the gases than predicted (Scaillet and Pichavant 2003; Burgisser and Scaillet 2007; Burgisser et al. 2008). These differences may be explained, in the case of intermediate to silicic systems, by the presence of pre-eruptive vapor. If the magma starts out with a significant fraction of pre-eruptive vapor, then the final vapor composition is essentially buffered by the starting composition of the vapor (Wallace et al. 1995). The effect of this is that the higher fraction of exsolved pre-eruptive vapor, the higher the C/S ratio of the volcanic gases at the surface. The exact path of vapor evolution will depend on storage conditions and the extent and depth of crystallization. C/S ratios of volcanic gases from arc volcanoes tend to fall between 1 and 30 (Fig. 8), which, if vapor and melt are in equilibrium during closed-system degassing, requires up to 1 wt% exsolved vapor in the magma reservoir prior to closed-system decompression and degassing (Scaillet and Pichavant 2003). Corroborating this interpretation is an abundance of evidence from other degassing studies for the existence of such a pre-eruptive vapor phase (Westrich and Gerlach 1992; Wallace and Gerlach 1994; Gerlach and McGee 1994; Gerlach et al. 1994, 2008; Scaillet et al. 1998; Wallace 2001, 2005). The modeling can be taken one step further by describing the equilibrium between SO2 and H2S under a variety of conditions, with or without pre-eruptive vapor (Burgisser et al. 2008). A wealth of volcanic gas data quantifying SO2/H2S exists for arc volcanoes (Fig. 8), and the results of such modeling show that the gas data compare well to models of closed-system equilibrium degassing for oxidized magmas (NNO to NNO+1) with 0.1-5 wt% vapor and a pre-eruptive C/S ratio of 0.3-3 (Fig. 8). Volcanic gases from mafic magmas, on the other hand, plot closer to the predicted vapor composition in equilibrium with melt at low pressures and do not require a large fraction of pre-eruptive vapor (Scaillet and Pichavant 2003). In fact, they commonly display lower C/S
228
Wallace & Edmonds
Figure 8. Molar C/S vs. molar SO2/H2S of natural volcanic gases (small symbols; Edmonds et al. 2010 and references therein) and volcanic gas compositions predicted from closed-system equilibrium degassing to 1 bar using thermodynamic models (calibrated for rhyolite melts; Burgisser et al. 2008). The results are contoured in two ways: 1) with respect to initial pre-degassing magma redox state (solid lines) and 2) with respect to initial melt C/S (dashed lines). The black lines are volcanic gas compositions resulting from closed-system degassing of magma starting with 5 wt% exsolved vapor; the grey lines with 0.1 wt% exsolved fluid.
contents than predicted, which indicates that melt-vapor separation must occur deeper in the crust, at pressures of a several hundred MPa and/or open-system degassing might play a larger role in these low-viscosity systems, where melt-gas segregation via bubble coalescence and rise is expected to be more efficient (e.g., Vergniolle 1996). The volcanic gas data then appear to be showing us that, for long-lived silicic magmas, fluids exsolve and accumulate in magma storage regions. The high viscosity and crystal content of these magmas would certainly slow the migration of fluid out of the magma body compared to mafic systems, where efficient meltvapor segregation results in simpler geochemical degassing trends. Effect of degassing on magma redox state. Further complicating a comparison of the models and gas data shown in Figure 8, which shows the predicted vapor composition on eruption, is the likelihood that magma and co-existing vapor change their redox state during decompression and degassing. A common observation in volcanic systems is that the erupted melt and the emitted gases do not display compatible oxidation states (Carmichael and Ghiorso 1986; Gerlach 2004; Burgisser and Scaillet 2007; Edmonds et al. 2010). Models have been parameterized to describe the specific case of the range of natural conditions for degassing of Fe-poor rhyolite in arc systems, to investigate how the redox state of the melt varies during closed-system, equilibrium decompression degassing (Burgisser and Scaillet 2007; Burgisser et al. 2008). The evolution of vapor composition during magma decompression is highly sensitive to the initial vapor mass fraction in the reservoir and also the amount of S in the system. Magmas with little pre-existing vapor in the reservoir show complex redox evolution (Fig. 9); for H2O-poor magmas with low C/S ratios in the melt, magma redox state tends to converge towards NNO. In S-poor systems, the magma diverges from the NNO buffer. In both scenarios, the H2S/SO2 ratio decreases as magma decompresses, by up to 3 orders of magnitude. With a high starting vapor fraction in the reservoir, the trends are simpler – the
Figure 8
The Sulfur Budget in Magmas
229
Figure 9. Evolution of the composition of an H-O-S gas during ascent of a rhyolite magma. Effect of the initial redox state on the ratio H2S/SO2 for three starting values of fO2, each with three different initial gas contents. Also shown are the natural range observed on active volcanoes in convergent settings, and the range calculated by phase equilibria experiments. Figure is modified after Burgisser et al. (2008).
pre-existing vapor tends to dominate over the effects of additional exsolved volatiles, magmas become more reduced during decompression regardless of C/S ratio and H2O content, and the H2S/SO2 ratio of the gases remains roughly constant (Fig. 9). The compositions of gases at the surface are therefore highly dependent on the depth at which melt-vapor separation occurs, the initial vapor content, the initial melt S and H2O content and the fO2. There is clearly a rich and diverse range of behavior for coexisting melts and gases during magma ascent and degassing in terms of oxidation state. These studies provide an explanation for why erupted magmas and gases often yield different oxidation states (with the magma commonly yielding an fO2 that is 1-2 log units lower than the gas, e.g., Gerlach 2004). A particular H2S/SO2 ratio in gases at the surface might result from the degassing of magma with a range of redox states and initial melt C/S, so it must be used in conjunction with other petrological information in order to reconstruct the degassing history.
“Excess sulfur” or more accurately, “excess volatiles” problem As discussed earlier in this chapter, a well-documented observation from many volcanoes is that there is an excess of S (and other volatiles) emitted in the gases, over that which can be supplied by degassing of the erupting melt alone (Wallace 2001, 2005; Scaillet and Pichavant Figure 2003; Shinohara 2008). It seems clear, from the aforementioned extensive compilations of sulfur emissions, that excess volatile emissions are characteristic of a number of eruption types and can be divided broadly into two groups based on degassing mechanisms: 1) effusive and explosive eruptions of evolved magmas, as well as persistent degassing during eruptive pauses (e.g., Soufriere Hills, Mount St. Helens, Pinatubo) and 2) persistently degassing mafic systems with little eruption of magma (e.g., Stromboli, Villarica, Erta Ale and Etna during non-eruptive periods). Explosive and large effusive eruptions of basaltic volcanoes, in contrast, appear to emit S that can be accounted for by the degassing of erupted melt, i.e., there is no excess sulfur (e.g., Mauna Loa, Laki; Wallace 2001; Shinohara 2008). Pre-eruptive vapor. The mechanisms for generating excess sulfur in volcanic systems have been discussed at great length in the literature and consensus has emerged as to the most important of these. Observations of sulfur excesses emitted during explosive eruptions at El
9
230
Wallace & Edmonds
Chichon (1982), Redoubt (1989), Pinatubo (1991) and Mount St. Helens (1980) in particular led to models invoking a pre-eruptive vapor phase as the source of excess sulfur (Luhr 1990; Westrich and Gerlach 1992; Wallace and Gerlach 1994; Gerlach and McGee 1994; Gerlach et al. 1994;) and further, how pre-eruptive vapor might be distributed in a magma body (Wallace et al. 1995; Wallace 2001). Experiments on silicic magmas under a range of pre-eruptive conditions pertaining to arc volcanism shows that S partitions into vapor from silicic melts at high pressures under oxidizing conditions with partition coefficients up to >1000 (Scaillet et al. 1998; Keppler 1999, 2010; see also Webster and Botcharnikov 2011, this volume). However, this mechanism only applies to oxidized systems, (redox state >NNO+1); under reducing conditions any addition of S to the system results in S being taken up by sulfide phases such as pyrrhotite and little addition of S to the vapor (Scaillet et al. 1998). Explosive eruptions of evolved oxidized magma therefore emit S that had exsolved into bubbles in a magma chamber prior to eruption, and the bubbles are then carried by the erupting magma. Further, explosive eruptions may erupt magmas that have accumulated bubbles at the roof of a magma chamber (Wallace 2001; Shinohara 2008). The magnitude of the excess sulfur will depend on the volume fraction of excess vapor coexisting with the magma at depth. Valuable insight into the role of excess vapor, how it may vary between magmas erupted from the same volcano and the effects on eruption style arises from comparative studies of S and other volatile emissions from Mount St. Helens during the 1980-1986 explosive and dome-building eruptions (Gerlach and Casadevall 1986; Gerlach and McGee 1994) and then during the recent 2004-2005 dome-building eruption (Gerlach et al. 2008). The recent eruption appears to involve “flat” magma, containing very little pre-eruptive vapor. The 2004-2005 eruption was characterized by low SO2 emission rates, comparable to the later stages of the 1980-1986 dome building eruptions (Fig. 10). The very large volatile emissions associated with the explosive eruptions in 1980, combined with the estimated erupted volumes and estimates of pre-eruptive dissolved volatiles, were used to infer a pre-eruptive vapor content of ~15 vol% or ~3 wt% (Wallace 2001, 2003). The 2004-2005 dacite is petrologically similar to the 1980-1986 dacite, albeit slightly more evolved. The pre-eruptive vapor content for the 2004-2005 magma was estimated using the same pre-eruptive H2O conditions as the 19801986 dacite and by comparing the amount of volatiles expected from the degassing of the erupted volumes with the amount measured in the gas plume. The pre-eruptive vapor content of the dacite ranged between 0.8 vol% at the beginning of the eruption in November 2004 and 1.35 vol% in January 2005 (Gerlach et al. 2008). The estimated volume of vapor is consistent with geodetic data, which is best described by models of magma reservoir contraction due to the removal of magma with <1.5 vol% compressible fluid (Mastin et al. 2009). These estimates for pre-eruptive vapor are well below the estimates for the 1980-1986 eruption (~15 vol%; Wallace 2001, 2003). During ascent and decompression, the vapor content of the 2004-2005 magma reached 30 vol% at around 2.5 km depth (Fig. 10), where open-system degassing began. Interestingly, the low vapor content of the magma also has implications for the H2S/SO2 ratio in the volcanic gases. Thermodynamic calculations suggest that at depths of 5.2 km, H2S/ SO2 > 40 and at depths of 0.5-1.0 km, H2S/SO2 is 3.3-6.5. Sulfur dioxide becomes dominant at depths of <200 meters in the conduit. Observations suggest that H2S is only a very minor constituent of the volcanic gas phase (H2S/SO2 <0.1), which is consistent with much of the H2S having sufficient time to convert to SO2 during the very slow magma ascent. In the case of the 2004-2005 eruption, the “excess sulfur” problem is less severe than with the 1980-1986 eruption. It is likely that sulfur resided in the pre-eruptive vapor at 200 MPa, consistent with experimental results that predict sulfur-melt partition coefficients of >45 for fO2 of NNO+1 (Keppler 1996; Scaillet et al. 1998; Pallister et al. 2008). The 2004-2005 Mount St. Helens magma was therefore relatively “flat” magma that may be, at least in part, a remnant of the magma body that fed the eruptions of the 1980s. These studies emphasize the importance of the pre-eruptive vapor content of magma in controlling the style and volume of an eruption.
The Sulfur Budget in Magmas
231
Figure 10. An illustration of the eruption of “flat magma” at Mount St. Helens, 2004-2005 (modified from Gerlach et al. 2008). (A) Time series plots of all SO2 emission rates measured during the 1980s (McGee and Casadevall 1994) and during 2004-2005 (Gerlach et al. 2008). Vertical lines indicate the starting dates of the 1980-1986 lava-dome eruptions. (B) Exsolved fluid content of dacite during closedsystem equilibrium degassing starting at 900 °C and 220 MPa to eruption. The pre-eruptive fluid content of the 1980 dacite is shown for comparison (Wallace 2003). Copyright USGS. Used by permission of United States Geological Survey, from Gerlach et al. 2008, US Geol Surv Prof Pap, Vol. 1750, Chapter 26, Fig. 16 and 23, p. 557-565.
Figure 10
232
Wallace & Edmonds
Mafic recharge supplies sulfur and other volatiles. One possibility for the source of preeruptive vapor is that it may be derived from second boiling of a parental magma, in which case the vapor fraction in the magma reservoir may develop very slowly during periods of dormancy between eruptions. Over long timescales, however, during cooling and exsolution of volatiles, magma will tend to crystallize extensively and lose its gas by bubble coalescence and/or rise, which, even in magmas with a high bulk viscosity, will occur on timescales comparable to inter-eruptive periods (Larsen et al. 2004). Alternatively, or in addition, vapor may be supplied to the system by recharging mafic magma, which mingles or mixes with resident magma. There are a number of lines of evidence to suggest that unerupted mafic magmas supply sulfur (and other volatiles) to long-lived intermediate to silicic magma reservoirs, with the transfer of heat and volatiles both triggering crystal-rich andesite or dacite to erupt and perhaps sustaining the eruptions. The mechanisms and timescales for such volatile transfer processes have until recently been unclear. The major questions that need to be addressed regarding this mechanism for explaining excess sulfur emitted to the atmosphere are: 1) is a single injection of mafic magma shortly before eruption of sufficient volume to provide all of the excess sulfur, or is the eruption-triggering injection the last in a series of smaller injections that gradually cause the magma body to become vapor-enriched during the inter-eruptive repose period, and 2) how does sulfur exsolve from the mafic magma and get transferred to the resident magma on eruptive timescales? Diffusion of sulfur through melt and bubble rise through viscous rhyolitic melts are clearly too slow to account for rapid volatile transfer on eruptive timescales (e.g., Sparks 1978; Baker et al. 2008; see Falcone et al. 2011, this volume). Conduit convection, which allows magma to degas at low pressure and then sink, is probably important in basaltic open-vent systems (e.g., Kazahaya et al. 1994), so a recharged magma body could potentially lose its dissolved volatiles by conduit convection involving the gassy, buoyant, mixed magma (Shinohara 2008). However, it has yet to be shown that conduit convection occurs in more silicic systems where the higher bulk viscosity of the magmas results in only very slow convection, unable to supply volatiles at a sufficient rate unless highly unrealistic conduit dimensions are invoked. The strongest case for conduit convection in a rhyolitic system has been made for Satsumi Iwojima, in which unusually high-temperature rhyolitic magma (~1000 °C) results in lower magma viscosity (Kazahaya et al. 1994). Gas transfer from depth through a permeable bubble network has also been considered as a possible mechanism for extracting deep vapor, but it is unlikely that sufficient underpressure can be generated in a 5-10 km depth magma chamber in order to create a permeable network of bubbles that might allow sulfur (and other volatiles) to migrate to the surface in the vapor phase (Wallace 2001). Recently, detailed studies of volcanic gas composition and magma petrology and geochemistry have shed some light on how mafic magmas intruding a magma reservoir at depth might influence the volatile budget. It seems likely, from the studies described below, that small but semi-continuous inputs of oxidized mafic magma mingle and/or mix to some degree with resident magmas, transferring volatiles to a hydrous vapor phase during the process but not altering the bulk composition of the magma reservoir to any significant degree (Fig. 11). Important observations at Popocatepetl volcano in Mexico have recently placed constraints on the contribution to the volatile budget from mafic magma and the mechanism by which gases are transported from deep-seated levels in the volcanic system to the surface, without needing to invoke conduit convection (Roberge et al. 2009). Popocatepetl erupts dacite magma that contains evidence for mixing with mafic magmas in the form of euhedral Mg-rich olivines, which contain basaltic to basaltic-andesite melt inclusions. The mafic melts are rich in volatiles, with H2O-CO2 data indicating vapor saturation pressures of up to 400 MPa, corresponding to depths of up to 15 km. Sulfur concentrations range from >2000 ppm in basaltic-andesite melt inclusions to below detection for rhyolitic matrix glasses. Measurements of volcanic gas fluxes constrain the vapor CO2/SO2 ratio, which is highly sensitive to vapor-melt separation depth.
The Sulfur Budget in Magmas
233
Figure 11. Mechanisms for volatile transfer during recharge of silicic magma chambers by oxidized mafic magmas. Magma mingling involves quenching, crystallization and vesiculation at the interface between the two magmas, with volatile exsolution due to second boiling. Vapor is fluxed through silicic magma, with sulfur derived from the mafic magma partitioning strongly into the vapor phases; there is little transfer of mass from the mafic magma. Erupted lavas may contain enclaves of mafic magma and evidence of disequilibrium caused by heating. Volatiles escaping during eruptive pauses require significant magma permeability or some other mechanism for advecting vapor to the surface (e.g., convection at depth). Magma mixing involves significant fractions of the mafic magma mixing with the silicic magma in the chamber, resulting in hybrid magmas that contain mafic-derived phenocrysts. Upon mixing, sulfur partitions into a vapor phase and then is released to the atmosphere after magma ascent, closed-system degassing and eruption. Modified from Edmonds et al. (2010).
Thermodynamic models suggest that this gas is released from magma at pressures of 150-350 MPa. Mafic magma ascends from depth, degasses due to first and second boiling (the latter during crystallization), and around 30% of it gets mixed with the resident dacite and erupted, degassing completely and advecting exsolved vapor to the surface. The rest quenches and solidifies in the sub-volcanic plumbing system, which is likely to be an array of dikes, sills and small pods of magma rather than a single large integrated reservoir. At Soufriere Hills Volcano, there is abundant evidence for mafic recharge triggering and sustaining the 1995-present eruption and supplying sulfur to the system (Edmonds et al. 2001). The crystal-rich erupting andesite contains decimeter-scale enclaves of basaltic-andesite bulk composition. Petrological evidence of heating, from amphibole and orthopyroxene rims and Figure other evidence of disequilibrium, such as rounded quartz and sieve-textured plagioclase, all point to heating of the andesite before eruption (Murphy et al. 2000). The timing of the heating, inferred to be due to intrusion of mafic magma, is constrained by Fe-Ti oxide diffusion profiles (based on the assumption that the profiles are due to temperature changes at a fixed fO2, which is probably not realistic during magma recharge), which suggests that heating takes place days to weeks before eruption of the andesite (Devine et al. 2003). Studies of volcanic gas flux and composition constrain the contribution of mafic magma-derived volatiles to the
11
234
Wallace & Edmonds
system (Edmonds et al. 2010). Soufrière Hills Volcano gases display a similar range in CO2/ SO2 and SO2/H2S to other arc volcanoes (Fig. 8). The strong inverse relationship between CO2/SO2 and SO2/H2S in Figure 8 does not allow for an important role for gas buffering at a constant SO2/H2S across the range of volcanic systems considered here (Giggenbach 1987). Instead, the spread of data suggest that melt-vapor separation at a range of pressures is the dominant mechanism controlling the ratios. In the case of Soufriere Hills, the gas data cannot be reconciled with models of simple closed- or open-system degassing of the erupting rhyolite, and an extra source of vapor is required. The data are best explained by a model of mixing between a vapor phase in equilibrium with the rhyolite melt at low pressures and a deep, mafic magma-derived vapor with high H2S/SO2 and CO2/SO2, which dominates during eruptive pauses, when persistent degassing takes place (Edmonds et al. 2010). The mafic magma quenches, crystallizes and vesiculates at the interface between the two magmas, liberating a vapor phase which is able to migrate to the surface through the rhyolite, through permeable networks of bubbles at depth and fractures closer to the surface (Fig. 11). One recent study suggests that oxidized mafic magmas transported large amounts of sulfur, mobilized from sulfides in the mantle source, to the shallow silicic magma reservoir prior to the climactic eruption of Pinatubo on 15 June 1991 (de Hoog et al. 2004). Pinatubo andesite dome lavas, erupted prior to the explosive dacitic eruption of 15 June, are the product of mixing between batches of basaltic andesite injected into the magma chamber and resident dacite. Melt inclusions in olivine contain variable amounts of sulfur, up to 1700 ppm, with 85% as sulfate, yielding an oxidation state of NNO+1.5 for the mafic magma prior to mixing, which is similar to the oxidation state calculated for the dacite melt of NNO + 1.5-1.7 (Scaillet and Evans 1999). Upon mixing with resident dacite, sulfur from the mafic magma was both fixed as anhydrite and sulfides in the dacite, and also partitioned into a H2O-CO2 vapor. This vapor may have migrated to the upper parts of the magma chamber and was readily available for discharge into the atmosphere during an explosive eruption (de Hoog et al. 2004). Controversy remains, however, over the role of mafic magma in contributing to the volatile budget of arc volcanoes. Microgranular enclaves contained within the andesite erupted prior to and after the explosive Pinatubo eruption of 15 June 1991 have bulk sulfur contents of 50-1000 ppm (locked in sulfides and late-formed apatite), indicating a variable efficiency in sulfur degassing during magma mingling (Di Muro et al. 2008). Based on the composition and morphology of sulfides in the enclaves, it has been proposed that intruding mafic melt was reduced, with sulfur partitioned between a Cu-poor sulfide (incorporated from country rocks during magma ascent towards the dacite reservoir) and the melt. Upon mixing with hydrous, cooler dacite, the Cu-poor sulfides partially transformed into Cu-rich sulfides and then Ba-Sr sulfates as the mafic melt was progressively oxidized. Melt inclusions record a range of sulfur concentrations, with highest S contents in the least and most evolved glasses, and the least S contents in the glasses with intermediate Mg number, proposed to represent glasses at the sulfate-sulfide redox boundary (Di Muro et al. 2008). The release of S from the melt at the redox boundary (as originally proposed by Kress 1997a, as the source of the massive sulfur release) was almost immediately counteracted by the absorption of S back into the melt, this time dissolved as sulfate, which was later partly incorporated into large amounts of late-stage apatite. Final magma ascent caused melt sulfur depletion through degassing. This study concludes that the mafic recharges are volumetrically insignificant, petrologically do not represent the parent to the dacite, and did not contribute to the huge S emissions associated with the 15 June eruption. The contribution of sulfur (and other volatiles) from crustal sediments is also likely to be important for some volcanic emitters. In particular, degassing of S- and C-rich evaporites, coal and limestones by metamorphic devolatilization and/or assimilation and melting could boost the volatile emissions of a particular volcano or volcanic region considerably. Digestion of sediments can be deduced using geochemical data for S and C isotopes in volcanic gases
The Sulfur Budget in Magmas
235
and for volatiles in melt inclusions and has been shown to be important at Vesuvius (IaconoMarziano et al. 2009). Magma convection and/or open system degassing in low viscosity volcanic systems. Persistent degassing with little eruption of magma at basaltic, basaltic-andesite or silica-undersaturated magma volcanoes is commonplace across the whole range of tectonic settings. The style of volcanic activity typically involves a narrow conduit and bubble bursting and/or strombolian eruptions (e.g., Villarica, Masaya, Stromboli), or a flared conduit and lava lake, with visible evidence of magma turnover (e.g., Erebus, Nyiragongo, Erta Ale, Kilauea summit 2008-present). These volcanoes are characterized, during persistent degassing, by a continuous gas plume rich in sulfur dioxide. The excess sulfur problem is, owing to the minimal magma erupted, extreme at these volcanoes. Measurements of the gas composition and flux have been carried out at many of these volcanoes and there is an abundance of recent work that now converges upon a consensus regarding the mechanisms of volatile supply to the gas plume (Sawyer et al. 2008; Aiuppa et al. 2009; Oppenheimer et al. 2009). The gas compositions are generally close to equilibrium with melt at low pressures and the gas contains large fractions of H2O and SO2. A remarkably steady gas composition and flux points to a magma convection regime, with shortterm fluctuations in gas composition, particularly the proportions of CO2 and halogens relative to SO2, pointing to a) periodic inputs of hotter, more gas-rich magma which disrupts convection (e.g., Oppenheimer et al. 2006, 2009) and/or b) open-system degassing at the topmost part of the conduit whereby either melt-vapor separation and bubble rise occurs (e.g., Edmonds and Gerlach 2007; Sawyer et al. 2008) or permeable bubble networks develop (Burton et al. 2007a). The relative importance of these three processes varies from volcano to volcano. Degassing at Stromboli takes place predominantly by open-system degassing, with gas slugs (with a high C/S ratio) rising from 1-3 km (Burton et al. 2007b). Convective magma degassing and overturn takes place in the conduit to supply a background flux of sulfur-rich gases. Prior to the effusive eruption and paroxysmal explosion in 2007, an increase in the C/S ratio of the gases was interpreted as the result of an enhanced number and rate of rise of deep-sourced (1-3 km) bubbles, perhaps caused by an injection of volatile-rich magma (Aiuppa et al. 2009). Lava lakes are an excellent natural laboratory for the study of magma turnover and the relative importance of open- versus closed-degassing processes. A remarkably steady SO2 flux and gas composition at Nyiragongo Volcano’s lava lake, measured using Fourier Transform infrared spectroscopy (FTIR), suggests a closed-system magma convection and degassing regime, whereby magma ascends and degasses, with melt and vapor in equilibrium until tens of meters depth. Rapid temporal fluctuations in the proportions of halogen species relative to sulfur in the plume indicate some melt-gas separation in the topmost part of the system, when bubbles achieve a large enough size through expansion and coalescence to rise relative to melt (Sawyer et al. 2008). Observations at both Erta Ale and Erebus Volcanoes suggest that SO2 flux scales linearly with lake area (Kyle et al. 1994; Oppenheimer et al. 2004). This suggests that the control or driving force for convection might be the sinking of cool, degassed, more crystalline magma, the volume of which controls the rate of turnover. At Villarica, Chile, reconstructions of the volatile budget suggest that magma convection operates to supply sulfur and other gases to the volcanic plume (Witter et al. 2004). Differential bubble rise was rejected as being the dominant mechanism supplying the gas plume. Typical bubble sizes and number densities (inferred from erupted products) and a 15-meter diameter conduit (from visual observations at the surface) could supply fluxes that would fall short of what is observed by 3-4 orders of magnitude (Witter et al. 2004). Degassed, dense magma is inferred to sink back down the conduit and not remain stored in the edifice, as there is no net deformation observed over time. Melt inclusion evidence for gas fluxing in mafic volcanic systems. There is an interesting parallel between the volcanic gas evidence for vapor transfer from mafic to silicic systems and the evidence from melt inclusions for large excesses of gas fluxing through mid- to upper-
236
Wallace & Edmonds
crustal magma reservoirs and conduit systems of mafic volcanic systems. Mafic magmas are typically vapor-saturated at middle- to upper-crustal pressures because of the relatively low solubility of CO2 in silicate melts (Métrich and Wallace 2008). Vapor saturation pressures for melt inclusions can be calculated using H2O and CO2 data, but these pressures usually represent minimum values because it can be difficult to correct for the amount of CO2 lost post-entrapment to a shrinkage bubble. Calculated pressures for melt inclusions from an individual volcano typically show evidence of having been trapped over a wide range of pressures (commonly from <50 MPa to 300 MPa; Métrich and Wallace 2008; Johnson et al. 2010). The H2O and CO2 variations probably represent trapping of variably degassed melts as magma ascended and crystallized within the upper crust. In detail, however, the data show significant scatter (well beyond analytical error) compared to equilibrium degassing models, and closed-system degassing models cannot successfully predict most of the H2O and CO2 variations (Métrich and Wallace 2008). A possible explanation for the wide range of H2O and CO2 values is that ascending magmas were affected by gas fluxing, in which CO2-rich vapor percolates through the system from below, where it is released by magma degassing deeper in the system (Anderson et al. 1989; Rust et al. 2004; Spilliaert et al. 2006b; Johnson et al. 2008, 2010; Vigouroux et al. 2008). The effect of this fluxing by CO2-rich vapor is to shift melt H2O contents to lower values along vapor saturation isobars as the melts attempt to re-equilibrate with the introduced vapor phase. The evidence for the pervasive effects of gas fluxing in volcanic conduit systems suggests that such magmas pond and degas CO2-rich vapor at lower crustal depths, probably as a result of storage and recharge in lower crustal sills (e.g., Annen et al. 2006). Additional gas could be supplied by magma as it ascends through the lower to middle crust provided that the magma transport system is ‘open’ to gas fluxing. An alternative possibility to explain observed H2O and CO2 variations in melt inclusions is variable post-entrapment loss of H by diffusion through the olivine host (Portnyagin et al. 2008; Gaetani et al. 2009; see also Métrich and Wallace 2008). This process is likely to be more important when magma ascent rates are slower (Johnson et al. 2010). The telltale sign of the proposed gas fluxing process is for many melt inclusions to have higher CO2/H2O than they would acquire during equilibrium closed- or open-system degassing. An obvious question then is whether melt inclusions have anomalously high S/ H2O. In most cases they do not, as evidenced by the agreement between DSvapor-melt calculated from melt inclusion data vs. empirical solubility models (see earlier section). This could be explained because the deeply derived, fluxing gas has relatively high CO2/H2O but low S/ H2O because of the increase of S solubility in hydrous melts with increasing pressure (Fig. 5). As a result, gas fluxing may not add significant dissolved S to melts in the middle to upper crust. If the melts are saturated with a sulfide phase (see earlier), then this would also limit the dissolved S content such that gas fluxing could not increase it. Melt inclusions from small volume, mafic cinder cones often show the greatest evidence of gas fluxing, and in some cases do not show as much S loss by degassing as is predicted by model calculations (Vigouroux et al. 2008; Johnson et al. 2010). In these cases it is possible that S contents in melts at shallow depths remain high as a result of strong gas fluxing.
Magmatic sulfur and ore deposits Sulfur plays an important role in the formation of ore deposits in several different ways (see also Simon and Ripley 2011, this volume). When magmas become saturated with immiscible Fe-S-O liquid or sulfide minerals, chalcophile elements (e.g., Pd, Ir, Au, Cu, Se), including many transition metals, partition preferentially into the sulfide phase (e.g., MacLean and Shimazaki 1976; Rajamani and Naldrett 1978; Jenner et al. 2010). Segregation of the dense sulfide liquid phase can form magmatic base metal sulfide deposits in layered mafic or ultramafic intrusions (Naldrett 2004). Chalcophile element distributions will also be strongly
The Sulfur Budget in Magmas
237
affected during formation of mafic magmas by whether residual sulfide is present in the mantle source during partial melting. Mantle sulfides have a particularly strong effect on the abundance and distributions of platinum group elements (e.g., Righter et al. 2008). Primary MORB magmas are likely to be sulfide saturated in their mantle source, and an immiscible sulfide component is therefore retained in the mantle residue. This causes preferential removal of base metals (e.g., Ni, Pb, Cu, Co, Zn), which have lower partition coefficients in the sulfide phase, and concentration in the mantle residue of precious metals (Au, Ag, Pt, Pd) because of their high partition coefficients for sulfide (Hamlyn et al. 1985). Remelting of highly refractory sources can then produce highly depleted, sulfide-undersaturated magmas such as boninites that are precious metal enriched (Hamlyn et al. 1985). Sulfur and magma oxidation state are also important in the formation of porphyry Cu systems. Porphyry deposits are closely related to underlying composite plutons (at paleodepths of 5 to 15 km) that supply fluids (vapor) and metals into vertically elongate (>3 km) stocks or dike swarms and associated mineralization (Sillitoe 2010). Both experimental studies and data on natural samples suggest that S along with Cl are important in complexation of Cu in magmatic vapor at the pressure-temperature conditions of felsic plutons and hydrothermal systems (Heinrich et al. 1992, 1999; Pokrovski et al. 2008; Simon et al. 2006; Zajacz et al. 2008; Nadeau et al. 2010). Dissolved S in silicate melt may also enhance the solubility of metals such as Au in the melt, so higher S can increase the ability of the melt to dissolve metals, allowing them to become more concentrated during igneous differentiation processes and then released into the vapor phase (Jego et al. 2010). Porphyry Cu systems are initiated by injection of oxidized magma saturated with S- and metal-rich, aqueous fluids from cupolas on the tops of the subjacent parental plutons. The high oxidation state is important because magmas with lower fO2 become saturated with sulfide early in their crystallization history, which causes metals to be removed from the melt because of their strong partitioning into the sulfide phase (Jenner et al. 2010; Sillitoe 2010). For example, studies of high temperature fluid inclusions in barren and mineralized granites and magmatic intrusions in porphyry systems show that the earliest fluids released from barren intrusions are usually metal poor whereas those from the mineralized systems are metal rich (Audetat et al. 2008). The implication of this is that the differences between barren and mineralized systems, while complex and involving many factors (particularly formation of immiscible brine), are at least in part inherited from earlier stages in the development (differentiation) of the systems. Early crystallization of sulfide could thus strongly decrease the metal content of magmas in highly evolved felsic plutons, whereas metals would become concentrated in residual melts in more oxidized systems. Given the evidence that recharge of mafic magma into the roots of intermediate to silicic magma chambers supplies S-bearing vapor and the fact that porphyry Cu systems represent large geochemical anomalies of S in the crust, it has been suggested that mafic magma recharge and vapor transfer are important to the metal budget of porphyry deposits (e.g., Hattori and Keith 2001; Nadeau et al. 2010). The ability of S to enhance the amount of Cu that partitions into the vapor phase released from mafic magmas is thus of potential importance to this process. Analysis of silicate melt and vapor inclusions in plagioclase from a basaltic andesite in Chile provide evidence of a S-rich, Cl-poor magmatic vapor phase in which Cu and Ag were strongly compatible, suggesting that vapor released from mafic magma could be important in transferring volatiles into more silicic magma systems (Zajacz and Halter 2009).
Recycling of sulfur in subduction zones The relatively high S contents of many mafic arc magmas require mantle source concentrations in the range of 250-500 ppm S (Métrich et al. 1999; de Hoog et al. 2001a), higher than MORB source estimates of 150-190 ppm S (Saal et al. 2002, and references
238
Wallace & Edmonds
therein) and the estimates of ~70-120 ppm S in primitive (fertile) mantle based on mantle xenoliths (Lorand et al. 2003) and 200 ± 40 ppm based on S data for komatiites (Palme and O’Neill 2004). These values suggest that the mantle wedge above subducting slabs is enriched in S. Coupled S-Os enrichments reported for some arc mantle xenoliths have been ascribed to high-temperature, oxidizing, aqueous fluids (McInnes et al. 1999; Lee 2002), and strongly metasomatized mantle xenoliths from arcs have as much as 230-1190 ppm S and contain abundant sulfides (McInnes et al. 2001; Lee 2002). Sulfur isotope data for fumarolic gases, submarine basaltic glasses, and subaerial whole rock samples from arcs and back-arcs have elevated d34S, suggest recycling of seawater sulfate (Ueda and Sakai 1984; Woodhead et al. 1987; Alt et al. 1993; Imai et al. 1993; Métrich et al. 1999; de Hoog et al. 2001b; Marini et al. 2011, this volume). In some cases, high d34S values could also reflect shallow degassing processes, so caution is required in interpreting d34S values in terms of subduction recycling processes (e.g., Mandeville et al. 1998, 2009; Goff et al. 2000). Fertile mantle lherzolite xenoliths tend to have d34S values close to 0‰, similar to meteorite values, but highly depleted (clinopyroxene-poor) xenoliths from southern Mongolia have d34S values of +5 to +7‰ accompanied by high Ba and K contents and high 87 Sr/86Sr ratios, all of which may be a result of interaction with fluids derived from subducted oceanic crust (Ionov et al. 1992). The high S contents of many arc basaltic magmas, elevated d34S values, and higher fO2 relative to MORB all suggest that percolation of slab-derived fluids in the mantle wedge causes both oxidation and addition of S to the mantle (McInnes et al. 2001; Wallace 2005; Kelley and Cottrell 2009). Geochemical modeling based on melt inclusions from the Chichináutzin volcanic field in central Mexico suggests that several weight percent S is present in the H2Orich component transferred from the slab to the wedge (Cervantes and Wallace 2003b). The high S contents of mafic arc magmas cannot be derived from low-fO2 MORB-type mantle because of solubility constraints imposed by sulfide saturation (Mavrogenes and O’Neill 1999). However, as described previously, at fO2 from FMQ+1 to ~FMQ+2.2, the sulfur content of sulfide-saturated melts increases exponentially such that oxidized subarc mantle could have residual sulfide present during melting and yield primitive melts with ≤ 1.4 wt% S (Jugo 2009). Numerous estimates have been published for the fluxes of the major volatiles, including S, returned from the mantle to the crust and hydrosphere by arc magmatism (Stoiber and Jepsen 1973; Ito et al. 1983; Varekamp et al. 1992; Andres and Kasgnoc 1998; Hilton et al. 2002; Wallace 2005). The flux of S from degassing of arc volcanoes is reasonably well known (except for the contribution from large and infrequent eruptions) because of the abundance of remote sensing data for volcanic SO2 emissions. The main uncertainty in this method is that measured fluxes have to be scaled to all active arc volcanoes worldwide (e.g., Hilton et al. 2002), but different scaling approaches have yielded fairly similar results (Wallace 2005). The best value for global arc S emission based on measured gas fluxes is probably 1×1013 g S/yr (Hilton et al. 2002) because it uses the most rigorous scaling analysis. This approach, however, only gives an estimate for the degassing flux and does not include the likely substantial amounts of S that remain locked in the arc crust in sulfide and/or sulfate minerals. The other basic approach for estimating fluxes of volatiles from arc magmatism is to use data on primary magmatic volatile concentrations based on melt inclusion and submarine glasses and combine it with the flux of mantle-derived magma to the crust in arc regions (e.g., Ito et al. 1983; Peacock 1990; Carmichael 2002; Wallace 2005). This approach has a large uncertainty in that the flux of mantle-derived magmas into the crust at convergent margins is not well known, primarily because the ratio of extrusive to intrusive rocks is poorly constrained (Crisp 1984; Reymer and Schubert 1984; Dimalanta et al. 2002). Using a global arc magma flux of 2.5 ± 0.5 km3/yr (Carmichael 2002) and an average S content of 1500 ± 500 ppm (Fig. 2) yields an arc S flux of 2.1±0.9 ×1013 g S/yr. This value is about a factor of 2 higher than the
The Sulfur Budget in Magmas
239
estimate based on measured SO2 fluxes from arc volcanoes but the values are largely within the mutual uncertainty of the two methods. The differences between the two estimates are permissive of ~50% of the total S returned to the crust via arc magmatism remaining locked up in sulfide and sulfate deposits, with the remainder being vented to the atmosphere. A major goal of subduction zone studies is to quantify the fluxes of volatiles subducted back into the mantle along subduction zones and returned from the mantle to the crust and hydrosphere via magmatism. Estimating the subduction input of S is difficult because of our poor knowledge of the bulk S content of altered oceanic crust. Sediment subduction probably contributes little S because of the lack of pyrite in present-day marine sediments (Canfield 2004). Another major source of uncertainty in comparing inputs with outputs is that data for outputs in back-arc regions are far too sparse to incorporate back-arc fluxes, which could be significant, into the global estimates for arc magmatism (Hilton et al. 2002). Using estimates of S input to subduction zones from Hilton et al. (2002) and the S outputs summarized above, the estimated flux of S returned to the mantle is about 4-7 times greater than the amount returned to the surface reservoir by arc magmatism (Wallace 2005; note, this calculation assumes that the bulk S content of the mantle is negligible before subduction-related enrichment). This is true whether one uses the arc S output based on magmatic S content and magma flux, which potentially accounts for S retained in sulfides and sulfates in the crust rather than degassed to the atmosphere, or the flux based on measured SO2 fluxes from arc volcanoes. The imbalance between input and output becomes even greater if the S content of the mantle before subduction input of S (based on mantle xenolith data) is taken into account. Oxygen isotope data in mid-ocean ridge basalts have recently been interpreted to indicate that the suboceanic mantle becomes enriched by small degree partial melts of subducted, dehydrated, altered oceanic crust (Cooper et al. 2004, 2009). Such a process provides a possible explanation for why the S flux due to arc magmatism seems to be a relatively small part of the overall S budget on the Earth. The oxygen isotope data suggest that substantial S might also be recycled from the slab to deeper parts of the upper mantle by partial melting of the slab at a deeper stage of subduction, after the slab has largely finished dewatering beneath arcs and back-arcs.
ACKNOWLEDGMENTS We would like to thank Maxim Portnyagin and Bruno Scaillet for thorough and careful reviews that led to many improvements in the final manuscript. PW would also like to thank Terry Plank for many helpful discussions regarding vapor-melt partitioning and degassing of S. Finally, we would like to thank Jim Webster and Harald Behrens for inviting this review and for their editorial comments.
REFERENCES Aiuppa A, Federico C, Giudice G, Giuffrida G, Guida R, Gurrieri S, Liuzzo M, Moretti R, Papale P (2009) The 2007 eruption of Stromboli volcano: Insights from real-time measurement of the volcanic gas plume CO2/ SO2 ratio. J Volcanol Geotherm Res 182:221-230 Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007) Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology 35:1115-1118 Allard P, Burton M, Muré F (2005) Spectroscopic evidence for a lava fountain driven by previously accumulated magmatic gas. Nature 43:407-410 Alt JC, Shanks WC, Jackson MC (1993) Cycling of sulfur in subduction zones – The geochemistry of sulfur in the Mariana-island arc and back-arc trough. Earth Planet Sci Lett 119:477-494 Anderson AT (1975) Some basaltic and andesitic gases. Rev Geophys Space Phys 13:37-55 Anderson AT, Newman S, Williams SN, Druitt TH, Skirius C, Stolper E (1989) H2O, CO2, Cl, and gas in Plinian and ash-flow Bishop rhyolite. Geology 17:221-225
240
Wallace & Edmonds
Andres RJ, Kasgnoc AD (1998) A time-averaged inventory of subaerial volcanic sulfur emissions. J Geophys Res 103:25251-25261 Andres RJ, Rose WI, Kyle PR, deSilva S, Francis P, Gardeweg M, Moreno Roa H (1991) Excessive sulfur dioxide emissions from Chilean volcanoes. J Volcanol Geotherm Res 46:323-329 Annen C, Blundy JD, Sparks RSJ (2006) The genesis of intermediate and silicic magmas in deep crustal hot zones. J Petrol 47:505-539 Audetat A, Pettke T, Heinrich CA, Bodnar RJ (2008) The composition of magmatic-hydrothermal fluids in barren and mineralized intrusions. Econ Geol 103:877-908 Baker DR, Freda C, Brooker RA, Scarlato P (2005) Volatile diffusion in silicate melts and its effects on melt inclusions. Ann Geophys 48:699-717 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Behrens H, Stelling J (2011) Diffusion and redox reactions of sulfur in silicate melts. Rev Mineral Geochem 73:79-111 Benjamin ER, Plank T, Wade JA, Kelley KA, Hauri EH, Alvarado GE (2007) High water contents in basaltic magmas from Irazù volcano, Costa Rica. J Volcanol Geotherm Res 168:68-92 Bezos A, Humler E (2005) The Fe3+/SFe ratios of MORB glasses and their implications for mantle melting. Geochim Cosmochim Acta 69:711-725 Blundy J, Cashman K, Humphreys M (2006) Magma heating by decompression-driven crystallization beneath andesite volcanoes. Nature 443:76-80 Brandon AD, Draper DS (1996) Constraints on the origin of the oxidation state of mantle overlying subduction zones: an example from Simcoe, Washington, USA. Geochim Cosmochim Acta 60:1739-1749 Burgisser A, Scaillet B (2007) Redox evolution of a degassing magma rising to the surface. Nature 445:194-197 Burgisser A, Scaillet B, Harshvardhan (2008) Chemical patterns of erupting silicic magmas and their influence on the amount of degassing during ascent. J Geophys Res 113:B12204 Burton M, Allard P, Muré F, La Spina A (2007b) Magmatic gas composition reveals the source depth of slugdriven Strombolian explosive activity. Science 317:227-230 Burton MR, Mader HM, Polacci M (2007a) The role of gas percolation in quiescent degassing of persistently active basaltic volcanoes. Earth Planet Sci Lett 264:46–60 Canfield DE (2004) The evolution of the Earth surface sulfur reservoir. Am J Sci 304:839-861 Carmichael ISE (1991) The redox states of basic and silicic magmas: a reflection of their source regions? Contrib Mineral Petrol 106:129-141 Carmichael ISE (2002) The andesite aqueduct: perspectives on the evolution of intermediate magmatism in west-central (105-99°W) Mexico. Contrib Mineral Petrol 143:641-663 Carmichael ISE, Ghiorso MS (1986) Oxidation-reduction relations in basic magma: a case for homogeneous equilibria. Earth Planet Sci Lett 78:200-210 Carroll MR, Rutherford MJ (1987) The stability of igneous anhydrite: experimental results and implications for sulfur behavior in the 1982 El Chichón trachyandesite and other evolved magmas. J Petrol 28:781-801 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation state: results from measured wavelength shifts of sulfur X-rays. Am Mineral 73:845-849 Cervantes P, Wallace P (2003a) Magma degassing and basaltic eruption styles: a case study of 2000 year BP Xitle volcano in Central Mexico. J Volcanol Geotherm Res 120:249-270 Cervantes P, Wallace P (2003b) Role of H2O in subduction-zone magmatism: new insights from melt inclusions in high-Mg basalts from central Mexico. Geology 31:235-238 Christopher T, Edmonds M., Humphreys MCS, Herd RA (2010) Volcanic gas emissions from Soufriere Hills Volcano, Montserrat 1995–2009, with implications for mafic magma supply and degassing. Geophys Res Lett 37:10.1029/2009GL041325 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulphur in hydrous rhyolitic melts. J Petrol 45 2171-2196 Cooper KM, Eiler JM, Asimow PD, Langmuir CH (2004) Oxygen isotope evidence for the origin of enriched mantle beneah the mid-Atlantic ridge. Earth Planet Sci Lett 220:297-316 Cooper KM, Eiler JM, Sims KWW, Langmuir CH (2009) Distribution of recycled crust within the upper mantle: insights from the oxygen isotope composition of MORB from the Australia-Antarctic Discordance. Geochem Geophys Geosys 10:Q12004 Crisp JA (1984) Rates of magma emplacement and volcanic eruption. J Volcanol Geotherm Res 20:177-211 Czamanske, GK, Moore JG (1977) Composition and phase chemistry of sulphide globules in basalt from the Mid-Atlantic Ridge rift valley near 37°N. Lat. Geol Soc Am Bull 88:587-599 Danyushevsky LV, Della-Pasqua FN, Sokolov S (2000) Re-equilibration of melt inclusions trapped by magnesian olivine phenocrysts from subduction-related magmas: petrological implications. Contrib Mineral Petrol 138:68-83
The Sulfur Budget in Magmas
241
Danyushevsky LV, McNeill AW, Sobolev AV (2002) Experimental and petrological studies of melt inclusions in phenocrysts from mantle-derived magmas: an overview of techniques, advantages and complications. Chem Geol 183:5-24 Davis AS, Clague DA, Schulz MS, Hein JR (1991) Low sulfur-content in submarine lavas – an unreliable indicator of subaerial eruption. Geology 19:750-753 de Hoog JCM, Hattori KH, Hoblitt RP. (2004) Oxidized sulfur-rich mafic magma at Mount Pinatubo, Philippines. Contrib Mineral Petrol 146:750-761 de Hoog JCM, Mason PRD, van Bergen MJ (2001a) Sulfur and chalcophile elements in subduction zones: Constraints from a laser ablation ICP-MS study of melt inclusions from Galunggung Volcano, Indonesia. Geochim Cosmochim Acta 65:3147-3164 de Hoog JCM, Taylor BE, van Bergen MJ (2001b) Sulfur isotope systematics of basaltic lavas from Indonesia: implications for the sulfur cycle in subduction zones. Earth Planet Sci Lett 189:237-252 Devine JD, Rutherford MJ, Norton GE, Young SR (2003) Magma storage region processes inferred from geochemistry of Fe-Ti oxides in andesitic magma, Soufriere Hills Volcano, Montserrat, W. I. J Petrol 44:1375-1400 Di Muro A, Pallister J, Villemant B, Newhall C, Semet M, Martinez M, Mariet C (2008) Pre-1991 sulfur transfer between mafic injections and dacite magma in the Mt. Pinatubo reservoir. J Volcanol Geotherm Res 17:517-540 Dimalanta C, Taira A, Yumul GP, Tokuyama H, Mochizuki K (2002) New rates of western Pacific island arc magmatism from seismic and gravity data. Earth Planet Sci Lett 202:105-115 Dixon JE, Clague DA, Stolper EM (1991) Degassing history of water, sulfur, and carbon in submarine lavas from Kilauea volcano, Hawaii. J Geol 99:371-394 Doyle CD, Naldrett AJ (1987) The oxygen content of “sulfide” magma and its effect on the partitioning of nickel between coexisting olivine and molten ores. Econ Geol 82:208-211 Ebel DS (2011) Sulfur in extraterrestrial bodies and the deep earth. Rev Mineral Geochem 73:315-336 Edmonds M, Aiuppa A, Humphreys M, Moretti R, Giudice G, Martin RS, Herd RA, Christopher T (2010) Excess volatiles supplied by mingling of mafic magma at an andesite arc volcano. Geochem Geophys Geosyst doi:10.1029/2009GC002781 Edmonds M, Gerlach TM (2007) Vapor segregation and loss in basaltic melts. Geology 35:751–754 Edmonds M, Pyle D, Oppenheimer C (2001) A model for degassing at the Soufriere Hills Volcano, Montserrat, West Indies, based on geochemical data. Earth Plan Sci Lett 186:159-173 Eichelberger JC (1995) Silicic volcanism: ascent of viscous magmas from crustal reservoirs. Ann Rev Earth Planet Sci 23:41-63 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Fonseca ROC, Campbell IH, O’Neill HSC, Fitzgerald JD (2008) Oxygen solubility and speciation in sulphiderich mattes. Geochim Cosmochim Acta 72:2619-2635 Freda C, Baker DR, Scarlato P (2005) Sulfur diffusion in basaltic melts. Geochim Cosmochim Acta 69:50615069 Gaetani GA, O’Leary JA, Shimizu N (2009) Mechanisms and timescales for reequilibration of water in olivinehosted melt inclusions. Eos Trans. AGU 90(52) Fall Meet. Suppl., Abstract V51E-1770 Gaetani GA, O’Leary JA, Shimizu N, Bucholz CE (2010) Decoupling of H2O, oxygen fugacity and incompatible elements in olivine-hosted melt inclusions by diffusive re-equilibration. Eos Trans. AGU Fall Meet. Suppl., Abstract V23E-06 Gerlach TM (2004) Comment on paper: ‘Morphology and compositions of spinel in Pu’u ‘O ‘o lava (1996– 1998), Kilauea volcano, Hawaii’—enigmatic discrepancies between lava and gas-based fO2 determinations of Pu’u ‘O ‘o lava. J Volcanol Geotherm Res 134:241-244 Gerlach TM, Casadevall TJ (1986) Evaluation of gas data from high temperature fumaroles at Mount St Helens, 1980-1982. J Volcanol Geotherm Res 28:107-140 Gerlach TM, McGee K (1994) Total sulfur dioxide emissions and pre-eruption vapor-saturated magma at Mount St. Helens, 1980-88. Geophys Res Lett 21:2833-2836 Gerlach TM, McGee KA, Doukas MP (2008) Emission Rates of CO2, SO2, and H2S, Scrubbing, and Preeruption Excess Volatiles at Mount St. Helens, 2004-2005. USGS Prof Pap 1750:543-571 Gerlach TM, Westrich HR, Casadevall TJ, Finnegan DL (1994) Vapor saturation and accumulation in magmas of the 1989-1990 eruption of Redoubt Volcano, Alaska. J Volcanol Geotherm Res 62:317-337 Gerlach TM, Westrich HR, Symonds RB (1996) Preeruption vapor in magma of the climactic Mount Pinatubo eruption: source of the giant stratospheric sulfur dioxide cloud. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Phillipines. Newhall CG and Punongbayan RS (eds) Univ Washington Press p 415-433 Giggenbach WF (1987) Redox processes governing the chemistry of fumarolic gas discharges from White Island, New Zealand. Appl Geochem 2:143-161
242
Wallace & Edmonds
Giggenbach WF (1997) Chemical composition of volcanic gases. In: Monitoring and Mitigation of Volcano Hazards. Scarpa R, Tilling RI (eds) Springer-Verlag, Berlin p 221-255 Goff F, McMurtry GM, Counce D, Simac JA, Roldan-Manzo AR, Hilton DR (2000) Contrasting hydrothermal activity at Sierra Negra and Alcedo volcanoes, Galapagos Archipelago, Ecuador. Bull Volcanol 62:34-52 Hamlyn PR, Keays RR, Warrington EC, Crawford, AJ, Waldron, HM (1985) Precious metals in magnesian lowTi lavas: implications for metallogenesis and sulfur saturation in primary magmas. Geochim Cosmochim Acta 49:1797-1811 Hattori KH, Keith JD (2001) Contribution of mafic melt to porphyry copper mineralization: evidence from Mount Pinatubo, Philippines, and Bingham Canyon, Utah, USA. Mineral Depos 36:799-806 Heinrich CA, Gunther D, Audetat A, Ulrich T, Frischknecht R (1999) Metal fractionation between magmatic brine and vapor, determined by microanalysis of fuid inclusions. Geology 27:755-758 Heinrich CA, Ryan CG, Mernagh TP, Eadington PJ (1992) Segregation of ore metals between magmatic brine and vapor—a fuid inclusion study using pixe microanalysis. Econ Geol 87:1566-1583 Hilton DR, Fischer TP, Marty B (2002) Noble gases and volatile recycling at subduction zones. Rev Mineral 47:319-370 Iacono-Marziano G, Gaillard F, Scaillet B, Pichavant M, Chiodini G (2009) Role of non-mantle CO2 in the dynamics of volcano degassing: the Mount Vesuvius example. Geology 37:319-322 Imai A, Listanco EL, Fuji T (1993) Petrologic and sulfur isotopic significance of highly oxidized and sulfur-rich magma of Mt. Pinatubo, Philippines. Geology 21:699-702 Ionov DA, Hoefs J, Wedepohl KH, Wiechert U (1992) Concentration and isotopic composition of sulfur in ultramafic xenoliths from Central Asia. Earth Planet Sci Lett 111:269-286 Ito E, Harris DM, Anderson AT (1983) Alteration of oceanic crust and geologic cycling of chlorine and water. Geochim Cosmochim Acta 47:1613-1624 Jego S, Pichavant M, Mavrogenes JA (2010) Controls on gold solubility in arc magmas: An experimental study at 1000 °C and 4 kbar. Geochim Cosmochim Acta 74:2165-2189 Jenner FE, O’Neill HSC, Arculus RJ, Mavrogenes JA (2010) The magnetite crisis in the evolution of arc-related magmas and the initial concentration of Au, Ag and Cu. J Petrol 51:2445-2464 Johnson ER, Wallace PJ, Cashman KV, Delgado Granados H (2010) Degassing of volatiles (H2O, CO2, S, Cl) during ascent, crystallization, and eruption of basaltic magmas in the central Trans-Mexican Volcanic Belt. J Volcanol Geotherm Res 197:225-238 Johnson ER, Wallace PJ, Cashman KV, Delgado Granados H, Kent A (2008) Magmatic volatile contents and degassing-induced crystallization at Volcan Jorullo, Mexico: Implications for melt evolution and the plumbing systems of monogenetic volcanoes. Earth Planet Sci Lett 269:477-486 Jugo PJ (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37:415-418 Jugo PJ, Luth RW, Richards JP (2005a) Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim Cosmochim Acta 69:477-503 Jugo PJ, Luth RW, Richards JP (2005b) An experimental study of the sulfur content in basaltic melts saturated with immiscible sulphide or sulfate liquids at 1300°C and 1.0 GPa. J Petrol 46:783-798 Kazahaya K, Shinohara H, Saito G (1994) Excessive degassing of Izu-Oshima volcano: magma convection in a conduit. Bull Volcanol 56:207-216 Kelley KA, Cottrell E (2009) Water and the oxidation state of subduction zone magmas. Science 325:605-607 Keppler H (1996) Constraints from partitioning experiments on the composition of subduction-zone fluids. Nature 380:237-240 Keppler H (2010) The distribution of sulfur between haplogranitic melts and aqueous fluids. Geochim Cosmochim Acta 74:645-660 Kress V (1997a) Magma mixing as a source for Pinatubo sulphur. Nature 389:591-593 Kress V (1997b) Thermochemistry of sulfide liquids I: The system O-S-Fe at 1 bar. Contrib Mineral Petrol 127:176-186 Kress V (2007) Thermochemistry of sulfide liquids II: Ni-bearing liquids at 1 bar. Contrib Mineral Petrol 154:191-204 Kress V, Greene LE, Ortiz MD, Mioduszewski L (2008) Thermochemistry of sulfide liquids IV: density measurements and the thermodynamics of O-S-Fe-Ni-Cu liquids at low to moderate pressures. Contrib Mineral Petrol 156:785-797 Kyle PR, Sybeldon LM, McIntosh WC, Meeker K, Symonds R (1994) Sulfur dioxide emission rates from Mount Erebus, Antarctica, in Kyle PR (ed) Volcanological and environmental studies of Mount Erebus, Antarctica. Am Geophys Union Geophys Monograph 66:69-82 Larocque ACL, Stimac JA (2000) Evidence for open-system behavior in immiscible Fe-S-O liquids in silicate magmas: implications for contributions of metals and sulfur to ore-forming fluids. Can Mineral 38:12331249 Larsen JF, Denis M-H, Gardner JE (2004) Experimental study of bubble coalescence in rhyolitic and phonolitic melts. Geochim Cosmochim Acta 68:333-344
The Sulfur Budget in Magmas
243
Le Roux PJ, Shirey SB, Hauri EH, Perfit MR, Bender JF (2006) The effects of variable sources, processes and contaminants on the composition of northern EPR MORB (8-10°N and 12-14°N): Evidence for volatiles (H2O, CO2, S) and halogens (F, Cl). Earth Planet Sci Lett 251:209-231 Lee CTA (2002) Platinum-group element geochemistry of peridotite xenoliths from the Sierra Nevada and the Basin and Range, California. Geochim Cosmochim Acta 66:3987-4005 Lee CTA, Leeman WP, Canil D, Zheng-Xue AL (2005) Similar V/Sc systematics in MORB and arc basalts: implications for the oxygen fugacities of their mantle source regions. J. Petrol 46:2313-2336 Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Lorand JP, Alard O, Luguet A, Keays RR (2003) Sulfur and selenium systematics of the subcontinental lithospheric mantle: inferences from the Massif Central xenolith suite (France). Geochim Cosmochim Acta 21:4137-4151 Lowenstern JB (1993) Evidence for a copper-bearing fluid in magma erupted at the Valley of Ten Thousand Smokes, Alaska. Contrib Mineral Petrol 114:409-421 Lowenstern JB (1995) Applications of silicate-melt inclusions to the study of magmatic volatiles. In: Magmas, Fluids, and Ore Deposits. Thompson JFH (ed) Mineral. Assoc. Canada Short Course p 23:71-99 Luhr JF (1990) Experimental phase relations of water- and sulfur-saturated arc magmas and the 1982 eruptions of El Chichón Volcano. J Petrol 31:1071-1114 Luhr JF (2001) Glass inclusions and melt volatile contents at Paricutin volcano, Mexico. Contrib Petrol Mineral 142:261-283 Luhr JF (2008) Primary igneous anhydrite: Progress since its recognition in the 1982 El Chichón trachyandesite. J Volcanol Geotherm Res 175:394-407 Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichón, Chiapas, Mexico: mineralogy and petrology of the anhydrite-bearing pumices. J Volcanol Geotherm Res 23:69-108 MacLean WH, Shimazaki H (1976) Partition of Co, Ni, Cu, and Zn between sulfide and silicate liquids. Econ Geol 6:1049-1057 Mallmann G, O’Neill HC (2009) The crystal/melt partitioning of V during mantle melting as a function of oxygen fugacity compared with some other elements (Al, P, Ca, Sc, Ti, Cr, Fe, Ga, Y, Zr and Nb). J Petrol 50:1765-1794 Mandeville CW, Sasaki A, Saito G, Faure K, King R, Hauri E (1998) Open-system degassing of sulfur from Krakatau 1883 magma. Earth Planet Sci Lett 160:709-722 Mandeville CW, Webster JD, Tappen C, Taylor BE, Timbal A, Sasaki A, Hauri E, Bacon CR (2009) Stable isotope and petrologic evidence for open-system degassing during the climactic and pre-climactic eruptions of Mt. Mazama, Crater Lake, Oregon. Geochim Cosmochim Acta 73:2978-3012 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Mastin LG, Lisowski M, Roeloffs E, Beeler N (2009) Improved constraints on the estimated size and volatile content of the Mount St. Helens magma system from the 2004–2008 history of dome growth and deformation. Geophys Res Lett 36, doi:10.1029/2009GL039863 Mathez EA (1976) Sulfur solubility and magmatic sulfide in submarine basalt glass. J Geophys Res 81:42694276 Mavrogenes JA, O’Neill HSC (1999) The relative effects of pressure, temperature and oxygen fugacity on the solubility of sulfide in mafic magmas. Geochim Cosmochim Acta 63:1173-1180 McGee KA, Casadevall TJ (1994) A compilation of sulfur dioxide and carbon dioxide emission-rate data from Mount St. Helens during 1980–88. U.S. Geological Survey Open-File Report 94-212, 24 p. McGee KA, Doukas MP, Gerlach TM (2001) Quiescent hydrogen sulfide and carbon dioxide degassing from Mount Baker, Washington. Geophys Res Lett 28:4479-4482 McInnes BIA, Gregoire J, Binns RA, Herzig PM, Hannington MD (2001) Hydrous metasomatism of oceanic sub-arc mantle, Lihir, Papua New Guinea: petrology and geochemistry of fluid-metasomatized mantle wedge xenoliths. Earth Planet Sci Lett 188:169-183 McInnes BIA, McBride JS, Evans NJ, Lambert DD, Andrew AS (1999) Osmium isotope constraints on ore metal recycling in subduction zones. Science 286:512-515 Métrich N, Bertagnini A, Landi P, Rosi M (2001) Crystallization driven by decompression and water loss at Stromboli volcano (Aeolian island) Italy. J Petrol 42:1471-1490 Métrich N, Schiano P, Clocchiatti R, Maury RC (1999) Transfer of sulfur in subduction settings: an example from Batan Island (Luzon volcanic arc, Philippines). Earth Planet Sci Lett 167:1-14 Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim Cosmochim Acta 60:4151-4160 Métrich N, Susini J, Galoisy L, Calas G, Bonnin-Mosbah M, Menez B (2003) X-ray microspectroscopy of sulphur in basaltic glass inclusions. Inference on the volcanic sulphur emissions. J Phys 4:104:393-398
244
Wallace & Edmonds
Métrich N, Wallace P (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions. Rev Mineral Geochem 69:363-402. Moore JG, Fabbi BP (1971) An estimate of the juvenile sulfur content of basalt: Contrib Mineral Petrol 33:118127 Moretti R, Baker DR (2008) Modeling the interplay of fO2 and fS2 along the FeS-silicate melt equilibrium. Chem Geol 256:286-298 Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts: the conjugated Toop-SamisFlood-Grjotheim (CTSFG) model. Geochim Cosmochim Acta 69:801-823 Moretti R, Papale P (2004) On the oxidation state and volatile behavior in multicomponent gas-melt equilibria. Chem Geol 213:265-280 Moretti R, Papale P, Ottonello G (2003) A model for the saturation of C-O-H-S fluids in silicate melts. In: Volcanic Degassing. Volume 213. Oppenheimer C, Pyle D, Barclay J (eds) Geol Soc Lond Spec Publ p 81-102 Moss R, Scott SD (2001) Gold content of eastern Manus Basin volcanic rocks: implications for enrichment in associated hydrothermal precipitates. Econ Geol 96:91-107 Murphy MD, Sparks RSJ, Barclay J, Carroll MR, Brewer TS (2000) Remobilization of Andesite Magma by Intrusion of Mafic Magma at the Soufrière Hills Volcano, Montserrat, West Indies. J Petrol 41:21-42 Nadeau O, Williams-Jones AE, Stix J (2010) Sulphide magma as a source of metals in arc-related magmatic hydrothermal ore fluids. Nature Geosci 3:501-505 Naldrett AJ (1969) A portion of the system Fe-S-O between 900 and 1080°C and its application to ore magmas. J Petrol 10:171-201 Naldrett AJ (2004) Magmatic Sulfide Deposits: Geology, Geochemistry and Exploration. Springer-Verlag, Berlin and Heidelberg, Germany Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2O-CO2 solution model written in Visual Basic for excel. Comput Geosci 28:597-604 O’Neill HSC, Mavrogenes JA (2002) The sulfide capacity and sulfur content at sulfide saturation of silicate melts at 1400°C and 1b. J Petrol 43:1049-1087 Oppenheimer C (2010) Ultraviolet sensing of volcanic sulfur emissions. Elements 6:87-92 Oppenheimer C, Bani P, Calkins JA, Burton MR, Sawyer GM (2006) Rapid FTIR sensing of volcanic gases released by Strombolian explosions at Yasur volcano, Vanuatu. Appl Phys B 85:453-460 Oppenheimer C, Lomakina AS, Kyle PR, Kingsbury NG, Boichu M (2009) Pulsatory magma supply to a phonolite lava lake. Earth Planet Sci Lett 284:392-398 Oppenheimer C, McGonigle AJS, Allard P, Wooster MJ, Tsanev V (2004) Sulfur, heat, and magma budget of Erta ‘Ale lava lake, Ethiopia. Geology 32:509-512 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Pallister JS, Thornber CR, Cashman KV, Clynne MA, Lowers HA, Mandeville CA, Brownfield IK, Meeker GP (2008) Petrology of the 2004–2006 Mount St. Helens lava dome—implications for magmatic plumbing and eruption triggering. USGS Prof Pap 1750:647-702 Palme H, O’Neill H St. C (2004) Cosmochemical estimates of mantle composition. In: Treatise on Geochemistry. Elsevier, Amsterdam 2:1-38 Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O + CO2 fluids in silicate melts. Chem Geol 229:78-95 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285314 Parkinson IJ, Arculus RJ (1999) The redox state of subduction zones: insights from arc peridotites. Chem Geol 160:409-423 Peacock SM (1990) Fluid processes in subduction zones. Science 248:329-337 Perfit MR, Fornari DJ, Malahoff A, Embley R (1983) Geochemical studies of abyssal lavas recovered by DSRV Alvin from eastern Galapagos Rift, Inca Transform, and Ecuador Rift 3. Trace element abundances and petrogenesis. J Geophys Res 88:10,551-10,572 Pichavant M, Scaillet B, Di Carlo I, Rotolo S, Métrich N (2006) Sulfur in hydrous oxidized basaltic magmas: phase equilibria and melt solubilities. Eos Trans AGU 87(36):V41C-02 Pokrovski GS, Borisova AY, Harrichoury JC (2008) The effect of sulfur on vapor-liquid fractionation of metals in hydrothermal systems. Earth Planet Sci Lett 266:345-362 Portnyagin M, Almeev R, Matveev S, Holtz F (2008) Experimental evidence for rapid water exchange between melt inclusions in olivine and host magma. Earth Planet Sci Lett 272:541-552 Rajamani V, Naldrett AJ (1978) Partitioning of Fe, Co, Ni, and Cu between sulfide liquid and basaltic melts and composition of Ni-Cu sulfide deposits. Econ Geol 73:82-93 Reymer A, Schubert G (1984) Phanerozoic addition rates to the continental crust and crustal growth. Tectonics 3:63-77
The Sulfur Budget in Magmas
245
Righter K, Chesley JT, Calazza CM, Gibson EK, Ruiz J (2008) Re and Os concentrations in arc basalts: the roles of volatility and source region fO2 variations. Geochim Cosmochim Acta 72:926-947 Ripley E, Li C, Moore CH, Elswick K, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem THIS VOLUME Roberge J, Wallace P, Delgado Granados H (2009) Mafic magma recharge supplies high CO2 and SO2 gas fluxes from Popocatépetl Volcano, Mexico. Geology 37:107-110 Roggensack K (2001) Unraveling the 1974 eruption of Fuego volcano (Guatemala) with small crystals and their young melt inclusions. Geology 29:911-914 Rose WI, Stoiber RE, Malinconico LL (1982) Eruptive gas compositions and fluxes of explosive volcanoes: Budget of S and Cl emitted from Fuego volcano, Guatemala. In: Andesites: Orogenic Andesites and Related Rocks. Thorpe RS (ed) Wiley, New York p 669-676 Ruscitto D, Wallace PJ, Johnson ER, Kent A, Bindeman IN (2010) Volatile contents of mafic magmas from cinder cones in the Central Oregon High Cascades: Implications for magma formation and mantle conditions in a hot arc. Earth Planet Sci Lett 298:153-161 Rust AC, Cashman KV, Wallace PJ (2004) Magma degassing buffered by vapor flow through brecciated conduit margins. Geology 32:349-352 Saal AE, Hauri EH, Langmuir CH, Perfit MR (2002) Vapor undersaturation in primitive mid-ocean ridge basalt and the volatile content of Earth’s upper mantle. Nature 419:451-455 Sawyer GM, Carn SA, Tsanev VI, Oppenheimer C, Burton M (2008) Investigation into magma degassing at Nyiragongo volcano, Democratic Republic of the Congo. Geochem Geophys Geosystems 9:doi:10.1029/2007GC001829 Scaillet B, Clemente B, Evans BW, Pichavant M (1998) Redox control of sulfur degassing in silicic magmas. J Geophys Res 103:23,937-23,949 Scaillet B, Evans BW (1999) The June 15, 1991 eruption of Mount Pinatubo. I. Phase equilibria and preeruption P-T- fO2- fH2O conditions of the dacite magma. J Petrol 40:381-411 Scaillet B, Luhr JF, Carroll MR (2003) Petrological and volcanological constraints on volcanic sulfur emissions to the atmosphere. In: Volcanism and the Earth’s Atmosphere. Robock A, Oppenheimer C (eds) Am Geophys Union Geophysical Monograph 139:11-40 Scaillet B, Pichavant M (2003) Experimental constraints on volatile abundances in arc magmas and their implications for degassing processes. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) London Geol Soc Spec Pub 213:23-52 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671-698 Shinohara H (2008) Excess degassing from volcanoes and its role on eruptive and intrusive activity. Rev Geophys 46:RG4005 Sillitoe RH (2010) Porphyry copper systems. Econ Geol 105:3-41 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Simon AC, Pettke T, Candela PA, Piccolli PM, Heinrich CA (2006) Copper partitioning in a melt-vapor-brinemagnetite-pyrrhotite assemblage. Geochim Cosmochim Acta 70:5583-5600 Sisson TW, Layne GD (1993) H2O in basalt and basaltic andesite glass inclusions from 4 subduction-related volcanoes. Earth Planet Sci Lett 117:619-635 Skinner BJ, Peck DL (1969) An immiscible sulfide melt from Hawaii. In: Magmatic Ore Deposits. Wilson HDB (ed) Econ Geol Monograph 4:310-322 Sparks RSJ (1978) The dynamics of bubble formation and growth in magmas: a review and analysis. J Volcanol Geotherm Res 3:1-37 Spilliaert N, Allard P, Métrich N, Sobolev A (2006b) Melt inclusion record of the conditions of ascent, degassing and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy). J Geophys Res 111:B04203, doi:10.1029/2005/JB003934 Spilliaert N, Métrich N, Allard P (2006a) S-Cl-F degassing pattern of water-rich alkali basalt: modelling and relationship with eruption styles on Mount Etna volcano. Earth Planet Sci Lett 248:772-786 Stoiber RE, Jepson A (1973) Sulfur dioxide contributions to the atmosphere by volcanoes. Science 182:577-578 Sun W, Arculus RJ, Kamenetsky VS, Binns RA (2004) Release of gold-bearing fluids in convergent margin magmas prompted by magnetite crystallization. Nature 431:975-978 Symonds RB, Rose WI, Bluth GJS, Gerlach TM (1994) Volcanic-gas studies: methods, results and applications. Rev Mineral 30:l-66 Ueda A, Sakai H (1984) Sulfur isotope study of Quaternary volcanic rocks from the Japanese island arc. Geochim Cosmochim Acta 48:1837-1848 Varekamp JC, Kreulen R, Poorter RPE, Van Bergen MJ (1992) Carbon sources in arc volcanism with implications for the carbon cycle. Terra Nova 4:363-373
246
Wallace & Edmonds
Vergniolle S (1996) Bubble size distribution in magma chambers and dynamics of basaltic eruptions. Earth Planet Sci Lett 140:269-279 Vigouroux N, Wallace P, Kent AJR (2008) Volatiles in high-K magmas from the western Trans-Mexican Volcanic Belt: Evidence for fluid-flux melting and extreme enrichment of the mantle wedge by subduction processes. J Petrol: doi:10.1093/petrology/egn039 Wade JA, Plank T, Melson WG, Soto GJ, Hauri AH (2006) Volatile content of magmas from Arenal volcano, Costa Rica. J Volcanol Geotherm Res 157:94-120 Wallace P (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma bodies. J Volcanol Geotherm Res 108:85-106 Wallace P (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. J Volcanol Geotherm Res 140:217-240 Wallace P, Anderson AT (1998) Effects of eruption and lava drainback on the H2O contents of basaltic magmas at Kilauea. Bull Volcanol 59:327-344 Wallace P, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874 Wallace P, Carmichael ISE (1999) Quaternary volcanism near the Valley of Mexico: Implications for subduction zone magmatism and the effects of crustal thickness variations on primitive magma compositions. Contrib Mineral Petrol 135:291-314 Wallace PJ (2003) From mantle to atmosphere: Magma degassing, explosive eruptions, and volcanic volatile budgets. In: Melt Inclusions in Volcanic Systems: Methods, Applications and Problems (Developments in Volcanology). Volume 5. De Vivo B, Bodnar RJ (eds) Elsevier Science, p 105-127 Wallace PJ, Anderson AT, Davis AM (1995) Quantification of pre-eruptive exsolved gas contents in silicic magmas. Nature 377:612-616 Wallace PJ, Carmichael ISE (1994) Sulfur speciation in submarine basaltic glasses as determined by measurements of SKa X-ray wavelength shifts. Am Mineral 79:161-167 Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions: evidence from Mount Pinatubo. Science 265:497-499 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Westrich HR, Gerlach TM (1992) Magmatic gas source for the stratospheric SO2 cloud from the June 15, 1991, eruption of Mount Pinatubo. Geology 20:867-870 Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov R, Kohn SC, Janousch M (2008) The origin of S4+ detected in silicate glasses by XANES. Am Mineral 93:235-240 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies of sulfur speciation in synthetic and natural glasses. Behrens H, Webster JD (eds) Rev Mineral Geochem THIS VOLUME Witter JB, Kress VC, Delmelle P, Stix J (2004) Volatile degassing, petrology, and magma dynamics of the Villarica lava lake, Southern Chile. J Volcanol Geotherm Res 134:303-337 Witter JB, Kress VC, Newhall CG (2005) Volcan Popocatépetl, Mexico. Petrology, magma mixing and immediate sources of volatiles for the 1994-present eruption. J Petrol 46:2337-2366 Woodhead JD, Harmon RS, Fraser DG (1987) O, S, and Pb isotope variations in volcanic rocks from the Northern Mariana Islands: implications for crustal recycling in intra-oceanic arcs. Earth Planet Sci Lett 83:39-52 Zajacz Z, Halter W (2009) Copper transport by high temperature, sulfur-rich magmatic vapor: Evidence from silicate melt and vapor inclusions in a basaltic andesite from the Villarrica volcano (Chile). Earth Planet Sci Lett 282:115-121 Zajacz Z, Halter W, Pettke T, Guillong M (2008) Determination of fluid/melt partition coefficients by LAICPMS analysis of co-existing fluid and silicate melt inclusions: controls on element partitioning. Geochim Cosmochim Acta 72:2169-2197
9
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 247-283, 2011 Copyright © Mineralogical Society of America
Distribution of Sulfur Between Melt and Fluid in S-O-H-C-Cl-Bearing Magmatic Systems at Shallow Crustal Pressures and Temperatures James D. Webster Dept. of Earth and Planetary Sciences, Division of Physical Sciences American Museum of Natural History Central Park West at 79th St. New York, New York 10024-5102, U.S.A.
[email protected]
Roman E. Botcharnikov Institut für Mineralogie, Callinstr. 3 Leibniz Universität Hannover 30167, Hannover, Germany INTRODUCTION Volatile constituents dissolved in magmatic fluids control fundamental geologic processes including magma ascent, degassing, and eruption; metasomatism; and hydrothermal mineralization in the shallow crust, and volatile compounds involving sulfur are vigorous agents of these fluids. Volcanic outgassing of magmatic volatiles, including SO2 and H2S, to the atmosphere has influenced atmospheric chemistry throughout Earth’s history (Arthur 2000), and of particular importance has been the effectiveness of SO2- and H2SO4-laden stratospheric aerosols to reflect sunlight and cool Earth’s surface (Siggurdson et al. 1990; Oppenheimer et al. 2011, this volume; Wallace and Edmonds 2011, this volume). In addition, many ore metals exhibit chalcophile behavior, i.e., a strong affinity to S. As a result, magmatically sourced S that is typically present as reduced sulfide species plays an essential role in processes of generating nickel sulfide, porphyry copper-gold-molybdenum, high-sulfidation precious metal, and volcanogenic massive sulfide deposits (Hedenquist and Lowenstern 1994; Burnham 1997; Vaughan and Craig 1997; Seo et al. 2009; Simon and Ripley 2011, this volume). Crucial and poorly understood aspects common to each of these processes are determining how and when multi-component fluids exsolve, evolve, and escape from magmas and how significantly S dissolves in such chemically complex fluids. The results of hydrothermal experiments, summarized herein, provide important new insights into these processes. The silicate melt phase (herein abbreviated si-mt) in magmas may contain variable concentrations of S, and S-isotopic data indicate that this S is mostly of magmatic origin (Rye et al. 1984; Wallace 2001). The melts of mafic and intermediate-silica-content magmas generally exhibit greater S concentrations (e.g., > several thousands of ppm S) than felsic melts (Ducea et al. 1994; Wallace and Anderson 2000; Cervantes and Wallace 2003; Métrich and Wallace 2008; Wallace and Edmonds 2011, this volume), and this observation indicates strong relationships between melt composition, temperature, and S solubility. It is also related to the sequestration of S from silicate melt by sulfide and sulfate minerals and/or by sulfide melts (herein abbreviated su-mt) (Mathez 1976; Luhr et al 1984; Wallace and Carmichael 1994) with or without involvement of H2S-, SO2-, SO3,- and potentially S3−- and SO42−-charged aqueous magmatic fluids during magma differentiation (Pallister et al. 1992; Gerlach et al. 1995; Métrich and Clocchiatti 1996; 1529-6466/11/0073-0009$05.00
DOI: 10.2138/rmg.2011.73.9
248
Webster & Botcharnikov
Mandeville et al. 1998; Wallace and Anderson 2000; Wallace 2001; Scaillet and Pichavant 2003; de Hoog et al. 2004; Jugo et al. 2005a; Newton and Manning 2005; Di Muro et al. 2008; Moune et al. 2007; 2009; Webster et al. 2009; Pokrovski and Dubrovinsky 2011). Alkali-enriched si-mt may contain higher S concentrations (e.g., up to thousands of ppm S) than metaluminous or peraluminous melts characterized by lower [(Na+K±Ca)/Al] ratios (Gorbachev 1990; Métrich and Clocchiatti 1996; Raia et al. 2000; Scaillet and Macdonald 2006). Interestingly, magmatic S abundances vary with tectonic setting, as mafic convergent-margin magmas tend to contain more S than other non-alkaline, basaltic magmas (Ducea et al. 1994; de Hoog et al. 2001; Wallace 2005). Silicate melt inclusions from subduction-related basalts, for example, have been reported to contain as much as 0.7 wt% S (Wallace and Edmonds 2011, this volume). Moreover, the concentrations of S and the proportion of S relative to other volatile constituents in arc magmas may vary as a function of subduction-zone depth across volcanic arcs as reported, for instance, for magmas of Kamchatka (e.g., Churikova et al. 2007; Portnyagin et al. 2007). Globally, the abundances of volatile constituents in volcanic gases of mafic to felsic magmas vary widely: H2O (30-90 mol%), CO2 (10-40 mol%), SO2 (5-50 mol%), H2S (<2 mol%), and H2 (<2 mol%) with lesser HCl and HF (Symonds et al. 1994; Williams-Jones and Rymer 2000; Delmelle and Stix 2000; Aiuppa et al. 2009a; Oppenheimer et al. 2011, this volume) (Fig. 1). The ranges in composition of volcanic vapors reflect the influences of tectonic setting; magma source(s); volatile solubilities in melts; pressure (P), temperature (T), fS2, and fO2 of initial volatile exsolution and partitioning between si-mt and fluid(s); and how the S-O-H-C-Cldominated gases evolve chemically with cooling, magma (i.e., melt) evolution, and ascent. The main S-bearing species occurring in natural magmatic fluids are SO2, H2S, S2, COS, and SO3 as determined by direct sampling/analysis and remote sensing of volcanic gas compositions (Symonds et al. 1994; McGonigle and Oppenheimer 2003 and references therein; Oppenheimer et al. 2011, this volume). The exact proportions of the gas species depend on the magmatic conditions, mainly on T, P, and redox state of the system (Giggenbach 1996, 1997). Hence, S speciation can change as gases exsolve from magma and ascend to the surface during pre-, syn-, or post-eruptive degassing (Burgisser and Scaillet 2007). Despite the fact that vapor
Figure 1. Schematic plot modified after Delmelle and Stix (2000) showing relative CO2, S, and HCl contents (normalized to an H2O-free basis; axes represent ratios of volatile components) of volcanichydrothermal gases from fumaroles representing shallowly degassed volatiles and volcanic-magmatic gases from 8 volcanoes representative of subduction-zone, oceanic-rift, and hot-spot tectonic environments. Data from Martini and Buccianti (1997), Delmelle and Stix (2000), and Aiuppa et al. (2009a).
Sulfur Distribution between Melts and Fluids in Magmatic Systems
249
compositions vary as a function of these and other variables and processes, volcanologists have managed to measure the abundances and ratios of volatiles in volcanic gases in order to interpret degassing and, potentially, to forecast eruptive behavior (Aiuppa et al. 2009b), since the amounts and proportions of volatile constituents in si-mt and fluids of mafic to felsic magmas are closely related to volcanic activity and in particular to the style of volcanic eruptions (Wallace 2001; Edmonds et al. 2003; Spilliaert et al. 2006a,b; Burton et al. 2007; Edmonds and Gerlach 2007; Edmonds and Herd 2007; Aiuppa et al. 2009b; Christopher et al. 2010; Oppenheimer et al. 2011, this volume). In addition, although not directly pertinent to the geological realm and out-of-scope for this review chapter, it is noteworthy that gas-melt interactions in S-bearing systems are important in technical processes of silicate glass manufacturing, because S plays important roles in the coloring and fining of molten glasses at ambient pressure (see Falcone et al. 2011, this volume; Müller-Simon 2011, this volume). In this regard, it is interesting that some commercial glass production involves T that far exceed magmatic T and, as a result, the SO42− dissolved in molten sodium-silicate glass decomposes and degasses as SO2 plus O2 at 1500 °C and strongly oxidizing conditions (Backnaes and Deubener 2011, this volume). This behavior contrasts with the behavior of oxidized S species (SO42−) in natural magmas as will be shown below. This chapter reviews the results of experimental, empirical, and thermodynamic studies on the dissolution of S in its various reduced and oxidized forms in chemically complex S-, O-, and H- ± C- and/or Cl-bearing si-mt and the partitioning of S between si-mt and fluids relevant to magmatic systems. This chapter also describes the role of S-bearing, multi-component magmatic fluids in processes leading to magma evolution, fluid exsolution, and volcanic eruption.
BACKGROUND It is important to clarify terminology used in this review. We define fluids to include all physically mobile, non-crystalline, polycomponent phases in volcanic and associated magmatic systems, but we do not, herein, treat si-mt as fluids even though they can be interpreted as such by this definition. At magmatic T and shallow crustal P, relatively low-density fluid phases are denoted as vapor, or synonymously as gas, and are typically dominated by H2O and/or CO2 with lesser SO2, H2S, H2, and acid halide species. Fluid phases of greater density are referred to as liquid and typically are either an aqueous phase (H2O-dominated); non-volatile molten sulfide; or H2O-bearing, electrolyte-bearing saline liquids variably enriched in molten chlorides, sulfates and/or carbonates. For all experiments discussed herein and for the corresponding natural systems addressed, the silicate melts may coexist with chemically complex fluid(s) that include: vapor, or hypersaline liquid, or vapor plus hypersaline liquid with or without S-bearing phases. In this chapter, we refer to the saturation concentrations of S in si-mt (i.e., melts that are saturated with respect to anhydrite or sulfide phases or su-mt) as S solubilities; all other reported S contents of si-mt (i.e., melts coexisting with mixed O-H-S±C±Cl fluids or si-mt-only systems) refer to S concentrations. All values of DSfl/mt, i.e., the partition coefficient for S, are reported as (wt% S in fluid(s)/wt% S in si-mt) unless specified otherwise. At equilibrium, the volatile concentrations of mineral-bearing and fluid-saturated si-mt are controlled by the chemical activities, ai, of volatile components, i, in the coexisting phases: fluid or fluids, sulfide minerals or su-mt, or other volatile-bearing minerals. For heterogeneous equilibria, the chemical potentials, mi, of all volatile components are equivalent for the coexisting phases: (1) misi-mt = mifl = mimin where f l represents vapor, hypersaline liquid, or aqueous liquid and min represents any mineral. The chemical activity of i is related to the chemical potential as:
250
Webster & Botcharnikov mi si -mt = mi °si -mt + RTlnai si -mt
(2)
where mi° represents the standard state chemical potential, T is the temperature and R is the universal gas constant. The activities of the volatile components in the coexisting phases are a function of the mole fraction of the volatiles, Xi, where: ai = Xigi (in each coexisting phase)
(3)
and gi is the activity coefficient. In addition, the activity of volatile component i is related to its fugacity in the fluid as: ai fl =
fi fl f °i fl
(4)
where fi fl is the fugacity of i in the fluid and f °i fl is the standard state fugacity of i in the fluid. The addition of another volatile, e.g., CO2, to the aqueous fluid phase will reduce the XH2Ofl by dilution, and assuming ideal activity-composition behavior in the system (i.e., gH2Ofl and gH2Osi-mt are equal to unity), CO2 addition should also reduce fH2Ofl and, hence, aH2Ofl and aH2Osi-mt. A related issue is how the addition of S to melts saturated in aqueous and aqueous-carbonic fluids influences aH2Ofl, aH2Osi-mt, aCO2fl, and aCO2si-mt. The experiments reviewed herein provide important new constraints on volatile mixing relationships for H2O, CO2, S, and Cl in fluidsaturated melts.
Experimental background Sulfur solubility in mineral- and/or fluid-saturated si-mt has been investigated through hydrothermal experimentation for nearly one hundred years, and it is supported by an extensive review literature (Vogt 1917; Haughton et al. 1974; Carroll and Webster 1994; Baker and Moretti 2011, this volume; Parat et al. 2011, this volume). Experiments on the partitioning of S between si-mt, fluids, and minerals, on the other hand, have been much more limited. The results of S-bearing hydrothermal experiments are used typically to interpret the behavior and role of S in geologic processes (Gerlach 1986; Gerlach et al. 1994, 1995; Scaillet et al. 1998; Core 2004). Experimental data like these are also crucial for the calibration of empirical and thermodynamic modeling of volcanic, magmatic, and magmatic-hydrothermal processes (Moretti et al. 2003; Moretti and Ottonello 2005; Scaillet and Pichavant 2005; Burgisser et al. 2008). Observations from hydrothermal experiments (Keppler 1999, 2010) and nature (Scaillet and Pichavant 2003) and associated thermodynamic modeling—for compositionally simple systems involving S and perhaps only one other volatile component—demonstrate that S is much more soluble in aqueous volcanic vapors and magmatic-hydrothermal fluids than in coexisting si-mt. Geologically and geochemically significant amounts of S partition in favor of fluid in relatively low-S magmas, so even magmas containing low S concentrations (e.g., hundreds of ppm) can have dramatic impacts on dynamics of magma degassing, volcanic eruption, and atmospheric chemistry. A primary limitation to our current understanding of S dissolution in si-mt and its degassing from magmas is that most hydrothermal experiments and models have focused on S- and H2Obearing si-mt and fluids, so current interpretations of magmatic, magmatic-hydrothermal, and related volcanic processes do not account for the complex influences of CO2, Cl, and other magmatic volatiles. Since magmas contain multiple primary volatile constituents: H2O, CO2, S, and Cl as well as other volatiles and fluxing components (e.g., F, B, and P), and as the addition of a third or fourth volatile component to two-volatile systems can alter volatile partitioning in a complex manner (Botcharnikov et al. 2004; Webster et al. 2009; Beermann et al. 2009), accurate interpretation of volatile behavior for natural magmas requires experiments involving all four primary volatiles. In addition, a wide variety of experimental and analytical techniques have been
Sulfur Distribution between Melts and Fluids in Magmatic Systems
251
employed in these studies because S is a particularly challenging volatile to work with and to analyze. Of particular note is that no experimental investigations have successfully constrained the S concentrations of fluid in equilibrium with si-mt by direct analysis at run conditions.
METHODS Experimentation: advantages and challenges Laboratory-based hydrothermal experiments are the primary means of directly determining S partitioning between si-mt, fluids, and minerals as a function of variable P, T, and system composition (see Table 1). Success with high-P and high-T experiments on S partitioning in magmatic systems can be achieved only if the relevant experimental parameters are known or can be accurately controlled. The most important parameters are melt composition, fO2, fS2, T, and P. Other experimental issues include the: (1) source of S used, (2) capsule material, and (3) fugacities of other volatile components in the experimental charge. Here, we discuss challenges related to and possible consequences of these experimental concerns. The sulfur source. Sulfur is added to starting experimental charges through various means. Sulfur concentrations in melts coexisting with fluids in 1-bar experiments have been determined, for example, involving gas-mixing apparatus (e.g., Fincham and Richardson 1954; Carroll and Webster 1994; Backnaes and Deubener 2011, this volume). In these runs, the chemical activity of sulfur (aS) in the system is controlled by the composition of the mixed gas phase flowing over a si-mt (e.g., by controlling fugacities of S-bearing species, S2, SO2, SO3, COS, and H2S and others, in CO-CO2-SO2 or H2-CO2-SO2 gas mixtures) (Buchanan et al. 1983; O’Neill 1987). Such an approach involves open-system conditions with fixed and easily monitored aS. The fugacities of the various gaseous species are calculated for given temperature using standard thermodynamic data (e.g., the JANAF thermodynamic tables, Chase 1998) and the experimental mixing ratios (Carroll and Webster 1994). Volatile-bearing experiments at P > 1 bar require conditions where the investigated system is isolated within a closed volume (capsule or container). Here, the source of S plays an important role because the complete consumption or conversion of the source or its stability over the entire duration of the experiment may significantly affect the behavior of S. In high-P experiments, S is charged into the capsule as elemental S; as sulfuric acid or sulfate solutions (e.g., Na2SO4, K2SO4, FeSO4⋅H2O); or as S compounds like sulfates (e.g., Na2SO4, CaSO4), other sulfur salts (e.g., Na2SO3, Na2S) and metal sulfides (e.g., monosulfides (MS): FeS, FeS2, PtS, Ag2S, PdS, etc. or intermediate sulfide solutions (ISS): FeNiS2, CuFeS2, etc.). The advantages of an elemental-S source are that it does not contribute any component other than S, it provides an easy way to calculate the initial proportions of components, and it is a simple technique for S loading into the capsule. The disadvantage is that elemental S, which is stable at ambient conditions, becomes a very reactive substance with increasing T and P and easily corrodes some capsule materials, especially if S is in direct contact with the capsule wall (see below). Another concern is that the addition of elemental S leads to loss of cations (alkalis in particular) from the si-mt to the fluid or fluids at relatively oxidizing conditions (e.g., Webster et al. 2009; Beermann 2010). The same complication is related to experiments with H2SO4. Although its application is relatively straightforward (e.g., Keppler 2010), H2SO4 adds both H2O and oxidized S and leads to acid-base exchange reactions that can cause significant extraction of cations from the si-mt. Cation leaching by such solutions may cause difficulties in the interpretation of the run products if the extent of cation extraction is unconstrained. The use of excess sulfates, sulfides, or other S salts (Jugo et al. 2005b; Liu et al. 2007; Backnaes et al. 2008; Webster et al. 2009; Beermann 2010) assures saturation of the system with that particular compound. However, the added cations may significantly change the
801-900
Peralkaline rhyolites
1150
1050
Basalt
Trachybasalt
100-200
25-400
500
200-300
200-800
200
217-398
200
150-156
200
50-300
Pressures (MPa)
S-O-H-Cl
S-O-H-C-Cl
S-O-H-Cl
S-O-H
S-O-H-C
S-O-H-Cl
S-O-H (one run with C)
S-O-H-Cl
S-O-H
S-O-H±C
S-O-H
Dominant Volatiles in Fluids
FMQ+0.7 to FMQ+3.2
FMQ+1.7 to FMQ+3.1
Not reported
FMQ-0.3 to FMQ+0.4
Not reported
NNO+0.5 to NNO+1.6
NNO to NNO+2.6
NNO
NNO-2 to NNO+3
NNO-0.4 to NNO+1.4
Co-CoO to NNO+1.0
Approximate log ƒO22
1-236
7-2804
Insufficient data
16-816
200-2000
2-1096
1-1564
246-1693
29-542
50-360
47-468
Approximate Range of Ds fl/mt 3
Beerman et al. (2010)
Lesne et al. (2011)
Gorbachev (1990)
Moune et al. (2009); Moune personal comm.
Teague et al. (2008)
Webster et al. (2009)
Scaillet et al. (1998)
Botcharnikov et al. (2004)
Scaillet and Macdonald (2006)
Webster et al. (in press)
Keppler (1999; 2010)
References
2
See text and figures for details on the influences of these variables on S partitioning behavior. Co-CoO is the Co-CoO oxygen buffer assemblage; NNO is the Ni-NiO oxygen buffer assemblage and FMQ is the fayalite-magnetite-quartz oxygen buffer assemblage; associated numbers show differences from these assemblages in log units. 3 DS fl/mt is the partition coefficient of S, i.e., DS fl/mt = wt% Sfluid /wt% Smelt.
1
1100
Basalt
1250-1300
Andesite
1050
896-1022
Phonolite to trachyte
Basaltic-andesite to basalt
776-899
Dacite
850
895-912
Haplogranite
Rhyodacite
750-850
Temperature (°C)
Haplogranite
Melt Composition
Table 1. Summary1 of run conditions and melt compositions of the primary experimental studies, reviewed herein, bearing on S partitioning between silicate melt and fluid(s).
252 Webster & Botcharnikov
Sulfur Distribution between Melts and Fluids in Magmatic Systems
253
composition of the si-mt making it, for instance, more alkali-rich (if Na2SO4 or CaSO4 are added) and change the extent of si-mt polymerization, i.e., si-mt structure, which is important for the incorporation of volatiles. Moreover, as explained below, the addition of metal- and S-bearing compounds may impose a specific fS2 in the system, which may be undesirable or lead to incorrect interpretation of S partitioning behavior. Precious metal capsule materials. Sulfur reacts aggressively with metals of highly siderophile character - in particular with Ag, Pt, and Pd - but to our knowledge, there has been no systematic study on the potential influences of chemical reactions between S and noble metal capsules to modify the solubility or f l-mt partitioning behavior of S. Nature, for example, demonstrates the reactive behavior of S in the stability of the minerals argentite (Ag2S), cooperite (PtS), and vysotskite (PdS). The chemical reaction between metals and S may lead to loss of S from the si-mt into the capsule material, corrosion of noble-metal capsules, and may cause experiment failure. Furthermore, the most severe effects occur at reduced conditions where metals form molten or solid sulfide phases. The formation of such phases may also change the fS2. For instance, this effect was observed in experiments where different Me-MeSx assemblages (with Me as Pt, Cu, and Ag) were used to buffer aS in the system (e.g., Mysen and Popp 1980). Since the metal-sulfide assemblage controls the prevailing fS2 in the system, the concentrations of S dissolved in the si-mt show a strong dependence on the composition of the Me-MeSx buffer. The highest concentrations of S in the si-mt correspond to the highest calculated fS2 in agreement with theoretical considerations. Another example is the experimental study of Beermann et al. (2011) indicating that the equilibrium concentration of S in fluid-saturated basaltic melt is significantly lower in the PdS-bearing system than that in FeS-bearing systems at similar fH2. These results provide evidence that the reaction of S with noble metals which are used as either the experiment container or as a buffer source material can strongly change the behaviour of S. The reaction of S with such metals to produce metal sulfides can control the fH2S and fS2 of the system and significantly influence the exchange reactions of S between si-mt, Fe-sulfides (solid or liquid), and fluid phase(s). It may also have unconstrained effects on the redox reactions, because geologically relevant volatile partitioning experiments typically involve numerous chemical components, contain few or no minerals, and hence involve multiple thermodynamic degrees of freedom. No known minerals composed of gold and sulfur have ever been found in nature, and no stable synthetic gold-sulfur compounds have been reported in the literature. Thus, gold is the metal that is typically employed as the capsule material in S-partitioning experiments, but there are discrepancies in the literature regarding potential reactions with and sequestration of S by gold capsules. For instance, F. Parat (pers. communication) observed that up to 0.05 wt% S could be dissolved in the wall of gold capsules for S-bearing andesitic systems at T of 900 °C, and these S concentrations are comparable with those in the silicate melt implying that the effect of S reaction with noble metal can be enormous owing to a typically much larger mass of capsule metal than that of the silicate melt plus fluid. From our own experience, a significant number of experiments in S-rich systems at T of 1050 °C failed because S apparently corroded the walls of gold capsules. Conversely, Keppler (2010) measured the S concentrations of fluids (the dilute, aqueous sulphuric acid starting solutions contained ≤ 6 wt% S) before and after exposure to gold capsules at 200 MPa and 850 °C (i.e., si-mt absent), and 5 of the 7 test capsules showed an apparent loss of S from the fluid that was ≤ 5% relative. The fluids of the other 2 capsules exhibited reduced S contents of 18 to 25% relative, and Keppler (2010) interpreted the missing S to reflect loss of gaseous S species while opening the quenched capsules. The observed differences can be attributed to the source of S, i.e., elemental S in experiments of Parat et al. (2008) and H2SO4 solutions in runs of Keppler (2010), and, hence, to the differences in initial fluid composition or/ and to the presence of Fe in the andesitic system investigated by Parat et al. (2008). The alloying between Au and Fe could potentially lead to a more effective reaction between the capsule
254
Webster & Botcharnikov
material and S than that in haplogranitic systems of Keppler (2010). Moreover, the extent of S sequestration by noble metal capsules is also a function of the rate of S diffusion through the capsule walls. Regarding the use of Au capsules, experimentalists should also be aware that, in practice, tubing made from recycled Au may not be pure due to possible prior reaction with material which contained various metals that alloy with Au. It implies that Au capsules can only be used at T ≤ ca. 1050 °C because the real or functional melting T are actually below the melting point of pure Au, i.e., ca. 1064 °C at 1 bar. Alternative capsule materials. In order to avoid or minimize problems associated with S and noble-metal capsules, several experimental approaches using a double-capsule technique with S-inert containers have been developed. For successful application of S-sustainable containers in volatile-rich magmatic systems, the requirements are: (1) the container should sustain high P and T without cracking and behave as an isolated/closed system to minimize loss of volatiles; (2) P inside the container must be the same as the outside conditions applied in the experiment; (3) the possible reaction between the container and the si-mt should not produce any exotic phases or dramatic changes in si-mt composition; and most importantly, (4) the container should be as inert as possible to reactions with S. Materials that have been used for such containers or included in the containers are graphite (used as a lining for noble-metal capsules, e.g., Jugo et al. 2005a,b; Liu et al. 2007), boron nitride (Mysen and Popp 1980), quartz glass (Fleet et al. 1991), olivine (Gaetani and Grove 1997; Holzheid and Grove 2002; Beermann et al. 2011), and others. It is noteworthy that Gorbachev (1990) reported unsatisfactory test results for Pt capsules lined with alundum, zirconium, molybdenum, and tungsten due to significant interaction of lining material with S, and that the best approach used a sintered mixture of olivine, orthopyroxene, chromite, and peridotite that was stable for a given magma composition. Although these materials are relatively inert to S, some present other disadvantages. Graphite-lined capsules, for example, establish very reducing conditions (log fO2 ≤ FMQ-1) in the system due to redox reactions among carbon species, and evacuated quartz glass tubes work at low-P conditions only. Moreover, other materials like boron nitride, sintered minerals, or olivine capsules saturate the si-mt with particular chemical components thus reducing the applicability of this approach to systems saturated in these phases. In summary, these materials may only be useful for restricted compositional systems and P-T conditions of interest. Constraining fO2, fS2, S speciation, and S diffusivity. The interpretation of experimental volatile partitioning data requires proper understanding of the practical limits to hydrothermal experiments and determination of intensive and extensive parameters that control volatile solubilities. Knowledge and/or control of fO2, fS2, and, S speciation/diffusivity behavior in experimental charges can be challenging. The partitioning of S between si-mt and fluid(s) varies strongly with redox conditions, because S has a variable valence state (i.e., S2−, S0, S4+, S6+) and can be present in multiple species in si-mt (e.g., S2−, SO42−) and coexisting fluids (e.g., SO2, H2S, S2, COS, S3−) or liquids (Carroll and Webster 1994; Clemente et al. 2004; Keppler 1999; 2010; Moretti and Baker 2008; Jugo et al. 2010; Pokrovski and Dubrovinsky 2011; Baker and Moretti 2011, this volume). Oxygen fugacity in experiments is usually controlled through the use of mixed H2-Ar pressure media and a Shaw membrane in experimental apparatus (Scaillet et al. 1992; 1998; Berndt et al. 2002) or through the use of capsules charged with oxygen buffers. The latter approach involves the incorporation of small, sealed precious-metal capsules containing H2O and solid phases representing one of the standard oxygen buffer assemblages (e.g., fayalitemagnetite-quartz, nickel-nickel oxide, hematite-magnetite) that are placed within or near the capsule containing the experimental sample charge. Hydrogen gas is generated by reaction of the buffer assemblage with H2O, and H2 diffuses through the capsule wall and interacts with the
Sulfur Distribution between Melts and Fluids in Magmatic Systems
255
experimental sample charge. With this method, fH2 is fixed and fO2 is indirectly controlled in the course of the experiment; strict fO2 control is challenging because it requires accurate constraints on aH2O in the system (Eqns. 5 and 6): H2 + ½ O2 = H2O aH 2 O K eq = ( fH2 )( fO 2 )1/ 2
(5) (6)
where Keq is the equilibrium constant of Equation (6). The activity-composition relationships for chemically complex C-O-H-S±Cl fluids are, as of yet, only poorly known, so it is challenging to fix the fO2 of complex multi-component experiments precisely. One can constrain the fO2 of experiments through the use of oxygen sensors (Taylor et al. 1992; Scaillet and Evans 1999; Clemente et al. 2004), but their implementation is not trivial because sensor calibration is limited to restricted T-P-fO2 conditions and because the sensor should be inserted directly into the capsule in order to measure fO2 (and not fH2). Apparent values of fO2 are also determined through electron microprobe analysis of the relative abundances of S2− and S6+ in run-product glasses and the application of relevant chemical equilibria (see Ripley et al. 2011, this volume). The multiple oxidation states of S in run products can make the investigation and interpretation of S behavior in hydrothermal experiments difficult. Detailed studies of S speciation have demonstrated that S may occur as S2−, S6+, or a combination of both species in natural and experimental silicate glasses (Fincham and Richardson 1954; Carroll and Rutherford 1988; Wallace and Carmichael 1994; Matthews et al. 1999; Fleet et al. 2005; Jugo et al. 2005b; Backnaes et al. 2008; Evans et al. 2009; Métrich et al. 2009; Wilke et al. 2011, this volume). Using microbeam XANES, Métrich et al. (2009) identified S4+ species as a stable minor component in Fe-free glasses synthesized at 0.4 to 1.6 GPa, and they suggested that sulfite can dissolve as molecular SO2 in some exotic melts under a very narrow range of redox conditions. Subsequently, Jugo et al. (2010) and Botcharnikov et al. (2011) confirmed that S2− and SO42− are the primary S species in experimental basaltic and andesitic glasses. The classical studies of Nagashima and Katsura (1973) and Katsura and Nagashima (1974) illustrate that si-mt, equilibrated at ambient pressure with H2-CO2-SO2 gas mixtures, have a minimum dissolved S content at log fO2 corresponding to ~FMQ+1 (where FMQ+1 is one order of magnitude, in fO2, greater than that of the fayalite-magnetite-quartz oxygen buffer). It is noteworthy that the observed change in S concentration in the si-mt with fO2 does not coincide with the proportions of H2S and SO2 (i.e., the dominant S species in the gas phase), but rather with the conditions where ƒH2S is overtaken by increasing fSO3 (Carroll and Webster 1994). The minimum S concentration of the si-mt corresponds to the dominance of SO2 as a fluid species. Furthermore, it indicates that DSfl/mt involves additional chemical reactions converting S2− and S6+ species of the melt to S2− and S4+ species of the fluid (Carroll and Webster 1994). The fS2 of experimental runs can be controlled by changing the amount of S in the bulk experimental charge and monitored by ensuring that the melt and fluid are in equilibrium with pyrrhotite which typically coexists with reduced magmatic melts at elevated T and P. However, the use of pyrrhotite (Fe1−xS, where x is a stoichiometric coefficient) to control fS2 in experiments is difficult because it has a variable proportion of Fe and S, and, hence, it does not fix the fS2 in the system to a single value. Moreover, a variable redox condition in the system can cause variations in fS2, in Fe1−xS composition, and hence, in the concentration of S dissolved in the si-mt (Wallace and Carmichael 1994; Botcharnikov et al. 2011). On the other hand, the composition of Fe1−xS is useful for constraining fS2 (Toulmin and Barton (1964). Additional details on this method are provided by Froese and Gunter (1976), Clemente et al. (2004), and Bockrath et al. (2004, modified by Liu et al. 2007). The diffusivity of S in si-mt of geological interest affects the extent of S partitioning between fluid and si-mt at disequilibrium conditions, and thus, S diffusivity influences syn-erup-
256
Webster & Botcharnikov
tive magmatic degassing. Experimental data indicate that S diffusion coefficients in anhydrous si-mt strongly depend on melt composition and T; these coefficients vary in the range of 10−16 to 10−11 m2/sec at T of 800 to 1600 °C (Baker et al. 2005, Behrens and Stelling 2011, this volume). The addition of H2O increases the diffusivity of S in rhyolitic and andesitic melts by up to 2 orders of magnitude, and for comparison, S diffuses 2 to 3 orders of magnitude slower than H2O in H2O-rich silicic melts. Details on S diffusivity in si-mt are given in Behrens and Stelling (2011, this volume).
Analytical: issues and challenges Extraction of experimental products from run capsules. Accurate analyses and correct interpretation of experimental results require that particular attention be paid to the complete extraction of experimental run products (e.g., su-mt, si-mt, and fluid) from run capsules. Since most si-mt are easily quenched in rapid-quench experimental apparatus, the major-element and volatile composition of the run-product glass can be readily analyzed using generally available analytical methods (described below). Note, however, that the speciation of volatile components in quenched glasses may resemble speciation at the glass transition T rather than that at applied experimental conditions (e.g., Nowak and Behrens (1995) for H2O). Although potential changes in the speciation of S during quench remain to be investigated (Métrich et al. 2009; Mueller-Simon 2011, this volume), recent results show that the proportions of SO42− and S2− species in quenched glasses correspond to those predicted from theoretical considerations (Jugo et al. 2010). Conversely, su-mt are difficult to quench due to very low viscosity; rather, they may crystallize to a multiphase assemblage particularly if intermediate solid solution compositions are initially used or produced during the run. Here, we address analytical methods for characterization of experimental products. Analysis of S in glasses and minerals. The concentrations of S in the glasses and minerals of the experiments addressed in this chapter were determined by electron microprobe (EPMA) and secondary ion mass spectrometry (SIMS) analysis; other investigators have employed laser ablation inductively coupled plasma-mass spectrometry (LA-ICPMS) for S analyses of silicate glasses (Kamenetsky et al. 1999). The spatial distribution of S in S-bearing phases can be mapped with these techniques and with other methods including backscattered-electron imaging scanning electron microscopy (BSE-SEM). Raman spectroscopy is useful for identifying S2− and S6+species in quenched glasses (McKeown et al. 2001; Tsujimura et al. 2004; Klimm and Botcharnikov 2010), but techniques for concentration quantification of individual S species remain to be improved (Lenoir et al. 2009). Individual stable isotopes of S in minerals and glasses of experimental charges have been determined with SIMS and LA-ICPMS (Mandeville 2010; Métrich and Mandeville 2010). Additional details on the analysis of S in glasses and minerals are detailed by Ripley et al. (2011, this volume). Recovery and analysis of S in fluid(s). Fluids are difficult to analyse directly after a run is quenched, because experiments typically involve changes in phase assemblage during the quench. The T change resulting from isobaric run termination leads to formation of single- or multi-phase quench products inside the capsule: for instance, gaseous species may exsolve from the experimental fluid during the quench. Alternatively, amorphous materials and/or crystalline solids may precipitate from the fluid(s) and may adhere to the interior walls of the capsule. If these precipitates are not recovered, the resulting incomplete extraction leads to incorrect determination of the fluid(s) composition at run conditions. Given these concerns, to determine the fluid compositions of the volatile-solubility experiments of Gorbachev (1990): (1) the capsules were weighed and opened; (2) the liquid at ambient conditions was carefully washed out of the capsule, evaporated over a water bath, and dried at 110 °C; and (3) the accumulated dry residue of the fluid was weighed, mixed with lithium metaborate, and fused to glass for subsequent analysis by EPMA.
Sulfur Distribution between Melts and Fluids in Magmatic Systems
257
It can be exceedingly difficult to establish the concentration of S in compositionally complex run-product fluids, because without special effort or attention, CO2, SO2, H2S and other gaseous components (at ambient conditions) will escape when the precious metal capsule is punctured. The loss of gaseous components from run products while opening quenched experimental capsules can theoretically be minimized and/or avoided through the use of cold traps that separate gases by freezing them. This approach has been applied successfully to gas-rich fluid inclusions (Roedder 1984; Salvi and Williams-Jones 2003 and references cited therein), but to our knowledge this method has not been used in practice for the recovery of experimental hydrothermal fluids. Alternatively, the mass proportions of volatile gaseous components in run-product fluids can be determined using classical weight-loss methods as has been applied to H2O- and CO2bearing fluids (Tamic et al. 2001; Botcharnikov et al. 2006; Webster et al. in press). With this approach: (1) the capsule is weighed; (2) the fluid phase is frozen by placing the capsule in liquid nitrogen; (3) the capsule is pierced with a needle; (4) after warming to ambient T, the capsule is weighed to determine the mass of CO2 in the fluid and (5) the capsule is placed into a drying furnace at 110 °C for 3-5 minutes and subsequently weighed to measure the mass of H2O lost from the capsule. As noted, this method works for simple two-volatile fluids, but the application of this approach to S-bearing multi-component fluid(s) has not been reported in the literature. Theoretically, a freezing/heating stage could be used to distinguish between different volatiles in multi-component fluids, because each has its own T of liquid/vapor or solid/liquid phase transition. For instance, at 1 bar: (1) SO2 condenses to liquid at −10.1 °C and freezes at −73 °C; (2) H2S condenses to liquid at -60.3° to −61 °C and freezes at −86 °C (the triple point is −85.6 °C); and (3) SO3 condenses at 45 °C and freezes at 16.9 °C. For comparison, the freezing point of CO2 is 78.5 °C, whereas CO condenses to a liquid at −191.5 °C and freezes at −205 °C. However, it must be noted that the various liquid and gas phases may react with each other (e.g., the oxidation state of S species may change) during these freezing and heating steps thus deleteriously affecting the results. These problematic issues often force experimentalists to apply the mass-balance technique to constrain the S content of fluids at run conditions. The concentrations of volatiles in the fluid are calculated on the basis of the mass proportions of different phases in the system, on the amounts of added volatile-bearing materials, and on the concentrations of volatiles measured in the experimental products. This approach requires quantitative data and constraints on the identities and quantities of all S-bearing condensed phases in the run products. Moreover, as noted previously, some cations are extracted from si-mt by S-bearing fluid(s) during fluid-melt interactions, and these constituents of the fluids must be accounted for (e.g., Webster et al. 2009; Beermann 2010). If S-bearing minerals are present at run conditions, one must know the abundances of these minerals (i.e., it can be difficult to determine accurate abundances of trace sulfide or sulfate minerals in run products). Their presence involves large imprecision on the computed concentrations of S in fluid, and hence, rigorous error calculations are required to verify these S concentrations. The presence of S-bearing minerals in experiments, on the other hand, can be useful. The utility of pyrrhotite has been discussed. Moreover, a recent study on S dissolution in fluids and si-mt employed a small apatite “canister and lid” assembly which contained a large anhydrite crystal that was directly used to determine the solubility of CaSO4 in NaCl-bearing aqueous fluids (Webster et al. 2009). During the runs the melt was in chemical communication with the fluid(s) and with the anhydrite crystal in the canister (via physical contact with the fluid(s)). After each experiment, the residual mass of anhydrite in the apatite canister was recovered and weighed, and the difference between the starting and final weights of the crystal determines the quantity of CaSO4 that dissolved into the fluids and melt during the run.
Webster & Botcharnikov
258
It follows that improved methods for the direct analysis of S in fluids are needed. In a recent study aimed at avoiding the problem of S loss while opening experimental capsules, Keppler (2010) first chilled his experimental capsules with liquid nitrogen, then pierced the capsules and submerged them in an aqueous solution of NaOH and H2O2 in polyethylene bottles, and sealed the bottles immediately afterward. This solution oxidized the reduced S to SO42−; subsequently, the SO42− was analyzed with ICP-AES. Moreover, Keppler (2010) used Raman spectroscopy to identify the S species in fluid inclusions trapped in the quenched experimental glasses. This latter method, however, only provided qualitative determination of primary S species in the fluid. It has been demonstrated with other studies of fluid inclusions in minerals (Guillong et al. 2008 and references therein), that the S concentrations of a fluid can be determined using LA-ICPMS analysis. This method was tested and calibrated on quartz-hosted fluid inclusions showing that the limits of detection for S correlate with the inclusion mass and are about 0.003 to 0.01 wt% for 60-mm diameter inclusions. It was also found that contamination can be induced by the ablation of the mineral host, and hence, a special baseline correction is required to improve analytical accuracy and precision. This method provides quantitative data on the bulk S concentration in the fluid phase and shows a promising applicability for experimental products, but it remains to be developed for the quantitative analysis of S in fluid inclusions in glass. The main requirement is the presence of fluid inclusions in solid run-product phases that are large enough for analysis. Alternatively, the concentrations of S-bearing constituents and other volatiles in a vapor phase can be analyzed with gas chromatography (Stevens et al. 1971; Salvi and Williams-Jones 2003 and references cited therein), and the more-dense residual liquid can be analyzed by liquid chromatography (Rethmeier et al. 1997). These approaches have not, yet, been used for the analysis of S in experimental hydrothermal fluids in prior published research.
EXPERIMENTAL RESULTS ON SULFUR PARTITIONING BETWEEN FLUID AND SILICATE MELT Felsic melts — S-H2O±CO2 Felsic melts — S-H2O. Sulfur dissolves strongly into aqueous fluids at P-T conditions representative of the shallow crust. Equilibrium hydrothermal experiments on S partitioning between aqueous fluids and felsic si-mt show that DSfl/mt ranges, broadly, from unity to 1564 with most values of DSfl/mt between 30 and 1000 (Scaillet et al. 1998; Keppler 1999, 2010; Scaillet and Macdonald 2006). These experimental studies were conducted at 50-400 MPa, 750-900 °C, with values of log fO2 approximately equivalent to the NNO-2 to NNO+3 range (where NNO is the Ni-NiO oxygen buffer), and the si-mt included synthetic subaluminous to peraluminous haplogranites, natural metaluminous and peralkaline granites, and a dacitic composition. The volatile components were H2O and S, but one exploratory run of Scaillet et al. (1998) included CO2. Depending on fO2, some runs involving natural starting materials included pyrrhotite and/ or anhydrite in the run products. The presence of these minerals and the use of mass-balance calculations resulted in increased error in computing the S contents of the fluid because of associated difficulties described previously. These studies demonstrate fundamental influences of fO2 and melt composition on the S concentration of felsic si-mt that are saturated in a S-bearing fluid- and/or mineral. The S concentrations of the fluid-saturated haplogranitic melt of Keppler (1999, 2010) ranged from 0.001 to 0.14 wt%, and those of the runs of Scaillet et al. (1998) and Scaillet and Macdonald (2006) varied from 0.024 to 0.23 wt%. Each of these investigations shows the S content of the si-mt varying with fO2; details on the S solubilities in si-mt are out of scope for this report and are provided in Carroll and Webster (1994) and Baker and Moretti (2011, this volume).
Sulfur Distribution between Melts and Fluids in Magmatic Systems
259
The partitioning of S between aqueous fluids and felsic si-mt varies strongly with si-mt composition. With all other variables held constant, DSfl/mt increases with increasing SiO2 in melts of hydrothermal experiments (Scaillet et al. 1998; Webster et al. 2009), and this strong correlation is verified by values of DSfl/mt constrained thermodynamically for natural magmas (Scaillet et al. 1998; Scaillet and Pichavant 2003). An important consequence of this relationship is that the S content of a magmatic fluid will increase with progressive magma evolution, and so felsic si-mt should be the most efficient emitters of volcanic S-bearing gases. DSfl/mt varies in a complex manner with fO2. Experiments of Scaillet et al. (1998) involving si-mt of Mt. Pinatubo dacite show that anhydrite-bearing runs at log fO2 ≥ NNO+1.2 exhibited significantly greater values of DSfl/mt (e.g., hundreds to thousands) than pyrrhotite-bearing runs at log fO2 < NNO+1 (e.g., DSfl/mt < 10). Increasing values of DSfl/mt from ca. 100 to 500 were also observed with si-mt of a natural, strongly alkaline rhyolite at 150 MPa and 800 °C as log fO2 increased to values > NNO+1, but in contrast two less-alkaline si-mt at the same P and T exhibited no significant change in DSfl/mt with log fO2 varying from NNO-1.6 to NNO+2.9 (Scaillet and Macdonald 2006). These observations differ dramatically with those of Keppler (2010) for haplogranite melt at 200 MPa and 850 °C. His runs showed DSfl/mt decreasing with increasing fO2. Values of DSfl/mt varied from 468 (at log fO2 equivalent to that of the Co-CoO buffer) to 47 (at log fO2 of NNO+0.5 to NNO+1), and an additional increase of log fO2 from ca. NNO+1 to that of the Fe2O3-Fe3O4 buffer showed no change in S partitioning (i.e., DSfl/mt was 49). In way of explaining these differences, Keppler (2010) noted that: (1) thermodynamically estimated data on S partitioning between fluid and si-mt do not vary significantly with fO2 (from ca. NNO-0.5 to NNO+1.7) for natural felsic to andesitic magmas (Scaillet and Pichavant 2003), and (2) that the low (fluid/melt) ratios of some of the experiments of Scaillet et al. (1998) may have caused significant error in the calculated DSfl/mt. For comparatively reducing conditions (ca. log fO2 < NNO+0.5), S is preferentially incorporated in favor of aqueous fluids that are in equilibrium with haplogranitic si-mt, but the S content of natural fluids will likely be limited by S sequestration due to pyrrhotite crystallization in felsic magmas at such fO2 (Keppler 2010). At higher fO2, however, the results on S partitioning are quite variable. With log fO2 > NNO+0.5, most thermodynamically estimated values of DSfl/mt for felsic magmas range from 100 to ca. 1000 (Scaillet and Pichavant 2003), experimental values of DSfl/mt for three alkaline si-mt at 150 MPa range from ca. 70-500 (Scaillet and Macdonald 2006), experimental values of DSfl/mt for molten Mt. Pinatubo dacite at 224 MPa are near 1000, and experimental values of DSfl/mt for calcium- and anhydrite-free haplogranite melt are near 50 (Keppler 2010). It should be noted, here, that the chemical associations of reduced and oxidized S species with H+ and the primary cations dissolved in aqueous fluids like these are poorly known at present. The issue of speciation may be pertinent to the differences in DSfl/mt observed for these haplogranitic versus those of the more chemically complex granitic melts. If the presence of Fe, Mg, and Ca in si-mt enhances the concentration of S in the fluid, then DSfl/mt should be greater with natural versus synthetic, haplogranitic melts at equivalent conditions of T, P, and fO2. It follows that future research should work to determine the speciation of S in geologically relevant, magmatic fluids. In contrast to the influences of si-mt composition and fO2, changes in T and P have less of an effect on S partitioning with felsic melts. Keppler’s (2010) haplogranitic runs at 750-850 °C and Scaillet and Macdonald’s (2006) runs at 800-900 °C show only minor changes in DSfl/mt with T. Keppler’s (2010) haplogranitic runs at 850 °C do show DSfl/mt varying from 58±3 to 94±9 to 47±4 to 68±5 as P increases from 50 to 100 to 200 to 300 MPa at constant fO2. It should also be noted that fS2 influences S partitioning, because the concentration of S in si-mt increases strongly with fS2 (Clemente et al. 2004). We note that the study of Scaillet et al. (1998) reports values of log fS2 for two runs (i.e., ranging from −3.1 to −1.2), but the influence of fS2 on DSfl/mt was not determined because the fO2 also varied between these runs. In general,
260
Webster & Botcharnikov
fS2 was not constrained for most of the investigations addressed above, and future experimental investigations should explore the influence of this parameter. Felsic melts — S-H2O-CO2. The influence of CO2 on DSfl/mt can be evaluated through comparison of the results for C-free runs of Keppler (1999, 2010) with observations on C-bearing runs of Webster et al. (2009; 2010) and Webster et al. (in press) (Fig. 2). All of these experiments involved fluid-saturated, subaluminous to peraluminous haplogranitic melts, and the former C-free runs were conducted at 200 MPa and 850 °C with log fO2 controlled by three external buffer conditions (e.g., Co-CoO and NNO+0.5 to NNO+1). The latter were run at 200 MPa, 895 to 912 °C, and log fO2 of NNO-0.4 to NNO+1.5, and most runs included CO2 but several C-free runs were conducted as well. In the studies of Webster et al. (2009) and Webster et al. (in press), the concentration of S in the run-product fluids was determined by mass balance, the CO2 concentration of most fluids was constrained by capsule freeze-and-puncture methods, and H2O in fluids was constrained by capsule heating and resulting weight-loss. For all runs, no sulfides were present, all S added to the starting charge apparently dissolved during the runs, and the mass of fluid in each run was sufficiently large such that the results of the mass-balance computations were robust and comparable. The S concentrations of the CO2-bearing and CO2-free melts are consistent for both sets of 200-MPa data (Fig. 2). The S concentrations of the CO2-free melts of Keppler (1999; 2010) range from <0.01 to 0.14 wt%, while those of the CO2-free si-mt of Webster et al. (in press) range from 0.02 to 0.08 wt% and those of the CO2-bearing runs of Webster et al. (in press) range from < 0.01 to 0.11 wt%. The log fO2 of most of the runs of Webster et al. (in press) varies from
Figure 2. Plot of the concentrations of S in fluids versus those in haplogranite melts at 200 MPa modified after Webster et al. (in press). Sulfur partitioning between fluid and melt varies strongly with oxygen fugacity (Keppler 2010). The consistency in S concentrations in melt and fluid of 900° C runs with and without CO2 suggest that S solubility in CO2-bearing fluids (with up to 50 mol% CO2 in fluid) is equivalent to that in aqueous (CO2-free) fluids (within the associated 1s precision). Data for the 900 °C experiments for O-H-S±C fluids are from Webster et al. (in press) and for the 850 °C runs with O-H-S fluids are from Keppler (2010).
Sulfur Distribution between Melts and Fluids in Magmatic Systems
261
NNO+0.1 to NNO+0.9, and they can be compared with the 200-MPa runs of Keppler (2010) conducted at Co-CoO to NNO+1. Moreover, the DSfl/mt of the former experiments ranges from 50-311, and DSfl/mt for the latter vary from 47-468. DSfl/mt for the CO2-bearing runs of Webster et al. (in press) also decreases with increasing fO2 as shown for the CO2-free runs of Keppler (2010). Within the precision shown, DSfl/mt does not appear to change as a function of the CO2 concentration of the fluids, even with as much as 50 mol% CO2 in the fluid. This observation for haplogranite melts is consistent with results of the single, exploratory C-O-H-S run involving dacite melt and an aqueous starting fluid containing 10 mol% CO2 mentioned previously (Scaillet et al. 1998). This CO2-bearing run (Scaillet et al. 1998) displayed no detectable influence of CO2 on DSfl/mt as compared with other CO2-free runs at the same P-T-fO2 conditions. The effect of CO2 on DSfl/mt for intermediate-silica content and mafic si-mt awaits experimental exploration.
Rhyodacitic melts — S-H2O-Cl Multi-component fluid – si-mt equilibria were investigated with a synthetic rhyodacitic melt corresponding to the groundmass composition of magmas erupted from Unzen volcano in 1991-1995 (Botcharnikov et al. 2004). The partitioning of S between rhyodacitic melt and a mixed fluid composed of H2O, S, and Cl was investigated experimentally at 850 °C and 200 MPa under NNO redox conditions. The solid experimental products consisted of quenched glass with or without small amounts of amphibole, orthopyroxene, plagioclase, and pyrrhotite. The solid phases were analyzed by EPMA, and the composition of the fluid was determined by mass-balance methods. The fS2 of the experiments was constrained from the compositions of pyrrhotite, and log fS2 was in the range from −2.4 to −0.9. Although the bulk concentrations of Cl in the system varied in a wide range, the chosen experimental conditions provided stability of a single fluid phase in all runs. Nevertheless, the partitioning of Cl between silicate melt and fluid (DClfl/mt) showed a strong deviation from ideal behavior indicating that conditions were very close to vapor-hypersaline liquid fluid immiscibility. The solubility of S in the melt is a complex function of the activity of fluid components, the FeOtotal content of the melt, and the speciation of S. With increasing S content in the system from 0 to 2 wt%, the amount of pyrrhotite increased (up to 2.2 vol%), the FeOtotal content of the melt decreased (from 2.8 to 0.9 wt%), and S solubility in the melt increased slightly (from 0.01 up to 0.02 wt%). The data indicate that the addition of Cl to the system may decrease DSfl/mt as illustrated in Figure 3, although the data scatter is quite large. Although the observed decrease in DSfl/mt was almost within the uncertainty of the experimental data, Botcharnikov et al. (2004) argued that the influence of Cl on DSfl/mt might be related to non-ideal mixing in the fluid, in particular to the increase in activity of S-bearing fluid species and that it could also be caused by changes in melt composition due to extraction of cations. It must be noted that increasing concentrations of S, in turn, have a detectable and positive effect on DClfl/mt in rhyodacitic systems, and this is analogous to that observed in phonolitic and basaltic systems (see below). Both the direction and magnitude of the observed correlations are comparable between basaltic (Beermann 2010), phonolitic (Webster et al. 2009), and more evolved (Botcharnikov et al. 2004) systems. The maximum effect of S (at saturation with a S-bearing phase) to increase DClfl/mt at studied conditions is ca. 25 to 30 rel%.
Phonolitic melts — S-H2O-Cl Volatile partitioning between fluids and melts, and anhydrite solubility in melts have been determined experimentally for alkali-enriched trachytic to phonolitic melts-H2O-S-Cl at 200 MPa and 896 to 1022 °C (Webster et al. 2009). The (S6+/S6++S2−) of the run-product glasses ranges from ca. 0.5 to 1 for those glasses analyzed by EPMA for their S Ka X-ray wavelength shifts, and this indicates that log fO2 ranges from ca. NNO+0.5 to NNO+1.6. The fS2 of the experiments was not constrained. All melts were saturated in either vapor or hypersaline
Webster & Botcharnikov
262 2000
Rhyodacite 850°C, 200 MPa, NNO
Ds, S fluid / S melt
1500
1000
500
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Cl concentration in melt, wt% Figure 3. Plot of the experimentally determined partition coefficient for S, DSfl/mt, (wt% S in fluid/wt% S in melt) for rhyodacitic melt as a function of Cl concentration in the melt in the experiments at 200 MPa, 850 °C, and under redox conditions corresponding to NNO oxygen buffer (after Botcharnikov et al. 2004). The large uncertainties in some DSfl/mt values are due to significant errors in the mass-balance calculations for the systems containing different mineral phases, including pyrrhotite. Precision shown is 1s.
liquid or vapor plus hypersaline liquid, and the S contents of these fluids were determined by measuring the weight loss of anhydrite crystals, due to anhydrite dissolution in the fluid(s), and by applying mass-balance calculations for S sequestration by the melt. Partitioning of S and other volatiles between melt and fluid(s). Sulfur is preferentially incorporated in the fluid(s) for all experiments, and DSfl/mt for most runs ranges from 40 to 300 although DSfl/mt for two runs exceed 500 (Webster et al. 2009). The values of DSfl/mt increase with decreasing T, the (S6+/S6++S2−) of the si-mt, and the CaO content of the si-mt. The (S6+/S6++S2−) of the si-mt decreases with fO2 through the log fO2 range of NNO+1.6 to NNO+0.5 (Fig. 4), and hence, DSfl/mt varies with fO2. The values of DSfl/mt for these Cl-bearing experiments are generally consistent with values of DSfl/mt determined in prior research involving Cl-bearing felsic melts (Botcharnikov et al. 2004; Scaillet and Macdonald 2006). Importantly, values of DSfl/mt for these experimental studies are consistent with those observed in Cl-free systems (Keppler 1999, 2010; Clemente et al. 2004) indicating that the presence of Cl does not influence S partitioning strongly. TheFig.correlation of DSfl/mt with melt composition is related to the dissolution of S in the 3. Webster and Botcharnikov si-mt (Baker and Moretti 2011, this volume). In particular, Webster et al. (2009) observed that the S concentration of the si-mt increases with the CaO content—for similar S contents in the coexisting fluid(s). Given this correlation and because of the dominance of SO42− in the oxidizing melts of most of these experiments, they hypothesized that SO42− species may be chemically associated with Ca2+ in the si-mt. Their interpretation follows similar observations in other studies (Jantzen et al. 2004; Parat and Holtz 2004; Li and Ripley 2009; Parat et al. 2008; Beermann 2010), but it should be noted that these correlations do not provide confirmation of the stability of Ca2+-SO42− species in comparatively oxidized Ca-bearing si-mt. Future research should test this hypothesis through spectroscopic analyses of S-bearing glasses and si-mt. It follows that for magmas undergoing differentiation, S partitioning between si-mt and fluid(s) will change depending on the extent of chemical evolution of the si-mt (Fig. 5). For
Sulfur Distribution between Melts and Fluids in Magmatic Systems
263
Figure 4. Plot of the concentrations of S in O-H-S-Cl fluids versus S contents of phonolitic to trachytic melts at 200 MPa modified after Webster et al. (2009). Open symbols represent melts with < 2.5 wt% CaO and (S6+/ S6++S2−) of 0.55 to 0.67 at 896-927 °C, closed squares signify melts with ≥ 2.5 wt% CaO and (S6+/S6++S2−) of 0.68-0.94 at 896-1022 °C. The line is an interpretive fit to the latter set of data. Precision shown is 1s.
Figure 5. Plot of the experimentally determined partition coefficient for S (wt% S in fluid/wt% S in melt) at 200 MPa and two temperature ranges (896-927° and 984-1022 °C) for trachytic to phonolitic melts and one or two, O-H-S-Cl-bearing fluid phases versus the Larsen Index for melt evolution (Eqn. 7) modified after Webster et al. (2009). Sulfur partitions more strongly in favor of aqueous fluids coexisting with chemically evolved felsic melts relative to fluid in equilibrium with less-evolved melts. Precision shown is 1s.
264
Webster & Botcharnikov
example, as si-mt evolve from intermediate- to high-silica compositions (i.e., with increasing values of the increasing Larsen Index) where on a wt% basis: Larsen Index = [(0.333.SiO2+K2O) − (CaO+MgO+FeO)]
(7)
the values of DSfl/mt increase. These experiments also display negative correlations between the S and Cl contents of the fluid(s)-saturated si-mt in agreement with the data of Botcharnikov et al. (2004) and Beermann (2010). Constraints on fluid compositions. To determine the S contents of the Cl-bearing aqueous fluids that were saturated with respect to phonolitic and trachytic melts at 200 MPa and ca. 920 °C (Webster et al. 2009), the solubility of anhydrite was determined as a function of the NaCl and KCl contents of the fluids (Eqn. 8). The resulting solubility relationship is: mCaSO4 = 71.52(X(Na,K)ClCaSO4−free fluids) − 1.61
(8)
where X(Na,K)ClCaSO4−free fluids is the mole fraction of NaCl + KCl in the starting fluids of the experimental charges (before any CaSO4 dissolved into the fluids) and mCaSO4 is the molality (i.e., the moles of anhydrite dissolved per kilogram of solvent) of anhydrite in the fluid at run conditions. The mass-balance calculations show that the fluids contained elevated concentrations of Na , K+, Ca2+, Fe3+, Fe2+, Cl−, and S. The presence of these components in experimental fluids returns our attention to the issue of how S speciates in experimental and geologically relevant fluids at oxidizing conditions. For hot volcanic gases at ambient P, the primary oxidized S species identified spectroscopically and confirmed thermodynamically are SO2 and SO3 (Einaudi et al. 2003; Oppenheimer et al. 2011, this volume). It is well established for high-T magmatic fluids at elevated P that the dielectric constant, e, for H2O has a very low value (i.e., e < 5) (Seward 1981 and references therein), and at these conditions, the majority of the constituents dissolved in aqueous solutions occur as chemically associated species. For oxidized fluids at magmatic T and elevated P, the dominant aqueous S species identified thermodynamically and spectroscopically include SO2, SO4−aq, and HSO4−aq (Newton and Manning 2005; Schmidt 2009; Jahn and Schmidt 2010). In this regard, Jahn and Schmidt (2010) determined MgSO4°aq species in H2O-MgSO4 solutions with in situ Raman spectroscopic study at 500 °C and elevated P. Given the stability and presence of MgSO4°aq and HSO4−aq, it is conceivable that charged NaSO4−aq and KSO4−aq as well as neutral, chemically associated Na2SO4°aq, K2SO4°aq, and other SO42−-based associated species may be dissolved in such fluid(s). In a recent study of andesites and dacites from Yanacocha, Peru, Chambefort et al. (2008) observed wormy inclusions of anhydrite (that contain vapor bubbles) in amphibole phenocrysts, and the authors interpreted the anhydrite to potentially represent the quenched products of a separate CaSO4-rich liquid phase. This observation is consistent with the presence of sulfate blebs in the run products of oxidized S-bearing basalt melts fused at 1300 °C and 1 GPa (Jugo et al. 2005a) and basaltic and andesitic melts synthesized at 1050 °C and 0.2 GPa (Beermann et al. 2011; Botcharnikov et al. 2011). Moreover, the previously described capacity of S in oxidizing fluids to extract alkalis from coexisting melts (and potentially render the si-mt peraluminous) implies that SO42− ions are available for association with alkalis in such fluids. The alternative interpretation, i.e., that the alkalis are associated with neutral SO2 species in the fluid(s), is less likely. Consequently, Webster et al. (2009) interpreted their relatively oxidizing experiments to include saline fluids comprised of Cl−- plus SO42−-enriched hydrosaline liquids, and at more reducing conditions, such fluid(s) would contain H2S and lesser HS− ± S2− and S2 (Holloway 1987; Newton and Manning 2005). It is conceivable that some aqueous iron-sulfide species, and potentially, aqueous alkalisulfide species may also have been present. These suggested chemical associations and resulting species, however, require proper identification by appropriate spectroscopic techniques. Such data are important not only to volcanic systems, but to mineralizing hydrothermal systems as well (Simon and Ripley 2011, this volume). +
Sulfur Distribution between Melts and Fluids in Magmatic Systems
265
Andesitic melts — S-H2O±CO2±B Data on the partitioning of S-bearing, multi-component volatiles between fluids and simt of intermediate composition are scarce. Gorbachev (1990) reported results on S behavior in B-bearing andesitic melts equilibrated with H2O-CO2 fluids, and he noted that S solubility is distinctly greater in B-bearing melts than in B-free melts. The preliminary experimental approach of Teague et al. (2008) focused on the partitioning of S between H2O-bearing but CO2-dominated vapor and molten Mt. Hood andesite at 1250 to 1300 °C and 200 to 800 MPa. The obtained results show that DSfl/mt ranges from approximately 200 at 800 MPa to 2000 at 200 MPa, demonstrating that, prior to saturation of the silicate melt with a S-rich phase, S displays an extreme affinity to the CO2-rich fluid phase The experiments of Parat et al (2008) with molten Huerto andesite (San Juan volcanic field, Colorado) and H-C-O-S-bearing fluids showed that the addition of S to the system decreases the proportions of H2O and CO2 in the fluid phase and can affect the crystallization of andesitic magmas. However, no data on DSfl/mt are provided, mainly because of the difficulty to perform mass-balance calculations for these highly crystallized systems containing poorly constrained fractions of S-rich minerals and S-bearing apatite crystals (the partitioning of S between silicate melt and minerals is discussed in detail by Parat et al. 2011, this volume). Recent work of Moune et al (2009) reported the solubility of S in basaltic andesites (as well as in basalts) at 1050 °C, 300 MPa, and relatively reducing conditions with log fO2 of FMQ-0.3 to FMQ+1.1. Using these experimental data and additional data on the amounts of added glass (Moune, personal communication) we have calculated the values DSfl/mt for the systems that were undersaturated in pyrrhotite. The DSfl/mt varies in the range from about 6 to 816 and shows an obvious correlation with the concentration of H2O dissolved in the melt at redox conditions varying in a very narrow range (with log fO2 from FMQ-0.3 to FMQ+0.4 for pyrrhotiteundersaturated systems) as illustrated in Figure 6. The explanation for this relationship needs further experimental investigation.
Basaltic melts — S-H2O±CO2-Cl Basaltic magmas are the most difficult systems to investigate with S experimentally. Firstly, the range in melting T of basalts readily exceeds the melting T of gold (i.e, for melts with low aH2O). In addition, the high T and relatively reduced redox conditions of mafic magmas readily provide the opportunity for S to react with noble metal capsules which may lead to failure of the experiment. Hence, there are only a few experimental studies on S solubility in basaltic melts available in the literature for two-volatile systems (Jugo et al. 2005a; Liu et al. 2007; Moune et al. 2009; Beerman et al. 2011; Botcharnikov et al. 2011), and up until recently there were no systematic studies reporting S partitioning between mafic si-mt and fluid phases. The recent work of Lesne et al. (2011) focused on the partitioning of multi-component H2O-CO2-S-Cl volatiles between synthetic basaltic melts of Masaya and Stromboli magmas and coexisting fluids. The experimental approach was devoted to simulating closed-system degassing processes of volatile-rich magma undergoing decompression from 400 to 25 MPa. The initial bulk amounts of volatiles (in particular, CO2) in the experimental charges exceed those of volatile-saturated basaltic and alkali-rich basaltic magmas at 400 MPa (Shishkina et al. 2010; Vetere et al. 2011), and hence, all systems were composed of basaltic melt and multicomponent fluid. The experimental results of Lesne et al. (2011) are illustrated in Figure 7 and indicate a very rapid and almost identical quantity of CO2 loss from basaltic melts in all experiments as P decreases from 400 to 25 MPa (Fig. 7a), whereas the decrease in H2O content is not so dramatic (Fig. 7b). Interestingly, the concentration of Cl remains constant and independent of P within the uncertainty of all experiments (Fig. 7c). Moreover, the results show that the
Webster & Botcharnikov
266 1000
Basaltic andesites
800
Ds, Sfluid / Smelt
1050°C, 300 MPa -0.3
Basalts
900
700 600 500 400 300 200 MPa
200 100 0
0
1
2
3
4
5
6
7
8
H2O in melt (wt%) Figure 6. The relationship between DSfl/mt (wt% S in fluid/wt% S in melt) and water content in basaltic and basaltic-andesite melts of experiments conducted at 1050 °C and 300 MPa (one run in basaltic-andesite system was conducted at 200 MPa) and under redox conditions in the range of log fO2 from FMQ-0.3 to FMQ+0.4. The DSfl/mt values are calculated for pyrrhotite-undersaturated systems using the data reported by Moune et al. (2009) and using additional information on mass proportions (mg of added glass) provided by S. Moune (personal communication). Note significant differences in DSfl/mt between H2O-rich and H2Opoor systems. Precision shown is 1s.
behavior of S in the system which is not saturated with any S-bearing melt or mineral phase is not significantly influenced by the composition of released fluid during decompression to about 100 to 150 MPa for all basaltic systems enriched or poor in bulk S (Fig. 7d). The P decrease from 150 to 25 MPa leads to a dramatic drop in S content of the si-mt to very low values with concomitant enrichment of the fluid phase with S. This, in turn, causes a significant decrease in the (S/Cl) ratio of the si-mt (Fig. 7e) indicating that shallow degassing of basaltic magma should produce fluids with high (S/Cl) ratios. The DSfl/mt for these systems with multicomponent fluids can increase from values <10 at high P to values >1000 as P decreases from ~100 to 25 MPa, indicating that S is preferentially partitioned into the fluid (Fig. 7f). The fluids released from the Masaya basaltic melt, which have lower S contents compared with those of Stromboli, characterized by higher enrichment in S relative to the fluids of the Stromboli Fig. 6.are Webster and Botcharnikov basalt experiments at similar conditions. This observation indicates that DSfl/mt is affected even by small variations in the composition of basaltic systems. The constant major-element composition of the si-mt and the absence of S-bearing phase(s) at high T and oxidizing conditions of the experiments indicate that the observed behavior of S should be attributed to changes in gSsi-mt and/or gSfl . The decrease in S concentration with decreasing concentration of H2O in the melt under constant fO2 and P has been reported in experimental studies of basaltic systems (Liu et al. 2007 and Moune et al. 2009, Fig. 6), although it was not confirmed by any systematic work. These observations indicate that H2O and S interact in magmatic phases and presumably change DSfl/mt. An increase in DSfl/mt can be also related to changes in the properties of the fluid phase, because the proportions of H2O and CO2 as major components of the fluid progressively increase with decreasing P as shown in the work of Lesne et al. (2011) for closed-system conditions. The very good agreement between the experimental results and the compositions of melt inclusions from natural rocks as reported by Lesne et al. (2011) indicates that the determined DSfl/mt values are not an experimental artifact.
Sulfur Distribution between Melts and Fluids in Magmatic Systems 2000
5 Stromboli A
1800
CO2 in the melt, μg/g
H2O in the melt, wt%
Stromboli B
1600
Masaya A
1400
Masaya B
1200 1000 800 600 400
(a)
200 0
100
200
300
Pressure, MPa
400
4 3 2 1
(b)
0
0
500
0
3000
100
200
300
Pressure, MPa
400
500
3200
2500
2800
S in the melt, μg/g
Cl in the melt, μg/g
267
2000 1500 1000 500
2400 2000 1600 1200 800 400
(c)
0
(d)
0 0
100
200
300
400
0
500
100
Pressure, MPa
200
300
400
500
Pressure, MPa
3
10000
1000
D, Sfluid / Smelt
S / Cl in the melt
2.5 2 1.5 1
100
10
0.5
(f)
(e)
0 0
100
200
300
Pressure, MPa
400
500
1 0
100
200
300
Pressure, MPa
400
500
Figure 7. Plots showing the closed-system evolution of volatile composition in synthetic basaltic melts relevant for magmas of Stromboli and Masaya volcanoes at 1150 °C and at redox conditions in the range from FMQ+1.7 to FMQ+3.1 as a function of pressure and initial concentrations of volatiles (modified after Lesne et al. 2011): (a) Botcharnikov CO2, (b) H2O, (c) Cl, and (d) S. The variation of (S/Cl) ratio in the melts is shown in Fig. 7. Webster and panel (e), whereas the DSfl/mt (wt% S in fluid/wt% S in melt) illustrated in panel (f). Precision shown is 1s.
However, the quantification and correct interpretation of the observed effect require additional future experimental study. It must be noted that the presence of small amounts of S in these basaltic systems and the partitioning of S in favor of the fluid at low P did not cause large changes in the behavior of other volatiles. The evolution of H2O and CO2 content with P is dictated largely by typical
268
Webster & Botcharnikov
partition mechanisms observed for S-free, H2O-CO2-basaltic melt systems (Dixon et al. 1995; Botcharnikov et al. 2005; Shishkina et al. 2010; Vetere et al. 2011). Since the bulk amount of Cl was quite low, the concentration of Cl in the melt remained almost constant over the entire P range leading to significant increases in the S/Cl ratio in the released fluids at low P. The low or negligible Cl content of the fluid in the investigated systems provides no apparent effect on the mixing properties of the fluid, and hence, on the partitioning of S and other volatiles (as observed for silicic systems with high-Cl concentrations, as described previously). Another recent study focused on a more classical experimental approach where a Mt. Etna trachybasaltic melt was equilibrated with excess fluid containing H2O, S, and Cl (Beermann et al. 2009; Beermann 2010). The experiments were conducted in an internally heated pressure vessel at 1050 °C and P of 100 and 200 MPa. The redox conditions inside the capsules varied from log fO2 of FMQ+0.7 to FMQ+3.2. The initial bulk concentrations of S and Cl in the system varied from 0.2 to 2 wt% and from 0.06 to 3.5 wt%, respectively (S was added as elemental sulfur and Cl as HCl solution with 5 to 30 wt% Cl). The data (Beermann 2010) show that at oxidizing conditions of FMQ+3.2, both P and Cl concentration of the system significantly decrease (P by approximately 20-25 rel%, whereas Cl by up to ~ 40 rel%) the S content of the si-mt resulting in increasing DSfl/mt. The DSfl/mt are always > 1 and can reach values of 35-40 and 70 at 200 and 100 MPa, respectively, depending on the amount of Cl in the system. The highest DSfl/mt values (up to 240) are observed for reduced systems at ~FMQ+0.7. The data indicate that increasing the concentration of Cl in the melt from 0.05 to ~3 wt% can increase DSfl/mt from ~160 to 240 under reducing conditions. However, it must be kept in mind that mass-balance calculations of DSfl/mt for systems containing sulfide phases can involve large uncertainties due to difficulties with determination of the extent of Fe loss and with quantification of the sulfide phase abundance. Nevertheless, the observed changes in DSfl/mt and the effect of Cl on S partitioning can be attributed to the possible change in activity coefficients of different S species in coexisting si-mt and fluids. The occurrence of different S species depends on magmatic conditions, in particular, on fO2 and on the phase compositions. For instance, the extraction of cations by Cl- and S-bearing aqueous fluids may have an effect on the composition of S- and Cl-bearing components of the fluid and hence on their activity coefficients. As shown by Beermann (2010), the change in S speciation from sulfide to sulfate leads to a significant extraction of Ca (as well as Na and K) by S from basaltic melt into the fluid phase with increasing fO2. Moreover, the efficiency of cation extraction increases with increasing Cl content of the fluid leading to possible variations in fluid and si-mt properties and changes in gS and gCl. On the other hand, at relatively reduced conditions the extraction of cations from the si-mt is significantly less, even in FeS-saturated systems, indicating that the mutual effects of volatiles are strongly dependent on the cation load of the fluid and complexing of volatile species with cations in coexisting phases. Beermann (2010) has also observed, as previously recorded for other systems, that the addition of S to the system increases DClfl/mt by approximately 25 rel% at oxidizing conditions but the effect of S on Cl partitioning is very small to negligible at relatively reduced conditions (Beermann 2010). Such differences may result from the more effective extraction of cations by oxidized S-bearing fluids. The reported correlation between the concentrations of dissolved S and cations in Clbearing and CaSO4-saturated si-mt is indicative of significant enrichment of S and cations in the fluid phase and is consistent with the experimental data on the solubility of CaSO4 in alkalichloride aqueous fluids (Newton and Manning 2005) and in Cl-bearing fluids coexisting with trachytic and phonolitic si-mt (Webster et al. 2009). The effect of Cl is theorized to reflect the change in S speciation and formation of alkali-sulfate components at the expense of the Casulfate component in the fluid according to the equilibrium equation:
Sulfur Distribution between Melts and Fluids in Magmatic Systems
269
2CaSO4 + NaCl + KCl (+H2O) = 2CaCl+aq + NaSO4−aq + KSO4−aq (+H2O)
(9)
as proposed by Webster et al. 2009 (see also Newton and Manning 2005). Furthermore, the increase in S partitioning into the alkali- and Cl-rich fluids implies that alkali-sulfate fluid components have lower activity coefficients relative to that of the Ca-sulfate component. It must be emphasized that recent experimental data for basaltic systems clearly illustrate that the partitioning of individual volatile components between basaltic melts and exsolving fluids is considerably influenced by the interrelated effects of volatiles and by properties of multi-component (volatile- and cation-bearing) aqueous fluids under given T, P, and redox conditions. Depending on the relative volatile abundances and magmatic conditions in basaltic systems, an increase in the Cl content of the system can enhance S partitioning into the fluid by approximately 30 to 40 rel%, whereas S is able to enhance Cl exsolution from the si-mt by ca. 25 to 30 rel%. Conversely, the experimental data also show that if the degassing processes occur in the magmas with S and Cl concentrations far from saturation with S- or Cl-bearing phase(s), the exsolution of significant S and Cl can be delayed to P of 150 to 100 MPa. Moreover, with further decompression, the S/Cl ratio in the fluid can increase dramatically due to much more effective partitioning of S in favor of the fluid.
Summary on S partitioning between fluids and rhyolitic to basaltic melts at crustal conditions The hydrothermal S partitioning experiments reviewed herein were conducted with rhyolitic to basaltic melts, at 50 to 800 MPa, and magmatic T, and they show DSfl/mt ranging from ca. 1 to 2800 (Table 1). Sulfur partitioning depends on and shows variable responses to changes in fO2 (and the valence state of S), fS2, melt composition, fluid composition, T, and P. It has been demonstrated for a variety of si-mt that the partitioning of S between si-mt and fluid varies with si-mt composition, and hence, DSfl/mt should vary with magma evolution (Scaillet and Pichavant 2003; Botcharnikov et al. 2004; Webster et al. 2009; Beermann 2010). For P > 100 MPa, DSfl/mt shows a general decrease with change in magma composition from basaltic to rhyolitic as illustrated in Figure 8, indicating that fluids associated with silicic magmas should be more enriched in S relative to melts as compared with basaltic systems. Furthermore, it also evident that, for a given composition, systems under relatively reduced conditions (log fO2
Webster & Botcharnikov
270
(a)
Ds, S fluid / S melt
1000
100
Relatively reduced < NNO
10 Relatively oxidized > NNO
1
0
1
2
3
4
5
6
7
MFM = (Na+K+2(Ca+Mg+Fe2+))/(Si+Al+Fe3+) [mole fractions]
(b)
1000
Ds, S fluid / S melt
8
100
10 Mafic magmas 1050-1150°C 100-400 MPa
Felsic magmas 750-1020°C 50-300 MPa
1 -10 0 10 20 30 Larsen Index = (0.333SiO2+K2O)-(CaO+MgO+FeO) [wt%] Basalt >NNO, B2010 Basalt
NNO, L2011* Basalt
Phonolite >NNO, W2009 Haplogranite >NNO, K2010 Haplogranite NNO, Wi.p. Rhyolite NNO, S&M2006
Figure 8. Plot (a) shows the DSfl/mt (wt% S in fluid/wt% S in melt) versus the MFM structural parameter of Fig.8 as Webster and Botcharnikov the silicate melt formulated by (Liu et al. 2007) where all cations are expressed in mole fractions. The Fe2+ and Fe3+ values were calculated using the model of Moretti (2005). Solid line in (a) approximately separates data at relatively reduced (≤NNO) and relatively oxidized (>NNO) conditions. Plot (b) shows the DSfl/mt versus the Larsen Index (see also Eqn. 7; the oxides are in wt%). The large range in DSfl/mt reflects the influences of differences in oxygen fugacity, S concentrations of experimental charges, melt and fluid composition, pressure, and temperature; see text for discussion. See legend for definition of system compositions and redox conditions. The data sources for melts include: B2010 as Beermann (2010) for basalts, L2011* as Lesne et al. (2011, *only data at P ≥ 100 MPa are shown) for basalts, M2009 as Moune et al. (2009, pers. comm.) for basalts and basaltic andesites, Bo2004 as Botcharnikov et al. (2004) for rhyodacites, W2009 as Webster et al. (2009) for phonolites and trachytes, K2010 as Keppler (2010) for H-O-S-bearing haplogranites, Wi.p. as Webster et al. (in press) for O-H-S±C-bearing haplogranites and iron-bearing granite, and S&M2006 as Scaillet and McDonald (2006) for alkali-rich rhyolites.
Sulfur Distribution between Melts and Fluids in Magmatic Systems
271
soluble in CO2-rich fluids as in aqueous fluids, but this observation demands further study. This observation is consistent with those of Teague et al. (2008) for CO2-bearing fluids in equilibrium with molten andesite at 200 to 800 MPa and for haplogranites at 200 MPa (Webster et al. in press). It should be noted for P of 1.1 to 2.5 GPa, however, Gorbachev (1990) observed that S solubility is greater in basaltic melts that are saturated in aqueous fluids than that with basaltic melts saturated in H2O-CO2 fluids, so DSfl/mt may be greater for C-O-H-S-bearing fluids than corresponding values of DSfl/mt with C-free fluids at mantle P conditions. The potential influence of Cl on DSfl/mt is important given that Cl exerts strong controls on fluid phase equilibria at shallow crustal P (e.g., Sourirajan and Kennedy 1962; Bowers and Helgeson 1983; Bodnar et al. 1985; Bischoff and Pitzer 1989; Driesner 2007). As an extreme example of this influence, volatile concentrations of fluids and coexisting si-mt can be buffered at fixed values due the influence of the aCl on the stability of vapor plus hypersaline liquid at sub-solvus P-T conditions (see Liebscher and Heinrich 2007; Webster and Mandeville 2007 for elaboration). The effect of Cl on DSfl/mt is not well determined, however. Research on fluid-saturated rhyodacitic si-mt (Botcharnikov et al. 2004) demonstrates that the addition of Cl increases the solubility of S in si-mt only marginally at 200 MPa and, hence, it reduces DSfl/mt. Conversely, experiments with basalts show Cl addition significantly enhances S partitioning in preference of fluid at reducing conditions (Beermann 2010). And in contrast with both of these studies, experiments with fluidsaturated phonolitic to tephritic si-mt show no clear relationships between the Cl content of the system and DSfl/mt. These discrepancies may be attributable to the differences in the bulk composition of the systems and to subtle differences in fO2. The experimentally constrained DSfl/mt values are generally consistent with values of calculated for basaltic and andesitic-rhyolitic si-mt. Scaillet and Pichavant (2003) used the modified Redlich-Kwong (MRK) equation of state to constrain S in the fluid for the C-OH-S system and determined the S contents of coexisting si-mt from silicate melt inclusion compositions. This study also models how DSfl/mt varies with P, T, and fO2, and in particular, these data indicate that DSfl/mt increases with decreasing P. Also for comparison, other investigators have attempted to calculate values of DSfl/mt for basaltic magmas by application of hypothetical degassing behavior to the compositions of melt inclusions (Sisson and Layne 1993; Métrich et al. 2001; Wade et al. 2006; Vigoroux et al. 2008; Wallace and Edmonds 2011, this volume), and the calculated values of DSfl/mt range from 5 to 100 in agreement with experimentally derived values. Thus, although experimental study of S-bearing systems is both complicated and challenging, experimentally determined values of DSfl/mt are consistent with these few, other constraints on S partitioning behavior. DSfl/mt
APPLICATION OF THE EXPERIMENTAL DATA TO PROCESSES OF FLUID EXSOLUTION AND THE EVOLUTION OF MAGMA AND MAGMATIC FLUIDS The compositions of volcanic gases and volatile components in melt inclusions and matrix glasses provide important constraints on the behavior and budget of S in magmatic systems. In particular, the fluxes of S and the relationships between S and other volatiles in subductionrelated magmas are indicators of the conditions of magma generation and serve as proxies for the recycling of S in subduction zones (Wallace and Edmonds 2011, this volume). However, the interpretation of measured fluxes and volatile ratios requires understanding of the extent to which different magmatic sources, conditions, and processes contribute to the geochemical behavior of S. The behavior of S can be significantly influenced by the volatile ratios inherited from the magma source. According to experimental data reviewed in this chapter, the most significant effects are expected for magmas generated at shallow subduction depths, i.e., where DSfl/mt
272
Webster & Botcharnikov
resembles the most significant variations with P. Further evolution of S behavior will be governed by conditions and processes such as crustal contamination (gain or loss of S), magma differentiation (increase in S concentration due to incompatible behavior or segregation of Sbearing phases), magma degassing (exsolution of S from the melt) and associated changes in fO2 and magmatic/hydrothermal interaction (remobilization of S) (Burgisser and Scaillet 2007; Burgisser et al. 2008). Each of these processes can change the entire budget of S en-route to the surface and the ratios of S to other volatile components in volcanic gases which are a direct manifestation of and indicators of magmatic degassing processes.
Magmatic gas composition as an indicator of magma and volcanic degassing activity Analyses of volcanic gas compositions show a wide range of concentrations and proportions of volatile components as a function of tectonic setting, magma storage conditions, and the evolution of magmatic systems (e.g., Symonds et al. 1994; Williams-Jones and Rymer 2000; Delmelle and Stix 2000; Aiuppa et al. 2009a). Moreover, the observed variations in ratios of volatile constituents may reflect changes in magmatic activity and efficiency of fluid exsolution at various depth conditions. For instance, correlations between the (S/C) and (S/Cl) ratios in volcanic gases and eruptive activity have been observed at several volcanoes (Aiuppa et al. 2002, 2004, 2009b; Burton et al. 2007; Edmonds and Gerlach 2007; Christopher et al. 2010) indicating that these ratios are closely linked to the conditions of magma degassing. Hence, these ratios were used to estimate the depths of magmatic sources responsible for passive degassing and for strombolian explosions illustrating that small explosive events typically occur at shallower depths than larger explosions (Burton et al. 2007). The estimations were based on the modeling of volcanic gas composition as a function of P or depth of magma degassing. However, the models used in the studies were calibrated against a very limited set of data bearing on volatile partitioning between fluids and si-mt. In particular, the data sets are limited for S behavior in multi-component volatile-bearing systems. The first systematic data on basaltic systems for a wide range of pressures (Lesne et al. 2011) should allow for quantitative testing on whether the previous assumptions on S partitioning were correct or not. Figure 9 presents the variation in molar (S/C) and (S/Cl) ratios in fluids exsolving from basaltic magma of Stromboli volcano undergoing decompression from 400 to 25 MPa. This chemical evolution of fluid composition with P represents closed-system degassing of basaltic magmas. Interestingly, although both ratios show some scatter they also show a pronounced minimum at about 150 to 200 MPa, indicating that similar volatile ratios can be observed in fluids released from both deep and shallow magmas (Fig. 9 a,b). Thus, these experimental results imply that, at least for the investigated condition of closed-system degassing behavior, the ratio of S to these other volatiles cannot be used as an ultimate constraint on the P of magmatic degassing. Nevertheless, these new experimental data do provide useful constraints on magma degassing conditions. The horizontal lines in Figure 9a and b correspond to (S/C) and (S/Cl) ratios measured in volcanic gases released during quiescent degassing and during explosive activity of Stromboli as measured using remote sensing techniques in 2002 (Burton et al. 2007) and in 2007 (Aiuppa et al. 2009b). The typically low volatile ratios of explosive events correspond to minimum values determined experimentally and can be interpreted as magma degassing at P of about 200 MPa. On the other hand, the ratios measured during quiet periods of volcanic activity do not explicitly point to any particular P of fluid release, except probably the very high (S/C) ratios measured in 2007 by Aiuppa et al. (2009b). The measured (S/C) value of 0.279 corresponds to a P of about 50 MPa. These estimates which are based on experiments with geologically relevant, multi-component systems point to higher P of gas release than that predicted by the models of Burton et al. (2007) and Aiuppa et al. (2009b). Moreover, owing to the lack of experimental data for open-system degassing, this is the only quantitative estimation of fluid exsolution behavior indicating that only relative variations in fluid composition can be used as a measure of volcanic activity change.
Sulfur Distribution between Melts and Fluids in Magmatic Systems
Molar ratio S / C in fluid
1
273
(a) Remote sensing: 2007, A2009 Passive degassing 2002, B2007
0.1
2002, B2007 Eruptions/explosions 2007, A2009 Experiments: Stromboli A; 0.2 wt% bulk S Stromboli B; 0.4 wt% bulk S
0.01 0
100
200
300
400
500
Pressure (MPa)
Molar ratio S / Cl in fluid
1000
(b)
100 Remote sensing:
10
Passive degassing 2002, B2007
1
Eruptions/explosions
0.1 0
100
200
300
400
500
Pressure (MPa)
Figure 9. Plots showing the variations in molar ratios: (a) S/C and (b) S/Cl in fluids coexisting with basaltic melts of Stromboli volcano as a function of pressure for closed-system conditions as determined by experiments (modified after Lesne et al. 2011). Circles and squares show the experimental data for systems containing different concentrations of S (see legend and Fig. 7 for details). The horizontal lines represent Fig. ratios 9. Webster and Botcharnikov molar measured using remote sensing techniques during periods of quiescent degassing and during eruptions/explosions: solid and dash-dotted lines in (a) are data after Aiuppa et al. (2009b; labeled as A2009) whereas dashed and dotted lines in (a) are data after Burton et al. (2007; labeled as B2007) for eruptive activity in 2007 and 2002, respectively. The dashed and dotted lines in (b) are data after Burton et al. (2007). Precision shown is 1s.
In addition, since recent experimental studies reviewed herein show that S partitioning is affected by the multi-component nature of magmatic fluids, the correct interpretation of magma degassing and volcanic activity, based on the composition of volcanic gases, requires further quantification via extensive experimental research on geologically relevant, multi-component fluids. This is particularly important because the changes in the regime of magma supply and degassing style (open- versus closed-system conditions) or in magma storage conditions at different depths may cause significant variations in the ratios of S and Cl as well as other volatile components in the magmatic fluids and volcanic gases.
274
Webster & Botcharnikov
Volatile mixing relationships and the influence of S on CO2, H2O, and Cl solubility in melt and new insights on vapor (fluid) saturation in felsic magmas. Water is the most abundant and soluble volatile constituent of magmas at shallow crustal P. The fH2O plays a dominant role in magmatic fluid exsolution, but the finite and low solubilities of C (Holloway 1976; Holloway and Blank 1994; Dixon 1997) and Cl (Webster 1992) in silicate melts can also drive fluid(s) saturation and exsolution in magmas containing far less H2O than that required for saturation in pure H2O. With felsic melts at 200 MPa, for example, CO2 solubility is as much as two orders of magnitude lower than that of H2O, and Cl solubility is a full order of magnitude lower. The behavior of S is less well known, but the emerging experimental data reviewed herein demonstrate complex mixing relationships and solubility behaviors for S, H2O, Cl, and CO2 in melts, and these correlations have fundamental bearing on processes of fluid exsolution. The addition of S to fluid-saturated granitic melts modifies the concentration of CO2 in the si-mt, and this relationship varies dramatically with changes in fO2 (Webster et al. in press). The CO2 concentrations of haplogranite and Fe-bearing granitic melts are reduced by > 50% at comparatively reduced conditions (i.e., log fO2 of NNO-0.4 to NNO+0.3) with as little as 2 wt% S in the fluid (Fig. 10); whereas the CO2 concentration in the si-mt changes little with the addition of S at relatively oxidized fO2 conditions (i.e., log fO2 > NNO+0.3). Moreover, the H2O concentration in the si-mt changes little with added S at these relatively reduced fO2 conditions.
Figure 10. Plot, modified after Webster et al. (in press), of experimentally determined concentrations of H2O and CO2 in granitic melts saturated in C-O-H±S-bearing fluid at 200 MPa and 900° C compared with modeled CO2 and H2O concentrations of S-free, vapor-saturated rhyolite melt at 900 °C and 200, 150, and 100 MPa (dotted, dashed, and solid curves, respectively, based on Newman and Lowenstern 2002). The presence of > 0.02 wt% S in melt and > 3 wt% S in fluid (with log fO2 of NNO-0.4 to NNO+0.3) reduce CO2 solubilities in 200-MPa melts to values as low as those for S-free, C-O-H-fluid-saturated melts at 150 MPa. Precision shown is 1s.
Sulfur Distribution between Melts and Fluids in Magmatic Systems
275
Although not definitive, these experiments also suggest that the concentration of H2O in the si-mt for several CO2-poor runs may increase (as a function of S addition to the system) at comparatively oxidized log fO2 > NNO+0.3 (Webster et al. in press). If correct, this latter observation is qualitatively consistent with prior experimental and theoretical work involving sulfur at relatively oxidizing conditions (Moretti et al. 2003; Botcharnikov et al. 2004) and, hence, requires further experimental investigation. An expected consequence of adding S to a mixed H2O- and CO2-bearing fluid is the reduction of the XH2Ofl and the XCO2fl by dilution. The resulting consequences for the XH2Osi-mt and XCO2si-mt depend on the mixing behavior of all volatiles in the fluid and si-mt as expressed by gifl and gisi-mt for H2O, CO2, and the primary S species. If the mixing behavior is ideal (i.e., gifl and gisi-mt for all volatile species are unity), the addition of S reduces XH2Osi-mt and XCO2simt , and this behavior is the most likely interpretation for the observation of decreasing CO2 concentrations of granitic si-mt at comparatively reduced conditions with S addition (Webster et al. in press). However, adding S did not reduce the H2O content of si-mt. In this regard it is noteworthy that these mixed S-H2O-CO2 runs contained up to 15 wt% (i.e., 6 mol%) S in the fluid, and because this level of dilution by S at relatively reduced conditions did not decrease the XH2Osi-mt measurably and following Equations (1) through (4), the gH2Ofl increased and/ or the gH2Osi-mt decreased because the (XH2Osi-mt gH2Osi-mt) is proportional to the (XH2Ofl gH2Ofl) at equilibrium. In addition, these S-bearing si-mt contained ≤ 10 micromoles of S per mole of simt and so it is less likely that the non-ideal behavior (which is required to explain the absence of reduced H2O contents of the glasses) occurred in the si-mt, but rather in the fluid which thus influenced gH2Ofl. The final observation of Webster et al. (in press), i.e., of no significant decrease in XCO2si-mt with increasing S in the system at oxidizing conditions, can be interpreted by Equation (10): XCO2si-mt gCO2si-mt = XCO2fl gCO2fl
(10)
to suggest that gCO2fl must have increased and/or the gCO2si-mt must have decreased with S addition. These solubility relationships for felsic systems have important implications for geobarometric interpretation of the P and corresponding depth of melt inclusion entrapment and for fluid exsolution in magmas. For example, the application of experimental constraints on H2O and CO2 solubilities for S-free rhyolitic si-mt (e.g., Newman and Lowenstern 2002) to melt inclusions representing S-bearing si-mt, at similar and geologically relevant values of fO2 would predict anomalously higher P of entrapment and/or fluid exsolution than those which are appropriate for C-O-H-S systems (Fig. 10). This is true because the presence of reduced S species shifts the isobaric H2O and CO2 solubility curves for felsic melts to lower CO2 values. Clearly, these relationships demand further investigation to determine how this behavior varies with P, T, and si-mt composition. Moreover, what is the influence of Cl on these solubility relationships? The strong influence of S to reduce Cl solubility in basaltic, phonolitic, trachytic, and rhyodacitic si-mt at log fO2 > NNO (Botcharnikov et al. 2004; Beermann et al. 2009; Webster et al. 2009; Beermann 2010) is geologically significant because the presence of even minute quantities of magmatic Cl may force the exsolution of hypersaline liquid or vapor plus hypersaline liquid from fluid-undersaturated si-mt, and in particular, for relatively anhydrous felsic si-mt (Webster 1992). This behavior is also significant because it applies broadly to melts ranging from 48 to 70 wt% SiO2. Thus, the presence of S in comparatively oxidized si-mt modifies Cl solubility in the si-mt which, in turn, controls fluid phase equilibria, fluid composition, and process of fluid exsolution (Webster et al. 2009), i.e. phenomena that are responsible for the dynamics of magma degassing.
276
Webster & Botcharnikov SUGGESTIONS FOR FUTURE RESEARCH
The volume fraction of pre-eruptive vapor (or other fluids) in magma is one parameter that serves to regulate volcanic eruptive styles, magnitudes, and durations (Wallace and Anderson 2000; Oppenheimer 2003). In addition, magmatic fluids range from low-density vapor to dense saline liquids and the type of fluid that first exsolves and the efficacy of that fluid in implementing eruptive processes vary strongly with fluid composition. The capacity of a multi-component fluid to expand and provide PDV work in the system varies with the partial molar volumes of the individual components and the molar volume and composition of the fluid mixture. For example, the addition of CO2 to an aqueous vapor at 200 MPa and T ≤ 700 °C can increase the partial molar volume of H2O and the molar volume of the mixture (Sterner and Bodnar 1991; Churakov and Gottschalk 2003). Conversely, the addition of NaCl to an aqueous vapor dramatically reduces the molar volume of the mixture at similar conditions (Bodnar 1985; Burnham 1997). The influence of S-dominated vapor species is less well known. Thermodynamic modeling indicates that the addition of H2S and SO2 to aqueous vapor may also increase the molar volume of the mixture (Shi and Saxena 1992), but supporting experimental data are sorely needed. This is particularly true for complex mixtures of C-O-H-S-Cl fluids. Finally, it should be emphasized that the results of experimental studies reviewed in this chapter clearly demonstrate that accurate and quantitative interpretation of natural processes responsible for the partitioning of S between magmatic phases and S budget in multi-component natural systems requires additional extensive experimental work bearing on natural multicomponent systems. The main focus should be the measurement of fluid compositions and the associated partitioning of volatiles in S-C-H-Cl-O systems with systematic variation in P, T, fO2, melt and fluid composition. In addition, new analytical techniques allowing more precise determination of fluid composition are required. Particular attention should also be directed to the behavior of volatile constituents in open versus closed systems at relevant magmatic conditions, because such data are important for the monitoring of changes in volcanic activity based on the composition of volcanic gases and melt inclusions. Another important aspect that should be addressed by future experiments is the kinetics of multi-component volatile exsolution from silicate liquids into the coexisting mixed fluids, which will provide important constraints on the dynamics of magma degassing. And finally, future experimental studies should take into account the saturation of magmas with volatile-rich phases, such as saline liquids, sulfides, sulfates and volatile-bearing silicate minerals, and the role of such phases in the behavior and budget of S in multi-component fluids.
ACKNOWLEDGMENTS We express our appreciation to Jake Lowenstern, Hans Keppler, and Harald Behrens for clear, detailed, and beneficial reviews, and N. Nicholson for assistance in editing. We would like to acknowledge O. Beermann, P. Lesne, and S. Moune for providing us with recent experimental data for basaltic systems in the form of manuscripts and theses that significantly improved the quality and depth of this review paper. Some research addressed herein was supported by National Science Foundation award EAR-0836741 to J.D.W. and by Deutsche Forschungsgemeinschaft (DFG) grants No378/4 and Bo2941/1 to R.E.B.
REFERENCES Aiuppa A, Baker DR, Webster JD (2009a) Halogens in volcanic systems. Chem Geol 263:1-18 Aiuppa A, Federico C, Paonita A, Pecoraino G, Valenza M (2002) S, Cl and F degassing as an indicator of volcanic dynamics: The 2001 eruption of Mount Etna. Geophys Res Lett 29(11), doi:10.1029/2002GL015032
Sulfur Distribution between Melts and Fluids in Magmatic Systems
277
Aiuppa A, Federico C, Giudice G, Gurrieri S, Paonita A, Valenza M (2004) Plume chemistry provides insights into mechanisms of sulfur and halogen degassing in basaltic volcanoes. Earth Planet Sci Lett 222(2):469483 Aiuppa A, Federico C, Giudice G, Giuffrida G, Guida R, Gurrieri S, Liuzzo M. Moretti R, Papale P (2009b) The 2007 eruption of Stromboli volcano: Insights from real-time measurement of the volcanic gas plume CO2/SO2 ratio. J Volcanol Geotherm Res 182(3-4):221-230 Arthur MA (2000) Volcanic contributions to the carbon and sulfur geochemical cycles and global change. In: Encyclopedia of Volcanoes. Sigurdsson H (ed) Academic Press, San Diego p 1045-1056 Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Backnaes L, Stelling J, Behrens H, Goettlicher J, Mangold S, Verheijen O, Beerkens GCR, Deubener J (2008) Dissolution mechanisms of tetravalent sulphur in silicate melts: Evidences from sulphur K edge XANES studies on glasses. J Am Ceram Soc 91(3):721-727 Baker DR, Barnes S-J, Simon G, Bernier F (2001) Fluid transport of sulfur and metals between sulfide melt and basaltic melt. Can Mineral 39:537-546 Baker DR, Freda C, Brooker RA, Scarlato P (2005) Volatile diffusion in silicate melts and its effects on melt inclusions. Ann Geophys 48(4-5):699-717 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Baker LL Rutherford MJ (1996) Sulfur diffusion in rhyolite melts. Contrib Mineral Petrol 123:335-344 Beermann O (2010) The solubility of sulfur and chlorine in H2O-bearing dacites of Krakatau and basalts of Mt. Etna. PhD Thesis, 109 p, Hannover Beermann O, Botcharnikov RE, Nowak M, Holtz F (2009) Redox control on S and Cl partitioning between basaltic melts and coexisting fluids: Experimental constraints at 1050 °C and 200 MPa. Eos Trans. AGU, 90(52), Fall Meet. Suppl. Abstract V43B-2230 Beermann O, Botcharnikov RE, Nowak M, Holtz F (2011) Temperature dependence of sulphide and sulphate solubility in olivine-saturated basaltic magmas. Geochim Cosmochim Acta, in press Behrens H, Stelling J (2011) Diffusion and redox reactions of sulfur in silicate melts. Rev Mineral Geochem 73:79-111 Berndt J, Liebske C, Holtz F, Freise M, Nowak M, Ziegenbein D, Hurkuck W, Koepke J (2002) A combined rapid-quench and H2-membrane setup for internally heated pressure vessels: Description and application for water solubility in basaltic melts. Am Mineral 87:1717-1726 Bischoff JL, Pitzer KS (1989) Liquid-vapor relations for the system NaCl-H2O: summary of the P-T-X surface from 300° to 500 °C. Am J Sci 289:217-248 Bockrath C, Ballhaus C, and Holzheid A (2004) Stabilities of laurite RuS2 and monosulfide liquid solution at magmatic temperature. Chem Geol 208:265-271 Bodnar RJ (1985) Pressure-volume-temperature-composition (PVTX) properties of the system H2O-NaCl at elevated temperatures and pressures. PhD Dissertation, The Pennsylvania State University, State College, Pennsylvania Bodnar RJ, Burnham CW, Sterner SM (1985) Synthetic fluid inclusions in natural quartz. III. Determination of phase equilibrium properties in the system H2O-NaCl to 1000 °C and 1500 bars. Geochim Cosmochim Acta 49:1861-1873 Botcharnikov RE, Behrens H, Holtz F, Koepke J, Sato H (2004) Sulfur and chlorine solubility in Mt. Unzen rhyodacite melt at 850 °C and 200 MPa. Chem Geol 213:207-225 Botcharnikov RE, Behrens H, Holtz F (2006) Solubility and speciation of C-O-H fluids in andesitic melt at T = 1100-1300 °C and P = 200 and 500 MPa. Chem Geol 229(1-3):125-143 Botcharnikov R, Freise M, Holtz F, Behrens H (2005) Solubility of C-O-H mixtures in natural melts: new experimental data and application range of recent models. Ann Geophys 48(4-5):633-646 Botcharnikov RE, Linnen RL, Wilke M, Holtz F, Jugo PJ, Berndt J. (2011) High gold concentration in sulphidebearing magma at oxidizing conditions. Nature Geosci, doi: 10.1038/NGEO1042 Bowers TS, Helgeson HC (1983) Calculation of the thermodynamic and geochemical consequences of nonideal mixing in the system H2O-CO2-NaCl on phase relations in geologic systems: Equation of state for H2O-CO2-NaCl fluids at high pressures and temperatures. Geochim Cosmochim Acta 47:1247-1275 Buchanan DL, Nolan J, Wilkinson N, De Villiers JPR (1983) An experimental investigation of sulfur solubility as a function of temperature in silicate melts. Spec Pub Geol Soc S Africa 7:383-391 Burgisser A, Scaillet B (2007) Redox evolution of a degassing magma rising to the surface. Nature 445:194197 Burgisser A, Scaillet B, Harshvardhan (2008) Chemical patterns of erupting silicic magmas and their influence on the amount of degassing during ascent. J Geophys Res 113(B12204):14p Burnham CW (1997) Magmas and hydrothermal fluids. In: Geochemistry of Hydrothermal Ore Deposits, 3rd edition. HL Barnes (ed) John Wiley & Sons, New York, p 63-123
278
Webster & Botcharnikov
Burton M, Allard P, Mure F, La Spina A (2007) Magmatic gas composition reveals the source depth of slugdriven strombolian explosive activity. Science 317(5835):227-230 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation state; results from measured wavelength shifts of sulfur X-rays. Am Mineral 73(7-8):845-849 Carroll MR, Webster JD (1994) Solubilities of sulfur, noble gases, nitrogen, chlorine, and fluorine in magmas. Rev Mineral 30:231-279 Cervantes P, Wallace PJ (2003) Role of H2O in subduction-zone magmatism: New insights from melt inclusions in high-Mg basalts from central Mexico. Geology 31:235-238 Chambefort I, Dilles J, Kent AJR (2008) Anhydrite-bearing andesite and dacite as a source for sulfur in magmatic-hydrothermal mineral deposits. Geology 36:719-722 Chase M (1998) NIST-JANAF thermochemical tables. Journal of Physical and Chemical Reference Data Monographs 9, 4th Edition. American Inst. of Physics Christopher T, Edmonds M, Humphreys MCS, Herd RA (2010) Volcanic gas emissions from Soufrière Hills Volcano, Montserrat 1995-2009, with implications for mafic magma supply and degassing. Geophys Res Lett 37:L00E04 Churakov SV, Gottschalk (2003) Perturbation theory based equation of state for polar molecular fluids: II. Fluid mixtures. Geochim Cosmochim Acta 67:2415-2425 Churikova T, Woerner G, Mironov N, Kronz A (2007) Volatile (S, Cl and F) and fluid mobile trace element compositions in melt inclusions: implications for variable fluid sources across the Kamchatka arc. Contrib Mineral Petrol 154(2):217-239 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulphur in hydrous rhyolitic melts. J Petrol 45:2171-2196 Core DP (2004) Oxygen and Sulfur Fugacities of Granitiods: Implications for Ore-Forming Processes. PhD dissertation, The University of Michigan, Ann Arbor, Michigan De Hoog JCM, Hattori KH, Hoblitt RP (2004) Oxidized sulfur-rich mafic magma at Mount Pinatubo, Philippines. Contrib Mineral Petrol 146:750-761 De Hoog JCM, Mason PRD, van Bergen MJ (2001) Sulfur and chalcophile elements in subduction zones: constraints from a laser ablation ICP-MS study of melt inclusions from Galunggung Volcano, Indonesia. Geochim Cosmochim Acta 65:3147-3164 Delmelle P, and Stix J (2000) Volcanic gases. In: Encyclopedia of Volcanoes. Sigurdsson H (ed) Academic Press, San Diego, p 803-815 Di Muro A, Pallister J, Villemant B, Newhall C, Semet M, Martinez M, Mariet C (2008) Pre-1991 sulfur transfer between mafic injections and dacite magma in the Mt. Pinatubo reservoir. J Volcanol Geotherm Res 175:517-540 Dixon JE (1997) Degassing of alkalic basalts. Am Mineral 82:368-378 Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in mid ocean ridge basaltic liquids. Part I: Calibration and solubility models. J Petrol 36:1607-1631 Driesner T (2007) The system H2O-NaCl. Part II. Correlations for molar volume, enthalpy, and isobaric heat capacity from 0 to 1000 °C, 1 to 5000 bar, and 0 to 1 XNaCl. Geochim Cosmochim Acta 71:4902-4919 Ducea M, McInnes B, Wyllie P (1994) Sulfur variations in glasses from volcanic rocks: Effect of melt composition on sulfur solubility. Int Geol Rev 36(8):703-714 Edmonds M, Gerlach TM (2007) Vapor segregation and loss in basaltic melts. Geology 35(8):751-754 Edmonds M, Herd RA (2007) A volcanic degassing event at the explosive-effusive transition. Geophys Res Lett 34:L21310 Edmonds M, Oppenheimer C, Pyle DM, Herd RA, Thompson G (2003) SO2 emissions from Soufrière Hills Volcano and their relationship to conduit permeability, hydrothermal interaction and degassing regime. J Volcanol Geotherm Res 124(1-2):23-43 Einaudi MT, Hedenquist JW, Inan EE (2003) Sulfidation state of fluids in active and extinct hydrothermal systems: transitions from porphyry to epithermal environments. In: Volcanic, Geothermal, and OreForming Fluids: Rulers and Witnesses of Processes within the Earth. Volume 10. Simmons SF, Graham I (eds) Econ Geol Spec Pub, p 285-313 Evans KA, O’Neill HSC, Mavrogenes JA, Keller NS, Jang LY, Lee JF (2009) XANES evidence for sulphur speciation in Mn-, Ni- and W-bearing silicate melts. Geochim Cosmochim Acta 73(22):6847-6867 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Fincham CJB Richardson FD (1954) The behavior of sulfur in silicate and aluminate melts. Phil Trans Roy Soc London A223:40-62 Fleet ME, Liu X, Harmer SL, King PL (2005) Sulfur K-edge XANES spectroscopy: Chemical state and content of sulfur in silicate glasses. Can Mineral 43(5):1605-1618 Fleet ME, Stone WE, Crocket JH (1991) Partitioning of palladium, iridium, and platinum between sulfide liquid and basalt melt - effects of melt composition, concentration, and oxygen fugacity. Geochim Cosmochim Acta 55(9):2545-2554
Sulfur Distribution between Melts and Fluids in Magmatic Systems
279
Froese E, Gunter AE (1976) A note on the pyrrhotite-sulphur vapor equilibrium. Econ Geol 71:1589-1594 Gaetani GA, Grove TL (1997) Partitioning of moderately siderophile elements among olivine, silicate melt, and sulfide melt: Constraints on core formation in the Earth and Mars. Geochim Cosmochim Acta 61(9):1829-1846 Gerlach TM (1986) Exsolution of H2O, CO2, and S during eruptive episodes at Kilauea Volcano, Hawaii. J Geophys Res 91:12177-12185 Gerlach TM, Westrich HR, Symonds RB (1995) Preeruption vapor in magma of the climactic Mount Pinatubo eruption: source of the giant stratospheric sulfur dioxide cloud. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) University of Washington Press p 415-433 Gerlach TM, Westrich HR, Casadevall TJ, Finnegan DL (1994) Vapor saturation and accumulation in magmas of the 1989-1990 eruption of Redoubt Volcano, Alaska. J Volcanol Geotherm Res 62:317-337 Giggenbach WF (1996) Chemical composition of volcanic gases. In: Monitoring and Mitigation of Volcano Hazards. Scarpa R, Tilling RI (eds) Springer, Berlin, p 221-256 Giggenbach WF (1997) The origin and evolution of fluids in magmatic-hydrothermal systems. In: Geochemistry of Hydrothermal Ore Deposits, 3rd Edition. Barnes HL (ed) John Wiley & Sons, New York, p 737-796 Gorbachev NS (1990) Fluid-magma interaction in sulfide-silicate systems. Int Geol Rev 32:749-836 Guillong M, Latkoczy C, Seo JH, Günther D, Heinrich CA (2008) Determination of sulfur in fluid inclusions by laser ablation ICP-MS. J Anal At Spectrom 23:1581-1589 Haughton DR, Roeder PL, Skinner BJ (1974) Solubility of sulfur in mafic magmas. Econ Geol 69:451-467 Hedenquist JW, Lowenstern JB (1994) The role of magmas in the formation of hydrothermal ore deposits. Nature 370:519-527 Holloway JR (1976) Fluids in the evolution of granitic magmas: consequences of finite CO2 solubility. Geol Soc Am Bull 87:1513–1518 Holloway JR (1987) Igneous fluids. Rev Mineral 17:211-233 Holloway JR, Blank JG (1994) Application of experimental results to C-O-H species in natural melts. Rev Mineral 30:187-230 Holzheid A, Grove TL (2002) Sulfur saturation limits in silicate melts and their implications for core formation scenarios for terrestrial planets. Am Mineral 87(2-3):227-237 Jahn S, Schmidt C (2010) Speciation in aqueous MgSO4 fluids at high pressures and high temperatures from ab initio molecular dynamics and Raman spectroscopy. J Phys Chem B 2010:15565-15572 Jantzen CM, Smith ME, Peeler DK (2004) Dependency of sulfate solubility on melt composition and melt polymerization (U). Symposium on Waste Management Technologies in Ceramic and Nuclear Industries. ACS. Report WSRC-MS-2004-00290, 14 p. Jugo PJ, Luth RW, Richards JP (2005a) An experimental study of the sulfur content in basaltic melts saturated with immiscible sulfide or sulfate liquids at 1300 °C and 1.0 GPa. J Petrol 46:783-798 Jugo PJ, Luth RW, Richards JP (2005b) Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim Cosmochim Acta 69(2):497-503 Jugo P, Wilke M, Botcharnikov RE (2010) Sulfur K-edge XANES analysis of natural and synthetic basaltic glasses: Implications for S speciation and S content as function of oxygen fugacity. Geochim Cosmochim Acta 74:5926-5938 Kamenetsky VS, Wolfe RC, Eggins SM, Mernagh TP, Bastrakov E (1999) Volatile exsolution at the Dinkidi Cu-Au porphyry deposit, Philippines: a melt-inclusion record of the initial ore-forming process. Geol 27:691-694 Katsura T, Nagashima S (1974) Solubility of sulfur in some magmas at 1 atmosphere. Geochim Cosmochim Acta 38(4):517-531 Keppler H (1999) Experimental evidence for the source of excess sulfur in explosive volcanic eruptions. Science 284:1652-1654 Keppler H (2010) The distribution of sulfur between haplogranite melts and aqueous fluids. Geochim Cosmochim Acta 74:645-660 Klimm K, Botcharnikov RE (2010) The determination of sulfate and sulfide species in hydrous silicate glasses using Raman spectroscopy. Am Mineral 95:1574-1579 Lenoir M, Grandjean A, Poissonnet S, Neuville DR (2009) Quantitation of sulfate solubility in borosilicate glasses using Raman spectroscopy. J Non-Cryst Solids 355(28-30):1468-1473 Lesne P, Kohn SC, Blundy J, Witham F, Botcharnikov RE, Behrens H (2011) Experimental simulation of basalt degassing at Stromboli and Masaya volcanoes. J Petrol, in press Li C, Ripley EM (2009) Sulfur contents at sulfide-liquid or anhydrite saturation in silicate melts: empirical equations and example applications. Econ Geol 104:405-412 Liebscher A, Heinrich CA (2007) Fluid-fluid interactions in the Earth’s crust – an introduction. Rev Mineral Geochem 65:1-13
280
Webster & Botcharnikov
Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71(7):1783-1799 Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichón volcano, Chiapas, Mexico: mineralogy and petrology of the anhydrite-bearing pumice. J Volcanol Geotherm Res 23:69-108 Mandeville CW (2010) Sulfur: a ubiquitous and useful tracer in Earth and planetary sciences. Elements 6:7580 Mandeville CW, Sasaki A, Saito G, Faure K, King R, Hauri E (1998) Open-system degassing of sulfur from Krakatau 1883 magma. Earth Planet Sci Lett 160:709-722 Martini M, Buccianti A (1997) The investigation of magmatic, hydrothermal and atmospheric contributions to the chemical composition of volcanic gases, as a guide to volcano surveillance (abs.) Jan. 1997 IAVCEI, Puerta Vallarta, Mexico Mathez EA (1976) Sulfur solubility and magmatic sulfides in submarine basalt glass. J Geophys Res 81(23):4269-4276 Matthews SJ, Moncrieff DHS, Carroll MR (1999) Empirical calibration of the sulphur valence oxygen barometer from natural and experimental glasses; method and applications. Mineral Mag 63(3):421-431 McGonigle AJS, Oppenheimer C. (2003) Optical sensing of volcanic gas and aerosol emissions. Geol Soc London Spec Pub 213(1):149-168 McKeown DA, Muller IS, Gan H, Pegg IL, Kendziora CA (2001) Raman studies of sulfur in borosilicate waste glasses: sulfate environments. J Non-Cryst Solids 288(1-3):191-199 Métrich N, Berry AJ, O’Neill HSC, Susini J (2009) The oxidation state of sulfur in synthetic and natural glasses determined by X-ray absorption spectroscopy. Geochim Cosmochim Acta 73(8):2382-2399 Métrich N, Bertagnini A, Landi P, Rosi M (2001) Crystallization driven by decompression and water loss at Stromboli Volcano (Aeolian Islands, Italy). J Petrol 42(8):1471-1490 Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim Cosmochim Acta 60:4151-4160 Métrich N, Mandeville CW (2010) Sulfur in magmas. Elements 6(2):81-86 Métrich N, Wallace P (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions. Rev Mineral Geochem 69:363-402 Moretti R (2005) Polymerisation, basicity, oxidation state and their role in ionic modelling of silicate melts. Ann Geophys 48:583-608 Moretti R, Baker DR (2008) Modeling the interplay of fO2 and fS2 along the FeS-silicate melt equilibrium. Chem Geol 256:286-298 Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts: The Conjugated ToopSamis-Flood-Grjotheim (CTSFG) model. Geochim Cosmochim Acta 69(4):801-823 Moretti R, Papale P, Ottonello G (2003) A model for the saturation of C-O-H-S fluids in silicate melts. Geol Soc London Spec Pub 213(1):81-101 Moune S, Holtz F, Botcharnikov R (2009) Sulphur solubility in andesitic to basaltic melts: implications for Hekla volcano. Contrib Mineral Petrol 157(6):691-707 Moune S, Sigmarsson O, Thordarson T, Gauthier P-J (2007) Recent volatile evolution in the magmatic system of Hekla volcano, Iceland. Earth Planet Sci Lett 255(3-4):373-389 Müller-Simon H (2011) Fining of glass melts. Rev Mineral Geochem 73:337-361 Mysen BO, Popp RK (1980) Solubility of sulfur in CaMgSi2O6 and NaAlSi3O8 melts at high-pressure and temperature with controlled fO2 and fS2. Am J Sci 280(1):78-92 Nagashima S, Katsura T (1973) Solubility of sulfur in Na2O-SiO2 melts under various oxygen partial pressures at 1100 °C, 1250 °C, and 1300 °C. Bull Chem Soc Japan 46(10):3099-3103 Newman S, Lowenstern JB (2002) VOLATILECALC: a silicate melt-H2O-CO2 solution model written in Visual Basic for excel. Comp Geosci 28:597-604 Newton RC, Manning CE (2005) Solubility of anhydrite, CaSO4, in NaCl-H2O solutions at high pressures and temperatures. J Petrol 46:701-716 Nowak M, Behrens H (1995) The speciation of water in haplogranitic glasses and melts determined by in situ near-infrared spectroscopy. Geochim Cosmochim Acta 59:3445-3450 O’Neill H St C (1987) The quartz-fayalite-iron and quartz-fayalite-magnetite equilibria and the free energies of formation of fayalite (Fe2SiO4) and magnetite (Fe3O4). Am Mineral 72:67-75 Oppenheimer C (2003) Volcanic degassing. In: The Crust. Treatise in Geochemistry. Volume 3. Rudnick RL (ed) Elsevier, p 123-166 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Pallister JS, Hoblitt RP, Reyes AG (1992) A basalt trigger for the 1991 eruptions of Pinatubo volcano? Nature 356:426-428 Parat F, Holtz F (2004) Sulphur partitioning between apatite and melt and effect of sulphur on apatite solubility at oxidizing conditions. Contrib Mineral Petrol 147:201-212
Sulfur Distribution between Melts and Fluids in Magmatic Systems
281
Parat F, Holtz F, Feig S (2008) Pre-eruptive conditions of the Huerto Andesite (Fish Canyon System, San Juan volcanic field, Colorado): influence of volatiles (C-O-H-S) on phase equilibria and mineral composition. J Petrol 49:911-935 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285-314 Pokrovski GS, Dubrovinsky LS (2011) The S3− ion is stable in geological fluids at elevated temperatures and pressures. Science 331(6020):1052-1054 Portnyagin M, Hoernle K, Plechov P, Mironov N, Khubunaya S (2007) Constraints on mantle melting and composition and nature of slab components in volcanic arcs from volatiles (H2O, S, Cl, F) and trace elements in melt inclusions from the Kamchatka Arc. Earth Planet Sci Lett 255(1-2):53-69 Raia F, Webster JD, DeVivo B (2000) Pre-eruptive volatile contents of Vesuvius magmas: contraints on eruptive history and behavior. I-the medieval and modern interplinian activities. Eur J Mineral 12:179-193 Rethmeier J, Rabenstein A, Langer M, Fischer U (1997) Detection of traces of oxidized and reduced sulfur compounds in small samples by combination of different high-performance liquid chromatography methods. J Chromat A 760:295-302 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Roedder E (1984) Fluid Inclusions. Reviews in Mineralogy Volume 12. Mineral Soc Am Rye RO, Luhr JF, Wasserman MD (1984) Sulfur and oxygen isotope systematics of the 1982 eruptions of El Chichón Volcano, Chiapas, Mexico. J Volcanol Geotherm Res 23:109-123 Salvi S, Williams-Jones AE (2003) Bulk analysis of volatiles in fluid inclusions. In: Fluid Inclusions: Analysis and Interpretation. Samson I, Anderson A, Marshall D (eds.) Min Assoc Canada Short Course 32:247-278 Scaillet B, Clemente B, Evans BW, Pichavant M (1998) Redox control of sulfur degassing in silicic magmas. J Geophys Res 103(B10):23,937-23,949 Scaillet B, Evans BW (1999) The June 15, 1991 eruption of Mount Pinabuto. I. Phase equilibria and preeruption P-T-fO2-ƒH2O conditions of the dacite magma. J Petrol 40:381-411 Scaillet B, Macdonald R (2006) Experimental and thermodynamic constraints on the sulphur yield of peralkaline and metaluminous silicic flood eruptions. J Petrol 47:1413-1437 Scaillet B, Pichavant M (2003) Experimental constraints on volatile abundances in arc magmas and their implications for degassing processes. In: Volcanic Degassing. Vol. 213. Oppenheimer C, Pyle DM, Barclay J (eds). Geol Soc Spec Pub p 23-52 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48(4-5):671-698 Scaillet B, Pichavant M, Roux J, Humbert G, Lefevre A (1992) Improvements of the Shaw membrane technique for measurement and control of ƒH2 at high temperatures and pressures. Am Mineral 77:647-655 Schmidt C (2009) Raman spectroscopic study of a H2O + Na2SO4 solution at 21-600 °C and 0.1 MPa to 1.1 GPa: relative differential v1SO42− Raman scattering cross sections and evidence of the liquid-liquid transition. Geochim Cosmochim Acta 73:425-437 Seo JH, Guillong M, Heinrich CA (2009) The role of sulfur in the formation of magmatic-hydrothermal copper-gold deposits. Earth Planet Sci Lett 282:323-328 Seward TM (1981) Metal complex formation in aqueous solutions at elevated temperatures and pressures. In: Physics and Chemistry of the Earth. Vol. 13+14. Richard DT, Wickman FE (eds). Pergamon Press, p 113-132 Shi P, Saxena SK (1992) Thermodynamic modeling of the C-H-O-S fluid system. Am Mineral 77:1038-1049 Shishkina T, Botcharnikov RE, Holtz F, Almeev RR, Portnyagin M (2010) Solubility of H2O- and CO2-bearing fluids in tholeiitic basalts at pressures up to 500 MPa. Chem Geol 277: 115-125 Siggurdson H, Carey S, Palais JM, Devine J (1990) Pre-eruptive compositional gradients and mixing of andesite and dacite magma eruption from Nevado del Ruiz volcano, Colombia in 1985. J Volcanol Geotherm Res 41:127-151 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Sisson TW, Layne GD (1993) H2O in basalt and basaltic andesite glass inclusions from 4 subduction-related volcanoes. Earth Planet Sci Lett 117:619-635 Sourirajan S, Kennedy GC (1962) The system H2O-NaCl at elevated temperatures and pressures. Am J Sci 260:115-141 Spilliaert N, Allard P, Métrich N, Sobolev AV (2006a) Melt inclusion record of the conditions of ascent, degassing, and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy). J Geophys Res-Solid Earth 111(B4):B04203 doi: 10.1029/2005JB003934 Spilliaert N, Métrich N, Allard P (2006b) S-Cl-F degassing pattern of water-rich alkali basalt: Modelling and relationship with eruption styles on Mount Etna volcano. Earth Planet Sci Lett 248(3-4):772-786
282
Webster & Botcharnikov
Sterner SM, Bodnar RJ (1991) Synthetic fluid inclusions. X: Experimental determination of P-V-T-X properties in the CO2-H2O system to 6 kb and 700 °C. Am J Sci 291:1-54 Stevens RK, Mulik JD, O’Keeffe AE, Krost KJ (1971) Gas chromatography of reactive sulfur gases in air at the parts-per-billion level. Anal Chem 43(7):827-831 Symonds RB, Rose WI, Bluth GJS, Gerlach TM (1994) Volcanic-gas studies: Methods, results and applications. Rev Mineral 30:1-66 Tamic N, Behrens H, Holtz F (2001) The solubility of H2O and CO2 in rhyolitic melts in equilibrium with a mixed CO2-H2O fluid phase. Chem Geol 174:333-347 Taylor JR, Wall VJ, Pownceby MI (1992) The calibration and application of accurate redox sensors. Am Mineral 77:284-295 Teague AJ, Kohn SC, Klimm K, Botcharnikov RE (2008) Sulphur solubility in Mount Hood andesites and CO2 fluids: Implications for volcanic degassing EOS Trans. AGU, 89(53), Fall Meet. Suppl. Abstract V21B-2086 Toulmin PI, Barton PBJ (1964) A thermodynamic study of pyrite and pyrrhotite. Geochim Cosmochim Acta 28:641-671 Tsujimura T, Xue X, Kanzaki M, Walter MJ (2004) Sulfur speciation and network structural changes in sodium silicate glasses: Constraints from NMR and Raman spectroscopy. Geochim Cosmochim Acta 68(24):5081-5101 Vaughan DJ, Craig JR (1997) Sulfide ore mineral stabilities, morphologies, and intergrowth textures. In: Geochemistry of Hydrothermal Ore Deposits, 3rd Edition. Barnes HL (ed) John Wiley & Sons, New York, p 367-434 Vetere F, Botcharnikov RE, Holtz F, Behrens H, De Rosa R (2011) Solubility of H2O and CO2 in shoshonitic melts at 1250 °C and pressures from 50 to 400 MPa: Implications for Campi Flegrei magmatic systems. J Volcanol Geotherm Res 202:251-261 Vigouroux N, Wallace P, Kent AJR (2008) Volatiles in high-K magmas from the western Trans-Mexican Volcanic Belt: Evidence for fluid-flux melting and extreme enrichment of the mantle wedge by subduction processes. J Petrol doi:10.1093/petrology/egn039 Vogt JHL (1917) Die sulphid-silikat-schmetlzlösungen. Norsk Geol Tides IV:43-51 Wade JA, Plank T, Melson WG, Soto GJ, Hauri AH (2006) Volatile content of magmas from Arenal volcano, Costa Rica. J Volcanol Geotherm Res 157:94-120 Wallace PJ (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma bodies. J Volcan Geotherm Res 108:85-106 Wallace PJ (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. J Volcanol Geotherm Res 140:217-240 Wallace PJ, Anderson AT, Jr (2000) Volatiles in magma. In: Encyclopedia of Volcanoes. Siggurdson H (ed) Academic Press, San Diego p. 149-170 Wallace PJ, Carmichael ISE (1994) S speciation in submarine basaltic glasses as determined by measurements of S Ka X-ray wavelength shifts. Am Mineral 79(1-2):161-167 Wallace PJ, Edmonds M (2011) The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Rev Mineral Geochem 73:215-246 Webster JD (1992) Fluid-melt interactions involving Cl-rich granites: experimental study from 2 to 8 kbar. Geochim Cosmochim Acta 56:679-687 Webster JD (2009) Advances in constraining solubilities of H-O-C-S-Cl-bearing fluids in silicate melts (abs). EOS V51-G02 Webster JD, Goldoff B, Shimizu N (2011) COHS fluids and granitic magma: how S partitions and modifies CO2 concentrations of fluid-saturated felsic melt at 200 MPa. Contrib Mineral Petrol, doi:10.1007/ s00410-011-0628-1 Webster JD, Mandeville CW (2007) Fluid immiscibility in volcanic environments. Rev Mineral Geochem 65:313-362 Webster JD, Mandeville CW, Goldoff B, Coombs ML, Tappen C (2010) Augustine volcano, Alaska: The influence of volatile components in magmas erupted AD 2006 to 2100 years before present. In: The 2006 Eruption of Augustine Volcano. Power J, Coombs ML, Freymueller J (eds) US Geol Surv Prof Paper 1769:383-423 Webster JD, Sintoni MF, De Vivo B (2006) The role of sulfur in promoting magmatic degassing and volcanic eruption at Mt. Somma-Vesuvius. In: Volcanism in the Campania Plain. Volume 9. De Vivo B (ed), Elsevier Amsterdam, p 219-233 Webster JD, Sintoni MF, De Vivo B (2009) The partitioning behavior of Cl and S in aqueous fluid- and salineliquid saturated phonolitic and trachytic melts at 200 MPa. Chem Geol 263:19-36 Wilke M, Jugo PJ, Klimm K, Susini J, Botcharnikov R, Kohn SC, Janousch M (2008) The origin of S4+ detected in silicate glasses by XANES. Am Mineral 93(1):235-240
Sulfur Distribution between Melts and Fluids in Magmatic Systems
283
Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Williams-Jones G, Rymer H (2000) Hazards of volcanic gases. In: Encyclopedia of Volcanoes. Sigurdsson H (ed) Academic Press San Diego, p 997-1004 Winther KT, Watson EB, Korenowski GM (1998) Magmatic sulfur compounds and sulfur diffusion in albite melt at 1 GPa and 1300-1500 degrees C. Am Mineral 83(11-12 Part 1):1141-1151
10
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 285-314, 2011 Copyright © Mineralogical Society of America
Sulfur-bearing Magmatic Accessory Minerals Fleurice Parat Géosciences Montpellier - Université Montpellier 2 Place E. Bataillon 34095 Montpellier cedex 5. France and Mineralogie-Geochemie, Institut für Geowissenschaften Albertstr. 23b 79104 Freiburg, Germany [email protected]
François Holtz Institut für Mineralogie, Callinstr. 3 Leibniz Universität Hannover 30167, Germany
Martin J. Streck Department of Geology Portland State University Portland, Oregon 97207-0751, U.S.A. IntroductIon: tHe occurrence oF MAgMAtIc SulFur-beArIng MInerAlS Almost all magmas contain sulfide- or sulfate-bearing phases. In most natural samples the sulfur-bearing phase is a sulfide, which typically is pyrrhotite (Fe1−xS) or pyrite (FeS2) although chalcopyrite (CuFeS2), pentlandite ((Fe,Ni)9S8), sphalerite (ZnS) or molybdenite (MoS2) may be present as well. Sulfate minerals are rare at magmatic conditions, and anhydrite (CaSO4) is the most common. Other magmatic SO4-bearing minerals include the sodalite group minerals (haüyne simplified formula: (Na,Ca)4−8(Al6Si6(O,S)24)(SO4,Cl)1−2), scapolite minerals (silvialite: (Ca,Na)4Al6Si6O24(SO4,CO3)), and S-bearing apatite (Ca5(PO4)3(F, Cl, OH)). Barite (BaSO4) has been mentioned in rare cases (Marchev 1991). Sulfur-bearing minerals usually constitute a negligible fraction of the mineral assemblage in magmatic rocks and thus can be classified as accessory minerals. The crystallization of sulfide, sulfate, and S-bearing minerals strongly depends on melt composition, temperature and pressure, and the S speciation in melt which, in turn, is strongly dependent on the prevailing oxygen fugacity (Baker and Moretti 2011, this volume; Wilke et al. 2011, this volume). Irrespectively of the low abundance of S-bearing minerals, the evolution of sulfur in magmas may be evaluated from the occurrence of sulfides and sulfates. These minerals are critical for estimating the activity of various sulfur-bearing species in magmas and can be used to constrain the oxygen fugacity and the S concentration in the melt (e.g., pre-eruptive sulfur concentration in melts). The presence of either sulfide or sulfate in silicate melt indicates that the predominant dissolved sulfur species in the melt are S2− or S6+, respectively. Typically one of these species is dominant, but there are conditions where S2− or S6+ are present in subequal proportions (Wilke et al. 2011, this volume). 1529-6466/11/0073-0010$05.00
DOI: 10.2138/rmg.2011.73.10
Parat, Holtz, Streck
286
In this chapter, the main S-bearing minerals occurring in magmatic rocks and their geological setting are described, and their utility in understanding sulfur geochemistry in magmas is explored. Because subsolidus reactions may affect S-bearing minerals in plutonic rocks, we focus on reviewing minerals observed in volcanic rocks.
Magmatic sulfides Magmatic sulfides are common in arc-related mafic and silicic igneous rocks, oceanic island basalts and mid-oceanic ridge basalts, whereas they are rare in alkaline magmas from intracontinental settings (e.g., Scaillet and Macdonald 2006). They are accessory minerals occurring as isolated crystals in glass or included in other mineral phases (such as magnetite, pyroxene, sphene, plagioclase). They rarely form euhedral crystals but instead commonly occur as sulfide globules or blebs (Fig. 1). Pyrrhotite is a stable sulfide in many arc related magmas (Figs. 1a and 2), whereas Cu-Ni-Fe-sulfides are abundant in oceanic island basalts and mid-oceanic island basalts (Figs. 1d and 3). Chalcopyrite, pentlandite and sphalerite are often reported in magmatic rocks, however, these sulfide phases formed after sulfide entrapment during cooling under subsolidus conditions. At magmatic conditions the stable sulfide phases are monosulfide solid solution (Fe1−xS-Ni1−xS, mss; Naldrett et al. 1967), intermediate solid solution (iss, high-temperature solid solution with compositions close to that of CuFeS2; see reviews by Raghavan 2004 and Fleet 2006), Cu-rich pyrrhotite, and a sulfide liquid. Upon cooling, monosulfide solid solution converts to pyrrhotite through exsolution of pentlandite,
d. b.
a. Po
CuFeS2
mag ilm
10 μ m
c.
10 μ m
d.
mag
icb
mss
100 μ m
mo
100 μ m
Figure 1. Microphotographs of magmatic sulfides. (a) Pyrrhotite (po) inclusions in magnetite taken with reflected light microscopy, Huerto Andesite; (b) Cu-Fe-sulfides in magnetite (mag) in Mount Pinatubo dacite with reflected light microscopy; ilm: ilmenite (Fournelle et al. 1996); (c) Triangular, semi-transparent crystal of molybdenite (mo) and coexisting melt inclusions from the Amalia Tuff, Latir volcanic field, New Mexico taken Figure with transmitted light microscopy (Audétat et al. 2011). (d) Sulfides in olivine phenocryst hosted bleb 1 of monosulfide solid solution (mss) with magnetite (mag) and subsolidus lamellar exsolution of isocubanite (icb) in pumice from the 1959 eruption of Kilauea (Stone and Fleet 1991). {Figure 1c was reproduced with permission of Oxford University Press from Audetat et al. (2011) Journal of Petrology, Vol. 54, p. 891-904}
Sulfur-bearing Magmatic Accessory Minerals
287
whereas the intermediate solid solution unmixes to form chalcopyrite and cubanite (CuFe2S3) (e.g., Mungall 2007; Nadeau et al. 2010). Sulfides are rare in peralkaline igneous rocks consisting of pyrrhotite, pyrite, and molybdenite (Scaillet and Macdonald 2006; Audétat et al. 2011) (Fig. 1). For example, pyrrhotite is present in peralkaline rhyolite of Tejeda volcano, Gran Canaria and in syenite from La Gomera Island, Canary Islands (Crisp and Spera 1987; Rodríguez-Losada and MartínezFrías 1998). Pyrrhotite and molybdenite are present in peralkaline rhyolite from Pantelleria, Italy (Lowenstern et al. 1993). Molybdenite may be more common in magmatic systems than recognized up to now. For example, Audétat et al. (2011) looked carefully for the presence of molybdenite in 27 continental felsic systems and identified this mineral as inclusions in quartz in 13 cases. Pyrrhotite (Fe1−xS) is the most abundant sulfide reported in continental arc-related magmatic rocks (e.g., Whitney and Stormer 1983; Luhr et al. 1984) and in island arc basalts and andesites (e.g., Heming and Carmichael 1973; Ueda and Itaya 1981). Magmatic pyrrhotite can crystallize at high temperature from reduced S-saturated melt (log fO2 (bar) < NNO+1, where NNO refers to the logarithm of the oxygen fugacity defined by the Ni-NiO buffer; Carroll and Rutherford 1987). At high oxygen fugacity, pyrrhotite may partially breakdown to a Fe-O-S liquid (Clemente et al. 2004). During cooling and devitrification, pyrrhotite can oxidize to form pyrite (FeS2) and magnetite (6FeS + 2O2 = 3FeS2 + Fe3O4) (Whitney 1984). Pyrite has been recorded in some volcanic examples (Table 1) but it is not clear if this mineral is magmatic or was produced by a subsolidus reaction. Both Cu-Fe sulfide and Ni-Fe sulfide have been identified as magmatic phases. For example, they are reported in island arc volcanic rocks from the 1991 Mount Pinatubo eruption (e.g., Pallister et al. 1992; Hattori 1996; McKibben et al. 1996), the Merapi volcano (Nadeau et al. 2010), and in continental arc andesites from Mount Shasta (Stone et al. 1989); pyrrhotite is rare in such occurrences. The composition of Cu- and Ni- sulfides ranges from pyrrhotite to bornite (Cu5FeS4) and from pyrrhotite to pentlandite ((Fe,Ni)9S8), respectively (Fig. 2). The mineral compositions plot in the fields of bornite solid solution, intermediate solid solution (iss), and pyrrhotite. Compositions intermediate between these three phases are not observed, in agreement with the existence of immiscibility gap (see phase relationships below). Curich minerals are ubiquitous inclusions in pyrrhotite, magnetite or in the matrix of some other arc-related magmas (Luhr et al. 1984; Lowenstern 1993; Chesner 1998; Costa et al. 2004) (Fig. 1b). Both iss and pyrrhotite have been reported as inclusions in silicate phenocrysts crystallized directly from silicate melt at high temperature (e.g., Lowenstern 1993), whereas small chalcopyrite or cubanite (CuFe2S3) crystals inside pyrrhotite were attributed to exsolution from the host pyrrhotite (e.g., Luhr et al. 1984). Phase equilibria suggest that the coexistence of Ni-bearing pyrrhotite and Cu-Fe-sulfide (e.g., iss) may be related to the miscibility gap between two high temperature CuFeS2 (iss) and Fe1−xS-Ni1−xS (mss) phases (Craig and Kullerud 1969; Raghavan 2006). However, textural observations and general occurrences of sulfides in mafic and silicic rocks suggest that the two types of sulfides do not crystallize contemporaneously but rather one after the other starting with early crystallization of Ni-pyrrhotite followed by crystallization of Cu-Fe-sulfides (along with anhydrite) as illustrated in the following examples. In volcanic rocks of Mount Pinatubo volcanics, in which both sulfate (anhydrite) and sulfides are present, sulfide phases are: (1) globular nickel-bearing (<4.3 wt% Ni) pyrrhotite as inclusions in early phenocrysts (olivine, augite) of basalt and andesite; and (2) irregularly shaped Cu-rich (up to 53 wt% Cu, Figs. 1b and 2) sulfides present as inclusions in late phenocrysts and in glass of basalt, andesite, and dacite (Hattori 1993; Fournelle et al. 1996; Hattori 1996; de Hoog et al. 2004). Hattori (1993, 1996) suggested a later crystallization for Cu-rich sulfide (and
Parat, Holtz, Streck
288
table 1. Sulfate-bearing magmatic rocks. locality
Age
rock type
Philippines Colombia Mexico Chile Peru Peru Russia Nouvelle Guinee U.S.A. U.S.A. U.S.A. U.S.A.
1991 1985 1982 26.5 ka 9.7-10.2 Ma 14.5-11.2 Ma 2001-2004 1951 1989-1990 1.36-1.59 Ma 28 Ma 63 Ma
dacite andesite-dacite trachyandesite andesite-dacite rhyolite andesite-dacite andesite andesite andesite andesite andesite rhyodacite
Chile Russia U.S.A. U.S.A. U.S.A.
5.7 Ma 251 Ma 63 Ma 71-75 Ma Cretaceous
microdiorite gabbro monzodiorite granodiorite tonalite-granodiorite
phonolite tephrite trachyphonolite trachyandesite phonolite nephelinite phonolite trachyte-phonolite phonolite ijolite-urtite trachyandesite tephrite-phonotephrite basanite, phonolite tephrite, phonotephrite phonolite hauynophyre syenitic nodule phonolitic leucitite phonotephrite phonotephrite phonolite latite-trachyte
log fo2
T (°c)
NNO+1.7 NNO+1.5 NNO+1 NNO+1.6 >>NNO NNO+1.5-2 n.d. NNO NNO+1.5-2 n.d. NNO+1.3 NNO+1.3-1.9
780 900 800 915 840 800-900 850 900 840-950 n.d. 900 730-760
n.d. n.d. >NNO+1 n.d. NNO+2
650 n.d. n.d. 800 700-800
NNO+2.3 NNO+1-2 NNO+1-2 n.d. NNO+0.5 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. NNO+2.2-2.9 NNO+2.2-2.9 n.d. n.d. n.d. n.d. n.d. n.d.
760 900 700 n.d. 800 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.
ANhydriTe-beAriNg roCks Volcanic Mount Pinatubo Nevado del Ruiz El Chichón Lascar Julcani Yanacocha Shiveluch Mount Lamington Redoubt Sutter Butes San Juan Volcanic Field Santa Rica
Intrusive El Teniente Kharaelakh Santa Rica Idaho Cajon Pass
sodAliTe-beAriNg roCks Laacher See Tephra Massif Central
Germany France
13 ka 65 Ma-3450 BP
Aiguillier Massif Tenerife Santo Antao Tahiti Wolf Rock, Cornwall Oldoinyo Lengai North Qiangtang North Qiangtang Vulture
France Canary Islands Cape Verde France England Tanzania Tibet Tibet Italy
2 Ma 0.6 Ma 0.7- 0.1Ma 0.5-1.2 Ma
Vulsini District
Italy
0.6-0.15 Ma
Vico Alban Hills, Tavolato Vesuvius Ventotene-Santo Stefano
Italy Italy Province Italy Italy
0.42-0.1 Ma 0.6-0.02 MA
0.8-0.58 Ma
ab: albite, aeg: aegirine, afs: alkalifeldspar, aln: allanite, am: amphibole, anh: anhydrite, an: anorthite, ap: apatite, bn: bornite, bt: biotite, ccp: chalcopyrite, cpx: clinopyroxene, cst: cassiterite, cup: cuprite, g: glass, grt: garnet, hbl: hornblende, hyn: haüyne, ilm: ilmenite, iss: intermediate solid solution, krs: kaersutite, lct: leucite, mag: magnetite, mel: melilite, mnz: monazite, mo: molybdenite, ne: nepheline, nsn: nosean, ol: olivine, opx: orthopyroxene, ox: FeTi-oxides, phl: phlogopite, pl: plagioclase, pn: pentlandite, po: pyrrhotite, py: pyrite, qtz: quartz, sa: sanidine, s-mei: sulfate meionite, sod: sodalite, spn: spinel, ttn: titanite, zrn: zircon. n.d.: not determined. NNO refers the logarithm of the oxygen fugacity defined by the Ni- NiO buffer.
Sulfur-bearing Magmatic Accessory Minerals
Paragenesis
pl, hbl, bt, qtz, anh, ap, ccp, ox pl, cpx, opx, bt, anh, ap, py, mag, ilm pl, cpx, hbl, bt, anh, ap, ttn, po, mag pl, cpx, opx, hbl, bt, qtz, ap, anh, po, zrn, ox, ilm pl, cpx, opx, hbl, bt, qtz, ttn, ap, anh, po, ox pl, cpx, opx, hbl, bt, qtz, ap, anh, po, ttn, mag, ilm pl, hbl, opx, cpx, ttn, cst, cup, bn, ilm, mag, ap, anh pl, cpx, opx, hbl, bt, mag, ap, zrn, anh, py pl, cpx, opx, hbl, anh, mag, ilm pl, hbl, bt, ap, mag, anh pl, cpx, opx, hbl, anh, ap, po, py, mag pl, hbl, bt, qtz, anh, ap, po, ccp, ttn, ilm, mag
pl, bt, qtz, anh, po, mag, ilm pl, ol, cpx, anh, po, pn, ccp pl, hbl, bt, qtz, anh, ap, ttn, mag pl, qzt, bt, afs, anh, ap, mnz, aln, zrn, py, ox pl, hbl, ttn, qtz, anh, ap, mag, ilm
sa, pl, am, phl, cpx, ttn, hyn, ap pl, afs, cpx, am, hyn, ap, mag pl, afs, cpx, am, bt, hyn, nsn, sod, zrn, spn, mag pl, cpx, am, hyn, nsn, ol, afs, ap, ttn, ox afs, bt, cpx, hyn, nsn, ap, ttn, mag, ilm sa, cpx, krs, hyn, ttn, ox pl, ne, sa, lct, cpx, hyn, nsn, ap, mag, ttn sa, ne, nsn, cpx, ox ne, cpx, grt, mag, ap, hyn, po cpx, lct, nsn, hyn, ne, sa, ap, ox ol, cpx, lct, ne, nsn, hyn, ttn, sa, ap, ox lct, ol, cpx, hyn, spl, afs, bt, ap, ox pl, cpx, ol, lct, am, afs, ne, bt, hyn, ap, mag sa, ne, grt, ttn, cpx, lct, hyn, ap, mag hyn, cpx, ne, lct, mag, sod, ap, mel afs, pl, cpx, am, hyn, mag, ttn cpx, bt, lct, pl, sa, hyn sa, cpx, hyn, ox, ap, ttn
po
iss anh hyn nsn
+ + + + + + + + +
+
+ + +
+ +
289
references
+ + + + + + + + + + + +
Pallister et al. 1992, 1996 Fournelle 1990 Luhr et al. 1984 Matthews et al. 1999 Drexler & Munoz 1985; Deen et al. 1994 Chambeford et al. 2008 Dirksen et al. 2006 Arculus et al. 1983 Swanson & Kearney 2008 Luhr 2008 Parat et al. 2002 Audétat and Pettke 2006
+ + + + +
Stern et al. 2007 Li et al. 2009 Audétat et al. 2004 Tepper & Kuehner 1999 Barth & Dorais 2000
+ + + + + + + + + + + + + + + + + + + +
+ + +
+ + + +
Wörner & Schmincke 1984 Downes 1987 Downes 1987 Morel & Gourgaud 1992 Bryan 2006 Holm et al. 2006 Tracy 2003 Tilley 1959; Harrison et al. 1977 Dawson et al. 1994; Zaitsev et al. 2009 Guo et al. 2006 Guo et al. 2006 De Fino et al. 1986; Bindi et al. 1999 De Fino et al. 1986; Bindi et al. 1999 De Fino et al. 1986; Bindi et al. 1999 De Fino et al. 1986; Bindi et al. 1999 Renzulli et al. 1998 Holm 1982 Perini et al. 2004 Peccerillo 2005 Peccerillo 2005 Peccerillo 2005
Parat, Holtz, Streck
290
Island arc; Heming and Carmichael 1973 Island arc; Hattori 1996 Island arc; Nadeau et al. 2010 Continental arc; Lowenstern 1993 Continental arc; Stone et al. 1989
S
S (at%) 20
80
a. 30
70
py
40
60
60 70
po
ccp
50
50
cu
40
bn
10 Cu (at%)
20
b.
30
40
50
S (at%)
20 30
60 Fe (at%)
80 70
py
40
60
vio
po
50
50
pn
60 70
10 Fe (at%)
30
20
30
40
40
50
60
30 Ni (at%)
Figure 2. Sulfide composition in arc-related rocks (individual analyses). (a) Cu-Fe-S and (b) Ni-Fe-S ternary plot of normalized atomic proportions in the sulfides. Sulfide compositions are either from reconstructed Figure 2 (Nadeau et al. 2010) or from sulfides without exsolution texture. bn: bornite, ccp: sulfide compositions chalcopyrite, cu: cubanite, pn: pentlandite, po: pyrrhotite, py: pyrite, vio: violarite.
anhydrite) from SO2-rich supercritical fluids released by underlying mafic magmas. Because the dacitic magma is more oxidized than the mafic magma, SO2 would be converted to H2S. The H2S produced by the dacitic magma would form sulfides in the dacitic melt together with elements transported by the fluids, such as Zn, Cd, Cu, and Se. An alternative scenario for the formation of Cu-rich sulfide and anhydrite in the Mount Pinatubo dacite may be due to the following (simplified) reaction: 2SO2 + CaO + FeO = CaSO4 + FeS + O2
(1)
In any case, the presence of Cu-rich pyrrhotite in silicic magmas such as Mount Pinatubo may be used as a proxy for the presence of Cu-bearing supercritical fluids. Evidence for a Cu-bearing fluid in magmas has been found by Lowenstern (1993) in form of Cu-rich vapor bubbles in melt inclusions (Valley of Ten Thousand Smokes). It should be noted that supercritical fluids may also contain chlorine. The presence of Cl leads to higher partition coefficients for S between solid phases and silicate melt (Botcharnikov et al. 2004) and lowers sulfide solubility in melt.
Sulfur-bearing Magmatic Accessory Minerals
291
Chlorine may thus play a role for sulfide saturation, as also evidenced for apatite and other sulfate-bearing minerals, which may contain appreciable Cl concentrations (see below). At Mount Shasta, complicated sulfide assemblages are present as inclusions in olivine and pyroxene megacrysts of andesitic tephras. They include mss, iss, pyrrhotite, pentlandite, chalcopyrite, violarite (FeNi2S4), idaite (Cu3FeS4), and pyrite (Stone et al. 1989). Ni-sulfides are common in olivine megacrysts, whereas chalcopyrite is present as inclusion in pyroxene megacrysts. The similarity of bulk sulfide (Fig. 2) and olivine compositions to experimental high-temperature partitioning data suggest an origin through sulfide liquid immiscibility, although high-temperature subsolidus processes such as degassing or S infiltration along fissures may also have been important (Stone et al. 1989). intermediate Ni-Cu-Fe sulfides are reported for mid-oceanic ridge basalts (e.g., Mathez 1976; Francis 1990) and oceanic island basalts (e.g., Stone and Fleet 1991). These sulfides occur as globules in glass or as inclusions in olivine or plagioclase phenocrysts. They are Ni- and Cu-rich, containing up to 58 wt% Cu (close to bornite composition) and 32 wt% Ni (pentlandite) (Fig. 3). The presence of Ni-Cu sulfides can be explained by direct crystallization from a silicate melt. However, the models commonly invoked to explain the crystallization of Ni-Cu-Fe sulfides include: i) formation of an immiscible sulfide liquid during the early magmatic stage of the host rocks, either by segregation (Naldrett 1981) or assimilation (Lesher and Groves 1986), ii) sulfide metasomatism during the late-magmatic stage, or iii) subsolidus processes (Stone et al. 1989). Molybdenite. Recently, Audétat et al. (2011) showed that molybdenite (MoS2) may be a common mineral in intraplate silicic magma systems. It occurs as small, brownish, semitransparent triangular or hexagonal platelets in quartz phenocrysts (Fig. 1c) (Lowenstern et al. 1993; Audétat et al. 2011). The presence of molybdenite in pantellerite is explained by Lowenstern et al. (1993) as result from a drop in temperature during magmatic differentiation of a Mo-rich peralkaline melt (12-25 ppm Mo (10000 ppm = 1 wt%)) at reducing conditions. Furthermore, the crystallization of molybdenite is thought to be favored by the absence of other Mo-bearing phases such as oxides (e.g., ilmenite) or titanite. To our knowledge, no exact composition of molybdenite occurring in volcanic rocks has been published (due to the small size of the minerals). However, the coexistence of molybdenite with pyrrhotite may be useful to determine fO2 and fS2 prevailing in magmas (Audétat et al. 2011; see below).
Magmatic sulfates and sulfate-bearing minerals Although magmatic sulfates and sulfate-bearing minerals are rare, reports of sulfatebearing minerals in igneous rocks (e.g., nosean, haüyne, silvialite) date back to early in the 20th century (Lindgren and Ransome 1906; Brauns 1914). The minerals crystallize from oxidized sulfate-bearing melts. Anhydrite was found in arc-related magmas, whereas sodalite-group minerals (haüyne and nosean) are present in alkaline magmas of intraplate settings. The absence of pyrrhotite in alkaline rocks and the presence of sulfate-bearing minerals may reflect the higher fO2 or lower Fe2+/Fe3+ when compared with metaluminous rocks (Scaillet and Macdonald 2006). Table 1 lists well-known natural examples from which sulfates and SO4-rich sodalitegroup minerals have been described, although Table 1 is meant to be neither a comprehensive list including all examples worldwide nor representative of the relative global distribution of sulfate-bearing volcanic rocks. Anhydrite (CaSO4) is rare in volcanic rocks and was recognized for the first time as a primary igneous phase in the 1982 El Chichón eruption products (Luhr et al. 1984). In El Chichón lavas, anhydrite forms euhedral phenocrysts surrounded by rhyolitic glass with no evidence of disequilibrium (Fig. 4a) but it is absent in more porous pumices. Its absence in some
Parat, Holtz, Streck
292
Continental arc; Lowenstern 1993 Continental arc, Stone et al. 1989
S
S (at%)
MORB; Francis 1990 MORB; Mathez 1976 MORB; Czamanske and Moore 1977 OIB; Stone and Fleet 1991 Island arc; Hattori 1996 Island arc; Nadeau et al. 2010
a.
po
10
90 Cu
Ni
80
20
cu vio pn
70
30
40
ccp
60 50
50
bn
40
60 70 Ni (at%)
10
20
30
40
50
60 Cu (at%)
b.
Fe
Fe (at%)
30
po
10
90 Cu
Ni
80
20
70
30
cu
40
60
pn
ccp
50
50
40
60 vio
70
30
80 10 Ni (at%)
20 20
30
40
50
60
70 Cu (at%)
Figure 3. Cu-Ni-Fe sulfide compositions. (a) Cu-Ni-S ternary plot of normalized atomic proportions in the sulfides projected from Fe. (b) Cu-Ni-Fe ternary plot of normalized atomic proportions in the sulfides projected from S. Sulfide compositions are from sulfides without exsolution texture, bulk analyses, or recon3 ccp: chalcopyrite, cu: cubanite, pn: pentlandstructed sulfide compositions (Nadeau et al. 2010).Figure bn: bornite, ite, po: pyrrhotite, py: pyrite, vio: violarite. MORB: mid-oceanic ridge basalt, OIB: oceanic island basalt.
Sulfur-bearing Magmatic Accessory Minerals
293
of the 1982 El Chichón eruptive products was interpreted to be due to dissolution by meteoritic water immediately subsequent to the eruption (Luhr et al. 1984). This highlights the high susceptibility of anhydrite to be erased from the geological record even if the rock otherwise may look quite pristine. Subsequent investigations identified primary anhydrite as inclusions in other minerals a. in a few recent, as well as in some older, intermediate to silicic arc-related anh eruptive products (Table 1). Anhydrite considered to be a primary igneous mineral has also been described from plutonic rocks ranging in composition ap from picritic gabbros to granitic rocks (Table 1). Anhydrite-bearing magmas are found in arc environments and usually contain mineral assemblages that record 50 µm hydrous, oxidized (>NNO) conditions and temperature between 650 and 900 °C (Table 1). In the volcanic and plutonic b. d. br examples listed in Table 1, pyrrhotite (or pyrite) or iss always is present with nsn anhydrite, with two exceptions: the Sutter Buttes andesite (Luhr 2008) and the Cajon Pass tonalite (Barth and Dorais 2000). Anhydrite often coexists with sulfur-rich hyn br apatite (from 50 ppm to >0.4 wt% S) in silicate melts, depending on temperature laz and fO2 conditions (Baker and Rutherford 1996). Particular attention has been paid 500 µm to the presence of anhydrite in volcanic systems in the last two decades because 10 mm c. an rim some recent explosive eruptions have produced one to two orders of magnitude ap glass more sulfur dioxide than could have been dissolved in the melt phase of the erupted magma. This is known as the “excess sulfur” problem and the breakdown of scapolite anhydrite has been evaluated as a possible source for this “excess” sulfur (Baker and Rutherford 1996; Rutherford and Devine 1996; Swanson and Kearney 2008).
Figure 4. Microphotographs of magmatic sulfates. (a) Backscattered electron images of isolated anhydrite (anh) crystals and S-rich apatite (ap) surrounded by vesiculated glass in 1982 El Chichón pumices (Luhr 2008); (b) Crystals of haüyne (hyn), lazurite (laz) and nosean (nsn) taken with transmitted light microscopy from haüynophyre lava from Mt. Vulture, Italy (Di Muro et al. 2004); (c) Thin section of scapolite in Figure 4 basanite from Chuquet-Genestoux, Chaîne des Puys (photo courtesy of P. Boivin) (Boivin and Camus 1981). an: anorthite, ap: apatite, br: outer opaque rim. {Figure 4a was reproduced with permission from Elsevier from Luhr (2008) Journal of Volcanology and Geothermal Research, Vol. 175, p. 394-407; Figure 4b was reproduced with permission from the Mineralogical Society of Great Britain and Ireland from Di Muro et al. (2004) Mineralogical Magazine, Vol. 68, p. 591-614, Fig. 2c}
Parat, Holtz, Streck
294
Magmatic minerals of the sodalite-group are haüyne (Na,Ca)4−8(Al6Si6(O,S)24)(SO4,Cl)1−2 and nosean Na8(Al6Si6O24)(SO4) (Fig. 4b). Haüyne and nosean are Si-poor, Al-rich tectosilicates that are distinguished chemically from pure sodalite (Na8(Al6Si6O24)Cl2) by the replacement of Cl by SO4 (Tracy 2003; Deer et al. 2004) (Fig. 5a). In addition, Ca is incorporated in various concentrations by nosean and especially haüyne (Fig. 5b). Haüyne and nosean are found as magmatic minerals in some alkaline (quartz-undersaturated) magmatic rocks and carbonatites (e.g., Oldoinyo Lengai, Dawson et al. 1994; Zaitsev et al. 2009). Furthermore, both haüyne and nosean have also been reported from intraplate trachyandesite, tephrite, phonotephrite and phonolite at several locations around the world (see typical examples in Table 1 and their composition in Fig. 5). Haüyne is a common mineral in rocks associated with alkaline volcanism from Central Italy (Table 1) (Cavarretta and Lombardi 1990). The Nosean phonolite of Wolf Rock, Cornwall, is well known for its abundant zoned phenocrysts of nosean (Harrison et al. 1977). The incorporation of both Cl and S in the minerals of the sodalite group is potentially useful to constrain the evolution of volatile concentrations in quartz-undersaturated magmatic systems. For example, Holm et al. (2006) noted that the Cl/SO4 ratio in haüyne increases with the Na/(Na+K+Ca) ratio, which is interpreted as reflecting the increase of the Cl/SO4 ratio in magmas with increasing degrees of differentiation from nephelinite to phonolite. However, use of Cl/SO4 in sodalite group minerals to constrain volatile concentrations in magmas remains Wörner and Schmincke 1984, Eifel Tracy 2003, Tahiti Bryan 2006, Canary Island Holm et al. 2006, Cape Verde Renzulli et al. 1998, Italy Stoppa et al. 2002, Italy De Fino et al. 1986, Italy Cavarretta and Lombardi 1990, Italy Di Muro et al. 2004, Italy + Zaitsev et al. 2009, Oldoinyo Lengai
a. 12 10
S wt%
8 6 4 +
2 0
0
0.5
1.0 1.5 Cl wt%
2.0
2.5 K
b.
K (at%)
70
30 Na
80
Ca
20
nosean haüyne
90
10
+ sodalite
Na (at%)
10
20
30
40
50
60
Ca (at%)
Figure 5. Sodalite-group mineral compositions. (a) Cl vs. S content (wt%) and (b) Na-Ca-K ternary diagram (at%) (after Lessing and Grout 1971).
Figure 5
Sulfur-bearing Magmatic Accessory Minerals
295
qualitative as long as no systematic investigations of the parameters controlling the ratio Cl/SO4 and the volatile fugacities in the fluid phase are available. silvialite (Ca,Na)4[(SO4,CO3)|(Al2Si2O8)3] is a sulfate-rich mineral of the scapolite-group (Na4Al3Si9O24Cl (marialite) - Ca4Al6Si6O24CO3 (meionite) series) (Fig. 4c). Silvialite has been found in some alkali-rich magmatic systems in which it precipitated from alkali basalt magma at 900-1000 °C and 0.8-1.2 GPa under high fSO2 and fO2 (Chappell and White 1968). Magmatic scapolite is rare and has been reported as phenocrysts in the Laacher See phonolite (Brauns 1914), in tephra of alkali basalt composition from the Massif Central, France (Fig. 4c) (Boivin and Camus 1981), in rocks of the Seqveika cinder cone, Hoggar, Algeria with amphibole + clinopyroxene + Fe-Ti oxide + apatite (scapolite with up to 1.38 wt% S, 3.20 wt% CO2) (Boivin and Camus 1981), in a latite dome in Arizona (scapolite with ~1.S74 wt% S, 1.75 wt% CO2, S/(S+C) of 0.58 (at%)) (Goff et al. 1982), and in a dyke from the West Greenland Tertiary igneous province with hyalophane + potassium feldspar + nepheline + analcime (Larsen 1981). Scapolite is also present as an interstitial primary igneous phase in cognate cumulate nodules from the alkali-basaltic Kula Volcanic Province in Western Turkey (Smith et al. 2009) and in upper-mantle garnet-granulite xenoliths hosted by olivine nephelinite, from McBride Province, North Queensland, Australia (Teertstra et al. 1999). The crystallization of sulfur-rich scapolite in magmas implies that high fSO2 and fCO2 are prevailing. The rarity of scapolite as a phenocrystic mineral suggests that high partial pressures of CO2 and SO2 are rare in the magmatic environment (Goff et al. 1982). When present, silvialite may be used to constrain T, P and fO2 of magmatic systems using experimental scapolite-plagioclase stability relations (Newton and Goldsmith 1976; Goldsmith and Newton 1977; Larsen 1981; Goff et al. 1982). s-bearing apatite (Ca5(PO4)3(F,Cl,OH)) is an important accessory mineral in oxidized magmas. Although sulfur is not incorporated into apatite as a major element, concentrations up to 0.8 wt% S and even higher have been reported (e.g., Luhr et al. 1984; Pallister et al. 1996; Streck and Dilles 1998; Tepper and Kuehner 1999; Parat et al. 2002; Broderick 2008; van Hoose et al. 2010). S-bearing apatite only crystallizes from oxidized magmas in which sulfur is present at least in part as S6+ because apatite incorporates sulfur as sulfate. Apatite has been reported with anhydrite in andesitic/dacitic arc-related magma (Luhr et al. 1984; Pallister et al. 1992; Parat et al. 2002) and with haüyne in alkaline intracontinental magma (Parat et al. 2011). The concentration of sulfur in apatite is correlated with the sulfur concentration in the melt at the time of apatite crystallization and, hence, it may be used as a proxy to trace the evolution of sulfur concentrations in magmas (Peng et al. 1997; Streck and Dilles 1998; Tepper and Kuehner 1999; Parat et al. 2002). A pre-requisite for this application is, however, that chemical exchange between apatite and melt is negligible after formation of apatite.
MInerAl StAbIlIty, PArAgeneSeS And MInerAl/Melt PArtItIonIng oF SulFur transition from sulfide to sulfate stability fields in silicate melts In magmatic systems at a given oxygen fugacity, the stability of sulfide and sulfate minerals is mainly dependent on the maximum sulfur concentration that can be dissolved in silicate melts. Thus, experimental data and models are required to predict the sulfur concentration of melts at sulfide saturation (SCSS) or at anhydrite saturation (SCAS). An overview of such data and models, and on the factors affecting SCSS and SCAS, is given by Baker and Moretti (2011, this volume). In addition to sulfur fugacity, T and P, sulfide and anhydrite saturation are known to be strongly dependent on oxygen fugacity and on the melt composition, especially the FeO and CaO contents (Ducea et al. 1999; O’Neill and Mavrogenes 2002).
Parat, Holtz, Streck
296
Pyrrhotite and anhydrite are the two most frequent minerals crystallizing in basaltic to rhyolitic systems. Figure 6 is based on a compilation of recent experimental data and can be used to estimate the sulfur concentration at which pyrrhotite and/or anhydrite are starting to crystallize from basaltic and intermediate to rhyolitic melts. Another model to estimate the crystallization conditions of pyrrhotite and anhydrite has been described by Luhr (1990), on the basis of the following reactions: 3FeS (po) + 2O2 (g) = Fe3O4 (mag)+ 3/2S2 (g)
(2)
CaSiO3 (cpx) + FeS (po) + 2O2(g) = CaSO4 (anh) + FeSiO3 (cpx)
(3)
CaSiO3 (cpx) + Fe3O4 (mag)+ 3/2S2(g)+ 4O2(g) = 3CaSO4 (anh)+ 3FeSiO3 (cpx)
(4)
For abbreviations of phases see Table 1. Luhr (1990) established a phase diagram showing the fO2 and fS2 range in which pyrrhotite and anhydrite are stable in El Chichón trachyandesite magma (Fig. 7). According to Figure 7, pyrrhotite and anhydrite only coexist along a univariant line, i.e., there is no domain at which both phases are coexisting. However, in natural systems (from peralkaline melts to andesites) both phases often coexist, as observed in natural rocks (Table 1) as well as in experiments (Fig. 8). Anhydrite coexists with pyrrhotite in experimental products for log fO2 close to NNO+1.5 at 800 °C and for NNO+0.5 < log fO2 < NNO+2.5 at 950 °C (Fig. 8). The oxygen fugacity at which both minerals do coexist corresponds to the fO2 range at which both S2− and S6+ species are observed in the silicate melt (see Wilke et al. 2011, this volume).
Stability range of sulfide phases in magmatic systems Fe sulfides. A detailed review of phase relations in the Fe-S system at high temperature is given by Fleet (2006) and this topic is not developed here. Whitney (1984) determined the stability fields of phases in the system Fe-O2-S2-SiO2 (Fig. 9) and the data are more relevant for natural silicate systems. The phase relationships are useful to understand, at least qualitatively, the evolution of the stability of Fe sulfide phases in magmatic systems as a function of T, fO2, and fS2. The phase diagrams of Whitney (1984) clearly show that pyrrhotite is stable over a large range of temperature (600-900 °C), whereas pyrite is not expected to be a stable phase at magmatic temperatures (>700 °C) and liquid sulfur is expected for high fS2 (Fig. 9). Using experiments on rhyolitic systems, Clemente et al. (2004) analyzed oxygen in sulfides with the electron microprobe and the results suggest that for log fO2 < NNO the sulfide phase is pyrrhotite, whereas above NNO, a Fe-S-O immiscible liquid ± pyrrhotite is formed. The FeS-O liquid was identified in the investigated temperature range of 800-1000 °C. The presence of oxygen in the sulfide phase may be used as evidence for a high temperature immiscible liquid phase (non-stoichiometry is more compatible with a liquid) at moderately reduced conditions. The identification of a liquid Fe-S-O phase is, however, difficult in quenched experimental samples and may be even harder to prove in natural samples because of the rapid crystallization of such a liquid during cooling. Often, the spherical to ellipsoidal shape of an inclusion is used as evidence for the existence of an immiscible liquid at high temperature but there are also numerous examples in which sulfides with rounded and wormy shapes were clearly entrapped as a solid (based on its composition, phase diagrams and independent temperature estimates). The determination of the exact composition of pyrrhotite in synthetic and natural phases can be extremely useful to constrain the sulfur fugacity, a crucial parameter which is difficult to estimate in magmas. Toulmin and Barton (1964) developed an equation relating the composition of pyrrhotite to fS2 and T: log= fS2
( 70.03 − 85.83N FeS )
1000 1/ 2 − 1 + 39.30 (1 − 0.9981N FeS ) − 11.91 T
(5)
where NFeS is the mole fraction of FeS in the system FeS-S2. This equation is widely used to de-
Sulfur-bearing Magmatic Accessory Minerals 2.0 1.8
S in melt (wt%)
1.6 1.4
297 a.
Liu et al. 2007; 1150-1430°C, 1 GPa Botcharnikov et al. 2011; 1050°C, 200 MPa Moune et al. 2009; 1050°C, 300 MPa Beermann 2010; 1050-1250°C, 200 MPa Jugo et al. 2005; 1300°C, 1 GPa
1.2 1.0 0.8 0.6 0.4
basalt
0.2 0 0.50 0.45
S in melt (wt%)
0.40 0.35
-3
-2
-1
0
1
2
3
ΔNNO b.
Liu et al. 2007; 1050-1450, 1 GPa Botcharnikov et al. 2011; 1050°C, 200 MPa Moune et al. 2009; 1050°C, 300 MPa Parat et al. 2008; 850-950°C, 400 MPa Carroll and Rutherford 1987; 850-1025, 200 MPa Luhr 1990; 800-1000°C, 200-400 MPa
0.30 0.25 0.20 0.15 0.10 0.05 0
andesite-rhyolite -3
-2
-1
0
1
2
3
ΔNNO
Figure 6. S content in silicate melt as a function of fO2 (expressed relative to the log fO2 of the Ni-NiO equilibrium (DNNO), fO2 in bar) in melts saturated with respect to pyrrhotite (white symbols) and anhydrite (black symbols) (gray symbols when both pyrrhotite and anhydrite present). (a) Basaltic composition and Figure 6 (b) rhyolitic and intermediate compositions. All experimental runs were performed in the temperature range 800-1300 °C and pressure range 200 MPa-10 GPa, at fluid-saturated conditions with 1 wt% S added except for basaltic compositions from Jugo et al. (2005) and Liu et al. (2007) where runs were performed with 1-2 wt% S and 0.5 wt% S added, respectively, at dry and fluid-undersaturated conditions.
termine fS2 in high temperature systems but should be used with caution (for a detailed review of the method, see Fleet 2006). The mole fraction of FeS in natural pyrrhotite may change during or after cooling (one may note the rapid diffusion of iron in pyrrhotite; see references in Schuessler et al. 2007) and small changes in the Fe concentration of pyrrhotite affects the calculated fS2. As an example, in a single volcanic rock sample, Parat et al. (2002) observed a variation of NFeS from 0.92 to 0.95, leading to a variation of fS2 from 0.04 to 3 bars (assuming a temperature of 900 °C). It should be also noted that the model of Toulmin and Barton (1964) was established on the basis of experiments performed at temperatures between 325 and 900 °C. Cu-Ni-bearing sulfides. The stability of sulfides in the systems Cu-Fe-S and Ni-Fe-S is reviewed by Fleet (2006). The main magmatic sulfides consist of mss in the system Ni-Fe-S (S > 55 at%, Figs. 10 and 11) and iss in the system Cu-Fe-S (at 850 °C, Fig. 12). At high
Parat, Holtz, Streck
298 NFeS 0.90
0
0.91 0.92 0.93
-2
pyrrhotite
2
anhydrite
0.94 0.95
1
0.96
3
-6 -8
fayalite
log f S2
-4
magnetite
-12 NNO
-14
NNO+1
NNO+3.7
-12 log f O2
hematite
-10
Figure 7. Fugacity diagram of S2 vs. O2 calculated for the quartzclinopyroxene-amphibole-vaporsaturated system O-H-S-Si-Ca-Fe-Ti at 800 °C and 200 MPa from Luhr (1990) (fS2 and fO2 in bar). See text for explanation of reactions 1-3. Stars indicate estimated S2 fugacities for magnetite-anhydrite equilibria at the manganosite-hausmanite (MNH) and magnetite-hematite (MTH) buffers (NNO+3.7 and NNO+4.4, respectively). NNO: Ni-NiO buffer. Tie dashed lines are for different NFeS values of pyrrhotite (mole fraction of FeS in the system FeS-S2).
-10
5 4
Figure 7
an
ΔNNO
3 2
an+po
1 NNO 0
po
-1 -2 -3 700
750
800
850
andesite trachyandesite
900 T(°C)
950
1000
1050
metaluminous rhyolite peralkaline rhyolite
Figure 8. Pyrrhotite (po) and anhydrite (an) stability as a function of temperature and oxygen fugacity (expressed relative to the log fO2 of the Ni-NiO equilibrium (∆NNO), fO2 in bar) from experiments performed at 150-400 MPa, water-saturated conditions and 1 wt% S added. Data sources: Trachyandesite from Luhr Figure 8 (1990); metaluminous andesite from Parat et al. (2008), peralkaline rhyolite from Scaillet and Mcdonald (2006) and metaluminous rhyolite from Baker and Rutherford (1996) and Clemente et al. (2004). Black, gray and white symbols for pyrrhotite, pyrrhotite+anhydrite, and anhydrite in experimental runs, respectively.
temperature (1000 °C and above), the system Cu-Ni-Fe-S contains large regions of mixed liquid and the dominant Cu-Ni-bearing sulfide phase is most probably a liquid (Craig and Kullerud 1969) (Fig. 13). To our knowledge, no experimental results are available to constrain the saturation of a silicate melt with respect to compositionally variable Cu-Ni-bearing sulfide (solid or liquid). Natural case studies reveal some limits on the concentration of sulfur in silicate
Sulfur-bearing Magmatic Accessory Minerals 2
900°C
2
FeSO4
0
-4
log f S2
log f S2
po
-2 mag
-6 hem
-8
fay
-4
mag
-6 hem
-8
fay
-10
-10 -14
700°C py
-12 -10 log f O2
-6
2
-4
hem fay
-6
py
FeSO4
-12 -10 log f O2
-8
-6
Figure 9. Isothermal, silicasaturated, log fO2 vs. logfS2 diagrams for the system FeO2-S2-SiO2 at ambient pressure from Whitney (1984) (fS2 and fO2 in bar). S(L) = sulfur liquid, po: pyrrhotite, fay: fayalite, mag: magnetite, hem: hematite, py: pyrite.
po mag
hem
-8 -10
-14
-8
S(L)
-6
mag
-10 -12 -16
600°C
-12 -10 log f O2
-2
po
-6
-14
0
FeSO4
-4
-8
-12 -16
S(L)
0 -2
-8
log f S2
-12 -16
log f S2
S(L) FeSO4
0
po
-2
2
800°C
299
-12 -16
fay
-14
-12 -10 log f O2
-8
-6
melts in equilibrium with a sulfide liquid. A sulfide liquid may be present at relatively low sulfur concentrations in tephriphonolitic melts (140-550 ppm S, Renno et al. 2004). In mid-oceanic Figure 9 ridge basalts, Czamanske and Moore (1977) reported also the presence of sulfide globules in glassy rims of pillows. In the case of oceanic island basalts from Hawaii, Cu-Ni-bearing sulfide liquids (see composition in Fig. 3) were found to coexist with basaltic to andesitic melts containing up to 0.2 wt% S (Stone and Fleet 1991). In the absence of adequate phase equilibria for complex natural systems, the crystallization of a Cu-Ni-sulfide liquid can be modeled, at least qualitatively, by using the quaternary phase diagrams (Cu-Ni-Fe-S) of Craig and Kullerud (1969) (Fig. 13): at the high temperatures typical of mid-oceanic ridge basalt (MORB) and oceanic island basalt (OIB), there is a large field of Cu-Ni-Fe-S compositions at which a sulfide liquid can exist (see plot at 1000 °C, Fig. 13a), whereas at the lower temperatures typical of arc magmas the compositional field of mixed CuNi-Fe-S melts is very restricted, and mss and iss are expected (see plot at 850 °C, Fig. 13b). Stone et al. (1989) used quaternary phase diagrams for the Mount Shasta tephra and concluded that cooling to approximately 1100 °C fosters crystallization of mss, consuming nearly all the Ni and leaving a residual sulfide liquid enriched in Cu from which iss may have crystallized later. Progressive cooling below 610 °C causes exsolution of pentlandite from the mss which eventually converts to pyrrhotite. Copper within the mss is partitioned into pentlandite or exsolves as chalcopyrite. Other examples using the phase equilibria in the simplified Cu-
Parat, Holtz, Streck
300
mss
Ni-Fe-S system or subsystems to discuss the crystallization of sulfide liquids in magmatic systems can be found in Stone and Fleet (1991) or Renno et al. (2004). The compositions of sulfides in arc magmas (Fig. 2) show that sulfides belong either to the Ni-Fe-S or to the CuFe-S system and thus phase equilibria in the subsystems Ni-Fe-S and Cu-Fe-S can be used to predict which solid sulfide crystallized from a sulfide melt. On the other hand, in mid-oceanic ridge basalt and oceanic island basalt systems, the Cu-Ni-Fe-S system needs to be taken into account because sulfides contain both Ni and Cu.
Figure 10. Phase diagram of the system Ni-Fe-S for Fe = Ni, S = 35-55 at% and T = 450-1100 °C after Sugaki and Kitakaze (1998), modified by Fleet (2006). mss: monosulfide solid solution; hpn: highform pentlandite, pn: pentlandite, L: liquid. Figure 10
Sulfide liquids exsolving early from mafic silicate melts are enriched in Ni and Cu (e.g., up to 36 wt% Ni and 18 wt% Cu in sulfide liquid, Brenan 2003), suggesting that the partition coefficients of these elements between sulfide liquid or mineral and silicate melt are high. Available data from natural systems in which Ni-Cu sulfides crystallize from Ni-Cu-bearing melts suggest partition coefficients for Ni (DNisulfide/silicate melt) and Cu (DCusulfide/silicate melt) in the range of 100 to 1000 and that the partition coefficients increase with magmatic differentiation, i.e., with increasing silica content of the
1100
1000 liquid mss+L
800
865
hpn+L γ(Fe,Ni) +L
746
hpn + mss
hpn
temperature (°C)
900
700 γ(Fe,Ni)+hpn
615
600
584
pn pn + mss
γ(Fe,Ni)+pn
500 35
40
50
45
sulfur at Fe=Ni (at%)
FeS
S
S (at%)
40 mss
60 NiS
hpn
60
liquid
Ni
Fe
40 30
70
20
80
10
90 Fe (at%) αFe 10
20
30
40
50 60 γ(Fe, Ni)
70
80
90
Ni (at%)
Figure 11. Phase relations in the metal-rich range of the Ni-Fe-S system at 850 °C after Sugaki and Kitakaze (1998), modified by Fleet (2006). Dotted lines are conodes, i.e., connect coexisting compositions. mss: monosulfide solid solution; hpn: high-temperature modification of pentlandite. See Figure 2 for comparison to natural Ni-Fe-sulfides. Figure 11
Sulfur-bearing Magmatic Accessory Minerals
301
S
S (at%)
30
70 py
iss+py+S(L)
40
po+py
ccp tal
bn+S(L) bn+iss
60
iss+po+py
iss+py
bn+iss +S(L)
bn
mh
cb
iss+po
hc iss
30
50
tr tr+Fe
40
bn+tr+Fe
bn
Cu (at%)
po
bn+po(tr)
bn+iss+po
Cu+bn10 Cu+bn+Fe20
Fe
60
iss+S(L)
50
Cu
40
50
60 Fe (at%)
Figure 12. Phase relations in the central part of the Cu-Fe-S system at 600 °C after Cabri (1973), modified by Fleet (2006). Solid lines mark the extensive fields of solid solution centered on bornite (bn), intermediate 12 pyrrhotite (po). Solid squares are stoichiometric compositions for bornite (bn), solid solutionFigure (iss), and chalcopyrite (ccp), talnakhite (tal), mooihoekite (mh), haycockite (hc), cubanite (cb), troilite (tr), and pyrite (py). See Figure 2 for the composition of natural Cu-Fe-sulfides. See Figure 2 for comparison to natural Ni-Fe-sulfides.
S
a.
S
b.
1000°C
850°C vs
vs (Ni, Fe)3S2 L
L
bn
Cu
Ni
ccp
s
Ni
ms
Cu
L
L ms s
bn
γ
γ
Fe
Fe
α
Figure 13. Cu-Ni-Fe-S phase equilibria at (a) 1000 °C, after Raghavan (2006) and (b) 850 °C, after Craig and Kellerud (1969). Tie lines to S liquid are omitted for clarity. Note that at 850 °C, the region of homogeneous sulfide liquid (grey region) has diminished compared to 13 1000 °C, mss spans the entire Ni-Fe-S face of the Figure system, and ccp and (Ni,Fe)3S2 are stable phases at 850 °C. bn: bornite; ccp: chalcopyrite; L: liquid; mss: monosulfide solid solution; vs: vaesite.
302
Parat, Holtz, Streck
magma (DNisulfide/silicate melt = 105 to 350 and DCusulfide/silicate melt = 178 to 1006 from komatiite to fractionated tholeiites: Francis 1990; Dsulfide/silicate melt = 274 for Ni, 245 for Cu and 80 for Co at 1255 °C in basaltic melt: Rajamani and Naldrett 1978). These partition coefficients have been determined from mineral assemblages in which Ni and Cu are present at trace levels and must be used with caution for mineral assemblages in which they are at wt% levels (Henry’s law is not valid). Considering that the Cu and Ni concentrations in magmas are lower than S concentrations, the crystallization of sulfides, even in very small proportions, leads to a strong decrease of Cu and Ni in the residual melts. Thus, sulfides crystallizing from magmatic systems become depleted in Cu and Ni with progressive crystallization (Francis 1990). Molybdenite. The stability range of molybdenite in silicate melts has not been studied in detail. The dissolution reaction of molybdenite in the melt (Lowenstern et al. 1993) can be described as: (6) 2MoS2 (mo) + 3O2 = 2MoO3 (g) + 2S2 indicating that fS2 and fO2 are important parameters controlling the occurrence of molybdenite in magmatic systems. Results from thermodynamic modeling by Audétat et al. (2011) show that, for a given oxygen fugacity buffer, molybdenite solubility increases with increasing temperature. Molybdenite has been observed in natural rocks in the log fO2 range from NNO1 to NNO+1.5. Thermodynamic constraints imply that molybdenite-saturated rhyolites are most probably also saturated with pyrrhotite. In this case, the Mo concentration in the melt (as measured in melt inclusions) can be used to estimate fO2 and fS2 in such magmas. Molybdenite has not been reported very often in natural rocks. However, this may be due to the fact that it was overlooked. If, indeed, this mineral is more common than previously thought, as suggested by the high number of positive identifications (13 out 27 examples) in the study by Audétat et al. (2011), the molybdenite-pyrrhotite fO2 and fS2 barometer may turn out to be very useful for estimating oxygen and sulfur fugacities of natural systems.
Sulfates and sulfate-bearing minerals Anhydrite stability. The stability of anhydrite in magmatic system has been discussed above and is mainly dependent on the oxygen fugacity and the concentration of sulfur in melts (assuming that Ca activity in the silicate melts is not the main parameter controlling anhydrite stability) (Figs. 6, 7, and 8) (Carroll and Rutherford 1988; Luhr 1990; Baker and Rutherford 1996; Clemente et al. 2004; Parat et al. 2008). Experimental data indicate that anhydrite is stable at log fO2 > NNO+1 for T = 700-1000 °C and P = 150-400 MPa (Fig. 8) over a large range of sulfur fugacity (Fig. 7). haüyne and nosean stability. Experimental studies of Van Peteghem and Burley (1963) demonstrated that complete solid solution exists between nosean and haüyne (6NaAlSiO4·Na2SO4 - 6NaAlSiO4·2CaSO4) at 600 °C and 100 MPa PH2O but only limited solid solution occurs in the systems sodalite-nosean (6NaAlSiO4·2NaCl - 6NaAlSiO4·Na2SO4) and sodalite-haüyne (6NaAlSiO4·2NaCl - 6NaAlSiO4·2CaSO4) under the same conditions. Considering that most natural minerals of the sodalite groups can be classified as nosean and haüyne (Fig. 5), a complete solid solution can be expected at magmatic conditions. The discussion of the stability of anhydrite vs. pyrrhotite in silicate melt indicates that fO2 is the dominant factor controlling the formation of sulfate or sulfide in mafic to rhyolitic (Cabearing) systems. In alkali-rich (relative Ca-poor) systems, sulfate-bearing minerals such as nosean or haüyne are favored not only by high oxygen fugacity but also by low silica activity and high peralkalinity (molar (Na + K)/Al >1) (Stormer and Carmichael 1971). A typical reaction accounting for the role of silica activity on the stability of nosean vs. sulfide has been given by Stormer and Carmichael (1971): Na8Al6Si6O24SO4 (nsn) + Fe3O4 (mag)+16SiO2 = FeS (po) + 2NaFeSi2O6 (aeg) + 6NaAlSi3O8 (ab) + 2O2
(7)
Sulfur-bearing Magmatic Accessory Minerals
303
According to this reaction, Na- and sulfate-bearing minerals also can be formed at relatively low fO2 because lower silica activity has also a strong control on the reaction (if silica activity decreases, the reaction will move to the left, provided that fO2 remains constant). This has relevance especially for silica-undersaturated systems. A case study in which the S-bearing stable phase changes from sulfide to sulfate as a result of compositional changes in the silicate melt has been described by Bryan (2006) for alkaline rocks from Tenerife (Canary Islands). There, pyrrhotite has been reported in mafic volcanic rocks (Wolff and Storey 1983), but as the magmas evolve to phonolitic compositions, SiO2 and/or Al2O3 activities decrease in response to a combination of kaersutite (a Ti-rich amphibole) crystallization from tephriphonolite to phonolite, followed by the crystallization of alkali feldspar and biotite in the phonolite magmas. Consequently, a reaction similar to that described above is driven to the left and sodalite was formed in the evolved Tenerife phonolites. Few experimental works have been performed to constrain the stability of haüyne at magmatic conditions in natural systems. Berndt et al. (2001) and Andújar et al. (2008) performed phase equilibrium experiments with phonolitic melt from Laacher See Volcano (Germany) and Tenerife (Canary Islands), respectively. Haüyne is stable at 200 MPa and 760-840 °C and 300-400 MPa and 760 °C with biotite, alkali feldspar ± plagioclase, sphene and magnetite in Laacher See phonolite (Berndt et al. 2001) and at 100 MPa and 700-875 °C with biotite, magnetite, clinopyroxene, feldspar, sphene, and apatite in phonolite from Tenerife (Andújar et al. 2008). Its crystallization is favored by high oxygen fugacity (above ~NNO+2.3) and its stability field increases with decreasing pressure and water activity. Because the sodalite group minerals incorporate S and Cl, these minerals are potentially interesting for determining fSO2/fCl2 of natural systems. On the basis of the exchange reaction: 2Na4Al3Si3O12Cl (sod) + SO2 + O2 = Na8Al6Si6O24(SO4) (nsn) + Cl2
(8)
Stormer and Carmichael (1971) estimated the limiting fugacity ratio fSO2/fCl2 for nosean and sodalite to be stable (assuming that SO2 is the dominant S-O species at magmatic temperatures; Fig. 14). Unfortunately, Figure 14 cannot directly be applied to deduce fSO2/fCl2 ratios in natural systems because natural sodalites also contain Ca and K (Fig. 5), the effect of which is not known. Figure 14 can only be used qualitatively for evaluating the role of fO2 and of the fSO2/fCl2 ratio for the stability of sodalites (cf. Tracy 2003; Di Muro et al. 2004). Additional experimental calibrations would be particularly useful to constrain the variation of S and Cl concentrations
12 Nosean
11
fO2
10
10-10
9
FMQ NN O
f Cl2
log
f SO2
8 7 6 5
10-5
4 3
HM
2
Sodalite
1 0
800
900 T (°C)
1000
1100
Figure 14. Equilibrium between nosean and sodalite plotted in terms of the ratio of SO2 to Cl2 and temperature for several oxygen fugacities (after Stormer and Carmichael 1971). Dashed curves for constant oxygen fugacities (fO2 in bar). Solid curves for oxygen fugacities buffered by the following assemblages: FMQ: fayalite-magnetite-quartz; NNO: nickel-nickel oxide; HM: hematite-magnetite.
304
Parat, Holtz, Streck
of compositionally zoned sodalites and trace degassing processes in magma chambers (e.g., Di Muro et al. 2004). scapolite minerals. Scapolite minerals are rare in magmatic rocks, whereas they are common minerals in metamorphic rocks (e.g., granulite, Lovering and White 1964). The stability of scapolite and plagioclase has been investigated at high T and P in the system NaAlSi3O8CaAl2Si2O8-CaCO3-CaSO4 by Goldschmidt and Newton (1977). Ca-SO4-rich scapolite (silvialite) has been crystallized experimentally in melts with a wide range of SiO2, CaO, and Na2O contents, at temperatures above 825 °C and pressures ranging from 300 MPa to 1.5 GPa (Newton and Goldsmith 1976; Goldsmith and Newton 1977). The equilibrium reaction is: 3CaAl2Si2O8 (an) + CaSO4 = Ca4Al6Si6O24SO4 (s-mei)
(9)
In a P-T diagram, the mineral assemblage anorthite + anhydrite was found to be stable up to 700 °C at 1.5 GPa and up to 1050 °C at 100 MPa. At higher P-T conditions, sulfate meionite was found to be the stable phase (Newton and Goldsmith 1976; Goldsmith and Newton 1977). s-bearing apatite. Solubility models for apatite in silicate melts have been developed by Watson (1980) and have been improved to account for the role of excess aluminum (i.e., formation of AlPO4 complexes in the silicate melt) by Pichavant et al. (1992) and Bea et al. (1992). Similar to S-free apatites crystallizing at reducing conditions and/or from S-free systems, S-bearing apatites are stable over a large range of temperature and pressure (Peng et al. 1997, 2008; Parat and Holtz 2004, 2005). Apatite incorporates sulfur as sulfate (S6+) and therefore S-bearing apatite crystallizes at high oxygen fugacity in the stability field of anhydrite, i.e., log fO2 > NNO+1. In general, a colinearity of Na with S and of Si with S at an atomic ratio of one is consistent with the S exchange reactions S6+ + Na+ ↔ P5+ + Ca2+ (Liu and Comodi 1993) and S6+ + Si4+ ↔ 2P5+ (Rouse and Dunn 1982). In REE- (Rare Earth Elements) free systems, the sulfur exchange reaction proposed by Liu and Comodi (1993) was confirmed by experimental data crystallization experiments with REE-free compositions (Parat and Holtz 2004; Parat and Holtz 2005). In REE-bearing systems, additional coupled exchange reactions involving REE can occur (Streck and Dilles 1998; Tepper and Kuehner 1999; Parat et al. 2008). The nonzero intercept and the slope slightly different from that expected for the substitution involving S, Si, Na and REE ((S6+ + Ce3+):(Na+ + Si4+) = 1) in Figure 15 show, however, that a single complex substitution mechanism (e.g., 2P5+ + 2Ca2+ ↔ S6+ + Si4+ + Na+ + REE3+) does not control the incorporation of sulfur but several independent substitutions are operative (Parat et al. 2011), such as: P5+ + Ca2+ ↔ S6+ + Na+ 2P ↔ S + Si 5+
6+
4+
(10) (11)
2Ca2+ ↔ Na+ + REE3+
(12)
P5+ + Ca2+ ↔ Si4+ + REE3+
(13)
with the Ca-site accommodating large cations (e.g., Na+, REE3+) and the P-site accommodating small highly charged cations (e.g., S6+, Si4+) (Piccoli and Candela 2002). Apatite incorporates sulfur as a trace element and is thus an excellent tracer of sulfur evolution in magmas. The sulfur content in apatite increases with increasing S-content in melt (Parat and Holtz 2004; Parat and Holtz 2005) (Fig. 16). The partitioning of sulfur between apatite and melt was first investigated experimentally by Peng et al. (1997) who showed that sulfur partitioning is dependent on temperature and pressure. Oxygen fugacity was found not to be an important parameter for their investigated range of oxidizing conditions (∆log fO2 = NNO+3.7 to NNO+4.5), which is nevertheless rather narrow and far beyond the sulfide/sulfate transition in melts. The sulfur contents in apatite reported by Peng et al. (1997) are, however,
Sulfur-bearing Magmatic Accessory Minerals
0.40
Natural trachyte Natural tephrite-phonolite Natural Huerto andesite Exp. Huerto andesite Exp. synthetic rhyolite
0.35
Si 4++ Na +
0.30
305
1:1
0.25 0.20 0.15 0.10 0.05
4+
2P5++ 2Ca 2+ = S 6+ + Si + Na + + REE 3+
0.00 0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
S 6+ + REE 3+
Figure 15. Correlation Figure 15 of ionic (S + Na) with ionic (S + REE) in apatite (n = 186). The superimposed slope 1:1 representing coupled ionic exchange reaction: 2P5+ + 2Ca2+↔ S6+ + Si4+ + Na+ + REE3+ (Tepper and Kuehner 1999) (see text). Apatites in natural andesite, trachyte, and tephrite-phonolite are from Parat et al. (2005), Parat et al. (2011), and Stoppa and Liu (1995), respectively. Apatites in phase equilibria experiments with andesite and rhyolite are from Parat et al. (2008) and Parat and Holtz (2004; 2005), respectively.
sulfate under -saturated
↑0.84
↑1.04
1.40
sulfate-saturated
↑
0.73
0.4 S ap = 0.0629 * ln S melt + 0.4513 (r2=0.68)
0.3
0.95
↑
0.2
↑
S in apatite (wt%)
0.5
↑0.92
↑
0.6
0.52
0.1 0
0
0.05
0.1 S in melt (wt%)
0.15
0.2
Crystallization experiments carbonatite; 1350-1380°C, 4-6 GPa (Hammouda et al. 2010) rhyolite; 900-1100°C, 200 MPa (Parat and Holtz 2004, 2005) andesite; 850-900°C, 400 MPa (Parat et al. 2008) dacite; 760°C, 220 MPa (Baker and Rutherford 1996) trachyandesite; 800-950°C, 200-400 MPa (Peng et al. 1997) Natural samples trachyte (Parat et al. 2011)
Figure 16. Sulfur content in apatite and melt from crystallization experiments and in situ data from natural samples. All experiments have been performed at high fO2. ap: apatite.
Figure 16
306
Parat, Holtz, Streck
relatively high compared to the sulfur contents reported in other experimental works performed at oxidizing conditions (NNO+1 to NNO+3.7: Baker and Rutherford 1996; Parat and Holtz 2004, 2005; Parat et al. 2008) (Fig. 16) as well as in natural S-rich apatites (e.g., Luhr et al. 1984; Pallister et al. 1996; Streck and Dilles 1998; Tepper and Kuehner 1999; Parat et al. 2002) and most probably reflect kinetic problems occurring during the experiments (Peng et al. 1997; Parat and Holtz 2004; Parat et al. 2011). Sulfur-rich melt inclusions in natural S-rich apatite are scarce and data are only reported for a trachytic composition from La Palma (Parat et al. 2011). The sulfur partition coefficient (DSapatite/silicate melt) estimated from the composition of apatite and trachytic melt decreases from 9.1 to 2.9 with increasing S of the melt inclusion (0.02 to 0.10 wt% S). The combination of experimental data and in situ measurement on natural samples (melt inclusion in apatite) reveals that the S partition coefficient (DSapatite/silicate melt) tends toward a value of 2 for high S content in the silicate melt (>0.08 wt% S). The partitioning behavior does not follow Henry’s Law. The experimental relationship can be expressed as: Sapatite (wt%) = 0.0629 × ln Smelt (wt%) + 0.4513 (r2 = 0.68)
(14)
considering natural and experimental data from Baker and Rutherford (1996), Parat and Holtz (2004, 2005), Parat et al. (2008), and Parat et al. (2011) (Fig. 16). One should note that this relationship is poorly constrained above 0.1 wt% S in the melt. Hammouda et al. (2010) performed crystallization experiments with S-rich apatite and S-rich carbonatitic melt (S = 0.52-1.4 wt%) but experimental data cannot be clearly correlated to those determined in silicate melt (Fig. 16). Most experimental and natural data cannot clearly constrain the effect of T, P and melt composition on the sulfur partitioning between apatite and melt. Nevertheless these parameters are important because, for a given fS2, the S content in silicate melts is a function of temperature, pressure, oxygen fugacity and melt compositions (e.g., Luhr 1990; Ducea et al. 1994; Clemente et al. 2004; Parat et al. 2008; Webster et al. 2009; Baker and Moretti 2011, this volume), which in turn affects the partition behavior (DSapatite/silicate melt) and thus reflects some influence of these intensive parameters. Up to now, experimental works in silicate systems were not able to reproduce the high S content observed in some natural apatites (0.4 to 0.8 wt% S; e.g., Streck and Dilles 1998; Parat et al. 2002; Streck et al. 2007). The close association of anhydrite and apatite may be a prerequisite for the formation of S-rich apatite (van Hoose et al. 2010; Parat et al. 2011) and the concomitant crystallization of anhydrite and apatite has been suggested in some natural case studies from petrographic observations of mutual inclusions of anhydrite and apatite (e.g., El Chichón and Mount Pinatubo; Luhr 2008) (Fig. 4a). Assuming that anhydrite and apatite are present close to each other in magmas, the formation of S-rich apatite may be related to local high S concentration in melts close to the interface with apatite and anhydrite. Dissolution of anhydrite, which may be caused by a change in magma storage conditions (e.g., degassing), could lead to a concomitant crystallization of apatite from a melt locally enriched in anhydriteforming elements (especially sulfur). Such a process would also explain the strong oscillatory variation of S concentration in apatites (Fig. 17) (e.g., Streck and Dilles 1998; Parat et al. 2002).
tHe IMPortAnce oF S-beArIng AcceSSory MInerAlS For decIPHerIng MAgMA reSerVoIr ProceSSeS Up to now, the use of sulfur-bearing accessory minerals as a tool to decipher petrogenetic processes related to the evolution of sulfur in a magma body is minimal compared to data obtained from melt inclusions. The main reason is that often sulfur-bearing minerals are not preserved or are altered. Most attention for the last ~15 years was given to S-bearing apatite and
Sulfur-bearing Magmatic Accessory Minerals
307
S wt% 0.24
0.06
0.13
0.08 0.15
grain margin
0.31
0.21
0.10 0.15 0.33 0.33
10 µm
0.06
Figure 17. Sulfur-zoned apatites from plutonic rocks of the Yerington batholith (Streck and Dilles 1998). Left panels shows S-Kα X-ray intensity map of two apatites measured with en electron microprobe; lighter shades correspond to higher count rates thus higher concentrations of S; outer grain margins of apatites are defined by P-Kα element map acquired simultaneously. Right panels shows same apatites with quantitative S (wt%) microprobe analyses with analysis spot size of ~5 μm (see Streck and Dilles (1998) for analytical details).
0.32
this mainly started subsequent to the documentation of such apatites in the magmatic products of the anhydrite-bearing and sulfur-rich eruptions of El Chichón (1982) and Mount Pinatubo Figure 17 (1991) (e.g., Pallister et al. 1995). These examples demonstrate that the sulfur content of apatite could be a function of the sulfur-content of the melt and that it could be used as tracer for the evolution of sulfur at oxidizing conditions (e.g., Baker and Rutherford 1996). In the last 15 years, several studies have routinely investigated the sulfur concentrations in apatite and used the results for interpreting pre-eruptive sulfur evolution histories within the magmatic reservoir systems. For example, strong variations of S contents in apatite (<0.04 to 0.56 wt% S), and concentric S zoning within individual apatite crystals (Fig. 17), as well as the overall decrease of S in apatite from intermediate to silicic magmas of the Yerington batholith led Streck and Dilles (1998) to postulate cryptic anhydrite crystallization (anhydrite has not yet been found) reducing melt sulfur content and leading to later porphyry-copper mineralization upon degassing and cooling. Nearly identical compositional ranges and zoning style of apatite compared to apatite from the Yerington batholith were found by Parat et al. (2002) in an Oligocene andesite from the San Juan volcanic field. In this case, anhydrite could be detected, but was only preserved as inclusions in amphibole, which was useful to correlate the S zoning patterns with the S evolution of this andesitic magma. It is emphasized that the S content of apatites coexisting with anhydrite does not need to be high. Crystallization of apatite with rather low-S content (~0.06 wt% S) at anhydrite saturation was documented in the form of an apatite-hosted anhydrite inclusion by Tepper and Kuehner (1999) in a plutonic rock from the Idaho Batholith (Table 1). This matches findings at El Chichón (Luhr 1990) and Mount Pinatubo (e.g., Pallister et al. 1996; van Hoose et al. 2010), in which anhydrite is a stable phase and where melt sulfur concentrations are low, i.e., low sulfur solubility in silicic melts can lead to anhydrite being a stable phase despite low sulfur concentrations in melt. From the examples discussed above and other studies, it emerges that sulfur concentrations in apatite from intermediate to silicic magmas are generally low (≤ 0.12 wt% S) whereas those from associated mafic magmas at comparable fO2 are generally higher (e.g., Humphreys et al. 2006; van Hoose et al. 2010), which can be attributed to the high S melt concentrations due to higher sulfur solubility at higher temperatures and in mafic melt compositions. Low sulfur contents of apatite from magmas of intermediate to silicic composition indicate low sulfur concentrations in the melts (if fO2 is high enough to stabilize mainly S6+ in the melt). However, intermediate (0.1-0.3 wt% S) to high (> 0.3 wt% S) S contents in apatite are ubiquitous in silicic magmas as well (Peng et al. 1997; Streck and Dilles 1998; Parat et al. 2002; Humphreys
308
Parat, Holtz, Streck
et al. 2006; Streck et al. 2007; Broderick 2008; Dietterich and de Silva 2010; van Hoose et al. 2010). This observation, combined with the apparent non-systematic distribution of anhydrite could lead to the conclusion formulated by Peng et al. (1997), emphasizing that “apatite fails as an indicator of ancient S-rich eruptions”. There are several points that suggest this view is too pessimistic. The apparent failure could be ascribed to source and timing of the sulfur supply (Westrich and Gerlach 1992; Di Muro et al. 2008; Dietterich and de Silva 2010) whereby the main erupted magma is essentially “by-passed”. As an example, Dietterich and de Silva (2010) deduced low pre-eruptive sulfur content of the melt from low S concentrations in apatites, which contrasts with the high stratospheric sulfur input of 16-55 Mt of S estimated from ice core data for the 1600 eruption of Huaynaputina, Peru. In this particular case, external sulfur derived from hydrothermally altered wall rocks is assumed to be the source for most of the released sulfur. Thus, apatite may faithfully record the equilibrium sulfur systematics in magmas, but is incapable to capture fast events, such as a high-sulfur flux during eruption that originated elsewhere (with one possible exception; see below). Post-eruptive dissolution also may explain cases where anhydrite should have crystallized but is not found. We contend that apatite may be a useful monitor for magmatic sulfur, yet there are still several outstanding issues to be resolved before apatite can yield its full potential (e.g., issue of partition coefficient, particularly where S <100 ppm in melt; issue of high-S apatite in silicic melt). Specific studies on magmatic systems with sulfur-rich apatites (i.e., apatites with higher S concentrations than reported from crystallization experiments (Streck et al. 2007; Broderick 2008; van Hoose et al. 2010) may be helpful to check if apatite can be used as a tracer for the streaming of S-rich fluids and/or as an indicator for a former existence of anhydrite crystals in magmas (magmas initially anhydrite-saturated but with subsequent anhydrite dissolution).
iN siTu SulFur ISotoPe In S-beArIng MInerAlS In situ sulfur isotope studies of igneous S-bearing minerals are limited but shed light on magmatic processes involving S-rich fluid (degassing) and place constraints on the source of magmatic volatiles (e.g., dissolution of anhydrite crystals, degassing of underplating mafic magma, whole-rock assimilation). Few sulfide minerals (mostly pyrrhotite) have been analyzed from igneous settings that are not associated with hydrothermal activity leading to mineralization. Most sulfur isotopic data for igneous systems are whole-rock and matrix glass data (see Marini and Moretti 2011, this volume, for discussion on sulfur isotopes in melts and magmas). In a recent work, Mandeville et al. (2009) analyzed the sulfur isotope ratios (d34S) in magmatic pyrrhotite and considered the lowest d34S in pyrrhotite (~0 ‰ d34S) as the initial d34S of the rhyodacite magma from Mount Mazama, in order to model the evolution of d34S in fluids and co-existing phases during degassing. Because pyrrhotite isotopic re-equilibration is not fast enough to record the syneruptive changes in d34S in the melt (Mandeville et al. 2009), the large ranges of d34S isotopes in pyrrhotite (~0 to +6 ‰ d34S) are considered as evidence of crystallization under vapor-saturated condition and are interpreted to record former degassing events of a relatively oxidizing magma (SO42− is the dominant species in the melt). d34S data for coexisting sulfide and sulfate-bearing minerals may be representative of d SSO4 and d34SH2S values of magmatic fluid phases (i.e., d34S in sulfate (SO4) and sulfide (H2S) species, respectively) and allow estimation of the temperature of crystallization, of the bulk sulfur isotopic compositions, and of the evolution of the SO42−/H2S of the parent fluids if the following conditions are met: (1) equilibrium was obtained between sulfate and sulfide aqueous species; (2) the mineral data approximate the d34SSO4 and d34SH2S of fluids; (3) post-depositional retrograde exchange in minerals did not occur (Rye 2005). 34
Sulfur-bearing Magmatic Accessory Minerals
309
In their ion-probe study of El Chichón trachyandesite, Luhr and Logan (2002) reported homogeneous d34S for individual anhydrite crystals but large crystal-to-crystal variations (+2.5 to +10.9 ‰ d34S, mean +6.4 ‰ d34S). They concluded that the observed sulfur isotopic variations in anhydrite and pyrrhotite (pyrrhotite: -0.1 to +2.7 ‰ d34S) can result from various proportions of (1) magmatic anhydrite and hydrothermal anhydrite, and (2) anhydrite and coexisting sulfide crystals which precipitated in different domains of a common magma reservoir. These domains may have been affected by different degrees of degassing and/or different degrees of crustal sulfur contamination. Similar results were obtained by McKibben et al. (1996) in their study of pumice from the 1991 eruption of Mount Pinatubo. At Pinatubo, there is also mineralogical evidence in form pyramidal surface growth features for hydrothermal anhydrite deposition (Jakubowski et al. 2002). Micrometer-scale sulfur isotopic variations recorded sulfur–isotope exchanges between sulfate and sulfide species in a degassing crystal–melt–fluid magmatic system as H2S exsolved preferentially and sulfate precipitated from a viscous degassing melt (Rye 2005).
concludIng reMArkS This chapter shows that the presence of high temperature S-bearing minerals may be extremely useful to provide information on magmatic conditions, especially to quantify fS2 and fO2 as well as the sulfur concentration in silicate melts. However, a considerable amount of work needs to be done to achieve this goal. At reducing conditions, sulfides are the main phases of interest and pioneering studies (Audétat et al. 2011) have demonstrated that fS2 and fO2 prevailing in magmas can be determined under certain circumstances. Nevertheless, the composition of sulfides coexisting with silicate melts is complex and the compositions and stability field of sulfides (solid phases, melts) are not well defined, especially in high temperature mafic systems. The composition of S-bearing phases other than sulfides or sulfates can also be used to trace magmatic processes (e.g., replenishment, mixing, degassing, kinetic processes). Some work has been done to model the partitioning of S between apatite and melts but there is little information for other S-bearing minerals that would be extremely useful to trace degassing processes in some magma chambers (e.g., for interpretation of the variation of S and Cl concentrations of compositionally zoned sodalites). The improvement of the analytical techniques should also open new ways to decipher magmatic and degassing processes based on sulfur isotopic fractionation between solid phases, melts and fluids.
AcknoWledgMentS This paper has greatly benefited from reviews by A. Audétat, M. Carroll, and H. Behrens and proof-reading by G. Morgan. We thank P. Boivin for the photograph of scapolite. The research conducted by F. Parat and F. Holtz described here has been supported by the German Science Foundation and the European Community. The research conducted by M. Streck has been supported by NSF-EAR grants 0337798 and 0838611.
reFerenceS Andújar J, Costa F, Martí J, Wolff JA, Carroll MR (2008) Experimental constraints on pre-eruptive conditions of phonolitic magma from the caldera-forming El Abrigo eruption, Tenerife (Canary Islands). Chem Geol 257:173-191 Arculus RJ, Johnson RW, Chappell BW, McKee CO, Sakai H (1983) Ophiolite-contaminated andesites, trachybasalts, and cognate inclusions of Mount Lamington, Papua New Guinea; anhydrite–amphibolebearing lavas and the 1951 cumulodome. J Volcanol Geotherm Res 18:215-247
310
Parat, Holtz, Streck
Audétat A, Pettke T, Dolejs D (2004) Magmatic anhydrite and calcite in the ore-forming quartz-monzodiorite magma at Santa Rita, New Mexico (USA): Genetic constraints on porphyry-Cu mineralization. Lithos 72:147-161 Audétat A, Pettke T (2006) Evolution of a porphyry-Cu mineralized magma system at Santa Rita, New Mexico (USA). J Petrol 47(10):2021-2046 Audétat A, Dolejs D, Lowenstern JB (2011) Molybdenite saturation in silicic magmas: occurrence and petrological implications. J Petrol 52(5):891-904, doi: 10.1093/petrology/egr1008 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Baker L, Rutherford MJ (1996) Crystallization of anhydrite-bearing magmas. In: Origin of Granites and Related Rocks. Brown M, Candela PA, Peck DL, Stephens WE, Walker RJ, Zen E-a (eds) Geol Soc Am Special Paper 315:243-250 Barth AP, Dorais MJ (2000) Magmatic anhydrite in granitic rocks: First occurrence and potential petrologic consequences. Am Mineral 85:430-435 Bea F, Fershtater G, Corretgé LG (1992) The geochemistry of phosphorus in granite rocks and the effect of aluminium. Lithos 29(1-2):43-56 Beermann O (2010) The solubility of sulfur and chlorine in H2O-bearing dacites of Krakatau and basalts of Mt. Etna. PhD thesis, Leibniz University of Hanover Berndt J, Holtz F, Koepke J (2001) Experimental constraints on storage conditions in the chemically zoned phonolitic magma chamber of the Laacher See volcano. Contrib Mineral Petrol 140(4):469-486 Bindi L, Cellai D, Melluso L, Conticelli S, Morra V, Menchetti S (1999) Crystal chemistry of clinopyroxene from alkaline undersaturated rocks of the Monte Vulture Volcano, Italy. Lithos 46:259-274 Boivin P, Camus G (1981) Igneous scapolite-bearing associations in the Chaîne-des-Puys, Massif Central (France) and Atakor (Hoggar-Algeria). Contrib Mineral Petrol 77:365-375 Botcharnikov RE, Behrens H, Holtz F, Koepke J, Sato H (2004) Sulfur and chlorine solubility in Mt. Unzen rhyodacitic melt at 850 °C and 200 MPa. Chem Geol 213(1-3):207-225 Botcharnikov RE, Linnen RL, Wilke M, Holtz F, Jugo P, Berndt J (2011) High gold concentrations in sulphidebearing magma under oxidizing conditions. Nature Geosci 4:112-115 Brauns R (1914) Skapolith führende Auswürflinge aus dem Laacher Seegebiet. Neues Jahrn Mineral Geol Palaeontol 39:79-125 Brenan JM (2003) Effects of fO2, fS2, temperature and melt composition on fe-ni exchange between olivine and sulfide liquid: implications for natural olivine-sulfide assemblages. Geochim Cosmochim Acta 67:26632681 Broderick CA (2008) Origin of S-rich apatite in calc-alkaline silicic rocks. Unpub MS thesis Portland State University, p 183 Bryan SE (2006) Petrology and geochemistry of the Quaternary caldera-forming, phonolitic Granadilla eruption, Tenerife (Canary Islands). J Petrol 47(8):1557-1589 Cabri LJ (1973) New data on phase relations in the Cu-Fe-S system. Econ Geol 68:443-454 Carroll MR, Rutherford MJ (1987) The stability of igneous anhydrite: experimental results and implications for sulfur behavior in the 1982 El Chichón trachyandesite and other evolved magmas. J Petrol 28:781-801 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation states: results from measured wavelength shifts of sulfur X-rays. Am Mineral 73:845-849 Cavarretta G, Lombardi G (1990) Origin of sulphur in the Quaternary perpotassic melts of Italy: Evidence from haüyne sulphur isotope data. Chem Geol 82:15-20 Chambefort I, Dilles JH, Kent AJR (2008) Anhydrite-bearing andesite and dacite as a source for sulfur in magmatic-hydrothermal mineral deposits. Geology 36(9):719-722 Chappell BW, White JR (1968) The X-ray spectrographic determination of sulfur coordination in scapolite. Am Mineral 53:1735-1738 Chesner CA (1998) Petrogenesis of the Toba Tuffs, Sumatra, Indonesia. J Petrol 39(3):397-438 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulphur in hydrous rhyolitic melts. J Petrol 45:2171-2196 Costa F, Scaillet B, Pichavant M (2004) Petrological and experimental constraints on the pre-eruption conditions of holocene dacite from the Volcan San Pedro (36°S, Chilean Andes) and the importance of sulfur in silicic subduction-related magmas. J Petrol 45(4):855-881 Craig JR, Kullerud G (1969) Phase relations in the Cu-Fe-Ni-S system and their application to magmatic ore deposits. Econ Geol Monograph 4:344-358 Crisp JA, Spera FJ (1987) Pyroclastic flows and lavas of the Mogan and Fataga formations, Tejeda Volcano, Gran Canaria, Canary Islands: mineral chemistry, intensive parameters, and magma chamber evolution. Contrib Mineral Petrol 96:503-518 Czamanske GK, Moore JG (1977) Composition and phase chemistry of sulfide globules in basalt from the midAtlantic Ridge rift valley near 37°N lat. Geol Soc Am Bull 88:587-599
Sulfur-bearing Magmatic Accessory Minerals
311
Dawson JB, Smith JV, Steele IM (1994) Trace-element distribution between coexisting perovskite, apatite and titanite from Oldoinyo Lengai, Tanzania. Chem Geol 117:285-290 De Fino M, La Volpe L, Peccerillo A, Piccarreta G, Poli G (1986) Petrogenesis of Monte Vulture volcano (Italy): inferences from mineral chemistry, major and trace element data. Contrib Mineral Petrol 92:135-145 de Hoog JCM, Hattori KH, Hoblitt RP (2004) Oxidized sulfur-rich mafic magma at Mount Pinatubo, Philippines. Contrib Mineral Petrol 146:750-761 Deen JA, Rye RO, Munoz JL, Drexler JW (1994) The magmatic hydrothermal system at Julcani, Peru: Evidence from fluid inclusions and hydrogen and oxygen isotopes. Econ Geol 89:1924-1938 Deer WA, Howie RA, Wise WS, Zussman J (2004) Rock-forming minerals: Framework Silicates: Silicate Minerals, Feldspathoids and the Zeolites. Geological Society, London Di Muro A, Bonaccorsi E, Principe C (2004) Complex colour and chemical zoning of sodalite-group phases in a hauynophyre lava from Mt. Vulture, Italy. Mineral Mag 68(4):591-614 Di Muro A, Pallister JS, Villemant B, Newhall CG, Semet M, Martinez M, Mariet C (2008) Pre-1991 sulfur transfer between mafic injections and dacite magma in the Mt. Pinatubo reservoir. J Volcanol Geotherm Res:517-540 Dietterich H, de Silva S (2010) Sulfur yield of the 1600 eruption of Huaynaputina, Peru: Contributions from magmatic, fluid-phase, and hydrothermal sulfur. J Volcanol Geotherm Res doi: 10.1016/j. jvolgeores.2010.1001.1003 Dirksen O, Humphreys MCS, Pletchov P, Melnik O, Demyanchuk Y, Spartks RSJ, Mahony S (2006) The 20012004 dome-forming eruption of Shiveluch volcano, Kamchatka: observation, petrological investigation and numerical modelling. J Volcanol Geotherm Res 155:201-226 Downes H (1987) Tertiary and Quaternary volcanism in the Massif Central, France. In: Fitton JG, Upton BGJ (eds) Alkaline Igneous Rocks. Geol Soc Spec Pub 30:517-530 Drexler JW, Munoz JL (1985) Highly oxidized pyrrhotite-anhydrite-bearing silicic magmas from the Julcani Ag-Cu-Bi-Pb-Au-W District, Peru: Physicochemical conditions of a productive magma. Canadian Institute of Mining Conference on Granite-Related Mineral Deposits, Halifax, September 15-17, 1985, Extended Abstracts, :87-100 Ducea MN, McInnes BIA, Wyllie PJ (1994) Sulfur variations in glasses from volcanic rocks: Effect of melt composition on sulfur solubility. Int Geol Rev 36:703-714 Ducea MN, McInnes BIA, Wyllie PJ (1999) Experimental determination of composition dependence of hydrous silicate melts on sulfate solubility. Eur J Mineral 11(1):33-43 Fleet ME (2006) Phase equilibria at high temperature. Rev Mineral Geochem 61:365-419 Fournelle J (1990) Anhydrite in Nevado del Ruiz November 1985 pumice; relevance to the sulfur problem. J Volcanol Geotherm Res 42:189-201 Fournelle J, Carmoby R, Daag AS (1996) Anhydrite-bearing pumices from the June 15, 1991, eruption of Mount Pinatubo: geochemistry, mineralogy, and petrology. In: FIRE and MUD: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, University of Washington Press, p 845-863 Francis RD (1990) Sulfide globules in mid-ocean ridge basalts (MORB), and the effect of oxygen abundance in Fe-S-O liquids on the ability of those liquids to partition metals from MORB and komatiite magmas. Chem Geol 85:199-213 Goff F, Arney BH, Eddy AC (1982) Scapolite phenocrysts in a latite dome, northwest Arizona, USA. Earth Planet Sci Lett 60(86-92) Goldsmith JR, Newton RC (1977) Scapolite-plagioclase stability relations at high pressures and temperatures in the system NaAlSi3O8-CaAl2Si2O8-CaCO3-CaSO4. Am Mineral 62:1063-1977 Guo Z, Wilson M, Liu J, Mao Q (2006) Post-collisional, potassic and ultrapotassic magmatism of the Northern Tibetan Plateau: constraints on characteristics of the mantle source, geodynamic setting and uplift mechanisms. J Petrol 47(6):1177-1220 Hammouda T, Chantel J, Devidal J-L (2010) Apatite solubility in carbonatitic liquids and trace element partitioning between apatite and carbonatite at high pressure. Geochim Cosmochim Acta 74:7220-7235 Harrison RK, Snelling NJ, Merriman RJ, Morgan GE, Goode AJJ (1977) The Wolf Rock, Cornwall: new chemical, isotopic age and paleomagnetic data. Geol Mag 114(4):249-328 Hattori KH (1993) High-sulfur magma, a product of fluid discharge from underlying mafic magma: evidence from Mount Pinatubo, Philippines. Geology 21:1083-1086 Hattori KH (1996) Occurrence and origin of sulfide and sulfate in the 1991 Mount Pinatubo eruption products. In: FIRE and MUD: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, University of Washington Press, p 807-824 Heming RF, Carmichael ISE (1973) High-temperature pumice flows from the Rabaul Caldera Papua, New Guinea. Contrib Mineral Petrol 38:1-20 Holm PM (1982) Mineral chemistry of perpotassic lavas of the Vulsinian district, the Roman Province, Italy. Mineral Mag 46:379-386
312
Parat, Holtz, Streck
Holm PM, Wilson JR, Christiensen BP, Hansen L, Hansen SL, Hein KM, Mortensen AK, Pedersen R, Plesner S, Runge MK (2006) Sampling the Cape Verde mantle plume: evolution of melt compositions on Santo Antao, Cape Verde Islands. J Petrol 47(1):145-189 Humphreys MCS, Blundy JD, Sparks RSJ (2006) Magma evolution and open-system processes at Shiveluch Volcano: insights from phenocryst zoning. J Petrol 47(12):2303-2334 Jakubowski R, Fournelle J, Welch S, Swope RJ, Camus P (2002) Evidence for magmatic vapor deposition of anhydrite prior to the 1991 climactic eruption of Mount Pinatubo, Philippines. Am Mineral 87:1029-1045 Jugo P, Luth RW, Richards JP (2005) An experimental study of the sulfur content in basaltic melts saturated with immiscible sulfide or sulfate liquids at 1300 °C and 10 GPa. J Petrol 46:783-798 Larsen JG (1981) Medium pressure crystallization of a monchiquitic magma-evidence from megacrysts of Drever’s block, Ubekendt Ejland, West Greenland. Lithos 14:241 Lesher CM, Groves DI (1986) Controls on formation of komatiite-associated nickel-copper sulfide deposits. In: Geology and Metallogeny of Copper Deposits. Friedrich GH, Genkin AD, Naldrett AJ, Ridge JD, Sillitoe RH, Vokes FM (eds) Springer-Verlag, Heidelberg, p 43-62 Lessing P, Grout CH (1971) Haüynite from Edwards, New York. Am Mineral 56:1096-1100 Li C, Ripley EM, Naldrett AJ, Schmitt AK, Moore CH (2009) Magmatic anhydrite-sulfide assemblages in the plumbing system of the Siberian Traps. Geology 37(3):259-262 Lindgren W, Ransome FL (1906) Geology and gold deposits of the Cripple Creek district, Colorado. U. S. Geological Survey Professional Paper 54, p 560 Liu Y, Comodi P (1993) Some aspects of the crystal-chemistry of apatites. Mineral Mag 57:709-719 Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Lovering JF, White JR (1964) The significance of primary scapolite in granulitic inclusions from deep-seated pipes. J Petrol 5:195-218 Lowenstern JB (1993) Evidence for a copper-bearing fluid in magma erupted at the Valley of Ten Thousand Smokes, Alaska. Contrib Mineral Petrol 114:409-421 Lowenstern JB, Mahood GA, Hervig RL, Sparks J (1993) The occurrence and distribution of Mo and molybdenite in unaltered peralkaline rhyolites from Pantelleria, Italy. Contrib Mineral Petrol 114:119-129 Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichón volcano, Chipas, Mexico: Mineralogy and petrology of the anhydrite-bearing pumice. J Volcanol Geotherm Res 23:69-108 Luhr JF (1990) Experimentally phase relations of water- and sulfur-saturated arc magmas and the 1982 eruption of El Chichón volcano. J Petrol 31:1071-1114 Luhr JF, Logan MA (2002) Sulfur isotope systematics of the 1982 El Chichón trachyandesite: An ion microprobe study. Geochim Cosmochim Acta 66(18):3303-3316 Luhr JF (2008) Primary igneous anhydrite: progress since its recognition in the 1982 El Chichón trachyandesite. J Volcanol Geotherm Res 175:394-407 Mandeville CW, Webster JD, Tappen CM, Taylor BE, Timbal A, Sasaki A, Hauri E, Bacon CR (2009) Stable isotope and petrologic evidence for open-system degassing during the climatic and pre-climatic eruptions of the Mt Mazama, Crater Lake, Oregon. Geochim Cosmochim Acta 73:2978-3012 Marchev P (1991) Primary barite in high-K dacite from the eastern Rhodope, Bulgaria. Eur J Mineral 3:10051008 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Mathez EA (1976) Sulfur solubility and magmatic sulfides in submarine basalt glass. J Geophys Res 81(23):4269-4276 Matthews SJ, Sparks RSJ, Gardeweg MC (1999) The Piedras Grandes-Soncor eruptions, Lascar Volcano, Chile; Evolution of a zoned magma chamber in the Central Andean upper crust. J Petrol 40(12):1891-1919 McKibben MA, Eldridge CS, Reyes AG (1996) Sulfur isotopic systematics of the June 1991 Mount Pinatubo eruptions: A SHRIMP ion microprobe study. In: FIRE and MUD: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, University of Washington Press, p 825-843 Morel J-M, Gourgaud A (1992) Pétrogenèse des trachy-andésites à haüyne du massif de l’Aiguiller (Monts Dore; Masif Central Francais): hypothèse de mélange magmatique entre trachy-phonolite et hawaiite. C R Acad Sci Paris 314(II):791-797 Moune S, Holtz F, Botcharnikov R (2009) Sulphur solubility in andesitic to basaltic melts: implications for Hekla volcano. Contrib Mineral Petrol 157:691-707 Mungall JE (2007) Crystallization of magmatic sulfides: An empirical model and application to Sudbury ores. Geochim Cosmochim Acta 71:2809–2819 Nadeau O, Williams-Jones AE, Stix J (2010) Sulphide magma as a source of metals in arc-related magmatic hydrothermal ore fluids. Nature Geosci 3:501-505
Sulfur-bearing Magmatic Accessory Minerals
313
Naldrett AJ, Craig JR, Kullerud G (1967) The central portion of the Fe-Ni-S system and its bearing on pentlandite exsolution in iron-nickel sulfide ores. Econ Geol 62:826-847 Naldrett AJ (1981) Nickel sulfide deposits: Classification, composition, and genesis. Econ Geol 75th Anniversary Volume, p 628-685 Newton RC, Goldsmith JR (1976) Stability of the end-member scapolites: 3NaAlSi3O8·NaCl, 3CaAl2Si2O8. CaCO3, 3CaAl2Si2O8.CaSO4. Z Kristallogr 143:333-353 O’Neill HSC, Mavrogenes JA (2002) The sulfide capacity and the sulfur content at sulfide saturation of silicate melts at 1400 degrees C and 1 bar. J Petrol 43(6):1049-1087 Pallister JS, Hoblitt RP, Reyes AG (1992) A basalt trigger for the 1991 eruptions of Pinatubo volcano? Nature 356:426-428 Pallister JS, Meeker GP, Luhr JF (1995) Recognizing ancient sulfur-rich eruptions: lessons from Pinatubo, El Chichón, and Mount Saint Helens. IUGG XXI General Assembly, A279 Pallister JS, Hoblitt RP, Meeker GP, Knight RJ, Siems DF (1996) Magma mixing at Mount Pinatubo; petrographic and chemical evidence from the 1991 deposits. In: FIRE and MUD: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, University of Washington Press, p 687-731 Parat F, Dungan MA, Streck MJ (2002) Anhydrite, pyrrhotite and sulfur-rich apatite: tracing the sulfur evolution of an Oligocene andesite (Eagle Mountain, Colorado, U.S.A.). Lithos 64:63-75 Parat F, Holtz F (2004) Sulfur partitioning between apatite and melt and effect of sulfur on apatite solubility at oxidizing conditions. Contrib Mineral Petrol 147:201-212 Parat F, Dungan MA, Lipman PW (2005) Contemporaneous trachyandesitic and calc-alkaline volcanism of the Huerto Andesite, San Juan Volcanic Field, Colorado, U.S.A. J Petrol 46(5):859-891 Parat F, Holtz F (2005) Sulfur partition coefficient between apatite and rhyolite: the role of bulk S content. Contrib Mineral Petrol 150(6):643-651 Parat F, Holtz F, Feig S (2008) Pre-eruptive conditions of the Huerto Andesite (Fish Canyon system, San Juan volcanic field, Colorado): Influence of volatiles (C-O-H-S) on phase equilibria and mineral composition. J Petrol 49(5):911-935 Parat F, Holtz F, Klügel A (2011) S-rich apatite-hosted glass inclusions in xenoliths from La Palma: constraints on the volatile partitioning in evolved alkaline magmas. Contrib Mineral Petrol doi: 10.1007/s00410-0110606-7 Peccerillo A (2005) Plio-Quaternary volcanism in Italy. Springer, Berlin-Heidelberg, p 365 Peng G, Luhr JF, McGee JJ (1997) Factors controlling sulfur concentrations in volcanic apatite. Am Mineral 82:1210-1224 Perini G, Francalanci L, Davidson JP, Conticelli S (2004) Evolution and genesis of magmas from Vico volcano, Central Italy: multiple differentiation pathways and variable parental magmas. J Petrol 45(1):139-182 Piccoli PM, Candela PA (2002) Apatite in igneous systems. Rev Mineral Geochem 48:255-292 Pichavant M, Montel J-M, Richard LR (1992) Apatite solubility in peraluminous liquids; experimental data and an extension of the Harrison-Watson model. Geochim Cosmochim Acta 56(10):3588-3861 Raghavan V (2004) Cu-Fe-S (Copper-Iron-Sulfur). J Phase Equilib Diffus 25(5):450-454 Raghavan V (2006) Cu-Fe-Ni-S (Copper-Iron-Nickel-Sulfur). J Phase Equilib Diffus 25:458-461 Rajamani V, Naldrett AJ (1978) Partitioning of Fe, Co, Ni, and Cu between sulfide liquid and basaltic melts and composition of Ni-Cu sulfide deposits. Econ Geol Bull Soc Econ Geol 73(1):82-93 Renno AD, Franz L, Witzke T, Herzig PM (2004) The coexistence of melts of hydrous copper chloride, sulfide and silicate compositions in a magnesiohastingsite cumulate, Tubaf Seamount, Papua New Guinea, Can Mineral 42:1-16 Renzulli A, Upton BGJ, Boyce A, Ellam RM (1998) Petrology of quartz syenite and hauyne syenite clasts from the Pitigliano Formation, Latera caldera, Vulsini District, Central Italy. Eur J Mineral 10:333-354 Rodríguez-Losada JA, Martínez-Frías J (1998) Ancient oxide- and sulphide-mineralization in the islands of Tenerife and La Gomera (Canary Archipelago, Spain). Mineral Deposita 33:639-643 Rouse RC, Dunn PJ (1982) A contribution to the crystal chemistry of ellestadite and the silicate sulfate apatites. Am Mineral 67(1-2):90-96 Rutherford MJ, Devine JD (1996) Pre-eruption pressure-temperature conditions and volatiles in the 1991 dacitic magma of Mount Pinatubo. In: FIRE and MUD: Eruptions and Lahars of Mount Pinatubo, Philippines. Newhall CG, Punongbayan RS (eds) Philippine Institute of Volcanology and Seismology, University of Washington Press, p 751-766 Rye RO (2005) A review of the stable-isotope geochemistry of sulfate minerals in selected igneous environments and related hydrothermal systems. Chem Geol 215:5-36 Scaillet B, Macdonald R (2006) Experimental and thermodynamic constraints on the sulphur yield of peralkaline and metaluminous silicic flood eruptions. J Petrol 47:1413-1437 Schuessler JA, Schoenberg R, Behrens H, von Blanckenburg F (2007) The experimental calibration of the iron isotope fractionation factor between pyrrhotite and peralkaline rhyolitic melt. Geochim Cosmochim Acta 71:417-433
314
Parat, Holtz, Streck
Smith GC, Holness MB, Bunbury JM (2009) Interstitial magmatic scapolite in glass-bearing crystalline nodules from the Kula Volcanic Province, Western Turkey. Mineral Mag 72(6):1243-1259 Stern CR, Funk JA, Skewes MA (2007) Magmatic anhydrite in plutonic rocks at the El Teniente Cu-Mo deposit, Chile, and the role of sulfur- and copper-rich magmas in its formation. Econ Geol 102:1335-1344 Stone WE, Fleet ME, MacRae ND (1989) Two-phase nickeliferous monosulfide solid solution (mss) in megacrysts from Mount Shasta, California: A natural laboratory for nickel-copper sulfides. Am Mineral 74:981-993 Stone WE, Fleet ME (1991) Nickel-copper sulfides from the 1959 eruption of Kilauea Volcano, Hawaii: Contrasting compositions and phase relations in eruption pumice and Kilauea Iki lava lake. Am Mineral 76:1363-1372 Stoppa F, Liu Y (1995) Chemical composition and petrogenetic implications of apatites from some ultra-alkaline Italian rocks. Eur J Mineral 7:391-402 Stoppa F, Woolley AR, Cundari A (2002) Extension of the melilite-carbonatite province in the Apennines of Italy: the kamafugite of Grotta del Cervo, Abruzzo. Mineral Mag 66(4):555-574 Stormer JC, Carmichael ISE (1971) The free energy of sodalite and the behavior of chloride, fluoride and sulfate in silicate magmas. Am Mineral 56:292-306 Streck MJ, Dilles JH (1998) Sulfur evolution of oxidized arc magmas as recorded in apatite from a porphyry copper batholith. Geology 26(6):523-526 Streck MJ, Broderick CA, Halter WE (2007) Origin of sulfur rich apatite in silicic, calc-alkaline magmas. Geochim Cosmoschim Acta 71(15):A979 Sugaki A, Kitakaze A (1998) High form of pentlandite and its thermal stability. Am Mineral 83:133-140 Swanson SE, Kearney CS (2008) Anhydrite in the 1989-1990 lavas and xenoliths from the Redoubt Volcano, Alaska. J Volcanol Geotherm Res 175:509-516 Teertstra DK, Schindler M, Sherriff BL, Hawthorne FC (1999) Silvialite, a new sulfate-dominant member of the scapolite group with an Al-Si composition near the I4/m-P42/n phase transition. Mineral Mag 63(3):321329 Tepper JH, Kuehner SM (1999) Complex zoning in apatite from the Idaho Batholith; a record of magma mixing and intracrystalline trace element diffusion. Am Mineral 84(4):581-595 Tilley CE (1959) A note on the Nosean Phonolite of the Wolf Rock, Cornwall. Geol Mag 96:503-504 Toulmin P, Barton PB (1964) A thermodynamic study of pyrite and pyrrhotite. Geochimica et Cosmochimica Acta 28:641-671 Tracy RJ (2003) Chemistry and origin of zoned haüyne in Tahitian phonolite, with implications for magmatic fractionation. In: Melt Inclusions in Volcanic Systems - Methods, Applications and Problems. De Vivo B, Bodnar RJ (eds) Developments in Volcanology 5:163-184 Ueda A, Itaya T (1981) Microphenocrystic pyrrhotite from dacite rocks of Satsuma-Iwojima, Southwest Kyushu, Japan and the solubility of sulfur in dacite magma. Contrib Mineral Petrol 1981:21-26 van Hoose AE, Streck MJ, Pallister JS (2010) Apatite sulfur systematics and crystal population in the 1991 Pinatubo magmas. AGU, San Francisco, Abstract #V41B-2278 Van Peteghem JK, Burley BJ (1963) Studies on solid solution between sodalite, nosean and haüyne. The Can Mineral 7:808-813 Watson EB (1980) Apatite and phosphorus in mantle source regions; an experimental study of apatite/ melt equilibria at pressures to 25 kbar. Earth Planet Sci Lett 51(2):322-335 Webster JD, Sintini MF, De Vivo B (2009) The partitioning behavior of Cl, S, and H2O in aqueous vapor±saline-liquid saturated phonolitic and trachytic melts at 200 MPa. Chem Geol 263:19-36 Westrich HR, Gerlach TM (1992) Magmatic gas source for the stratospheric SO2 cloud from the June 15, 1991, eruption of Mount Pinatubo. Geology 20(10):867-870 Whitney JA, Stormer JC (1983) Igneous sulfides in the Fish Canyon Tuff and the role of sulfur in calc-alkaline magmas. Geology 11:99-102 Whitney JA (1984) Fugacities of sulfurous gases in pyrrhotite-bearing silicic magmas. Am Mineral 69(1-2):6878 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Wolff JA, Storey M (1983) The volatile component of some pumice-forming alkaline magmas from the Azores and Canary Islands. Contrib Mineral Petrol 82:66-74 Wörner G, Schmincke H-U (1984) Mineralogical and chemical zonation of the Laacher See tephra Sequence (East Eifel, W. Germany). J Petrol 25:805-835 Zaitsev AN, Zaiteseva OA, Buyko AK, Keller J, Klaudius J, Zolotarev AA (2009) Gem-quality yellow-green haüyne from Oldoinyo Lengai volcano, northern Tanzania. Gems Gemology Fall 2009:200-203
11
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 315-336, 2011 Copyright © Mineralogical Society of America
Sulfur in Extraterrestrial Bodies and the Deep Earth Denton S. Ebel Department of Earth and Planetary Sciences American Museum of Natural History Central Park West at 79th Street New York, New York 10024, U.S.A. [email protected]
This chapter reviews the origin and fate of sulfur (S) in silicate melts in the solar system, experiments bearing on the role of S in element partitioning among melts and solids in planets, and finally our current understanding of silicate melts and the role of sulfur in planetary evolution. Sulfur is an important component of undifferentiated meteorites that are precursors to planets. When planetary bodies differentiated into cores and mantles, metal and/or sulfides were removed from silicates. This process can be traced. Then, iron-sulfide cores differentiated into metal and metal-sulfide fractions, some of which are preserved as iron meteorites. The iron meteorites probably fractionated from silicate mantles at much lower pressures than the cores of Earth or Mars. Understanding the role of sulfur in silicate melts is critical to unraveling the history of Earth, the terrestrial planets, and the differentiated asteroids that were once parts of early planetesimals.
COSMOCHEMISTRY OF SULFUR Silicate melts and sulfur in primitive source materials Primitive extraterrestrial samples available for laboratory study include 1-20 mm cometary grains collected by the United States’ (NASA) Stardust mission, asteroidal material collected by the Japanese (JAXA) Hyabusa mission, interplanetary dust particles (IDPs) collected from the stratosphere by airplanes, micrometeorites from various collection sites, and meteorites that fall to Earth and are recovered. Sources of primitive meteorites are parent bodies, primarily asteroids, that did not differentiate into silicate mantles and metal-rich cores. The oldest dated solar system rocks are not bulk meteorites, but are the high-temperature, melted components of undifferentiated meteorites, which are a kind of cosmic sedimentary rock. These “chondritic” meteorites are slightly younger than the components that accreted to form them. They have atomic ratios of rock-forming elements (e.g., Fe/Si) that are very similar to those measured in the solar photosphere using spectroscopy. The “primitive” nature of meteorites is established by their radiometric ages, and their lack of aqueous and thermal metamorphism on their parent asteroidal bodies prior to delivery to Earth. The most volatile-rich chondrites contain a high proportion of fine-grained mineral “dust”, or matrix, between the higher-temperature clasts (Fig. 1). The clast population is dominated by Mg-, Si-rich chondrules up to 1 mm diameter, which originated as free-floating droplets of silicate melt in the early solar system. Refractory Ca-, Al-rich inclusions (CAIs), mostly never melted, comprise up to 4% of clasts in some chondrite classes (CV, CO), but are very rare in others. The chondrites are divided into groups based on textures, chemical characteristics, and oxygen isotopic compositions (Weisberg et al. 2006). The abundances of elements in the most 1529-6466/11/0073-0011$05.00
DOI: 10.2138/rmg.2011.73.11
316
Ebel
Figure 1. Silicate melts in the early solar system. A range of materials from small grains to aggregates of small grains, to sintered and melted aggregates (both CAIs and chondrules) combine in different proportions to make chondrite parent bodies that differentiate upon melting, forming metallic cores and silicate mantles.
volatile-rich carbonaceous chondrites (Ivuna type or “CI”), particularly the meteorite Orgueil (Gounelle et al. 2006), agree well with spectroscopic measurements of their abundances in the solar photosphere, particularly volatile elements such as Zn, Na, and S (Palme and Jones 2003). The reference bulk composition of the solar system (atom ratios) is the canonical “chondritic abundance.” It is derived from bulk chemical analyses of the CI carbonaceous chondrites (Anders and Grevesse 1989). Meteorite data are combined with solar photosphere measurements that provide abundances of C, N, O and other gases. The resulting “solar” (original) composition is the basis for understanding subsequent chemical differentiation of major and minor elements, one from another. Cosmochemists often normalize the abundances of the elements to Si = 106 atoms, yielding solar abundances of oxygen 1.413×106, sulfur 4.49×105, Fe 8.38×105, Ni 4.78×104, and Co 2.323×103 (Lodders 2003). When small bodies accreted to some threshold size (e.g., 10 km, Lee et al. 1976) in the early solar system, radiogenic and impact heating caused them to melt. The resulting chondritic melts contained immiscible silicate and metal-rich (Fe, Ni, Co-rich) liquids and solids. Even in a small body, gravity causes differentiation of silicates from sulfur-bearing metallic liquids, forming a core and mantle (Fig. 1). The elements that go with the metal (siderophile) and those that partition into silicates (lithophile) are thus separated. But only 13.7% of present day meteorite falls come from differentiated bodies. Meteorites with low Fe/Si include basalts from Mars, samples thought to be from the asteroid Vesta, samples from unknown bodies, and anorthosites and basalts from the moon. These are complemented by iron and stony-iron meteorites from the interiors of planetary bodies that were disrupted long ago by impacts (Grady and Wright 2006; Weisberg et al. 2006). Recent work has established that parent bodies of some iron meteorites were differentiated into core and mantle within 1 Ma of formation of the oldest melted clasts, the CAIs (Burkhardt et al. 2008), and that the differentiation of Earth, including solidification of a planet-wide silicate magma ocean into the early mantle, was substantially complete by less than 100 Ma later (Nicholls 2006; Bourdon et al. 2008). Solar system abundances of the elements can be compared to abundances measured in meteorites and abundances inferred for the primitive upper mantle (PUM) of the Earth (Fig. 2; McDonough and Sun 1995; Palme and O’Neill 2003). The volatile-rich CI chondrites have much higher S/(Fe+Co+Ni) ratios than other chondrites, including the highly reduced enstatite chondrites. The rocky planets (Mercury, Venus, Earth, Mars) and most meteorite parent bodies
Sulfur in Extraterrestrial Bodies & Deep Earth
317
Figure 2. Atomic ratios of canonical “solar” composition, meteorites, and the primitive upper mantle (PUM) of Earth. Undifferentiated compositions include carbonaceous (Orgueil, CI - CH), ordinary (H - LL), R chondrites, and enstatite chondrites (EH - EL). For EH chondrites, S/O = 0.103, for PUM, K/S = 0.787. Sources (superscripts): 1Anders and Grevesse (1989), 2Lodders (2003), 3Wasson and Kallemeyn (1988), 4Lodders and Fegley (1998), 5McDonough and Sun (1995).
(asteroids) have either lost, or never accreted, a significant sulfur component. Most, but not all, lunar rocks are strongly depleted in sulfur, relative to Earth’s mantle. Initial results from the Stardust mission indicate a lower limit (2s) elemental ratio of S/Fe > 0.31 in the rocky portion of Comet Wild 2, the only comet ever sampled directly (Westphal et al. 2009). The most metal-rich ordinary chondrites (H) and volatile-depleted carbonaceous chondrites (CV, CO, CK, CR), have similar, low atomic S/(Fe+Co+Ni) ratios (Fig. 2). Figure 1 schematically summarizes the hierarchical accretion of planetary bodies from mm-sized clasts and sub-mm sized matrix to progressively larger bodies. Silicate melts appear to have been formed in the earliest phases of solar system evolution, and chondrules show evidence of repeated melting events. While the heating mechanisms for CAIs and chondrules are not known, the heating of bodies >20 km in diameter is understood to result from the decay of radioactive elements that were more abundant in the early solar system than they are today. Fully melted CAIs, rare components in CV chondrites, yield the oldest U/Pb ages of any rocks formed in the solar system. Silicate melts and sulfur are involved in every phase of planetary evolution, from accretionary impacts through core/mantle differentiation to volcanism and subduction (Fig. 1). In cosmochemistry, “volatile” elements include sulfur and others that remain in the vapor phase when a vapor of solar composition cools to below ~477 °C at a total pressure (Ptotal) of 10−4 bar. The solar abundance of S is preserved in at least the CI chondrites, and perhaps also in comets, suggesting that these bodies sampled solids equilibrated with a vapor of solar composition down to at least 431 °C , the temperature at which FeS becomes stable against evaporation at Ptotal = 10−4 bar (Lodders 2003). Imagine that the midplane of the protoplanetary disk (the rotating disk of gas and mineral dust from which the planets formed) around the sun concentrated dust rich in high-temperature silicate condensates. At the enhanced dust/gas ratios expected in the midplane of the protoplanetary disk, solid or liquid sulfides would form at higher temperatures (Ebel and Grossman 2000; Ebel 2006). These are chemical systems where the vapor pressure, primarily due to H2, is constant, but solid dust particles are more abundant, increasing the overall abundance of elements that are concentrated in dust (Si, Mg, Al, Ca, Fe, O). In the most S-rich meteorites, however, S is in sulfates and sulfides formed at low temperatures during parent body aqueous and thermal alteration. Davis (2006) summarizes the debate over volatile depletion in meteorites (including S).
Ebel
318 Sulfur content of the terrestrial planets
During and after planetary accretion, silicate, sulfide, and metallic melts form discrete but dynamically and chemically interacting reservoirs. Chemical composition, particularly sulfur content, strongly controls the distribution of elements among these reservoirs. Planetary core formation involves gravitational removal of metal and/or sulfides from silicates at pressures from ~4 GPa (the Moon’s core/mantle boundary), to 140 GPa (Earth’s core/mantle boundary). Fractional crystallization of cores of small bodies (e.g., parent bodies of iron meteorites) involves metal-sulfide liquid equilibria at low pressures. Planetary evolution is therefore strongly influenced by sulfur content. Albaréde (2009) points to volatile depletion, particularly of S (Dreibus and Palme 1996; O’Neill and Palme 1998), in arguing for the late delivery of water, which is also volatile, to all the terrestrial planets. Volatility is conveniently indexed by the temperature at which 50% of an element has condensed to solids from a cooling vapor of solar bulk composition. While the K/U ratio is a frequent proxy for volatile/refractory mixing in discussions of the compositions of planets (Fig. 2), S has a 50% condensation temperature (T50%) of 391 °C, 342 °C below that of K (potassium) in a vapor of solar composition at Ptotal = 10−4 bar (Lodders 2003). Because of its high solar system abundance, S should therefore unambiguously record volatile depletion or partial condensation in planetary bodies. Other volatile, moderately siderophile and chalcophile elements (Ga, Cu, P, Sn) are also likely to record pre-accretionary partial condensation (Humayun and Cassen 2000; Righter and Drake 2000), and not later, post-accretionary processes. Earth’s mantle composition can be inferred from upper mantle rocks sampled as xenoliths in magmas from deep sources (Becker et al. 2006; Carlson and Boyet 2008). McDonough (2003; cf. Boehler 1996) reviews our understanding of the composition of the Earth’s core, which cannot be measured directly, although the size and density of Earth’s core are well known. Curves comparing element volatility to upper mantle abundances for lithophile elements can be interpolated to S, C, and P to infer bulk Earth compositions (Fig. 3). The upper mantle deficits in S, C, and P must then be compensated by an excess in the core, to bring P, S, and C abundances up to the volatility trend line. Thus the core is inferred to contain ~90% of the Earth’s C and P, and ~98% of its S (McDonough 2003). But this C-P-S combination (6080% being S) can account for only ~2.5 mass% of the core, requiring another light element, probably Si, maybe O (Hillgren et al. 2000; Wade and Wood 2005; Wood et al. 2008). Several mechanisms have been proposed for volatile loss (including S) from the terrestrial planets and asteroids, and from the parent bodies of the iron-rich meteorites. Either bodies did not accrete volatiles at all (partial condensation), or volatiles were lost through magmatic processes, or heating by impacts. Data for K isotopes (Humayun and Clayton 1995) indicate that if planetary scale evaporative loss did occur, for example at the top of a magma ocean, it did not fractionate light from heavy K isotopes. Nor are any significant enrichments in heavy S isotopes observed among components of chondritic meteorites (Alexander et al. 2000; Alexander and Grossman 2005; Marini et al. 2011, this volume). These results indicate that S was only partially condensed, in isotopic equilibrium with condensates, into terrestrial planetary bodies, including most asteroids.
EXPERIMENTAL CONSTRAINTS Element partitioning Silicate melts are central to discussion of core-mantle differentiation and the histories of meteorite parent bodies. Once a molten core forms inside a planet, core solidification involves differentiation of solid crystalline metal from liquid metal-sulfide (Ohtani et al. 1997). Constraints on parent body processes can be provided by the equilibrium partitioning of highly siderophile elements (HSE): Ru, Rh, Pd, Re, Os, Ir, Pt, Au; moderately siderophile elements
Sulfur in Extraterrestrial Bodies & Deep Earth
319
Figure 3. Volatile-depletion trend and expectation for S and P content of bulk Earth. The abundance in the silicate portion of the Earth (McDonough 2003) relative to CI chondrites (Lodders and Fegley 1998), is plotted against the temperature (T50%, °C) at which 50% of the solar complement of each element would condense from a vapor of bulk solar composition at Ptotal = 10−4 bar (Lodders 2003). For carbon, T50% = −223°C. The dashed line approximates the expected bulk Earth volatility trend, and dashed vertical lines between that trend and measured mantle compositions (stars) illustrate how core P and S compositions may be inferred. See also O’Neill and Palme (2008, their Fig. 3; many other versions of this figure exist in the literature).
(MSE): Fe, Co, Ni, Cu, Mo, Ag, W; and chalcophile elements (Cu, Zn, Ga, Ge, As; Ag, Cd, In, Sn, Sb) between coexisting phases (e.g., silicate melt and metal-sulfide liquid) at high temperatures. The HSE strongly partition into planetary cores (Becker et al. 2006). HSE isotopes (e.g., of Os) may be tracers of mantle plume sources on Earth (Brandon and Walker 2005; Van Orman et al. 2008). MSE and chalcophile elements in planetary mantles provide other useful tracers of core formation, because these elements partition less completely into cores (Mann et al. 2009). In 1977, Kelly and Larimer lamented the omission of many siderophile elements from cosmochemical research, “owing to the almost complete lack of data on distribution coefficients in metal-sulfide-silicate systems, a void that hopefully can soon be filled.” Today, much of that void has been filled (Righter 2005). In part this is due to improved methods for traceelement analysis such as laser ablation inductively coupled plasma mass spectrometry (LAICPMS, e.g., Campbell and Humayun 1999), and nanoscale secondary ion mass spectrometry (SIMS, e.g., Stadermann et al. 1999). Improved techniques allow experiments that reach high temperature and pressure conditions, using multianvil apparatus (reaching the base of Earth’s upper mantle, ~25 GPa, ~2000 °C) and the diamond anvil cell (reaching Earth’s core/mantle boundary, ~135 GPa, >3200 °C; e.g., Campbell et al. 2007). Questions and challenges do remain, as reviewed by Righter and Drake (2003). These include limited partitioning data for moderately siderophile elements (e.g., Sb, As, Ag, Ge); gaps in the experimental database for metal-silicate melt partitioning; limitation of most experiments to basaltic melts, rather than the peridotitic melts of the early mantle; and insufficient high-temperature, high-pressure mineral/melt partitioning data to allow broad application of crystal chemical concepts such as elastic strain theory (Stewart et al. 2009). Distribution coefficients between solid metal and liquid metal-sulfide (DM/su‑m), sulfurfree metallic liquid and silicate melt (Dm‑m/si‑m), metal-sulfide melt and silicate melt (Dsu‑m/si‑m),
320
Ebel
or metal-sulfide melt and crystalline silicate (Dsu‑m/Sx) depend upon oxygen fugacity (fO2), temperature (T), pressure (P), and the compositions of coexisting phases (Baker and Moretti 2011, this volume; Parat et al. 2011, this volume). Distribution coefficients are a convenient way to parameterize equilibria that are more correctly described by chemical thermodynamics and the free energies of exchange reactions as functions of the activities of all participants in those reactions. Thus exchange reactions with strong volume changes will be more pressure dependent, and those with large entropy changes more temperature dependent. Righter and Drake (2003) present a thermodynamic approach for understanding the subtle interplay of fO2, T, P, and composition, particularly sulfur. Walker and Li (2008) demonstrate the superiority of a thermodynamic understanding in disentangling the effects of pressure and composition on the marginally siderophile element Mo, which becomes chalcophile (DM/su‑m(Mo) decreases) at 1100 °C from 0 to 6 GPa in the Fe-S system. The reader is referred to the contribution by Baker and Moretti (2011, this volume) for the thermodynamics of S partitioning between silicate melts and other phases. For most purposes, D values (ratios of concentrations in wt%) based on specific experiments or parameterizations (e.g., using the ratio of non-bridging oxygens to tetrahedral cations, NBO/T, to describe melt polymerization, rather than a full equation of state for the silicate melt) are used in the literature. The convention is, for example Dsu‑m/si‑m(X) = (wt% X in metal/sulfide melt) / (wt% X in silicate melt).
Liquid silicate - liquid metal-sulfide In planetary interiors undergoing differentiation, silicate liquids float and the denser immiscible metal-sulfide liquids sink. Experimental constraints allow modeling of this process based on observed abundances of elements in rocks from differentiated bodies, including the Earth. Pressure is very important in large bodies (e.g., Earth), and much less so in small ones (e.g., Mercury, asteroids). Sulfur strongly partitions into liquid metal-sulfide, with Dsu‑m/si‑m (S) increasing with increasing FeO activity in silicate melt, increasing pressure, and decreasing temperature (Rose-Weston et al. 2009, their Eqn. 12). Elements such as Ni, W, Os, and Re that partition strongly into solid metal rather than metal-sulfide liquid (high DM/su‑m) will also have Dsu‑m/si‑m values that decrease with increasing sulfur abundance in metal-sulfide liquid (Righter 2005). Thus sulfur abundance (and also C content) strongly influences element partitioning during differentiation (Li and Agee 2001; Li and Fei 2003). A vast literature exists on the partitioning of trace elements between metal-sulfide melt and silicate melt at various pressures, temperatures, and imposed fO2. Mann et al. (2009) recently investigated lithophile and weakly-siderophile element partitioning in C-, and S-bearing systems from 2 to 24 GPa, 1750-2600 °C, and log fO2 = IW-1.3 to IW-4.2, where IW refers to the iron-wüstite oxygen fugacity buffer curve. Applied to the Earth’s mantle abundances of those elements, their results indicate that the Earth’s core differentiated from the mantle at high pressures. Corgne et al. (2007) recently investigated K partitioning between peridotitic melt and Fe-Ni-S-C-O liquids at 1650-2200 °C and 1.0-7.7 GPa, showing that Dsu‑m/si‑m (K) remains very low even at high pressure and S content. Recent work has extended the experimental database to more volatile chalcophile elements, such as Sb (Corrigan et al. 2009; Righter et al. 2009b), and Te and Se (Rose-Weston et al. 2009). The laser-heated diamond anvil cell (DAC) allows experiments at deep-mantle pressures, albeit with more difficulty in the control of temperature gradients. Jephcoat et al. (2008) review results for Ni, Co, I, and noble gases in S-free systems, where Dm‑m/si‑m for Ni and Co converge to similar values of 20 < Dm‑m/si‑m(Ni, Co) < 30 above 30 GPa (to 52 GPa) at 2227-3263°C and log fO2 = IW-2.0. Increasing numbers of S-bearing experiments are being done that are relevant to S-mediated partitioning at very high pressures. Bass and Parise (2008) review new techniques being applied to understanding mineral-melt interactions at ultra-high pressures.
Sulfur in Extraterrestrial Bodies & Deep Earth
321
Many difficulties occur in partitioning studies of platinum group elements (PGEs) in silicate melts at high P and T. These include the low solubilities of PGEs in silicate liquids, and control of fO2 (Righter and Drake 2003). An additional issue is the interpretation of nanonuggets in run products as either quench products of initial heating to high temperature (Cottrell and Walker 2006), or part of the equilibrium metal assemblage at run conditions (Ertel et al. 2006, 2008). Cottrell and Walker (2006) provide strong evidence for the former in the case of Dsu‑m/si‑m(Pt). Recently, Righter et al. (2008) have overcome many of these problems in studying Dsu‑m/si‑m(Pd) in both graphite and MgO capsules followed by LA-ICPMS analysis. They found large decreases in Dsu‑m/si‑m(Pd) approaching 12 GPa and 1900 °C. They investigated both Fe-S and Fe-S-C liquids, with silicate liquid compositions closer to expected peridotitic Earth mantle compositions, in contrast to earlier work (Holzheid et al. 2000). The role of S in these experiments was critical, as S and C in the metallic liquid are known to decrease Dsu‑m/si‑m(Pd) in experimental systems (Chabot et al. 2006). Generally, Dsu‑m/si‑m of high-valence elements (W, Ga, P, Mo) increase with silicate melt polymerization, while low-valence elements (Co, Ni) are negligibly affected. Figure 4 illustrates the large effect on Dm‑m/si‑m for Ga, W and P in S-free experiments from polymerized rhyolitic melts to depolymerized peridotitic melts, in terms of the ratio of non-bridging oxygen atoms to tetrahedrally coordinated cations, NBO/T (Mysen 1991). Both C and S in metallic liquid can have large effects on Dsu‑m/si‑m (Jana and Walker 1997), as can the major element composition Ni/(Fe+Ni) in the metallic liquid.
Solid metal - liquid metal-sulfide Partitioning of metal-sulfide melts from silicate melts is followed by separation of solid metal from liquid metal-sulfide in planetary differentiation. Products of these metal-sulfide systems are represented by iron meteorites. Models for the fractional crystallization of iron meteorites, and inferences about their parent differentiated planetary bodies, depend on experimental data. In partially molten Fe-S-metal systems, S is strongly concentrated in the
Figure 4. Effect of silicate melt composition on S-free metal-liquid silicate-melt partition coefficients (Dm‑m/si‑m), after Righter and Drake (2003, their Fig. 6). Experimental data for Co, W, Ga from Jaeger and Drake (2000), P from Pak and Fruehan (1986).
Ebel
322
liquid. As the concentration of S in the liquid increases at low pressure, most siderophile elements partition more strongly into the solid metal (Fig. 5). Scott (1972) estimated distribution coefficients from measurements of concentrations in coexisting sulfide nodules (that were once sulfide liquid) and surrounding metal in iron meteorites. Jones and Malvin (1990) compiled experimental data, and devised a parameterization of DM/su‑m for many trace elements at 1 bar, based on the availability of metal domains in Fe-Ni-S-P-C liquids. That is, siderophile element preferences for liquid depend on the abundance of atomic-scale locations that are free of non-metals. This powerful method, revised by Chabot and Jones (2003), allows prediction of partitioning behavior in a wide range of circumstances. Table 1 gives fit parameters Do and b for partition coefficients of elements i, DM/su‑m(i), expressed as ratios of wt%: D M / su -m =
Do (Fe domains)b
(1)
Figure 5. Parameterizations and data for solid metal-liquid metal partition coefficients (DM/su‑m) of Chabot and Jones (2003). Experimental data for Au, Ge and Ir are as compiled by Chabot and Jones (2003) and plotted in Chabot and Haack (2006, their Fig. 8), shown for comparison to fitted curves (Table 1) for those elements, and fits for other elements (Re, W, Co, Ni, P). Fit for Ga is very close to Ge, and Os is very close to Re. Of the elements shown, only Cu is chalcophile.
Sulfur in Extraterrestrial Bodies & Deep Earth
323
Table 1. Fit parameters for DM/su‑m (wt% in solid metal / wt% in liquid metal sulfide) for siderophile elements at 1 bar pressure (Chabot and Jones 2003). Element
Do
b
As Au Co Ga Ge Ir Ni Os P Pd Pt Re W
0.22 0.25 1 0.78 0.66 1.5 0.86 2 0.06 0.43 0.81 2 1.2
2.2 2 1.2 2.6 3 4.9 0.6 5.1 2.8 1.1 4.4 5 3.6
where the fraction of Fe domains in the Fe-Ni-S system that are non-metal free is calculated as: (Fe domains) =
(1 − 2XS ) (1 − XS )
(2)
where XS is the mole fraction of S in liquid. In the Fe-Ni-S-P-C system the fraction of Fe domains is given as: (Fe domains) =
(1 − 2XS − 3X P − 3X C ) (1 − XS − 2X P − 2X C )
(3)
where XP and XC are the mole fractions of P and C in liquid, respectively. This is a logical framework, grounded in the success of associated solution theory in describing the thermodynamic behavior of sulfide melts (Chang and Hsieh 1987; Kress 2000, 2003). Associated solution theory is a general scheme for treating liquids as aggregates of associated and free atoms (e.g., Fe + Fe-S + S) in developing models for liquid thermodynamic properties. Partitioning in metal − metal-sulfide systems appears to be weakly pressure-dependent to moderate pressures (to 8 GPa, Jones and Walker 1991). Recent high-pressure experiments include work by Hayashi et al. (2009), Lazar et al. (2004), Stewart et al. (2009), Van Orman et al. (2008), Walker (2000), and Walker and Li (2008).
Sulfide saturation and immiscibility This topic is discussed in detail in this volume by Baker and Moretti (2011). A great deal of experimental work exists to constrain the S content at sulfide saturation (SCSS) of silicate liquids at pressures below 9 GPa. Liu et al. (2007) compared their model for SCSS with those of Holzheid and Grove (2002), Li and Ripley (2005), and Scaillet and Pichavant (2005). They report that sulfur solubility in silicate melts (in equilibrium with sulfide) increases with temperature and fO2, but decreases with pressure and the degree of melt polymerization. Holzheid and Grove (2002) calculated that up to 12 wt% S is stable in a metallic liquid in equilibrium with a magma ocean containing 200 ppm S, similar to the Earth’s mantle. That value allows S to be the dominant light element in the core, but only if the Earth’s S content does not follow the volatility trend seen in lithophile elements (Fig. 3).
Ebel
324
Corgne et al. (2007) have found that immiscible S-rich and C-rich liquid alloys coexist up to ~5.5 GPa, in experiments on a “chondrite-like” composition at conditions up to 2200 °C and 7.7 GPa. These exciting results suggest that in small planetary bodies, particularly with enhanced C and S contents, significant stratification of a molten core would have been likely to occur.
Rheology: wetting and deformation The rheology of silicate partial melts and the wetting behavior of coexisting sulfide melts are increasingly recognized as important influences on core/mantle differentiation and the eventual distribution of trace elements in planetary bodies. Experimental studies of wetting behaviors of Fe-Ni-S-O liquids show that migration of such liquids through an olivine-rich matrix would be highly inefficient below 11 GPa in reduced systems (Minarik et al. 1996; cf. Mungall and Su 2005). Gaetani and Grove (1999) found that sulfide liquids wet olivine, forming interconnected networks at high fO2 (FMQ) in 1 bar, 1350 °C experiments. In the late stages of core formation, and in some planetary environments, wetting therefore facilitates the transport of siderophile elements. Sulfide liquids may, therefore, variably fractionate more chalcophile Os and Pb from lithophile Re, U, and Th, depending upon how efficiently they wet surfaces of silicates. Deformation may have important effects while a planetary body is hot and actively differentiating into a core and mantle. Kohlstedt and Holtzman (2009) recently reviewed experiments designed to improve our understanding of deformation in the process of melt extraction, including systems containing metal-sulfide liquid (Bruhn et al. 2000; Holtzman et al. 2003). High-temperature shear deformation experiments on planetary precursor materials provide constraints on how cores might form in planetesimals subject to impacts. Rushmer et al. (2005) contrasted buoyancy mechanisms for core segregation with shear-induced segregation in response to impacts. The efficiency and rate of transport of metal-sulfide liquid depend upon both the degree of melting and the shear stress in Rushmer et al.’s (2005) model. The apparent solubility of S in Fe-rich metallic liquids segregating to the core of a chondritic planet is sensitive to the rate of liquid transport. Thus shear stress on a planetesimal can affect the partitioning of S and highly siderophile elements into the core, and can also drive return flow of siderophile elements back into the mantle. Sulfur may play an important role in mass transfer between metallic planetary cores and partially molten silicate mantles. Rushmer et al. (2000) reviewed the physical processes of core formation. If sulfide “wets” silicate grains in a partially molten mixture, the sulfide can form a continuous path between reservoirs of metallic material. Wetting behavior depends primarily on surface tension of liquids, and small amounts of electrical current can affect the surface properties of metal-sulfide liquids. Recently Kavner and Walker (2006; cf. Kavner et al. 2007) explored the “electro-wetting” of silicates by sulfide liquid under an imposed potential of 1 volt, to simulate possible effects of current flow across Earth’s core/mantle boundary. Experiments showed dramatic electrochemical effects, including enhanced wetting by sulfide, indicating that small electrical currents can enhance the percolation of sulfide liquids into silicate partial melts (core or mantle).
PLANETARY INTERIORS Iron meteorites Upon melting of a bulk planetesimal of chondritic composition, dense, metal-sulfide melts (siderophile) differentiate from less dense silicate melts (lithophile) to form cores and mantles of planetary bodies. These cores further fractionate as solid metal crystallizes from metal-sulfide liquid, progressively enriching that liquid in less siderophile elements. Thus the compositions of iron meteorites record the histories of the cores of the earliest planetary
Sulfur in Extraterrestrial Bodies & Deep Earth
325
bodies in the solar system, prior to their disruption by collisions (Uhlig 1954; Allan and Jacobs 1956). Improvements in isotope and trace-element measurements have made possible increasingly accurate chronologies of planetary accretion and differentiation (reviewed by Halliday and Kleine 2006; cf. Burkhardt et al. 2008; Cook et al. 2004), and inferences about the compositions of the silicate mantles that are missing from the iron meteorites. Rocks potentially from differentiated parent bodies represent less than 13.7% of meteorite falls, and only 5.2% of all falls are iron or stony iron, based on statistics of U.S. Antarctic finds (Zolensky 1998). Those percentages include a small number of stony irons produced in extraterrestrial impacts, and iron meteorites that were not formed by magmatic processes. Most iron meteorites contain between 5 and 10 wt% Ni, 0.5-1.0 wt% Co, 0.2-0.6 wt% S, 0.10.3 wt% P and 0.01-0.2 wt% C (originally as Fe-carbides, Lipschutz and Anders 1961). Iron meteorites survive impact and weathering longer and in larger pieces than stony meteorites, so are over-represented in non-Antarctic meteorite collections. Buchwald (1975) described nearly all known iron meteorites in detail, and recent reviews of non-chondritic meteorites (achondrites) include Mittlefehldt et al. (1998), Haack and McCoy (2003), Chabot and Haack (2006), McCoy et al. (2006). The sub-solidus exsolution of Fe-rich and Fe-poor crystalline metal phases (Widmanstätten patterns) and concentration profiles of trace elements in those phases in iron meteorites allow inferences of the cooling rates and parent body sizes of those meteorites (Goldstein et al. 2009). Sulfide liquid nodules in iron meteorites separated from the surrounding iron-nickel alloy during cooling. Because Pb is chalcophile and U is highly lithophile, magmatic sulfide (troilite, FeS) inclusions in iron meteorites record the least radiogenic Pb of any known solar system materials. The elements in these inclusions fractionated into iron liquid from silicate magmas, and then fractionated from the crystallizing iron cores of their parent planetary bodies. The primordial Pb composition of the solar system has been referenced against troilite from the Canyon Diablo iron meteorites for nearly forty years (Canyon Diablo troilite, CDT, Tatsumoto et al. 1973). Recently Blichert-Toft et al. (2010) revised these values based on a study of sulfide inclusions in a range of iron meteorites, and the discovery of more primitive Pb isotopic signatures in the Nantan iron meteorite. From correlations between Ni concentrations and trace elements, it is inferred that most iron meteorite compositions result from secondary fractional crystallization (also known as differentiation, or segregation) of solid metal alloy from metallic liquid. Dendritic crystallization is one likely mechanism (Haack and Scott 1993). Different geochemical groups of iron meteorites are distinguished by different trends of fractional crystallization, implying the existence of different parent body cores with different initial compositions of siderophile elements. Variations in Ga and Ge were first used to distinguish four “Ga-Ge” groups (Lovering et al. 1957). This taxonomy has been developed by Wasson and coworkers (e.g., Wasson 1967, Wasson 1974, Scott and Wasson 1975, Scott 1977a, Wasson 1985; Wasson and Kallemeyn 2002), to 13 groups based on ~600 iron meteorites (Wasson et al. 1989).
Sulfur in core fractionation Sulfur is the most chemically important non-metal in planetary cores differentiated from silicate melts. Solid metal/liquid metal partition coefficients (DM/su‑m) deduced from trace-element variations within iron meteorite groups are consistent with DM/su‑m measured experimentally. Non-equilibrium solidification assuming local equilibrium at a solid-liquid front, no backward diffusion of the solute, and complete liquid mixing can be modeled using the Schiel-Gulliver equation (Gulliver 1913; Schiel 1942). So the composition of the solid (Csolid) at time t is given by solid liquid C= D M / su -m × Cinitial × (1 − fS )( D t
M / su -m
−1)
(4)
326
Ebel
where fS is the fraction of liquid solidified at time t, and Cinitialliquid is the initial liquid composition. Trace element partition coefficients (DM/su‑m) between solid and liquid metal depend on concentrations of S, P, C (non-metals) in the liquid (Fig. 5, Table 1; Narayan and Goldstein 1982; Willis and Goldstein 1982; Malvin et al. 1986; Jones and Malvin 1990). The solar S/P ratio is 53 (Lodders 2003), so the effect of S is expected to dominate fractional crystallization (Chabot et al. 2006). The effect of phosphorus on partitioning is less wellconstrained at high pressure for many trace elements, and has been shown to be negligible at 1 bar if S is present (Corrigan et al. 2009). From parameterizations of the experimental data, the initial S contents of the parent body cores of the magmatic iron groups have been estimated by modeling of crystallization trends considering multiple elements partitioning simultaneously. In this way, Jones and Drake (1983) calculated S/Ni mass ratios ranging from ~0 in the parent body cores of IVB iron meteorites, to 0.67 in cores of IIIAB irons, to 3.4 in the cores of IIAB iron meteorites. In undifferentiated meteorites, the same ratio ranges from 5.51 in CI to 1.64 in CV carbonaceous chondrites, to 1.25 in H ordinary chondrites (Wasson and Kallemeyn 1988). Chabot (2004) has used partitioning data to closely match trends for three iron meteorite groups, yielding 12.0 ± 1.5 wt% S in IIIAB, 17.0 ± 1.5 wt% S in IIAB, and 1 ± 1 wt% S in IVB iron meteorite parent body cores. She accounted for S exclusion from solids, and resulting nonconstant partition coefficients, in modeling core crystallization. These S contents are consistent with other studies of fractional crystallization of the IIIAB parent body core (Haack and Scott 1993; Ulff-Møller 1998; Chabot and Drake 1999), but higher than those based on measurements of troilite nodule abundance in meteorites representing different crystallization stages. In the studies based on nodule measurements (e.g., Wasson and Choi 2003), measurements of S in a series of meteorites representing different stages of crystallization are extrapolated back to the S content at zero crystallization using models to estimate the fraction of trapped melt and degree of crystallization. The ratio of sulfur to the bulk metal+sulfide content, S/(S+Fe+Co+Ni) where values are in wt%, ranges from 23.4 in CI to 8.1 in CV to 6.4 in H chondrites (Fig 2, above; Wasson and Kallemeyn 1988). Thus there is wide latitude for assumptions about the initial S depletions in the parent magmas of the iron meteorites, but the IIIAB and IIAB groups appear to have retained a significant proportion of the solar S component in their cores. Yet, S-rich iron meteorites are rare and pallasites are also deficient in FeS. Ulff-Møller et al. (1998b) concluded that a FeS liquid escaped or became concentrated in pallasite domains that are underrepresented in the meteorite record. So far, no meteorites anomalously enriched in S have been discovered, but not all extraterrestrial material is represented in the meteorite record. Several mechanisms have been proposed for S removal from differentiated asteroids early in solar system history, including outgassing of S during differentiation, removal as immiscible liquids, segregation of S into metastable liquid layers by partial melting (Kracher and Wasson 1982), and removal by explosive volcanism (Keil and Wilson 1993). The latter two ideas involve formation of FeS-rich partial melts at the Fe,Ni - FeS cotectic. Haack and Scott (1993) used detailed trace-element observations, guided by experiments, to infer the accumulation of S-rich liquid at the top of asteroid cores and in structural traps during dendritic crystallization of the IIIAB parent body core, represented by ~200 meteorites including Cape York. Segregated S-rich layers would survive disruption and the space environment for much shorter timescales than the mantles of the iron meteorite parent bodies, which are also largely missing from the meteorite record (Burbine et al. 1996; Chabot and Haack 2006).
Earth core formation Compositional constraints on the Earth’s core result from (1) analysis of upper mantle rocks, (2) experiments on partitioning between Fe-Ni-S-P-O liquids and solid or liquid silicates,
Sulfur in Extraterrestrial Bodies & Deep Earth
327
and (3) comparisons with the meteorite record. Particularly with regard to (2), much progress has been made recently, as reviewed briefly above and by Wood (2008). Different mechanisms for Earth’s accretion history and core formation cannot be ruled out by the evidence so far. Thus, controversy remains between single stage models of steady “homogeneous accretion” modified by pre-accretionary volatile depletion, with subsequent fractionation in a magma ocean (Wänke 1981; Wänke et al. 1984; Righter 2003), and multi-stage models of “heterogeneous accretion” involving singular or stochastic material addition, or upward recycling of the core (Kimura et al. 1974; Morgan et al. 1981; O’Neill 1991; O’Neill and Palme 1998; Wade and Wood 2005; Wood et al. 2009; Rose-Weston et al. 2009). Heterogeneous accretion includes the widely accepted addition of siderophile elements to the mantle by a “late veneer” (Chou 1978), suggested by the near-chondritic ratios of HSE in the mantle, and the overabundances of many HSE compared to their expected partitioning behavior. The hypothesis is that after Earth’s core had been substantially differentiated (i.e., “late”) from a reduced precursor, a “veneer” of oxidized chondritic material accreted and remained in the mantle (Drake and Righter 2002). Yet another possibility is that high-velocity impacts during the late stage of planetary accretion caused large fractions of material to be lost from the Earth and other terrestrial planets (Agnor and Asphaug 2004; O’Neill and Palme 2008). A central issue in understanding the formation of the Earth’s core is that the abundances of Ni, Co, W, and HSEs (Os, Re, Ir, Ru, Pt, Rh, Pd, Au) in the primitive upper mantle, while very low, are higher than expected for a magma ocean in chemical equilibrium with the core, from experimental partition coefficient data. This controversy has been reviewed recently (e.g., Lorand et al. 2008). It is clear that sulfur is a crucial actor, based on all the experimental results mentioned above (Righter et al. 2008). Mass transfer across the core-mantle boundary could partly explain the composition of the mantle. This includes the possible return of core material to the mantle either in plumes from the early core-mantle boundary (McDonough 2003), or in asteroid and comet impacts during the late heavy bombardment of the inner solar system (Rushmer et al. 2005). Late or prolonged extraction of Pb to the core in sulfide liquid may have disturbed the U-Pb systematics of the mantle, relative to the W-Hf system (Hart and Gaetani 2006). Core-mantle communication may even influence the evolution of the Earth’s interior to the present (Walker 2000; Kavner and Walker 2006). Righter and Drake (2003) build on the conclusion of Righter et al. (1997) that mantle Ni, Co, Mo, W and P abundances are consistent with equilibration of S-bearing liquid metal with hydrous liquid silicate near 27 GPa and 1925 °C and log fO2 = IW-0.15, near the present-day upper/lower mantle boundary. This is known as the “deep magma ocean” hypothesis (Rubie et al. 2003). Righter et al. (2008) and Cottrell and Walker (2006) have shown by experiment that suprachondritic Pd/Ir and Pt/Ir, respectively, of the upper mantle (Becker et al. 2006) can be explained by metal sulfide melt - silicate melt equilibrium at the high pressure and temperatures of an early magma ocean. However, mantle Pd/Ir is not as high as experiments would predict for the even higher pressure and temperature conditions of a magma ocean suggested by Chabot et al. (2005) who find that core segregation conditions of 30-60 GPa, >1825 °C, and log fO2 = IW-2.2 are consistent with Dsu‑m/si‑m(Ni) and Dsu‑m/si‑m(Co) parameterizations at <10 GPa, confirmed by experiments at higher pressures (Cottrell et al. 2008; Bouhifd and Jephcoat 2003). It is not yet clear how a single set of pressure-temperature conditions can explain the mantle abundances of all siderophile elements. Heterogeneous accretion models provide a wider range of scenarios for the evolution of pressure, temperature, and fO2 in a magma ocean. If the Earth began to melt when it was only a small fraction of its present size, much lower pressures would have obtained, but also low initial fO2. Wade and Wood (2005) suggest a scenario in which the formation of high-pressure silicates in the mantle caused mantle fO2 to increase during accretion. More recently, Wood et al. (2008) parameterized Dsu‑m/si‑m for V, Cr and Nb from 1.5-8 GPa and 1480-2000 °C for peridotite
328
Ebel
liquids, and used these elements simultaneously to estimate core formation conditions in a deepening and progressively (with time) more oxidized mantle. Better partitioning data will enable better multi-stage and hybrid models, constrained by the evolution of the distribution of multiple trace elements during core formation from a silicate magma ocean.
Sulfur and lithophile element partitioning More than half of the Earth’s internal heat is generated by the decay of radioactive elements, primarily K, U, and Th that partition strongly into silicates (i.e., are lithophile). Sulfur has been invoked in arguments for significant abundances of radioactive K and U in the core, because highly reduced enstatite chondrites have Earth-like O isotopic compositions, and contain trace K-, and Na-, Cr-sulfides (Herndon 1996). However, they also contain abundant Mg-, and Mnsulfides, and CaS (oldhamite), that is rich in rare earth elements (REE). Although peridotitic silicate liquids with moderate S (1.9 wt%) and K (4.7 wt%) contents will not sequester more than a few tens of ppm K into metal liquid (Corgne et al. 2007), Fe-Ni-S-O liquids with >10 wt% S and ~5% wt% O can sequester up to 250 ppm K (Gessmann and Wood 2002), but such liquids would also remove Ca and other lithophile elements from the mantle into the core. Similarly, sulfides have been invoked in a highly reduced Earth to extract U and other lithophiles to the core, however Wheeler et al. (2006) showed that Dsu‑m/si‑m(U) does not exceed 0.0001 under conditions appropriate to a planetary magma ocean. There is no chemical or isotopic evidence relevant to the bulk silicate Earth to support removal of lithophile U, K or Th to the core.
MAGMAS OF OTHER SOLAR SYSTEM BODIES Meteorites, planetary missions, and experimental and theoretical studies provide a limited knowledge base for understanding silicate melts and sulfur contents of extraterrestrial bodies. The comprehensive review by the Basaltic Volcanism Study Project (BVSP 1981) summarized all then-current knowledge of silicate melts on all the planets, but it preceded conclusive identification of Martian meteorites. More recent reviews include Shearer et al. (1998) and Bottke et al. (2002). Our understanding of Mercury will be refined by imminent observations by the MESSENGER spacecraft. Recent flyby results (Prockter et al. 2010) reveal extensive volcanic flooding by silicate magmas that post-dates basin formation, and may be as young as 1 Ga. Observations are consistent with episodic local volcanism and many thin flow layers. Detection and measurement of S at Mercury requires orbital operations, and exosphere observations are robust only for Na, Ca, and Mg until MESSENGER commences orbital measurements (Vervack et al. 2010). In Mercury, S may have played a role in partitioning of normally lithophile elements to the core. Mercury’s core is >60% of its mass, unlike Venus and Earth, which have cores ~32% of their total mass. If Mercury is highly reduced, then Ca-, or REE-bearing S-rich liquids may have been extracted to its core, and the signature of such a history would be measurable by remote sensing of Mercury’s surface from space. However, the volatile-depletion trend of the terrestrial planets suggests that S was very strongly depleted from its solar system abundance in the solids that combined to form Mercury (Fig. 3). Our most recent knowledge of Venus derives from the Magellan mission, as reviewed in Bougher et al. (1998). Pyrite is likely present on the 390-470 °C hot Venusian surface (Hong and Fegley 1998). Kargel et al. (1993) synthesized Venera and VEGA mission results, and recalculated major-element oxide compositions (from X-ray fluorescence data) and K, U, Th abundances (from gamma ray spectroscopy). Unfortunately, sulfur was not measured. All seven landers reported mafic volcanic compositions, with particularly K-rich signatures suggesting an anhydrous, carbonated mantle. Kargel et al. (1993) conclude that the mantles of Earth and Venus are geochemically very similar in other (non-volatile) respects.
Sulfur in Extraterrestrial Bodies & Deep Earth
329
The chemical evolution of the Moon is thought to have been dominated by melting of the entire silicate mantle, creating a global magma ocean (Wieczorek et al. 2006). The liquid core that formed by differentiation in this period acted as a convecting dynamo (Garrick-Bethell et al. 2000). Although lunar magnetic activity has ceased, recent findings show the lunar core remains partially molten (Williams et al. 2001, 2004; Weber et al. 2011), and there are suggestions that the lunar mantle may still be degassing (Crotts and Hummels 2009). Weitz et al. (1999) found 600 ppm S in a melt inclusion from the source magma of lunar fire fountains, collected during the Apollo 17 mission. Condensates in vugs and vesicles yield lower limits on the sulfur content of erupted lavas (Fegley 1991). Sulfide replacement of silicates has been inferred from analysis of lunar breccias (Lindstrom and Salpas 1983; Shearer et al. 2010). The ratio of S to other chalcophile elements is higher in lunar basalts than terrestrial basalts (Heiken et al. 1991, p. 422), indicating that less sulfide was removed from the lunar mantle prior to mare basalt formation. The Moon’s S content and its expression in surface rocks deserve more exploratory attention. Martian meteorites, the Mars Exploration Rover missions, and recent remote sensing studies have brought Mars into sharp focus. Results from Mars Global Surveyor allowed determination of a range of radii for the Martian core (1520-1840 km, mean Mars radius 3390 km), and also ruled out a completely solid core (Yoder et al. 2003). Stewart et al. (2007) conclude from a study of the Fe-Ni-S system at 23-40 GPa, and earlier work, that the Martian core is completely molten near the (Fe,Ni)-S eutectic, assuming a core composition range of 10.6-16.2 wt% S. This range is consistent with many previous studies (e.g., Wänke and Dreibus 1988), but not all (e.g., Gaetani and Grove 1997). By experimental work at Mars interior P‑T conditions, and by comparison with the Martian basaltic meteorite Shergotty (low Al2O3, high-FeO liquid), Righter et al. (2009a) have shown that Martian magmas can dissolve up to ~4000 ppm S at sulfide saturation, easily accounting for the sulfate-rich soils and layered deposits recently discovered by the Mars Exploration Rovers (McSween et al. 2004). Such surface rocks record oxidizing, acidic conditions involving extensive flux of water in the Martian past. Ghosal et al. (1998) used basaltic meteorites to infer a reduced, Fe-depleted Martian mantle. In the absence of subduction, the speciation of volcanic gases on Mars is critically dependent on the redox state of the Martian mantle (J. Jones, pers. comm.). Much is yet to be learned about the composition and evolution of Mars. The largest bodies in the asteroid belt are almost certainly differentiated (Greenwood et al. 2005). There is increasing evidence for thermal stratification of the main belt, such that S and other volatile elements are increasingly abundant outward (Ghosh et al. 2006). Asteroid 4 Vesta is known to be differentiated, and is the likely source of the howardite-eucrite-diogenite meteorite suite (Keil 2002), including products of serial silicate volcanism (Yamaguchi et al. 1997). From meteoritic and oxygen isotopic evidence, Boesenberg and Delaney (1997) and Lodders (2000, her Table 1) both estimated that Vesta contains ~6 mass% FeS in bulk, and an atomic S/Si ratio of 0.12 (see Fig. 2). The Dawn mission will return a wide range of data from orbit around both Vesta and Ceres (Russell et al. 2007). Volcanism on Io was reviewed by Geissler (2003). Galileo mission observations provided a revolutionary new understanding of Jupiter’s moon (cf. Kargel et al. 1999). Io is the most volcanically active body in the solar system, with at least 152 active volcanic centers (Lopes et al. 2004). This activity is due to tidal forcing by Jupiter, which “kneads” its moon causing internal strain and heating. Volcanic exhalations on Io are very rich in sulfur, although their compositions are difficult to measure. Recent spectroscopic observations by the passing New Horizons mission revealed lower limits of magma temperatures of 880-1060 °C, consistent with basaltic lava enriched in volatile elements (e.g., F) or water (Spencer et al. 2007). Resurfacing of Io seems primarily to occur by plume eruptions, rather than large lava flows (Geissler et al. 2004). However, Radebaugh et al. (2004) concluded that the long-lived, variable over
330
Ebel
minutes, 1225-1325 °C, central region of Pele Patera is an overturning lava lake with a large, stable magma source. Pele’s is one of several very S2-rich plumes observed with the Hubble telescope, each of which erupts huge quantities of S onto the surface (Jessup et al. 2007).
CONCLUSIONS To a first order, O, Si, Mg and Fe dominate planetary chemistry. But S is ~0.5× the Si abundance (~0.15× Si+Mg+Fe)). Thus S is a primary actor in both condensation chemistry (in the vapor phase, SiO(g) and SiS(g) compete for Si), and in the partitioning behavior of many elements that enter the cores of planetary bodies at low pressures (sampled in asteroids and iron meteorites) and high pressures (the inaccessible cores of the moon and planets). Future advances in understanding the role of S in extraterrestrial and deep planetary silicate melts will depend upon better experimental data, as well as improved abilities to model many trace elements simultaneously in investigating the interactions of silicate melts, metal-sulfide melts, and minerals. Both orbiter and lander missions to other planets will gradually increase the pool of measurements necessary for a better understanding of the roles of silicate melts and sulfur in the solar system. The processes reviewed in this chapter formed the 30880 kg Ahnighito fragment of the Cape York iron meteorite, on display in the American Museum of Natural History, which contains ~250 g Pt and ~25 g Au. Eventually, a deeper understanding of extraterrestrial silicate magmas and the role of sulfur in differentiation will enable humans to mine the asteroid belt.
ACKNOWLEDGMENTS The author is grateful for thoughtful and thorough reviews by H. Behrens, N. Chabot, and H. Palme, and constructive suggestions by J. W. Webster. All greatly improved this paper. The author acknowledges support from the National Aeronautics and Space Administration Cosmochemistry program (grant NNX10AI42G), and the American Museum of Natural History. This research has made use of NASA’s Astrophysics Data System Bibliographic Services.
REFERENCES Agnor C, Asphaug E (2004) Accretion efficiency during planetary collisions. Astrophys J Lett 613:L157-L160 Albaréde F (2009) Volatile accretion history of the terrestrial planets and dynamic implications. Nature 461:1227-1233 Alexander CMO’D, Grossman JN (2005) Alkali elemental and potassium isotopic compositions of Semarkona chondrules. Meteorit Planet Sci 40:541-556 Alexander CMO’D, Grossman JN, Wang J, Zanda B, Bourot-Denise M, Hewins RH (2000) The lack of potassium-isotopic fractionation in Bishunpur chondrules. Meteorit Planet Sci 35:859-868 Allan DW, Jacobs JA (1956) The melting of asteroids and the origin of meteorites. Geochim Cosmochim Acta 9:256-272 Anders E, Grevesse N (1989) Abundances of the elements: Meteoritic and solar. Geochim Cosmochim Acta 53:197-214 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Basaltic Volcanism Study Project (1981) Basaltic Volcanism on the Terrestrial Planets. Pergamon Press, Inc., New York. 1286 pp. (http://adsbit.harvard.edu/books/bvtp/) Bass JD, Parise JB (2008) Deep Earth and recent developments in mineral physics. Elements 4:157-163 Becker H, Horan MF, Walker RJ, Gao S, Lorand J-P, Rudnick RL (2006) Highly siderophile element composition of the Earth’s primitive upper mantle: Constraints from new data on peridotite massifs and xenoliths. Geochim Cosmochim Acta 70:4528-4550
Sulfur in Extraterrestrial Bodies & Deep Earth
331
Blichert-Toft J, Zanda B, Ebel DS, Albarède F (2010) The solar system primordial lead. Earth Planet Sci Lett 300:152-163 Boehler R (1996) Melting temperature of the Earth’s mantle and core: Earth’s thermal structure. Annu Rev Earth Planet Sci 24:15-40 Boesenberg JS, Delaney JS (1997) A model composition of the basaltic achondrite planetoid. Geochim Cosmochim Acta 61:3205-3225 Bottke WF Jr, Cellino A, Paolicchi P, Binzel RP (eds) (2002) Asteroids III. University Arizona, Tucson, 785 pp Bougher SW, Hunten DM, Phillips RJ (1998) Venus II: Geology, Geophysics, Atmosphere, and Solar Wind Environment, U Arizona Press, 1362 pp Bouhifd MA, Jephcoat AP (2003) The effect of pressure on partitioning of Ni and Co between silicate and ironrich metal liquids: a diamond-anvil cell study. Earth Planet Sci Lett 209:245-255 Bourdon B, Touboul M, Caro G, Kleine T (2008) Early differentiation of the Earth and the Moon. Phil Trans R Soc A 366:4105-4128 Brandon AD, Walker RJ (2005) The debate over core-mantle interaction. Earth Planet Sci Lett 232:211-225 Bruhn D, Groebner N, Kohlstedt DL (2000) An interconnected network of core-forming melts produced by shear deformation. Nature 403:883-886 Buchwald, VF (1975) Handbook of Iron Meteorites, Their History, Distribution, Composition and Structure. U California Press, Berkeley, CA Burbine TH, Meibom AA, Binzel RP (1996) Mantle material in the main belt: Battered to bits? Meteorit Planet Sci 31:607-620 Burkhardt C, Kleine T, Bourdon B, Palme H, Zipfel J, Friedrich J, Ebel D (2008) Hf-W systematics of Ca-Alrich inclusions from carbonaceous chondrites: dating the age of the solar system and core formation in asteroids. Geochim Cosmochim Acta 72:6177-6197 Campbell AJ, Humayun M (1999) Trace element microanalysis in iron meteorites by laser ablation ICPMS. Anal Chem 71:939-946 Campbell AJ, Seagle CT, Heinz DL, Shen G, Prakapenka V (2007) Partial melting in the iron-sulfur system at high pressure: A synchrotron X-ray diffraction study. Phys Earth Planet Inter 162:119-128 Carlson RW, Boyet M (2008) Composition of the Earth’s interior: the importance of early events. Philos Trans R Soc A 366:4077-4103 Chabot NL (2004) Sulfur contents of the parental metallic cores of magmatic iron meteorites. Geochim Cosmochim Acta 68:3607-3618 Chabot NL, Agee CB (2003) Core formation in the Earth and Moon: new experimental constraints from V, Cr, and Mn. Geochim Cosmochim Acta 67:2077-2091 Chabot NL, Campbell AJ, Jones JH, Humayun M, Lauer HV Jr. (2006) The influence of carbon on trace element partitioning behavior. Geochim Cosmochim Acta 70:1322-1335 Chabot NL, Drake MJ (1999) Crystallization of magmatic iron meteorites: the role of mixing in the molten core. Meteorit Planet Sci 34:235-246 Chabot NL, Haack H (2006) Evolution of asteroidal cores. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson p 747-771 Chabot NL, Jones JH (2003) The parameterization of solid metal–liquid metal partitioning of siderophile elements. Meteorit Planet Sci 38:1425–1436 Chabot, NL, Draper, DS, Agee, CB (2005) Conditions of core formation in the Earth: constraints from nickel and cobalt partitioning. Geochim Cosmochim Acta 69:2141-2151 Chang YA, Hsieh K-C (1987) Thermochemical description of the ternary iron-nickel-sulphur system. 25th Annual Conference of Metallurgists (1986 CIM) Canadian Metallurgical Quarterly 26:311-327 Chou CL (1978) Fractionation of siderophile elements in the Earth’s upper mantle. Proceedings of the 9th Lunar and Planetary Science Conference, pp 219-230 Cook DL, Walker RJ, Horan MF, Wasson JT, Morgan JW (2004) Pt-Re-Os systematics of group IIAB and IIIAB iron meteorites. Geochim Cosmochim Acta 68:1413-1431 Corgne A, Wood BJ, Fei Y (2007) C- and S-rich molten alloy immiscibility and core formation of planetesimals. Geochim Cosmochim Acta 72:2409-2416 Corrigan CM, Chabot NL, McCoy TJ, McDonough WF, Watson HC, Saslow SA, Ash RD (2009) The ironnickel-phosphorous system: Effects on the distribution of trace elements during the evolution of iron meteorites. Geochim Cosmochim Acta 73:2674-2691 Cottrell E, Fei Y, Ricolleau A, Prakapenka V (2008) Nickel partitioning between liquid metal and liquid silicate in the LHDAC: Techniques for achieving reliable partition coefficients. Lunar Planet Sci XXXIX #2267 Cottrell E, Walker D (2006) Constraints on core formation from Pt partitioning in mafic silicate liquids at high temperatures. Geochim Cosmochim Acta 70:1565-1580 Crotts APS, Hummels C (2008) Lunar outgassing, transient phenomena, and the return to the Moon II: Predictions and tests for outgassing/regolith interactions. Astrophys J 707:1506-1523 Davis AM (2006) Volatile evolution and loss. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 295-307
332
Ebel
Drake MJ, Righter K (2002) Determining the composition of the Earth. Nature 416:39-44 Dreibus G, Palme H (1996) Cosmochemical constraints on the sulfur content in the Earth’s core. Geochim Cosmochim Acta 60:1125-1130 Ebel DS (2006) Condensaton of rocky material in astrophysical environments. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 253-277 Ebel DS, Grossman L (2000) Condensation in dust-enriched systems. Geochim Cosmochim Acta 64:339-366 Ertel W, Dingwell DB, Sylvester PJ (2008) Siderophile elements in silicate melts - A review of the mechanically assisted equilibration technique and the nanonugget issue. Chem Geol 248:119-139 Ertel W, Walter MJ, Drake MJ, Sylvester PJ (2006) Experimental study of platinum solubility in silicate melt to 14 GPa and 2273 K: Implications for accretion and core formation in Earth. Geochim Cosmochim Acta 70:2591-2602 Fegley B Jr (1991) Thermodynamic models of the chemistry of lunar volcanic gases. Geophys Res Lett 18:3073-3076 Gaetani GA, Grove TL (1997) Partitioning of moderately siderophile elements among olivine, silicate melt, and sulfide melt: Constraints on core formation in the Earth and Mars. Geochim Cosmochim Acta 61:1829-1846 Gaetani GA, Grove TL (1999) Wetting of mantle olivine by sulfide melt: Implications for Re/Os ratios in mantle peridotite and late - stage core formation. Earth Planet Sci Lett 169:147 - 163 Garrick-Bethell I, Weiss BP, Shuster DL, Buz J (2009) Early lunar magnetism. Science 323:356-359 Geissler PE (2003) Volcanic activity on Io during the Galileo era. Annu Rev Earth Planet Sci 31:175-211 Geissler PE, McEwen A, Phillips C, Keszthelyi L, Spencer J (2004) Surface changes on Io during the Galileo mission. Icarus 169:29-64 Gessmann CK, Wood BJ (2002) Potassium in the Earth’s core? Earth Planet Sci Lett 200(1–2):63-78 Ghosal S, Sack RO, Ghiorso MS, Lipschutz ME (1998) Evidence for a reduced, Fe-depleted martian mantle source region of shergottites. Contrib Mineral Petrol 130:346-357 Ghosh A, Weidenschilling SJ, McSween HY Jr, Rubin A (2006) Asteroidal heating and thermal stratification of the asteroid belt. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 555-566 Goldstein JI, Scott ERD, Chabot NL (2009) Iron meteorites: Crystallization, thermal history, parent bodies, and origin. Chemie Erde 69:293-325 Gounelle M, Shu FH, Shang H, Glassgold AE, Rehm KE, Lee T (2006) The irradiation origin of beryllium radioisotopes and other short-lived radionuclides. Astrophys J 640:1163-1170 Grady MM, Wright I (2006) Types of extraterrestrial material available for study. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 3-18 Greenwood RC, Franchi IA, Jambon A, Barrat JA, Burbine TH (2005) Widespread magma oceans on asteroidal bodies in the early Solar System. Nature 435:916-918 Gulliver GH (1913) The quantitative effect of rapid cooling upon the constitution of binary alloys. J Inst Metals 9:120-157 (accessed via http://books.google.com) Haack H, McCoy TJ (2003) Iron and Stony-iron Meteorites. In: Meteorites, Comets and Planets. Treatise on Geochemistry. Volume 1. Davis AM (ed), Elsevier, p 325-345 Haack H, Scott ERD (1993) Chemical fractionations in Group IIIAB iron meteorites: Origin by dendritic crystallization of an asteroidal core. Geochim Cosmochim Acta 57:3457-3472 Halliday AN, Kleine T (2006) Meteorites and the timing, mechanisms, and conditions of terrestrial planet accretion and early differentiation. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 775-802 Hart SR, Gaetani GA (2006) Mantle Pb paradoxes: the sulfide solution. Contrib Mineral Petrol 152:295-308 Hayashi H, Ohtani E, Terasaki H, Ito Y (2009) The partitioning of Pt-Re-Os between solid and liquid metal in the Fe-Ni-S system at high pressure: Implications for inner core fractionation. Geochim Cosmochim Acta 73:4836-4842 Heiken GH, Vaniman DT, French BM (eds) (1991) Lunar Sourcebook. Cambridge U Press, London Herndon JM (1996) Substructure of the inner core of the Earth. Proc Nat Acad Sci USA 93:646-648 Hillgren VJ, Gessmann CK, Li J (2000) An experimental perspective on the light element in Earth’s core. In: Origin of the Earth and Moon. Canup RM, Righter K (eds) U Arizona, Tucson, p 245-264 Holtzman BK, Groebner NJ, Zimmerman ME, Ginsberg SB, Kohlstedt DL (2003) Stress-driven melt segregation in partially molten rocks. Geochem Geophys Geosyst 4:8607 Holzheid A, Grove TL (2002) Sulfur saturation limits in silicate melts and their implications for core formation scenarios for terrestrial planets. Am Mineral 87:227-237 Holzheid A, Sylvester P, O’Neill HSt.C, Rubie DC, Palme H (2000) Evidence for a late chondritic veneer in the Earth’s mantle from high-pressure partitioning of palladium and platinum. Nature 406:396-399 Hong Y, Fegley B Jr (1998) The sulfur vapor pressure over pyrite on the surface of Venus. Planet Space Sci 46:683-690
Sulfur in Extraterrestrial Bodies & Deep Earth
333
Humayun M, Cassen P (2000) Processes determining the volatile abundances of the meteorites and terrestrial planets. In: Origin of the Earth and Moon. Canup RM, Righter K (eds) U Arizona Press, Tucson, p 3-23 Humayun M, Clayton RN (1995) Potassium isotope cosmochemistry: Genetic implications of volatile element depletion. Geochim Cosmochim Acta 59:2131-2148 Jaeger WL, Drake MJ (2000) Solubilities of W, Ga and Co in silicate melts as a function of melt composition. Geochim Cosmochim Acta 64:3887-3895 Jana D, Walker D (1997) The influence of sulfur on partitioning of siderophile elements. Geochim Cosmochim Acta 61:5255-5277 Jephcoat AP, Bouhifd MA, Porcelli D (2008) Partitioning experiments in the laser-heated diamond anvil cell: Volatile content in the Earth’s core. Phil Trans R Soc A 366:4295-4314 Jessup KL, Spencer JR, Yelle R (2007), Sulfur volcanism on Io. Icarus 192:24-40 Jones JH, Drake MJ (1983) Experimental investigations of trace element fractionation in iron meteorites, II: The influence of sulfur. Geochim Cosmochim Acta 47:1199-1209 Jones JH, Malvin DJ (1990) A nonmetal interaction model for the segregation of trace metals during solidification of Fe-Ni-S, Fe-Ni-P, and Fe-Ni-S-P alloys. Metall Trans 21B:697-706 Jones JH, Walker D (1991) Partitioning of siderophile elements in the Fe-Ni-S system: 1 bar to 80 kbar. Earth and Planetary Science Letters 105:127-133 Kargel JS, Delmelle P, Nash DB (1999) Volcanogenic Sulfur on Earth and Io: Composition and Spectroscopy. Icarus 142:249-280 Kargel JS, Komatsu G, Baker VR, Strom RG (1993) The volcanology of Venera and VEGA landing sites and the geochemistry of Venus. Icarus 103:253-275 Kavner A, Walker D (2006) Core/mantle-like interactions in an electric field. Earth Planet Sci Lett 248:301-314 Kavner A, Walker D, Sutton S, Newville M (2007) Externally-driven charge transfer in silicates at high pressure and temperature: A XANES study. Earth Planet Sci Lett 256:314-327 Keil K (2002) Geological history of asteroid 4 Vesta: The “smallest terrestrial planet”. In: Asteroids III. Bottke WF Jr, Cellino A, Paolicchi P, Binzel RP (eds) U Arizona, Tucson, p 573-584 Keil K, Wilson L (1993) Explosive volcanism and the compositions of cores of differentiated asteroids. Earth Planet Sci Lett 117:111-124 Kelly RW, Larimer JW (1977) Chemical fraction in meteorites-VII1 iron meteorites and the cosmochemical history of the metal phase. Geochim Cosmochim Acta 41:93-111 Kimura K, Lewis RS, Anders E (1974) Distribution of gold and rhenium between nickel-iron and silicate melts: implications for the abundance of siderophile elements on the Earth and Moon. Geochim Cosmochim Acta 38:683-701 Kohlstedt DL, Holtzman BK (2009) Shearing melt out of the Earth: An experimentalist’s perspective on the influence of deformation on melt extraction. Annu Rev Earth Planet Sci 37:561-593 Kracher A, Wasson JT (1982) The role of S in the evolution of the parental cores of iron meteorites. Geochim Cosmochim Acta 46:2419-2426 Kress V (2000) Thermochemistry of sulfide liquids II Associated solution model for sulfide liquids in the system O-S-Fe. Contrib Mineral Petrol 139:316-325 Kress V (2003) On the mathematics of associated solutions. Am J Sci 203:708-722 Lazar C, Walker D, Walker RJ (2004) Experimental partitioning of Tc, Mo, Ru, and Re between solid and liquid during crystallization in Fe-Ni-S. Geochim Cosmochim Acta 68:643-651 Lee T, Papanastassiou DA, Wasserburg GJ (1976) Demonstration of Mg-26 excess in Allende and evidence for Al-26. Geophys Res Lett 3:109-112 Li C, Ripley EM (2005) Empirical equations to predict the sulfur content of mafic magma at sulfide saturation and applications to magmatic sulfide deposits. Miner Deposita 40:218-230 Li J, Agee CB (2001) The effect of pressure, temperature, oxygen fugacity and composition on partitioning of nickel and cobalt between liquid Fe-Ni-S alloy and liquid silicate: implications for the Earth’s core formation. Geochim Cosmochim Acta 65:1821-1832 Li J, Fei Y (2003) Experimental Constraints on Core Composition. In: The Mantle and Core. Treatise on Geochemistry Volume 2. Carlson RW (ed), Elsevier, p 520-546 Lindstrom MM, Salpas PA (1983) Geochemical studies of feldspathic fragmental breccias and the nature of North Ray Crater ejecta. Proc Lunar Planet Sci Conf 13th, Pt 2, A671-A683 (J Geophys Res 88 Suppl) Lipschutz ME, Anders E (1961) The record in the meteorites - IV: Origin of diamonds in iron meteorites. Geochim Cosmochim Acta 24:83-88 Liu Y, Samaha N-T, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Lodders K (2000) An oxygen isotope mixing model for the accretion and composition of rocky planets. Space Sci Rev 92:341-354 Lodders K (2003) Solar system abundances and condensation temperatures of the elements. Astrophys J 591:1220-1247
334
Ebel
Lodders K, Fegley B Jr (1998) The Planetary Scientist’s Companion. Oxford U Press, New York Lopes RMC, Kamp LW, Smythe WD, Mouginis-Mark P, Kargel J, Radebaugh J, Turtle EP, Perry J, Williams DA, Carlson RW, Douté S, the Galileo NIMS and SSI Teams (2004) Lava lakes on Io: Observations of Io’s volcanic activity from Galileo NIMS during the 2001 fly-bys. Icarus 169:140-174 Lorand J-P, Luguet A, Alard O (2008) Platinum-group elements: A new set of key tracers for the Earth’s interior. Elements 4:247-252 Lovering JF, Nichiporuk W, Chodos A, Brown H (1957) The distribution of gallium, germanium, cobalt, chromium, and copper in iron and stony-iron meteorites in relation to nickel content and structure. Geochim Cosmochim Acta 11:263-278 Malvin DJ, Jones JH, Drake MJ (1986) Experimental investigations of trace element fractionation in iron meteorites III: elemental partitioning in the system Fe-Ni-S-P. Geochim Cosmochim Acta 50:1221-31 Mann U, Frost DJ, Rubie DC (2009) Evidence for high-pressure core-mantle differentiation from the metalsilicate partitioning of lithophile and weakly-siderophile elements. Geochim Cosmochim Acta 73:73607386 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 McCoy TJ, Mittlefehldt DW, Wilson L (2006) Asteroid Differentiation. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 733-745 McDonough WF (2003) Compositional model for the Earth’s core. In: The Mantle and Core. Treatise on Geochemistry, Volume 2. Carlson RW (ed) Elsevier, p 547-568 McDonough WF, Sun S-S (1995) The composition of the Earth. Chem Geol 120: 223-253 McSween HY, Arvidson RE, Bell JF III, Blaney D, Cabrol NA, Christensen PR, Clark BC, Crisp JA, Crumpler LS, DesMarais DJ, Farmer JD, Gellert R, Ghosh A, Gorevan S., Graff T, Grant J, Haskin LA, Herkenhoff KE, Johnson JR, Jolliff BL, Klingelhoefer G, Knudson AT, McLennan S, Milam KA, Moersch JE, Morris RV, Rieder R, Ruff SW, deSouza PA Jr, Squyres SW, Wänke H, Wang A, Wyatt MB, Yen A, Zipfel J (2004) Basaltic rocks analyzed by the Spirit Rover in Gusev Crater. Science 305:842-845 Minarik WG, Ryerson FJ, Watson EB (1996) Textural entrapment of core-forming melts. Science 272:530-533 Mittlefehldt DW, McCoy TJ, Goodrich CA, Kracher A (1998) Non-chondritic meteorites from asteroidal bodies. Rev Mineral36:4-1–4-195 Morgan JW, Wandless GA, Petrie RK, Irving AJ (1981) Composition of the Earth’s upper mantle: 1 Siderophile trace-elements in ultramafic nodules. Tectonophysics 75:47-67 Mungall JE, Su SG (2005) Interfacial tension between magmatic sulfide and silicate liquids: Constraints on kinetics of sulfide liquation and sulfide migration through silicate rocks. Earth Planet Sci Lett 234:135149 Mysen BO (1991) Relations between structure, redox equilibria of iron, and properties of magmatic liquids. In: Physical Chemistry of Magmas. Perchuk LL, Kushiro I (eds) Springer, New York, p 41-98 Narayan C, Goldstein JI (1982) A dendritic solidification model to explain Ge-Ni variations in iron meteorite chemical groups. Geochim Cosmochim Acta 46:259-268 Nicholls RH Jr (2006) Chronological Constraints on Planetesimal Accretion. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p 463-472 O’Neill HSC (1991) The origin of the moon and the early history of the earth—a chemical model. Part 2: The earth. Geochim Cosmochim Acta 55:1159-1172 O’Neill HSC, Palme H (1998) Composition of the silicate earth: implications for accretion and core formation. In: The Earth’s Mantle, Composition, Structure and Evolution. Jackson I (ed) Cambridge U Press, p 3-126 O’Neill HSC, Palme H (2008) Collisional erosion and the non-chondritic composition of the terrestrial planets. Phil Trans R Soc A 366:4205-4238 Ohtani E, Yurimoto H, Seto S (1997) Element partitioning between metallic liquid, silicate liquid, and lowermantle minerals: implications for core formation of the Earth. Phys Earth Planetary Inter 100:97-114 Pak JJ, Fruehan RJ (1986) Soda slag system for hot metal dephosphorization. Metall Trans 17B:797-804 Palme H, Jones A (2003) Solar system abundances of the elements. In: Meteorites, Comets, and Planets. Treatise on Geochemistry, Volume 1. Davis AM (ed) Elsevier, p 41-61 Palme H, O’Neill HSC (2003) Cosmochemical estimates of mantle composition. In: The Mantle and Core. Treatise on Geochemistry, Volume 2. Carlson RW (ed) Elsevier, p 1-38 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285-314 Prockter LM, Ernst CM, Denevi BW, Chapman CR, Head JW III, Fassett CI, Merline WJ, Solomon SC, Watters TR, Strom RG, Cremonese G, Marchi S, Massironi M (2010) Evidence for young volcanism on Mercury from the third MESSENGER flyby. Science 329: 668-671 Radebaugh J, McEwen AS, Milazzo MP, Keszthely LP, Davies AG, Turtle EP, Dawson DD (2004) Observations and temperatures of Io’s Pele Patera from Cassini and Galileo spacecraft images. Icarus 169:65-79 Righter K (2003) Metal-silicate partitioning of siderophile elements and core formation in the early Earth. Annu Rev Earth Planet Sci 31:135-174
Sulfur in Extraterrestrial Bodies & Deep Earth
335
Righter K (2005) Highly siderophile elements: constraints on Earth accretion and early differentiation. In: Earth’s Deep Mantle: Structure, Composition, and Evolution. van der Hilst RD, Bass J, Matas J, Trampert J (eds), AGU Geophysical Monograph Series 160:203-220 Righter K, Drake M, Yaxley G (1997) Prediction of siderophile element metal–silicate partition coefficients to 20 GPa and 2800 C: the effects of pressure, temperature, oxygen fugacity and silicate and metallic melt compositions. Phys Earth Planet Inter 100:115-134 Righter K, Drake MJ (2000) Metal/silicate equilibrium in the early Earth—new constraints from the volatile moderately siderophile elements Ga, Cu, P, and Sn. Geochim Cosmochim Acta 64:3581-3597 Righter K, Drake MJ (2003) Partition coefficients at high pressure and temperature. In: The Mantle and Core. Treatise on Geochemistry, Volume 2. Carlson RW (ed) Elsevier, p 425-449 Righter K, Humayun M, Campbell AJ, Danielson L, Hill D, Drake MJ (2009b) Experimental studies of metal– silicate partitioning of Sb: Implications for the terrestrial and lunar mantles. Geochim Cosmochim Acta 73:1487-1504 Righter K, Humayun M, Danielson L (2008) Partitioning of palladium at high pressures and temperatures during core formation. Nature online; doi: 101038/ngeo180 Righter K, Pando K, Danielson LR (2009a) Experimental evidence for sulfur-rich martian magmas: Implications for volcanism and surficial sulfur sources. Earth Planet Sci Lett 288:235-243 Rose-Weston LA, Brenan JM, Fei Y, Secco RA, Frost D (2009) Effect of Pressure, Temperature and Oxygen Fugacity on the Metal-Silicate Partitioning of Te, Se, and S: Implications for Earth Differentiation. Geochim Cosmochim Acta 73:4598-4616 Rubie DC, Melosh HJ, Reid JE, Liebske C, Righter K (2003) Mechanisms of metal-silicate equilibration in the terrestrial magma ocean. Earth Planet Sci Lett 205: 239-255 Rushmer T, Minarik WG, Taylor GJ (2000) Physical processes of core formation. In: Origin of the Earth and Moon. Canup RM, Righter K (eds), U Arizona, Tucson, p 227–244 Rushmer T, Petford N, Humayun M, Campbell AJ (2005) Fe-liquid segregation in deforming planetesimals: Coupling core-forming compositions with transport phenomena. Earth Planet Sci Lett 239:185-202 Russell CT, Capaccioni F, Coradini A, De Sanctis MC, Feldman WC, Jaumann R, Keller HU, McCord TB, McFadden LA, Mottola S, Pieters CM, Prettyman TH, Raymond CA, Sykes MV, Smith DE, Zuber MT (2007) Dawn Mission to Vesta and Ceres: Symbiosis between Terrestrial Observations and Robotic Exploration. Earth Moon Planets 101:65-91 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671-698 Scheil E (1942) Bemerkungen zur Schichtkristallbildung. Zeit Metallkunde 34:70-72 Scott ERD (1977) Composition, mineralogy and origin of group IC iron meteorites. Earth Planet Sci Lett 37:273-284 Scott ERD (1972) Chemical fractionation in iron meteorites and its interpretation. Geochim Cosmochim Acta 36:1205-1236 Scott ERD, Wasson JT (1975) Classification and properties of iron meteorites Rev Geophys Space Phys 13:527-546 Shearer CK, Burger PV, Guan Y, Papike JJ, Sutton S (2010) Vapor element transport in the lunar crust: Implications for lunar crustal processes, water on the Moon and lunar ore deposits. Annual LEAG Meeting, abstract #3031, Lunar and Planetary Institute Shearer CK, Papike JJ, Rietmeijer FJM (1998) The planetary sample suite and environments of origin. Rev Mineral 36:1-1–1-28 Spencer J R, Stern SA, Cheng AF, Weaver HA, Reuter DC, Retherford K, Lunsford A, Moore JM, Abramov O, Lopes RMC, Perry JE, Kamp L, Showalter M, Jessup KL, Marchis F, Schenk PM, Dumas C (2007) Io volcanism seen by New Horizons: A major eruption of the Tvashtar Volcano”. Science 318:240-243 Stadermann FJ, Walker RM, Zinner E (1999) Sub-mm isotopic measurements with the CAMECA NanoSIMS. Lunar Planet Sci XXX, Abstract #1407, LPI Houston Stewart AJ, Schmidt MW, van Westrenen W, Liebske C (2007) Mars: A new core-crystallization regime. Science 316:1323-1325 Stewart AJ, van Westrenen W, Schmidt MW, Günter D (2009) Minor element partitioning between fcc Fe metal and Fe-S liquid at high pressure: The role of crystal lattice strain. Earth Planet Sci Lett 284:302-309 Tatsumoto M, Knight RJ, Allègre CJ (1973) Time differences in the formation of meteorites as determined from the ratio of lead-207 to lead-206. Science 180:1279-1283 Uhlig HH (1954) Contribution of metallurgy to the origin of meteorites: Part I—Structure of metallic meteorites, their composition and the effect of pressure. Geochim Cosmochim Acta 6:282-301 Ulff-Møller F (1998) Effects of liquid immiscibility on trace element fractionation in magmatic iron meteorites: A case study of group IIIAB. Meteorit Planet Sci 33:207-220 Ulff-Møller F, Choi B-G., Rubin AE, Tran J, Wasson JT (1998) Paucity of sulfide in a large slab of Esquel: New perspectives on pallasite formation. Meteorit Planet Sci 33:221-227
336
Ebel
Van Orman JA, Keshav S, Fei Y (2008) High-pressure solid/liquid partitioning of Os, Re and Pt in the Fe–S system. Earth Planet Sci Lett 274:250-257 Vervack RJ Jr, McClintock WE, Killen RM, Sprague AL, Anderson BJ, Burger MH, Bradley ET, Mouawad N, Solomon SC, Izenberg NR (2010) Mercury’s complex exosphere: Results from MESSENGER’s third flyby. Science 329:672-675 Wade J, Wood BJ (2005) Core formation and the oxidation state of the Earth. Earth Planet Sci Lett 236:78-95 Walker D (2000) Core participation in mantle geochemistry: Geochemical Society Ingerson Lecture, GSA Denver, October 1999. Geochim Cosmochim Acta 64:2911-2987 Walker D, Li J (2008) Partitioning of molybdenum to 60 kbar along a warped Fe-FeS liquidus. Chemical Geology 248:166-173 Wänke H (1981) Constitution of terrestrial planets. Philos Trans R Soc A 303:287-302 Wänke H, Dreibus G (1988) Chemical composition and accretionary history of the terrestrial planets. Philos Trans R Soc A 325:545-557 Wänke H, Dreibus G, Jagoutz E (1984) Mantle chemistry and accretion history of the Earth. In: Archaean Geochemistry. Kröner A, Hanson GN, Goodwin AM (eds), Springer–Verlag, Berlin, p 1-24 Wasson JT (1967) The chemical classification of iron meteorites: I A study of iron meteorites with low concentrations of gallium and germanium. Geochim Cosmochim Acta 31:161-180 Wasson JT (1974) Meteorites: Classification and Properties. Springer-Verlag Wasson JT (1985) Meteorites, Their Record of Early Solar System History. W H Freeman, New York Wasson JT, Choi B-G (2003) Main-group pallasites: Chemical composition, relationship to IIIAB irons, and origin. Geochim Cosmochim Acta 67:3079-3096 Wasson JT, Kallemeyn GW (1988) Composition of chondrites. Phil Trans R Soc London A 325:535-544 Wasson JT, Kallemeyn GW (2002) The IAB iron-meteorite complex: A group, five subgroups, numerous grouplets, closely related, mainly formed by crystal segregation in rapidly cooling melts. Geochim Cosmochim Acta 66:2445-2473 Wasson JT, Ouyang X, Wang J, Jerde E (1989) Chemical classification of iron meteorites: XI Multi-element studies of 38 new irons and the high abundance of ungrouped irons from Antarctica. Geochim Cosmochim Acta 53:735-744 Weber RC, Lin P-Y, Garner EJ, Williams Q, Lognonne P (2011) Seismic detection of the lunar core. Science 331:309-312 Weisberg MK, McCoy TJ, Krot AN (2006) Systematics and evaluation of meteorite classification. In: Meteorites and the Early Solar System II. Lauretta D, McSween HY Jr (eds) U Arizona, Tucson, p19-52 Weitz CM, Rutherford MJ, Head JW III, McKay DS (1999) Ascent and eruption of a lunar high-titanium magma as inferred from the petrology of the 75001/2 drill core. Meteorit Planet Sci 34:527-540 Westphal AJ, Fakra SC, Gainsforth Z, Marcus MA, Ogliore RC, Butterworth AL (2009) Mixing fraction of inner solar system material in comet 81P/Wild2. Astrophys J 694:18-28 Wheeler KT, Walker D, Fei Y, Minarik WG, McDonough WF (2006) Experimental partitioning of uranium between liquid iron sulfide and liquid silicate: Implications for radioactivity in the Earth’s core. Geochim Cosmochim Acta 70:1537-1547 Wieczorek MA, Jolliff BL, Khan A, Pritchard ME, Weiss BP, Williams JG, Hood LL, Righter K, Neal C R, Shearer CK, McCallum IS, Tompkins S, Hawke BR, Peterson CD, Gillis JJ, Bussey B (2006) The Constitution and Structure of the Lunar Interior. Rev Mineral Geochem 60:221-364 Williams JG, Boggs DH, Ratcliff JT (2004) Lunar core and tides. Lunar Planet Sci XXXV #1398 Williams JG, Boggs DH, Yoder CF, Ratcliff JT, Dickey JO (2001) Lunar rotational dissipation in solid body and molten core. J Geophys Res – Planets 106:27933-27968 Willis J, Goldstein JI (1982) The effects of C, P, and S on trace element partitioning during solidification in Fe-Ni alloys. J Geophys Res 87:A435-A445 Wood BJ (2008) Accretion and core formation: constraints from metal-silicate partitioning. Philos Trans R Soc A 28:4339-4355 Wood BJ, Wade J, Kilburn MR (2008) Core formation and the oxidation state of the Earth: Additional constraints from Nb, V and Cr partitioning. Geochim Cosmochim Acta 72:1415-1426 Yamaguchi A, Taylor GJ, Keil K (1997) Metamorphic history of the eucritic crust of 4 Vesta. J Geophys Res 102:1381-13386 Yoder CF, Konopliv AS, Yuan DN, Standish EM, Folkner WM (2003) Fluid core size of Mars from detection of the solar tide. Science 300:299-303 Zolensky ME (1998) The flux of meteorites to Antarctica. In: Meteorites: Flux with Time and Impact Effects. Grady MM, Hutchison R, McCall GJH, Rothery DA (eds) Geological Society, London, Special Publications 140:93-104
12
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 337-361, 2011 Copyright © Mineralogical Society of America
Fining of Glass Melts Hayo Müller-Simon Hüttentechnische Vereinigung der Deutschen Glasindustrie (HVG) Siemensstraße 45 63071 Offenbach am Main, Germany [email protected]
INTRODUCTION The degassing of silicate melts plays an important role in both the Earth and glass sciences. Volcanic eruptions are driven by the expansion of exsolved volatiles. These volatiles are initially dissolved in magmas at high pressures and later released as one or more fluid phases during ascent of the magma to the Earth’s surface due to decompression (Webster and Botcharnikov 2011, this volume; Oppenheimer et al. 2011, this volume). During fining of glass melts the oversaturation of the silicate liquid with respect to dissolved volatiles is induced by heating. Although the process of volatile oversaturation occurring in geologic and industrial melts differs, the fundamental mechanisms controlling melt degassing are similar in both cases, particularly with respect to the nucleation and growth of gas bubbles in the melt and their upward migration driven by buoyancy. Another similarity of natural and technical melt degassing is the role of sulfur. Tremendous amounts of sulfur in form of SO2 and H2S are produced by volcanic eruptions at the surface or by the cryptic degassing of magmas ascending through the earth’s crust (see Oppenheimer et al. 2011, this volume). During the fining of industrial glass melts, one takes advantage of the solubility behavior of sulfur in silicate melts. Sulfate added to the melt batch as a raw material, decomposes to release gaseous compounds during fining that enhance the expansion of existing bubbles which can, as a result, ascend more readily to the melt surface. Hence, the release of gaseous sulfur dioxide is a major issue in both volcanic eruptions and industrial glass production. In this chapter we will have a detailed look at the process of fining in industrial mass glass production. Glass products are found in wide fields of daily life, such as automotive glazing, architecture, food and pharmaceutical packaging, and optical devices or opto-electronic components in computers and consumer electronics. The widespread usage of glass arises from the great variety of properties that can be specifically altered by changing the glass composition. Most industrial glasses belong to the group of soda-lime-silicate glasses, comprising container glass, flat glass, fiber glass and tableware. These so-called mass glasses account for far above 90% of the glass tonnage produced. All these mass glasses are fined by means of sulfur containing raw materials. After sulfur oxide emissions derived from fuel oil firing, the usage of fining agents produces the largest portion of sulfur oxide emissions in the glass industry. Being a substantial contributor to rain acidification, sulfur oxide emissions are thus limited by legal regulation. In order to satisfy these limits in the most economically effective way, it is important to understand in detail the main fining reactions utilized in industrial glass production. Environmental issues strongly influence the actual mass glass production through the need to use filter dust, which is formed when sulfur oxide is absorbed from the exhaust gas in the waste gas treatment plant (Kircher 1993; Krauß 1995), and waste glass which can be used by more than 90% as raw material, so-called recycled cullet, (Schaeffer 1996). Both substances provide a very varying composition making their use extremely difficult and both have a strong influence on the sulfur household in an industrial glass melting tank. 1529-6466/11/0073-0012$05.00
DOI: 10.2138/rmg.2011.73.12
Müller-Simon
338
The glass production process consists of the preparation of the batch from the raw materials, the melting of the batch, fining, homogenization, conditioning, and finally the forming as shown in Figure 1. The thoroughly mixed batch is fed into the glass melting tank, where the raw materials are heated rapidly to temperatures of more than 1200 °C. Upon heating the components (e.g., soda ash, lime stone and sand) react to form a glass melt (for details see Falcone et al. 2011, this volume). The primary melt contains many bubbles of air originally trapped between the grains of the raw materials, as well as CO2 formed through combustion of organic impurities and decomposition of carbonates. The presence of these bubbles in the glass yields a final product of unacceptable quality. In order to remove these bubbles, the melt is “fined” by further heating to temperatures between 1400 °C and 1600 °C (depending on the glass composition). This increase in temperature causes the bubbles to expand and ascend through the lower viscosity melt, where they are eventually released into the furnace atmosphere. The rising velocity v of a bubble can be estimated by the Stokes approximation (Mulfinger 1980) v=
2 gr 2 (ρ1 − ρ2 ) 9η1
(1)
where g is the acceleration of gravity, r is the bubble radius, r1 is the density of melt, r2 is the density of the bubble content and h1 is the melt viscosity. Figure 2 shows the dependence of the rising velocity as a function of the bubble diameter and the temperature for a melt of sodalime-silica composition. To a first approximation the gases in the bubbles will obey the ideal gas law; as such by applying the ideal gas law, the bubble diameter is calculated to enlarge by only 20% when the temperature is increased from 1200 °C to 1500 °C. This increase will not have a significant effect on the rising velocity of the bubble. Thus, in order to shorten the fining process, agents which release additional gas as a function of increasing temperature are added to the batch. This fining gas then diffuses into and enlarges the existing bubbles, thereby increasing their buoyancy. This type of chemical fining is the most widespread fining method utilized today. However, if chemical fining is not possible or efficient, there are alternate physical methods that can improve the fining of the glass melt. Physical fining may be accomplished with the application of ultrasonic waves, the application of subatmospheric pressure, and bubbling (i.e., introducing the fining gas through holes in the bottom of the glass melting tank) (Krämer 1999).
1 2 3
Figure 1. Sketch of the glass production process. Fig.1
Fining of Glass Melts
339
Figure 2. Rising velocity of bubbles as a function of bubble diameter and temperature in a soda-lime-silica melt.
After fining and homogenization, the melt is cooled to its forming temperature and fed into the forming machines (cf. Falcone et al. 2011, this volume). During cooling, fining bubbles that may still be present in the melt are resorbed. This process is sometimes called refining. Between refining and forming, melts for high quality optical glasses are additionally homogenized with a stirring system. This is done in order to minimize variations in chemical composition and related properties such as the refractive index. With the glass melt now achieved, the properties of the product are defined. After forming, most glass products undergo annealing in the lehr for residual stress to be removed (cf. Falcone et al. 2011, this volume).
AGENTS USED FOR CHEMICAL FINING Chemical fining exploits the fact that some compounds react to produce gas with increasing temperature (Cable 1960, 1961a,b; Krämer 1999). The fining process starts when the partial pressure of the fining gas in gaseous inclusions, e.g., O2 in Equation (3), is reached as temperature is increased (Beerkens 1995). The driving force for the diffusion of the fining gas into the bubbles is the difference between the fining gas concentration in the bubble/melt interface and the equilibrium concentration in the melt (at the given temperature). The bubble volume change with respect to temperature change is given by the differential equation (Cable 1961a; Krämer 1979): 2 dVi 4 πr Di ( cia − cib ) 1 1 = + ρ dt Di πt r
(2)
with the partial volume Vi of component i, time t, bubble radius r, diffusion coefficient Di of component i, concentration of component i in the melt cia and concentration of component i at the melt/bubble interface cib. After batch melting, gaseous inclusions consist predominantly
340
Müller-Simon
of CO2, CO, N2 and H2O. Thus, only small partial pressures of the fining gas are required in order to achieve a sufficient concentration gradient to start the growth of bubbles. The choice of the fining agent depends on the viscosity-temperature characteristics of the glass melt. The earliest release of the fining gas at increasing temperature should occur when the melt viscosity is low enough to allow bubbles to escape the melt. In other words, this should occur at the temperature where they have sufficient buoyancy and the lamellae between them and the furnace atmosphere collapse in a sufficiently short time. The most common fining agents are sodium chloride, antimony oxide and sodium sulfate. The most straightforward mechanism is oxygen fining, for example by the decomposition of antimony oxide Sb2 O5 ( m) → Sb 2O3 ( m) + O2 ( g )
(3)
If the glass melt is considered as an electrochemical solution, the fining action can be described by 1 Sb5 + ( m) + O2 − ( m) Sb3 + ( m) + O2 ( g ) 2
( 4)
concentration of pentavalent antimony
Figure 3 shows the general development of the oxygen partial pressure with increasing temperature, i.e., the temperature at which the fining pressure is reached depends on the Sb2O5/Sb2O3 ratio. During fining, not only CO2, CO, N2 and H2O trapped during the initial melting, but also the oxygen liberated from the fining reaction itself are expelled from the melt. As the fining reaction continues the Sb2O5/ Sb2O3 ratio shifts to lower values; thus, the fining temperature needs to be increased in order for the process to continue. Antimony fining starts at about 1300 °C. Alternative substances used for oxygen fin-6 ing are arsenic and cerium oxide. Sodium7 chloride sublimates at approximately 14408 °C at ambient pressure and is used for production of higher melting glasses such as borosilicate glass.
oxygen partial pressure
The O2− activity in the melt is mainly controlled by the addition of alkaline or alkaline earth oxides (i.e., by the bulk glass composition); therefore, the O2− activity of the melt is buffered with respect to Equation (4). The ionic description has some advantages when the interaction between pairs of polyvalent elements in the melt is considered (for instance see Eqn. 27).
fining pressure
T1
T2
temperature Fig. 3:
Figure 3. Principal temperature dependences of oxygen fining by means of antimony.
For the fining of industrial container glass, flat glass and fiber glass, sulfur-containing raw materials are used almost exclusively today. In addition to the fining effect, sodium sulfate and sodium sulfide form a low liquid eutectic that improves the dissolution of sand grains in the raw material; in other words they act as melting accelerator. Furthermore, sulfide ions are an important component in the amber color of glasses (cf. Falcone et al. 2011, this volume). The fining effect of sulfate was discovered accidentally in the second half of the 19th century when the production method of soda ash for the glass industry changed from the Leblanc process to the Solvay process (Thieme 1993). The new soda as produced by the Solvay
Fining of Glass Melts
341
process extended the fining time in many glass works, and enhanced the sensitivity of the melt to fluctuations of the firing conditions. Glass technologists soon noted that the contamination of soda ash produced by the Leblanc process with sulfate was unexpectedly advantageous. The sulfur fining process itself goes back to L. Bock, who performed extensive experiments in the Bohemian glass industry, eventually being granted a patent for the use of alkali sulfates in glass fining in 1902 (Weyl 1943). The fundamentals of sulfur chemistry in glass melts have been discovered as early as the 1950’s in the landmark study of Fincham and Richardson (1954). These findings have been reconfirmed for silicate glasses several times (e.g., Nagashima and Katsura 1973; Schreiber et al. 1987). Reviews of sulfur chemistry research in silicate glass melts have been given in several papers (Goldman 1985; Müller-Simon 1999; Beerkens 2007). A consistent description of sulfur fining in industrial glass production is not a simple task because sulfur acts in different capacities during different stages of the melting process. In the continuous melting tanks, which are common in the glass industry today, separation of the discrete sub-steps of the melting process is almost impossible. Thus, in order to understand industrial sulfur fining, knowledge about sulfur chemistry in glass melts must be drawn from laboratory experiments and subsequently applied to industrial conditions. The application of this knowledge has proven to be an intricate task. This is especially true because industrial glass melting furnaces are used for the production of high-quality glass products at specific conditions, which forbids parameter changes for the pursuit of basic scientific interest. The primary issue lies in the fact that the final state of the glass melt after fining cannot be easily assigned to either a steady or equilibrium state. For economic reasons the glass melting process is allowed a limited period only, and the fining is stopped before all bubbles can leave the melt. The flames and heating electrodes utilized in the melting process induce a temperature distribution in the glass bath. Together with the glass melt fluxes caused by temperature differences, the temperature distribution determines the total bubble volume at the end of fining. This distribution of solubility states is mixed during homogenization at the end of the melting process. Partial pressures inside the remaining gas volume and equilibria in the melt will adjust accordingly. In order to address to the dynamics of the sulfur dissolution in industrial glass melting, glass technologists often speak of sulfur retention instead of sulfur concentration. Thus, in the first section of this article the reactions of sulfur in glass melts will be identified based upon what is known from laboratory experiments. In the second section the information which can be deduced from the behavior of glass melting tanks about sulfur chemistry is collected. Comparison of the information from each of the two sections provides important insight into the fining action of sulfur in industrial glass melting tanks.
LABORATORY EXPERIMENTS ON SULFUR CHEMISTRY Laboratory experiments on sulfur chemistry can be divided into two types of experiments: equilibrium experiments where glass melts are equilibrated with an atmosphere or liquid of known composition, and melting experiments where a raw material batch is melted under defined heating characteristics.
Equilibrium experiments An atmosphere which contains sulfur and oxygen in the temperature range between 1000 and 1600 °C, is dominantly composed of the sulfur in the form of SO2 with minor amounts of SO3 or S2 if an oxygen excess and oxygen deficit occurs, respectively. Obviously, in the gas phase the dominant redox reactions are: 1 SO3 ( g ) SO2 ( g ) + O2 ( g ) 2
(5)
342
Müller-Simon
and 1 S2 ( g ) + O2 ( g ) SO2 ( g ) 2
(6)
If a glass melt is brought into equilibrium with such an atmosphere, only the oxidized and the reduced form of sulfur are present in the melt (see Wilke et al. (2011, this volume) for spectroscopic studies on sulfur speciation in glasses and melts) despite the fact that SO2 is present in a much larger concentration in the atmosphere than SO3 and S2. At constant temperature, glass composition, and oxygen partial pressure, the amount of dissolved sulfur in the melt increases with increasing sulfur concentration in the gas phase. Thus, sulfur concentration in the melt can be assumed to follow the solution behavior described by Henry’s law (Fincham and Richardson 1954; Nagashima and Katsura 1973; Schreiber et al. 1987). The concentration of dissolved sulfur increases also with increasing melt oxygen ion activity (Fincham and Richardson 1954; Nagashima and Katsura 1973), suggesting that the dissolution occurs via formation of sulfate anions and sulfide anions (Fincham and Richardson 1954; Nagashima and Katsura 1973): SO3 ( g ) + O2 − ( m) → SO24 − (m)
( 7)
SO2 → not soluble in glass melts
(8)
S2 ( g ) + 2O2 − ( m) → 2S2 − ( m) + O2 ( g )
(9)
The influence of the free oxygen ion activity in glass melts is referred as basicity. Since the chemical solubility of gases such as CO2, SO3 or H2O strongly depends on the basicity of the melt (Paul 1982), glass technologists have developed several means to quantify this property. The most common methods are based on the estimation of the strength of the cation-oxygen bond (Sun 1947, 1948; Krämer 1991) or the measurement of the position of the sp-transition of the 6s electrons of Pb2+. The property measured in the latter technique is known as optical basicity (Duffy and Ingram 1976). For glasses that do not contain lead, the optical basicity can be calculated by multiplying the optical basicities of the single oxides with their respective mole fractions in the glass (Duffy 1989). Although free oxygen ions exist in glass melts only in very small concentrations, they may nevertheless control the dissolution of sulfur in the melt because of the equilibrium between bridging oxygens O0, non-bridging oxygens O− and free oxygen ions O2− O0 (m) + O2 − (m) 2 O − (m)
(10)
(Toop and Samis 1962). Combination of the redox reactions in Equations (5) and (6) with the dissolution reactions in Equations (7) and (9) yields 1 SO24 − ( m) SO2 ( g ) + O2 ( g ) + O2 − ( m) 2
(11)
3 S2 − ( m) + O2 ( g ) SO2 ( g ) + O2 − ( m) 2
(12)
and
These findings are displayed in Figure 4. With increasing temperature the concentration of SO3 decreases and the concentration of S2 increases. Commonly polyvalent elements become more reduced with increasing temperature (Paul 1982). Applying the law of mass action to the reactions in Equations (11) and (12) yields the respective reaction constants: ∆HS0ox − T ∆SS0ox K Sox = exp − RT
[SO24 − ] = 1 2 pSO2 ⋅ pO2 ⋅ aO2−
(13 3)
sulfur concentration in the melt 10 11 12
S2(gas) SO2(gas) SO42-(melt) S2-(melt)
SO3(gas)
log (partial pressure of sulfur species in the gas phase)
Fining of Glass Melts
9
343
Figure 4. Equilibrium concentrations of SO3, SO2 and S2 in the gas phase calculated by the computer code FactSage and SO42− and S2− in glass melts according to Nagashima and Katsura (1973).
log (oxygen partial pressure) Fig. 4
and K Sred
∆HS0red − T ∆SS0red = exp − RT
[S2 − ] ⋅ pO22 = pSO2 ⋅ aO2− 3
(14)
where DH0 is the standard enthalpy and DS0 is the standard entropy for the respective reaction, T is the temperature, R is the gas constant, [S2−] and [SO42−] are the concentrations of the respective species, p is the respective partial pressures, and aO2– is the O2– activity. The temperature dependence in Equations (13) and (14) describes an Arrhenius behavior, which is found for all polyvalent elements in glass melts (Rüssel 1990). Rearranging Equation (13) and (14) yields 1 log [SO24 − ] = log pSO2 + log pO2 + log K Sox ⋅ aO2− 2
(15)
and 4
3 log [S2 − ] = log pSO2 − log pO2 + log K Sred ⋅ aO2− 2
(16)
In both silicate slags (Fincham and Richardson 1954) and glass melts (Nagashima and Katsura 1973) the experimental results agree well with these equilibria, showing the typical V-shape of sulfur solubility as a function of the logarithm of oxygen partial pressure (Fig. 4). According to Equation (15) and (16), the sulfur concentration should increase with increasing oxygen partial pressure with a slope of 0.5 on a logarithmic scale under oxidizing conditions and decrease with a slope of 1.5 under reducing conditions. This matches well in the oxidizing region but poorly under reducing conditions. Equations (15) and (16) imply that both S2− and SO42− have the same Henry’s law solubility, which cannot be assumed a priori. Moreover, the fact that no SO32− is found in the glass (Backnaes et al. 2008) does not prove that it does not exist in the melt. Combining Equations (11) and (12) and assuming the existence of SO32− in the melt yields 3 SO24 − ( m) + S2 − ( m) 4 SO32 − ( m)
(17)
Using the known reaction constants of Equations (11) and (12) (Müller-Simon 1998a; Beerkens 2003) demonstrates that the equilibrium in Equation (17) is shifted leftward during cooling; thus, SO32− may only be found by species sensitive method at high temperatures. Therefore it is possible that small concentrations of tetravalent sulfur may disturb the slope
Müller-Simon
344
of solubility curves predicted by Equations (15) and (16). However, the SO2 release when the glass fusion is finished proves that the SO2 solubility is much smaller than that of SO3 and S2 (Pajean et al. 1975; Klouzek et al. 2007; Collignon et al. 2010). Figure 5 shows the temperature dependence of dissolved sulfur under both oxidizing and reducing conditions (Nagashima and Katsura 1973). Obviously, reaction (11) is shifted to the right side with increasing temperature while reaction (12) is shifted to the left side with increasing temperature. Thus, under oxidizing conditions Equation (15) may serve as a description of sulfur refining. The description of sulfur refining under reducing conditions requires additional assumptions. A quantitative thermochemical description of sulfur chemistry in glass melts based on the above reactions is given in (Beerkens 2003).
sulfur concentration in wt.-% S
0,200
1250 °C
1300 °C
0,020
0,002 -14 13 14
-12
-10
-8
-6
-4
-2
log (oxygen partial pressure in bar) Fig. 5
Figure 5. Temperature dependence of sulfur dissolution according to Nagashima and Katsura (1973).
Instead of equilibrating the glass melt with a gaseous atmosphere, the melt can alternatively be equilibrated with a molten salt. If a soda-lime-silicate glass melt is exposed to a sodium sulfate melt, sulfur (from the sodium sulfate liquid) diffuses into the silicate melt until a maximum sulfur concentration of 0.25 wt% SO3 is attained (Brückner 1962). These phase relations indicate that there is a miscibility gap in the soda-lime-silicate glass melt/ sodium sulfate system. Thus, exposing the glass melt to a pure SO3 atmosphere results in the formation of a film of sodium sulfate, so-called sulfate gall, on the melt’s surface. In form of so-called salt bubbles this gall causes a severe quality reduction. Sulfur solubility also is influenced by the water concentration of the glass melt (Hanke and Scholze 1970). Fining bubbles contain SO2 and H2O, which dissolve during cooling according to their respective partial pressures and solubilities. If both components are present in the melt this results in a typical inverse solution behavior, as a function of the oxygen partial pressure, for industrial glasses (Geotti-Bianchini and De Riu 1998). It should be noted that under common fining conditions, only sulfur5 has different redox states, which govern the temperature dependent fining process, although water dissolved in the melt will considerably reduce the need for fining agents (Beerkens 2005, 2007).
Melting experiments Equilibrium experiments are often said to be somewhat academic, whereas melting experiments are preferred because they are more like the industrial melting process. However, the complexity of the melting reactions makes the interpretation of these melting experiments inherently difficult. Much of this difficulty lies in the fact that, under industrial conditions
Fining of Glass Melts
345
equilibrium behavior may only be valid for an infinitesimally small melt volume element. In such melting experiments, a batch of raw material is placed in a crucible, heated to the melting temperature and held for some time at maximum temperature before being cooled down. In order to investigate sulfur chemistry, the properties of the final glass such as redox state, sulfur concentration, or the bubble density are measured (Williams 1980). Figure 6 shows the results of some such experiments (Beerkens 2005). The relation between sulfur concentration and redox state differs definitely for equilibrium experiments (Fig. 5) and melting experiments (Fig. 6), especially in the region of the minimum of the sulfur solubility. More information can be obtained when the composition of the gases, which are released during heating, is measured by techniques such as evolved gas analysis (EGA) (Pajean et al. 1975; Kloužek et al. 2007; Collignon et al. 2010). Figure 7 shows typical EGA curves of the SO2 release from sulfur containing batches. A considerable reaction velocity is reached when the first liquid phases appear. This occurs at approximately 760 °C where some salts with low melting points, such as Na2CO3, 0,20 0,18
sulfur concentration in wt% S
0,16 0,14 0,12 0,10 0,08 0,06 0,04 0,02 0,00 ‐8
15 16
‐7
‐6
‐5
‐4
‐3
log oxygen partial pressure (in bar) at 1673 K
‐2
‐1
Fig. 6:
normalized gas release
Figure 6. Sulfur concentration from melting experiments according to Beerkens (2005).
800
900
1000
1100
1200
1300
1400
1500
1600
temperature 17 18 Figure
Pajean 1975, reduced
Klouzek 2007, SiO2+CaCO3+C+Na2SO4
Pajean 1975, oxidized
Klouzek 2007, SiO2+C+Na2SO4
Fig. 7: of SO2 release from different raw material combinations measured using evolved gas 7. Results analysis, according to Pajean et al. (1975) and Klouzek et al. (2007). 19 6
346
Müller-Simon
CaCO3 and NaNO3, have eutectics. This point can be considered the starting point for the batch reactions. At 800 °C the first metasilicates are formed: Na 2 CO3 (l ) + nSiO2 (s ) → Na 2 O ⋅ nSiO2 (m) + CO2 ( g )
(18)
where the CO2 reacts with the coke in the batch according to the Boudouard reaction CO2 ( g ) + C( s ) 2 CO( g )
(19)
According to the Boudouard equilibrium above 800 °C the ratio of CO/CO2 is >0.9. The presence of CO is essential for the reduction of the sodium sulfate (Flick and Nölle 1995). The most important sulfur-containing raw materials for industrial glass production are sodium sulfate, blast furnace slags, and filter dust which contains up to 50 wt% SO3. Each of these materials has a completely different melting behavior. Sodium sulfate is strongly involved in the batch reactions, while blast furnace slags hardly react in the batch but dissolve into the primary melt in the later stages of the glass melting process. The melting point of sodium sulfate is approximately 880 °C (Brückner 1962; Samadhi et al. 2001). The decomposition of sodium sulfate starts at 1600 °C (Samadhi et al. 2001) and takes place according to 1 Na 2SO 4 (l ) Na 2 O(l ) + SO2 ( g ) + O2 ( g ) 2
(20)
where Na2O is in equilibrium with gaseous Na and NaO, and SO2 is in equilibrium with SO3. As discussed above, the SO2/SO3 equilibrium is shifted nearly completely into the direction of SO2 at high temperature under typical oxygen partial pressures The reaction behavior of sodium sulfate changes, however, if it is combined with coke. At lower temperatures, considering Equation (19), sodium sulfate is reduced via a rapid solid-gas reaction. Na 2SO 4 ( s ) + 4CO( g ) Na 2S(s ) + 4CO2 ( g )
(21)
As soon as the melting point of sodium sulfate is reached, the liquid-solid reaction starting at 740 °C also becomes important. Na 2SO 4 (l ) + 2C(s ) Na 2S(l ) + 2CO2 ( g )
(22)
At this temperature, the system Na2SO4-Na2S has a eutectic and thus a liquid is formed at much lower temperatures than with pure sodium sulfate (this temperature is even lower than most carbonate eutectics). This liquid covers the sand grains and reacts according to Equation (18). Part of the sodium sulfate does not act according to reactions (21) and (22) but rather to the following reactions Na 2SO 4 (l ) + CO( g ) Na 2 O(s ) + SO2 ( g ) + CO2 ( g )
(23)
2 Na 2SO 4 (l ) + C( s ) 2 Na 2 O( s ) + 2 SO2 ( g ) + CO2 ( g )
(24)
and Under batch melting conditions the Na2SO4·Na2S liquid decomposes with increasing temperature according to 3 Na 2SO 4 (l ) + Na 2S(l ) 4 Na 2 O(l ) + 4 SO2 ( g )
(25)
At the end of the batch reactions a xNa2O·yCaO·zSiO2 melt exists (x,y,z represent the variable fractions of the different components) in which the sulfur is mainly dissolved as both SO42− and S2−(Beerkens 2003). Further increase of the temperature produces both SO2 and O2 according
Fining of Glass Melts
347
to Equation (15). With advanced fining (i.e., when most of the CO2, CO, H2O and N2 has been expelled from the melt), the relation pSO2 + pO2 = 1 bar may be assumed for the fining gas. These reactions can partially be identified in the EGA measurement results (Fig. 7). The first peak on the left side coincides with the occurrence of the first liquid phases, indicating that the sulfur release may be caused by reactions (23) and (24). The second peak from the left is most probably caused by reaction (25), and the peaks on right side can be attributed to reaction (20) (Collignon et al. 2011). The precise position of the fining peak can vary depending on the given composition (Klouzek et al. 2007). It is obvious that the sulfur incorporation into the melt is not only determined by the identity of the sulfur compounds, but also by carbon and its oxides (Bassine et al. 1987; Nölle and Al Hamdan 1990). Figure 8 shows the redox state of the glass given by the iron redox ratio as a function of the carbon addition (Bassine et al. 1987). The oxidation state of such a system is similar to a titration curve. The curve displays three characteristic properties: (1) in the case of sulfate excess the pO2 is close to 0.33 bar, i.e., the decomposition described by Equation (20); (2) in the case of coke excess the pO2 adjusts itself between the equilibrium with CO2 from the carbonate decomposition and the Boudouard equilibrium (Nölle and Al Hamdan 1990); and (3) the transition point between the two is given by the portion of coke which is consumed for the reduction of sodium sulfate according to Equation (24). This behavior shows the role of sulfates in glass batches as oxidizing agents, which is also often utilized for oxidation of the reducing contaminations of recycled cullet like paper and food relics. Interpretation of these types of batch melting experiments is difficult because melt temperature is not directly quantified with thermocouples (Conradt et al. 1994). Rather the temperature must be estimated from the rate of thermal input, furnace temperature, and the time scale. With this method the uncertainty in determining the reaction temperatures arises from unknown rates of thermal transfer from the furnace to the melting materials. The low melting liquids form a so-called batch foam. Before this batch foam is formed, evolving gases can easily leave the batch. During this phase of melting, the state of the batch is governed by the mass flow of the escaping gases. The batch foam hinders the released gases from leaving the batch (i.e., inside the batch foam an equilibrium of the melt with the released gas composition is established). In the interpretation of such experiments it is commonly
20 Fig. 8: of 8.21 Variation
Figure iron redox state as a function of carbon addition for soda lime silicate melts 22 various amounts of barium sulfate, according to Bassine et al. (1987). containing
Müller-Simon
348
assumed that the mass flow of gas in the beginning is small enough that the equilibria in the batch foam are not influenced. However, it must be assumed that the heating rate and related velocity of batch foam formation also impact the results of melting experiments. Because of the excess of CO2 and coke, the CO/CO2 ratio within the foam is defined by the Boudouard equilibrium. The redox state of the melt therefore strongly depends on the amount of CO which escapes before a sufficiently closed foam is formed. As to this point there may be considerable differences between laboratory melts and industrial glass melting furnaces (Conradt et al. 1994).
MONITORING OF THE REACTION PARAMETERS Equations (15) and (16) demonstrate that the sulfur solubility depends on the oxygen partial pressure, the activity of free oxygen ions as determined by the melt compositions, and the partial pressure of the gaseous sulfur compounds. For a complete description of the state of sulfur dissolution under laboratory conditions, as well as in industry, these parameters must be quantified together with the sulfur concentration in the glass. Under both sufficiently oxidizing and reducing conditions the concentrations of SO42− and of S2− in the melt are represented to a good approximation by the melt’s total sulfur concentration. Since sulfur species analysis is usually much more expensive, especially in industry, the total sulfur content is used for control purposes instead of the species concentration. For the determination of residual sulfur concentrations in glasses several physical methods such as X-ray fluorescence analysis (XFA) or wet-chemical methods such as inductively coupled plasma optical emission spectrometry (ICP/OES) in a digestion solution are available (Meckel 2000). Recently, a new device for sulfur monitoring based on an amperometric sensor has become available. This device allows the continuous measurement of the sulfur concentration under industrial glass production conditions in-situ (Müller-Simon and Bauer 2007). The influence of melt composition on the sulfur solubility is described by the characteristic basicity numbers (Duffy and Ingram 1976; Krämer 1991). Another method to evaluate the influence of composition is the determination of the sodium activity (Beerkens 2003). However, in mass glass production aO2− may be assumed constant in Equation (15) and (16). The pSO2 cannot be measured directly. As discussed above, under oxidizing conditions in gas bubbles, pO2 + pSO2 = 1 bar is valid, i.e., pSO2 = 1 bar-pO2. The decomposition of SO3 yields pSO2 = 0.66 bar. Under slightly reducing conditions, pSO2 ≈ 1 bar is valid, which applies for nearly all industrially melted glasses (Chopinet et al. 1982). The most important parameter is the oxidation state of the glass, which cannot be fixed by estimations or approximations. In order to give a suitable description of the oxidation state, different approaches have been presented which shall be discussed in the following.
Oxidation state of iron Iron is present in nearly all glasses as coloring agent or as contamination (cf. Falcone et al. 2011, this volume). If iron is the only polyvalent element in the glass melt, the Fe2+/Fe3+ ratio depends only on oxygen partial pressure pO2 for which the simple relation in Equation (26) applies: pO2
0 0 [ Fe3 + ] − ∆H Fe + ∆SFe = ex p 2+ RT [ Fe ]
4
(26)
With this equation, the Fe2+/Fe3+-ratio can be used to calculate the oxygen partial pressure at melting conditions. It is important to note that this relation is only valid if no physically dissolved oxygen or second polyvalent element is present in the melt as these elements react
Fining of Glass Melts
349
with the iron during the cooling process. However, the presence of more than one polyvalent element in concentrations of several 1000 ppm by weight is the normal case in industrial glass melting; these polyvalent elements can be coloring agents such as chromium in green glass or decoloring agents such as cerium, selenium or manganese in flint glass. If the temperature changes, between pairs of polyvalent elements electron exchange reactions take place according to (Lahiri et al. 1974; Müller-Simon 2007) 3 Fe 2 + ( m) + Cr 6 + ( m) 3 Fe3 + ( m) + Cr 3 + ( m)
(27)
These reactions can take place down to temperatures in the range of the glass transition by direct electron transfer and will freeze in, if the distance between the reaction partners becomes too large. In the case of chromium, Fe2+ is oxidized during cooling until all the Cr6+ is consumed (Dusdorf and Müller-Simon 1997). Sulfur is, of course, also a polyvalent element present at significant concentrations. The effect of sulfur on the Fe2+/Fe3+-ratios calculated with Equation (26) is manifested in a manner that depends on the oxidation state of the melt. In a closed system upon cooling, oxidized melts will have Fe2+/Fe3+-ratios larger than those predicted by Equation (26), whereas reduced melts will have Fe2+/Fe3+-ratios less than those predicted by Equation (26) (Müller-Simon 1994). As such, accurate calculation of the oxygen partial pressure from Equation (26) is only possible if the iron concentration is much larger than the concentration of other polyvalent elements. This condition is satisfied in typical green and amber glasses; however, in clear flint glass with 500 ppm Fe2O3 or lower, the Fe2+/Fe3+-ratio can be shifted by a factor of 2 or greater due to the reaction with sulfur (Müller-Simon 1994).
Oxygen sensors As the sulfur solubility depends on the redox state of the glass melt, it is reasonable to measure the oxygen partial pressure in-situ during production. Such measurements are possible by means of electrochemical oxygen sensors (Frey et al. 1980; Müller-Simon and Mergler 1988; Baucke et al. 1996; Laimböck and Beerkens 2001). In addition to oxygen partial pressure, it is also possible to measure the sulfur concentration continuously in the glass melt. Figure 9 shows the variation of the oxygen partial pressure and the sulfur concentration over a period of a few
sulfur concentration in wt.% S
-2,75
0,02
-3
-3,25
log oxygen partial pressure (in bar)
-2,5
0,03
-3,5
0,01 0
10
20
30
40
time in d 23 24
SWV
Fig. 9
XFA
pO2
Figure 9. Evolution of oxygen partial pressure and sulfur concentration during industrial green glass 25 production (Müller-Simon 1998b). Sulfur concentrations were measured by SWV (square wave voltammetry) and XFA (X-ray fluorescence analysis).
Müller-Simon
350
days in a green glass melt as measured with electrochemical sensors. For comparison the figure also shows the results of X-ray fluorescence analysis (XFA) of sulfur in the final glass products. As expected, the concentration of dissolved sulfur increases with increasing oxygen partial pressure. Measurements of both oxygen partial pressure and sulfur concentration are nowadays very reliable. The great advantage of these measurements is that continuous availability of data is possible without sample preparation. This method is also available as an offline measurement for raw material, especially for cullet examination (Beerkens et al. 1997). In this case melting experiments are conducted in a crucible during which oxygen partial pressure measurements are carried out by means of an electrochemical oxygen sensor. The measurement of the oxygen partial pressure provides a strongly temperature dependent sensor signal which adjusts to the redox equilibria of all present polyvalent elements. The temperature dependence of this signal is governed by the combination of all of the polyvalent elements in the melt. The concentration of physically dissolved oxygen is very small in glass melts (Lenhart and Schaeffer 1985; Kohl and Schaeffer 1987) so, only the oxygen which is chemically bonded to polyvalent ions is of interest in dealing with redox reactions. In the case of an industrial soda-lime-silica glass melt, the total chemically bonded oxygen can be written as: 1 1 4 6 8 [O2 ]chem = [ Fe3 + ] + [Cr 3 + ] + [Cr 6 + ] + [S4 + ] + [S6 + ] 4 4 4 4 4
(28)
where the species concentrations are multiplied by the electron transition to the most reduced state in the glass melt with respect to the ionization of O2. This quantity remains constant in the case of temperature variations which are fast compared to the time required for macroscopic oxygen diffusion. The total chemically bonded oxygen can be determined by species analysis in the final glass product. In contrast to oxygen diffusion, electron exchange reactions between pairs of polyvalent elements are possible down to temperatures in the range of the glass transition temperature. Since the equilibrium constants of the redox reactions of iron, chromium and sulfur are well known (Müller-Simon 2007) the redox ratio of every involved polyvalent element may be calculated at any temperature based on Equation (28). Especially the temperature variations of the Fe2+/Fe3+-ratio, which are inevitable under industrial conditions, may be corrected.
Redox number concepts The measuring methods considered above describe the state at the end of the process. This is often considered disadvantageous because the associated control loop, i.e., the time between a variation of the batch composition and the detection of the effect in the glass properties, is very long. Assuming that the redox state is mainly given by the batch composition, a characteristic number could already provide this information at the beginning of the process. For this purpose several redox number concepts have previously been developed. All these concepts are based on the carbon number concept of Manring and Hopkins, which was developed in the 1950s (Manring and Hopkins 1958). This concept is based on the simple assumption that sodium sulfate and carbon react according to the equilibria in Equation (24). It is assumed that the redox state of the batch is neutral if reaction (24) runs stoichiometrically. Under this condition, one mole of carbon will be neutralized by exactly two moles of sodium sulfate. Taking into account the molecular weights of carbon (12 g/mol) and sodium sulfate (142 g/mol), one weight unit of carbon is neutralized by 23.6 weight units of sodium sulfate. For practical purposes Manring and Hopkins (1958) introduced relative factors by which each component of the batch can be weighted with respect to its redox potential. Reducing raw materials like coke and blast furnace slags are assigned negative factors, whereas oxidizing agents such as sodium sulfate are assigned positive factors. In order to calculate the redox state
Fining of Glass Melts
351
of the batch, the portion of every batch component is multiplied with its respective redox factor and these figures are summed up over all batch components. The sum is related to a batch with 2000 kg of sand. A more detailed description of the calculation procedure and a list of coefficients is given in Falcone et al. (2011, this volume). The first tests of this approach showed that some theoretical factors did not agree well with the observed influence of the respective components. Especially those for coke, which is in practice, far less reducing than expected from the theoretical factor of the carbon number. This fact first appeared when the redox state of a melt was intended to be constant during a change from sulfate/coke to blast furnace slag addition (Simpson and Myers 1978). Simpson and Myers subsequently modified the factors of Manring and Hopkins (1958) by lowering the factor of coke from −16 to −6.7, while the factors for the oxidizing components remained constant. The redox number is still in use in many glass factories for the estimation of the redox state of the batch. One difficulty of this concept is the introduction of recycled cullet. Although the mean organic contamination can be estimated (Nix and Williams 1990), the control of short time variations requires additional effort. An alternative concept has been developed in which the reducing effect of the raw material is experimentally determined (Manring and Davis 1978; Manring and Diken 1980). The effective portion of carbon required by Equation (24) is determined by measurement of the chemical oxygen demand (COD). The equivalent amount of carbon is introduced in the calculation of the redox number with the theoretical efficiency of carbon according to Equation (24). This COD redox number concept is more laborious, but has the advantage that the redox state of the batch is experimentally verified. More accurate correlations can be determined using the COD redox number, if the reducing or oxidizing effect of raw materials with varying redox state such as filter dust or recycled cullet has to be considered.
Interdependence of redox related measurements The redox state of the glass melt is determined by the batch bulk composition, thus, there should be a strong relation between the oxygen partial pressure and the redox number of the batch. This can be found, as shown in Figure 10, for several industrially produced glasses. In most cases the COD of the cullet is estimated and the oxidizing power of filter dust is derived from its SO3 concentration.
0
log pO2 ( pO2 in bar)
-2 -4 -6 -8 -10
green
amber
flint
-12 -90
-50
-10
30
COD redox number 26
Figure 10. Correlation between batch redox number and oxygen partial pressure in the melt for different 27 10: glass meltsFig. (Müller-Simon 1999) [Used by permission of the Deutsche Glastechnische Gesellschaft, from Müller-Simon (1999) HVG-Fortbildungskurs 1999, Fig. 14].
Müller-Simon
352
MODELS OF INDUSTRIAL SULFUR FINING Depending on the particular requirements, different approaches have been worked out to describe the action of sulfur in the course of the melting process. In principle they reflect the view of the laboratory experiments, i.e., these concepts are based on the equilibria of polyvalent elements with oxygen in the glass melt or on the batch reaction. The quality of sulfur fining varies in a distinct manner over the entire redox range (Williams 1980). Under oxidizing conditions good fining quality may be achieved if the fining temperature is high enough. Slightly reducing conditions improve the fining action and a good quality product is already achieved at lower temperatures. At redox conditions corresponding to the sulfur solubility minimum, fining action becomes largely ineffective. This behavior predominantly affects the production of green glasses (Beutinger 1995). Amber glasses, which are produced under very reducing conditions, possess good fining quality; this behavior seems to be influenced by the iron concentration (Harding and Ryder 1970). However, if the oxygen partial pressure becomes too low, the amber color vanishes and no fining action is observed (MüllerSimon 1997).
Solubility concept The simplest concept of sulfur fining relies on the variation of the sulfur solubility as a function of oxygen partial pressure (see Figs. 4, 5 and 6). This concept assumes that the fining process is driven by the decrease in the solubility of sulfur compounds. In other words, the fining intensity increases in direction of the solubility minimum. The increased release of fining gas associated with this approach explains the improved fining behavior observed in slightly reducing flint glass melts. At constant additions of fining agents in the batch and more reducing conditions, less sulfur remains dissolved in the glass melt. According to this concept, the sulfur evolved as a result of the difference between the sulfur concentrations in the initial and the final melt acts as fining gas. However, according to reactions (18) to (25), this portion of SO2 is produced at a very early stage of the melting process therefore making no contribution to the fining process. This concept does not account for the basic observation that fining is temperature dependent. Furthermore, in contrast to the expected behavior glass melts produced near the sulfur solubility minimum tend to have poorer fining quality (Beutinger 1995). These considerations indicate that the solubility concept overestimates the amount of SO2 that contributes to fining.
Equilibrium concepts The basic problem with the concept discussed above, is that the role of temperature is not sufficiently considered. The role of temperature in fining was introduced in a concept proposed by Chopinet et al. (1983) and Chopinet and Barton (1986). This concept is based on the consideration that at the end of an effective fining process, the composition of bubbles should consist only of SO2 and O2, and that if such a melt is heated, the first bubbles occur at the temperature where pSO2 + pO2 = 1 bar. This temperature is called sulfur saturation temperature. Under this presumption Chopinet and Barton (1986) combined the decomposition reaction of sulfate (Eqn. 11) with the redox equilibrium of ferrous and ferric iron in silicate melts. 1 2 Fe3 + ( m) + O2 − ( m) 2 Fe 2 + ( m) + O2 ( g ) 2
(29)
Combining Equation (11) with Equation (29) yields SO24 − ( m) + 2 Fe 2 + ( m) SO2 ( g ) + 2 Fe3 + ( m) + 2 O2 − ( m)
Using the equilibrium constants for reaction (11)
(30)
Fining of Glass Melts K1 =
353
[SO24 − ] 1 pSO2 ⋅ pO22
(31)
and reaction (29) 2
[ Fe3 + ] 1 K2 = 1 2+ 2 [ Fe ] pO2
(32)
the equilibrium constant of reaction (30) can be expressed as 2
K′ =
pSO2 [ Fe3 + ] K = 2 2− 2+ K1 [SO 4 ] [ Fe ]
(33)
Chopinet and Barton (1986) obtained the thermochemical data of this reaction in fiber glass melts, studied the residual sulfur contents in samples from two fiber glass melting tanks, and plotted the sulfur contents against the melt’s redox state given by the square of the iron redox ratio. The plot of measured data and calculated data at two temperatures are shown in Figure 11. Although they found that the redox state varied over a wide range, the sulfur concentrations plotted according to Equation (33) lie close to the curves of constant temperature in Figure 11; in other words the temperature of sulfur release was primarily a property of the glass melting tank. This makes it easy to estimate the total process related sulfur emissions from the batch reactions and fining.
Dynamic equilibrium concepts The concepts which have been described above can only be applied to oxidizing melting conditions. For reducing conditions where reaction (12) takes place, no such exist. The oxygen
0,140
T = 1270 °C
sulfur concentration in wt% S
0,120
T = 1320 °C 0,100
0,080
0,060
0,040
0,020
0,000
0
1
2
3
4
5
6
(Fe2+/Fe3+)2 28
Figure 11.29Sulfur concentration as a function of redox state of iron in the glass. Solid lines are fits according Fig. 11: to Equation (33). Data are from Chopinet et al. (1986). 30
Müller-Simon
354
balance calculation allows for the calculation of SO2 in oxidized as well as in reduced glass melts as a function of temperature by means of Equation (28) (Müller-Simon 1994, 1997, 1998a). Figure 12 shows the temperature dependence of the calculated S4+ concentration (cf. Eqn 28) in two green container glasses which have been melted under different melting conditions. Because of the negligible small SO32− solubility it may be assumed that this S4+ immediately forms gaseous SO2 according to S4 + ( m) + 2O2 − ( m) SO2 ( g )
(34)
SO2 containing bubbles can only be dissolved if the S4+-concentration in a glass melt falls below the solubility limit at a given temperature. Forming of glasses is carried out at decreasing temperature between 1200 °C and 1000 °C. In this temperature range the dissolution of small bubbles requires typically a few minutes, which is longer than most forming processes. Especially in container glass production, bubbles will not dissolve after feeding the glass into the forming molds, i.e., in order to produce a bubble free glass the S4+ concentration must fall below a certain limit when the forming starts. According to Figure 12 this limit lies at about 0.02 wt% S, which is above typical sulfur solubilities in soda-lime-silicate melts (cf. Backnaes and Deubener 2011, this volume). However, this is not a chemical value but a quality criterion. A satisfactory green glass quality may still contain a small residual seediness. The most important advantage of this fining model is the possibility of describing fining under reducing conditions. One could assume that with increasing temperature S2 may be released, but there is no reason why the chemical solubility of S2− should decrease. On the other hand it is known that the fining action of sulfur under reducing conditions depends on a sufficient concentration of iron (Harding and Ryder 1970). It is also known that sulfur and iron interact in glass melts (Müller-Simon 1994). Thus, a fining reaction 6 Fe3 + ( m) + S2 − ( m) T → 6 Fe 2 + ( m) + S4 + ( m)
(35)
in combination with Equation (34) is most probable in iron-bearing glass melts under reducing conditions. The validity of reaction (35) is proved by the vanish of the amber color with 31
gob temperature
S4+ concentration in wt.% S
0,02
0,01
0,00 0
400
800
temperature in °C
1200
1600
with bubbles without bubbles 32 33 12. Calculated Fig. 12: concentrations of tetravalent sulfur as a function of temperature for two green Figure glasses 34 with different quality [Used by permission of Society of Glass Technology, from Müller-Simon
(1998a) Glastech. Ber. Glass Sci. Technol. 71, Fig. 8].
Fining of Glass Melts
355
increasing temperature (Müller et al. 1999). This reaction does not only describe the fining action, but it also explains why fining of over-reduced glass melts does not work (MüllerSimon 1997).
INVESTIGATIONS UNDER INDUSTRIAL CONDITIONS Over several years the Research Association of the German Glass Industry (HVG) has performed many investigations in industrial mass glass production, including the measurement of sulfur concentrations in gaseous emissions and glass melts. Data on sulfur mass flows are available for batch, fuel, waste gas, and glass. Additionally, redox data for the batch, the furnace atmosphere and the glass are even available for some cases. According to the considerations enumerated above, sulfur mass flows may be expected from batch reactions as well as from fining. In fossil fuel fired furnaces, the heat is produced by flames and mainly supplied to the glass melt via radiation. The most common type among these is the regenerative furnace in which the combustion air is pre-heated in regenerators. These consist of a ceramic checker work where in turn hot exhaust gas and cold air are passing through. For this purpose the furnace is heated from two sides alternately. While the exhaust gas is passing through the regenerator, the checker work is heated up. After the reversal of the firing side, this heat is transferred to the cold combustion air. Regenerative furnaces are used in the flat and the special glass fields as well as in the container glass production. They feature a relatively low specific heat demand but necessitate relatively high investment for the regenerators. If the ports where the fuel and the pre-heated air enter the furnace and the exhaust gas leaves the furnace are vis-á-vis one speaks of cross-fired regenerative furnaces, if they are side by side of end-fired furnaces. Another type are recuperative glass melting furnaces where the combustion air and the exhaust gas stream through separated cavities in steel or ceramic recuperators, thus, effecting counterflow heat exchange (Trier 1984). Figure 13 shows the relevant sulfur mass flows in an industrial glass melting furnace. If the balance boundaries are taken without the heat regenerators, sulfur is introduced into the process by the raw materials sbatch, the fuel sfuel, and a portion which is deposited in the regenerator checker work from the waste gas and which is released into the combustion air.
combustion
furnace atmosphere
waste gas
sfuel swaste
sair srelease sbatchemission sbatch batch
sfining smelt glass melt
sglass glass
35 13.13: Sketch of the sulfur mass flows in industrial glass melting tanks [Used by permission of German 36 FigureFig. 37
Society of Glass Technology, from Müller-Simon and Gitzhofer (2008) Glass Technol. Eur. J. Glass Sci. Technol. A 49, Fig. 1].
356
Müller-Simon
Sulfur leaves the process with the waste gas, swaste, and the produced glass, sglass. From Figure 13 balances for sulfur mass flows can be derived in industrial glass melting furnaces. sinput = sbatch + s fuel + sair
(36)
soutput = sglass + swaste
(37)
sbatch + s fuel + sair = sglass + swaste
(38)
and
The balance then yields
The compounds which are used as fining agents are sodium sulfate, blast furnace slags and filter dust. The amount of sulfur in the cullet depends on the type of cullet: amber, green, flint or mixed. The amount which is introduced with the fuel depends on the fuel quality. According to current EU regulations, heavy fuel oil can contain up to 1% S while light fuel contains up to 0.1% S. In the following section we will try to identify the reactions of sulfur release which have been compiled above with the measurable mass flows in glass melting furnaces. The total sulfur which is released from the melting process is srelease = swaste − s fuel − sair = sbatch − sglass
(39)
According to reactions (23) to (25) a sulfur mass flow from the batch into the furnace atmosphere, sbatch emissions, exists and according to the temperature dependence of reactions (11) (and 20, resp.) and (12) (and 35, resp.) a sulfur mass flow, sfining, from the melt into the furnace atmosphere. In cross-fired furnaces it is possible to measure sulfur mass flows through every port, sport, along the furnace axis, which makes estimation of how much sulfur is lost from the batch and how much from the fining area possible (Müller-Simon and Gitzhofer 2008). srelease = ∑ s port
( 40)
port
Sulfur enters the furnace chamber with the fuel and the combustion air and leaves it with the waste gas. The sulfur mass flow released with the waste gas is always larger than that which is introduced with the fuel and the combustion air. The difference between discharged sulfur and that introduced into the furnace is equal to the mass flow from the batch and the glass melt into the atmosphere. The results in Figure 14 from investigations in container glass furnaces are a demonstration of this balance. It may be assumed that the batch reaction takes place in the region of ports 1 and 2 and the fining reaction in the region of ports 3 and 4. For all melt compositions about 2/3 of the sulfur emissions come from the batch area and only 1/3 from the refining area. According to Figure 14, the sulfur release is mostly affected by the color of the produced glass, i.e., the redox state (flint glass is produced at pO2 = 10−1-10−2 bar, green at 10−3-10−4 bar and amber at 10−5-10−10 bar). Green and amber glass batches release considerably more sulfur compared to flint glass, although the sulfur addition in flint glass batches is larger compared to colored glasses. This is true for the batch reactions as well as for the fining. However, most of the sulfur is released by means of reactions (23) to (25), and smaller amounts by the fining reactions. This behavior is also in agreement with Figure 7. The batch reactions are also influenced by the furnace atmosphere. The two green glasses are melted under completely different combustion conditions. Oxygen excess in the atmosphere causes lower sulfur release in the region of the first port, where as an oxygen deficit causes increased sulfur release. The reason is that in Equation (21) not only the CO which is formed according to the Boudouard equilibrium, but also CO which has infiltrated from the furnace
Fining of Glass Melts
357
atmosphere becomes active in the reactions. Figure 15 shows the sulfur release into the flue gas, swaste, as a function of the redox state of the combustion. All the sulfur which is introduced with the batch and that is not dissolved in the glass melt must be released during glass melting: sbatch − sglass = srelease
( 41)
Figure 16 shows the total sulfur release of the batch and the glass melt as a function of the sulfur addition, with the batch less the sulfur mass flow with the produced glass. Both quantities are related to the tonnage of produced glass. The sulfur content of the fuel has an insignificant influence on the balance in Equation (41) (Müller-Simon and Gitzhofer 2008). In the case of flint and amber glass, the sulfur emissions clearly increase with increasing sulfur additions. The straight line in Figure 16 shows the best fit line which has a slope of one. This slope means that all the excess sulfur which is not dissolved in the melt is indeed emitted.
sulfur emissions in kg S/t glass
0,6
0,4
0,2
0,0
1
2 flint
38 39
3
port green
4 amber
Fig. 14:
Figure 14. Sulfur release out of the batch and the glass melt into the furnace atmosphere along the axis of 40 cross-fired furnaces for flint, green and amber container glass production [Used by permission of Society 41 Technology, from Müller-Simon and Gitzhofer (2008) Glass Technol. Eur. J. Glass Sci. Technol. of Glass A 49,42Fig. 3]. 2000
SO2 in flue gas in mg/m3
1500
1000
500
0 11
43
11,5
12
12,5
13
13,5
oxygen in flue gas in vol.%
44 15. Influence Fig. 15 of the oxygen content in the furnace atmosphere on sulfur release [Used by permission Figure of Society of Glass Technology, from Müller-Simon and Gitzhofer (2008) Glass Technol. Eur. J. Glass Sci. 45 Technol. A 49, Fig. 6].
Müller-Simon
358 1,8 1,6
flint
green
amber
srelease in kg/t glass
1,4 1,2 1 0,8 0,6 0,4 0,2 0 0
0,2
0,4
0,6
0,8
1
1,2
1,4
1,6
1,8
sbatch -sglass in kg/t glass
46 47 Fig Released 16. Figure 16. sulfur srelease as a function of sulfur mass flow with the batch less the sulfur mass flow
in the produced glass sbatch-sglass related to the load. Different filling of symbols refer to different redox state of the furnace atmosphere [Used by permission of Society of Glass Technology, from Müller-Simon and Gitzhofer (2008) Glass Technol. Eur. J. Glass Sci. Technol. A 49, Fig. 8].
The deviations from this behavior are probably caused by perturbations from the steady state. Figure 16 also contains an example of flint glass production under various furnace atmosphere oxidation states. This example demonstrates the effect of non-steady state behavior. Although the sulfur release may be caused by locally different mechanisms, in a steady state all the excess sulfur will be released. This means that the excess sulfur not released from the batch by the batch reaction or interaction with the furnace atmosphere is finally released during fining. This is a very important statement with respect to the environmental impact of a given batch composition and glass melting furnace. However, this does not implicitly result in an improvement of the fining action. In the case of a minor gas release with increasing temperature seeds or bubbles may result.
SUMMARY Fining is the most important step in industrial glass production. In this step the glass melt is cleared of bubbles and homogenized. Thus, fining is crucial for the quality of the final glass product. Agents added with the raw material batch release gas with increasing temperature enlarging gaseous inclusions to large bubbles. These bubbles develop enough buoyancy to reach the surface of the glass melt and escape in sufficiently short time. Special glasses are fined by means of oxygen fining mostly with16antimony oxide or sodium chloride fining. Mass glasses like container, flat and fiber glass, which account for the largest portion of produced tonnage, are fined with sulfur compounds. It can be shown that most of the sulfur is released in the course of the batch reactions due to the reaction with reducing components. An additional impact is caused by the interaction of the batch with the furnace atmosphere at this stage. Only one third of the released sulfur is caused by fining. Sulfur fining under oxidizing condition is due to the thermal decomposition of the sulfate anion. This decomposition can be improved at slightly reducing conditions or by addition of polyvalent elements for instance iron in green glass. Under very reducing conditions sulfur fining is achieved through oxidation of the sulfide anion through reaction with iron.
Fining of Glass Melts
359
Sulfur oxide emissions from industrial glass melting tanks are limited due to legal restriction. The knowledge of the basic reactions which are responsible for these emissions is required to optimize the batch composition and the melting process. However, the chemical interactions in the real glass melting process are complex. Thus, future application of this knowledge will increasingly rest upon model based control (Chmelar et al. 2000; Beerkens 2002).
ACKNOWLEDGMENT The author wishes to thank Don Baker and Mathi Rongen for reviewing this chapter. Further thanks go to Harald Behrens and Jim Webster for their constructive comments. All of them greatly assisted in getting this chapter across to geologists. Finally the author thanks Klaudia Jaenicke and Aaron Bell for proof-reading.
REFERENCES Backnaes L, Deubener J (2011) Experimental studies on sulfur solubility in silicate melts at near-atmospheric pressure. Rev Mineral Geochem 73:143-165 Backnaes L, Stelling J, Behrens H et al. (2008) Dissolution mechanism of tetravalent sulphur in silicate melts: evidence from sulphur K edge XANES studies on glass. J Am Ceram Soc 91:721-727 Bassine JF, Mestdagh MM, Rouxhet PG (1987) Redox buffering by sulphate and carbonate during the melting of reduced soda-lime-silica glasses. Glass Technol 28:50-56 Baucke FGK, Werner RD, Müller-Simon H, Mergler K (1996) Application of oxygen sensors in industrial glass melting tanks. Glastech Ber Glass Sci Technol 69: 57-63 Beerkens RGC (1995) The role of gases in glass melting processes. Glastech Ber Glass Sci Technol 68:369-380 Beerkens RGC (2002) Modelling of the melting process in industrial glass furnaces. In: Mathematical simulation in glass technology. Schott series on glass and ceramics, science, technology and applications. Krause D, Loch H (eds) Springer Verlag, p 17-72 Beerkens RGC (2003) Sulfate decomposition and sodium oxide activity in soda-lime-silica glass melts. J Am Ceram Soc 86:1893-1899 Beerkens RGC (2005) Sulphate decomposition and sulphur chemistry in glass melting processes. Glass Technol: Eur J Glass Sci Technol 46:39-46 Beerkens RGC (2007) Sulphur chemistry and sulphate fining and foaming of glass melts. Glass Technol: Eur J Glass Sci Technol A 48:41-52 Beerkens RGC, Faber A, Plessers J, Thontat T (1997) Methods for the characterization of redox state of cullet and batches in glass manufacturing. Glass 74:371-375 Beutinger M (1995) Einsatz von Recyclingglas in der Hohlglasschmelze. Glastech Ber Glass Sci Technol 68:N51-N58 Brückner R (1962) Zur Kinetik des Stoffaustausches an den Grenzflächen zwischen Silikatglas- und Salzschmelzen und des Stofftransportes in Silikatglasschmelzen unter besonderer Berücksichtigung des Verhaltens von Na2SO4 und seiner Zersetzungsprodukte. Teil III. Glastechn Ber 35:93-105 Cable M (1960) A study of refining. Part 1. Measurements of the refining of soda-lime-silica glass with and without refining agents. Glass Technol 1:144-154 Cable M (1961a) A study of refining. Part 2. Mechanism of refining. Glass Technol 2:60-70 Cable M (1961b) A study of refining. Part 3. The function of arsenious oxide during melting and refining. Glass Technol 2:151-158 Chmelar J, Bodi R, Muijsenberg E (2000) Supervisory advanced control of glass melters and forehearth by GS expert system. Glastech Ber Glass Sci Technol 73:276-284 Chopinet MH, Barton JL (1986) The Effect of melting temperature on the residual sulfate content of glass. Proc XIV. Int Congr on Glass {INCOMPLETE REFERENCE: need page numbers, publisher, editor, etc.} Chopinet MH, Massol JJ, Barton JL (1982) La relation entre la teneur en sulfate et l´état d´oxydation de verres fondus en presence de réducteurs. Rev Staz Sper Vetro 5:200–201 Chopinet MH, Massol JJ, Barton JL (1983) Factors determining the residual sulfate content of glass. Glastechn Ber 56K:596-601 (NOT USED?) Collignon J, Rongen M, Beerkens RGC (2010) Gas release during melting and fining of sulphur containing glasses. Glass Technol: Eur J Glass Sci Technol A 51:123-129 Conradt R, Suwannathada P, Pimkhaokham P (1994) Local temperature distribution and primary melt formation in a melting batch heap. Glastech Ber Glass Sci Technol 67:103-113 Duffy JA (1989) A common optical basicity scale for oxide and fluoride glasses. J Non-Cryst Solids 109:35-39
360
Müller-Simon
Duffy JA, Ingram MD (1976) An interpretation of glass chemistry in terms of the optical basicity concept. J Non-Cryst Solids 21:373-410 Dusdorf W, Müller-Simon H (1997) Investigations into the existence of hexavalent chromium in industrial glasses. Glastech Ber Glass Sci Technol 70:325-332 Falcone R, Ceola S, Daneo A, Maurina S (2011) The role of sulfur compounds in coloring and melting kinetics of industrial glass. Rev Mineral Geochem 73:113-141 Fincham CJB, Richardson FD (1954) The behaviour of sulphur in silicate and aluminate melts. Proc R Soc A223:1152, 40-63 Flick C, Nölle G (1995) Redox conditions during the melting of batch. Glastech Ber Glass Sci Technol 68:8183 Frey T, Schaeffer HA, Baucke F G K (1980) Entwicklung einer Sonde zur Messung des Sauerstoffpartialdrucks in Glasschmelzen. Glastechn Ber 53:116-123 Geotti-Bianchini F, De Riu L (1998) Water content of sulfate-fined industrial soda-lime glass and its influence on workability. Glastech Ber Glass Sci Technol 71: 230-242 Goldman DS (1985) Redox and sulfur solubility in glass melts. In: Gas bubbles in glass. Prepared by Technical Committee 14 of the International Commission on Glass. Institut National du Verre, Charleroi, 74-91 Hanke KP, Scholze H (1970) Einfluss des Wasserdampfes auf die Blasenbildung sulfathaltiger Glasschmelzen. Glastechn Ber 43:475-482 Harding FL, Ryder RJ (1970) Amber colour in commercial silicate glasses. J Can Ceram Soc 39:59-63 Kircher U (1993) Investigations concerning the recycling of filter dust from electrostatic precipitators in the glass industry. Glastech Ber 66:279-283 Kloužek J, Arkosiová M, Nèmec L, Cincibusová P (2007) The role of sulphur compounds in glass melting. Glass Technol:Eur J Glass Sci Technol A 48:176-182 Kohl R, Schaeffer HA (1987) Oxidation states of glass melts. Diffusion and Defect Data 53-54:325-334 Krämer F (1979) Mathematisches Modell der Veränderung von Gasblasen in Glasschmelzen. Glastechn Ber 52:43-50 Krämer F (1991) Contribution to basicity of technical glass melts in relation to redox equilibria and gas solubilities. Glastech Ber 64:71-80 Krämer F (1999) Theorie und Praxis des Läuterns technischer Gläser. In: Grundlagen des industriellen Glasschmelzprozesses. Verlag der Deutschen Glastechnischen Gesellschaft, Frankfurt/M., p 25-42 Krauß M, Lenhart A, Ratka H, Kircher U (1995) Characterization of filter dust in the glass container industry and definition of its influence on green glass melts. Glastech Ber Glass Sci Technol 65:278-284 Lahiri D, Mukherjee B, Majumdar RN, Bihar D (1974) Mechanismus der Wechselwirkung zweier RedoxOxide in Glas. Glastechn Ber 47:4-9 Laimböck P, Beerkens RGC (2001) On-line redox sensors in industrial glass melting tanks, Proc 62nd Conf on Glass Problems, p. 27-44 Lenhart A, Schaeffer HA (1985) Elektrochemische Messung der Sauerstoffaktivität in Glasschmelzen. Glastech Ber 58:139-147 Manring WH, Davis RE (1978) Controlling redox conditions in glass melting. Glass Ind 5:13-30 Manring WH, Diken GM (1980) A Practical approach to evaluating redox phenomena involved in the meltingfining of soda-lime glasses. J Non-Cryst Solid 38-39:813-818 Manring WH, Hopkins RW (1958) Use of sulfates in glass. Glass Ind 39:139-170 Müller M, Rüssel C, Claußen O (1999) UV-VIS spectroscopic investigations of amber glass at high temperatures. Glastech Ber Glass Sci Technol 72:362-366 Müller-Simon H (1994) On the interaction between oxygen, iron and sulfur in industrial glass melts. Glastech Ber Glass Sci Technol 67:297-303 Müller-Simon H (1997) Temperature dependence of the redox state of iron and sulfur in amber glass melt. Glastech Ber Glass Sci Technol 70:389-391 Müller-Simon H (1998a) Oxygen balance in sulfur-containing glass melts. Glastech Ber Glass Sci Technol 71:157-165 Müller-Simon H (1998b) Meßanordung zur Kontrolle der Farbe und des Läuterzustands industrieller Glasschmelzen. Final report of the IGF/AiF research project no 10397 Müller-Simon H (1999) Sulfatläuterung in Kalk-Natron-Silicatgläsern. In: Grundlagen des industriellen Glasschmelzprozesses. Verlag der Deutschen Glastechnischen Gesellschaft, Frankfurt/M, p 45-72 Müller-Simon H (2007) Elektronenaustausch zwischen Paaren polyvalenter Elemente in technischen Gläsern und seine Auswirkungen auf die industrielle Glasherstellung. Verlag der Deutschen Glastechnischen Gesellschaft, Offenbach/Main Müller-Simon H, Bauer J (2007) Elektrochemische Sensoren für Messungen in industriellen Glasschmelzen. Stand der Technik. dgg-journal 6:11-15 Müller-Simon H, Gitzhofer K (2008) Sulphur mass flow balances in industrial glass melting furnaces. Glass Technol Eur J Glass Sci Technol A 49:83-90
Fining of Glass Melts
361
Müller-Simon H, Mergler KW (1988) Electrochemical measurements of oxygen activity of glass melts in glass melting furnaces. Glastech Ber 61:293-299 Mulfinger HO (1980) Gase (Blasen) in der Glasschmelze. In: Glastechnische Fabrikationsfehler. JebsenMarwedel H, Brückner R (ed) Springer, Berlin, p 193-268 Nagashima S, Katsura T (1973) The solubility of sulfur in Na2O⋅SiO2 melts under various oxygen partial pressures at 1100 °C, 1250 °C and 1300 °C. Bull Chem Soc Japan 46:3099-3103 Nix M, Williams HP (1990) Calculation of the redox number of batches containing recycled cullet. 2nd Int Conf Fusion and Processing of glass. Glastech Ber 63K:271-279 Nölle G, Al Hamdan M (1990) Kohlenstoff in Glasrohstoffgemengen. Silikattechnik 41:192-193 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Pajean G, Delhopital G, Doumeng M (1975) Application de la chromatographie en phase gazeuse a l´étude de la decomposition des mélange vitrifiables. Verres Refract 29:231-235 Paul A (1982) Chemistry of Glasses. Chapman and Hall, London and New York Rüssel C (1990) The electrochemical behavior of some polyvalent elements in a soda-lime-silica glass melt. J Non-Cryst Solids 119:303-309 Samadhi TW, Elliott JC, Jones LE, Clare AG (2001) Sodium sulfate decomposition in dry atmospheres. Glastech Ber Glass Sci Technol 74:47-56 Schaeffer H A (1996) Recycling of cullet and filter dust in the German glass industry. Glastech Ber Glass Sci Technol 69:101-106 Schreiber HD, Kozak SJ et al. (1987) Sulfur chemistry in a borosilicate melt. Glastech Ber Glass Sci Technol 60:389-398 Simpson W, Myers DD (1978) The redox number concept and its use by the glass technologist. Glass Technol 19:4, 82-85 Sun K-H (1947) Fundamental condition of glass formation. J Am Ceram Soc 30:277-281 Sun K-H (1948) A scale of acidity and basicity in glass. Glass Ind 29:73-74, 98 Thieme C (1993) Sodium carbonate. In: Ullmann´s Encyclopedia of Industrial Chemistry. Elvers B et al. (eds) VCH Verlagsgesellschaft Weinheim Toop G W, Samis CS (1962) Some new ionic concepts of silicate slags. Trans Met Soc AIME 224:878-887 Trier W (1984) Glass Furnaces: Design, Construction and Operation. Society of Glass Technology, Sheffield. Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Weyl W (1943) The role of sodium sulfate in glass manufacture. Glass Ind 24:17-20, 39 Wilke M, Klimm K, Kohn SC (2011) Spectroscopic studies on sulfur speciation in synthetic and natural glasses. Rev Mineral Geochem 73:41-78 Williams HP (1980) Einfluss des Oxidationszustandes des Gemenges auf die Glasläuterung mit schwefelhaltigen Läutermitteln. Glastechn Ber 53:189-194
13
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 363-421, 2011 Copyright © Mineralogical Society of America
Sulfur Degassing From Volcanoes: Source Conditions, Surveillance, Plume Chemistry and Earth System Impacts Clive Oppenheimer1,2,3, Bruno Scaillet2 and Robert S. Martin4 1
2
Le Studium, Institute for Advanced Studies Orléans and Tours, France
Centre National de la Recherche Scientifique-Institut National des Sciences de l’Univers, Université d’Orléans
Université François Rabelais de Tours, Institut des Sciences de la Terre d’Orléans 1a rue de la Férollerie, Orléans 45071, France 3
Department of Geography, University of Cambridge Downing Place, Cambridge CB2 3EN, United Kingdom 4
School of Biological and Chemical Sciences Queen Mary, University of London, United Kingdom [email protected]
IntRODuCtIOn Despite its relatively minor abundance in magmas (compared with H2O and CO2), sulfur degassing from volcanoes is of tremendous significance. It can exert substantial influence on magmatic evolution (potentially capable of triggering eruptions); represents one of the most convenient opportunities for volcano monitoring and hazard assessment; and can result in major impacts on the atmosphere, climate and terrestrial ecosystems at a range of spatial and temporal scales. The complex behavior of sulfur in magmas owes much to its multiple valence states (−II, 0, IV, VI), speciation (e.g., S2, H2S, SO2, OCS and SO3 in the gas phase; S2−, SO42− and SO32− in the melt; and non-volatile solid phases such as pyrrhotite and anhydrite), and variation in stable isotopic composition (32S, 33S, 34S and 36S; e.g., Métrich and Mandeville 2010). Sulfur chemistry in the atmosphere is similarly rich involving gaseous and condensed phases and invoking complex homogeneous and heterogeneous chemical reactions. Sulfur degassing from volcanoes and geothermal areas is also important since a variety of microorganisms thrive based on the redox chemistry of sulfur: by reducing sulfur, thiosulfate, sulfite and sulfate to H2S, or oxidizing sulfur and H2S to sulfate (e.g., Takano et al. 1997; Amend and Shock 2001; Shock et al. 2010). Understanding volcanic sulfur degassing thus provides vital insights into magmatic, volcanic and hydrothermal processes; the impacts of volcanism on the Earth system; and biogeochemical cycles. Here, we review the causes of variability in sulfur abundance and speciation in different geodynamic contexts; the measurement of sulfur emissions from volcanoes; links between subsurface processes and surface observations; sulfur chemistry in volcanic plumes; and the consequences of sulfur degassing for climate and the environment.
Geodynamics and the geochemical behavior of sulfur The sulfur released by magmas in volcanic emissions may derive from three different sources: dissolved in the silicate liquid, present in a coexisting gas phase at depth, or from the breakdown of sulfur-bearing minerals. Both the amount of sulfur locked in solid compounds 1529-6466/11/0073-0013$10.00
DOI: 10.2138/rmg.2011.73.13
364
Oppenheimer, Scaillet, Martin
(essentially sulfates and sulfides) and that dissolved in silicate melt under pre-eruptive conditions can be accurately measured. The greatest unknown in assessing the budget of sulfur that can potentially be sourced from degassing magmas comes from the presence of an “excess” gas phase at depth, i.e., in the reservoir where the magma resides for a certain time prior to eruption (Shinohara 2008). This possibility introduces the difficulties of establishing the sulfur content of the gas phase in addition to the abundance of gas relative to minerals and silicate melt phases. Neither variable can be assessed by direct observation or geophysical investigations. We rely, therefore, on indirect means to estimate the abundance of gas and its sulfur content at depth. In the following subsections, we first consider the case for subduction zone magmas, concerning which considerable progress has been achieved in recent years, and then review the evidence for hot spot and ocean ridge environments.
Subduction zones Although prior research had already suggested the presence of a gas phase at depth in arc magmas (Rose et al. 1982), the first eruption for which the case for an “excess” gas phase was strongly articulated was that of El Chichón in 1982, thanks to the efforts of the late Jim Luhr. He carried out detailed petrological work on this sulfide- and sulfate-bearing andesitic magma (Luhr et al. 1984; Luhr 1990). Scaling estimates of pre-eruptive dissolved sulfur (measured in crystal-hosted melt inclusions) by the eruption magnitude suggested a sulfur release as much as two orders of magnitude lower than the SO2 measured by satellite remote sensing techniques (see later section). This great discrepancy between observations and petrologic estimates led Luhr et al. (1984) to suggest that the “excess” sulfur (i.e., that missing from the petrological calculation) was most probably stored in a coexisting gas phase in the reservoir. This approach revealed a similar picture for the next major sulfur-rich volcanic cloud, that released by the 1991 Plinian eruption of Mt. Pinatubo in the Philippines. Here again, comparison between remote sensing measurements and petrological calculations led Westrich and Gerlach (1992) to conclude the existence of an “excess” gas phase in which most of the sulfur was stored prior to eruption. An alternative scenario for sulfur release during the Pinatubo eruption involving anhydrite breakdown during decompression was proposed (Rutherford and Devine 1996), but such a mechanism is not supported on kinetic grounds (Gerlach et al. 1996). The evidence for excess sulfur stored in a gas phase has been reported since for a number of other active volcanoes, including Mt. St. Helens (USA, Gerlach and McGee 1994), Redoubt (USA, Gerlach et al. 1994), Nevado del Ruíz (Colombia, Williams et al. 1986; Sigurdsson et al. 1990), several Chilean volcanoes (Andres et al. 1991; Matthews et al. 1999) and Anatahan (Mariana Islands) during its 2003 eruption (de Moor et al. 2005). Thanks largely to satellite remote sensing estimates, there is, therefore, compelling evidence supporting the concept that arc magmas, in particular those of evolved composition, can hold a significant part of their sulfur budget in a gas phase at depth. The observed sulfur budgets point to a gas phase amounting to a few wt% of the magma, with a sulfur content (of the gas) of up to a few wt%. Another method used to retrieve gas abundance in magma reservoirs is based on the geochemical behavior of trace elements that are prevalent either in the silicate melt or in the gas phases (Wallace and Edmonds, 2011 this volume). If the partition coefficients of those elements are known, then it is possible to calculate the amount of a gas phase in the reservoir given a series of melt inclusions related to each other by a gas-melt fractionation process. Based on this assumption, it is possible to evaluate the amount of gas present in the various parcels of magma sampled by the melt inclusions. Such an approach has been applied successfully to the Bishop Tuff eruption (Wallace et al. 1995, 1999) and also to the 1991 Pinatubo eruption (Wallace and Gerlach 1994). Of particular importance is the latter case because the calculated abundance of gas in the reservoir (up to 5 wt%) corresponds to a sulfur yield similar to that measured independently by satellite remote sensing (around 9 Tg S).
Sulfur Degassing From Volcanoes
365
Consideration of percolation theory led Wallace (2001) to conclude that this value might represent a fluid mechanical threshold. In other words, a magma reservoir cannot sustain more than 5 wt% of gas, beyond which all excess gas arising from further crystallization or supplied from deeper levels to the upper regions of the reservoir, is lost via percolation of gas through a permeable bubble network to the top of the reservoir and into the hydrothermal system. The sulfur content of gas in a magma reservoir can be also estimated from thermodynamic calculations. Using standard concepts of homogeneous equilibria in the C-O-H-S system, as pioneered by Holloway (1977, 1987), it is possible to calculate the abundance and nature of volatile species present in the gas phase. Given the potentially wide range of application of such an approach, its basic principles are summarized here. The gas phase is modeled in the C-O-H-S system, which typically accounts for over 95 mol% of the bulk composition of volcanic gases. Halogens are next in abundance (Aiuppa et al. 2009) but though they potentially affect sulfur solubility in relatively oxidized silicate melts (Botcharnikov et al. 2004; Webster et al. 2009), their role in affecting sulfur behavior in magmatic gases is neglected here for simplicity. Species typically considered are H2O, H2, CO2, CO, CH4, S2, SO2, H2S and O2. The five coupled equilibria that govern the abundances of these species in the gas are: H 2= O CO +
H2 +
1 O2 2
1 O2 = CO2 2
CH 4 + 2O2=
2H 2O + CO2
(1) (2) (3)
1 S2 + O2 = SO2 2
(4)
1 S2 + H 2 = H 2S 2
(5)
with the corresponding equilibrium constants: K1 =
(f
1/ 2 O2
fH 2 O fCO2
K2 =
K3
(f (f = (f
K4 = K5 =
)
× fH 2
1/ 2 O2
) ) )
(7)
)
(9)
)
(10)
× fCO
CO2
× fH 2 O 2
CH 4
× fO2
(f (f
2
fSO2 O2
× fS2 1/ 2 fH 2 S
H2
× fS2 1/ 2
(6)
(8)
where fi is the fugacity of species i. In such a system at any fixed pressure (P) and temperature (T), the knowledge of three additional intensive parameters constrains species' proportions (Holloway 1987). The equilibrium constants Ki can be computed from standard thermodynamic
366
Oppenheimer, Scaillet, Martin
databases (Symonds and Reed 1993). The additional relationships needed are: f= Xi γ i P i
(11)
∑X
(12)
i
=1
where Xi is the mole fraction of species I, and γi the fugacity coefficient (a correction for nonideality; γi = 1 for an ideal gas) at pressure P and temperature T. Several equations of state (EOS) enable calculation of the fugacity coefficient of the main species of interest (see, for instance, Ferry and Baumgartner 1987) but, in the context of the present review, we focus on low-pressure conditions (<300 MPa) where most magmatic reservoirs reside, and for which there is thus little difference between the various EOS outcomes. In the following, use will be made of the classic MRK EOS introduced to geologists by John Holloway in 1977. The most straightforward way to apply the above approach is to rely on information given by melt inclusions regarding pre-eruptive dissolved volatiles. Conventional micro-analytical tools (FTIR, SIMS, EMPA) permit accurate determination of the most common volatile species dissolved in silicate glasses, notably the bulk contents of H2O, CO2 and S (Ripley et al. 2011, this volume). Once these quantities are known, it suffices to have appropriate thermodynamic models of volatile solubilities (e.g., Dixon et al. 1995; Zhang 1999; Moretti et al. 2003; Behrens et al. 2004; Clemente et al. 2004) to shift from concentrations to corresponding fugacities (i.e., fH2O, fCO2 or fS2), which are the input parameters needed to solve the above set of equations (e.g., Anderson et al. 1989; Scaillet and Pichavant 2003, 2005). In general, thermobarometry based on mineral-mineral or mineral-melt equilibria (see Putirka 2008) provides further constraints on both temperature and fO2. The main unknown variable with respect to most active volcanoes is the reservoir depth, which can be evaluated either from phase equilibria (e.g., Rutherford et al. 1985; Johnson and Rutherford 1989; Martel et al. 1998; Cottrell et al. 1999; Scaillet and Evans 1999; Costa et al. 2004; Di Carlo et al. 2006), or from gas saturation systematics as first shown by Anderson et al. (1989) for the Bishop tuff for the C-O-H system. In the latter method, one seeks the pressure value which, for a given set of T, fH2O, and fCO2 conditions, fulfills the constraint of ΣXi = 1. Application of this thermodynamic approach to the C-O-H-S sytem was first attempted by Scaillet and Pichavant (2003) for several recent arc eruptions, for which the key intensive variables mentioned above were reasonably well known. The results are shown in Figure 1, where it can be appreciated that the sulfur content in the gas amounts, at most, to 5-6 wt% (or less than 6 mol% of H2S and SO2) for the investigated samples. Another important aspect is the considerable variability between magmas of the studied eruptions, with sulfur content as low as 0.1 wt% (less than 0.1 mol%) in some cases. This is despite the fact that all considered eruptions represent a single tectonic environment, namely subduction-zone volcanism. The reasons for such variability are certainly complex and include source heterogeneity (in terms of sulfur content); the vagaries associated with the various fractionation mechanisms that can affect a magma between its source and the shallow reservoir; and the different types of interactions between host rocks and magma or during magma mixing, both of which are commonplace in arc environments. Whatever the causes, such diversity clearly indicates that evaluating the actual sulfur content of the gas phase present in a reservoir feeding any active volcano needs to be carried out on a case-by-case basis. It was also shown by Scaillet and Pichavant (2003) that in order to make the volcanic gas data match with computed gas composition at depth, the reservoir must already be saturated with at least 1 wt% gas phase, a conclusion consistent with evidence from remote sensing observations discussed above. The calculation of the deep-gas composition also reveals that it is, in general, somewhat richer in sulfur species' than corresponding volcanic exhalations, with the sum of sulfur-bearing species approaching a few mol% (for
Sulfur Degassing From Volcanoes
367
6 5
S (wt%) in gas
4 3 2
n oa in
M
e
Ka
Gr
ov
o Pi
ne
tm
a up
tin
Ta
pu
en
na
nz Hu
ay
t ra
tU
er ts
on M
M
ée el
on
tP M
ny
Ca
on sh
Ch
ich
Fi
au at
El
ak
le
ns Kr
He
ba
St
To
op sh
Bi
Pi
na
tu
bo
0
ai
1
Figure 1. Plots of sulfur concentration of the pre-eruptive gas phase for several historical eruptions in arc settings. The sulfur content has been calculated using a thermodynamic approach (see text and Scaillet and Pichavant 2003).
evolved magma compositions such as dacites), whilst in gases from volcanic arcs, sulfurbearing species abundances are less than 1 mol% (see Symonds et al. 1994). This difference is obviously related to the fact that, upon ascent and with reduced pressure, the water originally dissolved in the silicate liquid in the reservoir (generally in the range 4-7 wt%) is almost totally lost to the gas phase, which thus becomes very water rich. Another difference worthy of note is the fact that high-pressure gases are generally dominated by H2S, while the reverse is observed in volcanic gases, as a result of the pressure effect on equilibria governing the relative abundance of H2S and SO2 (see below). This same method has been applied to volcanoes in non-arc settings by Scaillet et al. (2003), with the goal of estimating atmospheric sulfur yields of several major volcanic eruptions capable of impacting climate at a regional-to-global scale. Figure 2 compares the amounts of sulfur released by several eruptions, as derived by the improved petrologic approach, with those obtained by remote sensing techniques. In the derivation of petrologically-based values it is assumed that the magma coexisted with 5 wt% gas. Clearly, the generally good agreement observed between the two data sets supports the idea that most magmas contain a significant proportion of gas prior to eruption. Except for some outliers, the extended database provided by Scaillet et al. (2003) confirms that the sulfur content of the gas phase in silicic-tointermediate arc magmas is always below 10 wt%, with an average of around 3 wt%. When plotted against silica content of the magma, only a weak negative correlation appears: mafic magmas showing a tendency to be slightly richer in sulfur than silicic ones (Fig. 3). The lack of a more pronounced negative correlation contradicts the common view that silicic magmas tend to be poorer in sulfur than mafic ones, because of their intrinsic lower sulfur solubility: the latter fact is more than compensated for by the propensity of the gas phase to hold any sulfur present in the system (e.g., Scaillet et al. 1998; Keppler 1999, 2010), especially under oxidized conditions (Scaillet et al. 1998). However, it must be pointed out that mafic arc magmas seldom erupt and among those rare cases, only a handful have had their pre-eruptive conditions well constrained. The currently
Oppenheimer, Scaillet, Martin TOMS/COSPEC/IC estimate of Sulfur (g)
368 10
16
10
15
10
14
10
13
10
12
10
11
10
10
melt
Toba
melt+gas
Lascar
Stromboli
109
arc magmas Pacaya
108 10
7
107 108
109
1010 1011 1012 1013 1014 1015 1016
Petrologic estimate of Sulfur (g)
Figure 2. Plot showing comparison of sulfur yields for several historical volcanic eruptions estimated from remote sensing data (Table 3) and petrological and thermodynamic constraints, assuming that the gas phase in the pre-eruptive reservoir amounts to 5 wt% (from Scaillet et al. 2003). The results for sulfur yields obtained only from melt degassing are also shown.
Fig. 2, Oppenheimer et al.
S in gas + melt (wt%)
0.70
arc non-arc
0.60 0.50 0.40
Figure 3. Plot of relationship between the bulk content of S in various magmas erupted at arc and non-arc volcanoes versus the magma SiO2 content (from Scaillet et al. 2003)
0.30 0.20 0.10 0.00
40
50
60
70
SiO2 bulk rock (wt%)
80
Oppenheimer available information onFig. the3, sulfur content of gasetinalmafic arc magmas relies on a very limited data base. One of the best understood arc basalts is that erupted by Stromboli volcano, whose plumbing system is constrained by both phase equilibria (Di Carlo et al. 2006; Pichavant et al. 2009), detailed melt inclusion work (Métrich et al. 2001; Bertagnini et al. 2003, 2008), and volcanic gas emission data (Allard et al. 1994; Burton et al. 2007a; Aiuppa et al. 2010; Allard 2010). Here, the observations point to a deep reservoir located at pressures of 200-300 MPa, and connected to a very shallow reservoir residing at pressures equivalent to around 10-30 MPa. This, in turn, feeds the persistent, mildly explosive activity that typifies Stromboli. The calculation of the gas phase composition in the deep reservoir at 200-300 MPa (Scaillet and Pichavant 2005) indicates a sulfur content in the range of 0.3-6 wt%, which is similar to that
Sulfur Degassing From Volcanoes
369
inferred for more silicic compositions. It is worth stressing that these calculations depend strongly on fS2, which remains difficult to constrain for H2O-rich arc basalt, owing to the lack of appropriate thermodynamic models for sulfur solubility in hydrous mafic liquids. This is clearly an area where more experimental work is needed. It is thus difficult to know whether the Stromboli case is widely representative of other mafic arc magmas in respect of their sulfur content and yield. Other basaltic volcanoes should have the nature of their plumbing systems better elaborated in order to answer this important question.
Ocean ridge environments Although most magmas by volume are produced and erupted at ocean ridges, very little is known about pre-eruptive gas composition for this tectonic setting. This is due in large part to the difficulties of observing and sampling submarine volcanism. One of the closest analogs is seen in the rifting, diking and volcanism of the Afar region (Ferguson et al. 2010), though it is also a challenging place to conduct fieldwork. Hence, despite collection of some relevant volcanic gas composition data (e.g., Giggenbach and Le Guern 1976; Gerlach 1980), we lack modern petrological data (volatiles in melt inclusions) for corresponding magmas, suitable for constraining system behavior at depth as achieved for subduction zones (e.g., Scaillet and Pichavant 2003). Icelandic volcanism has been better studied, with some detailed information on dissolved volatiles present in magma reservoirs (Métrich et al. 1991; Moune et al. 2009) but here, too, there are very few data on related magmatic gas compositions. Therefore, for this volumetrically important category of magmas we must calculate gas composition at depth using available constraints on dissolved volatiles, which have been well established for a number of dredged oceanic basalts (see Wallace and Edmonds 2011, this volume). One salient feature of mid-ocean ridge magmatism is its very dry character, dissolved H2O lying in the range of 0.1-0.5 wt%, or even less than 0.1 wt% (Saal et al. 2002). Dissolved sulfur contents are also well characterized, falling in the range of 800-1200 ppm for primitive end members (Wallace and Carmichael 1992; Saal et al. 2002). Carbon dioxide concentrations are more controversial: some argue that mid-ocean ridge basalts (MORB) are gas-undersaturated under pre-eruptive conditions (Saal et al. 2002), while others (e.g., Sarda and Graham 1990) argue for CO2 contents of magmas well above solubility values at pressures of magma storage in the upper oceanic crust (100-200 MPa). Since the case for undersaturation has been disputed on thermodynamic grounds (Scaillet and Pichavant 2004), we show in Figure 4 the calculated C-O-H-S gas phase composition as a function of fO2 at 1280 °C and 40 MPa for fixed melt H2O content of 0.08 wt% and sulfur melt content around 800 ppm, which are averages for primitive MORB melts (Saal et al. 2002) (note that the bulk CO2 content of the magma need not be known to carry out the calculations since fixing fH2O and fS2 (at a given P and T) automatically fixes fCO2, and hence XCO2, in the gas). These data can be taken to illustrate the possible sulfur contents of the gas phase coexisting with MORB in reservoirs associated with axial ridges. These calculations illustrate a well-known feature, that the species abundances in the gas phase vary a lot with fO2, with CO2 being the most abundant species in MORBs (Mathez 1984). Sulfur-bearing species are generally present below 1 mol% at fO2 below NNO–1, with corresponding sulfur concentration in the gas being below 1 wt% for this fO2 range, or even below 0.1 wt% at around NNO–3. In contrast, at fO2 greater than NNO–1, SO2 abundances increase quite significantly to the point that it becomes the second most abundant gas species (after CO2), with sulfur concentration exceeding 20 wt% at around NNO+1. This illustrates the important point that, besides fS2 (or melt sulfur content), the master variable required is precise knowledge of redox state. This remains an inssue: despite decades of effort by various groups (e.g., Christie et al. 1986; Carmichael 1991; Ballhaus 1993; Bézos and Humler 2002; Li and Lee 2005; Mallmann and O’Neill 2009) there is still no consensus on whether MORBs are either strongly (i.e., fO2 < NNO–2), or moderately (FMQ-NNO) reduced, or could even share the same redox state as mafic arc magmas (Mallmann and O’Neill 2009). If one accepts
Oppenheimer, Scaillet, Martin
370 1
0.1
0.01 XS2 XH2S
0.001
XSO2
Xi
XCO XCO2
0.0001
XH2O XH2
0.00001
0.000001
A 0.0000001 -3
-2
-1
0
1
NNO
100
S wt% in gas
10
1
0.1
B 0.01 -3
-2
-1
0
1
∆NNO
Figure 4. Plots of (a) the variation of fluid species abundances versus fO2 (in log units relative to the nickelnickel oxide buffer), calculated for volatile conditions of H2O = 0.06 wt% and 850 ppm dissolved sulfur, which correspond to primitive Mid Ocean Ridge Basalts (Saal et al. 2002). The calculations were made for a pressure of 40 MPa and a temperature of 1280 °C. (b) Evolution of the sulfur content of the gas phase with fO2, corresponding to the gas composition shown in (a).
that most MORBs have pre-eruptive fO2 at or below NNO–2, then the sulfur fraction stored in the gas is small, below 1000 ppm, with a negligible contribution to the bulk amount of sulfur degassed. In contrast, if the MORB fO2 is around NNO, sulfur contained in the gas approaches or may even exceed 10 wt%, in which case the gas phase contribution to eruptive loss of sulfur becomes important, or even dominant, relative to that dissolved in the melt phase, although we have few constraints on the amounts of gas present in the reservoir. Gas compositions collected from subaerial volcanoes associated with divergent plates (Ardoukoba, Erta ‘Ale, Surtsey; see Giggenbach and Le Guern 1976; Allard et al. 1977;
Sulfur Degassing From Volcanoes
371
Gerlach 1980; Sawyer et al. 2008a) have sulfur contents in the range of 10-30 wt%. If those volcanic centers are representative of their deep-oceanic counterparts, it may indicate that MORBs are more oxidized than generally believed. To conclude, the redox state of magmas at divergent plates still needs to be better evaluated—a conclusion that also applies to many, if not all, other magmas! This issue remains a priority for improving our understanding of sulfurrelated magmatic processes.
Hot spots The few hot spots where volcanism is presently active include Hawaii and Réunion. We also include Etna in this discussion although its tectonic setting is especially complex and contested (e.g., Schiano et al. 2001) with chemical characteristics of the magmas pointing to either hot spot or arc settings. Thus, any conclusions drawn from the Etna case should be viewed cautiously. Volcanic gas emissions from both Hawaii and Etna have been studied in considerable detail over the last half century, both using conventional approaches, as summarized below and by Symonds et al. (1994), and with more recently developed groundbased remote sensing tools (see below). Volcanic gases emitted at la Réunion are less well characterized as there is little degassing between eruptions (e.g., Toutain et al. 2002). In all three cases, a wealth of microanalytical work aimed at determining pre-eruptive dissolved volatile contents has been undertaken (e.g., Dixon et al. 1997; Bureau et al. 1998, 1999; Wallace and Anderson 1998; Dixon and Clague 2001; Wallace 2002; Spilliaert et al. 2006). Despite some significant variations between the three volcanoes, it appears that hot spots fall roughly in between oceanic and arc magmas, in terms of their pre-eruptive H2O contents (0.4-3 wt%) and redox state (NNO-NNO+1), although considerable uncertainties remain concerning redox state evaluation or significance (Gerlach 2004a; Roeder et al. 2004). The pre-eruptive melt sulfur contents are somewhat higher than for MORBs at comparable levels of FeO content, ranging between 1000 to 2000 ppm (Bureau et al. 1999; Dixon and Clague 2001), and reaching 3000 ppm in the H2O-rich variety of basaltic magma erupted at Etna (Spilliaert et al. 2006). The gas phase composition in the magma reservoir of hot-spot volcanoes can be computed, as illustrated for MORBs, assuming a pressure depth for the shallow reservoir in the range 100-200 MPa and various contents for key volatiles, based on petrological data summarized above. Figure 5 shows the variations of sulfur content of the gas phase as a function of fO2, for various H2O contents and fS2. For the latter, petrological investigations indicate that fS2 of hot-spot related basalts is within the range of 0.1-100 kPa (Wallace and Carmichael 1992). As already pointed out for MORBs, these calculations show clearly that, in general, the sulfur content of gas remains below 0.1 wt% at fO2 below NNO (hence, to a first approximation, the gas contribution to the bulk sulfur budget can be ignored). Above this fO2 threshold, the sulfur content can increase dramatically, even becoming the dominant constituent of the gas phase. Note that, under fairly reduced conditions (NNO–2), the gas can become quite sulfur-rich if the magma has a dissolved H2O content of 3 wt% or more. However, such elevated H2O levels have not been documented for reduced terrestrial basalts (but see Gaillard and Scaillet 2009 for the case of Mars). Volcanic gases measured at other hot spots are essentially similar to those of Kilauea (Gerlach 1980, 1993), which have sulfur contents in the range of 20-30 wt%, regardless of their more or less primitive character (i.e., degassed right after their arrival in the shallow reservoir or subsequently following equilibration with local conditions). Mt. Etna gases analyzed during the volcano’s 1970 eruption are the most sulfur-rich volcanic gases so far collected, with a total sulfur content exceeding 50 wt% and a SO2/H2S molar ratio of around 100 (Huntingdon 1973; Gerlach 1979). The more recent activity has shown a significant decline in sulfur content of emitted gases (Allard et al. 2005), although Etna remains today one of the main sustained volcanic sources of sulfur to the atmosphere (Allard et al. 1991; Allard 1997). These two
Oppenheimer, Scaillet, Martin
372 1
2 wt% H2O, fS2 = 0.1 2 wt% H2O, fS2 = 0.01 0.5 wt% H2O, fS2 = 0.1 0.5 wt% H2O, fS2 = 0.01
wt% S in gas
0.1
0.01
0.001
0.0001 -2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
NNO
Figure 5. Plot of variation in S content of gas versus fO2 (in log units relative to the nickel-nickel oxide buffer), calculated for 100 MPa total pressure for a range of volatile contents and corresponding fugacities (H2O and S) corresponding to the conditions inferred for present day hot spot related basalts (see text).
examples seem thus to suggest that hot-spot activity may coincide with elevated rates of sulfur emission. The high sulfur content of gases released also seems to indicate that magma redox state in the hot-spot context is somewhat oxidized relative to that of MORB. This evidence once again highlights the important role of redox conditions in gas phase composition. In terms of the sulfur yield of past eruptions to the atmosphere, little can be said without good constraints on redox. The problem can be even more complex when decompression-related effects on magma fO2 are taken into account. Recent modeling efforts, though rudimentary, indicate that volatile loss during decompression may strongly alter magma redox state (Moretti et al. 2003; Burgisser and Scaillet 2007; Burgisser et al. 2008; Gaillard and Scaillet 2009), adding further “noise” to the direct interpretation of volcanic gas data in terms of magma storage or generation conditions. Modeling the interplay between decompression and redox state of magmas and related gases is clearly another fruitful avenue for future research.
Flood basalts and silicic parts of large igneous provinces Interest in the potential global impacts of effusive volcanism has been fueled by the coincidence of several of the mass extinctions that punctuate the fossil record with the flood basalt episodes of large igneous provinces (LIPs; Rampino and Stothers 1988; Wignall 2001; Oppenheimer 2011). Estimated sulfur yields of such provinces are exceptional (e.g., Thordarson et al. 1996), and Ar-Ar chronometry has shown that the immense volumes of lava associated with flood basalts (> 106 km3) are erupted in comparatively short time periods For instance, the 30-Ma-old Ethiopian Plateau basalts erupted in under 1 Ma (Hofmann et al. 1997) while the main eruptive phase of the Deccan traps (65.5 Ma old) was over within 0.5 Ma (Courtillot and Renne 2003). Such rapid eruption rates imply correspondingly high fluxes of sulfur and other volatiles to the atmosphere. Self et al. (2008) analyzed melt inclusions in samples of the Deccan basalt and estimated that for each cubic kilometer of lava erupted, at least 3 Tg of SO2 were released into the atmosphere (plus a comparable amount of HCl). An individual pulse of 1000 km3 of magma
Sulfur Degassing From Volcanoes
373
would therefore release 4-5 Gt of SO2 (at least 200× as much as Pinatubo’s eruption in 1991). We have very little idea how long each flow took to erupt but estimates vary from years to a few decades. A 10-year duration would correspond to a mean annual eruption rate of 100 km3 of lava and 10-20× more sulfur than emitted by Pinatubo in 1991, whose sulfate aerosol veil resulted in a 0.3-0.4 °C global average cooling for 2-3 years (see below; Hansen et al. 1996). Such massive eruption intensities along with the expanse of active lava covering the surface are thought to have generated especially energetic fire fountains and vigorous atmospheric updrafts, enabling a substantial fraction of the released sulfur gases to reach the stratosphere. Here, they would have oxidized to form a sulfate aerosol veil capable of climate forcing. The estimates based on basaltic melt inclusions (Sharma et al. 2004; Self et al. 2008) set a lower limit on sulfur outgassing since, for equilibrium partitioning between melt and gas, any pre-eruptive magmatic gas must hold a finite amount of sulfur. In view of the preceding discussion of the importance of knowing fO2 in order to evaluate the distribution of sulfur between silicate melt and gas phases (oxidized magmas tend to store a significant quantity of their sulfur inventory in the gas phase), an important factor in assessing the environmental impacts of flood basalt events is their redox state. Flood basalts with fO2 in the range of NNO− to NNO+1 could yield sulfur outputs far higher than those estimated from melt inclusions alone. If, in contrast, the basalts have a lower redox state, then the sulfur content of the gas phase would be of negligible significance in calculating total sulfur emissions. Detailed investigations of volatile distribution have been carried out for the volcanoes of Hawaii, the modern archetype of the plume, or hot spot, regime viewed as parental to flood basalts. These have documented great regional variability in magma volatile contents and redox states (Dixon et al. 1988, 1991, 1997; Dixon and Clague 2001). If the Hawaiian plume can be considered as representative of hot spot magmatism through geological time, then the associated sulfur degassing is likely to have been very variable in time and space. A further point to be stressed is that, along with the enormous quantities of basalt, flood provinces can include a silicic component (e.g., the Ethiopia-Yemen (30-Ma-old) and Etendeka-Paraná (133-Ma-old) provinces). The behavior of sulfur in such environments suggests that the silicic end-members could contribute significantly to LIP sulfur budgets (Scaillet and Macdonald 2006). A critical factor is the Al/(Na+K) molar ratio of rhyolites. Peralkaline rhyolites have much higher sulfur solubilities than metaluminous magmas and, as a result, could convey comparable amounts of sulfur as the mafic counterpart of LIPs. In contrast, metaluminous rhyolites with much lower sulfur solubilities and different mode of production (i.e., by remelting previously dehydrated crust leading to particularly H2O- and S-poor magmas, see Scaillet and Macdonald 2006), could be expected to have diminished impacts on the atmosphere and climate for eruptions of the same magnitude. This may be one reason for the limited correlation apparent between the intensity of environmental changes and the magnitude of coincident flood magmatism (Wignall 2001).
MEaSuRInG VOlCanIC SulFuR EMISSIOnS Gas emissions from volcanoes are sourced within magmatic plumbing systems and their associated hydrothermal systems. Their vigor and chemical and isotopic composition can reveal much about subterranean processes, especially concerning the chemistry, differentiation, storage and transport of magmas and the relationships between degassing and eruptive style. The importance of gas emissions was recognized by early pioneers of volcanology, including Frank Perret and Thomas Jaggar, who bubbled fumarole gases into bottles of alkaline solution to trap the acidic species for subsequent analysis (Fig. 6a; Perret 1909, 1950; Jaggar 1940). This sampling method remains relevant today, with incondensable species held in the bottle’s head space also studied (Table 1).
374
Oppenheimer, Scaillet, Martin
(a)
(b)
Figure 6. Sampling volcanic emissions the traditional way: Frank Perret on Vesuvius and Patrick Allard on Erta ‘Ale.
Of course, since Perret’s time, analytical capabilities have developed tremendously, and this manner of gas collection continues to yield the most comprehensive analyses of the chemical and isotopic composition of volcanic gas emissions. However, such sampling is challenging for sustained geochemical surveillance, especially when fumaroles are inaccessible or levels of volcanic activity preclude safe approach. This section focuses on the principal sampling and sensing methods applied to measurements of sulfur emissions from volcanoes. It introduces direct sampling and sensing apparatus, ground-based or airborne ultraviolet spectroscopy
0.11
0
0
OCS
0.0025 0.000047
0.01797
0.04848
0.000014
<0.000005
CH4
0.086
0.60794
He
0.096
HF
1.01
0.0176
0.74
HCl
0.134
0.98
O2
0.39
H2S
6.98
0.0235
0.00009
2.05
SO2
0.016
5.56
ar
0.00084
CO
2.27
0.50059
3.6047
1.15
CO2
0.54
8.7
0.025
0.775
H2
84.77
803
n2
95.3
H 2O
870
Andesite, island arc
0
920
T, °C
Andesite, island arc
nH3
Basaltic andesite, island arc
0.000022
<0.003
0.00105
0.15
0.00003
0.25
0.68
1.02
0.0042
1.605
0.142
96
640
Andesite, island arc
Kudriavy, augustine, Merapi, Java, ngauruhoe, Kuril Islands uSa 1994 new Zealand
Magma type & environment
Volcano
0.00025
0.057
0.00071
0.024
0.053
0.0004
0.043
0.0033
1.2
0.63
98
800
Dacite, island arc
Showa Shinzan, Japan
0.000006
<0.0005
0.000058
0.008
0.00013
0.0004
0.0277
0.5306
0.164
0.818
0.00172
0.51
0.607
97.3
880
Rhyolite/ basalt, island arc
Satsuma Iwojima, Japan
0.000002
<0.0001
0.0001
0.0104
0.0003
0.00057
0.006
0.034
0.018
0.032
0.00005
0.36
0.2
99.3
690
Rhyolite, island arc
usu, Japan
0.00012
<0.0002
0.00012
0.1068
0.00056
0.00001
0.101
0.453
0.48
0.68
0.011
11.9
0.151
86.1
620
0.08
0.029
0.35
0.49
0.7
0.025
2.38
0.87
95.05
844
0.00005
<0.00005
0.00011
0.05529
0.000072
<0.00003
0.03839
0.5276
0.2335
0.01122
2.532
0.3516
96.25
747
Basalt, Basalt, continental arc continental arc
Momotombo, Momotombo, nicaragua, nicaragua, 1982 2002
table continuted on following page
Rhyolite, island arc
Vulcano, Italy
table 1. Representative compositions of high-temperature volcanic gas samples* in mol%. Data from compilations in Giggenbach (1996), Oppenheimer (2003), Fischer (2008) and from individual analyses in Gerlach (1980), Oppenheimer et al. (2002b), Oppenheimer and Kyle (2008) and Sawyer et al. (2009).
Sulfur Degassing From Volcanoes 375
Galeras, Colombia, 1991 Mt. St. Helens, uSa
Oldoinyo lengai, tanzania, 1999
0.000032
ar
0.1
0.0004
0.015
0.38
2.83
0.012
1.35
0.9
95
940
0.00032
0.041
0.062
0.716
0.5725
0.8415
0.0136
5.98
0.285
91.48
642
0.000025
0.12
0.00003
0.03
0.15
0.27
0.003
0.88
0.4
98.9
710
*except Lengai, Erebus and Nyiragongo, which were measured by open-path FTIR spectroscopy ** sample J-8 in Gerlach (1980) already adjusted for N2 and Ar and with Cl2 reported as HCl
OCS
He
O2
0.004
<0.0001
n2
nH3
CH4
0.25
0.28
0.09
HF
0.05
SO2
0.0012
1.42
0.0001
CO
0.12
1.06
0.04
CO2
0.59
HCl
0.55
H2
97.73
0.002
98
H 2O
1100
H2S
1020
T, °C
0.0197
0.0787
24.4
75.6
570-590
Magma Basalt, Basalt , Basalt/ Andesite, Dacite, Carbonatite, type & continental arc continental arc andesite, continental arc continental arc continental environment continental arc intraplate
Volcano
tolbachik, Kliuchevskoi, Poas, Costa Kamchatka, Kamchatka Rica 1976
table 1. Continued.
0.0023
0.11
0.26
4.55
0.86
23.68
70.54
1100
Basanite, continental intraplate
nyiaragongo, DR Congo, 2005
0.18
0.42
0.62
6.78
0.46
10.4
1.49
79.4
1130
Basalt, incipient oceanic ridge
Erta ‘ale, Ethiopia
0.01
1.27
0.69
1.4
2.33
36.4
57.9
>1000
Phonolite, continental intraplate
Erebus, antrarctica, 2004
0.08
11.87
1.5
48.91
0.49
37.11
<1200
Basalt, ocean island
Kilauea, uSa, 1918**
376 Oppenheimer, Scaillet, Martin
Sulfur Degassing From Volcanoes
377
(Correlation Spectrometer and successors), ground-based infrared spectroscopy (Fourier transform spectroscopy), and satellite remote sensing, including the important role of the Total Ozone Mapping Spectrometer, Ozone Monitoring Instrument, Infrared Atmospheric Sounding Interferometer and other ultraviolet and infrared sensors, for measuring the SO2 emissions from major eruptions.
Direct sampling Conventional analyses of volcanic gases have been made by collection of samples directly from fumarole vents using evacuated bottles and caustic solutions, and subsequent laboratory analysis (Symonds et al. 1994). The classic “Giggenbach bottle” (Fig. 6b, Giggenbach 1975; Giggenbach and Goguel 1989; Giggenbach and Matsuo 1991; Giggenbach et al. 2001) consists of an evacuated glass or quartz vessel partially filled with NaOH or, alternatively, ammonia (NH4OH) solution (Sortino et al. 2006). Sampling is carried out by bubbling gas through the solution via tubing inserted into the volcanic vent. Hydrogen sulfide and SO2 readily condense according to reactions such as: 4SO2(g) + 7OH − (aq) = 3SO 4 2 − (aq) + HS− (aq) +3H 2 O(l)
(13)
H 2S(g) + OH − (aq) =HS− (aq) + H 2 O(l)
(14)
The aqueous phase species can be analyzed by ion chromatography. The remaining gaseous species collect in the headspace and are usually analyzed by gas chromatography. A reagent such as Cd(OH)2 or AgNO3 can be used to separate H2S (which reacts with Cd2+ to precipitate CdS, or with Ag+ to precipitate Ag2S) for subsequent analysis (after conversion to sulfate) by ion chromatography (Picardi 1982; Montegrossi et al. 2001, 2008; Aiuppa et al. 2005b). Other solid phases, including Sx, can be quantified (e.g., by progressively oxidizing them and analyzing the anions by ion chromatography). Various activated substances such as silica gel have also been used to trap volcanic gas species for subsequent laboratory analysis, e.g., by gas chromatography (Naughton et al. 1963). Base-treated filters and diffusion tubes provide another means to trap acid species, and can be deployed around crater rims and in the vicinity of gas sources (e.g., Allen et al. 2002; Aiuppa et al. 2004; Martin et al. 2010b). These have been used in studies of SO2 plume dispersion and exposure (Allen et al. 2000; Delmelle 2002; Longo et al. 2005; Aiuppa et al. 2007a; Bhugwant et al. 2009) while rainwater collectors and sulfation plates have been used to map and investigate sulfur deposition (Delmelle et al. 2001, 2003; Aiuppa et al. 2006). Studies of the aerosol phase in volcanic emissions (of which sulfate is often the most abundant component) have been carried out using particle filters. Such filter-based methods have been extended to characterize aerosol size distribution and chemistry (e.g., Mather et al. 2004b; Martin et al. 2008; Ilyinskaya et al. 2010). The volatiles scavenged out of eruption clouds by ash particles, which then sediment to the ground, can also be studied analytically by leaching samples with distilled water (e.g., Edmonds et al. 2003b; Witham et al. 2005) Although such approaches offer very high sensitivity, measurements can be difficult and often hazardous to obtain, and complications can arise from post-collection reactions. Also, the typical delays in obtaining results (where laboratory analysis is required) can limit their value in volcanic crisis management.
In situ sensing One particularly promising approach to volcanic gas surveillance is the application of electrochemical sensors. These sensors contain an electrolyte, which is exposed to ambient air (and the volcanic emissions) by diffusion (with or without the aid of an air pump). The ensuing redox chemistry generates a current that is proportional to the target gas abundance. Their
378
Oppenheimer, Scaillet, Martin
disadvantages include sensor drift and imperfect specificity to target gases. This represents a significant problem because of the cocktail of gases typically found in a volcanic plume; for example, most commercially available H2S sensors are cross-sensitive to SO2 but because they are mass-manufactured for a wide-range of industrial and consumer applications they are cheap. They have been tested in a number of configurations, and appear capable of reliable SO2 and H2S (and other species) measurements (particularly when combined with, for instance, non-dispersive infrared sensors for CO2 measurement; e.g., Aiuppa et al. 2005a; Shinohara 2005; De Vito et al. 2007; Witt et al. 2008; Roberts et al. 2011). An important development is that long-term installations (using Wi-Fi or cell-phone networks, or satellite telemetry) are beginning to provide valuable and near-real time insights into the relationships between surface emissions and magmatic processes (e.g., Aiuppa et al. 2007b, 2010). Another promising field technique established some while ago, though yet to be more widely used in volcanology, is the portable gas chromatograph (Le Guern et al. 1982; Diaz et al. 2002). Its advantage is its sensitivity and the wide range of species that may be quantified, including H2S and H2.
ultraviolet spectroscopy A striking aspect of volcanic emission measurements is the prominence of observations of sulfur dioxide (Oppenheimer 2010). Considering that the most abundant volcanic vapors are water and carbon dioxide and that sulfur species (including hydrogen sulfide) typically account for less than 5 mol% of the gases emitted, it is not immediately obvious why sulfur has been such a focus of study. The reasons are the comparative ease with which SO2 can be measured in the atmosphere and because S is a useful tracer of magmatic processes The former aspect is due to its electronic absorption spectrum which provides a strong fingerprint against the ultraviolet background light in the daytime sky. Ultraviolet sensing using the sky as source eliminates one typical problem with optical spectroscopy—the need for precise alignment of the spectrometer with respect to the light source. All that is necessary is to collimate some sky light through a telescope and direct it into the spectrometer. The particular importance of ultraviolet sensing of volcanic plumes is that it represents the most straightforward means for estimating degassing rates (which are difficult to measure based on point sampling). The flux of SO2 from a vent or crater can be measured by recording the column density of all the SO2 molecules contained in a (generally though not necessarily) vertical section of the whole plume, and computing the product of the integrated column cross section of SO2 and the plume speed. In practice, the plume may be traversed roughly perpendicularly to the wind direction while pointing the spectrometer’s telescope up from the ground, or scanned from a fixed position. Thanks to the diffuse sky ultraviolet source, the pointing direction of the telescope is not critical. Acknowledging the significance of gas flux measurements in volcanic hazard assessment, this subsection will review in greater depth the development and application of ultraviolet remote sensing and spectroscopy in volcanology. The Correlation Spectrometer (COSPEC). The first remote sensing instrument to become widely used for volcanic plume monitoring was the Barringer Research “COSPEC”, or Correlation Spectrometer (Williams-Jones et al. 2008). The COSPEC was introduced four decades ago to measure industrial SO2 and NO2 sources (Moffat and Millán 1971; Hoff and Millán 1981), but its value for measuring volcanic emissions was quickly recognized (Stoiber and Bratton 1978; Stoiber et al. 1983). Early volcanological interest addressed the question of whether changes in SO2 flux could be associated with magma migration and eruptive style. For instance, Malinconico (1979) reported a positive correlation between increasing SO2 flux and eruptive vigor at Mt. Etna. The Correlation Spectrometer went on to play a prominent monitoring role during many volcanic crises, crucially helping to discriminate magmatic origins of volcanic unrest. Of particular note, COSPEC played a vital role in civil
Sulfur Degassing From Volcanoes
379
emergency planning during the unrest of Mt. Pinatubo in 1991. As seismic unrest increased two weeks before the magmatic denouement, so did the SO2 flux, supporting the view that a substantial magma body was connecting with the surface. This interpretation prompted the civil evacuation credited with saving tens of thousands of lives (Daag et al. 1996). COSPEC measurements have also been influential in recognizing the importance of magma convection, permeability and gas separation (e.g., Francis et al. 1993; Kazahaya et al. 1994; Oppenheimer et al. 2002a). Furthermore, as COSPEC observations were made at more and more volcanoes around the world, the compiled data yielded another important result, namely the first estimates of the global emission of volcanic SO2 (discussed in a later section). The COSPEC design goal was to build a system capable of minimizing all “noise” (including those due to light absorption and scattering by other atmospheric constituents) in order to measure a single species (e.g., SO2). The result was an instrument that yields an estimate of the quantity of SO2 in the field of view (measured, for instance, in units of mass per unit area of the atmospheric column). While this greatly simplified data retrieval and processing, it made it difficult to assess potential errors arising from wavelength shifts, light scattering and solar-elevation effects, thermal and mechanical distortions, and so on. Also, the COSPEC response depended on the correlation mask used and the abundance of gases present. An alternative approach is to measure wide-band spectra with sufficient spectral resolution to be able to model trace-gas abundances. With the COSPEC long out of production now, a new generation of ultraviolet spectrometers has come to the fore in the realm of volcano surveillance (Oppenheimer 2010). Ultraviolet Differential Optical Absorption Spectroscopy (DOAS). The common approach to recording an ultraviolet spectrum is use of a diffraction grating to disperse the light received by an optical telescope coupled to the spectrometer. Spectra are typically collected using a CCD detector array onto which the dispersed light is focused. Analysis of the spectra and determination of trace-gas abundances are widely carried out with a methodology known as differential optical absorption spectroscopy, or DOAS (Platt and Stutz 2008). A key approach of DOAS is the removal of fluctuations in the recorded spectrum that result from molecular and aerosol scattering in the atmosphere. Because their spectral dependence is of lower frequency than that associated with the electronic structure of absorbing molecules, appropriate signal processing yields absorption spectra suitable for modeling. Another trick widely used in DOAS practice is to obtain atmospheric spectra with and without the emissions of interest present. In the case of volcanic emissions, this can be achieved by collecting “background” or “clear sky” spectra from either side of the plume. All the plume spectra are then divided by the out-of-plume background spectrum. This reduces interferences caused by background atmospheric absorption and the solar spectral structure (Fraunhofer lines). Specific gases can be identified by their characteristic absorption spectra, and their abundances derived from the strength of the absorption, following the BeerLambert formula: I (λ )= I 0 (λ )exp( −σ(λ ) NL )
(15)
where I(λ) is the observed intensity of radiation at wavelength λ, I0(λ) is the original intensity of radiation before interaction with the sample, σ(λ) is the absorption cross section of the absorbing molecule at wavelength λ, and N is the mean concentration of the species over the path-length L of the sample. In general practice, the logarithm of the ratio spectrum is computed, and then SO2 (and other gas species) amounts can be calculated by scaling reference spectra of the gases of interest (obtained via laboratory experiments: Fig. 7) until they match the observed spectrum. The scaling factors thus identified, along with the known abundances for the reference spectra, determine the gas column amounts in the plume.
Oppenheimer, Scaillet, Martin
380 10000
BrO
s(l) (cm2 molec-1 1020)
1000
IO OClO
100 SO2
NO2
10 HONO
1
300
320
340
360
380 400 Wavelength (nm)
420
440
460
480
Figure 7. Plots of absorption cross sections in the ultraviolet and visible regions of the spectrum for several gases: SO2, NO2, BrO, HONO, IO and OClO (convolved with a Gaussian function of 1.21 nm full-widthat-half-maximum). Data courtesy of Vitchko Tsanev (University of Cambridge).
The first volcanological DOAS observations were made in 1992 from a scientific vessel that cruised the Mediterranean downwind of Etna, Stromboli and Vulcano (Edner et al. 1994; Weibring et al. 1998, 2002). A bulky ultraviolet spectrometer was employed, and it was operated alongside a COSPEC and a lidar (light detection and ranging) system that was also capable of SO2 measurements (see below). At that time, COSPEC represented the state of the art for remote geochemical surveillance of volcanoes, and its pre-eminence was not challenged by the expensive and more operationally complex DOAS and laser-based instruments. But this picture changed in early in the 2000s owing to the commercial availability of ultraviolet spectrometers built around low-cost CCD detectors and mass-produced optical benches. They were cheaper, smaller and less power consumptive than anything else on the market and could be controlled via USB connection to a portable computer. The first volcanological measurements with this new generation of spectrometer were made at Masaya (Nicaragua) in 2001 (McGonigle et al. 2002; Galle et al. 2003). Several competing spectrometers are now available (Kantzas et al. 2009), and there has been tremendous innovation in the application of the new technology as volcanologists have been quick to seize on the potential of such cheap, adaptable and capable devices. One of the first tasks required was to confirm good agreement between DOAS-based estimates of SO2 emissions and those obtained by COSPEC. This was particularly important for volcano observatories switching to the new technology (Elias et al. 2006). The next development was the more imaginative deployment of the miniature ultraviolet spectrometers (Fig. 8). Observational platforms for traverse-style measurements have included pedestrians (McGonigle et al. 2002; Mori et al. 2006), fixed-wing aircraft, helicopters and ultra-light aircraft (Grutter et al. 2008) and a remote-controlled helicopter (McGonigle et al. 2008). With respect to gas-flux measurements, considerable attention has to be given to the estimation of plume transport speeds. It is widely acknowledged that uncertainty in plume
Sulfur Degassing From Volcanoes
381
speed (typically a few tens of %) dominates errors in measurements of SO2 and other gas fluxes. A consistent approach to constraining plume speed is therefore essential. Absorptioncorrelation methods using spatially distributed instruments (McGonigle et al. 2005; WilliamsJones et al. 2006) or a multi-beam instrument (e.g., Johansson et al. 2009b; Boichu et al. 2010) are arguably the most rigorous means to record plume velocity accurately (see below). By virtue of recording spectra, broadband UV spectroscopy has the capability of detecting multiple gas species. Measurements of H2S/ SO2 ratios have been accomplished at Vulcano, Italy, using a compact ultraviolet spectrometer and an ultraviolet lamp (O’Dwyer et al. 2003; Aiuppa et al. 2005a). The use of an artificial source along with comparatively short pathlengths is required because the H2S-absorption feature is in the shorter-wavelength region of the ultraviolet spectrum that is not transmitted over long paths in the atmosphere. Another species that can be measured by DOAS in the ultraviolet region (but at longer wavelengths) is BrO. This species is of significance for its impact on atmospheric ozone and measured SO2/BrO ratios combined with SO2 flux estimates for a number of volcanic plumes worldwide are beginning to provide a picture of the regional and global importance of volcanogenic reactive bromine emissions (Oppenheimer et al. 2006; Bobrowski and Platt 2007; Bobrowski et al. 2007; Bani et al. 2009; Boichu et al. 2011). Operational surveillance. The sustained surveillance carried out by volcano observatories demands a high degree of automation of monitoring systems. This saves the time and expense of repeated fieldwork; enables networks of sensors to be developed; protects staff from prolonged exposure in hazardous areas; Figure 8. Photographs of portable ultraviolet and infrared sensing systems for plume measurements: (a) INGV-Catania’s Mini-COSPEC in operation on Mt. Etna; (b) compact ultraviolet spectrometer set up for walking traverses on Erta ‘Ale volcano (the spectrometer is near the laptop’s keyboard; (c) one of the first scanning systems for ultraviolet sensing of SO2 emissions, in use on Stromboli; (d) FTIR spectrometer during measurements on Erebus volcano, Antarctica.
382
Oppenheimer, Scaillet, Martin
and permits more time to be spent on data analysis, synthesis and interpretation. The low power requirements and cost of the new generation of ultraviolet spectrometers have spurred development of autonomous SO2 monitoring stations. However, to measure SO2-degassing rates, fixed installations must mimic the traverse method of flux estimation by scanning the sky sequentially in a plane that intersects the plume. This can be readily achieved by rotating the field of view of the spectrometer around the sky by use of a stepper motor to which a plane mirror is attached. Combined with a telemetry system and powered by solar energy, such a station may operate autonomously, transmitting spectra back to the volcano observatory. The first scanner development applied to volcanic-plume measurements was reported by McGonigle et al. (2002, 2003). This was a portable system designed for rapid, campaign installation, but it was soon superseded by Montserrat Volcano Observatory’s autonomous ultraviolet DOAS stations (Edmonds et al. 2003a). Subsequently, scanner networks were installed at Tungurahua volcano, Ecuador (Arellano et al. 2008), Stromboli (Burton et al. 2009) and Etna (Salerno et al. 2009a,b). Two European projects (DORSIVA and NOVAC; www.novac-project.eu/) have taken this concept even further through the development of a scanning system that has now been installed at seventeen volcanoes worldwide (Galle et al. 2010). One complication in obtaining a flux measurement with a scanning instrument is that uncertainty in the plume height corresponds directly to uncertainty in the flux estimation. Since the measurements are made radially within a plane that intersects the plume, the measurement from a single instrument cannot discriminate between a high, wide plume and a low, narrow plume (with correspondingly lower SO2 flux), even if the two have identical SO2 mixing ratios and thicknesses. Thus, plume height must be determined accurately if the scanner is to yield a reliable measurement of flux. This can be achieved crudely through simple scaling between wind speed and plume height, through more sophisticated dispersion modeling, or, ideally, from the observations themselves. Arguably the most robust procedure is the tomographic reconstruction of the plume’s SO2 cross section, demonstrated by Wright et al. (2008), Kazahaya et al. (2008) and Johansson et al. (2009a). Nevertheless, tomographic approaches rely on assumptions about distribution of SO2 in the cross section, and are computationally costly to implement. A further important benefit of scanning systems is their capability for rapid flux measurements. In practice, plume scans may be made within a few minutes, providing a high time resolution. This can reveal rapid variability in source emissions, which might, for instance, relate to fluctuating magma flow to the surface. When two-dimensional CCD detectors are used (one dimension representing the spectral information), it is possible to build an image with a scanning system, yielding valuable information on gas distribution within a plume (e.g., Louban et al. 2009). But even higher time resolution is possible with ultraviolet cameras (the use of which, however, entails sacrificing much spectral information) since an entire plume can be captured instantaneously with the imposition of filters to provide the spectral contrast needed for SO2 measurement (Mori and Burton 2006, 2009; Bluth et al. 2007; Yamamoto et al. 2008; Dalton et al. 2010; Kantzas et al. 2010). A variation on this approach is the use of telescopes with distorted optical fields of view. Boichu et al. (2010) designed a device with two spectrometers attached to telescopes with cylindrical lenses. The two fields of view were parallel but offset by a small angle, achieved with a goniometer. The advantage of the system is that the stretched field of view records all the SO2 rising out of the crater simultaneously (without recourse to scanning). Further, the angular offset between the two telescopes permits calculation of the plume speed by correlating the time series of SO2 obtained from the two spectrometers. This approach can provide accurate SO2 flux measurements at very high time resolution depending on the correlation length needed to resolve plume speed and could, in principle, provide a real-time SO2 flux meter. Reliable measurements are the cornerstone of volcano monitoring and are vital for effective hazard evaluation. These new spectroscopic approaches, which constrain plume speed and
Sulfur Degassing From Volcanoes
383
height, are particularly suited to delivering accurate, high time-resolution, SO2flux measurements. Work remains to be done to establish the extent to which atmospheric radiative transfer effects, such as multiple scattering in dense plumes and in-scattering of light in front of plumes when observed from long distances, affect measurements (Kern et al. 2010), but undoubtedly, SO2 observations will increasingly contribute to routine observatory work and to a deeper understanding of magma degassing and volcanic activity.
Broad-band infrared spectroscopy Infrared absorption spectroscopy is widely used in the analytical sciences for measuring the rotational structure of vibrational spectral bands. Naughton et al. (1969) appear to be the first to have applied the technique in an open-path configuration to field measurement of volcanic gas emissions. They were able to measure SO2, CO2 and H2O at Kilauea using a fire fountain as the infrared source. However, it was not until the first half of the 1990s that field portable Fourier transform infrared (FTIR) spectrometers became more widely used in volcanology. FTIR spectrometers are based on Michelson interferometers. Collimated light from an infrared source is admitted to the instrument and divided into two beams using an optical beam-splitter, which also recombines these beams after they are reflected at mirrors. One of the mirrors is translated back and forth along the optical axis, introducing a variable path difference between the two beams, resulting in a time-variable signal from the single broadband detector due to interference of recombined beams. Application of an inverse Fourier transform to the temporal signal yields absorption spectra, from which the column amounts of volcanic gases present in the optical path can be retrieved (following the Beer-Lambert law; Eqn. 15). A range of infrared light sources: direct sunlight, fire fountains, lava lakes, hot rocks, and infrared lamps have been used in volcanological work highlighting the flexibility of the approach to adapt to the circumstances of activity, access and terrain. Love et al. (1998, 2000) have shown it is also possible to measure volcanic gases emitting infrared radiation against a cold sky background. At Popocatépetl volcano, Mexico, Love et al. (1998) observed a steady increase in SiF4/SO2 ratio prior to an eruption in February 1997, followed by a tenfold decrease within a few hours. These results suggested a cooling of the gas prior to the eruption, attributed to adiabatic gas expansion on release of a solidified magma plug in the conduit. The first FTIR spectroscopy of volcanic emissions was carried out in 1991 at Asama volcano, Japan (Notsu et al. 1993), by a team that has subsequently reported measurements at Unzen, Japan (SO2 and HCl; Mori et al. 1993); Aso, Japan (CO, OCS, CO2, SO2 and HCl; Mori and Notsu 1997); and Vulcano, Italy (SO2 and HCl; Mori et al. 1995). Benefits of open-path infrared spectroscopy include its adaptability to diverse field circumstances (Francis et al. 1995, 1998; Oppenheimer et al. 1998a) and the ability to detect and measure abundances of several gases of interest (including H2O, SO2, HCl, HF, CO2, CO, SiF4, and OCS). Measurements of the relative abundance for different gases can be made with a temporal sampling rate of about 1 Hz (Fig. 9). Data can even be collected during more vigorous eruptive episodes. This provides significant insights into the dynamics of magma transport and degassing of magma (e.g., Allard et al. 2005; Burton et al. 2007b; Oppenheimer et al. 2009, 2011). Although H2S has a well-characterized infrared spectrum, its detection in a volcanic plume by infrared absorption spectroscopy does not appear to have been reported. The accessible bands are relatively weak and overlap with regions prone to stronger signals from H2O and CO2. Nor can S2 (or H2) be measured by conventional infrared spectroscopic methods since they are homonuclear and therefore their vibrational modes are not infrared active. This is unfortunate, since measurement of combinations of oxidized and reduced species enables modeling of the degree of equilibrium represented by the volcanic emission. However, CO and OCS can be measured under suitable conditions, and their relative abundances probed
Oppenheimer, Scaillet, Martin
384
(a)
(b)
200 180 160
SO2/OCS /mol mol-1
140 120 100 80 60 40 20 0 03:00
04:00
05:00
UTC 15 December 2005
Figure 9. (a) Photograph of Erebus volcano showing strong degassing from the anorthoclase phonolite lava lake within the summit crater; (b) plot of molar ratio of SO2/OCS retrieved from absorption spectra of the plume collected with a FTIR spectrometer sited at the crater rim and viewing the lava lake (the infrared source). Raw interferograms were collected with a time-step of ~1 s. Note the gas signature of an explosion 03:15 UT (much lower SO2/OCS) and the quasi-periodic variation in the gas ratio (period around 10-20 min), which is thought to reflect pulsatory magma supply to the lake (Oppenheimer et al. 2009).
according to the following redox equation: 3CO + SO2 = 2CO2 + OCS
(16)
Factors that affect the balance of reactants and products include oxidizing capacity of the melt, temperature and pressure. Infrared spectroscopy has succeeded in measuring these four species at a few volcanoes including Aso (Mori and Notsu 1997, 2008), Stromboli (Burton et al. 2007b), Nyiragongo (Sawyer et al. 2008b) and Erebus (Oppenheimer and Kyle 2008).
Sulfur Degassing From Volcanoes
385
laser spectroscopy To date, the most commonly applied laser technique for volcanic plume measurements is lidar, in which a pulsed laser beam is directed towards the plume. Recording the temporally varying intensity of backscattered light provides information about the atmospheric composition as a function of propagation distance along the beam’s path. Lidar has been used to measure concentrations and fluxes (via traverses) of sulfate aerosol (Casadevall et al. 1984; Porter et al. 2002), as well as ash. Parallel gas sampling and aerosol measurements can enable estimation of gas to particle conversion rates (e.g., for SO2 to SO42−; Stith et al. 1978; Radke 1982; Rose et al. 1986). A variation on lidar known as differential absorption lidar (DIAL) involves rapid switching the frequency of laser pulses on- and off-resonance of an absorption feature of the gas of interest. By ratioing the lidar signals (returned signal versus height) obtained at the two wavelengths and applying the Beer-Lambert law (Eqn. 15), range-resolved gas mixing ratios may be derived revealing detailed plume structure, in contrast to the column amounts obtained from FTIR, COSPEC and DOAS. The technique has been applied to the Southern Italian volcanoes using ultraviolet lasers (Edner et al. 1994; Weibring et al. 1998, 2002), in parallel with COSPEC and an ultraviolet spectrometer, revealing SO2 column amounts up to 50% higher in the former case as a consequence of scattering-induced errors in the passive techniques (Kern et al. 2010). However, this DIAL apparatus was costly, heavy and bulky. Whilst DIAL offers unique capabilities for volcanology, it requires further innovation in order to become a suitable tool for routine observatory use. Alternative sensing strategies using near- and mid-infrared diode-based lasers (Gianfrani et al. 2000; Richter et al. 2002; Rocco et al. 2004; Kassi et al. 2006) have been evaluated by monitoring the laser beam’s absorption following numerous transits of a multi-pass cell, into which the volcanic gas sample is pumped. Thanks to very narrow spectral line-widths, lasers can make sensitive measurements of isotopic abundances. They can even resolve isotopomers that would be very difficult to discriminate by isotope ratio mass spectrometry. Richter et al. (2002) and Weidmann et al. (2003) have described a mid-infrared laser system capable of measuring all isotopes of CO2, while Gianfrani et al. (2003) have tested a diode laser spectrometer able to measure water isotopes in the near-infrared. Castrillo et al. (2004) measured 13C/12C in the field with a diode-based laser spectrometer. Christensen et al. (2007) described a tunable infrared laser spectrometer capable of measuring the isotopic composition of SO2.
Satellite remote sensing The larger releases of volcanic volatiles to the atmosphere defy synoptic measurements from the ground. This is where space-borne Earth observation methods play a preeminent role. Indeed, major advances in our understanding of explosive volcanism and its impact on the atmosphere and climate have been achieved thanks to satellite observations. Again, sulfur dioxide is the most readily measured species—its abundance in volcanic clouds has been measured using space-borne sensors operating in the ultraviolet (electronic structure), infrared (roto-vibrational structure) and microwave (rotational structure) regions of the spectrum. The earliest measurements were made by the ultraviolet-sensing Total Ozone Mapping Spectrometer (TOMS), which was operational (in various guises and with data gaps) between 1978 and 2006 (Krueger 1983; Krueger et al. 1995, 2008; Carn et al. 2003). It detected many larger silicic and intermediate composition explosive eruptions, and some mafic eruptions, notably those from Nyiragongo and Nyamuragira in the Great Lakes region of central Africa (Carn et al. 2003). TOMS along with the infrared TOVS (TIROS Optical Vertical Sounder) provided early estimates of the initial sulfur yield to the stratosphere of the 1991 Mt. Pinatubo eruption (around 9±2 Tg of S; Guo et al. 2004) and other instruments, including NASA’s Microwave Limb Sounder (MLS), were able to track the SO2 clouds as they were depleted
386
Oppenheimer, Scaillet, Martin
over the following weeks (Read et al. 1993). As discussed earlier, these observations provided crucial data on SO2 yields for comparison with estimates derived from petrological arguments. More recently-launched space-borne ultraviolet instruments have significantly improved capabilities for measuring volcanic SO2 emissions. In particular, wider wavelength coverage has lowered detection limits enabling measurements of lesser volcanic clouds. These devices include the Ozone Monitoring Instrument (OMI: Yang et al. 2007, 2009a,b; Carn et al. 2009; Krueger et al. 2009), Global Ozone Monitoring Experiment-2 (GOME-2: Heue et al. 2010) and the SCanning Imaging Absorption SpectroMeter for Atmospheric CHartographY (SCIAMACHY: Lee et al. 2009a,b). Several space-borne infrared sensors including the Infrared Atmospheric Sounding Interferometer (IASI), the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and the Atmospheric InfraRed Sounder (AIRS) can also quantify volcanic SO2 emissions (e.g., Prata and Bernardo 2007; Prata and Kerkmann 2007; Clerbaux et al. 2008; Doutriaux-Boucher and Dubuisson 2008; Eckhardt et al. 2008; Corradini et al. 2009; Campion et al. 2010) and in some cases, simultaneously, the H2SO4 aerosol abundance (Watson et al. 2004; Karagulian et al. 2010).
IntERPREtatIOn OF SulFuR-EMISSIOn Data Time-series of gas flux, chemical and isotopic measurements can be interpreted with respect to magma and volcano behavior, and the interrelationships between degassing, eruptive character, and other geophysical and geodetic parameters (Table 1). Gas geochemistry therefore plays an important role in volcanic hazard assessment. Temporal changes in gas chemistry and flux, in particular, are widely regarded as diagnostic of future volcanic activity, and many volcano observatories worldwide measure plume and fumarole emissions as part of their monitoring efforts. The basic tasks are to identify volatile sources, magma-hydrothermal system interactions, the dynamics of degassing, and changes in these through time. However, interpretation of the observations is highly challenging because of the multiple intensive parameters that control magmatic volatile content (mantle melting, slab contributions, wallrock assimilation, etc.), exsolution and gas separation of different volatile species from magma, and the subsequent chemical and physical interactions of the exsolved fluids, for example, with crustal rocks and hydrothermal fluids, as they ascend to the surface (Giggenbach 1996; Oppenheimer 2003).
Proportions of sulfur species In general terms, the relative proportions of sulfur species in a volcanic emission are dictated by temperature-, pressure- and redox-controlled partition coefficients between vapor and melt, the mixing or mingling of different magma sources, the dynamics of degassing, and interactions with crustal rocks and magmatic-hydrothermal fluids (e.g., Giggenbach 1987, 1996, 1997; Gerlach and Nordlie 1975; Gerlach 1993; Symonds et al. 1994). Pertinent redox equilibria affecting volcanic gas composition include: SO2 + 3H 2 = H 2S + 2H 2 O
(17)
SO log 2 = logK1 − 3RH − log( ftot X H2 O ) H 2S
(18)
for which:
The SO2/H2S ratio thus increases with fO2 (since −RH = log(H2O/H2), decreasing total pressure (ftot) and decreasing mole fraction of H2O (XH2O). Further, since the forward reaction of Equation (17) is endothermic, the equilibrium constant (K1) increases with temperature (Giggenbach 1987; Martin et al. 2009). Also relevant is the following equation:
Sulfur Degassing From Volcanoes
387
1 1 FeO(melt) += S2 FeS(melt/solid) + O2 2 2
(19)
This is most likely to represent a major gas redox buffer because of the comparable abundances of the two S-bearing gas species and its strong pressure dependence. The complexity of sulfur degassing in part reflects the multiple valences of sulfur and variety of sulfur species. In magmas, the key species are S2− and SO42− in the melt and H2S and SO2 in the gas phase (Baker and Moretti 2011, this volume; Webster and Botcharnikov 2011, this volume). Surface emissions may also include S2, OCS, SO3, H2SO4, SO, HS, metal sulfides and sulfates and so on, in a combination of gaseous, aqueous or other solid/liquid phases. At low temperatures where sulfur can condense, the following generalized reactions become important: SO2 + 2H 2S =
3 S x ( solid / liquid ) + 2H 2O x
SO2 + H 2S + H 2 =
2 S x ( solid / liquid ) + 2H 2 O x
(20a ) (20 b)
In crater-lake environments, polythionates (SxO62–, where x = 4-15) may also form according to activity of subaqueous fumaroles (Takano et al. 1997): H 2S + 3SO2 = S4 O6 2 – + 2H + Sx O6 2 – + HSO3 –= 0
S
Sx –1O6 2 – + S2O32 – + H +
+ HSO3 → S2 O3 + H –
(liquid)
(21a)
2–
+
(21b) (21c)
Although thermodynamical codes have been developed to restore observed analyses of gas composition to equilibrium compositions (e.g., Gerlach 1993; Symonds et al. 1994), unraveling these complex and highly non-linear processes from a patchy record of surface observations, and identifying precise magmatic and hydrothermal processes, remain real challenges (Burgisser and Scaillet 2007; Burgisser et al. 2008).
Sulfur fluxes Some of the volcanoes with the most sustained SO2 flux monitoring include the basaltic systems of Etna (Caltabiano et al. 1994; Salerno et al. 2009), Kilauea (Sutton et al. 2001, 2003) and Miyakejima (Kazahaya et al. 2004, 2008). For this long-term monitoring, the SO2 flux datasets have contributed to inferences on lava effusion rates, the fluid dynamics of magma conduits and in particular understanding of the decoupling of gas and melt on magma ascent. At Miyakejima, regular COSPEC observations reveal remarkable and ongoing decadal-scale decay in SO2 flux since a caldera-forming episode in 2000 (http://staff.aist.go.jp/kazahaya-k/ SO2average.htm). However, even a casual perusal of the literature on gas geochemical monitoring of intermediate to silicic volcanoes will reveal conflicting interpretations of ostensibly similar observations—for instance, decreasing SO2 fluxes have been ascribed to: (i) depletion of volatiles in a magma body, and (ii) decreased permeability (e.g., due to sealing of bubble networks by hydrothermal precipitation). Process (i) might be taken to indicate diminished eruption likelihood, while (ii) might act to pressurize the magma and increase the chance of eruption (e.g., Stix et al. 1993; Fischer et al. 1994; Watson et al. 2000; Edmonds et al. 2003c). Thus the same observation can be interpreted in different ways with contradictory hazard implications. Scientists at observatories often have a better feel for how to interpret changes based on expertise gained from extended monitoring a particular volcano.
388
Oppenheimer, Scaillet, Martin
Nevertheless, high SO2 fluxes remain a reliable indicator of the presence of magma during episodes of unrest at volcanoes, and help to discriminate between magmatic, and tectonic or hydrothermal causes of unrest. In particular, COSPEC measurements are generally recognized as having contributed significantly to hazard assessment prior to the 1991 eruption of Mt. Pinatubo (Daag et al. 1996; Hoff 1992). Immediately prior to the eruption the measured SO2 flux increased by an order of magnitude over two weeks, mirroring seismic unrest. These observations were interpreted as evidence of shallow intrusion of magma, suggesting increased likelihood of an impending eruption. Observations of elevated SO2 emission rates (exceeding 500 Mg d−1) at Soufrière Hills Volcano (Montserrat) also pointed to sustained magmatic unrest during a hiatus in dome growth in 1998-1999, at a time when geodetic and seismic anomalies were diminished (Oppenheimer et al. 2002a). They also provided evidence for significant permeability in the magma plumbing system to permit sulfur sourced from mafic magma at depth to reach the surface. Observations at other intermediate and silicic volcanoes have revealed decreasing post-effusive SO2 emission rates, including Augustine (Symonds et al. 1990), Redoubt (Casadevall et al. 1994) and Galeras (Zapata et al. 1997). At Mt. St. Helens, both C and S contents of magmatic gases decreased through time following the 1980 Plinian eruption (Gerlach and Casadevall 1986), in parallel with decreasing SO2 emission rate (Swanson et al. 1983; McGee 1992). These indications, along with observed decreases in eruption rate, have been interpreted as evidence for exhausted magma supply at depth. Elsewhere, measurements of SO2 flux combined with geochemical indices such as the C/S ratio in fumarole or plume gases have yielded insights into magma-hydrothermal processes, with implications for understanding and forecasting phreatic and phreatomagmatic eruptions (e.g., at Ruapheu volcano, New Zealand: Werner et al. 2008; Christenson et al. 2010). The new networks of ground-based ultraviolet spectrometers discussed above are now highlighting short-term variability in SO2 emission rates. At Soufrière Hills volcano, SO2 flux increases occur following rockfalls and during episodes of seismic tremor, consistent with release of pockets of gas from the dome and uppermost region of the conduit (Edmonds et al. 2003a). Observations made with an ultraviolet camera system enabled Mori and Burton (2009) to determine the mass of SO2 released by individual explosions at Stromboli. At Erebus volcano, measurements with a dual-field of view ultraviolet spectrometer system revealed a cyclic degassing thought to be related to episodic arrival of bubbly magma in the volcano’s lava lake (Oppenheimer et al. 2009; Boichu et al. 2010).
Sulfur isotopes Isotopic signatures can be of great value in discriminating mantle, crustal and hydrothermal contributions to volcanic fluids (Luhr and Logan 2002); in investigation of degassing processes, for example, open- vs. closed-system mechanisms (Métrich and Mandeville 2010); and in understanding the atmospheric transport of volcanic clouds (Savarino et al. 2003a,b). The extent of fractionation during degassing itself is influenced by temperature and the speciation of sulfur in melt and gas phases, and hence redox conditions (Mandeville et al. 2009; de Moor et al. 2010). The four stable isotopes of sulfur can be used to identify contributions of magmatic volatiles, assuming that isotopic compositions of surface reservoirs are constrained (See summary of Marini et al. 2011, this volume). For example, δ34SSO2 measurements of gas samples collected from the lava lake of Erta ‘Ale volcano (Ethiopia) and Asal Rift (Djibuti) indicated the mantle origin of emitted sulfur (δ34S between −5 and 1‰; Allard et al. 1977; Allard 1979, 1981), compared with high-temperature fumaroles on arc volcanoes that display more variable δ34S (0 to +10‰) due to contamination by slab and crustal sources (e.g., Ohba et al. 2008). Further, δ34S values of S in H2S and SO2 are especially sensitive to redox conditions and fH2O (Hubberten et al. 1975; Sakai et al. 1982; Marini et al. 1998, 2002; Menyailov et al. 1986), and disproportionation in crater lakes (Oppenheimer 1992; Kusakabe et al. 2000).
Sulfur Degassing From Volcanoes
389
VOlCanIC SulFuR EMISSIOn tO tHE atMOSPHERE Most estimates of the volcanic source strength of sulfur to the atmosphere are based on compilations of COSPEC and related observations of lesser emissions from individual volcanoes (many exhibiting long-term degassing), and satellite measurements of the larger, spontaneous releases of SO2 to the upper troposphere and stratosphere associated with large explosive or effusive eruptions (Tables 2-4). These data sources are patchy and go back to the 1970s only, and the statistical distribution of emissions in time and space is thus poorly constrained. At any one time, a few volcanoes appear responsible for substantial fractions of the total volcanic source strength (e.g., Bani et al. 2009). Recently, these have included Nyiragongo (Democratic Republic of Congo), Miyakejima (Japan), Ambrym (Vanuatu), Anatahan (Mariana Islands), Popocatépetl (Mexico) and Mt. Etna (Italy), the latter of which is exceptional not just within Europe but globally as one of the most prodigious sources of volcanic gases to the troposphere. Its average SO2 emission rate is of order 5000 Mg d−1 (Caltabiano 1994) and it is thought to contribute to elevated levels of tropospheric sulfate in southern Italy (Graf et al. 1998). It has been suggested that these emissions have caused pollution events in mainland Italy, and even that they have accelerated deterioration of Roman monuments (Camuffo and Enzi 1995). There is a reasonable consensus regarding the magnitude of annual volcanic source strengths of sulfur, though difficulties arise in time-averaging the sporadic but large magnitude releases to the stratosphere from explosive eruptions, and in extrapolating field data for a comparatively small number of observed tropospheric volcanic plumes to the global volcano population (see Wallace and Edmonds 2011, this volume). Historically, the most widely used global dataset is that compiled for the Global Emissions Inventory Activity (GEIA) by Andres and Kasgnoc (1998). This arrives at a global flux of S from all sources that exceeds 10.4 Tg yr−1. More recently, Halmer et al. (2002) estimated the global volcanic SO2 emission to the atmosphere as 15-21 Tg yr−1 for the period 1972-2000. Their figures for all S species add considerable uncertainty to the total volcanic sulfur flux (9-46 Tg yr−1 of S) mainly because of a very large uncertainty in the H2S emission (1.4-35 Tg yr−1 of S). For comparison, the IPCC (2001) estimates of emission rates of other sources of sulfur include anthropogenic (76 Tg yr−1), biomass burning (2.2 Tg yr−1) and dimethyl sulfide (25 Tg yr−1). Most of the volcanic source strength is from the continuous degassing of many volcanoes worldwide. Using the TOMS dataset and a scaling based on ice core records, Pyle et al. (1996) estimated the medium-term (~100 yr) flux of volcanic sulfur to the stratosphere to be about 1 Tg yr−1 (range of 0.3-3 Tg yr−1). An important point that has emerged from additional work is the disproportionate contribution of volcanic sulfur emissions to the global atmospheric sulfate budget compared with other sources of sulfur including anthropogenic and oceanic emissions. Episodic, large magnitude eruptions are the principal perturbation to stratospheric aerosol levels (e.g., the 30 Tg of sulfate generated by the 1991 Mt Pinatubo eruption). In the troposphere, the picture is less clear, but modeling suggests that up to 40% of the global tropospheric sulfate burden may be volcanogenic (Graf et al. 1997), though Stevenson et al. (2003) and Chin and Jacob (1996) obtain lower figures (14 and 18%, respectively), in part due to use of a lower volcanic sulfur source strength. In any case, these figures all exceed the fraction of the sulfur source to the atmosphere that is volcanogenic (around 10%) because of the generally higher altitudes of entrainment of volcanic sulfur compared with biogenic (DMS) or anthropogenic sources, and hence the longer residence time of volcanic SO2 compared with other sources. This results in lower deposition rates, more conversion of SO2 to sulfate, and a longer residence time of the higher altitude aerosol. Sulfate aerosol plays a significant role in the Earth’s radiation budget because it may both backscatter incoming short-wave solar radiation and absorb outgoing long-wave radiation; the competition between these processes depending strongly on particle
450 380 150
64 38
28 23 17
11.4
7.4
Ambrym, Vanuatu
Popocatépetl, México
Etna, Italy
Nyiragongo, DRC
Láscar, Chile
Bagana, PNG
Tungurahua, Ecuador
Kilauea, USA
Ulawun, PNG
(kg s−1)
SO2 flux
Miyakejima, Japan
Volcano
640
980
1460
2000
2400
3280
5500
13000
32800
38880
(Mg d−1)
0.23
0.36
0.54
0.73
0.88
1.20
2.02
4.73
12.0
14.2
(tg a−1)
Equivalent SO2 flux
2003; arc basalt, frequent eruptions
Average for combined output from East Rift and Summit for 1979-1997; ocean island basalt
2004-2006; andesite composition
2003; andesite; active since 1972
McGonigle et al. 2004b
Sutton et al. 2001
Arellano et al. 2008
McGonigle et al. 2004b
Mather et al. 2004a or b?
Sawyer et al. 2008 Carn 2004
Lava lake degassing in 2005 (peaks up to 185 kg s−1); melilitite / nephelinite lava lake 2003; andesite lava dome / open vent
Caltabiano et al. 1994
Delgado-Granados et al. 2001; Roberge et al. 2009
Bani et al. 2009
Nakada et al. 2004
References
1987-1991 average; alkali basalt
Following explosive eruption in 1996; 9 Tg emitted between 1995 and 1997 (lava domes & Vulcanian eruptions); basaltic degassing, andesite lavas and tephra
Peak measured in 2005; variable emissions; arc basalt
Average for late 2000-mid 2001; arc basalt
Remarks
table 2. Examples of large SO2 emitters via non-explosive (passive) degassing during the last 10-15 year.s
390 Oppenheimer, Scaillet, Martin
15 June 1991
28 March-4 April, 1982
26 February 1981-11 January
Pinatubo, Phillippines
El Chichón, Mexico
Nyamuragira, DRC
8 May-13 June 2004
17-31 October 1998
25 July-2 August 2002
24 April-18 May 1989
5 April -19 September 1982
6 February-7 March 2001
23 October 2005
7-8 August 2008
8-15 August 1991
5-16 July 1994
13 November 1979
April 27, 1981
August 28, 1982
March 25, 1984
24 February-4 March 1984
Nyamuragira, DRC
Nyamuragira, DRC
Nyamuragira, DRC
Nyamuragira, DRC
Galunggung, Java
Nyamuragira, DRC
Sierra Negra, Galápagos
Kasatochi, Alaska
Cerro Hudson, Chile
Nyamuragira, DRC
Sierra Negra, Galápagos
Alaid, Kuril islands
Wolf Galápagos
Mauna Loa, Hawaii
Nyamuragira, DRC
1982
Eruption date(s)
Volcano
1
1
1.1
1.1
1.2
1.47
1.5
1.7
1.7
1.73
1.73
2.1
2.32
2.54
2.6
4.11
7
18
SO2 yield (tg)
Phonolitic tephrite lavas
Lava flows; tholeiitic basalt
Fire fountains and lava flows; tholeiitic basalt
Explosive basaltic eruption
Lava flows; tholeiitic basalt
Phonolitic tephrite lavas
Andesitic Plinian eruption
Explosive andesitic eruption
Lava fountains and flows; tholeiitic basalt
Phonolitic tephrite lavas
Multiple eruptions; beginning with andesites ending up with high magnesium basalts
Phonolitic tephrite lavas
Phonolitic tephrite lavas
Phonolitic tephrite lavas
Phonolitic tephrite lavas
phohnolitic tephrite lavas
14 eruptions; 24.5 Tg emitted in total from 1979-2005;
Trachyandesite eruption
20.2 Tg total for 12-15 June 1991; dacite eruption
Remarks
table 3. Large eruptive emissions of SO2 retrieved from satellite remote sensing observations.
Bluth and Carn 2008
Sharma et al. 2004
Carn et al. 2003
Bluth et al. 1997
Bluth et al. 1997
Bluth and Carn 2008
Krueger et al. 1995
Karagulian et al. 2010
Yang et al. 2009a or b?
Bluth and Carn 2008
Bluth et al. 1994
Bluth and Carn 2008
Bluth and Carn 2008
Bluth and Carn 2008
Bluth and Carn 2008
Bluth and Carn 2008
Bluth et al. 1997
Guo et al. 2004
References
Sulfur Degassing From Volcanoes 391
Oppenheimer, Scaillet, Martin
392
table 4. Estimated volcanic emissions of sulfur species to the atmosphere (all figures are for sulfur mass only). Source
Sulfur as SO2 (tg a−1)
Other S species
Pyle et al. 1996
0.45-1.5 stratospheric
Andres and Kasgnoc 1998
1.9 stratospheric 4.84 tropospheric 6.7 total
3.7 (OCS, H2S, H2SO4)
Halmer et al. 2002
1.2-2 stratospheric 7.5-10.5 total
S as H2S 0.09-4.7 stratospheric S as H2S 1.4-35 total
size. In addition, sulfate aerosol can have a secondary radiative effect by promoting cloud condensation or modification of the microphysical properties and longevity of existing clouds (Graf et al. 1997; Gassó 2008). Changes in time and space in this “background” emission could represent an important forcing factor that has yet to be characterized. A more recent database on volcanic SO2 emissions compiled by Diehl (2009) is based on the Global Volcanism Program data set on volcanic eruptions (http://www.volcano.si.edu/) and prior work of Berresheim and Jaescke (1983), Andres and Kasgnoc (1998), Halmer et al. (2002), and Blake (2003) covers the period 1979-2009.
Ice cores One of the most valuable archives of past atmospheric and climatic conditions on Earth is found in the polar regions and some lower latitude ice caps and glaciers in the Central Andes (e.g., Kellerhals et al. 2010), the Himalayas and Tibet. Ice cores contain a deposition signal (representing the fallout of atmospheric gases and particles to the surface) and a postdepositional signal (compaction, chemical diffusion, wind deflation and re-deposition). Oxygen isotopes and the calcium ion record are essential indicators of climatic variability, while peaks in sulfate ions (SO42−) and in electrical conductivity of the ice indicate volcanic aerosol fallout. The pioneering work on the volcanic acid layers preserved in ice cores was carried out by Claus Hammer and colleagues (e.g., Hammer 1977; Hammer et al. 1980). Glaciochemical stratigraphy thereby reveals a vast amount of information on past volcanism (Fig. 10). One of the striking revelations of the ice core record is the evidence for numerous great eruptions, which have not otherwise been recognized in tephra records. One caveat to the approach is that although the dating of the ice core by counting of seasonal layers is fairly robust, it is not fail-safe. The greater the depth from which the core is retrieved, the more likely it is to have suffered deformation (e.g., Baillie 2010). One of the ice cores most closely studied to date for its volcanic record is the Greenland Ice Sheet Project 2 (GISP2) core. Initial efforts took the record back to 7000 BCE with the core measured at biennial resolution (Zielinski et al. 1994). All ice core records of volcanism are affected to greater or lesser extent by the proximity of the core sites to active volcanoes. While some large explosive eruptions in the tropics have sourced fallout identified in cores from both polar regions (Gao et al. 2007, 2008), records are generally biased. For instance, the Greenland records provide better records of eruptions in Iceland, Alaska, the Aleutians and Kamchatka. Particles from high-northern latitude eruptions tend to reach the Arctic ice sheets within months of eruption, while acid fallout from tropical eruptions peaks up to 1-2 years later depending on prevailing circulation of the upper atmosphere. Fallout from eruptions south of the tropics is unlikely to make it to the Arctic (and similarly, fallout from boreal eruptions is unlikely to reach the Antarctic). The GISP2 team later extended the ice core record of volcanism back to 110 ka ago (Zielinski et al. 1996). More recently established, the
Sulfur Degassing From Volcanoes
393
Pinatubo 1991
Cosigüina 1835
10
Krakatau 1883 Katmai 1912
20
Laki 1783 Mystery eruption 1809 & Tambora 1815
30
Huaynaputina 1600
40
?Kuwae »1452
50
Mystery eruption»1257/1258
60
Mystery eruption 529 & 541
Total sulphur yield to the global atmosphere (megatonnes)
70
0
Year (CE)
Figure 10. Plot of GISP2 volcanic sulfate markers for the past two millennia based on statistical analysis (Zielinski et al. 1994). Several large anomalies have not been traced to the responsible volcanoes, including the prominent AD 640 and 1259 peaks. Data provided by the National Snow and Ice Data Center, University of Colorado at Boulder, and the WDC-A for Paleoclimatology, NGDC, Boulder, CO, USA.
European Project for Ice Coring in Antarctica (EPICA) record goes back to around 800 ka ago (EPICA 2004) though the portion suitable for reliable identification of volcanic sulfate may cover only the last 400 ka (Traversi et al. 2009). An important application of the ice core record is in providing estimates of the sulfur output of large eruptions (Table 5). The most widely used calibration for sulfate concentration in an ice core is based on measurements of radioactivity due to fallout from atmospheric nuclear weapons tests carried out in the 1950s and 1960s. Because the yield of these various bombs is known, a correlation can be established between the quantities of radionuclides released into the atmosphere and the concentration of radioactive fallout measured in the ice. This same correlation can then be applied to the volcanic markers, assuming that the volcanic sulfate was transported by comparable atmospheric circulation patterns, from similar latitude, and so on, which is difficult to gauge for ancient eruptions (as mentioned above, the responsible volcano is often unidentified). Very low snow deposition rates and post-depositional processes that affect the ice, such as densification, diffusion, and wind deflation or re-deposition, and the contributions from sources of sulfur (including the oceans), also complicate interpretation of ice core sulfate records. Nevertheless, multiple estimates of stratospheric aerosol loading of eruptions are reasonably convergent, and are consistent with estimates obtained by other methods (e.g., based on the petrological methods discussed earlier), providing some confidence in the approach (Fig. 2).
Oppenheimer, Scaillet, Martin
394
table 5. Estimated sulfur yield from selected major historic and prehistoric eruptions (magnitude > 1013 kg). Eruption year and volcano
Magnitude (kg)a
Sulfur yield (tg of S)b
Reference
~ 73 ka BP Toba, Indonesia
7×1015
35-3300
Summarized in Oppenheimer 2002; Chesner and Luhr 2010
~230 CE Taupo, New Zealand
7.7×1013
?6.5
~935 CE Baitoushan (aka Chanbaishan, Pektusan), China / North Korea
5.8×10
13
b
Zielinksi et al. 1996
>2
Horn and Schmincke 2000; Nakamura et al. 2007
1257/8 CE, Unknown
?
>100
Oppenheimer 2003
1452 cKuwae, Vanuatu
7.5-15×1013
>6.5-13
Scaillet et al. 2003; Witter and Self 2007
1600 Huaynaputina, Peru
2.1×1013
26-55
Costa et al. 2003
1815 Tambora, Indonesia
7.4×1013
28
Self et al. 2004
1883 Krakatau, Indonesia
3.0×1013
15
b
Zielinksi et al. 1996
13
d
11
b
Zielinksi et al. 1996
2.5×1013
10
b
Zielinksi et al. 1996
1902 Santa Maria, Guatemala 1912 Katmai, Alaska 1991 Pinatubo, Philippines
2.2×10
1.3-1.8×10
13
9
Guo et al. 2004
a Total eruption magnitude for multiple phases of eruption and combining tephra deposits from both Plinian and coignimbrite clouds and associated pyroclastic density current deposits where applicable; bStratospheric sulfur yield based on ice core sulfate deposition; cbut see Németh et al. (2007) for an alternative viewpoint on the Kuwae caldera; d since this estimate is based on ice core sulfate deposition, it may reflect the cumulative aerosol fallout of other notable 1902 eruptions, i.e., Mont Pelée (Martinique) and Soufrière (St. Vincent) as well as the Santa Maria event.
atMOSPHERIC anD ClIMatIC IMPaCtS OF SulFuR DEGaSSInG There is considerable interest in volcanic emissions of sulfur compounds because of their role in atmospheric radiation and climate, the hydrological cycle, acid precipitation, and air quality. The main protagonist in volcanic forcing of climate is the sulfate aerosol formed by oxidation of sulfur gases released to the upper atmosphere. This section focuses on the impacts of major, sulfur-rich explosive eruptions on the stratosphere and climate, and of passive degassing on tropospheric chemistry and ecosystems.
Chemical schemes relevant to volcanic sulfur emissions Volcanic plumes obviously affect the composition of the atmosphere they pass through. Similarly, the low-temperature, oxidizing environment of the atmosphere results in profound changes in the brew of volcanic gases that leave the vent. As a result, the atmospheric chemistry of volcanic plumes is complex. The mixture of ash particles, liquid droplets, soluble salts, ice crystals, and volcanic gases can absorb and scatter solar radiation, resulting in temperature changes. Since the rates of many chemical reactions in the atmosphere are temperaturedependent, and as temperature controls the phase—solid, liquid, gas—of atmospheric constituents, these effects influence atmospheric composition. Changes in shortwave radiation also affect photochemical processes contingent on levels of ultraviolet and visible light. In the stratosphere, the most significant chemical role of volcanic aerosols is in providing surfaces on which atmospheric gases can react (heterogeneous reactions). Multi-phase reactions involving diffusion into a liquid droplet and reaction within the liquid are also important.
Sulfur Degassing From Volcanoes
395
High-temperature chemistry at the vent. The presence of highly oxidized species in volcanic emissions—near-source SO42− particles (Allen et al. 2002; Mather et al. 2003a,b), ClO and OClO (Bobrowski et al. 2007), BrO (Bobrowski et al. 2003; Oppenheimer et al. 2006), NOy (Mather et al. 2004a; Oppenheimer et al. 2005, 2010)—is not predicted by equilibrium models for high-temperature magmatic gases (e.g., Symonds and Reed 1993). Even relatively oxidized magmatic gases do not thermodynamically support the formation of these species. Instead, magmatic gases may be partially oxidized at the vent by atmospheric gases. A simple thermal model of binary mixing (Fig. 11; heat capacities are assumed equal) between magmatic gas (e.g., volume = VM, T = 1273 K, fO2 = 10−8 bar) and atmospheric gas (e.g., volume = VA, T = 273 K, fO2 = 0.21 bar) shows that tremendous increases in fO2 may occur with minimal change in the temperature. Such a thermal model predicts that highly oxidized, high-temperature mixtures will form at vents. However, a limitation is that chemistry is not considered and reduced species in the magmatic gas (e.g., H2S, H2, CO) would buffer the oxidation state of mixture to some extent. A kinetic model would be desirable to understand better the chemistry occurring at the vent, but relevant mechanisms and reaction rates remain unknown despite the development of kinetic models in other high-temperature systems (e.g., combustion and chemical synthesis). A simplified approach to investigate vent chemistry relies on the fact that reaction rates increase with increasing temperature, promoting the rapid attainment of equilibrium. With the assumption that full equilibrium is attained and maintained within the high-temperature mixture of magmatic and atmospheric gases, modeling with an averaged magmatic gas composition (Gerlach 2004b; Martin et al. 2006) predicts significant changes in fO2 and sulfur speciation as mixing occurs (Fig. 11). The main features are losses of reduced S species (H2S, S2, OCS) and gains in oxidized S species (SO3, H2SO4) around a compositional discontinuity (i.e., a step increase in fO2; Gerlach and Nordlie 1975). Equivalent trends have been shown for other elements, e.g., C, N, Br, Cl. The location of the compositional discontinuity can be calculated for a magmatic gas composition simply by the amount of O2 required to oxidize all H2S and H2 (Martin et al. 2006), implying that more oxidized magmatic gases with lower H2S and H2 are more likely to the support the formation of oxidized species. There are, however, indications from field measurements that full equilibrium may not be attained at the vent since the model predicts mutual exclusivity of H2S and oxidized species, which is inconsistent, for
1300
-1
-1
-2
-2
-3
-3 -5
(with stable H2S)
-6
equilibrium model (full)
1100 0
0.05 VA/VM
0.1
SO2 H2S
-4 log (fX)
-4
equilibrium model
1200
log (fO2)
Temperature (K)
thermal model
0
-5
-6
-7
-7
-8
-8
-9
-9
SO3
S2 OCS
H2SO4
0
0.05
0.1
VA/VM
Figure 11. Plots of (a) thermal and equilibrium model calculations for temperature and fO2 in mixtures forming at the volcanic vent as function of the volume fraction of mixing, VA/VM. (b) S speciation assuming full thermodynamic equilibrium.
396
Oppenheimer, Scaillet, Martin
instance, with the presence of both H2S (Aiuppa et al. 2005b) and BrO (Oppenheimer et al. 2006) in Mt. Etna’s plume. Further, the measurements of the atmospheric plume emitted from Mt. Etna reveal a SO2/H2S ratio consistent with the predicted magmatic gas phase (Aiuppa et al. 2005b) and so H2S may not always react rapidly at high temperature (see also discussion on tropospheric chemistry below). Kinetic stability of H2S at high temperature would lower the compositional discontinuity, increase fO2 and increase the abundance of oxidized species in the mixture since atmospheric O2 is no longer consumed by reaction with H2S (Fig. 11; Martin et al. 2009). This problem highlights the key importance of H2S/SO2 measurements, and measurements of other redox pairs (e.g., H2/H2O, CO/CO2), in investigating the chemistry occurring at volcanic vents. Equilibrium models show that the highly oxidized, high-temperature conditions of the vent play an important role in converting relatively unreactive species (e.g., H2S, HCl, HBr, N2) into more reactive species, such as SO3, Cl, Br and NO (Fig. 11; Mather 2008). However, the most abundant “reactive” species at the vent are generally different from those measured in the plume (e.g., SO42−, ClO, OClO, BrO, NO2, HNO3), so further chemical processing must occur beyond the vent. Sulfate ions may form simply by reaction of hygroscopic SO3 with water, as indicated by isotopic studies of near-source SO42− at Masaya volcano (Mather et al. 2006) that show δ18O is close the magmatic values. Additionally, SO42−/SO2 ratios in near-source volcanic emissions are typically on the order of 10−2 (Allen et al. 2002; Mather et al. 2004b; Martin et al. 2008, 2010b) and so in good agreement with the results from equilibrium modeling. Nevertheless, much remains to be done to reconcile observations and models and advance our understanding of the first seconds of the chemical and physical evolution of magmatic volatiles on exposure to the atmosphere. Tropospheric chemistry. The major fates of tropospheric SO2 are dry and wet deposition, and oxidation to sulfate. In the gas phase, hydroxyl rapidly reacts with SO2 by addition: OH + SO2 + M → HOSO2 + M
(22)
where M represents a molecule (usually N2) that absorbs excess kinetic energy from the reactants (Hewitt 2001). The reaction rate is second order in the troposphere. In unpolluted air, OH is produced by the photolysis of ozone and the subsequent reaction of oxygen atoms with water vapor. In polluted air, the photolysis of nitrous acid (HONO) and hydrogen peroxide (H2O2) yields OH directly (Hewitt 2001). Once formed, the free radical HOSO2 reacts rapidly with oxygen to form SO3, which in turn reacts rapidly with water vapor to form sulfuric acid, H2SO4: HOSO2 + O2 → HO2 + SO3
(23)
SO3 + H 2O → H 2SO 4
(24)
Möller (1980) calculated mean tropospheric e-folding times of 12 d for homogenous gasphase oxidation by OH radicals (i.e., the time taken for SO2 abundance to decay exponentially to 1/2.718 of its initial level). However, much faster oxidation can occur through heterogeneous reactions. Sulfur dioxide can dissolve in aqueous aerosol, establishing an equilibrium with its ionic products (SO2·H2O, bisulfite ions (HSO3−) and sulfite ions (SO32−): SO2 + H2O = SO2·H2O SO2·H2O = HSO3− + H+ HSO3− = SO32− + H+
(25) (26) (27)
The pH of the aerosol controls the equilibrium and hence speciation. The oxidation of S(IV) to S(VI) typically involves ozone or hydrogen peroxide (e.g., Martin and Damschen 1981): S(IV) + O3 → S(VI) + O2 S(IV) + H2O2 → S(VI) + H2O
(28) (29)
Sulfur Degassing From Volcanoes
397
The reaction of S(IV) with ozone is rapid in the aqueous phase but is more pH dependent (and less effective for pH < 4) than the reaction with H2O2 (Eatough et al. 1994; Hewitt 2001). S(IV) can also be oxidized by metal catalysts such as Mn2+ and Fe2+, which could have significant abundance in some volcanic aerosol. Möller (1980) calculated mean tropospheric e-folding times of just 0.1-10 d for oxidation on the surfaces of particles, and < 6 h for oxidation within aqueous particles by H2O2 and O3. The short e-folding time of tropospheric SO2 does not itself pose a problem for traverse or scanning measurements of SO2 fluxes since these are generally measured close to source (plume age < 1 h). However, as discussed above, recent studies have highlighted the oxidizing nature of volcanic plumes compared to the background troposphere (von Glasow et al. 2009). In the specific case of SO2, the observation is that measured fluxes may decrease significantly over relatively small distances, suggesting e-folding times much shorter than 6 h. The shortest reported measured e-folding time is 17 min at Soufrière Hills volcano on Montserrat (Oppenheimer et al. 1998b) and a more recent measurement reported an e-folding time of 3 h (Rodriguez et al. 2008). Measurements elsewhere include Mt. St. Helens (>100 d; Martin et al. 1986), Masaya (28 h; McGonigle et al. 2004a) and Mt. Erebus (~1 d; Oppenheimer et al. 2010). Rodriguez et al. (2008) suggest that the apparently shorter e-folding times at Soufrière Hills volcano relate to the humid conditions promoting SO2 oxidation in aqueous particles; if true, this may suggest a systematic underestimation of SO2 fluxes from other volcanoes in comparable climates. The presence of volcanic sulfate (both near-source and formed by low temperature SO2 oxidation) has a number of implications for tropospheric chemistry. Volcanic sulfate aerosols act as cloud condensation nuclei (e.g., Mather et al. 2003a), which enable condensation of water vapor to form clouds. Similarly, the volcanic sulfate aerosols may lead to physical changes in existing clouds by the introduction of additional cloud condensation nuclei. The first satellite observations of this phenomenon showed decreased particle size and increased particle number in impacted clouds due to increased competition of water vapor (Gassó 2008). Additionally, the presence of sulfate would cause acidification of existing aerosol (e.g., von Glasow et al. 2009), potentially leading to increased rates of acid-catalyzed aqueous reactions and the revolatilization of acidic gases (e.g., HCl, HBr, HF, HNO3). The major fate of tropospheric H2S is H-abstraction by OH radicals (with a mean e-folding time of 12 h; Jaeschke et al. 1980). While OH radicals are formed through high-temperature chemistry at the vent (e.g., Gerlach 2004b), kinetic model studies at low temperature (Aiuppa et al. 2007b) have shown that the OH concentrations would be too low to result in significant losses of H2S over short timescales (i.e., ~1 h). The analogous reactions between H2S and halogen radicals (Cl, Br) are predicted to be much faster when high-temperature vent chemistry is considered, with H2S being removed by Cl after a few seconds (Aiuppa et al. 2007b). These results are inconsistent with measurements showing a stable H2S/SO2 ratio in Mt. Etna’s plume, consistent with magmatic degassing and no overprinting by subsequent kinetic processing. Possible explanations for this mismatch are given as either that Cl production is exaggerated by the model for high-temperature vent chemistry or a missing loss process for Cl. The latter possibility was not thought likely since high Cl concentrations were required to explain observation of high ClO by differential optical absorption spectroscopy (Bobrowski et al. 2007). However, it now appears that these early ClO measurements are unreliable (Kern et al. 2009) so a missing loss process for Cl remains an important possibility. As previously indicated, the tropospheric chemistry of volcanic S does not occur in isolation from other volcanic emissions. Much of the oxidizing nature of volcanic plumes may stem from the high concentrations of sulfate aerosol, not as a reactant in the strict sense, but as a medium for other reactions to take place within and on the aerosol. The importance of the aerosol has been emphasized in kinetic models (e.g., Roberts et al. 2009; von Glasow 2010) for Br, Cl and NO oxidation. Additionally, recent work by von Glasow (2010) highlights the
398
Oppenheimer, Scaillet, Martin
potential role of SO2 in Hg speciation in volcanic emissions. This is an issue of key importance because while Hg(0) is an unreactive insoluble gas that is transported globally, Hg(II) occurs as reactive gases and particles that deposit rapidly and proximally. This may result in elevated and potentially harmful levels of Hg deposition around active volcanoes. The key reactions are as follows (where X = Br, Cl): Hg + 2X → HgX2 HgX2 + SO2 → Hg + X2 + SO2
(30) (31)
High-temperature models of vent chemistry (Bagnato et al. 2007) predict that Hg(II) comprises 1% of total Hg, while kinetic modeling (von Glasow 2010) shows that the amount of Hg(II) would increase due to oxidation by halogen radicals (Br, Cl). However, at high concentrations of SO2, any HgX2 (i.e., Hg(II)) formed would be rapidly converted back to Hg. While the prospect of local Hg deposition depends on a series of uncertain rate constants, it is clearly that case that S chemistry could play a critical role in determining Hg speciation after emission.
Impacts of tropospheric sulfur emissions from volcanoes From local to regional scales, volcanic emissions can result in significant environmental and environmental health consequences, including destruction of agricultural crops, contamination of pasture, and human respiratory morbidity and cardiovascular mortality. Large lava eruptions, such as that of Laki (Iceland) in 1783-1784, which released an estimated 120 Tg of SO2, have resulted in major pollution episodes responsible for regional-scale extreme weather, agricultural losses, and elevated human morbidity and mortality (Thordarson et al. 1996; Witham and Oppenheimer 2005; Oppenheimer 2011). Thanks largely to the work following the 1991 eruption of Mt. Pinatubo, there is now a good understanding of the stratospheric impacts of volcanic eruption clouds, at least for emissions on this scale. In contrast, the impacts of tropospheric volcanic plumes are rather poorly known (Table 2). This partly stems from a wide spectrum of emissions and emission styles (e.g., continuous degassing, minor eruptions), rapid transport of some components to the Earth’s surface, and the greater concentration and variability of H2O in the troposphere. Reactive S, Cl and F compounds may be present in both the gas and particle phase in volcanic plumes, and are typically co-emitted with many other volatile species, including water vapor, and with silicate ash particles. Once in the troposphere, physical and chemical processes convert gaseous SO2 to sulfate, while gaseous HF and HCl establish equilibria with aqueous phase H+, F− and Cl−. Increasing interest in the impacts of tropospheric volcanic emissions on terrestrial and aquatic ecology and on human and animal health is driving more research into the tropospheric chemistry and transport of volcanic plumes. This will be of particular importance in understanding the long-range pollution impacts of volcanic clouds (that might be expected from future eruptions of the scale as that of Laki in 1783-1784). Impacts on agriculture, ecosystems and air quality. Volcanic volatiles emitted into the atmosphere are ultimately wet or dry deposited at the Earth’s surface. The chemical and physical form in which they are deposited, and the spatial and temporal distribution of deposition, are strongly controlled by atmospheric chemistry and transport of the volcanic plume. Various components of volcanic emissions (including acid species and heavy metals; e.g., Martin et al. 2010a) can be taken up by plants, and can have both harmful and beneficial effects. The detrimental effects are generally either mediated through acidification of soils (by dry or wet deposition) or by direct fumigation of foliage (e.g., respiration of acid gases through stomata). Individual passively degassing volcanoes can also represent major pollution sources. For example, Mt. Etna continuously emits of order 70 kg s−1 SO2, and substantially elevates tropospheric sulfate in southern Italy (Graf et al. 1998). Very fine (sub-micron) and highly acidic (pH about 1) aerosol is also emitted from open-vent volcanoes such as Mt. Etna and
Sulfur Degassing From Volcanoes
399
Masaya (Allen et al. 2002, 2006; Mather et al. 2003b). Adverse environmental and health impacts are observed downwind of many degassing volcanoes, including Masaya (Nicaragua), Kilauea (Hawai`i), Poás (Costa Rica), Miyakejima (Japan) and Popocatépetl (Mexico) (e.g., Baxter et al. 1982; Mannino et al. 1996; Delmelle et al. 2003; Fujita et al. 2003). Downwind of Masaya volcano, SO2 levels exceed background levels at the surface across an area of 1250 km2, and an estimated 50,000 people are exposed to concentrations of SO2 exceeding WHO air quality standards (125 ppbv over 24 h; 50 ppb over 1 year; Baxter et al. 1982; Baxter 2000; Delmelle et al. 2001, 2002). Symptoms of respiratory illness are commonly reported anecdotally, though formal epidemiological studies have yet to be undertaken. Measurements in Mexico City have indicated the direct influence of Popocatépetl volcano’s emissions on urban air quality (e.g., Raga et al. 1999; de Foy et al. 2009). This hints at the potential impacts of future eruptions on the scale of the Laki 1783-1784 episode, with long-range transport of SO2 and sulfate aerosol reaching European cities, where air quality standards are already surpassed at certain times of the year. Air quality in Hawai’i is affected by “vog” (volcanic fog) associated with SO2 and sulfate aerosol from Kilauea’s plume (Sutton and Elias 1993; Longo and Yang 2008; Longo et al. 2008, 2010; Longo 2009). However, a study of healthy residents on the Big Island of Hawai’i found no appreciable effects of the “vog” on the autonomic nervous system (Chow et al. 2010). A study on Réunion revealed elevated exposure to SO2 in coastal regions during eruptive episodes but limited clinical evidence for respiratory consequences of the volcanic emissions (Viane et al. 2009; Bhugwant et al. 2009). Chemical burning of leaves and flowers of vegetation downwind of the crater is a common observation during degassing crises of Masaya volcano, with substantial economic impact from the loss of coffee crops (Delmelle et al. 2001, 2003). A further common response to increased gas concentrations is defoliation, as was recently observed close to the April 2007 eruption on Réunion (Martin et al. 2010a). More minor responses to exposure to SO2 including changes in leaf permeability (Percy and Baker 1988) and stomatal conductance (Smith 1990), which affect the uptake of aerosol and gas into the leaf. SO2 can cause direct damage to plants once it is taken up by the foliage (Smith 1990). The response, however, can be very variable, depending on dosage, atmospheric conditions, leaf type and pre-existing plant health (e.g., Linzon et al. 1979; Winner and Mooney 1980; Smith 1990). In general, chronic exposures to SO2 concentrations of a few 10s or 100s ppb are sufficient to affect plant ecosystems, decrease agricultural productivity and cause visible foliar chlorosis and necrosis (Winner and Mooney 1980). The impacts of sulfate deposition on soils have been investigated widely in the context of anthropogenic pollution, indicating that the SO42− retention capacity of soils varies widely and can be expected to dictate the ecosystem disturbance of volcanogenic sulfur deposition. Anion sorption in soils is directly related to the soil mineralogy and hence, to the soil parent material. Delmelle et al. (2001) estimated that the amount of SO2 and HCl dry-deposited within 44 km of Masaya volcano generates an equivalent H+ flux ranging from <1 to 30 mg m−2 day−1. Sustained acid loading at these rates can severely impact soil chemistry, reflected downwind of Masaya in low pH and depressed base-saturation contents of soils (Parnell 1986) as metals are displaced from the soil complex by H+.
the atmospheric and climatic impact of the 1991 eruption of Mt. Pinatubo The principal source of sulfur to the stratosphere is the episodic injection of SO2 (and/or H2S) from major eruptions. Once in the stratosphere, SO2 is oxidized to sulfuric acid at rates determined by the availability of OH radicals via the reaction scheme outlined in Equations (22-24). The sulfuric acid vapor spontaneously nucleates particles or condenses on existing aerosol (modifying the size distribution of existing aerosol. We understand these processes as well as we do largely thanks to studies of the aftermath of a single eruption—that of Pinatubo (Republic of the Philippines) in 1991. The coincident emission of sulfur represented the largest
400
Oppenheimer, Scaillet, Martin
perturbation to stratospheric aerosol abundance in decades and resulted in unambiguous atmospheric and climatic change. This section reviews what we have learned from the studies of the volcanic cloud and Earth system responses. Stratospheric SO2 cloud and sulfate formation. The first clear signs of unrest at Mt. Pinatubo were observed on 2 April 1991, when steam explosions occurred near the summit. Pinatubo had not previously been recognized as a potentially active volcano. On the afternoon of 15 June 1991, and following a crescendo in activity (Fig. 12), the volcano disgorged around 10 km3 of ash and pumice. Its magnitude, Me, of 6.1 (based on the expression Me = log10(mass of ejecta (in kg)) − 7) makes it the largest eruption in a century (since that of Katmai in Alaska in 1912). The stratosphere was reached both by the Plinian eruption column, and by co-ignimbrite clouds rising from pyroclastic currents. Due to the vast scale of Pinatubo’s eruption plume, satellite observations were central in tracking its growth and dispersal around the globe. By 16:40 local time, the umbrella cloud forming in the stratosphere had attained a diameter of 500 km. It covered an area of 300,000 km2 at 19:40 hours, and reached a maximum diameter of over 1100 km (Fig. 13). The umbrella cloud was 10-15 km thick and its top reached 35 km above sea level. Thirty-six hours after the start of the eruption, the cloud covered an area of 2.7×106 km2. By this stage the plume was travelling west-southwest with the ash concentrated at an altitude of 16-18 km, and fallout occurring across the South China Sea.
Figure 12. Photograph of eighteen-km-high eruption plume of Mt. Pinatubo on 12 June 1991, three days before the climactic eruption. Photographed from Clark Air Base (20 km east). Credit: Rick Hobblitt, US Geological Survey Cascades Volcano Observatory.
(b)
(d)
(a)
(c)
Figure 13. Images of umbrella cloud of the 15-16 June, 1991 Mt. Pinatubo eruption, Philippines, as seen by the GMS weather satellite. The cross marks the volcano’s location. (a) 13:31 local time, and the beginning of the climactic phase. The diameter of the umbrella cloud is about 90 km. Ash is being transported southwest. The white cloud is associated with Typhoon Yunya; (b) 14:31, the umbrella cloud is expanding so fast it is propagating upwind as well as downwind at up to 70 km h−1; (c) 15:31, ash is being dispersed by winds to the southwest beneath the umbrella cloud; (d) 16:31, the umbrella cloud has reached its maximum diameter of over 1000 km. Images processed by R.E. Holasek (Holasek et al. 1996).
Sulfur Degassing From Volcanoes 401
402
Oppenheimer, Scaillet, Martin
The best estimate yet published of the release of SO2 to the stratosphere is around 18 Tg (or 9 Tg of S; Guo et al. 2004), which generated around 30 Tg of sulfate aerosol. Ground-based lidar observations established that the bulk of Pinatubo’s aerosol veil was at an altitude of 20-27 km. Over the following months, the cloud expanded into both hemispheres though remaining trapped within a band between latitudes 30°N and 20°S through mid-1991. Strengthening of the Brewer-Dobson circulation eventually helped to mix the volcanic aerosol to higher latitudes. Stratospheric ozone depletion. Sulfate aerosol promotes numerous heterogeneous and multi-phase reactions that act in such a way as to shift stratospheric chlorine from stable compounds (HCl and ClONO2) into more reactive ones (i.e., hypochlorous acid, HOCl) that can destroy ozone. In theory, any increase in stratospheric HOx, NOx, ClOx or SOx could deplete stratospheric ozone levels. A few months after the Pinatubo eruption, global stratospheric ozone levels began to show a strong downturn. The clearest picture of the impacts was provided by the space-borne TOMS instrument (see section on satellite remote sensing above). Ozone levels, integrated through the depth of the atmosphere, decreased 6-8% in the tropics in the first months after the eruption. These figures actually mask local depletions of up to 20% at altitudes of 24-25 km. By mid-1992, ozone abundance in the stratosphere was lower than at any time in the preceding 12 years, reaching a low point in April 1993 when the global deficit was about 6% compared with the average. Losses were greatest in the northern hemisphere: total column ozone above the USA dropped 10% below average with the strongest depletion observed between 13 and 33 km altitude. Radiative and climatic impacts. The effects of stratospheric aerosol veils on electromagnetic radiation are highly complex (Fig. 14) because the haze can consist of variable proportions of very fine ash and sulfate aerosol of different compositions, sizes and shapes (and hence optical properties). The different components also accumulate and sediment out at different rates according to their masses and shapes, so that any effects on the Earth’s heat budget can be expected to change through time. Once most of the gaseous sulfur has been
Figure 14. Schematic diagram of the radiative effects of volcanic aerosol and some associated processes. The background photograph was taken from the Space Shuttle (mission STS 43) above central Africa and shows a double layer of aerosol generated after the 1991 eruption of Mt. Pinatubo. The layers are at heights of approximately 20 and 25 km above sea level. Photograph from the Image Science and Analysis Laboratory, NASA-JSC.
Sulfur Degassing From Volcanoes
403
converted, little more aerosol is formed and the total stratospheric aerosol load decreases as the particles subside into the troposphere, from which they are rapidly transferred to the surface by rainfall and other depositional processes. As the total aerosol burden decreases, the particle size distribution changes since larger particles have greater terminal velocities. As a result, the mean radius of the aerosol veil decreases over time. The particles scatter incoming shortwave radiation but can also absorb both upwelling and downwelling radiation. The net effect on the Earth’s radiation budget is not straightforward to determine but, broadly, for warming to outbalance cooling, the aerosol’s effective radius (a measure weighted by the particles’ surface area) should exceed 2 μm. Prior to Pinatubo, the effective radius of stratospheric aerosol was about a tenth of this value. Pinatubo’s contribution elevated the effective radius of stratospheric aerosol to about 0.5 μm. Measures of the effects of the aerosol veil were obtained from the spaceborne Earth Radiation Budget Experiment (ERBE). By August 1991, ERBE revealed that the backscattering of solar radiation by the aerosol had increased the global albedo by around 5% (Minnis et al. 1993; Wong et al. 2006). This corresponds to a deficit in direct sunlight of around 25-30%, which was partly compensated for by an increase in the diffuse light from the sky. (Solar and sky measurements at the Mauna Loa Observatory, Hawai’i, indicated a decrease in the direct flux from about 520 to 400 W m−2, but an increase in the diffuse flux from 40 to 140 W m−2; Robock 2005). ERBE also revealed that the albedo increase was pronounced in normally “dark”, cloud-free regions, including the Australian deserts and the Sahara, and in typically “bright” regions associated with deep convective cloud systems in the tropics including the Congo Basin. By July 1991, the outgoing shortwave flux increased dramatically over the tropics. This corresponded to a change in the net flux of up to −8 W m−2 in August 1991, twice the magnitude of any other monthly anomaly. The net forcing for August 1991 amounted to −4.3 W m−2 for the region between 40°S and 40°N, indicating a globally averaged volcanic forcing of at least −2.7 W m−2. The net flux returned to normal levels by March 1993. Pinatubo’s forcing effect on Earth’s heat budget exceeded for two years the positive forcing due to anthropogenic greenhouse gases. The observed global cooling was initially rapid but punctuated by a warming trend, predominantly over northern land masses, between January and March 1992 (Parker et al. 1996). Cooling resumed and, by June 1992, amounted to about 0.5 °C. This figure hides much wider regional variations that included pockets of abnormally strong heating as well as cooling. For example, the Siberian winter was 5 °C warmer than normal, while the north Atlantic was 5 °C cooler than average. The overall cooling is also a fraction of a degree stronger if the warming effect of the prevailing El Niño is accounted for. Globally, there was a significant drop in rainfall over land during the year following the eruption, making it the driest period in the half-century for which good records were available (Trenberth et al. 2007). However, there were strong regional patterns, including winter warming anomalies in Scandinavia, Siberia and central North America (Graf et al. 1993; Marshall et al. 2009; Stenchikov 2002). These temperature anomalies were associated with marked departures in sea-level pressure patterns in the first and second boreal winter associated with amplification of the positive phase of the North Atlantic Oscillation. This acts to transport warm, moist air to northern Eurasia. The temperature and precipitation patterns following the Pinatubo eruption have been fitted reasonably well by Intergovernmental Panel on Climate Change (IPCC) models, though they reproduce the summer cooling better than the winter warming (Broccoli et al. 2003; Stenchikov et al. 2006; Thomas et al. 2009a,b). The reduced shortwave solar radiation falling on the sea surface resulted in a net deficit of 5×1022 J in the surface ocean (Stenchikov et al. 2009). Globally averaged sea-surface temperatures dropped by around 0.4 °C. This effect is now recognized for numerous earlier eruptions based on studies of corals and tree rings (D’Arrigo et al. 2009) and acts to lower sea level.
404
Oppenheimer, Scaillet, Martin
Requirements for a climate-forcing eruption Based in large part on our understanding of the impacts of the 1991 Pinatubo eruption, it is possible to summarize a number of factors likely to distinguish an eruption capable of hemispheric to global scale climate impact (Oppenheimer 2011). We have seen that sulfur emission is crucial: it is the release of sulfur gases into the atmosphere during an eruption that leads to the formation of sulfate aerosol that may then perturb the Earth’s heat budget. Thus the melt composition, sulfur content, redox state, eruption magnitude and so on, will play a large role in controlling the sulfur yield of an eruption (see section on geodynamics and geochemical behavior of sulfur above; Baker and Moretti 2011, this volume). However, the climate response does not scale in any simple way with the sulfur yield of an eruption. In particular, greater releases of sulfur lead to non-linearities in the Beer-Lambert relationship (Eqn. 15), limiting light scattering, and resulting in growth of larger particles that are less effective in scattering short-wave radiation (H2SO4 vapor generated by oxidation of SO2 will tend to condense on existing particles rather than nucleate new small particles). They also settle out of the stratosphere more rapidly (Timmreck et al. 2009, 2010). The injection height of sulfur into the atmosphere represents another important determinant of climate impact. More intense eruptions, i.e., those with higher magma discharge rates, are more likely to loft the reactive sulfur gases into the stratosphere where they can generate climatically effective aerosol. Co-ignimbrite clouds are capable of entraining material into the stratosphere, and have been responsible for some of the largest tephra fall deposits. However, little is known about their efficiency, compared with Plinian eruption columns, in injecting sulfur into the stratosphere. Differences in the dynamics of the two plume types may influence the transfer of sulfur to the stratosphere (Herzog et al. 2010). The location of a volcano strongly influences the geographic distribution of atmospheric heating and the development of planetary waves that affect air circulation (especially in the northern hemisphere). Another relevant factor is that the height of the tropopause varies with latitude—at the tropics it is around 16-17 km above sea level but descends to 10-11 km at high latitudes. In general terms, an explosive eruption requires a greater intensity (magma discharge rate) to cross the tropopause in the tropics than at mid to polar latitudes. However, there are two factors that limit this effect. The first is that a high-latitude eruption will have a more limited effect than a low-latitude one because further from the tropics there is less solar energy to intercept. Secondly, atmospheric circulation works in a way to limit the effects of high latitude eruptions. A tropical eruption that pumps aerosol into the stratosphere results in localized heating. This increases the temperature difference in the middle atmosphere between the equator and high latitudes, and thereby enhances meridional air flows that spread aerosol into both hemispheres, promoting climate forcing at a worldwide scale. In contrast, volcanic aerosol injected into the stratosphere from high latitude volcanoes will tend to have the opposite effect on the temperature gradient, acting to stagnate meridional air flow. Very little, if any, of the stratospheric aerosol formed as a result of eruption of a high latitude volcano will reach the opposing hemisphere. Nevertheless, both historical and modeling evidence suggests that high-latitude volcanic eruptions can have significant hemispheric climatic effects (Schneider et al. 2009). A climate model run for the 1912 Katmai eruption (Alaska), and the same event scaled up three times, confirmed pooling of the aerosol veil to high-boreal latitudes (Oman et al. 2005). It also suggested that radiative impacts should dominate over the effects on circulation of the atmosphere. However, the results of such modeling efforts are very sensitive to assumed sizes of aerosol and parameterizations of the microphysical processes that occur in clouds. Linked to spatial influences on the climate forcing credentials of eruptions are seasonal considerations. These are likely to be most significant for high-latitude eruptions. With a strong polar vortex in winter, SO2 emissions could be trapped at high latitudes, and it would take
Sulfur Degassing From Volcanoes
405
longer to form sulfuric acid aerosol because of limited sunlight and oxidizing agents. One study that has explored seasonal influences on a mid-latitude eruption is the simulation of a supereruption of 1700 Tg of SO2 (100× Pinatubo) from Yellowstone volcano (44°N; Timmreck and Graf 2006). In the model, the aerosol veil of a summer eruption traveled west and southward, driven by the Aleutian high pressure system, compared with east and northward transport for a winter event, when westerlies dominate circulation. For the summertime case, more aerosol was transported to the southern hemisphere. This could be expected to exert a strong influence on radiative forcing at the Earth’s surface.
SuMMaRy anD COnCluSIOnS Although this review has focused on sulfur, its behavior and impacts should always be regarded as part of a larger picture of chemical and physical interactions, notably in magmas where interactions with other volatile species are important. Nevertheless, sulfur is a truly remarkable element in its own right. Its behavior in magmatic, hydrothermal, volcanic, atmospheric, terrestrial and aquatic environments is complex owing to a range of factors including its important role in redox chemistry (and biogeochemistry). In the atmosphere, the oxidation of sulfur gases to sulfuric acid, either in the aqueous aerosol phase in the troposphere or via homogenous reactions in the stratosphere (with sulfuric acid condensing under the prevailing conditions), is of particular significance. The short-term (months-to-years) impacts of volcanism on the atmosphere, climate and environment are strongly controlled by location, timing, flux, magnitude and emission height of sulfur gases (principally SO2 or H2S depending strongly on the physical conditions of the magmatic source). Episodic explosive eruptions represent the principal perturbation to stratospheric aerosol (though the atmospheric effects of sulfur degassing associated with continental flood basalts might well be more profound). In the troposphere, the picture is less clear but a significant part of the global tropospheric sulfate burden may be volcanogenic. Sulfate aerosol influences the Earth’s radiation budget by scattering and absorption of shortwave and long-wave radiation, and by acting as cloud condensation nuclei. When they are brought to the boundary layer and Earth’s surface, clouds containing volcanic sulfur in both gaseous and aerosol phases can result in profound environmental and health impacts. The global impacts of only one large eruption (magnitude > 1013 kg) have been studied with any instrumental detail—the 1991 eruption of Mt. Pinatubo. As the twentieth anniversary of that event looms (at the time of writing), it is remarkable to note that new findings concerning its climatic, environmental and ecological consequences are still emerging. Despite the tremendous insights afforded by this event, it represents a very small sample of the range of volcanic eruption styles, geographic locations and atmospheric states that could combine to produce significant perturbations to atmospheric radiation and dynamics. An important issue is the readiness of the scientific community to record the next major climate-forcing eruption as effectively as possible. More generally, substantial further work is required to constrain the temporal and spatial distribution of gas and particle emissions (including sulfur, halogens and trace metal species) to the atmosphere from all erupting and dormant volcanoes. In addition to improved observational data on the spatial and temporal distributions of volcanic volatiles to the atmosphere, further studies are required to characterize the physical and chemical interactions of gases and particles in the atmosphere. This will be essential for the realistic application of numerical models describing the transport and chemical evolution of plumes, and will contribute to a better understanding of volcanogenic pollution and improved mitigation of its effects. Surveillance of gas composition and flux are essential for interpretation of volcanic activity, since degassing is intimately linked with the physical and chemical environments of magma storage and the dynamics of magma transport. A revolution in spectroscopic and
406
Oppenheimer, Scaillet, Martin
electrochemical techniques has taken place over the past decade providing greatly improved means for volcanic gas surveillance at all scales. In the near future, we can anticipate further revolutions in the ability to measure volcanic volatile emissions in the field thanks to current developments of laser spectroscopy and lidar systems (enabling in situ determination of isotopic compositions of C, O, H, S, Cl, etc., and remote measurement of CO2 fluxes). Satellite remote sensing methods and the use of unmanned aerial vehicles will also lead to improved measurements of volcanic plumes and better inventories of the temporal and spatial distribution of sulfur (and carbon and halogen) fluxes. Unfortunately, the modeling frameworks for interpretation of geochemical data remain underdeveloped, impeding the application of such data in hazard assessment. Advances in this area will benefit from development and validation of comprehensive physico-chemical models for volcanic degassing and associated magma storage and transport based on the integration of results from experiments on the controls on distribution of volatiles in synthetic and natural melts (with an emphasis on understanding redox controls), analysis of dissolved volatiles preserved in melt inclusions, and observed volcanic gas geochemistry. Ultimately, such models can be applied to integrated inversions of geophysical, geodetic and geochemical monitoring data to support hazard assessment.
aCKnOwlEDGMEntS We thank Tobias Fischer, Mike Burton and Jim Webster for their careful and constructive reviews of the original manuscript. CO acknowledges support from the NERC via the National Centre for Earth Observation (“Dynamic Earth” theme) and the project “Magma dynamics at persistently degassing basaltic volcanoes: a novel approach to linking volcanic gases and magmatic volatiles within a physical model”, the European Research Council (for the DEMONS project “Deciphering eruptions by modeling outputs of natural systems”), and the EC 7th Framework Programme for support via the MIAVITA project “Mitigate and assess risk from volcanic impact on terrain and human activities.” Le Studium Institute for Advanced Studies receives support from the European Regional Development Fund.
REFEREnCES Aiuppa A, Baker DR, Webster JD (2009) Halogens in volcanic systems. Chem Geol 263:1-18 Aiuppa A, Bellomo S, Brusca L, D’Alessandro W, Di Paola R, Longo M (2006) Major-ion bulk deposition around an active volcano (Mt Etna, Italy). Bull Volcanol 68:255-265 Aiuppa A, Bellomo S, D’Alessandro W, Federico C, Ferm M, Valenza M (2004) Volcanic plume monitoring at Mount Etna by diffusive (passive) sampling. J Geophys Res 109, doi:101029/2003JD004481 Aiuppa A, Bertagnini A, Métrich A, Moretti R, Di Muro A, Liuzzo A, Tamburello G (2010) A model of degassing for Stromboli volcano. Earth Planet Sci Lett 295:195-204 Aiuppa A, Federico C, Giudice G, Gurrieri S (2005a) Chemical mapping of a fumarolic field: La Fossa Crater, Vulcano Island (Aeolian Islands, Italy). Geophys Res Lett 32, doi:101029/2005GL023207 Aiuppa A, Inguaggiato S, McGonigle AJS, O’Dwyer M, Oppenheimer C, Padgett MJ, Rouwet D, Valenza M (2005b) H2S fluxes from Mt Etna, Stromboli and Vulcano (Italy) and implications for the global volcanic sulfur budget. Geochim Cosmochim Acta 69:1861-1871 Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007a) Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology 35:1115-118 Aiuppa A, Franco A, von Glasow R, Allen AG, D’Alessandro W, Mather TA, Pyle DM, Valenza M (2007b) The tropospheric processing of acidic gases and hydrogen sulphide in volcanic gas plumes as inferred from field and model investigations. Atmos Chem Phys 7:1441-1450 Allard P (1979) 13C/12C and 34S/32S ratios in magmatic gases from ridge volcanism in Afar. Nature 282:56-58 Allard P (1981) Stable isotope composition of hydrogen, carbon and sulphur in magmatic gases from rift and island arc volcanoes. Bull Volcanol 45:269-271 Allard P (1997) Endogeneous magma degassing and storage at Mount Etna. Geophys Res Lett 24:2219-2222 Allard P (2010) A CO2-rich gas trigger of explosive paroxysms at Stromboli basaltic volcano, Italy. J Volcanol Geoth Res 189:363-374
Sulfur Degassing From Volcanoes
407
Allard P, Burton M, Muré F (2005) Spectroscopic evidence for a lava fountain driven by previously accumulated magmatic gas. Nature 433:407-410 Allard P, Carbonnelle J, Dajlevic D, Le Bronec J, Morel P, Robe MC, Maurenas JM, Faivre-Pierret R, Martin D, Sabroux JC, Zettwoog P (1991) Eruptive and diffuse emissions of CO2 from Mount Etna. Nature 351:387390 Allard P, Carbonnelle J, Métrich N, Loyer H, Zettwoog P (1994) Sulphur output and magma degassing budget of Stromboli volcano. Nature 368:326-329 Allard P, Le Guern F, Sabroux JC (1977) Themodynamic and isotopic studies in eruptive gases. Geothermics 5:37-40 Allen AG, Baxter PJ, Ottley CJ (2000) Gas and particle emissions from Soufrière Hills volcano, Montserrat, WI: Characterization and health hazard assessment. Bull Volcanol 62:6-17 Allen AG, Mather TA, McGonigle AJS, Aiuppa A, Delmelle P, Davison B, Bobrowski N, Oppenheimer C, Pyle DM, Inguaggiato S (2006) Sources, size distribution, and downwind grounding of aerosols from Mount Etna. J Geophys Res 111, doi:101029/2005JD006015 Allen AG, Oppenheimer C, Ferm M, Baxter PJ, Horrocks L, Galle B, McGonigle AJS, Duffell HJ (2002) Primary sulphate aerosol and associated emissions from Masaya volcano, Nicaragua. J Geophys Res, doi 101029/2002JD002120 Amend JP, Shock EL (2001) Energetics of overall metabolic reactions in thermophilic and hyperthermophilic Archaea and bacteria FEMS. Microbiol Rev 25:175-243 Anderson AT, Newman S, Williams SN, Druitt TH, Skirius C, Stolper E (1989) H2O, CO2, Cl gas in Plinian and ash-flow Bishop rhyolite. Geology 17:221-225 Andres RJ, Kasgnoc AD (1998) A time-averaged inventory of subaerial volcanic sulfur emissions. J Geophys Res 103(D19):25,251-25,261 Andres RJ, Rose WI, Kyle PR, DeSilva S, Francis P, Gardeweg M, Moreno Roa H (1991) Excessive sulfur dioxide emissions from Chilean volcanoes. J Volcanol Geotherm Res 46:323-329 Arellano S, Hall M, Samaniego P, Le Pennec J-L, Ruiz A, Molina I, Yepes H (2008) Degassing patterns of Tungurahua volcano (Ecuador) during the 1999-2006 eruptive period, inferred from remote spectroscopic measurements of SO2 emissions. J Volcanol Geotherm Res 176:151-162 Bagnato E, Aiuppa A, Parello F, Calabrese S, D’Alessandro W, Mather TA, McGonigle AJS, Pyle DM, Wängberg I (2007) Degassing of gaseous (elemental and reactive) and particulate mercury from Mount Etna volcano (Southern Italy). Atmos Env 41:7377-7388 Baillie MGL (2010) Volcanoes, ice-cores and tree rings: one story or two? Antiquity 84:202-215 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Ballhaus C (1993) Redox states of lithospheric and asthenospheric upper mantle. Contrib Mineral Petrol 114: 331-348 Bani P, Oppenheimer C, Tsanev VI, Carn SA, Cronin SJ, Crimp R, Calkins JA, Charley D, Lardy M Roberts TR (2009) Surge in sulphur and bromine degassing from Ambrym volcano, Vanuatu. Bull Volcanol 71:11591168 Baxter PJ (2000) Impacts of eruptions on human health. In: Encyclopedia of volcanoes. Sigurdsson H, Houghton BF, McNutt SR, Rymer H, Stix J (eds) Academic Press, San Diego, p 1035-1043 Baxter PJ, Stoiber RE, Williams SN (1982) Volcanic gases and health: Masaya Volcano, Nicaragua. Lancet 2:150-151 Behrens H, Ohlhorst S, Holtz F, Champenois M (2004) CO2 solubility in dacitic melts equilibrated with H2OCO2 fluids: Implications for modelling the solubility of CO2 in silicic melts. Geochim Cosmochim Acta 68:4687-4703 Berresheim H, Jaeschke W (1983) The contribution of volcanoes to the global atmospheric sulfur budget. J Geophys Res 88(C6):3732-3740 Bertagnini A, Métrich N, Francalanci L, Landi P, Tommasini S, Conticelli S (2008) Volcanology and magma geochemistry of the present-day activity: constraints on the feeding system. In: Learning from Stromboli. Calvari S, Inguaggiato S, Puglisi G, Ripepe M, Rosi M (eds) Am Geophys Union, Geophys Mon 182:1938 Bertagnini A, Métrich N, Landi P, Rosi M (2003) Stromboli volcano (Aeolian Archipelago, Italy): an open window on the deep-feeding system of a steady state basaltic volcano. J Geophys Res 108(B7): 2336, doi:101029/2002JB002146 Bézos A, Humler E (2002) The Fe3+/ΣFe ratios of MORB glasses and their implications for mantle melting. Geochim Cosmochim Acta 69:711-725 Bhugwant C, Sieja B, Bessafi M, Staudacher T, Ecormier J (2009) Atmospheric sulfur dioxide measurements during the 2005 and 2007 eruptions of the Piton de La Fournaise volcano: Implications for human health and environmental changes. J Volcanol Geotherm Res 184:208-224
408
Oppenheimer, Scaillet, Martin
Blake S (2003) Correlations between eruption magnitude, SO2 yield, and surface cooling. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Special Publication of the Geological Society of London 213:177-202 Bluth G, Shannon JM, Watson IM, Prata AJ, Realmuto VJ (2007) Development of an ultra-violet digital camera for volcanic SO2 imaging. J Volcanol Geotherm Res 161:47-56 Bluth GJS, Carn SA (2008) Exceptional sulphur degassing from Nyamuragira volcano, 1979-2005. Int J Remote Sens 29:6667-6685 Bluth GJS, Casadevall TJ, Schnetzler CC, Doiron SD, Walter LS, Krueger AJ, Badruddin M (1994) Evaluation of sulfur dioxide emissions from explosive volcanism: the 1982-1983 eruptions of Galunggung, Java, Indonesia. J Volcanol Geotherm Res 63:243-256 Bluth GJS, Rose WI, Sprod IE, Krueger AJ (1997) Stratospheric loading of sulfur from explosive volcanic eruptions. J Geol 105:671-684 Bobrowski N, Honninger G, Galle B, Platt U (2003) Detection of bromine monoxide in a volcanic plume. Nature 423:273-276 Bobrowski N, Platt U (2007) SO2/BrO ratios studied in five volcanic plumes. J Volcanol Geotherm Res 166:147160 Bobrowski N, von Glasow R, Aiuppa A, Inguaggiato S, Louban I, Ibrahim OW, Platt U (2007) Reactive halogen chemistry in volcanic plumes. J Geophys Res 112:D06311, doi:101029/2006JD007206 Boichu M, Oppenheimer C, Tsanev V, Kyle PR (2010) High temporal resolution SO2 flux measurements at Erebus volcano, Antarctica. J Volcanol Geotherm Res 190:325-336 Boichu M, Oppenheimer C, Tsanev V, Kyle PR (2011) On bromine, nitrogen oxides and ozone depletion in the tropospheric plume of Erebus volcano (Antarctica). Atmos Env doi:10.1016/j.atmosenv.2011.03.027 Botcharnikov RE, Behrens H, Holtz F, Koepke J, Sato H (2004) Sulfur and chlorine solubility in Mt Unzen rhyodacitic melt at 850 °C and 200 MP. Chem Geol 213:207-225 Broccoli AJ, Dixon KW, Delworth TL, Knutson TR, Stouffer RJ, Zeng F (2003) Twentieth-century temperature and precipitation trends in ensemble climate simulations including natural and anthropogenic forcing. J Geophys Res 108(D24):4798, doi:101029/2003JD003812 Bureau H, Métrich N, Semet MP, Staudacher T (1999) Fluid-magma decoupling in a hot spot volcano. Geophys Res Lett 26:3501-3504 Bureau H, Pineau F, Métrich N, Semet M, Javoy M (1998) A melt and fluid inclusion study of the gas phase at Piton de la Fournaise volcano (Réunion Island). Chem Geol 147:115-130 Burgisser A, Scaillet B (2007) Redox evolution of a degassing magma rising to the surface. Nature 445:194-197 Burgisser A, Scaillet B, Harshvardhan (2008) Chemical patterns of erupting silicic magmas and their influence on the amount of degassing during ascent. J Geophys Res 113:B12204, doi:101029/2008JB005680 Burton M, Allard P, Murè F, La Spina A (2007a) Magmatic gas composition reveals the source depth of slugdriven Strombolian explosive activity. Science 317:227-230 Burton MR, Caltabiano T, Murè F, Randazzo D (2009) SO2 flux from Stromboli during the 2007 eruption: Results from the FLAME network and traverse measurements. J Volcanol Geotherm Res 182:214-220 Burton MR, Mader HM, Polacci M (2007b) The role of gas percolation in quiescent degassing of persistently active basaltic volcanoes. Earth Planet Sci Lett 264:46-60 Caltabiano T, Romano R, Budetta G (1994) SO2 flux measurements at Mount Etna (Sicily). J Geophys Res 99:12,809-12,819 Campion R, Salerno GG, Coheur P-F, Hurtmans D, Clarisse L, Kazahaya K, Burton M, Caltabiano T, Clerbaux C, Bernard A (2010) Measuring volcanic degassing of SO2 in the lower troposphere with ASTER band ratios. J Volcanol Geotherm Res 194:42-54 Camuffo D, Enzi S (1995) Impact of clouds of volcanic aerosols in Italy during the last seven centuries. Nat Hazards 11(2):135-161 Carmichael ISE (1991) The redox state of basic and silicic magmas: a reflection of their source regions? Contrib Mineral Petrol 106:129-141 Carn SA (2004) Eruptive and passive degassing of sulphur dioxide at Nyiragongo Volcano (D. R. Congo): the 17th January 2002 eruption and its aftermath. Acta Vulcanol 14(15):75-86 Carn SA, Krueger AJ, Bluth GJS, Schaefer SJ, Krotkov NA, Watson IM, Datta S (2003) Volcanic eruption detection by the Ozone Mapping Spectrometer (TOMS) instruments: a 22-year record of sulphur dioxide and ash emissions. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc Lond Spec Pub 213:177-202 Carn SA, Krueger AJ, Krotkov NA, Yang K, Evans K (2009) Tracking volcanic sulfur dioxide clouds for aviation hazard mitigation. Nat Hazards 51:325-343 Casadevall TJ, Doukas MP, Neal CA, McGimsey RG, Gardner CA (1994) Emission rates of sulfur dioxide and carbon dioxide from Redoubt Volcano during the 1989-1990 eruptions. J Volcanol Geotherm Res 62:519530
Sulfur Degassing From Volcanoes
409
Casadevall TJ, Rose WI Jr, Fuller WH, Hunt WH, Hart MA, Moyers JL, Woods DC, Chuan RL, Friend JP (1984) Sulfur dioxide and particles in quiescent volcanic plumes from Poás, Arenal, and Colima volcanos, Costa Rica and Mexico. J Geophys Res 89(D6):9633-9641 Castrillo A, Casa G, Burgel MV, Tedesco D, Gianfrani L (2004) First field determination of the 13C/12C isotope ratio in volcanic CO2 by diode-laser spectrometry. Opt Express 12:651 Chesner CA, Luhr JF (2010) A melt inclusion study of the Toba Tuffs, Sumatra, Indonesia. J Volcanol Geotherm Res 197:259-278 Chin M, Jacob DJ (1996) Anthropogenic and natural contributions to tropospheric sulfate: a global model analysis. J Geophys Res 101:18691-18699 Chow DC, Grandinetti A, Fernandez E, Sutton AJ, Elias T, Brooks B, Tam EK (2010) Is volcanic air pollution associated with decreased heart-rate variability? Heart Asia 2:36-41 Christensen LE, Brunner B, Truong KN, Mielke RE, Webster CR, Coleman M (2007) Measurement of sulfur isotope compositions by tunable laser spectroscopy of SO2. Anal Chem 79:9261-9268 Christie DM, Carmichael ISE, Langmuir CH (1986) Oxidation states of mid-ocean ridge basalt glasses. Earth Planet Sci Lett 79:397-411 Clemente B, Scaillet B, Pichavant M (2004) The solubility of sulphur in rhyolitic melts. J Petrol 45:2171-2196 Clerbaux C, Coheur P-F, Clarisse L, Hadji-Lazaro J, Hurtmans D, Turquety S, Bowman K, Worden H, Carn SA (2008) Measurements of SO2 profiles in volcanic plumes from the NASA Tropospheric Emission Spectrometer (TES). Geophys Res Lett 35:L22807, doi:101029/2008GL035566 Corradini S, Merucci L, Prata AJ (2009) Retrieval of SO2 from thermal infrared satellite measurements: correction procedures for the effects of volcanic ash. Atmos Meas Tech 2:177-191 Costa F, Scaillet B, Gourgaud A (2003) Massive atmospheric sulfur loading of the AD 1600 Huaynaputina eruption and implications for petrologic sulfur estimates. Geophys Res Lett 30(2):1068, doi:101029/2002GL016402 Costa F, Scaillet B, Pichavant M (2004) Petrological and experimental constraints on the pre-eruption conditions of Holocene dacite from Volcán San Pedro (36°S, Chilean Andes) and the importance of sulphur in silicic subduction-related magmas. J Petrol 45:855-881 Cottrell E, Gardner J, Rutherford MJ (1999) Petrologic and experimental evidence for the movement and heating of the pre-eruptive Minoan rhyodacite (Santorini, Greece). Contrib Mineral Petrol 135:315-331 Courtillot VE, Renne PR (2003) On the ages of flood basalt events. C R Geosci 335:113-140 D’Arrigo R, Wilson R, Tudhope A (2009) The impact of volcanic forcing on tropical temperatures during the past four centuries. Nature Geosci 2:51-56 Daag AS, Tubianosa BS, Newhall CG, Tuñgol NM, Javier D, Dolan MT, Delos Reyes PJ, Arboleda RA, Martinez ML, Regalado TM (1996) Monitoring sulfur dioxide emission at Mount Pinatubo. In: Fire and mud: eruptions and lahars of Mount Pinatubo Philippines. Newhall CG, Punongbayan RS (eds), Philippine Institute of Volcanology and Seismology, Quezon City, and University of Washington Press, Seattle. p 409-434 Dalton MP, Waite GP, Watson IM, Nadeau PA (2010) Multiparameter quantification of gas release during weak Strombolian eruptions at Pacaya Volcano, Guatemala. Geophys Res Lett 37:L09303, doi:101029/2010GL042617 de Foy B, Krotkov NA, Bei N, Herndon SC, Huey LG, Martínez A-P, Ruiz-Suárez LG, Wood EC, Zavala M, Molina LT (2009) Hit from both sides: tracking industrial and volcanic plumes in Mexico City with surface measurements and OMI SO2 retrievals during the MILAGRO field campaign. Atmos Chem Phys 9:9599-9617 de Moor JM, Fischer TP, Hilton DR, Hauri E, Jaffe LA, Camacho JT (2005) Degassing at Anatahan volcano during the May 2003 eruption: Implications from petrology, ash leachates, and SO2 emissions. J Volcanol Geotherm Res 146:117-138 de Moor JM, Fischer TP, Sharp ZD, Hauri EH, Hilton DR, Atudorei V (2010) Sulfur isotope fractionation during the May 2003 eruption of Anatahan volcano, Mariana Islands: Implications for sulfur sources and plume processes. Geochim Cosmochim Acta 74:5382-5397 De Vito S, Massera E, Quercia L, Di Francia G (2007) Analysis of volcanic gases by means of electronic nose. Sensors Actuators 127: 36-41 Delgado-Granados H, González LC, Sánchez NP (2001) Sulfur dioxide emissions from Popocatépetl volcano (Mexico): case study of a high-emission rate, passively degassing erupting volcano. J Volcanol Geotherm Res 108:107-120 Delmelle P, Delfosse T, Delvaux B (2003) Sulfate, chloride and fluoride retention in Andosols exposed to volcanic acid emissions. Environ Pollut 126:445-457 Delmelle P, Stix J, Baxter P, Garcia-Alvarez J, Barquero J (2002) Atmospheric dispersion, environmental effects and potential health hazard associated with the low altitude gas plume of Masaya volcano, Nicaragua. Bull Volcanol 64:423-434 Delmelle P, Stix J, Bourque CPA, Baxter P, Garcia-Alvarez J, Barquero J (2001) Dry deposition and heavy acid loading in the vicinity of Masaya volcano, a major sulfur and chlorine source in Nicaragua. Environ Sci Technol 7:1289-1293
410
Oppenheimer, Scaillet, Martin
Di Carlo I, Pichavant M, Rotolo SG, Scaillet B (2006) Experimental crystallization of a high-K arc basalt: The golden pumice, Stromboli volcano (Italy). J Petrol 47(7):1317-1343 Diaz JA, Giese CF, Gentry WR (2002) Mass spectrometry for in-situ volcanic gas monitoring. Trends Anal Chem 21:498-514 Diehl T (2009) A global inventory of volcanic SO2 emissions for hindcast scenarios. http://www-lscedods.cea. fr/aerocom/AEROCOM_HC/volc/ Dixon JE, Clague DA (2001) Volatiles in basaltic glasses from Loihi seamount, Hawaii: evidence for a relatively dry plume component. J Petrol 42:627-654 Dixon JE, Clague DA, Stolper E (1991) Degassing history of water, sulfur, and carbon in submarine lavas from Kilauea volcano, Hawaii. J Geol 99:371-394 Dixon JE, Clague DA, Wallace P, Poreda R (1997) Volatiles in alkalic basalts from the North Arch volcanic field, Hawaii: extensive degassing of deep submarine-erupted alkalic series lavas. J Petrol 38:911-939 Dixon JE, Stolper E, Delaney JR (1988) Infrared spectroscopic measurements of CO2 and H2O in Juan de Fuca Ridge basaltic glasses. Earth Planet Sci Lett 90:87-104 Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in mid-ocean ridge basaltic liquids Part I: Calibration and solubility models J Petrol 36:1607-1631 Doutriaux-Boucher M, Dubuisson P (2008) Detection of volcanic SO2 by spaceborne infrared radiometers. Atmos Res 92:69-79 Eatough DJ, Caka FM, Farber RJ (1994) The conversion of SO2 to sulfate in the atmosphere. Isr J Chem 34:301314 Eckhardt S, Prata AJ, Seibert P, Stebel K, Stohl A (2008) Estimation of the vertical profile of sulfur dioxide injection into the atmosphere by a volcanic eruption using satellite column measurements and inverse transport modeling. Atmos Chem Phys 8:3881-3897 Edmonds M, Herd RA, Galle B, Oppenheimer C (2003a) Automated, high time-resolution measurements of SO2 flux at Soufrière Hills Volcano, Montserrat. Bull Volcanol 65:578-586 Edmonds M, Oppenheimer C, Pyle DM, Herd R (2003b) Rainwater and ash leachate analysis as a proxy for plume chemistry at Soufrière Hills Volcano, Montserrat. In: Volcanic degassing. Geol Soc Spec Pub 213:203-218 Edmonds M, Oppenheimer C, Pyle DM, Herd RA, Thompson G (2003c) SO2 emissions from Soufrière Hills Volcano and their relationship to conduit permeability, hydrothermal interaction and degassing regime. J Volcanol Geotherm Res 124:23-43 Edner H, Ragnarson S, Svanberg S, Wallinder E, Ferrera R, Cioni R, Raco B, Taddeucci G (1994) Total fluxes of sulfur dioxide from the Italian Volcanoes Etna, Stromboli and Vulcano measured by differential absorption lidar and passive differential optical absorption spectroscopy. J Geophys Res 99:18827-18838 Elias T, Sutton AJ, Oppenheimer C, Horton KA, Garbeil H, Tsanev V, McGonigle AJS, Williams-Jones G (2006) Intercomparison of COSPEC and two miniature ultraviolet spectrometer systems for SO2 measurements using scattered sunlight. Bull Volcanol 68:313-322 EPICA Community Members (2004) Eight glacial cycles from an Antarctic ice core. Nature 429:623-628 Ferguson DJ, Barnie TD, Pyle DM, Oppenheimer C, Yirgu G, Lewi E, Kidane T, Carn S, Hamling I (2010) Recent rift-related volcanism in Afar, Ethiopia. Earth Planet Sci Lett 292:409-418 Ferry JM, Baumgartner L (1987) Thermodynamic models of molecular fluids at the elevated pressures and temperatures of crustal metamorphism. Rev Mineral 17:323-365 Fischer TP, Morrissey MM, Calvache ML, Gómez M, Torres R, Stix J, Williams SN (1994) Correlations between SO2 flux and long-period seismicity at Galeras volcano. Nature 368:135-137 Fischer TP (2008) Volatile fluxes (H2O, CO2, N2, HCl, HF) from arc volcanoes. Geochem J 42:21-38 Francis P, Burton M, Oppenheimer C (1998) Remote measurements of volcanic gas compositions by solar FTIR spectroscopy. Nature 396:567-570 Francis P, Maciejewski A, Oppenheimer C, Chaffin C, Caltabiano T (1995) SO2:HCl ratios in the plumes from Mt Etna and Vulcano determined by Fourier transform spectroscopy. Geophys Res Lett 22:1717-1720 Francis PW, Oppenheimer C, Stevenson D (1993) Endogenous growth of persitently active volcanoes. Nature 366:554-557 Fujita S, Sakurai T, Matsuda K (2003) Wet and dry deposition of sulfur associated with the eruption of Miyakejima volcano, Japan. J Geophys Res 108(D15):4444, doi:101029/2002JD003064 Gaillard F, Scaillet B (2009) The sulfur content of volcanic gases on Mars. Earth Planet Sci Lett 279:34-43 Galle B, Johansson M, Rivera C, Zhang Y, Kihlman M, Kern C, Lehmann T, Platt U, Arellano S, Hidalgo S (2010) Network for Observation of Volcanic and Atmospheric Change (NOVAC)—A global network for volcanic gas monitoring: Network layout and instrument description. J Geophys Res 115:D05304, doi:101029/2009JD011823 Galle B, Oppenheimer C, Geyer A, McGonigle A, Edmonds M, Horrocks LA (2003) A miniaturised ultraviolet spectrometer for remote sensing of SO2 fluxes: a new tool for volcano surveillance, J Volcanol Geotherm Res 119:241-254
Sulfur Degassing From Volcanoes
411
Gao C, Oman L, Robock A, Stenchikov GL (2007) Atmospheric volcanic loading derived from bipolar ice cores: Accounting for the spatial distribution of volcanic deposition. J Geophys Res 112:D09109, doi:101029/2006JD007461 Gao C, Robock A, Ammann C (2008) Volcanic forcing of climate over the past 1500 years: An improved ice core-based index for climate models. J Geophys Res 113, D23111, doi:101029/2008JD010239 Gassó S (2008) Satellite observations of the impact of weak volcanic activity on marine clouds. J Geophys Res 113:D14S19, doi:101029/2007JD009106 Gerlach T (2004b) Volcanic sources of troposphere ozone-depleting trace gases. Geochem Geophys Geosyst 5:Q09007, doi:10.1029/2004GC000747 Gerlach TM (1979) Evaluation and restoration of the 1970 volcanic gas analyses from Mount Etna, Sicily. J Volcanol Geotherm Res 6:165-178 Gerlach TM (1980) Investigation of volcanic gas analyses and magma outgassing from Erta ‘Ale lava lake, Afar, Ethiopia. J Volcanol Geotherm Res 7:415-441 Gerlach TM (1993) Oxygen buffering of Kilauea volcanic gases and the oxygen fugacity of Kilauea basalt. Geochim Cosmochim Acta 57:795-814 Gerlach TM (2004a) Comment on paper: “Morphology and compositions of spinel in Pu’u’O’o lava (19961998), Kilauea volcano, Hawaii—enigmatic discrepancies between lava and gas-based fO2 determinationsof Pu’u’O’o lava. J Volcanol Geotherm Res 134:241-244 Gerlach TM, Casadevall TJ (1986) Evaluation of gas data from high temperature fumaroles at Mount St Helens, 1980-1982. J Volcanol Geotherm Res 28:107-140 Gerlach TM, McGee KA (1994) Total sulfur dioxide emissions and pre-eruption vapor‐saturated magma at Mount St Helens, 1980-88. Geophys Res Lett 21(25):2833-2836 Gerlach TM, Nordlie BE (1975) The C-O-H-S gaseous system Part II: Temperature, atomic composition and molecular equilibria in volcanic gases. Am J Sci 275:377-394 Gerlach TM, Westrich HR, Casadevall TJ, Finnegan DL (1994) Vapor saturation and accumulation in magmas of the 1989-1990 eruption of Redoubt volcano, Alaska. J Volcanol Geotherm Res 62:317-37 Gerlach TM, Westrich HR, Symonds RB (1996) Pre-eruption vapor in magma of the climactic Mount Pinatubo eruption: source of the giant stratospheric sulfur dioxide cloud. In: Newhall CG, Punongbayan RS (eds) Fire and Mud Eruptions and Lahars of Mount Pinatubo, Philippines, University of Washington Press, p 415-434 Gianfrani L, De Natale P, De Natale G (2000) Remote sensing of volcanic gases with a DFB-laser-based fiber spectrometer. Appl Phys B 70:467-470 Gianfrani L, Gagliardi G, van Burgel M, Kerstel E (2003) Isotope analysis of water by means of near-infrared dual-wavelength diode laser spectroscopy. Opt Express 11(13):1566-1576 Giggenbach WF (1975) A simple method for the collection and analysis of volcanic gas samples. Bull Volcanol 39: 132-145 Giggenbach WF (1987) Redox processes governing the chemistry of fumarolic gas discharges from White Island, New Zealand. Appl Geochem 1987(2):143-161 Giggenbach WF (1996) Chemical composition of volcanic gases. In: Scarpa M, Tilling RJ (eds) Monitoring and mitigation of Volcanic Hazards Springer; Heidelberg, Germany. p 221-256 Giggenbach WF (1997) The origin and evolution of fluids in magmatic-hydrothermal systems. In: Geochemistry of Hydrothermal Ore Deposits. 3rd Edition. Barnes HL (ed) Wiley, New York, NY, USA. p 737-796 Giggenbach WF, Goguel RL (1989) Collection and analysis of geothermal and volcanic water and gas discharges. Chemical Division of DSIR Report. CD 2401, 81 pp Giggenbach WF, Le Guern F (1976) The chemistry of magmatic gases from Erta’Ale, Ethiopia. Geochim Cosmochim Acta 40:25-30 Giggenbach WF, Matsuo S (1991) Evaluation of results from Second and Third IAVCEI field workshops on volcanic gases, Mt Usu, Japan, and White Island, New Zealand. Appl Geochm 6:125-141 Giggenbach WF, Tedesco D, Sulistiyo Y, Caprai A, Cioni R., Favara R, Fischer TP, Hirabayashi J-I, Korzhinsky M, Martini M, Menyailov I, Shinohara H (2001) Evaluation of results from the fourth and fifth IAVCEI field workshop on volcanic gases, Volcano island, Italy and Java, Indonesia. J Volcanol Geoth Res 108:157172 Graf H-F, Feichter J, Langmann B (1997) Volcanic sulfur emissions: Estimates of source strength and its contribution to the global sulfate distribution. J Geophys Res 102:10,727-10,738 Graf H-F, Kirchner, I Robock A, Schult I (1993) Pinatubo eruption winter climate effects: Model versus observation. Clim Dyn 9:81-93 Graf H-F, Langmann B, Feichter J (1998) The contribution of Earth degassing to the atmospheric sulfur budget. Chem Geol 147:131-145 Grutter M, Basaldud R, Rivera C, Harig R, Junkerman W, Caetano E, Delgado-Granados H (2008) SO2 emissions from Popocatépetl volcano: emission rates and plume imaging using optical remote sensing techniques. Atmos Chem Phys 8:6655-6663
412
Oppenheimer, Scaillet, Martin
Guo S, Bluth GJS, Rose WI, Watson IM, Prata AJ (2004) Re-evaluation of SO2 release of the 15 June 1991 Pinatubo eruption using ultraviolet and infrared satellite sensors. Geochem Geophys Geosyst 5:Q04001, doi:101029/2003GC000654 Halmer MM, Schmincke H-U, Graf H-F (2002) The annual volcanic gas input into the atmosphere, in particular into the stratosphere: a global data set for the past 100 years. J Volcanol Geotherm Res 15:511-528 Hammer CU (1977) Past volcanism revealed by Greenland ice sheet impurities. Nature 270:482-486 Hammer CU, Clausen HB, Dansgaard W (1980) Greenland ice sheet evidence of postglacial volcanism and its climatic impact. Nature 288:230-235 Hansen J, Ruedy R, Sato M, Reynolds R (1996) Global surface temperature in 1995: return to pre-Pinatubo level. Geophys Res Lett 23:1665-1668 Herzog M, Graf HF (2010) Applying the three-dimensional model ATHAM to volcanic plumes: Dynamic of large co-ignimbrite eruptions and associated injection heights for volcanic gases. Geophys Res Lett 37, L19807, doi:10.1029/2010GL044986 Heue K-P, Brenninkmeijer CAM, Wagner T, Mies K, Dix B, Frieß U, Martinsson BG, Slemr F, van Velthoven PFJ (2010) Observations of the 2008 Kasatochi volcanic SO2 plume by CARIBIC aircraft DOAS and the GOME-2 satellite. Atmos Chem Phys 10:4699-4713 Hewitt CN (2001) The atmospheric chemistry of sulphur and nitrogen in power station plumes. Atmos Env 35:1155-1170 Hoff RM (1992) Differential SO2 column measurements of the Mt. Pinatubo volcanic plume. Geophys Res Lett 19:175-178 Hoff RM, Millán MM (1981) Remote SO2 mass flux measurements using Cospec. J Air Pollut Control Assoc 31:381-384 Hofmann C, Courtillot V, Feraud G, Rochette P, Yirgu G, Ketefo E and Pik R (1997) Timing of the Ethiopian flood basalt event and implications for plume birth and global change. Nature 389:838-841 Holasek RE, Self S, Woods AW (1996) Satellite observations and interpretation of the 1991 Mount Pinatubo eruption plumes. J Geophys Res 101(B12):27,635-27,655 Holloway JR (1977) Fugacity and activity of molecular species in supercritical fluids. In: Thermodynamics in Geology. Fraser D (ed) Dordrecht, Reidel Publishing Company. p 161-181 Holloway JR (1987) Igneous fluids. Rev Mineral 17:211-232 Horn S, Schmincke HU (2000) Volatile emission during the eruption of Baitoushan Volcano (China/North Korea) ca. 969 AD. Bull Volcanol 61:537–55 Hubberten HW, Nielsen H, Puchelt H (1975) The enrichment of 34S in the solfataras of the Nea Kameni volcano, Santorini Archipelago, Greece. Chem Geol 16:197-205 Huntingdon AT (1973) The collection of volcanic gases from Mount Etna. Phil Trans R Soc London 274:119128 Ilyinskaya E, Oppenheimer C, Mather TA, Kyle P (2010) Chemistry and size distribution of water-soluble aerosol in the plume of Mt Erebus volcano, Antarctica. Geochem Geophys Geosys 11:Q03017, doi:101029/2009GC002855 Jaeschke W, Claude H, Herrmann J (1980) Sources and sinks of atmospheric H2S. J Geophys Res 85(C10):56395644 Jaggar TA (1940) Magmatic gases. Am J Sci 238:313-353 Johansson M, Galle B, Rivera C, Zhang Y (2009a) Tomographic reconstruction of gas plumes using scanning DOAS. Bull Volcanol 71:1169-1178 Johansson M, Galle B, Zhang Y, Rivera C, Chen D, Wyser K (2009b) The dual-beam mini-DOAS technique: Measurements of volcanic gas emission, plume height and plume speed with a single instrument. Bull Volcanol 71:747-751 Johnson MC, Rutherford MJ (1989) Experimentally determined conditions in the Fish Canyon Tuff, Colorado, magma chamber. J Petrol 30:711-737 Kantzas EP, McGonigle AJS, Bryant RG (2009) Comparison of low cost miniature spectrometers for volcanic SO2 emission measurements. Sensors 9:3256-3268 Kantzas EP, McGonigle AJS, Tamburello G, Aiuppa A, Bryant RG (2010) Protocols for UV camera volcanic SO2 measurements. J Volcanol Geotherm Res 194:55-60 Karagulian F, Clarisse L, Clerbaux C, Prata AJ, Hurtmans D, Coheur PF (2010) Detection of volcanic SO2, ash, and H2SO4 using the Infrared Atmospheric Sounding Interferometer (IASI). J Geophys Res 115:D00L02, doi:101029/2009JD012786 Kassi S, Chenevier M, Gianfrani L, Salhi A, Rouillard Y, Ouvrard A, Romanini D (2006) Looking into the volcano with a Mid-IR DFB diode laser and Cavity Enhanced Absorption Spectroscopy. Opt Express 14:11442-11452 Kazahaya K, Shinohara H, Saito G (1994) Excessive degassing of Izu-Oshima volcano: magma convection in a conduit. Bull Volcanol 56:207-216 Kazahaya K, Shinohara H, Uto K, Odai M, Nakahori Y, Mori H, Iino H, Miyashita M, Hirabayashi J (2004) Gigantic SO2 emission from Miyakejima volcano, Japan, caused by caldera collapse. Geology 32:425-428
Sulfur Degassing From Volcanoes
413
Kazahaya R, Mori T, Kazahaya K, Hirabayashi J (2008) Computed tomography reconstruction of SO2 concentration distribution in the volcanic plume of Miyakejima, Japan, by airborne traverse technique using three UV spectrometers. Geophys Res Lett 35:L13816, doi:101029/2008GL034177 Kellerhals T, Tobler L, Brütsch S, Sigl M, Wacker L, Gäggeler HW, Schwikowski M (2010) Thallium as a tracer for preindustrial volcanic eruptions in an ice core record from Illimani, Bolivia. Env Sci Technol 44:888–893 Keppler H (1999) Experimental evidence for the source of excess sulfur in explosive volcanic eruption Science 284:1652-1654 Keppler H (2010) The distribution of sulfur between haplogranitic melts and aqueous fluids. Geochim Cosmochim Acta 74:645-660 Kern C, Deutschann T, Vogel L, Wöhrbach M, Wagner T, Platt U (2010) Radiative transfer corrections for accurate spectroscopic measurements of volcanic gas emissions. Bull Volcanol 72:233-247 Kern C, Sihler H, Vogel L, Rivera C, Herrera M, Platt U (2009) Halogen oxide measurements at Masaya Volcano, Nicaragua using active long path differential optical absorption spectroscopy. Bull Volcanol 71:659-670 Krueger AJ (1983) Sighting of El Chichón sulfur dioxide clouds with the Nimbus 7 total ozone mapping spectrometer. Science 220:1377-1379 Krueger AJ, Krotkov N, Carn S (2008) El Chichón: the genesis of volcanic sulfur dioxide monitoring from space. J Volcanol Geotherm Res 175:408-414 Krueger AJ, Krotkov NA, Yang K, Carn S, Vicente G, Schroeder W (2009) Applications of satellite-based sulfur dioxide monitoring. IEEE J Sel Top Appl Earth Obs Remote Sens 2:293-298 Krueger AJ, Walter LS, Bhartia PK, Schnetzler CC, Krotkov NA, Sprod I, Bluth GJS (1995) Volcanic sulfur dioxide measurements from the Total Ozone Mapping Spectrometer (TOMS) instruments. J Geophys Res 100:14057-14076 Kusakabe M, Komoda Y, Takano B, Abiko T (2000) Sulfur isotopic effects in the disproportionation reaction of sulfur dioxide in hydrothermal fluids: implicaitons for the δ34S variations of dissolved bisulfate and elemental sulfur from active crater lakes. J Volcanol Geotherm Res 97:287-307 Le Guern F, Gerlach TM, Nohl A (1982) Field gas chromatograph analyses of gases from a glowing dome at Merapi volcano, Java, Indonesia, 1977, 1978, 1979. J Volcanol Geotherm Res 14:223-245 Lee C, Martin RV, van Donkelaar A, O’Byrne G, Krotkov N, Richter A, Huey LG, Holloway JS (2009a) Retrieval of vertical columns of sulfur dioxide from SCIAMACHY and OMI: Air mass factor algorithm development, validation, and error analysis. J Geophys Res 114:D22303 doi:101029/2009JD012123 Lee C, Richter A, Weber M, Burrows JP (2009b) SO2 Retrieval from SCIAMACHY using the Weighting Function DOAS (WFDOAS) technique: comparison with Standard DOAS retrieval. Atmos Chem Phys 8:6137-6145 Li ZXA, Lee CTA (2005) The constancy of upper mantle fO2 through time inferred from V/Sc ratios in basalts. Earth Planet Sci Lett 228:483-493 Linzon SN, Temple PJ, Pearson RG (1979) Sulfur concentrations in plant foliage andrelated effects. J Air Pollut Cont Assoc 29:520-525. Longo BM (2009) The Kilauea volcano adult health survey. Nursing Res 58:23-31 Longo BM, Grunder A, Chuan R, Rossignol A (2005) SO2 and fine aerosol dispersion from the Kilauea plume, Kau district, Hawaii, USA. Geology 33: 217-220 Longo BM, Rossignol A, Green JB (2008) Cardiorespiratory health effects associated with sulphurous volcanic air pollution. Public Health 122: 809-820 Longo BM, Yang W (2008) Acute bronchitis and volcanic air pollution: a community-based cohort study at Kilauea Volcano, Hawai’i, USA. J Toxicol Environ Health A71:1565-1571 Longo BM, Yang W, Green JB, Longo AA, Harris M, Biblione R (2010) An indoor air quality assessment for vulnerable populations exposed to volcanic vog from Kilauea Volcano. Fam Community Health 33:21-31 Louban I, Bobrowski N, Rouwet D, Inguaggiato S, Platt U (2009) Imaging DOAS for volcanological applications. Bull Volcanol 71:753-765 Love SP, Goff F, Counce D, Siebe C, Delgado H (1998) Passive infrared spectroscopy of the eruption plume at Popocatepetl volcano, Mexico. Nature 396:563-567 Love SP, Goff F, Schmidt SC, Counce D, Pettit D, Christenson BW, Siebe C (2000) Passive infrared spectroscopic remote sensing of volcanic gases: ground-based studies at White Island and Ruapehu, New Zealand, and Popocatepetl, Mexico. In: Remote Sensing of Active Volcanism. Mouginis-Mark P, Crisp J, Fink J (Eds) Geophys Monographs 116:117-138 Luhr JF (1990) Experimental phase relations of water- and sulfur- saturated arc magmas and the 1982 eruptions of El Chichòn volcano. J Petrol 31:1071-1114 Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichòn Volcano, Chiapas, Mexico: mineralogy and petrology of the anhydrite-bearing pumices. J Volcanol Geotherm Res 23:69-108 Luhr JF, Logan MAV (2002) Sulfur isotope systematics of the 1982 El Chichon trachyandesite: An ion microprobe study. Geochim Cosmochim Acta 66:3303-3316
414
Oppenheimer, Scaillet, Martin
Malinconico LL (1979) Fluctuations in SO2 emission during recent eruptions of Etna. Nature 278:43-45 Mallmann G, O’Neill HC (2009) The crystal/melt partitioning of V during mantle melting as a function of oxygen fugacity compared with some other elements (Al, P, Ca, Sc, Ti, Cr, Fe, Ga, Y, Zr and Nb). J Petrol 50:1765-1794 Mandeville CW, Webster JD, Tappen C, Taylor BE, Timbal A, Sasaki A, Hauri E, Bacon CR (2009) Stable isotope and petrologic evidence for open-system degassing during the climactic and pre-climactic eruptions of Mt. Mazama, Crater Lake, Oregon. Geochim Cosmochim Acta 73:2978-3012 Mannino DM, S Ruben F C Holschuh, T C Holschuh, M D Wilson, T Holschuh (1996) Emergency department visits and hospitalizations for respiratory disease on the island of Hawaii, 1981 to 1991. Hawaii Med J 55:48-53 Marini L, Chiappini V, Cioni R, Cortecci G, Dinelli E, Principe C, Ferrara G (1998) Effect of degassing on sulfur contents and δ34S values in Somma-Vesuvius magmas. Bull Volcanol 60:187-194 Marini L, Gambardella B, Principe C, Arias A, Brombach T, Hunziker J C (2002) Characterization of magmatic sulfur in the Aegean island arc by means of the δ34S values of fumarolic H2S, elemental S, and hydrothermal gypsum from Nisyros. Earth Planet Sci Lett 200:15-31 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Marshall AG, Scaife AA, Ineson S (2009) Enhanced seasonal prediction of European winter warming following volcanic eruptions. J Climate 22:6168-6180 Martel C, Pichavant M, Bourdier J-L, Traineau H, Holtz F, Scaillet B (1998) Magma storage conditions and control of eruption regime in silicic volcanoes: experimental evidence from Mt. Pelée. Earth Planet Sci Lett 156:89-99 Martin D, Ardouin B, Bergametti G, Carbonnelle J, Faivre-Pierret R, Lambert G, Le Cloarec MF, Sennequier G (1986) Geochemistry of sulfur in Mount Etna plume. J Geophys Res 91(B12):12,249-12,254 Martin LR, Damschen DE (1981) Aqueous oxidation of sulfur dioxide by hydrogen peroxide at low pH. Atmos Environ 15: 1615-1621 Martin RS, Mather TA, Pyle D, Power M, Allen AG, Aiuppa A, Horwell CJ, Ward EPW (2008) Compositionresolved size distributions of volcanic aerosols in the Mt. Etna plumes. J Geophys Res 113, D17211, doi:10.1029/2007JD009648 Martin RS, Mather TA, Pyle DM (2006) High-temperature mixtures of magmatic and atmospheric gases. Geochem Geophys Geosys 7, Q04006, doi:10.1029/2005GC001186 Martin RS, Mather TA, Pyle DM, Day JA, Witt MLI, Collins SJ, Hilton RG (2010a) Major and trace element distributions around active volcanic vents determined by analyses of grasses: Implications for element cycling and biomonitoring. Bull Volcanol 72:1009-1020 Martin RS, Roberts TJ, Mather TA, Pyle DM (2009) The implications of H2S and H2 stability in high-T mixtures of magmatic and atmospheric gases for the production of oxidized trace species (e.g., BrO and NOx). Chem Geol 263:143-150 Martin RS, Sawyer GM, Spampinato L, Salerno GG, Ramirez C, Ilyinskaya E, Witt MLL, Mather TA, Allen AG, Watson IM, Phillips JC, Oppenheimer C (2010b) A total volatile inventory for Masaya volcano, Nicaragua, J Geophys Res 115:B09215, doi:10.1029/2010JB007480 Mather T, Pyle DM, Oppenheimer C (2003a) Tropospheric volcanic aerosol. In: Volcanism and the Earth’s atmosphere. Robock A, Oppenheimer C (eds), Am Geophys Union Monograph 139 189-212 Mather TA (2008) Volcanoes and the atmosphere: the potential role of the atmosphere in unlocking the reactivity of volcanic emissions. Phil Trans Roy Soc A 366:4581-4595 Mather TA, Allen AG, Oppenheimer C, Pyle DM, McGonigle AJS (2003b) Size-resolved particle compositions of the tropospheric plume of Masaya volcano, Nicaragua. J Atmos Chem 46(3):207-237 Mather TA, Tsanev VI, Pyle DM, McGonigle AJS, Oppenheimer C, Allen AG (2004c) Characterization and evolution of tropospheric plumes from Lascar and Villarica volcanoes, Chile. J Geophys Res 109, D21303, doi:10.1029/2004JD004934 Mather TA, Allen AG, Pyle DM, Davison BM, Oppenheimer C and McGonigle AJS (2004a) Nitric acid from volcanoes. Earth Planet Sci Lett 218:17-30 Mather TA, McCabe JR, Rai VK, Thiemens MH, Pyle DM, Heaton THE, Sloane HJ, Fern GR (2006) The oxygen and sulfur isotopic composition of volcanic sulfate aerosol at the point of emission. J Geophys Res 111:D18205, doi:10.1029/2005JD006584 Mather TA, Oppenheimer C, Allen AG, McGonigle AJS (2004b) Aerosol chemistry of emissions from three contrasting volcanoes in Italy. Atmos Environ 38:5637-5649 Mathez EA (1984) Influence of degassing on oxidation states of basaltic magmas. Nature 310:371-375 Matthews SJ, Sparks RSJ, Gardeweg MC (1999) The Piedras Grandes-Soncor eruptions, Lascar volcano, Chile; evolution of a zoned magma chamber in the Central Andean upper crust. J Petrol 40:1891-1919 McGee KA (1992) The structure, dynamics and chemical composition of non-eruptive plumes from Mt. St. Helens, 1980-88. J Volcanol Geotherm Res 51:269-282
Sulfur Degassing From Volcanoes
415
McGonigle AJS, Aiuppa A, Giudice G, Tamburello G, Hodson AJ, Gurrieri S (2008) Unmanned aerial vehicle measurements of volcanic carbon dioxide fluxes. Geophys Res Lett 35:L06303, doi:10.1029/2007GL032508 McGonigle AJS, Delmelle P, Oppenheimer C, Tsanev VI, Delfosse T, Horton H, Williams-Jones G (2004a) SO2 depletion in tropospheric volcanic plumes. Geophys Res Lett 31:L13201, doi:101029/2004GL019990 McGonigle AJS, Hilton DR, Fischer TP, Oppenheimer C (2005) Plume velocity determination for volcanic SO2 flux measurements. Geophys Res Lett 32:L11302, doi:101029/2006GL022470 McGonigle AJS, Oppenheimer C, Galle B, Mather T, Pyle D (2002) Walking traverse and scanning DOAS measurements of volcanic gas emission rates. Geophys Res Lett 29(20):1985 doi: 101029/2002GL015827 McGonigle AJS, Oppenheimer C, Hayes AR, Galle B, Edmonds M, Caltabiano T, Salerno G, Burton M, Mather TA (2003) Sulphur dioxide fluxes from Mount Etna, Vulcano, and Stromboli measured with an automated scanning ultraviolet spectrometer. J Geophys Res 108(B9):2455, doi:101029/2002JB002261 McGonigle AJS, Oppenheimer C, Tsanev VI, Saunders S, Mulina K, Tohui S, Bosco J, Nahou J, Kuduon J and Taranu F (2004b) sulphur dioxide fluxes from Papua New Guinea’s volcanoes. Geophys Res Lett 31:L08606, doi:101029/2004GL019568 Menyailov IA, Nikitina LP, Shapar VN, Pilipenko VP (1986) Temperature increase and chemical change of fumarolic gases at Momotombo volcano, Nicaragua, in 1982-1985: are these indicators of a possible eruption? J Geophys Res 91(B12):12,199-12,214 Métrich N, Bertagnini A, Landi P, Rosi M (2001) Crystallisation driven by decompression and water loss at Stromboli volcano (Aeolian Islands, Italy). J Petrol 42:1471-1490 Métrich N, Mandeville CW (2010) Sulfur in magmas. Elements 6:81-86 Métrich N, Sigurdsson H, Meyer PS, Devine JD (1991) The 1783 Lakagigar eruption in Iceland: geochemistry, CO2, and sulfur degassing. Contrib Mineral Petrol 107:435-447 Minnis P, Harrison EF, Stowe LL, Gibson GG, Ddenn FM, Doelling DR, Smith WL Jr. (1993) Radiative climate forcing by the Mount Pinatubo eruption. Science 259 1411-1415 Moffat AJ, Millán MM (1971) The application of optical correlation techniques to the remote sensing of SO2 plumes using skylight. Atmos Environ 5:677-690 Möller D (1980) Kinetic model of atmospheric SO2 oxidation based on published data. Atmos Env 14:10671076 Montegrossi G, Tassi F, Minissale AA, Vaselli O, Buccianti A (2008) Natural fluctuation of sulfur species in volcanic fumaroles. J Non-Equilib Thermodyn 33:75-102 Montegrossi G, Tassi F, Vaselli O, Buccianti A, Garofalo K (2001) Sulfur species in volcanic gases. Anal Chem 73:3709-3715 Moretti R, P Papale and G Ottonello (2003) A model for the saturation of C-O-H-S fluids in silicate melts. In: Volcanic degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc Spec Pub 213:81-101 Mori T, Burton M (2006) The SO2 camera: A simple, fast and cheap method for ground-based imaging of SO2 in volcanic plumes. Geophys Res Lett 33, L24804, doi:101029/2006GL027916 Mori T, Burton MR (2009) Quantification of the gas mass emitted during single explosions on Stromboli with the SO2 imaging camera. J Volcanol Geotherm Res188: 395-400 Mori T, Kazahaya K, Oppenheimer C, McGonigle AJS, Tsanev V, Olmos R, Ohwada M Shuto T (2006) Sulfur dioxide fluxes from the volcanoes of Hokkaido, Japan. J Volcanol Geotherm Res 158: 235-243 Mori T, Notsu K (1997) Remote CO, COS, CO2, SO2, HCl detection and temperature estimation of volcanic gas. Geophys Res Lett 24(16), 2047-2050, doi:101029/97GL52058 Mori T, Notsu K (2008) Temporal variation in chemical composition of the volcanic plume from Aso volcano, Japan, measured by remote FT-IR spectroscopy. Geochem J 42:133-140 Mori T, Notsu K, Tohjima Y, Wakita H (1993) Remote detection of HCl and SO2 in volcanic gas from Unzen Volcano, Japan. Geophys Res Lett 20(13), 1355-1358, doi:101029/93GL01065 Mori T, Notsu K, Tohjima Y, Wakita H, Nuccio PM, Italiano F (1995) Remote detection of fumarolic gas chemistry at Vulcano, Italy, using an FT-infrared spectral radiometer. Earth Planet Sci Lett 134:219-224 Moune S, Holtz F, Botcharnikov RE (2009) Sulphur solubility in andesitic to basaltic melts: implications for Hekla volcano. Contrib Mineral Petrol 157:691-707 Nakada S, Nagai M, Kaneko T, Nozawa A, Suzuki-Kamata K (2005) Chronology and products of the 2000 eruption at Miyakejima. Bull Volcanol 67:205–218 Nakamura T, Okuno M, Kimura K, Mitsutani T, Moriwaki H, Ishizuka Y, Kim KH (2007) Application of 14C wiggle-matching to support dendrochronological analysis in Japan. Tree-Ring Res 63:37-46 Naughton JJ, Derby JV, Glover RB (1969) Infrared measurements on volcanic gas and fume: Kilauea eruption, 1968. J Geophys Res 74(12):3273-3277 Naughton JJ, Heald EF, Barnes IL (1963) The chemistry of volcanic gases 1 Collection and analysis of equilibrium mixtures by gas chromatography. J Geophys Res 68(2):539-544 Németh K, Cronin SJ, White JDL (2007) Kuwae caldera and climate confusion. Open Geol J 1:7-11 Notsu K, Mori T, Igarashi G, Tohjima Y, Wakita H (1993) Infrared spectral radiometer: A new tool for remote measurement of SO2 of volcanic gas. Geochem J 27:361-366
416
Oppenheimer, Scaillet, Martin
O’Dwyer M, Padgett MJ, McGonigle AJS, Oppenheimer C, Inguaggiato S (2003) Real-time measurement of volcanic H2S and SO2 concentrations by UV spectroscopy. Geophys Res Lett 30(12):1652, doi 101029/2003GL017246 Ohba T, Nogami K, Hirabayashi J, Mori T (2008) Isotopic fractionation of SO2 and H2S gases during the absorption by KOH solution, with the application to volcanic gas monitoring at Miyakejima Island, Japan. Geochemical J 42:119-131 Oman L, Robock A, Stenchikov G, Schmidt GA, Ruedy R (2005) Climatic response to high-latitude volcanic eruptions. J Geophys Res 110:D13103, doi:101029/2004JD005487 Oppenheimer C (1992) Sulphur eruptions at Volcán Poás, Costa Rica. J Volcanol Geotherm Res 49:1-21 Oppenheimer C (2002) Limited global change due to largest known Quaternary eruption, Toba ≈74 kyr BP? Quat Sci Rev 21:1593-1609 Oppenheimer C (2003) Ice core and palaeoclimatic evidence for the great volcanic eruption of 1257. Int J Climatol 23:417-426 Oppenheimer C (2010) Ultraviolet sensing of volcanic sulfur emissions. Elements 6:87-92 Oppenheimer C (2011) Eruptions That Shook the World. Cambridge University Press Oppenheimer C, Burton MR, Durieux J, Pyle DM (2002a) Open-path Fourier transform spectroscopy of gas emissions from a carbonatite volcano: Oldoinyo Lengai, Tanzania. Optics Lasers Eng 37:203-214 Oppenheimer C, Edmonds M, Francis P, Burton MR (2002b) Variation in HCl/SO2 gas ratios observed by Fourier transform spectroscopy at Soufrière Hills Volcano, Montserrat. In: The eruption of Soufrière Hills Volcano, Montserrat, from 1995 to 1999. Druitt TH, Kokelaar P (eds) Geol Soc Lond Mem 21:621-639 Oppenheimer C, Francis P, Burton M, Maciejewski A, Boardman L (1998a) Remote measurement of volcanic gases by Fourier transform infrared spectroscopy. Appl Phys B 67:505-515 Oppenheimer C, Francis P, Stix J (1998b) Depletion rates of SO2 in tropospheric volcanic plumes. Geophys Res Lett 25:2671-2674 Oppenheimer C, Kyle PR (2008) Probing the magma plumbing of Erebus volcano, Antarctica, by open-path FTIR spectroscopy of gas emissions. J Volcanol Geotherm Res 177:743-754 Oppenheimer C, Kyle PR, Tsanev VI, McGonigle AJS, Mather TA, Sweeney D (2005) Mt Erebus, the largest point source of NO2 in Antarctica. Atmos Env 39:6000-6006 Oppenheimer C, Lomakina A, Kyle PR, Kingsbury NG, Boichu M (2009) Pulsatory magma supply to a phonolite lava lake. Earth Planet Sci Lett 284:392-398 Oppenheimer C, Moretti R, Kyle P, Eschenbacher A, Lowenstern J, Hervig R (2011) Mantle to surface degassing of alkalic magmas at Erebus volcano., Antarctica. Earth Planet Sci Lett, doi:10.1016/j.epsl.2011.04.005 Oppenheimer C, Kyle P, Eisele JF, Crawford GJ, Huey DG, Tanner SD, Kim LS, Mauldin DL, Blake AD, Beyersdorf MA, Buhr M, Davis D (2010) Atmospheric chemistry of an Antarctic volcanic plume. J Geophys Res, 115 D04303, doi:101029/2009JD011910 Oppenheimer C, Tsanev VI, Braban CF, Cox RA, Adams JW, Aiuppa A, Bobrowski N, Delmelle P, Barclay J, McGonigle AJS (2006) BrO formation in volcanic plumes. Geochim Cosmochim Acta 70:2935-2941 Parker DE, Wilson H, Jones PD, Christy JR, Folland CK (1996) The impact of Mount Pinatubo on world-wide temperatures. Int J Climatol 16:487-197 Parnell RA (1986) Processes of soil acidification in tropical durandepts, Nicaragua. Soil Sci 42:43-55 Percy KE, Baker EA (1988) Effects of simulated acid rain on leaf wettability, rain retention and uptake of some inorganic ions. New Phytol 108: 75-82 Perret FA (1909) Vesuvius: characteristics and phenomena of the present repose-period. Am J Sci 28:413-430 Perret FA (1950) Volcanological observations. Carnegie Inst Wash Publ 549: 162pp Picardi G (1982) Fumaroles gas collection and analysis. Bull Volcanol 45:257-260 Pichavant M, Di Carlo I, Le Gac Y, Rotolo S, Scaillet, B (2009) Experimental constraints on the deep magma feeding system at Stromboli volcano, Italy. J Petrol 50:601-624 Platt U, Stutz J (2008) Differential Optical Absorption Spectroscopy: Principles and applications. SpringerVerlag, Heidelberg, 272 pp Porter JN, Horton K, Mouginis-Mark P, Lienert B, Lau E, Sutton AJ, Elias T, Oppenheimer C (2002) Sun photometer and lidar measurements of the plume from the Hawaii Kilauea volcano Pu’u ‘O’o vent: estimates of aerosol flux rates and SO2 lifetime. Geophys Res Lett doi:101029/2002GL014744 Prata AJ, Bernardo C (2007) Retrieval of volcanic SO2 column abundance from Atmospheric Infrared Sounder data. J Geophys Res 112, D20204, doi:101029/2006JD007955 Prata AJ, Kerkmann J (2007) Simultaneous retrieval of volcanic ash and SO2 using MSG-SEVIRI measurements. Geophys Res Lett 34, L05813, doi:101029/2006GL028691 Putirka KD (2008) Thermometers and barometers for volcanic systems. In: Putirka KD, Tepley F (eds). Rev Mineral Geochem 69:61-120 Pyle DM, Beattie PD, Bluth GJS (1996) Sulphur emissions to the stratosphere from explosive volcanic eruptions. Bull Volcanol 57:663-671 Radke LF (1982) Sulphur and sulphate from Mt Erebus. Nature 299:710-712
Sulfur Degassing From Volcanoes
417
Raga GB, Kok GL, Baumgardner D (1999) Evidence for volcanic influence on Mexico City aerosols. Geophys Res Lett 26:1149-1152 Rampino MR, Stothers R (1988) Flood basalt volcanism during the past 250 million years. Science 241:663-668 Read WG, Froidevaux L, Waters JW (1993) Microwave Limb Sounder measurements of stratospheric SO2 from the Mt. Pinatubo eruption. Geophys Res Lett 20:1299-1302 Richter D, Erdelyi M, Curl RF, Tittel FK, Oppenheimer C, Duffell HJ, Burton M (2002) Field measurement of volcanic gases using tunable diode laser based mid-infrared and Fourier transform infrared spectrometers. Opt Lasers Eng 37:171-186 Roberge J, Delgado-Granados H, Wallace PJ (2009) Mafic magma recharge supplies high CO2 and SO2 gas fluxes from Popocatépetl volcano, Mexico. Geology 37:107-110 Roberts TJ, Braban CF, Martin RS, Oppenheimer C, Adams JW, Cox RA, Jones RL, Griffiths PT (2009) Modelling reactive halogen formation and ozone depletion in volcanic plumes. Chem Geol 263:110-121 Roberts TJ, Braban CF, Oppenheimer C, Freshwater R, Martin RS, Dawson DH, Griffiths PT, Cox RA, Safell J, Jones RL (2011) Electrochemical sensing of volcanic gases. Chem Geol, in revision Robock A (2005) Cooling following large volcanic eruptions corrected for the effect of diffuse radiation on tree rings. Geophys Res Lett 32, L06702, doi:101029/2004GL022116 Rocco A, De Natale G, De Natale P, Gagliardi G, Gianfrani L (2004) A diode-laser-based spectrometer for insitu measurements of volcanic gases. Appl Phys B 78:235-240 Rodriguez LA, Watson IM, Edmonds M, Ryan G, Hards V, Oppenheimer C, Bluth GJS (2008) SO2 loss rates in the plume emitted by Soufriere Hills volcano, Montserrat. J Volcanol Geotherm Res 173:135-147 Roeder PL, Thornber C, Grant A (2004) Reply to comment on paper: “Morphology and composition of spinel in Pu’u ’O’o lava (1996-1998), Kilauea volcano, Hawaii”—enigmatic discrepancies between lava and gasbased fO2 determinations of Pu’u ’O’o lava. J Volcanol Geotherm Res 134:245-248 Rose WI, Chuan RL, Giggenbach WF, Kyle PR, Symonds RB (1986) Rates of sulfur dioxide and particle emissions from White Island volcano, New Zealand, and an estimate of the total flux of major gaseous species. Bull Volcanol 48:181-188 Rose WI, Stoiber RE, Malinconico LL (1982) Eruptive gas compositions and fluxes of explosive volcanoes: budget of S and Cl emitted from Fuego volcano, Guatemala. In: Andesite: orogenic andesites and related rocks. Thorpe RS (ed) Wiley & Sons. p 669-676 Rutherford MJ, Devine JD (1996) Pre-eruption pressure-temperature conditions and volatiles in the 1991 Mount Pinatubo magma. In: Fire and Mud: Eruptions and Lahars of Mount Pinatubo. Quezon City: Philippine Institute of Volcanology and Seismology; Seattle, WA. Newhall CG, Punongbayan RS (eds) University of Washington Press, p. 751-766 Rutherford MJ, Sigurdsson H, Carey S (1985) The May 1, 1980 eruption of Mount St Helens: 1, Melt composition and experimental phase equilibria. J Geophys Res 90:2929-2947 Saal AE, Hauri EH, Langmuir CH, Perfit MR (2002) Vapor undersaturation in primitive midocean ridge basalt and the volatile content of Earth’s upper mantle. Nature 419:451-455 Sakai H, Casadevall TJ, Moore JG (1982) Chemistry and isotope ratios of sulfur in basalts and volcanic gases at Kilauea Volcano, Hawaii. Geochim Cosmochim Acta 46:729-738 Salerno GG, Burton M, Oppenheimer C, Caltabiano T, Randazzo D, Bruno N, Longo V (2009b) Three-years of SO2 flux measurements of Mt. Etna using an automated UV scanner array: comparison with conventional traverses and uncertainties in flux retrieval. J Volcanol Geotherm Res 183:76-83. Salerno GG, Burton MR, Oppenheimer C, Caltabiano T, Tsanev VI, Bruno N, Randazzo D (2009a) Novel retrieval of volcanic SO2 abundance from ultraviolet spectra. J Volcanol Geotherm Res 181:141-153 Sarda P, Graham D (1990) Mid-ocean ridge popping rocks: implications for degassing at ridge crests. Earth Planet Sci Lett 97:268-289 Savarino J, Bekki S, Cole-Dai J, Thiemens MH (2003a) Evidence from sulfate mass independent oxygen isotopic compositions of dramatic changes in atmospheric oxidation following massive volcanic eruptions. J Geophys Res 108(D21), 4671, doi:10.1029/2003JD003737 Savarino J, Romero A, Cole-Dai J, Bekki S, Thiemens MH (2003b) UV induced mass-independent sulfur isotope fractionation in stratospheric volcanic sulphate. Geophys Res Lett 30(21), 2131, doi:10.1029/2003GL018134 Sawyer GM, Oppenheimer C, Tsanev VI, Yirgu G (2008a) Magmatic degassing at Erta ‘Ale volcano, Ethiopia. J Volcanol Geotherm Res 178:837–846 Sawyer GM, Carn SA, Tsanev VI, Oppenheimer C, Burton M (2008b) Investigation into magma degassing at Nyiragongo volcano, Democratic Republic of the Congo. Geochem Geophys Geosyst 9, Q02017, doi:101029/2007GC001829 Scaillet B, Clemente B, Evans BW, Pichavant M (1998) Redox control of sulfur degassing in silicic magmas. J Geophys Res 103:23937-23949 Scaillet B, Evans BW (1999) The 15 June 1991 eruption of Mount Pinatubo. I. Phase equilibria and pre-eruption P-T-fO2-fH2O conditions of the dacite magma. J Petrol 40:381-411.
418
Oppenheimer, Scaillet, Martin
Scaillet B, Luhr JF, Carroll MC (2003) Petrological and volcanological constraints on volcanic sulfur emissions to the atmosphere In: Volcanism and the Earth’s Atmosphere. Robock A, Oppenheimer C (eds) Geophysical Monograph 139:11-40 Scaillet B, Macdonald R (2006) Experimental and thermodynamic constraints on the sulphur yield of peralkaline and metaluminous silicic flood eruptions. J Petrol 47:1413-1437 Scaillet B, Pichavant M (2003) Experimental constraints on volatile abundances in arc magmas and their implications for degassing processes. In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc Lond Spec Pub 213:23-52. Scaillet B, Pichavant M (2004) Earth Science: Role of fO2 in fluid saturation of oceanic basalt. Nature 430, doi:10.1038/nature02814 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671-698 Schiano P, Clocchiatti R, Ottolini L (2001) Transition of Mount Etna lavas from a mantle plume to an island-arc magmatic source. Nature 412:900-904 Schneider DP, Ammann CM, Otto-Bliesner BL, Kaufman DS (2009) Climate response to large, high-latitude and low-latitude volcanic eruptions in the Community Climate System Model. J Geophys Res 114:D15101, doi:101029/2008JD011222 Self S, Blake S, Sharma K, Widdowson M, Sephton S (2008) Sulfur and chlorine in Late Cretaceous Deccan magmas and eruptive gas release. Science 319:1654-1657 Self S, R Gertisser, T Thordarson, M R Rampino, J A Wolff (2004) Magma volume, volatile emissions, and stratospheric aerosols from the 1815 eruption of Tambora. Geophys Res Lett 31:L20608, doi:101029/2004GL020925 Sharma K, Blake S, Self S, Krueger AJ (2004) SO2 emissions from basaltic eruptions, and the excess sulfur issue. Geophys Res Lett 31:L13612, doi:101029/2004GL019688 Shinohara H (2005) A new technique to estimate volcanic gas composition: plume measurements with a portable multi-sensor system. J Volcanol Geotherm Res 143:319-333 Shinohara H (2008) Excess degassing from volcanoes and its role on eruptive and intrusive activity. Rev Geophys 46:RG4005, doi:101029/2007RG000244 Shock EL, Holland M, Meyer-Dombard D’A, Amend JP, Osburn GR, Fischer TP (2010) Quantifying inorganic sources of geochemical energy in hydrothermal ecosystems, Yellowstone National Park, USA. Geochim Cosmochim Acta 74:4005-4043 Sigurdsson H, Carey S, Palais JM, Devine J (1990) Pre-eruption compositional gradients and mixing of andesite and dacite magma erupted from Nevado del Ruiz Volcano, Colombia in 1985. J Volcanol Geotherm Res 4:127-151 Smith WH (1990) Air Pollution and Forests: Interaction Between Air Contaminants and Forest Ecosystems. Springer-Verlag, New York, pp 618 Sortino F, Nonell A, Toutain JP, Munoz M, Valladon M, Volpicelli G (2006) A new method for sampling fumarolic gases: Analysis of major, minor and metallic trace elements with ammonia solutions. J Volcanol Geotherm Res 158:244-256 Spilliaert N, Allard P, Métrich N, Sobolev AV (2006) Melt inclusion record of the conditions of ascent, degassing, and extrusion of volatile-rich alkali basalt during the powerful 2002 flank eruption of Mount Etna (Italy). J Geophys Res 111(B4) B04203, doi: 10.1029/2005JB003934 Stenchikov G, Delworth TL, Ramaswamy V, Stouffer RJ, Wittenberg A, Zeng F (2009) Volcanic signals in oceans. J Geophys Res 114:D16104, doi:101029/2008JD011673 Stenchikov G, Hamilton K, Stouffer RJ, Robock A, Ramaswamy V, Santer B, Graf H-F (2006) Arctic Oscillation response to volcanic eruptions in the IPCC AR4 climate models. J Geophys Res 111:D07107, doi:101029/2005JD006286 Stenchikov G, Robock A, Ramaswamy V, Schwarzkopf M D, Hamilton K, Ramachandran S (2002) Arctic Oscillation response to the 1991 Mount Pinatubo eruption: effects of volcanic aerosols and ozone depletion, J Geophys Res 107(D24):4803, 101029/2002JD002090 Stevenson DS, Johnson CE, Collins WJ, Derwent RG (2003) The tropospheric sulphur cycle and the role of volcanic SO2. In: Volcanic Degassing (Oppenheimer C, Pyle DM, Barclay J (eds) Geol Soc Lond Spec Pub 213:295-305 Stith JL, Hobbs PV, Radke LF (1978) Airborne particle and gas measurements in the emissions from six volcanoes. J Geophys Res 83(C8):4009-4017 Stix J, Zapata JA, Calvache M, Cortes GP, Fischer TP, Gomez D, Narvaez L, Ordonez M, Ortega A, Torres R, Williams SN (1993) A model of degassing at Galeras volcano, Colombia, 1988–1993. Geology 21:963–96 Stoiber RE, Bratton G (1978) Airborne correlation spectrometer measurements of SO2 in eruption clouds of Guatemalan volcanoes. EOS Trans Am Geophys Union 59:122 Stoiber RE, Malinconico Jr, LL, Williams SN (1983) Use of the Correlation Spectrometer at Volcanoes. In: Forecasting Volcanic Events, Tazieff H, Sabroux J-C (eds), Elsevier, Amsterdam. p. 425-424
Sulfur Degassing From Volcanoes
419
Sutton AJ, Elias T (1993) Volcanic gases create air pollution in the Island of Hawaii. Earthquakes Volcanoes 24:178-196 Sutton AJ, Elias T, Gerlach TM, Stokes JB (2001) Implications for eruptive processes as indicated by sulfur dioxide emissions from Kilauea Volcano, Hawai’i, 1979-1997. J Volcanol Geotherm Res 108:283-302 Sutton AJ, Elias T, Kauahikaua J (2003) Lava-effusion rates for the Pu'u 'O'o-Küpaianaha eruption derived from SO2 emissions and very low frequency (VLF) measurements. USGS Prof Paper 1676:137-148 Swanson DA, Casadevall TJ, Dzurisin D, Malone SD, Newhall CG, Weaver CS (1983) Predicting eruptions at Mount St. Helens, June 1980 through December 1982. Science 221:1369-1376 Symonds RB, Reed MH (1993) Calculation of multicomponent chemical-equilibria in gas-solid-liquid systems: Calculation methods, thermochemical data, and applications to studies of high temperature volcanic gases with examples from Mount St Helens. Am J Sci 293:758-864 Symonds RB, WI Rose, GJS Bluth, TM Gerlach (1994) Volcanic gas studies: methods, results, and applications. Rev Mineral 30:1-60 Takano B, Koshida M, Fujiwara Y, Sugimori K, Takayanagi S (1997) Influence of sulfur-oxidizing bacteria on the budget of sulfate in Yugama crater lake, Kusatsu-Shirane volcano, Japan. Biogeochem 38:227-253 Thomas MA, Giorgetta MA, Timmreck C, Graf H-F, Stenchikov G (2009a) Simulation of the climate impact of Mt Pinatubo eruption using ECHAM5 - Part 2: Sensitivity to the phase of the QBO and ENSO. Atmos Chem Phys 9:3001-3009 Thomas MA, Timmreck C, Giorgetta MA, Graf H-F, Stenchikov G (2009b) Simulation of the climate impact of Mt Pinatubo eruption using ECHAM5 - Part 1: Sensitivity to the modes of atmospheric circulation and boundary conditions. Atmos Chem Phys 9:757-769 Thordarson T, Self S, Oskarsson N, Hulsebosch T (1996) Sulfur, chlorine, and fluorine degassing and atmospheric loading by the 1783-1784 AD Laki (Skaftar Fires) eruption in Iceland. Bull Volcanol 58:205-225 Timmreck C, Graf H-F (2006) The initial dispersal and radiative forcing of a Northern Hemisphere mid-latitude super volcano: a model study. Atmos Chem Phys 6:35-49 Timmreck C, Graf H-F, Lorenz SJ, Niemeier U, Zanchettin D, Matei D, Jungclaus JH, Crowley TJ (2010) Aerosol size confines climate response to volcanic super-eruptions. Geophys Res Lett 37:L24705, doi:10.1029/2010GL045464 Timmreck C, Lorenz SJ, Crowley TJ, Kinne S, Raddatz TJ, Thomas MA, Jungclaus JH (2009) Limited temperature response to the very large AD 1258 volcanic eruption. Geophys Res Lett 36:L21708, doi:101029/2009GL040083 Toutain JP, Baubron J-C, François L (2002) Runoff control of soil degassing at an active volcano: The case of Piton de la Fournaise, Réunion Island. Earth Planet Sci Let 197:83-94 Traversi R, Becagli S, Castellano E, Marino F, Rugi F, Severi M, de Angelis M, Fischer H, Hansson M, Stauffer M, Steffensen JP, Bigler M, Udisti R (2009) Sulfate spikes in the deep layers of EPICA-Dome C ice core: evidence of glaciological artifacts. Environ Sci Technol 43:8737-8743 Trenberth KE, Dai A (2007) Effects of Mount Pinatubo volcanic eruption on the hydrological cycle as an analog of geoengineering. Geophys Res Lett 34, L15702, doi:101029/2007GL030524 Viane C, Bhugwant C, Sieja B, Staudacher T, Demoly P (2009) Comparative study of the volcanic gas emissions and the hospitalizations for asthma of the Reunion island population between 2005 and 2007. Revue Française d’Allergologie 49:346-351 von Glasow R (2010) Atmospheric chemistry in volcanic plumes. Proc Natl Acad Sc 107:6594-6599 von Glasow R, Bobrowski N, Kern C (2009) The effects of volcanic eruptions on atmospheric chemistry. Chem Geol 262:131-142 Wallace PJ (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magma bodies. J Volcanol Geotherm Res 108:85-106 Wallace PJ (2002) Volatiles in submarine basaltic glasses from the Northern Kerguelen plateau (ODP site 1140): implications for source region compositions, magmatic processes, and plateau subsidence. J Petrol 43:1311-1326 Wallace PJ, Anderson AT (1998) Effects of eruption and lava drainback on the H2O contents of basaltic magmas at Kilauea. Bull Volcanol 59:327-344 Wallace PJ, Anderson AT, Davis AM (1995) Quantification of pre-eruptive exsolved gas contents in silicic magmas. Nature 377:612-5 Wallace PJ, Anderson AT, Davis AM (1999) Gradients in H2O, CO2, and exsolved gas in a large-volume silicic magma system: Interpreting the record preserved in melt inclusions from the Bishop Tuff. J Geophys Res 104(B9):20,097-20,122 Wallace PJ, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1683-1874 Wallace PJ, Edmonds M (2011) The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Rev Mineral Geochem 73:215-246 Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions: evidence from Mount Pinatubo. Science 265:497-499
420
Oppenheimer, Scaillet, Martin
Watson IM, Oppenheimer C, Voight B, Francis PW, Clarke A, Stix J, Miller A, Pyle DM, Burton MR, Young SR, Norton G, Loughlin S, Darroux B, MVO Staff (2000) The relationship between degassing and deformation at Soufriere Hills volcano, Montserrat. J Volcanol Geotherm Res 98:117-126 Watson IM, Realmuto VJ, Rose WI, Prata AJ, Bluth GJS, Gu Y, Bader CE, Yu T (2004) Thermal infrared remote sensing of volcanic emissions using the moderate resolution imaging spectroradiometer. J Volcanol Geotherm Res 135:75-89 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Webster JD, Sintoni MF, De Vivo B (2009) The partitioning behavior of Cl and S in aqueous fluid- and salineliquid saturated phonolitic and trachytic melts at 200 MPa. Chem Geol 263:19-36 Weibring P, Edner H, Svanberg S, Cecchi G, Pantani L, Ferrara R, Caltabiano T (1998) Monitoring of volcanic sulphur dioxide emissions using differential absorption lidar (DIAL), differential optical absorption spectroscopy (DOAS), and correlation spectroscopy (COSPEC). Appl Phys B 67:419-426 Weibring P, Swartling J, Edner H, Svanberg S, Caltabiano T, Condarelli D, Cecchi G, Pantani L (2002) Optical monitoring of volcanic sulphur dioxide emissions—comparison between four different remote-sensing spectroscopic techniques. Opt Lasers Eng 37:267-284 Weidmann D, Roller C, Tittel FK, Curl RF, Uehara K, Oppenheimer C, De Natale P (2003) Development of a 43 μm quantum cascade laser based 13CO2/12CO2 isotopic ratio sensor. Abstract, ISI 2003, Second International Symposium on Isotopomers, Stresa, 4-7 November 2003 Westrich HR, Gerlach TM (1992) Magmatic gas source for the stratospheric SO2 cloud from the June 15, 1991 eruption of Mount Pinatubo. Geology 20:867-70 Wignall PB (2001) Large igneous provinces and mass extinctions. Earth Sci Rev 53:1-33 Williams SN, Stoiber RE, Garcia NP, Londono AC, Gemmell JB, Lowe DR, Connor CB (1986) Eruption of the Nevado del Ruiz volcano, Colombia on 13 november 1985: gas flux and fluid geochemistry. Science 233:964-967. Williams-Jones G, Horton KA, Elias T, Garbeil H, Mouginis-Mark PJ, Sutton AJ, Harris AJL (2006) Accurately measuring volcanic plume velocity with multiple UV spectrometers. Bull Volcanol 68:328-332 Williams-Jones G, Stix J, Hickson C (eds) (2008) The COSPEC cookbook: making SO2 gas measurements at active volcanoes. IAVCEI Methods in Volcanology 1. Published online: http://www.iavcei.org/IAVCEI_ publications/COSPEC/COSPEC_Cookbook.html Winner WE, Mooney HA (1980) Responses of Hawaiian plants to volcanic sulfur dioxide: stomatal behavior and foliar injury. Science 210:789-791 Witham CS, Oppenheimer C (2005) Mortality in England during the 1783-4 Laki Craters eruption. Bull Volcanol 67:15-26 Witham CS, Oppenheimer C, Horwell CJ (2005) Volcanic ash-leachates: a review and recommendations for sampling methods. J Volcanol Geotherm Res 141:299-326 Witt MLI, Mather TA, Pyle DM, Aiuppa A, Bagnato E, Tsanev VI (2008) Mercury and halogen emissions from Masaya and Telica volcanoes, Nicaragua. J Geophys Res 113:B06203, doi:10.1029/2007JB005401 Witter JB, Self S (2007) The Kuwae (Vanuatu) eruption of AD 1452: potential magnitude and volatile release. Bull Volcanol 69:301-318 Wong T, Wielicki BA, Lee RB, Smith GL, Bush K (2006) Re-examination of the observed decadal variability of Earth Radiation Budget using altitude-corrected ERBE/ERBS nonscanner WFOV data. J Climate 19:4028-4040 Wright TE, Burton M, Pyle DM, Caltabiano T (2008) Scanning tomography of SO2 distribution in a volcanic gas plume. Geophys Res Lett 35: L17811, doi:10.1029/2008GL034640 Yamamoto H, Watson IM, Phillips JC, Bluth GJ (2008) Rise dynamics and relative ash distribution in vulcanian eruption plumes at Santiaguito Volcano, Guatemala, revealed using an ultraviolet imaging camera. Geophys Res Lett 35, L08314, doi:101029/2007GL032008 Yang K, Krotkov NA, Krueger AJ, Carn SA, Bhartia PK, Levelt PF (2007) Retrieval of large volcanic SO2 columns from the Aura Ozone Monitoring Instrument: Comparison and limitations. J Geophys Res 112:D24S43 doi:101029/2007JD008825 Yang K, Krotkov NA, Krueger AJ, Carn SA, Bhartia PK, Levelt PF (2009a), Improving retrieval of volcanic sulfur dioxide from backscattered UV satellite observations. Geophys Res Lett 36:L03102, doi:10.1029/2008GL036036 Yang K, Liu X, Krotkov NA, Krueger AJ, Carn SA (2009b) Estimating the altitude of volcanic sulfur dioxide plumes from space borne hyper-spectral UV measurements. Geophys Res Lett 36:L10803, doi:101029/2009GL038025 Zapata G, Calvache ML, Cortés GP, Fischer TP, Garzon V, Gómez MD, Narváez ML, Ordoñez VM, Ortega EA, Stix, J, Torres CR, Williams SN (1997) SO2 fluxes from Galeras volcano, Colombia 1989–1995: Progressive degassing and conduit obstruction of a Decade Volcano. J Volcanol Geotherm Res 77:195208
Sulfur Degassing From Volcanoes
421
Zhang Y (1999) H2O in rhyolitic glasses and melts: measurement, speciation, solubility, and diffusion. Rev Geophys 37:493-516 Zielinski GA, Mayewski PA, Meeker LD, Whitlow S, Twickler MS (1996) A 110,000-yr record of explosive volcanism from the GISP2 (Greenland) ice core. Quat Res 45:109-118 Zielinski GA, Mayewski PA, Meeker LD, Whitlow S, Twickler MS, Morrison M, Meese DA, Gow AJ Alley RB (1994) Record of volcanism since 7000 BC from the GISP2 Greenland ice core and implications for the volcano-climate system. Science 264:948-952
14
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 423-492, 2011 Copyright © Mineralogical Society of America
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, and Magmas Luigi Marini Laboratory of Geochemistry, Dip.Te.Ris University of Genova, Corso Europa 26, I-16132 Genova, Italy and Institute of Geosciences and Georesources, CNR Area della Ricerca, Via G. Moruzzi 1, I-56124 Pisa, Italy [email protected]
Roberto Moretti Dipartimento di Ingegneria Civile, Seconda Università degli Studi di Napoli Real Casa Santa dell’Annunziata Via Roma 29, I-81031 Aversa (CE), Italy and Istituto Nazionale di Geofisica e Vulcanologia, sezione di Napoli ‘Osservatorio Vesuviano’ Via Diocleziano 328, I-80124 Naples, Italy [email protected]
Marina Accornero D’Appolonia S.p.A, Via S.Nazaro 19, I-16145 Genova, Italy [email protected] Supplemental tables can be found at http://www.minsocam.org/MSA/RIM.
INTRODUCTION The first studies on sulfur isotope geochemistry were performed approximately sixty years ago (e.g., Thode et al. 1949; Trofimov 1949). Since these early measurements on terrestrial and extra-terrestrial materials, S isotopes have been used to highlight and investigate several processes, such as the evolution of the solar system; the oxidation of the Earth’s atmosphere and hydrosphere; the genesis of ore deposits and fossil fuels (coal, oil, and gases); the origin and provenance of S species in different natural fluids, including groundwater, rainwater, as well as present-day and ancient marine waters (as recorded by evaporite deposits); the S isotope fractionation in bacterially-mediated processes; the impact of anthropogenic activities, for instance mining and related acid drainage; and others. The sulfur isotopic compositions of volcanic rocks, magmatic gases, and closely related hydrothermal fluids were the subject of numerous investigations. Among these, a considerable contribution was provided by Sakai and coworkers, who elucidated the importance of degassing and sulfide separation and the different effects of these processes, depending on the redox state of the melt and, in the case of degassing, of the separated magmatic gases as well. All of these applications were made possible due to the experimental and theoretical works devoted to the determination of the fractionation factors between different S-bearing species and compounds and their temperature dependence. 1529-6466/11/0073-0014$10.00
DOI: 10.2138/rmg.2011.73.14
Marini, Moretti, Accornero
424
This review is devoted to the magmatic-hydrothermal environment as a whole, focusing on active systems and their past analogues, which are represented by different types of ore deposits. Special emphasis is given to the use of S isotopes to investigate the effects of degassing and separation of sulfide and sulfate minerals from silicate melts. For the sake of clarity, we recall the meaning of the following terms: •
Fluid is a general term including both liquids and gases.
•
Vapor is a substance in the gas phase at a temperature lower than its critical point. Therefore the vapor can be condensed to a liquid or to a solid by increasing its pressure without reducing the temperature.
•
Steam is the gas phase of water (water vapor).
BASIC PRINCIPLES Terminology of sulfur isotope systematics Sulfur (S) has 25 isotopes, four of which are stable, 32S, 33S, 34S, and 36S. All the radioactive isotopes of sulfur are short-lived, apart from 35S, which is formed from cosmic ray spallation of 40Ar in the atmosphere and has a half-life of 87.51 days. The atomic masses of the stable isotopes of sulfur are A32S = 31.972 070 73(15) amu, A33S = 32.971 458 54(15) amu, A34S = 33.967 866 87(14) amu, and A36S = 35.967 080 88(25) amu, according to the Commission on Atomic Weights and Isotopic Abundances, CAWIA, of the International Union of Pure and Applied Chemistry, IUPAC (Coplen et al. 2002). The sulfur representative isotopic composition (Atom %) is 32S = 94.93(31), 33S = 0.76(2), 34S = 4.29(28), 36S = 0.02(1) (Rosman and Taylor 1997). Physical processes inducing isotope fractionation, such as gaseous diffusion of molecules or ions and ultrafiltration, are characterized by mass-dependent rates of diffusion. Isotope fractionation caused by chemical processes can be grouped in the two following types: (i) equilibrium isotope fractionation, when the rate of the forward isotope-exchange reaction is equal to that of the corresponding backward reaction; (ii) kinetic isotope fractionation, which is determined by unidirectional reactions whose rates are generally mass dependent. Usually, in equilibrium isotope reactions, the heavier isotope is concentrated in the ions, atoms, or molecules with the higher oxidation state and/or in the more condensed state (e.g., Urey 1947; Tudge and Thode 1950). For instance, 34S is enriched in sulfate relative to sulfite, in sulfite relative to elemental sulfur, and in sulfur relative to sulfide; besides 18O and 2H are generally enriched in liquid water relative to water vapor. However, this rule is not always true in that the 34S fractionation between sulfate minerals and dissolved sulfate is so small that it is usually neglected (Kusakabe and Robinson 1977). In chemical processes causing kinetic isotope fractionation, the lighter isotope is more reactive and is enriched in reaction products, as predicted by statistical mechanics. Consequently, the heavier isotope is enriched in reactants. Many biologically mediated reactions cause kinetic isotope fractionation, which has variable magnitude and may even be in the direction opposite to that expected for equilibrium isotope fractionation between the same chemical entities. The distribution of isotopes in two entities X and Y is described by the isotope fractionation factor αX,Y. For sulfur, the isotope ratio 34S/32S is generally adopted: α X-Y
( S = ( S 34
32
34
32
) S) S
X Y
(1)
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 425 Variations of abundance ratios of stable isotopes are generally small. Therefore, for describing stable isotope ratios, it is convenient to use the delta-notation, in per-mil (also spelled per mil, permil, or per mille) units (‰), which is defined by the following equation: ( 34 S / 32 S)sample − ( 34 S / 32 S)standard = δ 34 S(‰) × 1000 ( 34 S / 32 S)standard
(2)
Since 1962, the internationally accepted standard of reference has been the troilite (FeS, named for Domenico Troili, who described, in 1766, a meteorite which fell near Modena, Italy) contained in the Canyon Diablo meteorite, with a 34S/32S isotope ratio of 1/22.22 (Jensen and Nakai 1962), although the abundances of 32S (95.02%) and 34S (4.21%) reported by McNamara and Thode (1950) are consistent with a 34S/32S ratio of 1/22.57. Thirty years later, the IAEA advisory group on reference materials for stable isotope measurements suggested a V-CDT (Vienna CDT) scale, defined by the reference material IAEA-S-1 (Ag2S previously known as NZ-1), which is characterized by a δ34SV-CDT value of −0.3‰ (Robinson 1993). Absolute 32S/33S and 32S/34S ratios of V-CDT, IAEA-S-1 and other two reference materials (known as IAEA-S-2 and IAEA-S-3) were determined by Ding et al. (2001) and are reported in Table 1. In principle, data produced after the definition of the V-CDT scale should be reported as δ-values with respect to V-CDT, whereas data obtained before should be reported as δ-values with respect to CDT. In practice, there is a negligible to nil difference between the two scales, since a special effort was made to have the isotope composition of V-CDT identical to that of CDT (Robinson 1993). Consequently, all the data are reported as δ-values with respect to V-CDT in this review. Table 1. Absolute 32S/33S and 32S/34S ratios of the V-CDT standard and the IAEA-S-1, IAEA-S-2, and IAEA-S-3 reference materials (from Ding et al. 2001). Standard / Reference Material
32
S/33S
32
S/34S
V-CDT
126.948 (±0.047)
22.6436 (±0.0020)
IAEA-S-1
126.942 (±0.047)
22.6504 (±0.0020)
IAEA-S-2
125.473 (±0.055)
22.1424 (±0.0020)
IAEA-S-3
129.072 (±0.032)
23.3933 (±0.0017)
δ33S and δ36S values The δ33S and δ36S values of terrestrial materials were the subject of a limited number of investigations owing to both: (i) experimental difficulties, caused by the low abundances of 33S and 36S, and (ii) acceptance of the theory of mass-dependent isotope fractionation (Bigeleisen and Mayer 1947), which predicted the following linear relations among different δ values: δ33S = 0.515 × δ34S
(3)
δ36S = 1.89 × δ34S
(4)
making the determination of δ S and δ S values useless. 33
36
Recently, the usefulness of these measurements for terrestrial materials was understood when Farquhar et al. (2000) found that some sulfide and sulfate minerals hosted in sedimentary rocks older than ~2.0 Ga do not obey the mass-dependent fractionation Equations (3) and (4). Farquhar et al. (2001) underscored that SO2 photolysis generated elemental S and SO42− ions with strong mass-independent fractionation of sulfur isotopes, and suggested that the presence
Marini, Moretti, Accornero
426
or absence of this signature in sedimentary rocks was closely linked with the evolution of atmospheric oxygen. In spite of the considerable interest of these recent findings and other related results of other authors, δ33S and δ36S values will not be considered in the following discussion.
Equilibrium fractionation factors Let us take into account the isotope exchange reaction: H234S + 32SO2 = H232S + 34SO2
(5)
whose concentration-based equilibrium constant is: K=
[34 SO 2 ] [32SO 2 ] [H 234S] [H 232S]
(6)
In the case of Equation (5), the fractionation factor, αSO2-H2S, is given by the relation: (34 S / 32 S)SO2 δ34SSO2 + 1000 αSO= = 2 − H2S (34 S / 32 S) H2S δ34SH2S + 1000
(7)
In the case of a pair of simple molecules, containing one atom of the isotope-exchanging element, α is equal to K, but when the number of atoms of the element of interest, n, is different from 1, α = K1/n. Since α is normally very close to unity, the parameters 1000⋅lnα and ε = (α−1)⋅1000 are usually used in the pertinent literature instead of α. It is easy to verify that ε ≈ 1000⋅lnα, as long as α−1 is less than 0.01-0.02. Referring again to the SO2-H2S case (Eqn. 5), it can be easily shown that: δ34SSO + 1000 εSO2 -H2 S =(αSO2 -H2 S − 1) × 1000 = 34 2 − 1 × 1000 δ SH S + 1000 2 34 34 34 ≈ δ SSO2 − δ SH2 S = Δ SSO2 -H2 S ,
(8)
at least when δ34SH2S << 1000. Therefore, the following approximated relation is valid when the difference between the delta values is small: 1000 × ln αSO2 -H2S ≈ ε ≈ δ34SSO2 − δ34SH2S = Δ 34SSO2 -H2S
(9)
Concentration-based equilibrium constants of isotope exchange reactions may be formulated in terms of partition functions, Q, of the involved isotopic molecules. For instance, in the case of the SO2-H2S isotope exchange reaction (Eqn. 5): K =
[ 34 SO2 ] [ 32SO2 ] Q 34SO2 Q 32SO2 = [H 2 34S] [H 2 32S] QH 2 34S QH 2 32S
(10)
Since ratios of partition functions depend on their fundamental vibrational frequencies only, the equilibrium constant of the SO2-H2S isotope exchange reaction (Eqn. 5) can be easily obtained by means of Equation (10). The foundations of this theoretical approach were laid down by Urey and Greiff (1935), Urey (1947) and Bigeleisen and Mayer (1947) with further contributions and developments by other authors including Tudge and Thode (1950), Sakai (1957, 1968), and Thode et al. (1971). For further details on the theoretical prediction of isotope equilibrium constants and adopted strategies (e.g., the Born-Oppenheimer approximation) see Richet et al. (1977) and references therein. To tabulate fractionation factors, it is useful to refer to exchange reactions between a given molecule and the isotope-exchanging element, such as:
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 427 32
SO2 + 34S = 34SO2 + 32S
(11)
H232S + 34S = H234S + 32S
(12)
whose fractionation factors, called β factors, are: [34SO ] [32SO ] βSO2 -S =34 2 32 2 [ S] [ S]
(13)
[H 2 34S] [H 2 32S] βH2 S-S = [ 34S] [ 32S]
(14)
In terms of β factors, Equation (7) can be rewritten as: βSO2 -S αSO2 -H2S = βH2S-S
(15)
Based on the β factors reported by Richet et al. (1977) for several gaseous molecules (Table 2), the fractionation factor αSO2-H2S, that is the isotope equilibrium constant of the SO2-H2S exchange reaction (Eqn. 5), decreases from 1.0338 at 20 °C to 1.0013 at 1300 °C. Therefore, at isotope equilibrium, the δ34SSO2 value will be heavier than the δ34SH2S value, by 33.8‰ at 20 °C and by 1.3‰ at 1300 °C. Table 2. β factors for sulfur exchange (from Richet et al. 1977). Temperature (°C)
SO3
SO2
COS
CS2
CS
S2
H2S
20 100 150 200 250 300 350 400 450 500 600 700 800 900 1100 1300
1.0834 1.0573 1.0467 1.0388 1.0327 1.0279 1.0241 1.0210 1.0185 1.0164 1.0132 1.0108 1.0091 1.0077 1.0059 1.0047
1.0456 1.0314 1.0256 1.0212 1.0179 1.0152 1.0131 1.0114 1.0100 1.0088 1.0070 1.0057 1.0047 1.0039 1.0029 1.0022
1.0188 1.0125 1.0100 1.0082 1.0069 1.0058 1.0050 1.0043 1.0037 1.0033 1.0026 1.0021 1.0018 1.0015 1.0011 1.0008
1.0198 1.0133 1.0107 1.0089 1.0074 1.0063 1.0055 1.0048 1.0042 1.0037 1.0030 1.0024 1.0020 1.0017 1.0013 1.0010
1.0173 1.0121 1.0099 1.0082 1.0069 1.0059 1.0051 1.0044 1.0038 1.0034 1.0027 1.0022 1.0018 1.0015 1.0011 1.0008
1.0130 1.0085 1.0067 1.0054 1.0045 1.0038 1.0032 1.0028 1.0024 1.0021 1.0017 1.0013 1.0011 1.0009 1.0007 1.0005
1.0114 1.0084 1.0071 1.0061 1.0053 1.0047 1.0041 1.0037 1.0033 1.0030 1.0024 1.0020 1.0017 1.0015 1.0011 1.0009
It is evident from Table 2 that isotope fractionation depends on the oxidation state of sulfur in the considered gaseous molecules as well as on temperature. Other factors influencing the extent of equilibrium isotope fractionation are pressure, crystal structure, and cationic substitution. However, the pressure effect is generally negligible, at least for pressures lower than 10 kbar. The temperature dependence of the equilibrium isotope fractionation factors is often expressed through polynomial functions of 1/T (K−1) of the following type (e.g., Ohmoto and Rye 1979): 1000 = ln α
A × 106 B × 103 + +C T2 T
(16)
428
Marini, Moretti, Accornero
Several experimental studies were performed to investigate the equilibrium fractionation factors between sulfur compounds and the order of 34S enrichment obtained through theoretical prediction was confirmed experimentally. Conversely, there are some inconsistencies between the fractionation factors derived by different authors. For example, considering the sphaleritegalena couple, Grootenboer and Schwarcz (1969) suggested experimental fractionation factors fitting the relation (T in K): 0.63 × 106 1000ln α ZnS-PbS =2 T
(17)
whereas Kajiwara et al. (1969) proposed higher values, defining the equation (T in K): 0.90 × 106 1000 ln α ZnS-PbS =2 T
(18)
These and similar disagreements between fractionation factors proposed by different authors are chiefly due to both (i) production of experimental data for a limited number of temperatures only and (ii) the influence of low-temperature values, which are usually affected by large uncertainties as isotope equilibrium is not easily attained under these conditions. Starting from these premises, Ohmoto and Rye (1979) carried out a critical review of available raw experimental data (Grootenboer and Schwarcz 1969; Kajiwara et al. 1969; Rye and Czamanske 1969; Grinenko and Thode 1970; Schiller et al. 1970; Kajiwara and Krouse 1971; Salomons 1971; Thode et al. 1971; Kiyosu 1973; Robinson 1973; Czamanske and Rye 1974; Igumnov 1976; Igumnov et al. 1977; Smith et al. 1977; Sakai and Dickson 1978). They took into account the following points: (i) attainment of isotope equilibrium; (ii) measurement errors; (iii) minimum and maximum fractionation factors, if isotope equilibrium was not attained and (iv) compatibility with the fractionation factors evaluated from other experimental datasets. Available data may be used to evaluate scarcely known fractionation factors. For example, it is possible to combine the fractionation factors for the couples HSO4−(aq)-FeS2 (Nakai 1970), HSO4−(aq)-S(g) (e.g., Robinson 1973), and PbS-S(g) (Grootenboer and Schwarcz 1969) to obtain the FeS2-PbS fractionation factors; these are expected to be in agreement, within experimental uncertainties, with those directly measured by Kajiwara and Krouse (1971). In addition, it is possible to use the fractionation factors for the couples ZnS-S(g) (Grootenboer and Schwarcz 1969), H2S-S(g) (Grinenko and Thode 1970), and ZnS-HS−(aq) (Kiyosu 1973) to obtain the HS−(aq)-H2S fractionation factors, which were the subject of very few direct determinations. Finally, the order of 34S enrichment theoretically predicted for several sulfur compounds (Sakai 1968; Bachinski 1969) may be utilized to constrain fractionation factors in the absence of experimental results. Ohmoto and Lasaga (1982) re-evaluated the fractionation of sulfur isotopes between dissolved sulfate and sulfide taking into consideration the experimental data by Robinson (1973), Bahr (1976), Igumnov et al. (1977), and Sakai and Dickson (1978). Some of the experimental investigations carried out afterwards were reviewed by Seal (2006), who estimated the fractionations factors with respect to H2S for: (i) argentite (Ag2S), covellite (CuS), and digenite (Cu9S5), based on the results obtained by Hubberten (1980); and (ii) bismuthinite (Bi2S3), on the basis of the data produced by Bente and Nielsen (1982). Few experimental measurements were carried out to constrain the fractionation factors between sulfate minerals and dissolved sulfate (e.g., Ault and Kulp 1959; Thode et al. 1961; Thode and Monster 1965; Holser and Kaplan 1966; Taylor et al. 1984; Raab and Spiro 1991), obtaining values from 0 to +2.4‰ for the gypsum-aqueous sulfate pair at room temperatures. However, Ohmoto and Rye (1979), Ohmoto and Lasaga (1982), and Ohmoto and Goldhaber
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 429 (1997) assumed that there is negligible fractionation among sulfate minerals and dissolved sulfate above 100 °C, as suggested by Sakai (1968). The equilibrium fractionation factors with respect to H2S of the sulfur compounds mentioned above are listed in Table A-1 and some are depicted in Figure 1. The theoretical fractionation factors for sulfide minerals calculated by Li and Liu (2006) are also reported in Table A-1.
Figure 1. Temperature dependence of the equilibrium isotope fractionation factors for selected sulfur compounds and species with respect to H2S, either gaseous or aqueous (data from Ohmoto and Rye 1979, unless otherwise indicated). Solid lines refer to experimentally determined data, whereas dashed lines identify theoretically predicted data.
Of particular interest for the present review are the sulfur isotope fractionation factors between sulfate and sulfide which were determined experimentally by Miyoshi et al. (1984) through decomposition of anhydrous Na2SO3 in melts of NaCl and a LiCl-KCl composition, at temperatures of 600 to 1000 °C. Obtained results fit the equation: 1000ln α sulfate-sulfide =
7.4 × 106 − 0.19 T2
(19)
which describes the temperature variation, in the experimental interval, of the fractionation factor between bulk sulfate (comprising SO42− and NaSO4−, in the NaCl melt) and bulk sulfide (including HS−, S2−, NaHS and NaS−, in the NaCl melt). The fractionation factor between dissolved sulfate and gaseous H2S in the temperature range 600-1000 °C is closely approximated by the following relationship: 6.5( ±0.3) × 106 1000 ln αsulfate-H2 S = 2 T
(20)
Marini, Moretti, Accornero
430 Isotope geothermometry
In principle, the equilibrium isotope fractionation factors between two substances of interest can be used to estimate the temperature at which they equilibrate. In practice, there is no certainty on the attainment of isotope equilibrium between the two compounds in all the systems of interest. Specific pressure and temperature conditions as well as the presence of other reactants and/or catalysts are usually needed to promote a certain reaction and the possible attainment of chemical equilibrium. For instance, in the case of the SO2-H2S isotope exchange reaction (Eqn. 5), both H2O and high temperatures promote the progress towards equilibrium (e.g., Thode 1991). Otherwise, the system may acquire isotopic characteristics intermediate between those of the initial state and those expected for final equilibrium, at a given temperature. In spite of these uncertainties, the temperature-dependence of sulfur isotope fractionation between either two minerals or two aqueous species (e.g., SO42−(aq) and H2S(aq)) has been widely used for sulfur isotope geothermometry (e.g., Ohmoto and Rye 1979; Seal 2006 and references therein).
SULFUR ISOTOPIC COMPOSITION OF NATURAL SAMPLES Mantle-derived materials and igneous rocks Sulfide inclusions in mantle-derived materials, comprising primitive igneous rocks, mantle xenoliths, and diamonds provide indications on the δ34S value of the mantle which is the source of most magmas that reach the crust (Fig. 2). Early studies (e.g., Vinogradov 1958; Thode et al. 1961) recognized that the sulfur isotopic composition of the mantle is close to 0 ± 2‰ and similar to that of meteoritic sulfur. A review of available data carried out by Seal (2006) pointed out the following average δ34S values and related standard deviations (data from Sakai et al.
Figure 2. Range of sulfur isotope values for sulfides from meteorites, mantle xenoliths, diamonds, igneous rocks and modern sediments (data from Sasaki and Ishihara 1979; Chambers 1982; Rye et al. 1984; Sakai et al. 1984; Chaussidon et al. 1987, 1989; Ishihara and Sasaki 1989; Torssander 1989; Eldridge et al. 1991; Santosh and Masuda 1991; Sælen et al 1993; Strauss 1997; Farquhar et al. 2002; Luhr and Logan 2002). The δ34S value of dissolved sulfate in present-day oceans is also shown (Rees et al. 1978).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 431 1984; Chaussidon et al. 1987, 1989; Torssander 1989; Eldridge et al. 1991; Farquhar et al. 2002): (i) +1.3 ± 3.8‰ for sulfide inclusions in mantle xenoliths; (ii) −0.3 ± 2.3‰ for sulfides in mid-ocean ridge basalts (MORB); (iii) +1.0 ± 1.9‰ for sulfides in ocean island basalts (OIB); (iv) +1.2 ± 5.6‰ for sulfide inclusions in diamonds, although the range is very large, from −11 to +14‰. In contrast, an extensive analytical study of sulfur concentrations and isotope ratios in ultramafic xenoliths from Central Asia (Ionov et al. 1992) shows that the continental lithospheric mantle may be characterized by sulfur concentrations <50 ppm and positive δ34S values, up to 7‰. As regards the sulfide inclusions in diamonds, there is a remarkable difference between peridotitic diamonds and eclogitic diamonds. The first ones are strictly relatable to the mantle and are characterized by sulfide inclusions with δ34S values in the restricted range from −5 to +5‰. The second ones are derived from biogenic carbon and contain small amounts of biogenic sulfur, as suggested by δ34S values defining the whole range recalled above, i.e., from −11 to +14‰. According to Farquhar et al. (2002) these elements were introduced into the mantle through the subduction process in the Archean. Sulfur isotope data also imply that this element is not well mixed in the source regions of diamonds, permitting reconstruction of the Archean sulfur cycle and possibly giving insight into the nature of mantle convection through time. Interestingly, the whole-rock δ34S values of continental and island arc basalts and gabbros, +1.0 ± 3.2‰, are superimposed to those of MORB and OIB, whereas andesites exhibit weakly higher δ34S values, +2.6 ± 2.3‰ (Rye et al. 1984; Luhr and Logan 2002). The average sulfur isotope composition of granitoid rocks, +1.0 ± 6.1‰, is not very different from the meteoritic value, but their δ34S values distribute in a large range, from −11 to +14.5‰, probably due to assimilation and/or partial melting of either sedimentary sulfides with low δ34S values or evaporite sulfates with high δ34S values (Sasaki and Ishihara 1979; Ishihara and Sasaki 1989; Santosh and Masuda 1991).
Sulfide and sulfate minerals from magmatic, magmatic-hydrothermal and related ore deposits Magmatic sulfide deposits. The geochemistry of sulfur and sulfur isotopes in magmatic sulfide deposits has been reviewed by Ohmoto (1986), Taylor (1987) and Ripley and Li (2003). This type of ore deposits is considered to be produced through separation of immiscible sulfide liquids during the crystallization of magma batches (Simon and Ripley 2011, in this volume). Examples of these ore deposits are the Cu-Ni mineralizations related to mafic magmas. These can be further distinguished as sulfur-rich magmatic systems, representing important resources of Cu and Ni, versus sulfur-poor magmatic systems, which are mined for platinum-groupelements, i.e., PGE (Ripley and Li 2003). Several aspects of both magmatic Cu-Ni-(PGE) deposits and PGE deposits (characteristics and classification, resource and grade, processes involved in ore genesis, source magmas, transport of sulfide liquids) are presented by Simon and Ripley (2011, this volume). The δ34S values of sulfide minerals contained in sulfur-poor magmatic systems, such as the Merensky Reef of the Bushveld Complex, South Africa, and the J-M Reef of the Stillwater Complex, Montana, span a restricted interval centered on the mantle value, 0‰ (Buchanan et al. 1981; Zientek and Ripley 1990; Ripley and Li 2003). In contrast, the sulfur isotope composition of sulfide minerals contained in sulfur-rich magmatic systems, such as the Duluth Complex, Minnesota, and Noril’sk, Russia, span a wide interval with a positive average. These characteristics were attributed to crustal contamination of magma (Ripley and Al-Jassar 1987; Taylor 1987; Li et al. 2003; Ripley et al. 2003). However, where the country rocks have δ34S values close to zero, as is the case of the Archean metasedimentary rocks situated in the footwall of the Sudbury deposit, Ontario, it is difficult to assess the role of crustal contamination (Thode et al. 1962; Schwarcz 1973).
432
Marini, Moretti, Accornero
Porphyry copper deposits. The sulfur isotope geochemistry of the hydrothermal systems of this type was reviewed by Ohmoto (1986), Seal et al. (2000), and Rye (2005). Several aspects of porphyry ore deposits (tectonic setting, composition and oxidation state of associated magmas, origin of S and Cu, composition of ore-forming fluids, source of fluid components, processes governing the deposition of metal sulfides) are thoroughly discussed by Simon and Ripley (2011 this volume). Porphyry deposits are characterized by large tonnages and low grades and represent important resources of Cu, Mo, Au, Ag, etc. They are generated by hydrothermal fluids which, in turn, are mixtures of fluids largely released from granitic magmas with lesser meteoric waters. Hydrothermal systems related to fluids of comparatively low pH and relatively oxidizing magmas, such as I-type granitoids, usually exhibit significant variations of sulfur isotopes, owing to the presence of both SO2 and H2S, in nearly equal amounts, in these kinds of magmatic fluids. In contrast, the isotopic characteristics of the deposits associated with S-type granitoids are governed by the dominance of H2S in parent hydrothermal fluids (Burnham and Ohmoto 1980). A powerful tool for the interpretation of the δ34S values of coexisting sulfate and sulfide minerals is represented by Figure 3. The basics needed for the construction of this type of diagram are presented by Fifarek and Rye (2005). These δsulfate-δsulfide plots can provide indications on the temperature of isotope equilibrium, the bulk sulfur isotope values (δ34SΣS), and the SO42−(aq)/H2S ratio of the parent fluids if the following conditions are satisfied: (1) equilibrium was attained between dissolved sulfate and sulfide species, (2) the SO42−(aq)/H2S ratio of the parent fluids remained constant during mineral precipitation, (3) the minerals did not experience post-depositional isotope exchange. Based on these hypotheses: (i) the isotope equilibrium temperature of mineral pairs is given from a series of lines of unit slope; (ii) the SO42−(aq)/H2S molar ratio of the parent fluids can
Figure 3. Plot of δ34SSulfate vs. δ34SPyrite values for mineral assemblages from selected porphyry copper deposits (data from Field and Gustafson 1976; Eastoe 1983; and Field et al. 2005, see text and supplemental Table E-1). Also shown are the equilibrium fractionation lines at selected temperatures in the 250-700 °C range (see Table A-1) and the lines for δ34SΣS = 0 and δ34SΣS = +5‰ and variable SO42−/H2S ratio of the parent fluids.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 433 be inferred from the lines of variable negative slopes, comprised between the horizontal lines (SO42−(aq)/H2S = 0) and the vertical lines (SO42−(aq)/H2S = ∞); (iii) the δ34SΣS can be evaluated from the intersection of a given alignment of data with the line of unit slope characterized by nil sulfate-sulfide fractionation (Δ = 0‰). If these conditions are not satisfied, the δsulfate-δsulfide diagrams may be used to interpret the consequences of disequilibrium relations. The δ34S values of coexisting sulfate and sulfide minerals from some porphyry copper ore deposits are listed in supplemental Table E-1 and those of sulfate-pyrite pairs are plotted in Figure 3. If the hypotheses listed above are satisfied, data plotted in Figure 3 indicate that the δ34SΣS values of these porphyry copper magmatic-hydrothermal systems may be close to +5‰ approximately, whereas sulfate-pyrite equilibrium temperatures are generally in the 300-650 °C interval, in agreement with independent evidence provided by fluid inclusion studies (Rye 2005). Also shown in Figure 3 is the variability of the SO42−(aq)/H2S molar ratio of the parent fluids. Although these sulfate-pyrite data are well correlated and define linear trends, scattered samples are sometimes observed, probably due to retrograde re-equilibration (Rye 2005). Epithermal deposits. The sulfur isotope geochemistry of the hydrothermal systems of this type was reviewed by Ohmoto (1986), Seal et al. (2000), and Rye (2005). Epithermal deposits mineralized in Cu, Au, Ag, Pb, and Zn comprise both adularia-sericite deposits and acid-sulfate deposits (Heald et al. 1987), often representing the uppermost portions of porphyry copper hydrothermal-magmatic systems. In addition the hydrothermal fluids originating epithermal deposits are mixed magmatic-meteoric waters, similar to those of porphyry copper deposits (Rye 1993). Isotopic data of alunite and coexisting pyrite are compiled for acid-sulfate epithermal deposits (see Fig. 4 and supplemental Table E-2). In contrast to porphyry copper deposits, sulfate-sulfide mineral pairs from epithermal deposits indicate lower equilibrium temperatures, chiefly situated in the 150-300 °C range
Figure 4. Plot of δ34SAlunite vs. δ34SPyrite values from selected acid-sulfate epithermal deposits (data from Rye et al. 1992, Arribas et al. 1995, Bethke et al. 2005, Fifarek and Rye 2005, Juliani et al. 2005, and Rainbow et al. 2005; see text and supplemental Table E-2). Also shown are the equilibrium fractionation lines at selected temperatures in the 200-700 °C range (see Table A-1) and the lines for δ34SΣS = 0 and δ34SΣS = +5‰ and variable SO42−/H2S ratio of the parent fluids.
434
Marini, Moretti, Accornero
(Hedenquist and Lowenstern 1994), which are in line with those evaluated by means of different approaches, e.g., fluid inclusions. In addition, for most epithermal deposits, it is difficult to constrain the δ34SΣS values of parent hydrothermal fluids due to the spread of data and lack of linear trends in the δsulfate-δsulfide plot (Fig. 4), which are usually due to occurrence of steam separation (boiling), causing substantial changes in both the concentrations of volatile sulfur species and the δ34SΣS values of parent hydrothermal fluids. Modern seafloor hydrothermal systems. The stable isotope geochemistry of these systems located along mid-ocean ridges has been reviewed by Seal et al. (2000) and Shanks (2001). The heat released from magmatic activity along the submarine spreading centers promotes hydrothermal convection of seawater through a large volume of rocks, comprising igneous and sedimentary lithotypes. δ34S values are available for several sulfide minerals, comprising pyrite, marcasite, pyrrhotite, chalcopyrite, sphalerite, wurtzite, bornite (Cu5FeS4), and cubanite (CuFe2S3), as well as for H2S in vent fluids. Dissolved sulfate may be largely removed from marine water in the downwelling leg of these circuits, in response to its progressive heating, owing to the retrograde solubility of anhydrite and other sulfate minerals (Bischoff and Seyfried 1978; Seyfried and Bischoff 1981). Sulfide minerals and H2S of vent fluids in modern seafloor hydrothermal systems are generated by the following processes (Janecky and Shanks 1988; Shanks 2001): (i) simple adiabatic mixing of seawater and hydrothermal fluid at the seafloor, which produces H2S with maximum δ34S values close to +4.5‰; (ii) thermochemical reduction of marine sulfate (before its removal by precipitation), through interaction with magnetite and Fe(II)-silicates, which generates H2S with δ34S values up to +15‰; (iii) dissolution of biogenic sulfides contained in sedimentary rocks (previously generated through bacterial sulfate reduction), which produces H2S with negative δ34S values. In most mid-ocean ridge systems characterized by an absence or scarcity of sediments, H2S of vent fluids has δ34S values ranging from 0 to +6‰, whereas systems with sediments show wider ranges of δ34S values. Thermochemical reduction of marine sulfate, found to occur in systems affected by shallow seawater entrainment, leads to higher δ34S values of vent-fluid H2S and, consequently, precipitating sulfide minerals. Low-pH fluids, probably reflecting disproportionation of SO2, occur in back-arc seafloor hydrothermal systems (Herzig et al. 1993; Gamo et al. 1997), which can be considered a sort of equivalent of acid-sulfate epithermal systems (see previous section). Reaction path models of seawater-basalt interaction at mid-ocean ridges were developed by Bowers and Taylor (1985) and extended to include sulfur isotopic exchange equilibria by Bowers (1989). The effects of variable contribution of S from basalt and of high- and lowtemperature reduction of sulfate in the fluids were considered. Two limiting cases were taken into account, assuming either equilibrium between sulfate and sulfide species or disequilibrium. In the second case sulfate and sulfide are treated as two distinct elements, with individual mass balances and distinct isotopes; consequently, an initial δ34S value for both sulfate and sulfide in both the starting solution and the reactants must be specified in reaction path modeling.
Sulfur isotopes in magmatic and magmatic-hydrothermal systems Volcanic gases. High-temperature (>500 °C) gas mixtures discharged from fumaroles typically found in the craters of active volcanoes are mainly constituted of H2O, CO2, SO2, and H2S, in decreasing order of importance (e.g., Symonds et al. 1994; Oppenheimer et al. 2011, this volume). In addition to these constituents, fumarolic gases emitted from volcanoes situated along convergent plate boundaries are also generally richer in HCl with respect to volcanic gases from divergent plate boundaries and hot spots (Symonds et al. 1994; Giggenbach 1997). Gas speciation in fumarolic gases can be predicted with equilibrium equations like: H2S(g) +3/2 O2 (g) = SO2(g) + H2O(g)
(21)
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 435 whose thermodynamic equilibrium constant can be rearranged as follows: log
fSO2 fH 2 S
XSO2 3 log K − log fH2 O + log fO2 ≅ log = 2 X H2 S
(22)
where fi and Xi indicate fugacity and mole fraction of the i-th gas, respectively. The fugacity ratio fSO2 /fH2S is nearly equal to the mole fraction ratio XSO2 /XH2S as long as the fugacitycoefficient-ratio is close to 1. This approximation is certainly valid at high temperatures and low total pressures, where real gases approach ideal-gas behavior. However, since the critical point of H2S (TC = 100 °C, PC = 89.37 bar) is not very different from that of SO2 (TC = 157.6 °C, PC = 78.84 bar), it is reasonable to extend this approximation to relatively low temperatures and comparatively high total pressures. The temperature dependence of log K of Equation (21) is (thermodynamic data from Stull et al. 1969; temperature in K): log= K (21)
27377 − 3.986 T
(23)
The equilibrium fO2 values calculated for high-temperature volcanic gases from different platetectonic settings range between two log units above the NNO buffer and one log unit below the QFM buffer (Symonds et al. 1994). Computed fO2 values agree with in situ fO2 measurements carried out at some volcanoes, including Mount St. Helens (Gerlach and Casadevall 1986) as well as Merapi and Momotombo (Bernard 1985) and equilibrium fO2 values are similar to those expected for their respective lavas (see Symonds et al. 1994). The XSO2 /(XH2S + XSO2) ratio computed at redox conditions fixed by the NNO and QFM buffers is plotted as a function of temperature and water fugacity (which closely approximates total pressure as H2O is by far the major component of volcanic gases and as long as the fugacity coefficient of water does not depart significantly from unity) in Figure 5. It shows that H2S is stabilized by decreasing temperatures and increasing pressures and SO2 is favored by increasing temperatures and decreasing pressures. In addition to redox conditions, temperature, and pressure, the XSO2 /(XH2S + XSO2) ratio is sensitive to magma degassing, owing to the different solubility of H2S and SO2 in silicate melts.
Figure 5. Plot of the XSO2/(XH2S + XSO2) ratio as a function of temperature and water fugacity (which closely approximates total pressure) at oxygen fugacity fixed by (a) the NNO buffer and (b) the FMQ buffer.
436
Marini, Moretti, Accornero
Other secondary, non-magmatic processes potentially affecting the concentrations of S species in volcanic gases have been discussed by Giggenbach (1996). They comprise: (i) deposition and revolatilization of elemental sulfur due to thermal effects; (ii) disproportionation of SO2 according to the equations: 4 SO2(g) + 4 H2O(g,l) → 3 HSO4−(aq) + 3 H+(aq) + H2S(g) 3 SO2(g) + 2 H2O(g,l) → 2
HSO4−(aq)
+2
H+(aq)
+ S(l,s)
(24) (25)
followed by dissolution of the produced sulfuric acid into associated brines; (iii) addition of H2S, produced through interaction of high-temperature volcanic gases with earlier deposited elemental S, as indicated by the reaction: 4 S(l,s) + 4 H2O(g) → HSO4−(aq) + H+(aq) + 3H2S(g)
(26)
again with absorption of sulfuric acid in brines; (iv) addition of both H2S and SO2, generated through interaction of volcanic gases with elemental S, according to the reaction: 3 S(l,s) + 2 H2O(g) → SO2(g) + 2H2S(g)
(27)
(v) oxidation of H2S to SO2 driven by atmospheric O2 or other oxidants; (vi) reduction of SO2 to H2S controlled by excess, higher temperature H2 or other reducing agents; (vii) mixing of relatively oxidized, high-C/S magmatic gases with more reduced, lowC/S gases coming from lateral hydrothermal environments. The latter process was recognized, for instance, to occur at White Island (Giggenbach 1987) and Vulcano Island (Chiodini et al. 1993). All of these secondary, non-magmatic processes may influence not only the contents of S species in volcanic gases but also the δ34S values of SO2 and H2S. Assuming, for the purpose of the following discussion, that the effects of secondary, non-magmatic processes are negligible and that the two main S species present in volcanic gases are in chemical and isotopic equilibrium, the δ34S values of SO2 and H2S are related to the δ34S value of total gaseous sulfur by the following mass balance: δ34SΣS = δ34SSO2 × YSO2(g) + δ34SH2 S × (1 − YSO2(g) )
(28)
where YSO2(g) = XSO2(g) / (XSO2(g) + XH2S(g)). Provided that SO2 and H2S are in isotopic equilibrium, the values of δ34SSO2 and δ34SH2S are linked by the equation: 1000 ln αSO2 -H2 S = δ34SSO2 − δ34SH2 S
(29)
where 1000lnαSO2-H2S represents the equilibrium fractionation factor between gaseous SO2 and gaseous H2S, whose temperature dependence, from 350 to 1050 °C, is reported in Table A-1 (following Ohmoto and Rye 1979). Consequently, if the chemical composition of the gaseous mixture and at least one isotopic value, either δ34SΣS or δ34SSO2 or δ34SH2S, are known, Equations (28) and (29) constrain the other two isotopic values, allowing quantification of the distribution of S isotopes in the considered volcanic gas. Vent temperature was used in these calculations. Values of δ34SΣS, δ34SSO2, δ34SH2S, and outlet temperature for high-temperature volcanic gases are reported in supplemental Table E-3, also including some entries for lower-temperature fumaroles (200-500 °C). These data refer to different plate-tectonics settings. It must be noted that δ34SSO2 values that are very similar to the value listed in supplemental Table E-3 for Kilauea, +0.7‰, were measured in 1979 by Sakai et al. (1982) for two solfataric
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 437 fumaroles of the August 1971 and July 1974 eruptive fissures, +0.9 ± 0.2‰ (n = 15) and +0.8 ± 0.2‰ (n = 7), respectively, in spite of the much lower temperature of these two vents, 128 ± 2 °C and 145 ± 5 °C, respectively. This similarity is probably due to the virtual absence of H2S in these volcanic gases. The δ34SSO2 and δ34SH2S values of the volcanic gases listed in supplemental Table E-3 are compared in Figure 6, with the expected cooling paths for three hypothetical volcanic gases with initial δ34SΣS value of 0‰, which is the accepted value for the pristine mantle sulfur, and different SO2/H2S ratios. These cooling paths were computed using Equations (28) and (29) and assuming that SO2 and H2S concentrations remain constant during cooling (to simplify calculations). Strictly speaking, constant species ratio is valid only for volcanic gases with H2 contents significantly lower than those of gaseous S species, such that no multi-component reactions take place. In spite of the untested hypothesis of chemical and isotopic equilibrium, most samples plot between the two extreme cooling paths for SO2/H2S ratios of 0.11 and 9. Only the gases from Satsuma-Iwojima, with temperatures of 702-885 °C; the fluid from Galeras, with temperature of 358 °C; and the sample from Kuju at 500 °C, deviate significantly from the hypothetical, isochemical cooling paths for volcanic gases with initial δ34SΣS value of 0‰. Among the various processes which may be invoked to explain these deviating values, the most probable are: (i) reduction of sulfate minerals, to explain high δ34SSO2 and δ34SH2S values; adopting anhydrite as the sulfate mineral of interest and assuming that H2S and SO2 are
Figure 6. Plot of δ34SSO2 vs. δ34SH2S values for fumarolic discharges from several volcanoes in different plate-tectonics frameworks (see references in supplemental Table E-3). Black symbols = divergent plate margins; grey symbols = hot spots; open symbols = convergent plate margins. Also shown are the H2S-SO2 equilibrium fractionation lines at selected temperatures, and the expected cooling paths (dashed lines) for three volcanic gases with initial δ34SΣS value of 0‰ and different SO2/H2S ratios.
438
Marini, Moretti, Accornero produced in equal molar amounts, the process is described by the following reaction: 2 CaSO4(s) + 5 H2(g) → SO2(g) + H2S(g) + 2 H2O(g,l) + 2 Ca[OH]2(s)
(30)
where Ca[OH]2(s) represents a Ca-bearing mineral, most likely a Ca-Al-silicate such as wairakite, clinozoisite, prehnite, etc. This reaction may occur if large amounts of H2 are present in the gas mixture or if another reducing agent is available in the considered system. Reduction of dissolved sulfate present in external waters is possible and formally equivalent to reduction of sulfate minerals, however the concentration of SO42−(aq) in high-temperature aqueous solutions is generally low, owing to the strong decrease of the solubility of sulfate minerals with increasing temperature. Another potential process is: (ii) decomposition of pyrite, to explain low δ34SSO2 and δ34SH2S values, as indicated by the reaction: 3 FeS2(s) + 5 H2O(g,l) → SO2(g) + 5 H2S(g) + 3 FeO(s)
(31)
where FeO refers to either iron(II) oxide or an unspecified Fe(II)-bearing solid phase, e.g., an Fe-Al-silicate; as the reaction involving CaSO4 (Eqn. 30) is a disproportionation process, it does not require the involvement of either oxidizing or reducing agents. To complement the plot of δ34SSO2 vs. δ34SH2S values, it is convenient to recall the statistical parameters of the δ34SΣS value, including both measured data and data computed by using Equation (28), but excluding the evidently anomalous samples from Satsuma-Iwojima, Galeras, and Kuju (see above). The remaining 25 samples are situated in the interval from −3.1 to +5.6‰, with a median of +2.30‰, an average of +2.05‰, and a standard deviation of 2.45‰. These values are somewhat higher than that of pristine mantle sulfur, 0‰, evidently due to the occurrence of interfering processes. A well-documented case history is represented by the crater fumaroles of Vulcano Island, Aeolian Archipelago (Italy). In particular, sulfur isotopes were measured by Cortecci et al. (1996) in 166 fumarolic gas samples collected in the period October 1979 thru April 1995. Among these data, we selected 42 analyses which reported the δ34SΣS value, chemical composition, and temperature (see supplemental Table E-4). As before, δ34SSO2 and δ34SH2S values were computed by means of Equations (28) and (29) and compared through the same kind of plot of Figure 6 (Fig. 7) with the expected cooling paths for three hypothetical volcanic gases with initial δ34SΣS value of 0‰, the value of pristine mantle sulfur, and different SO2/H2S ratios. Before inspecting Figure 7, it is worth noting that the chronogram of the δ34SΣS value for fumarole F5 (not reported), for which data are available since September 1973, shows constant values close to +3.0 ± 0.3‰ until late 1984, followed by a drop to +0.2‰ in May 1985 and, then, by a less pronounced decline to δ-values as low as −2.0‰ in early 1995, with fluctuations between −2.7 and +1.4‰. The relatively high δ34S values of total fumarolic S detected in the initial period are possibly due to remobilization of previously deposited elemental sulfur, as observed by Giggenbach (1987) at White Island. Apart from these effects, the δ34SΣS values of crater fumarolic discharges, are chiefly controlled by mixing between a magmatic component richer in 34S and a hydrothermal component richer in 32S, as proposed by Cortecci et al. (1996), in line with the dry model of Cioni and D’Amore (1984), later refined by Chiodini et al. (1991, 1993, 1995). This mixing process also governs the δD and δ18O values of H2O and the concentrations of non-reactive and major components (e.g., H2O, CO2, N2, He). Also shown in Figure 7 are the δ34SSO2 and δ34SH2S values expected for reduction of anhydrite driven by H2 to generate SO2 and H2S in equal molar amounts (Eqn. 30), hydrolysis of sulfur (Eqn. 27), and decomposition of pyrite (Eqn. 31). Excluding one sample of fumarole F5 and one sample from vent FA, which plot in the field of S hydrolysis, all the other gases
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 439
Figure 7. Plot of δ34SSO2 vs. δ34SH2S values for the crater fumaroles of Vulcano Island, Italy (data from Cortecci et al. 1996). Also shown are the equilibrium fractionation lines at selected temperatures (solid lines), the expected cooling paths (dashed lines) for three volcanic gases with initial δ34SΣS value of 0‰ and different SO2/H2S ratios, as well as the δ34SSO2 and δ34SH2S values expected for anhydrite reduction, sulfur hydrolysis and pyrite decomposition.
discharged by the crater fumaroles of Vulcano Island are situated within the cooling lines for mantle S, indicating both the presence of a magmatic component and the magmatic derivation of the S species present in the hydrothermal component, whereas the contributions of S from decomposition of pyrite, reduction of anhydrite, and remobilization of elemental S appear to be minor or totally negligible. The magmatic affinity of hydrothermal S-bearing gases is not surprising, as they are mainly produced through input of magmatic gases into an hydrothermal aquifer, a process determining preferential scrubbing of SO2 (Symonds et al. 2001; Marini and Gambardella 2005). However, hydrothermal H2S and SO2 may be generated through steam-driven hydrolysis reactions taking place in the dry, hot zones surrounding the central magmatic-gas column (Chiodini et al. 1993). Sulfur isotope compositions of fumarolic gases discharged at White Island, New Zealand, were first investigated by Rafter et al. (1958). They found that the isotopic composition of δ34SΣS was close to +3‰ and the elemental sulfur deposited around the fumarolic vents was enriched in 32S. Subsequently, from 1971 to 1976, an extensive sulfur isotope study was made by Giggenbach and Robinson, who collected samples of SO2, H2S, elemental S, and SO42−(aq) (Robinson 1987). During this interval of time White Island experienced two heating cycles, with maximum outlet temperatures of 800 °C (1971-1972 and 1975), separated by a cooler period, with maximum vent temperatures lower than 300 °C. The δ34SΣS varied from +2 to +9‰, with low values during the two hot periods and high values during the cooler interval. According
440
Marini, Moretti, Accornero
to Robinson (1987), these high δ34SΣS values are due to the separation of isotopically light elemental sulfur upstream of the sampling point during the cooler periods. In particular, SO2H2S fractionation indicates apparent equilibrium temperatures from 270 °C to 960 °C, around an average value of 590 °C, although kinetic isotope effects during separation of elemental S shift the δ34S values of remaining S-bearing gases towards greater values. In the peripheral fumaroles, in contrast, the δ34S values of gases decrease due to the formation of sulfate. An interesting set of volcanic gas samples discharged from low-temperature fumaroles situated in 12 active volcanoes of the Cascade Range and Aleutian Arc (including Mount Shasta, Mount Hood, Mount St. Helens, Mount Rainier, Mount Baker, Augustine Volcano, Mount Griggs, Trident, Mount Mageik, Aniakchak Crater, Akutan, and Makushin) was obtained between 1992 and 1998 by Symonds et al. (2003). Maximum outlet temperatures are 173 °C at Mount Mageik and 108-134 °C at Mount Baker, whereas all other fumaroles have temperatures close to or slightly lower than 100 °C. The δ34S of total S for the fumarolic gas samples collected in 1996 was +4.7 to +5.3‰ at Mount Hood, +3.8‰ at Mount Shasta, +1.0 to +1.1‰ at Makushin, −0.1to −4.7‰ at Akutan, and −7.3 to −9.8‰ at Mount Baker. As recognized by Symonds et al. (2003), these δ34S values are within the reported interval of volcanic gases and consistent with chemical and isotopic modifications caused by scrubbing (Symonds et al. 2001; Marini and Gambardella 2005). In fact, the overall interval in δ34S values, from +5.3‰ to −9.8‰, may be related to differences in the SO2/H2S ratio in the volcanic gases and variable effects of scrubbing processes, especially hydrolysis of SO2 to form SO42−(aq) or HSO4−(aq) ions (Eqns. 24 and 25). The H2S/ SO2 pair very often acts as homogeneous redox buffer for re-equilibration reactions occurring during the ascent (cooling) of volcanic gases from the magma to the vent (e.g., Giggenbach 1987; Chiodini et al. 1993; Symonds et al. 1994). The H2S/SO2 magmatic redox buffer then becomes inactive upon removal of SO2 by external groundwaters, if present. Mather et al. (2006) investigated the isotopic compositions of O and S of the aerosol and magma at Masaya volcano, Nicaragua. They found that the δ34S value of aerosol sulfate has an average of +7.7 ± 0.8‰, which compares with that of the magma, +6.6 ± 0.2‰. Isotopic data suggest that sulfate is produced by either high-temperature equilibration of the magmatic gases with small quantities of atmospheric oxygen or by direct emission of sulfate from the magma. Hydrothermal aqueous solutions. Hydrothermal aqueous solutions are generally characterized by the coexistence of: (i) dissolved sulfide, with prevalence of either H2S(aq) or HS−(aq) or S2−(aq), depending on pH; and (ii) dissolved sulfate, with dominance of either HSO4−(aq) or SO42−(aq), again as a function of pH. In addition, both sulfate and sulfide form several aqueous complexes (e.g., CaSO4°(aq), NaSO4−(aq), FeHS+(aq); Seward and Barnes 1997 and references therein), whose stability constants are partially known as a function of the temperature, pressure conditions prevailing in hydrothermal systems. Based on this picture, it is obvious that the rate of exchange of S isotopes between sulfate and sulfide in hydrothermal liquids is essential for the interpretation of the isotope ratios of sulfur in hydrothermal environments. This topic was the subject of several studies (e.g., Igumnov 1976; Robinson 1978; Sakai and Dickson 1978; Kamada et al. 1980; Ohmoto and Lasaga 1982; Sakai 1983), suggesting that the rate of sulfate/sulfide isotope exchange is a function of temperature, pH, and total dissolved sulfur. Temperature seems to be the main controlling parameter, considering that Sakai (1983) evaluated the following half-times for the exchange reaction between SO42−(aq) and H2S(aq): (a) 200 years in well OR-3 of the Hachimantai geothermal reservoir, Japan, at a SiO2-temperature (i.e., the equilibrium temperature calculated based on the solubility of a silica mineral like quartz or chalcedony) of 188 °C; (b) 50 years in well KG-8 of the Krafla geothermal reservoir, Iceland, at a SiO2-temperature of 210 °C; (c) 22 years in well Wairakei-28, New Zealand, at a SiO2-temperature of 250 °C; (d) 0.4 years in well KJ-15, again at Krafla, at a SiO2-temperature of 300 °C.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 441 In addition, Sakai (1983) showed that sulfate and sulfide are not in isotopic equilibrium at these sites, through comparison of fractionation factors and experimental data. This disequilibrium may be attributed to (Sakai 1983): (i) although sulfate and sulfide might have attained isotope equilibrium in the roots of these geothermal systems, at temperatures close to 350 °C, at least at Hachimantai and Wairakei, sulfur isotopes of sulfate and sulfide did not re-equilibrate in the shallower reservoirs tapped by the wells, due to comparatively short residence times, (ii) the presence of one reservoir only, in which geothermal liquids remained for insufficient times to reach isotopic equilibrium, and (iii) the occurrence of sulfide oxidation, and consequent lowering of the δ34S of sulfate, due to exploitation or other reasons. Data for these and other systems including geothermal wells are reported in supplemental Table E-5 and displayed in Figure 8. Although most δ34SSO42− values are in the interval from +13 to +23‰ (Fig. 8), there are also a few isotopically lighter sulfate samples, including the extremely low values of the Icelandic geothermal wells, which span the range from −2.0 to +3.8‰. In contrast, the δ34SH2S values are situated in a much narrower interval, usually from −4 to +2‰, apart from the two exceptions of Zunil and Wairakei. δ34SH2S values are generally close to 0‰ and are consistent with a magmatic origin. Based on the data in supplemental Table E-5, the δ34S values of dissolved sulfate expected for isotope equilibrium in the hydrothermal reservoir fluid, δ34SSO42−, can be computed by means of the following relationship: δ34= SSO24− 1000 ln αSO24− -H2 S + δ34SH2 S
(32)
where δ34SH2S is the measured isotope value of H2S and 1000lnαSO2-H2S represents the equilibrium fractionation factor between dissolved sulfate and H2S. The temperature dependence of the fractionation factor, in the temperature range of 200-400 °C, is conveniently described by the polynomial equation suggested by Ohmoto and Lasaga (1982; see Table A-1).
Figure 8. Plot of δ34SSO42− vs. δ34SH2S values for deep geothermal wells from different fields (data from Sakai et al. 1980; Sakai 1983; Giggenbach et al. 1992; Bayon and Ferrer 2005; Matsuda et al. 2005; González-Partida et al. 2005). Also shown are the equilibrium fractionation lines at selected temperatures.
442
Marini, Moretti, Accornero
Apart from a few wells (H-28 at Hatchobaru, TM2 and TO1 at Mt. Apo, NJ-6D at Palinpinon, M14 at Mahanagdong, and PAL-10D at Bacon-Manito), whose δ34SSO42−,eq values differ from the measured counterparts by less than one per-mil unit, most measured δ34SSO42− values are significantly lower than expected for isotope equilibrium under reservoir conditions, suggesting possible isotope equilibration at higher temperatures, as indicated by Figure 8. Alternatively, the reader may see that apparent equilibration temperatures displayed by Figure 8 are higher than temperatures listed in supplemental Table E-5 in most cases. This is consistent with the conclusions of previous research (e.g., Sakai et al. 1980; Sakai 1983; Giggenbach et al. 1992; Bayon and Ferrer 2005; Matsuda et al. 2005), although other explanations are possible (see above). Interestingly, dissolved sulfate and sulfide species are generally close to isotope equilibrium in acidic liquids, suggesting that low-pH values and high temperatures might increase the exchange rates of sulfur isotopes, favoring the attainment of isotopic equilibrium. Assuming that the dissolved sulfate of deep geothermal fluids derives from disproportionation of magmatic SO2 (Eqn. 24), it is reasonable to hypothesize that: (i) isotopic re-equilibration occurs as long as the aqueous solution is acidic, whereas (ii) it dramatically decreases upon neutralization through water-rock interaction. During this process, a geothermal liquid may remain acidic for one or more unspecified reasons (e.g., low neutralizing capacity of the rocks); if so, sulfur isotopes may re-equilibrate at the comparatively low temperatures of the geothermal reservoir. As described above, the very small fractionation between dissolved sulfate and sulfide in the Icelandic fields of Namafjall and Krafla was attributed (Sakai et al. 1980; Sakai 1983) to the very short residence time of fluids in these geothermal reservoirs (i.e., less than 1 year to a few years), which prevented the attainment of isotopic equilibrium. These extremely short residence times may be due to the high-convective flow rate in these systems, which lie along one of the most active sites of shallow magma movement in Iceland. Light sulfates with δ34SSO42− values relatively close to 0‰, typically found in near-surface geothermal environments, might derive their isotopic signature by H2S oxidation driven by atmospheric oxygen. Descent of these shallow aqueous solutions, whose dissolved sulfate is depleted in 34S, and mixing with deep geothermal liquids, whose dissolved sulfate is enriched in 34S, may generate sulfate with intermediate δ34S values, that would imply unrealistically high temperatures of apparent isotopic equilibrium. This process takes place, for instance, at Mt. Apo, Philippines (Bayon and Ferrer 2005). A similar mechanism was invoked by Martinez Serrano et al. (1996) to explain both the large variability of the δ34SSO42− values in the Los Humeros geothermal reservoir, from +1.7 to +13.5‰, and the similarity of δ34S values of dissolved sulfate in wells H9, H12, H13, H24, and H28 and δ34S values of H2S and pyrite. Sulfate-rich waters are considered to be produced by dissolution of a H2S-rich vapor phase at shallow levels into meteoric water or geothermal water, followed by rapid oxidation of H2S to sulfate, which results in enrichment in 32S. Hydrothermal gases. Hydrothermal gases are characterized by the virtual absence of SO2 and the presence of H2S as unique S-bearing gaseous species (Giggenbach 1980). Therefore, the isotopic composition of H2S can be very important for establishing the origin of S. For instance, average values and standard deviations of sulfur isotope analyses of H2S in different geothermal areas of New Zealand are shown in supplemental Table E-6 (from Robinson 1987). The global average and standard deviation of the entire dataset are +4.7 and ±1.0‰, respectively. As a starting point for the interpretation of these data, Giggenbach (1977) noted that hydrothermal discharges in the Taupo Volcanic Zone have C/S ratio of 35, a value much higher than that of local magmatic discharges, which exhibit a C/S ratio close to 2. This suggests that only a small portion of the sulfur entering the hydrothermal systems from below actually reached the surface, probably because of absorption of sulfur gases in the aqueous phase and consequent precipitation of large amounts of pyrite and other sulfide minerals at depth (e.g.,
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 443 Tui Mine, Te Aroha, where δ34Spyrite lies between +3 and +5‰; Robinson 1987). These values are comparable with those of the H2S released from the geothermal areas, owing to the small fractionation factors between sulfide minerals and H2S gas. Similarly, the δ34S value of H2S is representative of the isotopic composition of the sulfur source, as long as sulfur is lost from the fluid through separation of sulfide minerals and/or as gaseous H2S (e.g., boiling), although the total quantity of sulfur present in the surface discharges was evaluated to be less than 5% of that initially present in the parent fluid (Robinson 1987). In geothermal systems, part of the sulfur is present as dissolved sulfate, but the quantity of sulfate produced through oxidation of H2S is generally small and has limited influence on the δ34SH2S value. For instance, the geothermal reservoir liquids of Broadlands and Kawerau are characterized by high H2S(aq)/SO42−(aq) ratios of 21 and 41 respectively, and high δ34SH2S values of +5.5 and +6.4‰, respectively. In contrast, at Wairakei and Ngawha, where the H2S(aq)/ SO42−(aq) ratios are low, 0.6 and 4 respectively, the δ34SH2S values are also low, +4.3 and +3.3‰, respectively. This is probably due to exchange of S isotopes between H2S(aq) and SO42−(aq), with 34 S preferentially incorporated in the latter. Summing up, it appears that sulfur in New Zealand geothermal systems derives from magmatic H2S, with a δ34S value of +3 to +5‰, and that most of it is separated as sulfide minerals at depth. The remaining H2S is either partly oxidized to sulfate or escapes to the surface. Another interesting hydrothermal-magmatic system is that of Nisyros Island, Greece, whose fumarolic discharges have been the subject of several studies (Ambrosio et al. 2010 and references therein). These investigations produced a significant series of data, including the δ34S value of H2S in fumarolic gases (Marini and Fiebig 2005). The average δ34SH2S of all available data is +3.7‰, with a standard deviation of 0.6‰ (N = 73). Some possibly significant changes of the δ34S value of H2S with time are evident, but their oscillating nature casts doubt on their meaning. In contrast, vent-to-vent changes are small, suggesting that the effects of shallow and relatively shallow processes (e.g., steam separation and condensation) on the δ34S value of H2S are nil to negligible. Interestingly, the δ34S value of H2S at Nisyros is comparable with the isotopic composition of H2S from the geothermal systems of New Zealand. Utilizing the δ34S values of H2S at Nisyros Island, Marini et al. (2002) carried out reaction path modeling to simulate the irreversible water-rock mass exchanges involving the parent geothermal liquid, as well as the progressive changes in the concentrations and δ34S values of S-bearing solutes and minerals. The ultimate aim was to constrain the isotopic composition of sulfur in the magmatic gases entering the shallow hydrothermal system from below. Based on the conceptual geochemical model of the hydrothermal system of Nisyros Island (Ambrosio et al. 2010 and references therein), fumarolic discharges are produced through single-step steam separation from a parent geothermal liquid, deriving from mixing of local seawater (∼30%) with arc-type magmatic water (∼70%), as indicated by H and O isotopes of water (Brombach et al. 2003). Therefore, geochemical modeling of Marini et al. (2002) involved three steps: (1) reaction of seawater with andesitic rocks; (2) neutralization of magmatic water through interaction with andesitic volcanics; and (3) mixing of modified seawater and neutralized magmatic water in the proportions dictated by stable water isotopes. Simulations were carried out under closedsystem conditions, through step-by-step dissolution of andesite in 1 kg of water assuming bulk dissolution of andesite. Temperature was fixed at 350 °C, in line with measurements in deep geothermal wells and geothermometric indications, and PCO2 was assumed to remain constant at 126 bar, the full-equilibrium value dictated by coexistence of Ca-Al-silicates and calcite (Giggenbach 1984). Both the andesite-seawater interaction and mixing of modified seawater and neutralized magmatic water were modeled under a constant fO2 fixed by the FeO-FeO1.5 hydrothermal buffer (Giggenbach 1987). In contrast, the neutralization of magmatic water was simulated under a constant fO2 corresponding to the H2S-SO2 magmatic gas buffer (Giggenbach 1987). This reaction path modeling exercise resulted to be very useful to reproduce, to a first
444
Marini, Moretti, Accornero
approximation, the processes occurring in the root of the hydrothermal system of Nisyros Island and substantiate the interpretation of the sulfur isotopic composition of H2S. It must be noted that this quantitative approach represents a sort of evolution of the qualitative model proposed by Giggenbach (1977) and Robinson (1987) for the geothermal systems of New Zealand. In contrast to the isotopic composition of H2S from previous sites, hydrogen sulfide from the ridge-centered Icelandic magmatic-hydrothermal systems has much lower δ34S values. In particular: (i) the geothermal wells of Krafla have a mean of +0.5 ± 0.3‰, N = 5 (Ármannsson et al. 1982); this geothermal system is directly fed by magmatic gases, as highlighted by the frequent occurrence of fissure eruptions in this area; (ii) the geothermal wells and fumaroles of Nesjavellir as well as the fumaroles of Hengill and Hveragerdi have an average of −0.2 ± 0.8‰, N = 16 (Marty et al. 1991). As pointed out by Ármannsson et al. (1982) and Marty et al. (1991), these sulfur isotope ratios span a restricted range close to 0‰, typical of mantle sulfur (e.g., Grinenko et al. 1975; Sakai et al. 1984). Only the data from the Hveragerdi fumaroles (+1.7 and +1.0‰) are slightly higher, possibly due to a slight contribution of marine water (Marty et al. 1991). Also, the fumarolic H2S of the Solfatara, Campi Flegrei, Italy, is very similar to the meteoritic standard and to the values typical of mantle-derived sulfur, with a mean of −0.3 ± 0.3‰, N = 17 (Allard et al. 1991). Consequently, the fumarolic H2S discharged at Solfatara is considered to be derived from an igneous source, including either deep magma degassing or leaching of sulfur-bearing minerals contained in the pyroclastic rocks filling the Campi Flegrei caldera (Allard et al. 1991). Data reported by Allard et al. (1991) were obtained on gas samples collected between 1983 and 1988, i.e., during and soon after the 1982-1984 seismo-volcanic crisis (Barberi et al. 1984). In contrast, 19 samples of gases, collected from the Solfatara fumaroles in the period 19982001, have an average δ34SH2S value of +1.7 ± 0.3‰ (Gambardella, pers. comm.), suggesting the possible occurrence of progressive degassing from a unique magma batch, without the supply of new magma. Alternatively, the higher δ34SH2S values in the period 1998-2001 could be attributed to the separation of isotopically light elemental sulfur upstream of the sampling point. This is similar to that observed at White Island (Robinson 1987), although the outlet temperature of the Bocca Grande fumarole remained fixed at 155 ± 5 °C from 1983 to 1987 (Allard et al. 1991) and at 160 ± 4 °C from 1999 to 2006 (Caliro et al. 2007). Changes in the δ34SH2S values of fumarolic gases with time, even higher than those of the Solfatara, were recorded at the Baia di Levante beach, Vulcano Island, Italy from 1975 to 1995 (Fig. 9; Cortecci et al. 1996). At this site, geothermal steam at temperatures of 84-105 °C is discharged along the beach through water-soaked sand and shallow seawater. All H2S samples collected before 1984 have δ34S values from −5.4 to −1.5‰, whereas most samples obtained after 1984 have δ34S values from +1 to +5‰. In 1984, a concurrent but reversed change was recorded in the δ34S values of total sulfur in the high-temperature volcanic fumaroles of the La Fossa crater (see above). According to Cortecci et al. (1996), H2S is supplied to the geothermal system of the Baia di Levante beach by both: (i) crater-type fluids, modified by interaction with shallow groundwater, and (ii) reduction of marine sulfate. The contribution of the second source seems to have increased after 1984. Hydrothermal sulfide and sulfate minerals of deep geothermal boreholes. The sulfur isotopic composition of hydrothermal sulfides (pyrite and pyrrhotite) and sulfates (anhydrite and alunite) from deep wells of geothermal systems of New Zealand were analyzed by Steiner and Rafter (1966). They found that: (i) the average δ34S values of pyrite are + 6‰ at Wairakei, +4.7‰ at Waiotapu, and +4.5‰ at Kawerau, with an overall mean value of +5.4‰; and (ii) the average δ34S values of pyrrhotite are +4.0‰ at Wairakei and +4.1‰ at Waiotapu. In general, these sulfide minerals are not in mutual isotopic equilibrium at the present-day in-hole temperatures. The sulfur isotope ratios of these sulfide minerals, as well as those of supergene alunite samples (with a mean value of +5.6‰) are close to that of H2S (see previous section).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 445
Figure 9. Chronogram of δ34SH2S values for the fumaroles of the Baia di Levante Beach, Vulcano Island (data from Cortecci et al. 1996).
A detailed study of the sulfur isotopic composition of mineral sulfides from drill cores in the Broadlands geothermal field was performed by Browne et al. (1975). In borehole Br16, average δ34S values are: +5.5‰ for pyrite, +4.6‰ for sphalerite, and +2.7‰ for galena, whereas the corresponding means for well Br7 are +3.9‰, +3.7‰, and +1.4‰, respectively. Although textural evidence suggests occurrence of chemical equilibrium among these sulfide minerals and the geothermal fluids at temperatures of 160-285 °C, isotope equilibrium temperatures calculated for the sulfide mineral pairs are generally higher than measured inhole temperatures, probably due to incomplete achievement of isotope equilibrium. Only two sphalerite-galena pairs from high-permeability zones appear to be in overall equilibrium with isotope temperatures (∼280 °C) close to measured temperatures (267 and 274 °C). It appears that the exchange of sulfur isotopes with the aqueous solution is faster for galena than for pyrite, pyrrhotite, and sphalerite (Robinson 1987). The incorporation of seawater-derived sulfate in geothermal systems has been invoked in numerous studies. Sakai et al. (1980) measured the sulfur isotope ratios in 33 pyrite samples and 2 gypsum samples, which were taken at different depths from different Icelandic geothermal fields (see supplemental Table E-7). They found a remarkable difference between: (i) the geothermal systems fed by meteoric water (e.g., Nàmafjall, Krafla, Reykholt, Reykir and Leira), in which the δ34S values of hydrothermal pyrite are close to the magmatic value, suggesting that basaltic rocks (or perhaps magmatic gases, in our opinion) represent the main source of sulfur, and (ii) the geothermal fields fed by both meteoric water and marine water (e.g., Reykjanes and Svartsengi), where the δ34S values of hydrothermal pyrite is enriched in 34 S by up to 8‰, with respect to the magmatic value, suggesting that both basaltic rocks (or perhaps magmatic gases, in our opinion) and seawater act as sulfur sources. However, things are not so simple, in that (i) at Reykjanes, the contribution of S from seawater sulfate becomes less important with decreasing depth; (ii) in contrast, at Reykir and other low-temperature systems, mixing of light sulfur produced by near-surface processes and basaltic (or magmatic) sulfur from depth seems to occur. Liou et al. (1985) investigated the sulfur isotope composition of hydrothermal pyrites from the Onikobe geothermal system (Japan). They found that δ34S values range from +2.05
446
Marini, Moretti, Accornero
to +3.59‰ at shallow depths (179-333 m), whereas a somewhat higher value (+5.43‰) was measured at a depth of 704 m (supplemental Table E-7). These δ34S values are very similar to those of pyrites hosted in shallow (< 800 m) altered basaltic rocks in the Reykjanes geothermal field, Iceland, where they are attributed to partial contribution of seawater (see above). Since infiltration of present-day seawater is considered unlikely at Onikobe, Liou et al. (1985) suggested the contribution of fossil marine sulfur from Miocene sediments. As part of an isotopic-mineralogical study of hydrothermal alteration at Arima Spa, Japan, Masuda et al. (1986) observed that sulfur isotopic compositions of pyrite (supplemental Table E-7) and sphalerite are within a restricted interval, from +2.2 to +3.9‰, with the δ34S values of pyrite somewhat higher than those of sphalerite. Possible seawater infiltration and introduction of marine sulfate into the Arima hydrothermal system, when Arima Spa was close to sea level, was invoked by Masuda and coworkers to interpret these data. Boiling can have significant effects on the isotopic composition of precipitated pyrite in geothermal systems. The isotopic compositions of sulfur in over 105 samples of pyrite from five deep wells of the Los Humeros geothermal field, Mexico, was determined by Martinez Serrano et al. (1996) along with the δ34S values of dissolved sulfide and sulfate. Pyrite samples were separated from cuttings collected at variable depths. The δ34S values of pyrite and H2S are very similar and vary between a minimum of −4.5 and a maximum of +4.5‰, around an average of +0.02‰, suggesting the provenance of sulfur involves a magmatic (mantle) source. However, the δ34S values of pyrite from four wells, with an average of −0.4‰, are slightly more negative than those from the fifth borehole, with a mean of +2‰. The largest changes in the isotopic characteristics of pyrite were observed at depths where boiling and mixing phenomena occur. Both processes might impose sulfur isotope fractionation, finally affecting the δ34S values of deposited pyrite (Martinez Serrano et al. 1996). González-Partida et al. (2005) report the isotopic compositions of hydrothermal pyrite in the Los Azufres geothermal field, Mexico. Two different data groups can be identified: (i) one comprises the pyrite samples coming from deep zones, where boiling does not occur, with δ34S values of −1.6 to −3.0‰; and (ii) the other includes pyrites sampled from steam-dominated, shallow levels, with δ34S values of −4 to −4.8‰. The average δ34S value of pyrite, −2.5‰, is similar to that of H2S, −2.0‰. The authors suggest that: (i) the initial isotopic value for sulfur in the geothermal reservoir, −2.3‰, implies a magmatic (mantle) source; and (ii) the values strongly depleted in 34S observed in pyrites from shallow levels are produced by fractionation during boiling of hydrothermal fluids. In their comprehensive study of the four Philippines geothermal fields of Palinpinon, Mahanagdong, Mt. Apo, and Bacon-Manito, Bayon and Ferrer (2005) present sulfur isotope ratios for dissolved sulfate and H2S (see above), and for anhydrite and pyrite collected at variable depths in the geothermal boreholes (supplemental Table E-7). At Palinpinon, Mahanagdong, and Bacon-Manito, sulfur isotope ratios of anhydrite are similar to those of dissolved sulfate, and the δ34S values of pyrite are comparable with those of H2S. These similarities suggest that dissolved sulfate and H2S acted as sulfur sources for secondary anhydrite and pyrite, respectively. In addition, the isotopic composition of both sulfides and sulfates are consistent with the supply of sulfur from magma degassing. In contrast, at Mt. Apo, sulfur isotope ratios of pyrite vary from 0 to +3‰ and are significantly higher than the average δ34S value of H2S of −3‰. According to Bayon and Ferrer (2005), secondary pyrite reflects the δ34S ratio of the original, unboiled hydrothermal fluid, whereas continuous boiling has caused the progressive depletion in 34S in H2S remaining in the separated liquid. If so, the δ34S values of pyrite could suggest that both magmatic gases and seawater act as sulfur sources, similar to what was observed at Reykjanes and Svartsengi (see above). Although data are available only for the four Philippines geothermal fields and two samples from Reykjanes (representative of gypsum and an unspecified sulfate mineral), it is interesting
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 447 to plot these data (Fig. 10). This figure shows that: (i) the hydrothermal pyrites and anhydrites from the geothermal fields of Palinpinon, Mahanagdong, and Bacon-Manito are consistent with the deposition from parent fluids characterized by a δ34SΣS close to 0‰ and low SO42−/H2S ratio at temperatures of 200-400 °C. (ii) In contrast, to explain the isotopic data of the pyrites and anhydrites from the other Philippinian geothermal field of Mt. Apo, it is necessary to invoke parent liquids with δ34SΣS close to 5‰ and variable SO42−/H2S ratios (from 0 to 1) at apparent equilibrium temperatures ≥300 °C. (iii) The data available for Reykjanes are consistent with parent fluids having δ34SΣS close to 5‰ and low SO42−/H2S ratios, at temperatures of 350-400 °C, in line with the interpretation of Sakai et al. (1980). Acid sulfate-chloride groudwaters and the active crater lakes. The acid sulfate-chloride waters, also called volcanic waters (Giggenbach 1988) originate through absorption of hightemperature volcanic gases containing HCl and SO2 in groundwater. The bisulfate contained in these waters (which generally prevails over sulfate ion at the very low pH values of these aqueous solutions) is generated by disproportionation of magmatic SO2 (Eqns. 24 and 25), as recognized by Iwasaki and Ozawa (1960), Ellis and Mahon (1977), and Kusakabe et al. (2000). Bisulfate is generally enriched in 34S, due to the exchange of sulfur isotopes during the disproportionation of SO2. Sulfur, which is typically present in active crater lakes, is produced through both disproportionation of SO2 (Eqn. 25) and oxidation of H2S driven by atmospheric O2 (e.g., Kusakabe et al. 2000). Acid sulfate-chloride thermal waters are generally characterized by HSO4−(aq) concentrations greater than H2S(aq) concentrations. Nevertheless, their HSO4−(aq)/H2S(aq) ratio may deviate significantly from the value of 3, which is expected based on the stoichiometry of SO2 disproportionation (Eqn. 24), due to occurrence of interfering processes (e.g., precipitation of S-bearing minerals, loss of H2S through degassing, etc.). The chemical and isotopic
Figure 10. Plot of δ34SAnhydrite vs. δ34SPyrite values for deep geothermal wells from different fields (data from Sakai et al. 1980 and Bayon and Ferrer 2005). Also shown are the equilibrium fractionation lines at selected temperatures and the lines for δ34SΣS = 0 and δ34SΣS =+5‰ and variable SO42−/H2S ratios of the parent fluids.
448
Marini, Moretti, Accornero
characteristics of these waters were the subject of several investigations (e.g., Williams et al. 1990; Shinohara et al. 1993; Sturchio et al. 1993). In particular, the acid sulfate-chloride thermal waters present in the volcanic areas of northeastern Japan were studied chemically and isotopically by Kiyosu and Kurahashi (1983, 1984). The latter investigation reports the sulfur isotope ratios of coexisting HSO4−(aq) and H2S(aq), whereas in most other studies only the δ34S values of sulfate were determined. The δ34S values of bisulfate and hydrogen sulfide of these Japanese thermal waters range from +3.8 to +31.1‰ and from −16.8 to −1.2‰, respectively, with δ34S values of total dissolved sulfur from +2.5 to +31.1‰ (supplemental Table E-8, listing only the samples for which the isotopic composition of both HSO4−(aq) and H2S(aq) are available). These samples were interpreted to have formed by reaction of volcanic gases having S/Cl ratios of 4 to 7 and δ34SΣS values of total sulfur close to 0‰ with groundwater at 200 °C, with or without precipitation of sulfide minerals. Further indications are provided by the correlation plot of δ34SHSO4− vs. δ34SH2S values (Fig. 11), in which the lines for δ34SΣS = 0‰ and variable SO2/H2S ratio of the parent magmatic gases were drawn based on the isotope mass balances describing the quantitative disproportionation of SO2 (Eqns. 24 and 25). The chemical and isotopic characteristics of active crater lakes have received considerable attention, chiefly for reconstructing their conceptual model and the cycle of chemical elements, including sulfur, with the ultimate aim of mitigating the natural hazards associated with these reservoirs of water and sometimes of liquid sulfur as well. In particular, different sulfurbearing materials of several active crater lakes have been analyzed for the 34S/32S isotope ratios (supplemental Table E-9). In addition to SO2 and H2S discharged by the fumarolic vents, which are typically present in volcanic crater areas, they include both: (i) dissolved bisulfate and (ii) elemental sulfur, which is present as sulfur spherules floating on the water surface and fragments of quenched molten sulfur. These forms of sulfur are considered to be derived from
Figure 11. Plot of δ34SHSO4− vs. δ34SH2S values for acid sulfate-chloride groundwaters of Japan (data from Kiyosu and Kurahashi 1983, 1984). Also shown are the equilibrium fractionation lines at selected temperatures and the δ34SHSO4− and δ34SH2S values expected for disproportionation of SO2, according to Equations (24) (solid lines) and (25) (dashed lines), in magmatic fluids of variable SO2/H2S ratio and initial δ34SΣS = 0‰.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 449 a pool of molten sulfur situated at the bottom of active crater lakes (e.g., Giggenbach 1974; Oppenheimer 1992; Takano et al. 1994). The δ34SHSO4− value ranges from +0.5‰ at El Chichón to +28.3‰ at lake Kawah Putih, Patuha volcano, whereas the δ34SS° value spans the interval from −11.4‰ at Poás to +4.6‰ at El Chichón. The Δ34SHSO4−-S° varies from −4.1‰ at El Chichón to +28.5‰ at Keli Mutu, crater lake 2. Average δ34SHSO4− and δ34SS° values are compared in the correlation plot of Figure 12. High δ34SHSO4− values accompanied by large Δ34SHSO4−-S° (e.g., Maly Semiachik; Poás; Copahue; Banmnoe Lake; El Chichón in 1983; Yugama; Ruapehu; Keli Mutu; Kawah Putih; Kawah Ijen) suggest that bisulfate ion and reduced sulfur species were formed through disproportionation of SO2 in the magmatic-hydrothermal systems presumably situated below the crater lakes (e.g., Rye et al. 1992). In contrast, low δ34SHSO4− values accompanied by small Δ34SHSO4−-S° (e.g., O-Yunuma, Taisho-Jigoku; Khloridnoe Lake; El Chichón in 1995-1997) are ascribed to bacterially mediated aerobic oxidation of elemental sulfur and sulfides (Nakai and Jensen 1964), which usually have low δ34S values (e.g., Kusakabe et al. 2000). Sulfur oxidizing bacteria (e.g., Thiobacillus thiooxidans) are active in warm (10-60 °C) and acidic conditions (pH close to 3), but they cannot grow in strongly acidic environments, i.e., pH<1 (Takano et al. 1997). In addition to disproportionation of SO2 in the magmatic-hydrothermal system and bacterial, inorganic oxidation of H2S driven by atmospheric O2, other reactions may produce both bisulfate ion and elemental sulfur. For instance, hydrolysis of polythionate ions was suggested by Takano and Watanuki (1990): 3 S3O62−(aq) + 2 H2O(g,l) + H+(aq) = 4 S°(l,s) + 5 HSO4−(aq)
(33)
Unfortunately, it is difficult to recognize the pathway controlling the origin of elemental sulfur based on its isotopic signature. However, bacterial oxidation of H2S is unlikely at pH lower than 1.
Figure 12. Plot of δ34SHSO4− vs. δ34SSulfur values for active crater lakes (see references in the text and in supplemental Table E-9). Also shown are the HSO4−-S° equilibrium fractionation lines at selected temperatures and the lines for δ34SΣS = 0 and variable SO2/H2S ratio of the parent magmatic fluids experiencing SO2-disproportionation according to either Equations (25) and (35) (solid line) or Equations (24) and (34) (dashed line), followed by O2-driven oxidation of H2S to elemental sulfur. MS =Maly Semiachik; PO = Poás; CO = Copahue; BL = Banmnoe Lake; CH = El Chichón; YU =Yugama; RU= Ruapehu; KM = Keli Mutu; KP = Kawah Putih; KI = Kawah Ijen; OY = O-Yunuma; TJ = Taisho-Jigoku; KL = Khloridnoe Lake; TA = Taal; NI = Niseko.
450
Marini, Moretti, Accornero
The isotopic values of HSO4−(aq) and S° may be linked to the SO2/H2S ratio of magmatic gases through the following isotope balances: 3 3 δ34SΣS = δ34SHSO− × YSO2 (g) + δ34SS° × 1 − YSO2 (g) 4 4 4
(34)
2 2 δ34SΣS = δ34SHSO−4 × YSO2(g) + δ34SS° × 1 − YSO2(g) 3 3
(35)
= where YSO2(g) XSO2(g) / ( XSO2(g) + X H2 S(g) ) .
Equations (34) and (35) assume total conversion of SO2 through Equations (24) and (25), respectively, as well as total oxidation of H2S through the following reaction: H2S(g) + ½ O2(g) → S(l,s) + H2O(l)
(36)
Equations (34) and (35) were used to draw lines for δ34SΣS = 0 and variable SO2/H2S ratio of the parent magmatic gases in Figure 12. For the crater lakes with high δ34SHSO4− values, comparison of the average isotope values with these theoretical lines indicates that the parent fluids circulating in the sublimnic magmatic-hydrothermal systems are characterized by: (i) very low SO2/H2S ratios at El Chichón in 1983, Kawah Putih, Banmnoe Lake, Yugama, and Kawah Ijen; (ii) SO2/H2S ratios near 1 at Ruapehu and Keli Mutu; and (iii) SO2/H2S ratios somewhat higher than 1 at Maly Semiachik, Copahue, and Poás. Also shown in Figure 12 are the equilibrium fractionation lines between bisulfate ion and elemental sulfur, at selected temperatures. Again, for the crater lakes with high δ34SHSO4− values, comparison of the average analytical isotope values with these equilibrium fractionation lines suggests the attainment of isotope equilibrium between bisulfate ion and elemental sulfur at temperatures of 220-400 °C. Such temperatures are characteristic for magmatic-hydrothermal systems and are often encountered within the sublimnic zone of active crater lakes. Elemental sulfur. Yellow crystals and crusts of solid elemental sulfur are present around the fumarole outlets in all the active volcanoes of our planet. Deposition of solid elemental sulfur is attributed to different redox reactions, possibly bacterially-mediated, as the reverse of Equation (27), the O2-driven oxidation of H2S (Eqn. 36), and the hydrolysis of polythionate ions (Eqn. 33; Takano and Watanuki 1990). In active volcanoes, elemental liquid sulfur occurs, in different locations, such as: (i) at the bottom of several active crater lakes, and (ii) in fumarolic conduits and other parts of the volcanic edifice where temperatures exceed the melting points. Elemental sulfur may involve coexisting molecular forms in the system Sx, with 1 < x < 106 (Schmidt and Siebert 1973). Only a few reports concerning large bodies of liquid sulfur are present in the volcanological literature, the most important of which was the effusion of a 1.4-km-long sulfur flow at SiretokoIôsan volcano, Japan, in 1936 (Watanabe 1940). Other, less-voluminous sulfur flows occurred at Volcán Azufre, Galapagos Islands (Colony and Nordlie 1973); Mauna Loa, Hawaii (Skinner 1970; Greeley et al. 1984); Lastarria, Chile (Naranjo 1985); and Momotombo, Nicaragua (Smithsonian Institution 1990). At all sites, this phenomenon was ascribed to melting and remobilization of fumarolic sulfur deposits. Since 1989, sulfur flows, sulfur pools, and 1-3 m high sulfur volcanoes ejecting sulfur tephra and accretionary lapilli were observed inside the crater of Poás volcano, Costa Rica, due to desiccation of the crater lake (Oppenheimer and Stevenson 1989; Oppenheimer 1992). Values of δ34S ranging between −12.3 and −9.4‰ were obtained on five samples of elemental sulfur from Poás volcano by Oppenheimer (1992). This range of δ34S values is similar to that reported by Rowe (1994) for eight samples (−11.4 to −9.4‰; supplemental Table E-9).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 451 A large set of δ34S values was produced by Ueda et al. (1979), for elemental sulfur samples collected from 44 different volcanic sites in Japan, also including the data previously obtained by Sakai (1957). The whole dataset comprises 91 entries, spanning the range from −9 to +7‰, with an average of −2.2‰ and a standard deviation of 3.6‰. A total of 36 samples of elemental sulfur from the solfataras of Nea Kameni Islet, Santorini Island (Greece), extending over two craters formed in 1940, were reported by Hubberten et al. (1975). The sulfur isotopes are distributed in the interval from 0 to +10‰, with a mean of +5.1‰ and a standard deviation of 2.6‰. The δ34S values of 11 samples of elemental sulfur collected from the crater fumaroles of Vulcano Island, Italy, between 1973 and 1981 range from −5.4 to −1.1‰, with an average of −3.7‰ and a standard deviation of 1.5‰ (Cortecci et al. 1996). The probability plots (Sinclair 1974, 1986) for the δ34S value of elemental sulfur samples from the Japanese volcanic areas (data from Ueda et al. 1979), all the active crater lakes (data in supplemental Table E-9), and the solfataras of Nea Kameni Islet, Santorini Island, Greece (Hubberten et al. 1975) (Fig. 13) show that: (i) the Japanese samples can be fit by a single straight line and are therefore representative of a single population; (ii) the Nea Kameni dataset presents a flex at 32% probability and, therefore, it comprises two populations; and (iii) the crater lake set exhibits two inflection points, at 25% and 79% probability and, consequently, it includes three populations. The medium and high populations of the crater lake dataset partly overlap the Japanese volcanoes population, whereas all the Nea Kameni samples and the low population of the crater lake set appear to depart significantly from the the Japanese volcanoes population. These distinct populations reflect the different sources of sulfur as well as the fractionation effects of the processes involved in sulfur deposition (e.g., the reverse of Eqn. 27, and Eqns. 36 and 33).
Figure 13. Probability plots for the δ34S value of elemental sulfur samples from the Japanese volcanic areas (data from Ueda et al. 1979), the active crater lakes (see supplemental Table E-9), and the solfataras of Nea Kameni Islet, Santorini Island, Greece (data from Hubberten et al. 1975). Inflection points indicated by arrows.
452
Marini, Moretti, Accornero SULFUR STABLE ISOTOPES AND THE POTENTIAL FOR PROBINg DEgASSINg AND CRySTALLIzATION PROCESSES AND SULFUR SOURCES IN MAgMAS
As for many other stable isotope ratios, sulfur isotopes measured in bulk igneous rocks may be used to recognize the origin and evolution of magmas, and mixing of different sources and/ or assimilation of country rocks, provided that degassing and crystallization effects do not mask other processes. Despite the relatively narrow range of mantle sulfur isotope values (0 ± 2‰; see above), igneous rocks and their constituent minerals show a much wider variability. This testifies to the occurrence of other processes in addition to simple mantle melting, such as rock contamination and assimilation (Shima et al. 1963; Cheney and Lange 1967; Gorbachev and Grinenko 1973; Ripley 1981; Ohmoto 1986; Coleman 1979), as well as interaction with seawater (Grinenko et al. 1975; Sakai et al. 1978). However, sulfur isotope values of SO2 and H2S discharged by fumarolic effluents and hydrothermal minerals in active geothermal systems and ore deposits demonstrate that the magmatic-derived sulfur entering the ore-forming hydrothermal environments experiences very large fractionations (see above). Indeed, very simple physical phenomena accompanying magmatic production, differentiation, ascent and emplacement, such as (1) degassing of SO2 and/or H2S and (2) crystallization, particularly of S-bearing phases, may have a dramatic effect on the variations of δ34S values observed in rocks, sulfide and sulfate phases (either solid or liquid), and gases. These effects must be addressed, chemically and isotopically, to understand the evolution of the magmatic system. In fact, mixing is one of the possibly occurring phenomena, but secondary with respect to sulfur isotope fractionation induced by the phase separation processes recalled above, which take place under the high-temperature conditions of magmatic systems. The main consequence of the occurrence of these separation processes is that the δ34SΣS value of a given rock is not representative of the original (source) value, because the investigated rock is severely affected by the separation style as well as by pressure, temperature, bulk system composition and oxygen fugacity. Therefore, measured data must be tested first for the effects of fractionation due to degassing and/or crystallization. Once degassing and/or crystallization effects have been recognized, mixing between different sources can be inferred based on both the concurrent evaluation of several isotopic ratios and chemical data analysis (e.g., trace elements). Many interpretations based on mixing of different sources should be revised, if not dismissed, then simplifying the conceptual model of the magmatic systems under investigation. A good example of the correct approach to the problem is given by Faure et al. (1984), Ueda and Sakai (1984), Zheng (1990), and Marini et al. (1994) amongst others. The case of sulfur is particularly complicated because of its high reactivity, mainly due to to the numerous formal oxidation states (−2 to +6) it can have in a magmatic system which largely determines its speciation in both melts and gas phases (see Webster and Botcharnikov 2011, this volume). Coleman (1979) already observed that the isotopic composition in studied granitic complexes may be related to the magma-forming oxidation state. Moreover, the high reactivity of sulfur also determines the separation of different solid phases, not only sulfides and sulfates, usually solid solutions dominated by FeS and CaSO4, respectively, (see Baker and Moretti 2011, this volume), but also Fe-O-S liquids coexisting with silicate melts. All these phase separation possibilities give rise to a large spectrum of effects on the sulfur content and isotope signature of the parental melt. The separation style (closed- vs. open-system) further complicates the picture. To reconstruct how sulfur and sulfur isotopes partition among coexisting phases, different models can be applied (Baker and Moretti 2011, this volume). The choice of the speciation and saturation model to use is left to the readers, who must be aware that this choice may strongly affect (i) the SO2/H2S ratio in the gas phase, (ii) the saturation state of silicate melt with respect to solid/liquid phases, and (iii) the δ34S values of the gas and melt phases.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 453 In order to simplify mass and isotope balances, the modeling possibilities shown below will assume, to a first approximation, that S-bearing sulfide phases can be modeled as pure FeS (pyrrhotite), either liquid or solid, in igneous systems. Similarly, anhydrite will be considered the only relevant solid sulfate separating from silicic oxidized melts.
Magmatic degassing and sulfur isotope fractionation The theoretical variations in the bulk δ34S value of the melt, δ34SΣS, caused by closed- and open-system degassing are described by (e.g., Marini et al. 1994): δ34SΣS, f = δ34SΣS,i − (1 − F ) × 1000 ln α gas-melt
δ34SΣS , f = δ34SΣS,i + 1000 × ( F
αgas-melt −1
− 1)
(37)
(38)
respectively, where F = CΣS,melt,f /CΣS,melt,i is the fraction of S remaining in the melt. Equations (37) and (38) depict two extreme scenarios, which bracket all the conditions of gas separation from the melt. Rayleigh distillation is a process in which instantaneous equilibrium is achieved between gas and melt prior to gas removal. Note that Equation (38) strictly holds for a constant fractionation factor through the whole degassing process. Based on the calculus theorem, an average value can be considered in case αgas-melt varies along the considered fractionation path (e.g., Criss 1999). Intermediate possibilities exist between closed-system degassing and Rayleigh distillation. Discrete, open-system, degassing may be well described by a multi-step process involving a sequence of closed-system degassing episodes, each ending with complete or partial gas extraction. The gas-melt equilibrium S isotope fractionation factor is constrained by the following equations (Sakai et al. 1982): 1000 ln α gas − melt ≅ δ34SΣS,gas − δ34 SΣS,melt = = YSO2 × δ34SSO2 + (1 − YSO2 ) × δ34 S H2 S − YSO24− × δ34SSO24− + (1 − YSO24− ) × δ34SS2− = YSO2 × 1000 ln αSO2 − H2 S + YSO24− × 1000 ln αS2− -SO24− + 1000 ln α H2 S-S2−
(39)
where YSO42− and YSO2 are defined as: YSO42− = XSO42−/(XSO42− + XS2−); YSO2 = XSO2/(XSO2 + XH2S). According to Taylor (1986), the temperature (K) dependence of the equilibrium fractionation factors for the couples SO2-H2S and H2S-S2− is conveniently approximated by the following relations: 3
2
103 103 103 1000 ln αSO2 -H2 S = 4.367 0.105 −0.42 × + × − × − 0.41 T T T
(40)
2
1000 ln α H2 S-S2−
103 1.1 × = − 0.19 T
(41)
Equations (40) and (41) are somewhat different from those proposed by Ohmoto and Rye (1979), which are reported in Table A-1. Equation (40) is valid in the range 400-1300 °C, whereas Equation (41) strictly holds in the interval 600-1000 °C (Taylor 1986). The temperature dependence of the sulfate-sulfide equilibrium fractionation factor is described by Equation (19), which was experimentally derived by Miyoshi et al. (1984) and is strictly valid from 600 to 1000 °C. However, Equations (41) and (19) will be used up to 1200 °C for the purpose of this review. It will be apparent, from the following discussion, that degassing can change the sulfur isotopic composition of the residual melt and the separated gas in different directions and in various extents, depending on the speciation of S in both phases. This is due to the different values of the fractionation factor 1000 lnαgas-melt computed through Equation (39). It is
Marini, Moretti, Accornero
454
then important to understand how YSO42− and YSO2 are related to pressure, temperature, and compositional variables.
Separation of S-bearing liquids and/or solid phases Sulfide separation. The theoretical effects of sulfide separation on the δ34SΣS values of magmas under closed- and open-system conditions and constant temperature can be computed by means of Equations (37) and (38), in which α stands for the fractionation factor of S isotopes between the coexisting sulfide phase and silicate melt. Taking pyrrhotite as the most obvious analogue of the sulfide phase, the mineral-melt fractionation may be expressed as (Gambardella 2000): 1000 ln α FeS-melt ≅ δ34SFeS − δ34SΣS,melt = = δ34SFeS − YSO24− × δ34SSO24− + (1 − YSO24− ) × δ34SS2−
(42)
YSO24− × 1000 ln αS2− -SO24− + 1000 ln α FeS-H2 S + 1000 ln α H2 S-S2− =
The temperature dependence of the equilibrium fractionation factor between FeS and H2S, in the 200-600 °C temperature range, is (Ohmoto and Rye 1979, T in K, see Table A-1): 103 1000lnα FeS-H2 S = 0.10 × T
2
(43)
A coefficient of 0.25 for the right-side term was proposed by Li and Liu (2006), over the temperature range 0 to 1000 °C (see Table A-1). The temperature dependence of the equilibrium fractionation factors for the couples SO42−-S2− and H2S-S2− is discussed above. To a first approximation, the same Equations (42) and (43) will be adopted also for describing the fractionation of S isotopes occurring upon separation of an immiscible S-bearing liquid, for simplicity taken as a pure, oxygen-free, FeS liquid. Similar to degassing, the δ34SΣS variations greatly depend on the way the attainment of sulfur saturation is computed with respect to the other variables such as melt composition, pressure, temperature and fO2 (Baker and Moretti 2011, this volume). Anhydrite separation. Similarly, the isotopic fractionation induced by anhydrite separation from the melt can be computed by: 1000 ln α CaSO4 -melt ≅ δ34SCaSO4 − δ34SΣS,melt = = δ34SCaSO4 − YSO2− × δ34SSO2− + (1 − YSO2− ) × δ34SS2− 4 4 4 YSO2− × 1000 ln α S2− -SO2− + 1000 ln α CaSO4 -H2 S + 1000 ln α H S-S2− = 4
4
(44)
2
The temperature dependence of the equilibrium fractionation factor between CaSO4 and H2S is conveniently described by Equation (20). This equation, which was experimentally obtained by Miyoshi et al. (1984), is strictly valid in the temperature interval 600-1000 °C, but it will be extended up to 1200 °C for the aims of this review. The temperature dependence of the equilibrium fractionation factors SO42−-S2− and H2S-S2− has already been described.
Parametric assessment It is clear from the above reactions and equations, that temperature, oxygen fugacity, and phase separation style (closed- vs. open-system behavior) are, together with the initial composition, the main variables governing the δ34SΣS values of residual melts and separated gas, liquid, and solid phases. It is evident that all these parameters operate simultaneously during magmatological and volcanological processes.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 455 Let us now consider different degassing and mineral separation scenarios, in which these parameters either vary independently or are interrelated by well-defined speciation-redoxsolubility functions available from the literature. In particular, we will focus on the effect of chemistry and particularly H2O, in order to show how the total (dissolved+exsolved) volatile content plays a role in shifting δ34SΣS values. The δ34SΣS values of the residual melt and separated phases may shift due to the loss of: (i) S2− by H2S degassing: H2O(g) + S2− ↔ H2S(g) + O2−
(45)
(ii) S2− by sulfide separation, either liquid FeS or pyrrhotite (po): S2− + Fe2+ ↔ FeS(liq or po)
(46)
2−
(iii) SO4 induced by SO2 degassing: SO42− ↔ SO2(g) + O2− + ½ O2(g)
(47)
(iv) SO42− due to anhydrite separation: SO42− +Ca2+ ↔ CaSO4(anh) 2−
(48)
2−
(v) S and SO4 through disproportionation of sulfate and sulfide ions and degassing of SO2, as expressed by the reaction: and
3 SO42− + S2− ↔ 4 SO2(g) + 4 O2−
(49)
(vi) S2− and SO42− on joint degassing of SO2 and H2S in a 1:1 proportion: SO42− + S2− + H2O(g) ↔ H2S(g) + SO2(g) + 2O2− + ½ O2(g)
(50)
which is obtained by combining Equations (45) and (47). Figure 14 reports the expected covariation of sulfide and sulfate concentrations in the melt phase dictated by these reactions. It is worth noting that Equation (50), producing the 1:1 slope shown on the diagram of Figure 14, could explain the typical SO2/H2S ratio in the gas phase of “andesitic” volcanoes, already reported by Giggenbach (1987, 1996). Conversely, Equation (49) produces a 3:1 slope. In the complex frame of volcanic processes, many intermediate mechanisms involving multi-step sequences of several processes may occur. Figure 14 can then be used to evaluate the main process of sulfur loss that takes place in a given magma. For example, one can plot the sulfur speciation data collected from melt inclusions and measured via XANES (e.g., Métrich et al. 2002) or via electron microprobe through the ΔSKα technique (Carroll and Rutherford 1988). Visual analysis of samples for the presence of FeS or CaSO4 will help discriminate between the occurrence of sulfide separation or H2S degassing as well as between SO2 degassing and CaSO4 separation. Let us then see how the above processes determine variations of δ34S in both melt and gas phases by considering different ways to assess YSO42− in the melts, also depending on the adopted way to describe the oxidation state of the system. Figures 15 to 23 represent theoretical frames for the interpretation of analytical isotopic data. In all figures, the initial δ34S value of the parental magma was assumed to be 0‰, the value inferred for the primordial mantle (see above). Figure 15 reports the effect of H2S outgassing according to Equation (45). Two different temperatures were considered: 1200 °C and 900 °C, grossly typical of basaltic and rhyolitic magmatism, respectively. Two general conditions are considered: (i) all the dissolved sulfur is present as sulfide ion, S2− (YSO42− = 0; Fig. 15 a, c), and (ii) sulfur is partitioned at 50% between sulfide ion, S2−, and sulfate ion, SO42− (YSO42− = 0.5; Fig. 15 b, d). The oxidation state is not
456
Marini, Moretti, Accornero
Figure 14. Diagram sketching the main processes yielding sulfur loss in the melt upon degassing and separation of sulfides and anhydrite. Sulfur is present as sulfide anion (S2−) and as sulfate anion (SO42−) which contains hexavalent sulfur cation (S6+) covalently bonded to oxygen atoms. The case of a melt with initial composition of bulk sulfur equal to 2000 ppm, and S2−/S6+ = 1 is shown. See text for details.
fixed and the interconversion between S2− and SO42− is not allowed; so H2S is produced up to S2− exhaustion in the silicate melt. Although geologically unlikely, this latter assumption is included to provide a clear starting basis for developing the modeling. In both cases YSO2 is equal to zero. In the first case, the gas (gray lines) is enriched in the heavier isotope, whereas in the second case the opposite situation is observed, with the gas being depleted in the heavier isotope. Furthermore, if only S2− is present (YSO42− = 0), the δ34S value of both gas and silicate melt decreases progressively with sulfur loss, whereas it tends to increase with sulfur loss for initial YSO42− of 0.5, with the exception of the gas phase under closed-system conditions. Figure 16 shows the fractionation effects due to SO2 outgassing for the same temperatures considered above but for sulfur present either as sulfate only (YSO42− = 1; Fig. 16 a,c), or as both sulfate and sulfide in equal proportions (initial YSO42− = 0.5; Fig. 16 b,d). YSO2 is equal to one in all cases, and again, no constraint is imposed on the redox conditions. Isotope fractionation between the melt and gas phase takes place as a consequence of the reaction of sulfate to SO2 according to Equation (47). For YSO42− = 1, SO2 outgassing produces an increase in the δ34S value of both gas and silicate melt under both closed- and open-system degassing conditions. Fractionation effects are, of course, larger at 900 °C than at 1200 °C. On the other hand, for initial YSO42− of 0.5, SO2 release produces a decrease in the δ34S value of the silicate melt, whereas the 34S isotope content of the gas phase (i) increases constantly for closed-system degassing conditions and (ii) increases up to a maximum around F = 0.6 for open-system gas separation. Figure 17 considers the simultaneous loss of S2− and SO42− via Equation (49), yielding SO2 outgassing (Fig. 17 a,c), or via Equation (50), producing SO2 and H2S in equal proportions (Fig. 17 b,d). In both cases, initial YSO42− is set to 0.5. In the case of SO2 outgassing upon S2−-SO42− disproportionation (Eqn. 49; Fig. 17 a,c), YSO2 is obviously equal to 1. The limiting lowest value of F (0.35) is imposed by mass balances
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 457
Figure 15. Effects of H2S outgassing at (a) 1200 °C and YSO42− = 0, (b) 1200 °C and YSO42−= 0.5, (c) 900 °C and YSO42− = 0, and (d) 900 °C and YSO42− = 0.5. F is the fraction of sulfur remaining in the melt. Black lines indicate the silicate melt phase; gray lines represent the gas phase. Dashed and solid lines represent openand closed-system conditions, respectively. δ34S values were computed through Equations (37), (38), and (39). H2S outgassing takes place via reaction (45) only (i.e. loss of S2-), without any involvement of SO42-.
determined by the stoichiometry of Equation (49). This case is quite similar to that of Figures 16b,d, with a maximum δ34S value of the gas around F = 0.65 under open-system separation. If H2S and SO2 are simultaneously released in 1:1 proportion, SO42− and S2− are consumed in the same 1:1 proportion, as dictated by Equation (50). In this case, we can observe how sulfur isotope fractionation takes place in a degassing scenario which explains many gas compositions observed at andesitic volcanoes (Giggenbach 1987, 1996). Starting from an equivalent distribution of S2− and SO42− (Fig. 17 b,d), both YSO42− and YSO2 keep the value of 0.5 throughout the whole degassing process, and δ34S values of both gas and silicate melt tend to increase. The silicate melt is always enriched in the heavier sulfur isotope with respect to the separated gas. When condensed phases are separated, such as stoichiometric pyrrhotite (or the
458
Marini, Moretti, Accornero
Figure 16. Effects of SO2 outgassing at (a) 1200 °C and YSO42− = 1, (b) 1200 °C and YSO42− = 0.5, (c) 900 °C and YSO42− = 1, and (d) 900 °C and YSO42− = 0.5. Black lines indicate the silicate melt phase; gray lines represent the gas phase. Dashed and solid lines represent open- and closed-system conditions, respectively. δ34S values were computed through Equations (37), (38), and (39). SO2 outgassing takes place via Equation (47) only (i.e. loss of SO42−), without any involvement of S2−.
corresponding liquid) and anhydrite, the isotopic composition of the silicate melt phase becomes progressively more negative with decreasing values of F. The magnitude of isotope fractionation is larger for FeS separation than for CaSO4 separation. Closed-system separation of either sulfide or sulfate is likely not to be detected in most cases and only apparent under extreme open-system behavior. These results, shown in Figure 18, refer to Equations (46) and (48), respectively, and apply to silicate melts in which all sulfur is present as either S2−, when FeS separation occurs, or as SO42− when CaSO4 is considered. The fractionation of sulfur isotopes is evidently controlled by the values of YSO42− and YSO2, depending on the chemical mechanism occurring (e.g., Eqns. 45 to 50). Simple patterns can then be assessed by imposing YSO42− and assuming that YSO2 takes the same values, which holds if SO42− contributes to SO2 and S2− contributes to H2S. The results reported in Figure 19 show
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 459
Figure 17. Effects of SO2 outgassing coupled with S2−-SO42− disproportionation (equation 49) at (a) 1200 °C and (c) 900 °C and effects SO2-H2S outgassing (1:1 proportion) following Equation (50) at (b) 1200 °C and (d) 900 °C. Initial YSO42− is set to 0.5 in all cases. Black lines denote the melt phase; gray lines represent the gas phase. Dashed and solid lines represent open- and closed-system conditions, respectively. δ34S values were computed through Equations (37), (38), and (39).
the strong control exerted by YSO42− on the δ34S values of both the silicate melt and gas phases and clearly demonstrate the important role of assessing dependencies of sulfur speciation (hence YSO42−) on fO2, especially when sulfur speciation is not directly measured. We can then evaluate the effects of isotope fractionation dictated by degassing of SO2 and H2S under conditions governed by Equation (21), keeping in mind that PH2O affects YSO2 via this reaction. Figures 20 (at 1200 °C) and 21 (at 900 °C) are based on this equilibrium, coupled with the dependence of the S2−/SO42− ratio on fO2 proposed by Wallace and Carmichael (1994): X 2− log SO4 X 2− S
25410 1.02 × log fO2 + − 10 = T
(51)
460
Marini, Moretti, Accornero
Figure 18. Effects of FeS separation from silicate melt (Equation 46) at (a) 1200 °C and (c) 900 °C for YSO42− = 0 and effects of CaSO4 separation (reaction 48) at (b) 1200 °C and (d) 900 °C for YSO42− = 1. Black lines denote the melt phase; gray lines represent the gas phase. Dashed and solid lines represent open and closed system conditions, respectively. Calculations were computed on behalf of Equations (42) and (44), respectively, who were incorporated into Equations (37) and (38) after modifying these to describe meltFeS and melt-CaSO4 systems.
Different redox conditions are considered with respect to the NNO buffer (Huebner and Sato 1970) and for different partial pressures of water. Figures 20 and 21 have very similar features, although at the two investigated temperatures some differences in fractionation trends emerge at around NNO, a redox buffer that preferentially prevails in magmas encountered in subduction settings (Symonds et al. 1994). The different modeled conditions show that under reduced conditions the δ34S value becomes more negative with increasing sulfur release from melt (Fig. 20 a, b, c, f, g, h). Due to the non-linearities arising from the ensemble of adopted equations, at 1200 °C, the largest negative enrichment occurs at NNO-1, rather than at NNO-2 (Fig. 20 a, b, f, g). At NNO, trends are still negative, but become positive at NNO+1 (Fig. 20 d,i). When this occurs, the PH2O (from
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 461
Figure 19. H2S and SO2 outgassing from silicate melts at 1200 °C under fixed YSO42− and YSO2. Dashed and solid lines represent open- and closed-system conditions, respectively.
1 to 1000 bar) does not exert any appreciable role on modeled isotope fractionation (Fig. 20 d, e, i, j). On the contrary, the effect of changing PH2O is visible for redox conditions ≤ NNO, especially around NNO-1 (Fig. 20 b, g). The similar behaviors observed at 900 °C (Fig. 21) do not change appreciably the conceptual frame, despite the tendency to have a positive isotope fractionation at more reduced conditions. In fact, the inversion from negative to positive δ34S patterns is observed at NNO (Fig. 21 c, h), and appears to be clearly a function of PH2O, in the light of its effect on YSO2. As in the previous case at 1200 °C, the highest negative isotope fractionation is observed at NNO-1.
462
Marini, Moretti, Accornero
Figure 20 (on facing page and above). Effects of H2S-SO2 outgassing from silicate melts at 1200 °C following Wallace and Carmichael (1994) for different oxygen fugacities relative to the NNO solid buffer. Dashed and solid lines represent open- and closed-system conditions, respectively. Black and gray lines denote the fractionation taking place in the melt and gas phase, respectively. Total sulfur (ST = 2000 ppm).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 463
Figure 20. caption on facing page
However, these dependencies are extrapolated from an equation which does not consider the effect of composition on YSO42−, which was clearly enlightened by Moretti and Ottonello (2005) and clearly emerges from the many available studies on the speciation and solubility of sulfur in melts (see Baker and Moretti 2011 in this volume). Alternative formulations can be based on equations: log( XSO2− / XS2− ) = −38.48CH2 O + 16.348 + 2 log fO2
(52)
log( XSO24− / XS2− ) = −62.60CH2 O + 25.459 + 2 log fO2
(53)
4
for the two extreme compositions of tholeiitic basalt at 1200 °C and rhyolite at 900 °C, respectively. In this approach, YSO42− is not a simple function of temperature and fO2, but of water content as well. Because we are addressing degassing systems, water contents are those at saturation for the pressures (PH2O) here considered, i.e., 1 to 1000 bar. Corresponding water contents at saturation (basalt: 0.007 to 3.78 wt%; rhyolite: 0.01 to 3.81 wt%) were computed via the Papale et al. (2006) model. With respect to the previous approach, we see that δ34S trends similar to those displayed in Figures 20 and 21 are already obtained at more reduced conditions (about 1 log unit; Fig. 22 and 23). As in the previous case, YSO42− and YSO2 increase with fO2, obliterating the effect of PH2O (Fig. 22 c, d, e, h, i, j and Fig. 23 d, e, i, j). In case of rhyolitic melts at 900 °C, the inversion of
464
Marini, Moretti, Accornero
Figure 21 (on facing page and above). Effects of H2S-SO2 outgassing from silicate melts at 900 °C following Wallace and Carmichael (1994) for different oxygen fugacities relative to the NNO solid buffer. Dashed and solid lines represent open- and closed-system conditions, respectively. Black and gray lines denote the fractionation taking place in the melt and gas phase, respectively. Total sulfur (ST = 2000 ppm).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 465
Figure 21. caption on facing page
δ34S degassing trends from decreasing to increasing ones is observed at NNO-1, and depends on the non-linear interplay of redox conditions with water contents and pressures.
Comparison of analytical S and δ34SΣS data from selected volcanic systems and theoretical models of degassing and separation of sulfides and anhydrite Available data. The first S isotopic studies of the volcanic environment (e.g., Sakai 1957; Ault and Kulp 1959; Thode et al. 1961) were focused on separated sulfide and sulfate minerals, native sulfur, gas species, etc. whereas volcanic rocks and melt inclusions were not considered yet. Based on this limited information, it is not possible to reconstruct the effects of degassing and phase separation during magmatic differentiation and magma ascent towards the surface. Among the first pioneering studies on the sulfur isotope composition of volcanic products is that by Smitheringale and Jensen (1963) who measured the δ34SΣS values of several Triassic igneous rocks from the eastern United States, including the basaltic lavas of the Newark group. A few years later, Schneider (1970) measured the δ34S values of sulfide-S, sulfate-S, and total (bulk)-S of Tertiary basaltic rocks from Northern Hessia and Southern Lower Saxony, Germany. However none of these studies measured the concentrations of either bulk S or sulfide-S and sulfate-S. Both Smitheringale and Jensen (1963) and Schneider (1970) discussed the fractionation effects due to S loss, but their interpretation remained at a qualitative level
466
Marini, Moretti, Accornero
Figure 22 (on facing page and above). Effects of H2S-SO2 outgassing constrained by the function for tholeiitic basalts at 1200 °C (Eqn. 52); water contents were computed at saturation with the Papale et al. (2006) model. Dashed and solid lines represent open- and closed-system conditions, respectively. Black and gray lines denote the fractionation taking place in the melt and gas phase, respectively. Total sulfur (ST = 2000 ppm).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 467
Figure 22. caption on facing page
owing to the lack of concentration data. In the same years, δ34SΣS values were measured in fourteen lavas from Santorini Island (Greece), without determining S contents, by Hubberten et al. (1975), who totally neglected fractionation effects induced by phase separation. Grinenko et al. (1975) were probably the first who analyzed the δ34S values of sulfide-S and sulfate-S, as well as their concentrations, in igneous and metamorphic rocks from the Mid Atlantic Ridge, the Indian Ocean Ridge and the East Pacific Rise, but interpretation of these data was also qualitative. Since the 1980s, researchers began to determine systematically the wholerock S concentrations and 34S/32S isotope ratios from different volcanic systems including: (i) the subaerial basalts of Kilauea, Hawaii, whose δ34SΣS values vary from −2.2 to +0.4‰; these are very similar to those of the local S-undegassed submarine basalts, which have δ34SΣS values from −0.6 to +0.8‰ (Sakai et al. 1982); (ii) the Quaternary volcanoes of the Japanese Island Arc, which exhibit δ34SΣS values from −0.4 to +18‰ (Ueda and Sakai 1984); (iii) the Jurassic tholeiites of the Kirkpatrick Basalt Group from both Solo Nunatak and Mt. Falla, Northern Victoria Land, in Antarctica, whose δ34SΣS values vary from −4.01 to +3.41‰ (Mensing et al. 1984) and from −1.45 to +11.73‰ (Faure et al. 1984), respectively;
468
Marini, Moretti, Accornero
Figure 23 (on facing page and above). Effects of H2S-SO2 outgassing constrained by the function for rhyolite melt at 900 °C (Eqn. 53); water contents were computed at saturation with the Papale et al. (2006) model. Dashed and solid lines represent open- and closed-system conditions, respectively. Black and gray lines denote the fractionation taking place in the melt and gas phase, respectively. Total sulfur (ST = 2000 ppm).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 469
Figure 23. caption on facing page
(iv) the submarine volcanic rocks of the Mariana Island Arc and back-arc trough, which have δ34SΣS values from +0.1 to +10.3‰ (Alt et al. 1993), and the Quaternary subaerial lavas from the northern Mariana Island, which exhibit δ34SΣS values from +2.0 to +20.7‰ (Woodhead et al. 1987); the latter authors did not analyzed S concentrations and, therefore, their data will not be considered in the following discussion; (v) the tholeiitic, alkaline, and transitional alkaline volcanic rocks from Iceland, with δ34SΣS values from −2.0 to +4.2‰ (Torssander 1989); and (vi) the young calc-alkaline volcanic rocks from the Taupo Volcanic Zone, New Zealand (Robinson and Graham 1992), which exhibit δ34SΣS values from +0.1 (in basalts) to +12.5‰ (in andesites). The two studies of Sakai et al. (1982) and Ueda and Sakai (1984) provide a complete characterization of the volcanic rocks from the chemical and isotopic point of view, in that the authors measured the concentration and δ34S values of both sulfide and sulfate species. Sakai and coworkers carried out other very interesting, high-quality studies on the ocean-floor basalts from several sites (Galapagos Ridge, FAMOUS area, Cayman Trough, Kilauea east rift, Juan de Fuca Ridge, and Reykjanes Ridge), whose δ34SΣS values vary from −4.2 to +0.4‰ (Sakai et al. 1980, 1984). These submarine volcanic products did not experience significant S-loss and may represent a starting point for modeling gas separation from basaltic systems. In these papers,
Marini, Moretti, Accornero
470
the effects of S-loss were extensively discussed, also considering isotope balances and isotope equilibrium constraints, but geochemical modeling of relevant processes (sulfide precipitation and degassing) was not attempted. A step forward in this direction was made by Marini and coworkers, through the investigation of the silica-undersaturated alkaline magmas of Mt. Vulture (with δ34SΣS values from +4 to +7.8‰; Marini et al. 1994) and Vesuvius volcanoes (with δ34SΣS values from −0.8 to +7.9‰; Marini et al. 1998), as well as the lavas and 122 B.C. Plinian fall deposit of Mt. Etna, with δ34SΣS values ranging from +2.0 to 4.1‰ (Gambardella 2000; Moretti et al. 2005). In these works, the mass and isotope balances proposed by Sakai et al. (1982) were taken into account, together with a thermodynamic description of S speciation in the gas phase and an empirical reconstruction of S speciation in the melt phase (Wallace and Carmichael 1992), in order to model the effects of degassing and sulfide separation, under closed-system and open-system conditions (see above). Plots of δ34SΣS vs. bulk S contents were used to compare theoretical paths with analytical data for the Italian volcanoes mentioned above. In subsequent years, the approach of Marini and coworkers was used to investigate several volcanic systems. These new studies were focused on: (i)
the tephra from the catastrophic 1883 eruption of Krakatoa, Indonesia (Mandeville et al. 1998), in which δ34SΣS values vary from +2.6 to +17.6‰;
(ii) several basaltic rocks from Indonesia, which exhibit δ34SΣS values from +1.7 to +7.8‰ (de Hoog et al. 2001); (iii) glass inclusions of tholeiitic to alkali basaltic composition in Miocene basaltic hyaloclastites of Gran Canaria (Canary Islands), analyzed in situ for δ34SΣS values (from −1.0 to 8.5‰, with an average reproducibility of ± 1.5‰), sulfur concentration (800-2100 ppm) and speciation (YSO42− from 0.40 to 0.87) (Gurenko et al. 2001); (iv) the pumice rhyolitic fall deposits of 31 ka B.P. of Yali Island, Greece, with δ34SΣS values up to ∼17‰ (Gambardella, unpublished data); (v) the climactic rhyodacitic eruption of 7700 a B.P. of Mount Mazama, western North America, that formed the 10-km diameter caldera now containing Crater Lake, whose pyroclastic fall products have δ34SΣS values from +2.4 to 14.8‰ (Mandeville et al. 2009). Few studies have dealt in a complete way with both melt and separated sulfur-bearing phases. For example, Arculus et al. (1983) studied lava samples from Mount Lamington, Papua New Guinea. Two anhydrite-pyrite-bearing lava samples from the southern crater wall of Mount Lamington were analyzed for δ34S values and their concentrations of sulfide and sulfate sulfur (supplemental Table E-10). Three significant features are shown by these data. First, there is a predominance of sulfide over sulfate in these samples which agrees with the relative abundances of pyrite and anhydrite in the rocks. Second, the permil fractionation between sulfate and sulfide is 9.8‰ in sample L15 and 11.1‰ in sample L16. These values correspond (assuming isotopic equilibrium) to temperatures of about 800 and 700 °C, respectively, and are compatible with slight cooling from near-solidus temperatures. Thirdly, although the δ34S values of total sulfur are not the same for both samples (-1.6 and +2.7‰), they are close to 0‰, and this excludes the possibility that marine sulfate (average δ34S value = +20.3‰) was involved in the genesis of the anhydrite. Rye et al. (1984) and Luhr and Logan (2002) investigated the anhydrite- and pyrrhotitebearing trachyandesites of the 1982 eruptions of El Chichón Volcano, Mexico. Rye et al. (1984) measured the sulfur isotope compositions of bulk anhydrite and pyrrhotite separates by means of conventional techniques, obtaining δ34S values of +9.0 to +9.2‰ for anhydrite and of +2.7 and +3.6‰ for pyrrhotite.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 471 Luhr and Logan (2002) used the ion microprobe analysis to determine the sulfur isotope compositions of anhydrite crystals as well as of pyrrhotite, chalcopyrite, and sulfide solid solutions. They obtained average δ34S values of +6.4‰ for anhydrite and of +0.7‰ for sulfide minerals. These values are significantly lower than those presented by the previous authors. Luhr and Logan (2002) proposed three viable mixing models involving (i) magmatic and hydrothermal anhydrite, as well as anhydrite and coexisting sulfide minerals precipitated in different portions of the same magma reservoir, which were affected to different extents by either (ii) degassing or (iii) addition of crustal S. Nevertheless, none of these models can explain all observations, as noted by the authors. When going through these studies, one can recognize a general trend to move from a bulk approach on whole rocks to analyses on suites of melt inclusions hosted in crystals. Melt inclusions, in fact, provide a picture of the full differentiation taking place from the parental melt to more evolved terms. This progressive change of approach demands a full characterization of samples and benefits from the evolution of microanalytical techniques, such as SIMS. Sulfur, one the most important and also easily accessible elements that can be studied by miocroanalytical techniques while working on melt inclusions, greatly benefited from these improvements especially with the possibility of in situ isotopic analyses. A good example of this evolution is the very recent study of Bouvier et al. (2008) on basaltic melt inclusions from St. Vincent (Lesser Antilles Arc), in which δ34SΣS analyses by SIMS were carried out along with a nearly full chemical characterization of the products, including S and H2O concentrations, composition of melt inclusions, stable isotopes of oxygen, lithium and boron and redox estimates (Bouvier et al. 2008). The completeness of datasets allows one to relate δ34SΣS variations and patterns of S-loss to other variables describing the physico-chemical evolution of the system in terms of pressure (i.e., due to magma ascent), temperature, oxidation state, as well as to identify possible mixing and assimilation phenomena. The available data for volcanic products recalled so far are presented in supplemental Table E-10. Figure 24 displays these data on correlation diagrams of S content vs. δ34S values, grouped in three big categories: “basaltic”, “andesitic” sensu Giggenbach (1996), and “potassic”. The latter family is very atypical and available data are restricted to the Italian volcanism (panel c). In this latter plot, data from Mt. Etna (also reported in panel a) are included as well: they fall along the general trend, which starts with the S-rich products of Mt. Vulture. In Figure 24a, for basaltic systems, data spread around the mantle value (0‰), the largest deviations being associated with SIMS measurements of sulfur-rich (> 700 ppm) melt inclusion samples (Gurenko et al. 2001; Bouvier et al. 2008). Data from Mt. Etna (volatile-rich hawaiites) fall largely above the mantle array. Data for “andesitic” volcanism (Fig. 24b) are, with few exceptions, all largely positive, with the highest δ34S values observed for samples having from few to ~200 ppm. The three panels seem to indicate that: (1) excluding melt inclusion data, most basaltic systems show relatively limited deviations from the mantle value, evidently due to the small extent of degassing and precipitation of condensed phases, if any; (2) andesitic systems experience δ34S enrichment during their evolution, which likely reflects occurrence of S loss, through phase separation, prior to the onset of explosive eruptions, although the possible 34 S-enrichment of the magma source, through injection into the zones of melt generation of fluids released from the subducting slab, cannot be excluded; and (3) Italian volcanoes have quite an enigmatic evolution, combining fluid-enriched sources with an evolution intermediate between that of “basaltic” and “andesitic” volcanism (sensu Giggenbach 1996). Let us now examine in detail some particular case histories, in order to evaluate the potential that sulfur isotopes have for the study of pre-eruptive volcanic processes. We will use sulfur isotope data to obtain insights into the dynamics occurring at Vesuvius (Italy), Mazama (USA) and Etna (Italy) volcanoes. The treatments developed for these volcanoes will show the increasing interpretative complexity from Vesuvius, through Mazama, to Etna.
472
Marini, Moretti, Accornero
Figure 24. Systematics of available δ34S data for magmatic products. Most data are whole-rock S concentrations and 34S/32S isotope ratios, apart from the studies of Gurenko et al. (2001), Bouvier et al. (2008), and Mandeville et al. (2009) that produced melt inclusion data. Panel a) reports data associated with “basaltic” magmatism (sensu Giggenbach 1996). Panel b) reports data associated with “andesitic” volcanism (sensu Giggenbach 1996). Panel c) reports data from Italian volcanism of potassic affinity. Etna data were reported on both panels a) and c).
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 473 Mt. Vesuvius. This volcano has been quiescent since 1944. Vesuvius activity has been characterized by relatively frequent Strombolian eruptions and emissions of lavas and less frequent sub-Plinian and Plinian eruptions (Arnò et al. 1987; Arrighi et al. 2001; Principe et al. 2004). The activity of the first type took place under open-vent conditions and involved small volumes (order of 0.001 km3 or less) of primitive magmas rapidly ascending from considerable depths, possibly >30 km for the last eruption which occurred in 1944 (Dolfi and Trigila 1978). Plinian eruptions involved larger volumes (from 0.1 to >1 km3) of evolved magmas, which differentiated in relatively shallow magma chambers at depths of 3 to 5 km (Santacroce 1983; Barberi et al. 1981) before their discharge, Marini et al. (1998) measured whole-rock S contents and S isotope ratios of recent lavas and pumices of Vesuvius. To provide a theoretical framework for the interpretation of these analytical data, they solved the system of Equations (37) to (39) for temperatures of 800 and 1100 °C and different redox conditions, expressed as ΔNNO. YSO42− was constrained adopting the approach of Wallace and Carmichael (1994), Equation (51), whereas YSO2 was fixed by means of Equations (21) to (23), setting fH2O at 1 bar. Effects of water fugacity on YSO2 were proven to be insignificant up to fH2O close to 1 kbar at 1100 °C and some tens of bars at 800 °C. The initial δ34S value of the melt was assumed to be equal to the typical mantle value (0‰) and an initial S concentration of 3000 ppm was adopted, based on the most primitive sample and an empirical relation proposed by Ducea et al. (1994). The results (Fig. 25) led the authors to conclude that sulfur contents and δ34S values of Somma-Vesuvius volcanics are largely governed by syneruptive, open-system degassing at different temperatures. The key parameter governing the 34 32 S/ S ratio in Somma-Vesuvius volcanic rocks seems then to be the average magma oxidation state, which generally varies from +0.85 to +1.20 ΔNNO units for lavas and from +1.20 to +1.40 ΔNNO units for pumices, which then are more oxidized than lavas. Moreover, Plinian pumices degassed less extensively than lavas, at temperatures of 800-850 °C, consistent with storage and cooling of these magmas in shallow chambers and the cessation of degassing upon syneruptive quenching, which may also explain why the Plinian pumices are richer in S than the lavas. On the other hand, most lavas experienced more extensive open-system degassing than Plinian pumices, at temperatures of 1050-1300 °C, most likely at near surface or very shallow conditions. This is partly justified by the rapid ascent of basic magma towards the surface, without any significant cooling, the delayed vesiculation, and the easy outgassing of lavas due to their enhanced fluidity. Lavas are indeed relatively S-poor compared to the more differentiated magmas which are related to explosive activity yielding pumices. Consequently, the joint interpretation of S contents and δ34S values of magmas constitute a potentially valuable tool to estimate not only their average redox conditions, but also to understand more about the dynamics of volcanic eruptions. Mt. Mazama. Mt. Mazama (Cascade arc) is a 10-km diameter caldera related to the 7700 years B.P. climactic eruption of Crater Lake (Mandeville et al. 2009 and references therein). Compositions of eruption deposits vary from rhyodacite through andesite to mafic cumulate compositions (Williams 1942; Bacon and Druitt 1988; Druitt and Bacon 1989). Mandeville et al. (2009) provided an interesting and complete study of Mt. Mazama volcanics from climactic rhyodacitic to andesitic tephra and from deposits of the pre-climactic and postcaldera eruptions. The authors determined water and sulfur contents in matrix glasses, whole rocks and melt inclusions, sulfur speciation in melt inclusions and δ34S of glasses and whole rocks. δ34S values of the latter range from +2.4 to +14.8‰. Moreover, the occurrence of coexisting sulfide precipitates suggests that their separation plays a pivotal role in controlling the δ34S values of the melt. Mandeville et al. (2009) also analyzed (via SIMS) these sulfides and found that their δ34S varies from +0.4 to +5.8‰. The presence of sulfide phases causes increases of sulfur content in the surrounding environment, equilibrated with the sulfide, accompanied by possible local changes in the δ34S value (due to the “nugget” effect). For example, the crystallization and separation of only 0.314 wt% of
474
Marini, Moretti, Accornero
Figure 25. Plot of δ34S values vs. the fraction of S remaining in the magma upon degassing (F). The theoretical effects of open-system (A) and closed-system (B) isothermal degassing are shown for a water fugacity of 1 bar, at temperatures of 1100 °C (solid lines) and 800 °C (dashed lines), and specified ΔNNO values. Circles and squares refer to lavas and pumices from Mt. Vesuvius, respectively (data from Marini et al. 1998).
a stoichiometric pyrrhotite, with a S content of 11,400 mmol/kg (365,000 ppm), reduces the S content of the residual silicate melt from 46.8 mmol/kg (1500 ppm) to 11.1 mmol/kg (355 ppm), which corresponds with the maximum value observed in Mazama inclusions. Therefore, both sulfide separation and sulfur degassing must be used to explain isotopic values observed in Mazama rhyodacitic and andesitic samples. As observed by Mandeville et al. (2009), pyrrhotite grains in contact with silicate glass are rare in Mazama samples, whereas they are present as inclusions in all the phenocryst phases, and are especially common in Fe-Ti oxides. This suggests that pyrrhotite separation occurred essentially before syn-eruptive degassing, as noted by Mandeville et al. (2009). Based on these observations, the evolution of the δ34S value of the silicate melt will be modeled starting from an assumed, but reasonable, initial content of 1500 ppm, in response to early separation of pure pyrrhotite, followed by syn-eruptive degassing. Mt. Mazama data offer the opportunity to couple degassing information from melt inclusions with those from sulfur isotopes. In particular, it is possible to assess the conditions for the joint degassing of S and H2O by means of the multicomponent saturation model of Moretti et al. (2003) and Moretti and Papale (2004) (see also Baker and Moretti 2011, this volume). Figure 26 was drawn (i) by taking a redox state fixed at NNO+1 based on mineral equilibria (Druitt and Bacon 1989), and (ii) by setting initial total /(exsolved+dissolved) sulfur contents at 400 ppm and 170 ppm, the shift between these two values being produced by FeS separation. The validity of the model approach to reproduce reasonable degassing paths in H2O-S compositional space (Fig. 26) under the experimentally observed constraints given by composition, redox state and total contents of sulfur and water is confirmed when comparing computed with measured YSO42−. Computed values of YSO42− vary from 0.62 at high pressure and high water contents, up to 1 at 1 bar. This covers exactly the same range as has been measured for YSO42− by Mandeville et al. (2009). The following model-generated relationship can be established for a Mazama rhyodacite (Mandeville et al. 2009), to assess oxidation state at 900 °C:
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 475
Figure 26. H2O-S covariation computed for Mt. Mazama melt inclusions and matrix glasses. Solid lines denote closed-system degassing; dashed lines open-system. Computations were carried at NNO+1. Two initial conditions were considered: 1) H2OT= 7 wt% and ST = 400 ppm; 2) H2OT = 7 wt% and ST = 170 ppm. The loss of initial sulfur is ascribed to FeS separation (data from Mandeville et al. 2009)
log( XSO24− / XS2− ) = −55.60CH2 O + 25.372 + 2 log fO2
(54)
The procedure followed here is therefore much more suitable than that based on a much more empirical description of YSO42− dependence on redox state only. Figure 27 shows how the δ34S value of the melt varies in response to open-system sulfide separation and degassing. As initial δ34S, we set the typical mantle value (0‰) for an initial sulfur content of 1500 ppm. FeS separation was modeled by assuming a value of 0.11 for the initial YSO42−. As precipitation occurs, by subtraction of S2− only from the melt, YSO42− increases up to 1. This explains the concave-upward curvature of the FeS separation curve in the semilogarithmic diagram of Figure 27. The YSO42− range measured by Mandeville et al. (2009) is well covered along the FeS separation curve. It is worth noting that sulfide separation yields δ34Smelt = +0.75‰ for a melt sulfur content of 300 ppm. This value is in line with that used by Mandeville et al. (2009) as the initial pole for their degassing calculations and corresponds to the δ34S value (determined via SIMS) of the most primitive pyrrhotite grains in rhyodacite from the climactic eruption. Moreover, δ34S values of separated FeS range between −0.03 and +7.20‰, which are in good agreement with measured values. Degassing trends start from the main curve of FeS separation, and give rise to straight lines (on the semilogarithmic diagram) whose slope is determined by inherited YSO42− values, and which satisfy relation (54) for a given water content and agree with the water-sulfur covariations of Figure 26. The evolution of the δ34S value and S content in Mt. Mazama volcanic products (data from Mandeville et al. 2009) proposed here is based on a geochemical model incorporating sulfur speciation and saturation in the silicate melt (Moretti et al. 2003). A refined appraisal of quantities such as the initial amount of S prior to sulfide separation, which can probably be larger than 1500 ppm only, may improve modeling further. For example, very low amounts of
Marini, Moretti, Accornero
476
Figure 27. δ34S evolution of Mt. Mazama volcanic products (data from Mandeville et al. 2009). Black lines denote SO2-H2S outgassing at NNO+1 but computed for different YSO42− values, calculated from Equation (54) under different plausible pre-eruptive H2O contents (2.95 and 7 wt%). Dashed black lines refer to open-system degassing; solid ones to closed-system degassing. The gray dashed line represents opensystem precipitation of FeS(po) from a parent melt with initial sulfur content of 1500 ppm.
sulfur (<200 ppm) are known for dacites and andesites of Montserrat (Lesser Antilles), although degassing needs a source of some 2500 ppm of sulfur (Edmonds et al. 2010). It should be noted that in felsic melts (such as the Mazama rhyodacite) sulfide crystallization may occur even at low concentrations of sulfur in the melt (tens of ppm) at NNO+1 (Moretti and Baker 2008). Mt. Etna. The case of Mt. Etna represents an interesting application of sulfur degassing and sulfur isotopic evolution related to basaltic explosive volcanism (Parfitt 2004). In the case of Etna hawaiites, adoption of the multicomponent saturation model of Moretti et al. (2003) and Moretti and Papale (2004) (see also Baker and Moretti 2011, this volume) leads to the following relation (at 1200 °C): log( XSO2− / XS2− ) = −37.40CH2 O + 16.1576 + 2 log fO2 4
(55)
Spilliaert et al. (2006) and unpublished data (Moretti 2002) on Mt. Etna volcanics report the presence of sulfide globules in olivine-hosted melt inclusions. Some of these sulfides coexist with glasses having up to 2600 ppm of sulfur (Moretti 2002). This suggests that the total contents of sulfur in the magma may be unexpectedly high, up to more than ~1 wt%, unless such high abundances reflect anomalous enrichments during crystallization of host olivines (Baker 2008). Moretti (2002) measured up to 4000 ppm of S dissolved in Etnean melt inclusions. Available isotopic data are from whole rocks (Gambardella 2000), plus averaged SIMS data on primitive melt inclusions from Allard et al. (2006). It is thus possible to exploit the theoretical evolution of S concentrations and δ34S values in comparison with analytical data. In particular, the highest δ34S values (Gambardella 2000) belong to a Plinian basaltic eruption that occurred at Mt. Etna on 122 BC (Coltelli et al. 1999; Fig. 27). Moretti (2002) investigated Etnean melt inclusions for sulfur speciation via the ΔSKa technique based (Carroll and Rutherford 1988), and found that 122BC melt inclusions are more oxidized (YSO42− ranging
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 477 from 0.6 to 0.9) than other products from much more recent eruptions (1997 to 1999), with the exception of a S-rich (~ 4000 ppm) sample from the 1997 mid-strombolian eruption of Bocca Nuova crater, which yields YSO42− =0.72. This latter value, together with those from 122 BC melt inclusions, are quite in line with the evidences for high YSO42− (around 1), hence high oxidation state, of the most primitive, Sand H2O-rich, Etnean melt inclusions (Métrich et al. 2009). The three isotopic data from 122 BC are well explained by an open-system degassing line (Fig. 28) that, accounting for error bars, also goes through the SIMS datum of Allard et al. (2006). The main features of this opensystem degassing line are: (i) an original sulfur content of 4000 ppm, (ii) a δ34S value of the parent melt of +2‰, suggesting a fluid-contaminated mantle source (e.g., Allard et al. 2006), and (iii) a YSO42− value of 0.64, in line with above evidences for an oxidized parent melt, and (iv) a consequent YSO2 of 0.61. By using Equation (55), and a mean water content of 3.2 wt% (PH2O ≈ 1000 bar) for primitive inclusions (e.g., Allard et al. 2006), a relative oxygen fugacity at NNO can be computed, which is in good agreement with previous findings about the water-rich Etnean magmas (Métrich and Clocchiatti 1996) and plume measurements (Aiuppa et al. 2004, 2007). When these values are “quenched,” probably in response to the fast eruption dynamics leading to fragmentation, linear trends such as AB are produced, due to open-system degassing, in a semi-logarithmic correlation plot between δ34S and S content (Fig. 28a), similar to what has been previously computed for Mt. Mazama volcanics (Fig. 27). S-isotopic data from other historical eruptions (gray points in Fig. 28a) can be explained by starting from the same parent value (S = 4000 ppm, δ34S = +2‰), via either open- or closed-system degassing, and simply assuming that gas can escape through the magma and reequilibrate at the various depths along the saturation curve of the degassing magma. Degassing relations (Fig. 28c) were computed via the Moretti et al. (2003) model, in its updated version for the revised H2O-CO2 saturation model of Papale et al. (2006), and have been used to compute redox parameters and related quantities (Fig. 28b). Calculation of the degassing evolution was performed by adopting total concentrations for H2O, CO2 and S of 3.4, 2 and 0.32 wt% (trend AC), respectively, based on the study from Aiuppa et al. (2007). This way water saturation, hence PH2O, determines how YSO2 and δ34S will change throughout the degassing history. By keeping YSO42− fixed at 0.64, we observe H2O, PH2O, and YSO2 changing progressively from 3.36 wt%, 1654 bar, 0.64, down to 0.02 wt%, 0.96 bar, and 1 respectively, at 1 bar total pressure (Fig. 28b). Consequently, both open-system and closed-system degassing curves AC and AC′ assume a convex curvature also in semi-logarithmic diagrams (Fig. 28a). For open-system degassing, the curved path is caused by the variable gas-melt fractionation factor, in contrast to the straight line expected for a constant gas-melt fractionation factor (as was observed in the Vesuvius case, see above). The same exercise can be done by adopting YSO42− = 0.5 (trend AD), a value commonly found by Moretti (2002). In this case, H2O, PH2O, and YSO2 change progressively from 3.36 wt%, 1654 bar, 0.64, down to 0.02 wt%, 0.96 bar, at 1 bar total pressure, respectively (Fig. 28b). It is worth noting that keeping a constant YSO42− value corresponds to a well-defined physical condition: the SO42−/S2− couple acts as an effective redox buffer of the magmatic system (likely in conjunction with Fe2+/Fe3+ pair), hence determining variations of YSO2 and fO2. This could be the case for an ascending magma batch, in which gas is released quite efficiently, thus allowing fO2 re-equilibration. Other paths of sulfur loss involve (liquid) FeS separation. We omit here paths that can be generated from any degassing line, but focus our attention on a possible path that can explain both the measured S concentration of parent melt and also its isotopic composition without any involvement of mantle contamination. The most primitive melt inclusions representing S-undegassed magma contain 0.29-0.34 wt% S with a mean δ34S of +2.4±0.4‰, comparable to that of 800-1100 °C Etna volcanic gases (Allard 1986). Water in these products is about 3.2 wt%.
478
Marini, Moretti, Accornero
Figure 28. Panel a) Plot of δ34S values vs. total S concentration for Etna whole rocks (data from Gambardella 2000), plus one datum on “primitive” melt inclusions (Allard et al. 2006). The diagram also reports theoretical changes expected for different combinations of SO2-H2S degassing and separation of immiscible sulfides. Symbols are: open boxes = 122 BC Plinian eruption; circles = Strombolian products from the SE crater, February 1999, and Strombolian products from the Bocca Nuova, 1997; diamond = “primitive” melt inclusions. Trends calculated as described in the text at 1200 °C. The initial pole for degassing trends is at δ34S = 2‰ and S = 4000 ppm (A). Note that the δ34S-S path for 122 BC requires YSO42− = 0.64. Dashed lines (AB, AC, AD) denote open-system degassing; solid lines (AC′, AD′) closed-system degassing. Dashed thick line AE represents opensystem separation of FeS(liq) from an hypothetical parent melt with 1.3 wt% of initial sulfur. Panel b) Relations between YSO42− (open triangles), YSO2 (solid triangles) and log fO2 (given as ΔNNO; gray lines) with sulfur contents for the same degassing trends reported in panel a. Panel c) Relations between dissolved H2O (wt%) and PH2O with sulfur contents for the same trends. In panels b) and c) curves AC′ and AD′ were not reported because indistinguishable from AC and AD respectively. Fractionation factors (given as 103ln α) are 0.27 in E and −1.65 in A.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 479 This path can be produced by starting from an initial sulfur content of 1.3 wt% (point E in Fig. 28a), characterized by YSO42− = 0.2, thus quite reduced, and a mantle δ34S of 0‰. Continuous removal of S2− during pyrrhotite open-system separation leads to point A, the one taken as representative of parent melts from which extruded Etnean magmas evolve. Along path EA, the corresponding δ34SFeS increases from −0.3‰ to +0.3‰. The initial 1.3 wt% of sulfur in the system would explain some peculiarities, such as the aforementioned presence of sulfide globules in S-rich melt inclusions. However, it is very difficult to have natural melt compositions reaching such a large amount of sulfur content at sulfide saturation (SCSS), unless admitting oxidized conditions of the source melting region (e.g., Moretti and Baker 2008). Mt. Etna again: a good example about the role of crystallization. A process that deserves special attention is multi-phase-crystallization, leaving the residual melt enriched in S. Enrichment of sulfur in the residual melt can be expected even if degassing or sulfide/sulfate separation occurs simultaneously. As a consequence, fractional crystallization might impose additional constraints on the sulfur isotopic composition of the residual melt. Throughout the following, we will turn again to Etnean S-bearing volcanics and elaborate to what extent their sulfur isotopic composition might have been affected by fractional crystallization. Let us assume that the parent melt has a mantle δ34S signature (0‰) and is extremely enriched in sulfur (1.3 wt%), before starting precipitating FeS. Although such high concentrations were never measured, it is possible that initial S concentrations are as high as 2900 to 3200 ppm (Spilliaert et al. 2006; Aiuppa et al. 2007). This is sustained on the basis of sampled products, and accounts for crystallization of a possibly “undegassed” parent melt (e.g., Spialliaert et al. 2006). At first approximation, crystallization may be modeled as a function of the total pressure again accounting for the presence of a relevant amount of CO2 in the system, i.e., 2 wt% (Aiuppa et al. 2007 and references therein). If we consider that in the range 1-4000 bar (the upper value referring to the entrapment pressure of most primitive and undegassed melt inclusions) some 30 wt% of the initial magma is crystallized (Spillieart et al. 2006), the following empirical function may be used to relate the extent of crystallization to pressure during magma ascent: mXX (wt%)= −0.0075P + 30
(56)
where mXX is the mass of separated crystals and P is the pressure in bars. Let us now insert this expression in the closed-system degassing model made by Aiuppa et al. (2007), performed at NNO and based on initial volatile content of H2O, CO2 and S of 3.4, 2 and 0.32 wt%, respectively. Degassing plots in the H2O-S and H2O-CO2 diagrams (Fig. 29c,d) clearly show that both H2O and S tend to concentrate in the melt despite degassing is going on since the beginning, at 400 MPa (mainly because of CO2 exsolution), similar to what is observed in submarine basalts (Taylor 1986). By considering sulfur enrichment due to crystallization, we actually perform a multistep degassing characterized by a continuous sequence of closed-system degassing conditions (Eqn. 37) in which the concentration of initial sulfur takes at any step the value corrected for crystallization as if it were an incompatible element. Degassing coupled to crystallization from 400 MPa implies a rapid positive evolution of δ34Smelt, up to 1‰ (Fig. 29a), accompanied by a slight increase of S concentration to 3600 ppm (Fig. 29c), when the melt reaches 17 wt% of crystallization. At this stage, the amount of sulfur lost to the gas is 2500 ppm (or 75% of the initial amount), supplied by an oxidized melt with YSO42− decreasing progressively. After this point, both YSO42− and δ34Smelt increase and sulfur concentration decreases. This evolution reflects the imposed crystallization-degassing ratio, of course, and must be taken only as indicative of a general behavior which is characterized by (i) an initial increase in both sulfur content and δ34Smelt, followed by (ii) a decrease in sulfur content and a concurrent increase in the δ34Smelt value.
480
Marini, Moretti, Accornero
Figure 29. Panel a) Plot of δ34S values vs. total S concentration for Etna whole rocks (data from Gambardella 2000), plus one datum on “primitive” melt inclusion (Allard et al. 2006). The diagram also reports theoretical changes expected for SO2-H2S degassing and separation of immiscible sulfides for a crystallizing Etnean magma (see text). Symbols are same as in panel a) of Figure 28. The dashed thick line represents opensystem separation of FeS(liq) from a hypothetical parent melt with 0.32 wt% of initial sulfur. Multiple crystallization takes place since the beginning, leading toward points A′ or A′′ depending on the relative proportion of sulfur precipitating as FeS with respect to the amount of sulfur increase because of its incompatible behavior during magma crystallization. The amount of sulfur precipitating over that gained upon melt enrichment is 10% along trend E′-A′ and 25% along E′A′′. Both trends lead to YSO42− = 0.64-0.65 when approaching the “primitive” melt inclusion condition and the initial degassing pole defined in Figure 28. A different behaviour is related to the deep-degassing + crystallization trend E′F. Panel b) Relations between melt sulfur content, YSO42− (open triangles), YSO2 (solid triangles) for the deepdegassing + crystallization trend E′F. Panel c) Relations between melt sulfur content, dissolved H2O and PH2O for the deep-degassing + crystallization trend E′F. Panel d) Relations between dissolved water and carbon dioxide (left axis) during the course of crystallization (right axis). Fractionation factors (given as 103ln α) are −1.65 in E and −1.70 in A.
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 481 Separation of sulfide without degassing (trends E′A′ and E′A′′A′′′ in Fig. 29a) during the same crystallization path brings even more interesting results, because yields nearly straight patterns connecting point E′ (S = 0.32 wt%; δ34S = 0‰) with the SIMS datum in the primitive melt inclusion (δ34S = 2.4‰; S=0.315 wt%) or in its close vicinity. This suggests that crystallization may be an effective process able to explain isotopic values of the source without any contamination. Differences between the two trends reported in Figure 29a (E′A′ and E′A′′A′′′) are a function of the relative amounts of sulfur precipitating as FeS and sulfur enriched through silicate crystallization (10% for E′A′ and 25% for E′A′′A′′′, respectively). Both trends start at relatively oxidizing conditions with YSO42− = 0.63. δ34SFeS varies between −0.65 and +0.30‰ for both E′A′ and E′A′′ paths. Although multiple magma crystallization offers an interesting alternative to both (i) extremely S-rich (S = 1.3 wt%) but reduced (YSO42− = 0.2) parental sources with mantle δ34S value (0‰), and (ii) contaminated (δ34S = +2‰; Allard et al. 2006) and oxidized (YSO42− > 0.6) parental sources with initial sulfur contents corresponding to the maximum values observed in melt inclusions (~ 0.3 wt%), it is not possible yet to solve ambiguities about the S-isotopic signature of the parental sources. Answers aimed at solving these ambiguities may derive from the analysis of δ34SFeS in most primitive sulfide globules.
Conclusions and future research Sulfur loss due to degassing and/or separation of sulfide-bearing solid or liquid phases, for example FeS-dominated or anhydrite (CaSO4), can drastically change the δ34S value of silicate melts during their evolution, especially if open-system separation (Rayleigh distillation) occurs. For example, very high δ34SΣS values of S-depleted volcanic products are very often associated with open-system degassing under oxidized conditions (e.g., Fig. 24b). These considerable variations in sulfur isotopes are reminiscent of the changes in δD accompanying rhyolitic volcanism (e.g., Taylor 1988, 1991) or in δ13C during basaltic volcanism (e.g., Gerlach and Taylor 1990). All of the controlling factors occur simultaneously in magmas, and they cannot be taken independently from each other. Consequently, non-linear modeling of multicomponent degassing coupled with suitable equations for isotopic fractionation are required. There can be no simple rule of thumb, because of the strong dependence of sulfur isotope fractionation on volcanic degassing via redox equilibria and multicomponent volatile saturation. The latter, in particular, is highly dependent on the total volatile content of the magmatic system of interest, which is not a general feature for a given magmatic type (e.g., basaltic, rhyolitic, etc.) but depends on deep and shallow geochemical processes, subduction history, geodynamics, regional tectonics, and other factors. The joint interpretation of volatile degassing data and sulfur isotope evolution may be very useful to discriminate between different fractionation mechanisms, particularly hypotheses about contamination processes such as those involving interaction with seawater or fluid contamination of a mantle source. However, we also suggest that crystallization of magmas involves so many important effects on sulfur concentrations and isotopic composition that such effects should be clearly accounted for (via chemical analyses and geochemical modeling) before reaching unwanted conclusions about the isotopic value of the parental magmatic source. Applications to volcanology show that the interpretation of sulfur isotope data may be relevant for surveillance techniques and also for the comprehension of the dynamics of preeruptive and syn-eruptive processes. In particular, the main volcanological question which arises from the sulfur isotope evidence from studied explosive products is: why do volcanoes experience relatively oxidized open-system degassing during or before highly explosive events? Recent developments of analytical tools for micro-scale isotopic determinations (e.g., Bendall et al. 2006; Bouvier et al. 2008; Ripley et al. 2011 in this volume) provide enormous possibilities to explore how degassing takes place in volcanic systems. In this respect, the role
482
Marini, Moretti, Accornero
played by sulfur isotopes is pivotal. However, these advancements must be based on 1) accurate assessments of the homogeneity scale of samples in order to have representative data and error estimates not affected by secondary phenomena, and 2) reliable tools and models to link isotopic variation to the magmatological and volcanological process, degassing and multiple phase crystallization, particularly. This latter aspect, however, necessitates physico-chemical approaches that account well for the role of compositional variables (including the redox state) on saturation properties of coexisting phases at equilibrium. It cannot be based on empirical formulations that bear, at best, on limited compositional domains under very limited P-T-fO2 conditions. It is worth noting that this is particularly true for sulfur isotopes, in light of the high reactivity of sulfur species in magmatic systems, and this approach should be extended to all stable isotopes. The few considerations and techniques we have shown here are examples of how to jointly approach chemical and isotope exchanges. Such a joint approach is a major step on the path towards a global simulator of the physico-chemical aspects of volcanism, from magma production, deep gas exsolution, to the injection of separated gases into the hydrothermal systems that feed surface discharges.
ACkNOwLEDgMENTS We are greatly indebted to Jim Webster, Harald Behrens, Jens Fiebig, and Bruce Taylor for their constructive and appreciated comments and suggestions that considerably improved the first version of this review manuscript.
REFERENCES Aiuppa A, Inguaggiato S, McGonigle AJS, O’Dwyer M, Oppenheimer C, Padgett MJ, Rouwet D, Valenza M (2004) H2S fluxes from Mt. Etna, Stromboli, and Vulcano (Italy) and implications for the sulfur budget at volcanoes. Geochim Cosmochim Acta 69:1861-1871 Aiuppa A, Moretti R, Federico C, Giudice G, Gurrieri S, Liuzzo M, Papale P, Shinohara H, Valenza M (2007) Forecasting Etna eruptions by real-time observation of volcanic gas composition. Geology 35:1115-1118 Allard P (1979) 13C/12C and 34S/32S ratios in magmatic gases from ridge volcanism in Afar. Nature 282:56-58 Allard P (1986) Isotopic composition and origin of water, carbon, and sulfur in volcanic gases: rift zones, continental margins and island arcs. PhD Thesis, Paris University Allard P, Maiorani A, Tedesco D, Cortecci G, Turi B (1991) Isotopic study of the origin of sulfur and carbon in Solfatara fumaroles, Campi Flegrei caldera. J Volcanol Geotherm Res 48:139-159 Allard P, Métrich N, Deloule E, Belhadj O, Mandeville C, Spilliaert N (2006) First ion microprobe determination of water and sulfur isotopic ratios in melt inclusions of olivines at Mount Etna. American Geophysical Union, Fall Meeting 2006, abstract #V13D-08 Alt JC, Shanks WC. Jackson MC (1993) Cycling of sulfur in subduction zones – The geochemistry of sulfur in the Mariana-Island arc and back-arc tough. Earth Planet Sci Lett 119:477-494 Ambrosio M, Doveri M, Fagioli MT, Marini L, Principe C, Raco B (2010) Water-rock interaction in the magmatic-hydrothermal system of Nisyros Island (Greece). J Volcanol Geotherm Res 192:57-68 Arculus RJ, Johnson RW, Chappell BW, McKee CD, Sakai H (1983) Ophiolite-contaminated andesites, trachybasalts, and cognate inclusions of Mount Lamington, Papua New Guinea: anhydrite-amphibolebearing lavas and the 1951 cumulodome. J Volcanol Geotherm Res 18:215–247 Ármannsson H, Gíslason G, Hauksson T (1982) Magmatic gases in well fluids aid the mapping of the flow pattern in a geothermal system. Geochim Cosmochim Acta 46:167-177 Arnò V, Principe C, Rosi M, Santacroce R, Sbrana A, Sheridan MF (1987) Eruptive history. In: SommaVesuvius. Santacroce R (ed) Quaderni de La Ricerca Scientifica, CNR, Roma 114, Vol. 8 p 53–103 Arribas AJr, Cunningham CG, McKee EH, Rye RO, Rytuba JJ, Tosdal RM, Wasserman, MD, Aoki M (1995) Compilation of sample preparation and analytical methods and results of chemical, isotopic, and fluid inclusion analyses, Rodalquilar gold-alunite deposit, Spain. U.S. Geological Survey Open-File Report 95-221, p 1-33 Arrighi S, Principe C, Rosi M (2001) Violent strombolian and subplinian eruptions at Vesuvius during post-1631 activity. Bull Volcanol 63:126-150
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 483 Ault WU, Kulp JL (1959) Isotopic geochemistry of sulphur. Geochim Cosmochim Acta 16:201-235 Ayres DE, Burns MS, Smith JW (1981) A sulphur isotope reconnaisance study of the porphyry copper deposits at Panguna and Mt. Fubilan, Papua New Guinea. Pacific Geol 15:37-50 Bachinski DJ (1969) Bond strength and sulfur isotope fractionation in coexisting sulfides. Econ Geol 64:56-65 Bacon CR, Druitt TH (1988) Compositional evolution of the zoned calcalkaline magma chamber of Mount Mazama, Crater Lake, Oregon. Contrib Mineral Petrol 98:224-256 Bahr JR (1976) Sulfur isotopic fractionation between H2S, S and SO42− in aqueous solutions and possible mechanisms controlling isotopic equilibrium in natural systems. MSc Thesis, Pennsylvania State University Baker DR (2008) The fidelity of melt inclusions as records of melt composition. Contrib Mineral Petrol 156:377395 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Barberi F, Bizouard H, Clocchiatti R, Métrich N, Santacroce R, Sbrana A (1981) The Somma-Vesuvius magma chamber: a petrological and volcanological approach. Bull Volcanol 44:295-315 Barberi F, Corrado G, Innocenti F, Luongo G (1984) Brief chronicle of a volcano emergency in a densely populated area. Bull Volcanol 47:175-185 Bayon FEB, Ferrer H (2005) Sulphur isotope ratios in Philippine geothermal systems. In: Use of isotope techniques to trace the origin of acidic fluids in geothermal systems. IAEA-TECDOC-1448, p 111-132 Bendall C, Lahaye Y, Fiebig J, Weyer S, Brey GP (2006) In situ sulfur isotope analysis by laser ablation MCICPMS. Appl Geochem 21:782-787 Bente K, Nielsen H (1982) Experimental S isotope fractionation studies between coexisting bismuthinite (Bi2S3) and sulfur (S°). Earth Planet Sci Lett 59:18-20 Bernard A (1985) Les méchanismes de condensation des gaz volcaniques. PhD Thesis. University of Bruxelles Bethke PM, Rye RO, Stoffregen RE, Vikre PG (2005) Evolution of the magmatic-hydrothermal acid-sulfate system at Summitville, Colorado: integration of geological, stable-isotope, and fluid-inclusion evidence. Chem Geol 215: 281-315 Bigeleisen J, Mayer MG (1947) Calculation of equilibrium constants for isotopic exchange reactions. J Chem Phys 15:261-267 Bischoff JL, Seyfried WE (1978) Hydrothermal chemistry of seawater from 25° to 350 °C. Am J Sci 278:838860 Bouvier AS, Métrich N, Deloule E (2008) Slab-derived fluids in the magma sources of St.Vincent (Lesser Antilles Arc): volatile and light element imprints. J Petrol 49:1427-1448 Bowers TS (1989) Stable-isotope signatures of water-rock interaction in mid-ocean ridge hydrothermal systems: sulfur, oxygen, and hydrogen. Jour Geophys Res 94-B5:5775-5786 Bowers TS, Taylor HPJr (1985) An integrated chemical and stable-isotope model of the origin of midocean ridge hot spring systems. J Geophys Res 90-B14:12583-12606 Brombach T, Caliro S, Chiodini G, Fiebig J, Hunziker JC, Raco B (2003) Geochemical evidence for mixing of magmatic fluids with seawater, Nisyros hydrothermal system, Greece. Bull Volcanol 65:505-516 Browne PRL, Rafter TA, Robinson BW (1975) Sulphur isotopic variations in nature- Part II. Sulphur isotope ratios of sulphides from the Broadlands geothermal field, New Zealand. New Zealand J Sci 18:35-40 Buchanan DL, Nolan J, Suddaby P, Rouse JE, Viljoen MJ, Davenport JWJ (1981) The genesis of sulfide mineralization in a portion of the Potgietersrus Limb of the Bushveld Complex. Econ Geol 76:568-579 Burnham CW, Ohmoto H (1980) Late-stage process of felsic magmatism. Soc Mining Geol Japan Spec Issue 8:1-11 Caliro S, Chiodini G, Moretti R, Avino R, Granieri D, Russo M, Fiebig J (2007) The origin of the fumaroles of La Solfatara (Campi Flegrei, South Italy). Geochim Cosmochim Acta 71:3040-3055 Carroll MR, Rutherford MJ (1988) Sulfur speciation in hydrous experimental glasses of varying oxidation state: Results from measured wavelength shifts of sulfur X-rays. Am Mineral 73:845-849 Casadevall TJ, De La Cruz-Reyna S, Rose WI Jr, Bagley S, Finnegan DL, Zoller WH (1984) Crater lake and post-eruption hydrothermal activity, El Chichón volcano, Mexico. J Volcanol Geotherm Res 23:169-191 Chambers LA (1982) Sulfur isotope study of a modern intertidal environment, and the interpretation of ancient sulfides. Geochim Cosmochim Acta 46:721-728 Chaussidon M, Albarède F, Sheppard SMF (1987) Sulphur isotope heterogeneity in the mantle from ion microprobe measurements of sulphide inclusions in diamonds. Nature 330:242-244 Chaussidon M, Albarède F, Sheppard SMF (1989) Sulphur isotope variations in the mantle from ion microprobe analyses of micro-sulphide inclusions. Earth Planet Sci Lett 92:44-156 Cheney ES, Lange IM (1967) Evidence for sulfurization and the origin of some Sudbury-type ore. Minera Deposita 2:80-94 Chiodini G, Cioni R, Guidi M, Marini L (1991) Geochemical variations at Fossa Grande crater fumaroles (Vulcano Island, Italy) in summer 1988. Acta Vulcanol 1:179-192 Chiodini G, Cioni R, Marini L (1993) Reactions governing the chemistry of crater fumaroles from Vulcano Island, Italy, and implications for volcanic surveillance. Appl Geochem 8:357-371
484
Marini, Moretti, Accornero
Chiodini G, Cioni R, Marini L, Panichi C (1995). Origin of the fumarolic fluids of Vulcano Island, Italy, and implications for volcanic surveillance. Bull Volcanol 57:99-110 Cioni R, D’Amore F (1984) A genetic model for the crater fumaroles of Vulcano Island (Sicily, Italy). Geothermics 13:375-384 Coleman ML (1979) Isotopic analyses of trace sulfur from some S- and I-type granites: heredity or environment? In: Origin of Granite Batholits – Geochemical Evidence. Atherton MP, Tarney J (eds) Shiva Pub Co, Orpington, p 129-133 Colony WE, Nordlie BE (1973) Liquid sulfur at Volcán Azufre, Galapagos Islands. Econ Geol 68:371-380 Coltelli M, Del Carlo P, Vezzoli L (1999) Discovery of a Plinian basaltic eruption of Roman age at Etna volcano, Italy. Geology 26:1095-1098 Coplen TB, Bohlke JK, De Bièvre P, Ding T, Holden NE, Hopple JA, Krouse HR, Lamberty A, Peiser HS, Révész K, Rieder SE, Rosman KJR, Roth E, Taylor PDP, Vocke RDJr, Xiao YK (2002) Isotope-abundance variations of selected elements. Pure Appl Chem 74:1987-2017 Cortecci G, Ferrara G, Dinelli E (1996) Isotopic time-variations and variety of sources for sulfur in fumaroles at Vulcano island, Aeolian archipelago, Italy. Acta Vulcanol 8:147-160 Criss RE (1999) Principles of Stable Isotope Distribution. Oxford University Press Czamanske GK, Rye RO (1974) Experimentally determined sulfur isotope fractionations between sphalerite and galena in the temperature range 600° to 275 °C. Econ Geol 69:17-25 de Hoog JCM, Taylor BE, van Bergen MJ (2001) Sulfur isotope systematics of basaltic lavas from Indonesia: implications for the sulfur cycle in subduction zones. Earth Planet Sci Lett 189:237-252 Delmelle P, Bernard A, Kusakabe M, Fischer TP, Takano B (2000) Geochemistry of the magmatic–hydrothermal system of Kawah Ijen volcano, East Java, Indonesia. J Volcanol Geotherm Res 97:31-53 Delmelle P, Kusakabe M, Bernard A, Fischer T, de Brouwer S, del Mundo E (1998) Geochemical and isotopic evidence for seawater contamination of the hydrothermal system of Taal Volcano, Luzon, the Philippines. Bull Volcanol 59:562-576 Ding T, Valkiers S, Kipphardt H, De Bièvre P, Taylor PDP, Gonfiantini R, Krouse R (2001) Calibrated sulfur isotope abundance ratios of three IAEA sulfur isotope reference materials and V-CDT with a reassessment of the atomic weight of sulfur. Geochim Cosmochim Acta 65:2433-2437 Dolfi D, Trigila R (1978) The role of water in the 1944 Vesuvius eruption. Contrib Mineral Petrol 67:297-304 Druitt TH, Bacon CR (1989) Petrology of the zoned calcalkaline magma chamber of Mount Mazama, Crater Lake, Oregon. Contrib Mineral Petrol 101:245-259 Ducea MN, McInnes BI, Wyllie PJ (1994) Sulfur variations in glasses from volcanic rocks: effect of melt composition on sulfur solubility. Int Geol Rev 36:703-714 Eastoe CJ (1983) Sulfur isotope data and the nature of the hydrothermal systems at the Panguna and Frieda porphyry copper deposits, Papua New Guinea. Econ Geol 78:201-213 Edmonds M, Aiuppa A, Humphreys M, Moretti R, Giudice G, Martin RS, Herd RA, Christopher T (2010) Excess volatiles supplied by mingling of mafic magma at an andesite arc volcano. Geochem Geophys Geosys 11, Q04005, doi:10.1029/2009GC002781 Eldridge CS, Compston W, Williams IS, Harris JW, Bristow JW (1991) Isotope evidence for the involvement of recycled sediments in diamond formation. Nature 353:49-653 Ellis AJ, Mahon WAJ (1977) Chemistry and Geothermal Systems. Academic Press, New York Farquhar J, Bao H, Thiemens M (2000) Atmospheric influence of Earth’s earliest sulfur cycle. Science 289:756758 Farquhar J, Savarino J, Airieau S, Thiemens MH (2001) Observation of wavelength-sensitive mass-independent sulfur isotope effects during SO2 photolysis: implications for the early atmosphere. J Geophys Res 106 (E12):32829-32839 Farquhar J, Wing BA, McKeegan KD, Harris JW, Cartigny P, Thiemens MH (2002) Mass-independent sulfur of inclusions in diamond and sulfur recycling on early Earth. Science 298:2369-2372 Faure G, Hoefs J, Mensing TM (1984) Effect of oxygen fugacity on sulfur isotope compositions and magnetite concentrations in the Kirkpatrick Basalt, Mount Falla, Queen Alexandra Range, Antarctica. Chem Geol 46:301-311 Field CW, Gustafson LB (1976) Sulfur isotopes in the porphyry copper deposit at El Salvador, Chile. Econ Geol 71:1533-1548 Field CW, Zhang L, Dilles JH, Rye RO, Reed MH (2005) Sulfur and oxygen isotopic record in sulfate and sulfide minerals of early, deep, pre-Main Stage porphyry Cu-Mo and late, shallow Main Stage base-metal mineral deposits, Butte district, Montana. Chem Geol 215:61-93 Fifarek RH, Rye RO (2005) Stable-isotope geochemistry of the Pierina high-sulfidation Au-Ag deposit, Peru: influence of hydrodynamics on SO42−-H2S sulfur isotope exchange in magmatic-steam and steam-heated environments. Chem Geo 215:253-279 Gambardella B (2000) Abbondanza e composizione isotopica dello zolfo nelle vulcanite etnee. Quantificazione degli equilibri ossido-riduttivi in atto. MSc Thesis. University of Genova
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 485 Gamo T, Okamura K, Charlou JL, Urabe T, Auzende JM, Ishibashi J, Shitashima K, Chiba H (1997) Acidic and sulfate-rich hydrothermal fluids from Manus back-arc basin, Papua New Guinea. Geology 25:139-142 Gerlach TM, Casadevall TJ (1986) Evaluation of gas data from high-temperature fumaroles at Mount St. Helens, 1980-1982. J Volcanol Geotherm Res 28:107-140 Gerlach TM, Taylor BE (1990) Carbon isotope constraints on degassing of carbon dioxide from Kilauea Volcano. Geochim Cosmochim Acta 55:2051-2058 Giggenbach FW (1974) The chemistry of Crater Lake, Mt. Ruapehu (New Zealand) during and after the 1971 active period. New Zealand J Sci 17:33-45 Giggenbach WF (1977) The isotopic composition of sulphur in sedimentary rocks bordering the Taupo Volcanic Zone. Geochemistry 1977, DSIR Bull 218:57-64 Giggenbach WF (1980) Geothermal gas equilibria. Geochim Cosmochim Acta 44:2021-2032 Giggenbach WF (1984) Mass transfer in hydrothermal alterations systems. Geochim Cosmochim Acta 48:26932711 Giggenbach WF (1987) Redox processes governing the chemistry of fumarolic gas discharges from White Island, New Zealand. Appl Geochem 2:143-161 Giggenbach WF (1988) Geothermal solute equilibria. Derivation of Na-K-Mg-Ca geoindicators. Geochim Cosmochim Acta 52:2749-2765 Giggenbach WF (1996) Chemical composition of volcanic gases. In: Monitoring and Mitigation of Volcanic Hazards. Scarpa R, Tilling RI (eds) Springer, Berlin, p 221-256 Giggenbach WF (1997) The origin and evolution of fluids in magmatic-hydrothermal systems. In: Geochemistry of hydrothermal ore deposits, 3d Edition. Barnes HL (ed) Wiley, New York, p 737-796 Giggenbach WF, Le Guern F (1976) The chemistry of volcanic gases from Erta’Ale, Ethiopia. Geochim Cosmochim Acta 40:25-30 Giggenbach WF, Paniagua De Gudiel D, Roldan Manzo AR (1992) Isotopic and chemical composition of water and gas discharges from the Zunil geothermal system, Guatemala. In: Geothermal Investigations with Isotope and Geochemical Techniques in Latin America. IAEA-TECDOC-641, p 245-278 Goff F, McMurtry GM (2000) Tritium and stable isotopes of magmatic waters. J Volcan Geotherm Res 97:347396 González-Partida E, Carrillo-Chávez A, Levresse G, Tello-Hinojosa E, Venegas-Salgado S, Ramirez-Silva G, Pal-Verma M, Tritlla J, Camprubi A (2005) Hydro-geochemical and isotopic fluid evolution of the Los Azufres geothermal field, Central Mexico. Appl Geochem 20:23-39 Gorbachev NS, Grinenko LN (1973) Origin of the Octobe sulfide ore deposits, Noril’sk region, in the light of sulfide and sulfate sulfur isotope composition. Geochem Int 12:132-137 Greeley R, Theilig E, Christensen P (1984) The Mauna Loa sulfur flow as an analog to secondary sulfur flows (?) on Io. Icarus 60:189-199 Grinenko VA, Dmitriev LV, Migdisov AA, Sharaskin AY (1975) Sulfur contents and isotope composition for igneous and metamorphic rocks from Mid-Ocean Ridges. Geochem Int 12:132-137 Grinenko VA, Thode HG (1970) Sulfur isotope effects in volcanic gas mixtures. Can J Earth Sci 7:1402-1409 Grootenboer J, Schwarcz HP (1969) Experimentally determined sulfur isotope fractionation between sulfide minerals. Earth Planet Sci Lett 7:162-166 Gurenko AA, Chaussidon M, Schmincke HU (2001) Magma ascent and contamination beneath one intraplate volcano: Evidence from S and O isotopes in glass inclusions and their host clinopyroxenes from Miocene basaltic hyaloclastites southwest of Gran Canaria (Canary Islands). Geochim Cosmochim Acta 65:43594374 Heald P, Foley NK, Hayba DO (1987) Comparative anatomy of volcanic-hosted epithermal deposits: acidsulfate and adularia-sericite types. Econ Geol 82:1-26 Hedenquist JW and Lowenstern JB (1994) The role of magmas in the formation of hydrothermal ore deposits. Nature 370: 519-527 Herzig PM, Hannington MD, Fouquet Y, von Stackelberg U, Petersen S (1993) Gold-rich polymetallic sulfides from the Lau back arc and implications for the geochemistry of gold in sea-floor hydrothermal systems of the southwest Pacific. Econ Geol 88:2182-2209 Holser WT, Kaplan IR (1966) Isotope geochemistry of sedimentary sulfates. Chem Geol 1:93-135 Hubberten HW (1980) Sulfur isotope fractionation in the Pb-S, Cu-S and Ag-S systems. Geochem J 14:177-184 Hubberten HW, Nielsen H, Puchelt H (1975) The enrichment of 34S in the solfataras of the Nea Kameni volcano, Santorini Archipelago, Greece. Chem Geol 16:197-205 Huebner JS, Sato M (1970) The oxygen fugacity-temperature relationships of manganese oxide and nickel oxide buffers. Am Mineral 55:934-952 Igumnov SA (1976) Sulfur isotope exchange between sulfide and sulfate in hydrothermal solutions. Geokhimiya 4:497-502 Igumnov SA, Grinenko VA, Poner NB (1977) Temperature dependence of the distribution coefficient of sulfur isotope between H2S and dissolved sulfate in the temperature range 260-400 °C. Geokhimiya 7:1085-1087
486
Marini, Moretti, Accornero
Ionov DA, Hoefs J, Wedepohl KH, Wiechert U (1992) Content and isotopic composition of sulphur in ultramafic xenoliths from central Asia. Earth Planet Sci Lett 111:269-286 Ishihara S, Sasaki A (1989) Sulfur isotopic ratios of the magnetite-series and ilmenite-series granitoids of the Sierra Nevada batholith – A reconnaissance study. Geology 17:788-791 Iwasaki I, Ozawa T (1960) Genesis of sulfate in acid hot spring. Bull Chem Soc Japan 33:1018-1019 Janecky DR, Shanks WC III (1988) Computational modeling of chemical and sulfur isotopic reaction processes in seafloor hydrothermal systems: chimneys, massive sulfides, and subjacent alteration zones. Can Mineral 26:805-825 Jensen ML, Nakai N (1962) Sulfur isotope meteorite standards, results and recommendations. In: Biogeochemistry of Sulfur Isotopes, NSF Symposium, p 30-35 Juliani C, Rye RO, Nunes CMD, Snee LW, Correa Silva RH, Monteiro LVS, Bettencourt JS, Neumann R, Neto AA (2005) Paleoproterozoic high-sulfidation mineralization in the Tapajós gold province, Amazonian Craton, Brazil: geology, mineralogy, alunite argon age, and stable-isotope constraints. Chem Geol 215:95125 Kajiwara Y, Krouse HR (1971) Sulfur isotope partitioning in metallic sulfide systems. Can J Earth Sci 8:13971408 Kajiwara Y, Krouse HR, Sasaki A (1969) Experimental study of sulfur isotopic fractionation between coexistent sulfide minerals. Earth Planet Sci Lett 7:271-277 Kamada E, Sakai H, Kishima N (1980) Experimental study of sulfur isotope exchange between SO42− and H2S (aqueous) at 400 °C and 1000 bars water pressure. Okayama Daigaku Onsen Kenkyusho Hokoku 50:l-15 (in Japanese) Kiyosu Y (1973) Sulfur isotopic fractionation among sphalerite, galena and sulfide ions. Geochem J 7:191-199 Kiyosu Y, Kurahashi M (1983) Origin of sulfur species in acid-sulfate-chloride thermal waters; NE Japan. Geochim Cosmochim Acta 47:1237-l245 Kiyosu Y, Kurahashi M (1984) Isotopic geochemistry of acid thermal waters and volcanic gases from Zao Volcano, Japan. J Volcanol Geotherm Res 21:313-331 Kusakabe M, Komoda Y, Takano B, Abiko T (2000) Sulfur isotopic effects in the disproportionation reaction of sulfur dioxide in hydrothermal fluids: implications for the δ34S variations of dissolved bisulfate and elemental sulfur from active crater lakes. J Volcanol Geotherm Res 97:287-307 Kusakabe M, Robinson BW (1977) Oxygen and sulfur isotope equilibria in the BaSO4-HSO4−-H2O system from 110 to 350 °C and applications. Geochim Cosmochim Acta 41:1033-1040 Li C, Ripley EM, Naldrett AJ (2003) Compositional variations of olivine and sulfur isotopes in the Noril’sk and Talnakh intrusions, Siberia: implications for ore-forming processes in dynamic magma conduits. Econ Geol 98:69-86 Li Y, Liu J (2006) Calculation of sulfur isotope fractionation in sulfides. Geochim Cosmochim Acta 70:17891795 Liou JG, Seki Y, Guillemette RN, Sakai H (1985) Compositions and parageneses of secondary minerals in the Onikobe geothermal system, Japan. Chem Geol 49:1-20 Luhr JF, Logan MA (2002) Sulfur isotope systematics of the 1982 El Chichón trachyandesite: An ion microprobe study. Geochim Cosmochim Acta 66:3303-3316 Mandeville CW, Sasaki A, Saito G, Faure K, King R, Hauri E (1998) Open-system degassing of sulfur from Krakatau 1883 magma. Earth Planet Sci Lett 160:709-722 Mandeville CW, Webster JD, Tappen C, Taylor BE, Timbal A, Sasaki A, Hauri E, Bacon CR (2009) Stable isotope and petrologic evidence for open-system degassing during the climactic and pre-climactic eruptions of Mt. Mazama, Crater Lake, Oregon. Geochim Cosmochim Acta 73:2978-3012 Marini L, Chiappini V, Cioni R, Cortecci G, Dinelli E, Principe C, Ferrara G (1998) Effect of degassing on sulfur contents and δ34S values in Somma-Vesuvius magmas. Bull Volcanol 60:187-194 Marini L, Fiebig J (2005) Fluid geochemistry of the magmatic-hydrothermal system of Nisyros (Greece) In: The Geology, Geochemistry and Evolution of Nisyros Volcano (Greece). Implications for the Volcanic Hazards. Hunziker JC and Marini L (eds) Memoires de Géologie (Lausanne) 44:121-163 Marini L, Gambardella B (2005) Geochemical modeling of magmatic gas scrubbing. Ann Geophys 48:739-753 Marini L, Gambardella B, Principe C, Arias A, Brombach T, Hunziker JC (2002) Characterization of magmatic sulfur in the Aegean island arc by means of the δ34S values of fumarolic gases, elemental sulfur, and hydrothermal gypsum of Nisyros and Milos islands. Earth Planet Sci.Lett 200:15-31 Marini L, Paiotti A, Principe C, Ferrara G, Cioni R (1994). Isotopic ratio and concentration of sulphur in the undersaturated alkaline magmas of Vulture volcano (Italy). Bull Volcanol 56:487-492 Martinez Serrano RG, Jacquier B, Arnold M (1996) The δ34S composition of sulfates and sulfides at the Los Humeros geothermal system, Mexico and their application to physicochemical fluid evolution. J Volcanol Geotherm Res 73:99-118 Marty B, Gunnlaugsson E, Jambon A, Oskarsson N, Ozima M, Pineau F, Torssander P (1991) Gas geochemistry of geothermal fluids, the Hengill area, southwest rift zone of Iceland. Chem Geol 91:207-225
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 487 Masuda H, Sakai H, Chiba H, Matsuhisa Y, Nakamura T (1986) Stable isotopic and mineralogical studies of hydrothermal alteration at Arima Spa, Southwest Japan. Geochim Cosmochim Acta 50:19-28 Mather TA, McCabe JR, Rai VK, Thiemens MH, Pyle DM, Heaton THE, Sloane HJ, Fern GR (2006) Oxygen and sulfur isotopic composition of volcanic sulfate aerosol at the point of emission. J Geophys Res, 111, doi:10.1029/2005JD006584 Matsubaya O, Sakai H, Kusakabe M, Sasaki A (1980) An isotopic study of hot springs in Nagano Prefecture (in Japanese). Papers of Inst Thermal Spring Res, Okayama Univ 50:17-24 Matsuda K, Shimada K, Kiyota Y (2005) Isotope techniques for clarifying origin of SO4 type acid geothermalfluid. Case studies of geothermal areas in Kyushu, Japan. In: Use of isotope techniques to trace the origin of acidic fluids in geothermal systems. IAEA-TECDOC-1448, p 83-95 McNamara J, Thode HG (1950) Comparison of the isotope constitution of terrestrial and meteoritic sulfur. Phys Rev 78:307-308 Mensing TM, Faure G, Jones LM, Bowman JR, Hoefs J (1984) Petrogenesis of the Kirkpatrick basalt, SoloNunatak, Northern Victoria Land, Antarctica, based on isotopic compositions of strontium, oxygen and sulfur. Contrib Mineral Petrol 87:101-108 Métrich N, Berry AJ, O’Neill HSC, Susini J (2009) The oxidation state of sulfur in synthetic and natural glasses determined by X-ray absorption spectroscopy. Geochim Cosmochim Acta 73:2382-2399 Métrich N, Bonnin-Mosbah M, Susini J, Menez B, Galoisy L (2002) Presence of sulfite (SIV) in arc magmas: Implications for volcanic sulfur emissions. Geophys Res Lett 29, doi: 10.1029/2001GL014607 Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim Cosmochim Acta 60:4151-4160 Miyoshi T, Sakai H, Chiba H (1984) Experimental study of sulfur isotope fractionation factors between sulfate and sulfide in high temperature melts. Geochem J 18:75-84 Mizutani Y, Hayashi S, Sugiura T (1986) Chemical and isotopic compositions of fumarolic gases from KujuIwoyama, Kyushu, Japan. Geochem J 20:273-285 Mizutani Y, Sugiura T (1982) Variations in chemical and isotopic compositions of fumarolic gases from Showashinzan volcano, Hokkaido, Japan. Geochem J 16:63-71 Moretti R (2002) Volatile solubility in Silicate Melts with particular regard to sulphur species: Theoretical Aspects and Application to Etnean Volcanics. PhD Thesis, Pisa University Moretti R, Baker DR (2008) Modeling of the interplay of fO2 and fS2 along the FeS-silicate melt equilibrium. Chem Geol 256:286-298 Moretti R, Gambardella B, Marini L, Métrich N (2005) Effects of sulfur degassing and sulfide separation in some products of Mt. Etna volcano (Sicily, Italy). Geochim Cosmochim Acta, Goldschmidt Conference Abstracts 2005, 69/10, Supplement 1, A154 Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts: The Conjugated Toop-SamisFlood-Grjotheim (CTSFG) model. Geochim Cosmochim Acta 69:801-823 Moretti R, Papale P (2004) On the oxidation state and volatile behavior in multicomponent gas-melt equilibria. Chem Geol 213:265-280 Moretti R, Papale P, Ottonello G (2003) A model for the saturation of C-H-O-S fluids in silicate melts, In: Volcanic Degassing. Oppenheimer C, Pyle DM, Barclay J (eds), Geol Soc London Spec Publ 213:81-101 Nakai N (1970) Isotopic ratios of sulfide compounds in sulfide ore deposits. Chikuykagaku 1:31-34 (in Japanese) Nakai N, Jensen ML (1964) The kinetic isotope effect in the bacterial reduction and oxidation of sulfur. Geochim Cosmochim Acta 28:1893-1912 Naranjo JA (1985) Sulphur flows at Lastarria volcano in the North Chilean Andes. Nature 313:778-780 Ohmoto H (1986) Stable isotope geochemistry of ore deposits. Rev Mineral 16:491-559 Ohmoto H, Goldhaber MB (1997). Sulfur and Carbon Isotopes. In: Geochemistry of hydrothermal ore deposits, 3d Edition. Barnes HL (ed) Wiley, New York, p 517-611 Ohmoto H, Lasaga AC (1982) Kinetics of reactions between aqueous sulfates and sulfides in hydrothermal systems. Geochim Cosmochim Acta 46:1727-1745 Ohmoto H, Rye RO (1979) Isotopes of sulfur and carbon. In: Geochemistry of hydrothermal ore deposits, 2nd Edition. Barnes HL (ed) Wiley, New York, p 509-567 Ohsawa S, Takano B, Kusakabe M, Watanuki K (1993) Variation in volcanic activity of Kusatsu-Shirane volcano as inferred from δ34S in sulfate from the Yugama crater lake. Bull Volcanol Soc Jpn 38:95-99 Oppenheimer C (1992) Sulphur eruptions at Volcán Poás, Costa Rica. J Volcanol Geotherm Res 49:1-22 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Oppenheimer C, Stevenson D (1989) Liquid sulphur lakes at Poás volcano. Nature 342:790-793 Papale P, Moretti R, Barbato D (2006) The compositional dependence of the saturation surface of H2O+CO2 fluids in silicate melts. Chem Geol 229:78-95 Parfitt EA (2004) A discussion of the mechanisms of explosive basaltic eruptions. J Volcanol Geotherm Res 134:77-107
488
Marini, Moretti, Accornero
Principe C, Tanguy JC, Arrighi S, Paiotti A, Le Goff M, Zoppi U (2004) Chronology of Vesuvius’ activity from A.D. 79 to 1631 based on archeomagnetism of lavas and historical sources. Bull Volcanol 66:703-724 Raab M, Spiro B (1991) Sulfur isotopic variations during seawater evaporation with fractional crystallization. Chem Geol 86:323-333 Rafter TA, Wilson SH, Shilton WB (1958) Sulphur isotopic variations in nature, part 6 – Sulphur isotopic measurements on the discharge from fumaroles on White Island. New Zealand J Sci 1:154-171 Rainbow A, Clark AH, Kyser TK, Gaboury F, Hodgson CJ (2005) The Pierina epithermal Au-Ag deposit, Ancash, Peru: paragenetic relationships, alunite textures, and stable-isotope geochemistry. Chem Geol. 215:235-252 Rees CE, Jenkins WJ, Monster J (1978) The sulphur isotope geochemistry of ocean water sulphate. Geochim Cosmochim Acta 42:377-382 Richet P, Bottinga Y, Javoy M (1977) A review of hydrogen, carbon, nitrogen, oxygen, sulphur, and chlorine stable isotope fractionation among gaseous molecules. Ann Rev Earth Planet Sci 5:65-110 Ripley EM (1981) Sulfur isotopic studies of the Dunka Road Cu-Ni deposit, Diluth complex, Minnesota. Econ Geol 76:610-620 Ripley EM, Al-Jassar TJ (1987) Sulfur and oxygen isotope studies of melt-country rock interaction, Babbitt CuNi deposit, Duluth Complex, Minnesota. Econ Geol 82:87-107 Ripley EM, Li C (2003) Sulfur isotope exchange and metal enrichment in the formation of magmatic Cu-Ni(PGE) deposits. Econ Geol 98:635-641 Ripley EM, Lightfoot PC, Li C, Elswick ER (2003) Sulfur isotopic studies of continental flood basalts in the Noril’sk region: Implications for the association between lavas and ore-bearing intrusions. Geochim Cosmochim Acta 67:2805-2817 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Robinson BW (1973) Sulfur isotope equilibrium during sulfur hydrolysis at high temperatures. Earth Planet Sci Lett 18:443-450 Robinson BW (1978) Isotopic equilibria between sulfur solute species at high temperature. In: Stable Isotopes in the Earth Sciences. Robinson BW (ed) DSIR Bull 220:203-206 Robinson BW (1987) Sulphur and sulphate-oxygen isotopes in New Zealand geothermal systems and volcanic discharges. Studies on sulphur isotope variations in nature. Proceedings of an advisory group meeting, Vienna, 17-20 June 1985, IAEA, Vienna, p 31-48 Robinson BW (1993) Sulfur isotope standards. Reference and inter-comparison materials for stable isotopes of light elements. Proceedings of a consultants meeting, Vienna, 1-3 December 1993, IAEA, Vienna, p 13-30 Robinson BW, Graham IJ (1992) The sulphur isotopic composition of mafic-intermediate volcanic rocks, Taupo Volcanic Zone, New Zealand. In: Water-Rock Interaction WRI-7. Kharaka Y, Maest A (eds) Balkema, Rotterdam, p 975-978 Rosman KJR, Taylor PDP (1997) Isotopic Composition of the elements 1997. International Union of Pure and Applied Chemistry Rowe GL Jr (1994) Oxygen, hydrogen, and sulfur isotope systematic of the crater lake system of Poás Volcano, Costa Rica. Geochem J 28:263-287 Rye RO (1993) The evolution of magmatic fluids in the epithermal environment: the stable isotope perspective. Econ Geol 88:733-753 Rye RO (2005) A review of the stable-isotope geochemistry of sulfate minerals in selected igneous environments and related hydrothermal systems. Chem Geol 215:5-36 Rye RO, Bethke PM, Wasserman MD (1992) The stable isotope geochemistry of acid sulfate alteration. Econ Geol 87:225-262 Rye RO, Czamanske GK (1969) Experimental determination of sphalerite-galena sulfur isotope fractionation and application to ores at Providencia, Mexico. Geol Soc Am Abstr 7:195-196 Rye RO, Luhr JF, Wasserman MD (1984) Sulfur and oxygen isotopic systematics of the 1982 eruptions of El Chichón Volcano, Chiapas, Mexico. J Volcanol Geotherm Res 23:109-123 Sælen G, Raiswell R, Talbot MR, Skei JM, Bottrell SH (1993) Heavy sedimentary sulfur isotopes as indicators of super-anoxic bottom-water conditions. Geology 21:1091-1094 Sakai H (1957) Fractionation of sulphur isotopes in nature. Geochim Cosmochim Acta 12:150-169 Sakai H (1968) Isotopic properties of sulfur compounds in hydrothermal processes. Geochem J 2: 29-49 Sakai H (1983) Sulfur isotope exchange rate between sulfate and sulfide and its application. Geothermics 12:111-117 Sakai H, Casadevall TJ, Moore JG (1982) Chemistry and isotope ratios of sulfur in basalts and volcanic gases at Kilauea volcano, Hawaii. Geochim Cosmochim Acta 46:729-738 Sakai H, Desmarais DJ, Ueda A, Moore JG (1984) Concentrations and isotope ratios of carbon, nitrogen and sulfur in ocean-floor basalts. Geochim Cosmochim Acta 48:2433-2441
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 489 Sakai H, Dickson FW (1978) Experimental determination of the rate and equilibrium fractionation factors of sulfur isotope exchange between sulfate and sulfide in slightly acid solutions at 300 °C and 1000 bars. Earth Planet Sci Lett 39:151-161 Sakai H, Gunnlaugsson E, Tòmasson J, Rouse JE (1980) Sulfur isotope systematics in Icelandic geothermal systems and influence of seawater circulation at Reykjanes. Geochim Cosmochim Acta 44:1223-1231 Sakai H, Matsubaya O (1977) Stable isotopic studies of Japanese geothermal systems. Geothermics 5:97-124 Sakai H, Nagasawa H (1958) Fractionation of sulphur isotopes in volcanic gases. Geochim Cosmochim Acta 15:32-39 Sakai H, Ueda A, Field CW (1978) δ34S and concentration of sulfide and sulfate sulfurs in some ocean – floor basalts and serpentinites. In: Short papers of the Fourth International Conference on Geochronology, Cosmochronology and Isotope Geology. Zartman RE (ed) US Geological Survey Open File Report 78701 p 372-374 Salomons W (1971) Isotope fractionation between galena and pyrite and between pyrite and elemental sulfur. Earth Planet Sci Lett 11:236-238 Santacroce R (1983) A general model for the behaviour of the Somma-Vesuvius volcanic complex. J Volcanol Geotherm Res 17:237-248 Santosh M, Masuda H (1991) Reconnaissance oxygen and sulfur isotopic mapping of Pan-African alkali granites and syenites in the southern Indian Shield. Geochem J 25:173-185 Sasaki A, Ishihara S (1979) Sulfur isotopic composition of the magnetite-series and ilmenite-series granitoids in Japan. Contrib Mineral Petrol 68:107-115 Schiller WR, von Gehlen K, Nielsen H (1970) Hydrothermal exchange and fractionation of sulfur isotopes in synthesized ZnS and PbS. Econ Geol 65:350-352 Schmidt M, Siebert W (1973) Sulphur. In: Comprehensive Inorganic Chemistry. Vol 2. Bailar JC, Emeléus HJ, Nyholm R, Trotman-Dickenson AF (eds) Pergamon Press, Oxford, p 795-933 Schneider A (1970) The sulfur isotope composition of basaltic rocks. Contrib Mineral Petrol 25:95-124 Schwarcz HP (1973) Sulfur isotope analyses of some Sudbury, Ontario ores. Can J Earth Sci 10:1444-1459 Seal II RR (2006) Sulfur isotope geochemistry of sulfide minerals. Rev Mineral Geochem 61:633-677 Seal II RR, Alpers CN, Rye RO (2000) Stable isotope systematics in sulfate minerals. Rev Mineral Geochem 40:541-602 Seward TM, Barnes HL (1997) Metal transport by hydrothermal ore fluids. In: Geochemistry of hydrothermal ore deposits, 3d Edition. Barnes HL (ed) Wiley, New York, p 435-486 Seyfried WE, Bischoff JL (1981) Experimental seawater-basalt interaction at 300 °C, 500 bars, chemical exchange, secondary mineral formation and implications for the transport of heavy metals. Geochim Cosmochim Acta 45:135-149 Shanks WC III (2001) Stable isotopes in seafloor hydrothermal systems: vent fluids, hydrothermal deposits, hydrothermal alteration, and microbial processes. Rev Mineral Geochem 43:469-525 Shima M, Gross WH, Thode HG (1963) Sulfur isotope abundances in basic sills, differentiated granites, and meteorites. J Geophys Res 68:2835-2847 Shinohara H, Giggenbach WF, Kazahaya K, Hedenquist JW (1993) Geochemistry of volcanic gases and hot springs of Satsuma-Iwojima, Japan: following Matsuo. Geochem J 27:271-285 Simon AC, Ripley RM (2011) The role of magmatic sulfur in the formation of ore deposits. Rev Mineral Geochem 73:513-578 Sinclair AJ (1974) Selection of thresholds values in geochemical data using probability graphs. J Geochem Explor 3:129-149 Sinclair AJ (1986) Statistical interpretation of soil geochemical data. In: Exploration geochemistry: design and interpretation of soil surveys. FletcherWK, Hoffman SJ, Mehrtens MB, Sinclair AJ, Thompson I (eds) Rew Econ Geol, Society of Economic Geologists, El Paso, 3:97-115 Skinner BJ (1970) A sulfur lava flow on Mauna Loa. Pacific Sci 24:144-145 Smith JW, Doolan S, McFarlanc EF (1977) A sulfur isotope geothermometer for the trisulfide system galenasphalerite-pyrite. Chem Geol 19:83-90 Smitheringale WG, Jensen ML (1963) Sulfur isotopic composition of the Triassic igneous rocks of eastern United States. Geochim Cosmochim Acta 27:183-1207 Smithsonian Institution (1990). Momotombo. Bull Global Volcanism Network 15:6-8 Spilliaert N, Métrich N, Allard P (2006) S-Cl-F degassing pattern of water-rich alkali basalt: Modelling and relationship with eruption styles on Mount Etna volcano. Earth Planet Sci Lett 248:772-786 Sriwana T, van Bergen MJ, Varekamp JC, Sumarti S, Takano B, van Os BJH, Leng MJ (2000) Geochemistry of the acid Kawah Putih lake, Patuha Volcano, West Java, Indonesia. J Volcanol Geotherm Res 97:77-104 Steiner A, Rafter TA (1966) Sulfur isotopes in pyrite, pyrrhotite, alunite, and anhydrites from steam wells in the Taupo Volcanic Zone, New Zealand. Econ Geol 61:1115-1129 Strauss H (1997) The isotopic composition of sedimentary sulfur through time. Palaeogeogr Palaeoclimat Palaeoecol 132:97-118
490
Marini, Moretti, Accornero
Stull DR, Westrum EF, Sinke GG (1969) The Chemical Thermodynamics of Organic Compounds. Wiley, New York Sturchio NC, Williams SN, Sano Y (1993) The hydrothermal system of Volcan Puracé, Colombia. Bull Volcanol 55:289-296 Symonds RB, Gerlach TM, Reed MH (2001) Magmatic gas scrubbing: implications for volcano monitoring. J Volcanol Geotherm Res 108:303-341 Symonds RB, Janik CJ, Evans WC, Ritchie BE, Counce D, Poreda RJ, Iven M (2003) Scrubbing masks magmatic degassing during repose at Cascade-Range and Aleutian-Arc volcanoes. Open-File Report 03435, U.S. Geological Survey Symonds RB, Rose WI, Bluth GJS, Gerlach TM (1994) Volcanic-gas studies: methods, results, and applications. Rev Mineral 30:1-66 Takano B, Koshida M, Fujiwara Y, Sugimori K, Takayanagi S (1997) Influence of sulfur-oxidizing bacteria on the budget of sulfate in Yugama crater lake Kusatsu–Shirane volcano, Japan. Biogeochemistry 38:227-253 Takano B, Saitoh H, Takano E (1994) Geochemical implications of subaqueous molten sulfur at Yugama crater lake Kusatsu–Shirane volcano, Japan. Geochem J 28:199-216 Takano B, Watanuki K (1990) Monitoring of volcanic eruptions at Yugama crater lake by aqueous sulfur oxyanions. J Volcanol Geotherm Res 40:71-87 Taran Y, Fischer TP, Pokrovsky B, Sano Y, Aurora Armienta M, Macias JL (1998) Geochemistry of the volcanohydrothermal system of El Chichón Volcano, Chiapas, Mexico. Bull Volcanol 59:436-449 Taylor BE (1986) Magmatic volatiles: isotopic variation of C, H, and S. Rev Mineral 16:185–225 Taylor BE (1987) Stable isotope geochemistry of ore-forming fluids. In: Stable Isotope Geochemistry of Low Temperature Processes. Kyser TK (ed) Mineral Assoc Canada Short Course Handbook 13:337-445 Taylor BE (1988) Degassing of rhyolitic magmas: Hydrogen isotope evidence and implications for magmatic– hydrothermal ore deposits. In: Granites and Related Mineral Deposits. Taylor RP, Strong DF (eds), CIM Special Publication, p 33-49 Taylor BE (1991) Degassing of Obsidian Dome rhyolite, Inyo volcanic chain, California. In: Stable Isotope Geochemistry: A Tribute to Samuel Epstein. Taylor HP Jr, O’Neil JR, Kaplan R (eds), The Geochemical Society, Special Publication No. 3, p 339-353 Taylor BE, Wheeler MC, Nordstrom DK (1984) Oxygen and sulfur compositions of sulphate in acid mine drainage: evidence for oxidation mechanisms. Nature 308:538-541 Thode HG (1991) Sulphur isotopes in nature and the environment: an overview. In: SCOPE 43 - Stable Isotopes: Natural and Anthropogenic Sulphur in the Environment. Krouse HR, Grinenko VA (eds) Wiley, Chichester, p 1-26 Thode HG, Cragg CB, Hulston JR, Rees CE (1971) Sulfur isotope exchange between sulfur dioxide and hydrogen sulfide. Geochim Cosmochim Acta 35:35-45 Thode HG, Dunford HB, Shima M (1962) Sulfur isotope abundances of the Sudbury district and their geological significance. Econ Geol 57:565-578 Thode HG, McNamara J, Collins CB (1949) Natural variations in the isotopic content of sulphur and their significance. Can J Res B27:361-373 Thode HG, Monster J (1965) Sulfur isotope geochemistry of petroleum, evaporites and ancient seas. AAPG Mem 4:367-377 Thode HG, Monster J, Dunford HB (1961) Sulphur isotope geochemistry. Geochim Cosmochim Acta 25:159174 Torssander P (1989) Sulfur isotope ratios of Icelandic rocks. Contrib Mineral Petrol 102:18-23 Trofimov AV (1949) Isotopic composition of sulfur in meteorites and terrestrial objects. Dokl Akad Nauk SSSR 66:181-184 (in Russian) Tudge AP, Thode HG (1950) Thermodynamic properties of isotopic compounds in sulphur. Can J Res 28B:567578 Ueda A, Sakai H (1984) Sulfur isotope study of Quaternary volcanic rocks from the Japanese Islands Arc. Geochim Cosmochim Acta 48:1837-1848 Ueda A, Sakai H, Sasaki A (1979) Isotopic composition of volcanic native sulfur from Japan. Geochem J 13:269-275 Urey HC (1947) The thermodynamic properties of isotopic substances. J Chem Soc 1947:562-581 Urey HC, Greiff LJ (1935) Isotopic exchange equilibria. J Am Chem Soc 57:321-327 Varekamp JC, Kreulen R (2000) The stable isotope geochemistry of volcanic lakes, with examples from Indonesia. J Volcanol Geotherm Res 97:309-327 Varekamp JC, Ouimette AP, Kreulen R (2004) The magmato-hydrothermal system at Copahue volcano, Argentina. In: Water-Rock Interaction WRI-1. Wanty RB, Seal RR II (eds) Taylor & Francis Group, London p 215-218 Vinogradov AP (1958) Isotopic composition of sulphur in meteorites and in the earth. In: Radioisotopes in Scientific Research. Extermann RC (ed) Pergamon, New York, 2:581-591
Sulfur Isotopes in Magmatic-Hydrothermal Systems, Melts, & Magmas 491 Wallace PJ, Carmichael ISE (1992) Sulfur in basaltic magmas. Geochim Cosmochim Acta 56:1863-1874 Wallace PJ, Carmichael ISE (1994) Speciation in submarine basaltic glasses as determined by measurements of SKα X-ray wavelength shifts. Am Mineral 79:161-167 Watanabe T (1940) Eruptions of molten sulphur from the Siretoko-Iôsan volcano, Hokkaido, Japan. Jpn J Geol Geogr 17:289-310 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Williams H (1942) The Geology of Crater Lake National Park, Oregon. Carnegie Inst Wash Publ 540, 162 p Williams SN, Sturchio NC, Calvache VML, Mendez RF, Londoño CA, Garcia NP (1990) Sulfur dioxide from Nevado del Ruiz volcano, Colombia; total flux and isotopic constraints on its origin. J Volcanol Geotherm Res 42:53-68 Woodheaed JD, Harmons RS, Fraser DG (1987) O, S, Sr, and Pb isotope variations in volcanic rocks from the northern Marian Islands – Implications for crustal recycling in intraoceanic arcs. Earth Planet Sci Lett 83:39-52 Zheng Y (1990) Sulfur isotope fractionation in magmatic systems: models of Rayleigh distillation and selective flux. Chinese Journal of Geochemistry 9:27-45 Zientek ML, Ripley EM (1990) Sulfur isotope studies of the Stillwater Complex and associated rocks, Montana. Econ Geol 85:376-391
Marini, Moretti, Accornero
492
Table A-1. Equilibrium isotope fractionation factors of sulfur compounds with respect to H2S. A, B, and C represent the coefficients of the polynomial Equation (16). Compound
A
Dissolved sulfates and sulfate minerals Dissolved sulfates and sulfate minerals Dissolved sulfates and sulfate minerals Sulfites SO2 COS CaS, SrS, BaS MgS MoS2 FeS2 CoS2, NiS2, MnS2 ZnS FeS MnS, CoS, NiS CuFe2S3 CuFeS2 S, S8 HSCu5FeS4 CdS, CuS SnS PbS HgS Cu2S, Sb2S3 Ag2S S2CuS Cu2S Ag2S Bi2S3 ZnS (Sphalerite) PbS (Galena) FeS (Pyrrhotite) CuFeS2 (Chalcopyrite) CdS (Greenockite) Cu5FeS4 (Bornite) Cu3VS4 (Sulvanite) CuFe2S3 (Cubanite) FeNi2S4 (Violarite)
6.463 5.26 8.0 4.12 4.70 0.67 0.60 ± 0.10 0.50 ± 0.10 0.45 ± 0.10 0.40 ± 0.08 0.40 ± 0.10 0.10 ± 0.05 0.10 ± 0.05 0.10 ± 0.05 0.05 ± 0.05 −0.05 ± 0.08 −0.16 −0.06 ± 0.15 −0.25 ± 0.10 −0.40 ± 0.10 −0.45 ± 0.10 −0.63 ± 0.05 −0.70 ± 0.10 −0.75 ± 0.10 −0.80 ± 0.10 -0.21 0.04 -0.06 -0.62 -0.67 0.10 -0.64 0.25 0.05 0.18 -0.07 0.03 0.04 -0.27
B
5.82 0.43
C
T (°C) interval
0.56 ± 0.5 6.0 ± 0.5 ± 1.0 −5.0 −0.5 ± 0.5 −1.15
200-400 200-350 >400 >25 350-1050 >25
200-700 50-705 200-600
± 0.5 −0.6
−1.23
200-600 200-400 50-350
−1.23 280-490 510-630 280-700 150-600
References: (1) Ohmoto and Lasaga 1982; (2) Ohmoto and Rye 1979; (3) Seal 2006; (4) Li and Liu 2006.
Refs. (1) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (2) (3) (3) (3) (3) (4) (4) (4) (4) (4) (4) (4) (4) (4)
15
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 493-511, 2011 Copyright © Mineralogical Society of America
Interactions Between Metal and Slag Melts: Steel Desulfurization Jean Lehmann and Michèle Nadif ArcelorMittal Global R&D Maizieres Process Process Engineering and Steelmaking BP 30320, 57283 Maizières-lès-Metz Cedex, France [email protected]
INTRODUCTION Knowledge of the partitioning behavior of sulfur between molten metals and silicates is of interest for understanding processes in the early solar system and in the deep Earth at the mantle/core boundary, but also in technical applications such as the production and purification of steel. Thermodynamic aspects relevant to processes in the Earth and in the early solar system are presented by Baker and Moretti (2011) and Ebel (2011) of this review volume. Here we focus on interactions between molten metals and slags which are the basis for desulfurization of molten metals during steel production. Desulfurization is an important step in the refining of high-quality steels for various applications. The production of liquid steel can be described schematically as follows. Steel can be produced directly from iron ore which is reduced to hot metal in a blast furnace (BF) or by melting scraps in an electric arc furnace (EAF). The hot metal is transformed into steel by oxidizing dissolved carbon in a basic oxygen furnace (BOF). During this process, the phosphorus content is reduced as well. Before the BOF operations, a first desulfurization is performed in the transfer vessel from BF to BOF. After scrap melting or refining in the BOF, steel is “tapped” in a ladle for secondary steelmaking operations, i.e., further decarburization (especially for Ultra-Low Carbon steels), deoxidation (to reduce the oxygen content), alloying (to reach the target composition) and further desulfurization, if needed. In solid steel, sulfur is mainly present as manganese sulfide (MnS) inclusions. MnS inclusions affect the processing and properties of steel. Their volume fraction, size, shape and distribution depend on many factors. The most important factors are the S-content, the solidification rate, the degree of hot and cold deformation and the hot working temperatures. Since the inclusions are more plastic than steel, they act as crack initiation sites and zones of weakness during deformation. Therefore, S is detrimental to bendability, ductility, toughness, formability, weldability and corrosion resistance. The maximum S-content acceptable for high quality steel grades—such as Advanced High Strength Steels sheets, wheels and tubes—has decreased over the years, in a context of always more severe conditions applied to steel due to thickness reduction and increase in deformation rate. The most demanding applications of steel especially for flat products (used for car body, wheels, packaging applications, etc.) require very low levels of S (<0.003 to <0.001 wt%). It is necessary to decrease the S-content during secondary steelmaking operations (i.e. during ladle refining, before casting) in order to satisfy the ultra-low S specifications. Figure 1 shows the favorable effect of low S-content of steel on the increase of resistance to ductile tearing in transverse direction. 1529-6466/11/0073-0015$05.00
DOI: 10.2138/rmg.2011.73.15
Lehmann & Nadif Resistance to ductile tearing (J)
494
25 20 15 10
Thickness 1.9 mm
5
Thickness 2.2 mm
0
0
15
30
45
60
75
S (wt% x 10000)
figure 1. Scheme of resistance measurement (left) and plot of resistance to ductile tearing in transverse direction vs. steel S-content (right).
However, S is beneficial to machinability. Thus, higher S-levels are required for applications which are extensively machined during manufacturing. Inclusion population must be carefully adjusted, so that oxides and sulfides that precipitate remain plastic and protect carbide cutting tools from abrasion during steel product manufacturing.
PaRTITIONINg Of SULfUR BeTweeN SLag aND MeTaL MeLTS Sulfide capacity During secondary steelmaking operations (i.e. alloying operations following the transformation of hot metal into steel in converters), the control of the steel S-content at very low levels requires knowledge of the properties of metallurgical slags, i.e., the oxide liquid phases which cover the liquid steel during steelmaking applications. At the low oxygen fugacities prevailing during iron and steel making operations, the presence of sulfates can be excluded (Richardson and Fincham 1954), and the reaction controlling the S partitioning between steel and slag can then be written as: MpO (slag) + S (metal) → MpS (slag) + O (metal)
(1)
where M stands for any cation present in the slag, p is the stoichiometric coefficient of the corresponding oxide/sulfide in the slag and O(metal) and S(metal) are dissolved oxygen and sulfur, respectively, in the liquid metal. The capacities of a slag to extract S from a gas phase can be expressed through a reaction analogous to (1) but involving gas species: MpO (slag) + ½ S2 (g) → MpS (slag) + ½ O2 (g)
(2)
Fincham and Richardson (1954) have introduced the concept of sulfide capacity CS defined as: P CS = O 2 PS 2
12
( wt% S)slag
(3)
where PO2 and PS2 are the partial pressures of O2 and S2 in the gas phase. For steelmaking applications, this quantity is often modified to refer to diluted solutions containing 1 wt% of the respective species in liquid iron: a CS′ = O ( wt% S)slag aS
( 4)
Interactions Between Metal & Slag Melts: Steel Desulfurization
495
where ai is the activity of species i in the liquid metal. The relationship between the two quantities depends only on the temperature (Rist et al. 1974): 2154 CS′ = exp − + 3.166 × CS T
(5)
The usefulness of the definition appears when expressing the sulfur partition coefficient at slag/metal equilibrium DS: DS =
( wt% S)slag [ wt% S]metal
= CS′
fS aO
(6)
where fS is the activity coefficient for sulfur in the liquid metal, defined as fS = aS/[wt% S]metal, and aO is the oxygen activity prevailing at the slag/metal interface where the desulfurization reaction takes places. This relation enlightens the role of the different factors controlling the reaction: •
The sulfide capacity depends only on the slag composition and on the temperature.
•
fS depends only on the metal composition and on the temperature. This coefficient is close to 1 for low-alloyed steel but can differ markedly from this value for other metal liquids. For example for “hot” metal (C-saturated liquid iron), this factor has a value close to 6 around 1500 °C due to the effect of carbon on S activity. Such a large value favors the desulfurization reaction.
•
In low alloyed steel, aO value is practically equal to the content of dissolved oxygen in the steel expressed in wt%. This activity of oxygen aO depends both on metal and slag compositions and on temperature. It emphasizes that an efficient desulfurization can only take place under very reducing conditions.
Modeling Different approaches have been proposed to correlate the sulfide capacity with the composition of slags. Among the most popular models are those that are based on the optical basicity concept (Sosinsky and Sommerville 1986) which expresses the sulfide capacity as a linear function of slag compositions. The merit lies in the simplicity of the mathematical expressions. The use on large composition domains however can lead to large prediction errors (Gaye et al. 1989). Other models use more rigid thermodynamic concepts, e.g., the model proposed by Nzotta et al. (2000), which takes into account the possible interactions among the different components of the melt. Zhang and Toguri (1987) built their approach on the concept of Flood et al. (1953), which restricts the application domain to basic slags that contain iron oxides in equilibrium with liquid iron. Reddy and Blander (1987) used the Flory polymerization model to describe the structure of the melt and to derive the expression of the sulfide capacity. This approach was later modified by Pelton et al. (1993). The most precise and promising models are those that are based on a comprehensive thermodynamic approach as in Lehmann and Gaye (1992-1993) and Kang and Pelton (2009), in which the sulfide capacity is derived from a formulation of the Gibbs free energy of mixing in the melt. The major characteristic is that these models provide a description of the short range ordering (SRO) phenomenon occurring in this kind of ionic solution. Depending on the model, this SRO phenomenon can be depicted by cells containing one central anion surrounded by two cations as in the cell model (Lehmann et al. 2008) or by quadruplets containing two cations and two anions as in the Modified Quasichemical Model (Pelton et al. 1993). These models use the quasichemical approach (Guggenheim et al. 1952) to obtain an approximation of the mixing entropy. The cell model is included in the proprietary computer code CEQCSI (Chemical EQuilibrium Calculations for the Steel Industry) for its use inside the
496
Lehmann & Nadif
group ArcelorMittal and is not available to the scientific community. Most of the models and calculations presented in this paper are based on this program, which is particularly adapted to problems relevant for the steel industry. Commercial programs such as FactSage (http://www. factsage.com/) may be used for some specific questions. Figure 2 is an illustration of how these concepts are able to reproduce the typical shape of activity curves encountered in slag systems through the description of cation-cation interactions they provide. When introducing SiO2 molecules in a pure lime system, the formation energy of asymmetric Si-O-Ca cells corresponding to the reaction ½ Ca-O-Ca + ½ Si-O-Si → Si-O-Ca
(7)
is so high that all the cations Si4+ introduced in the melt are used to form these asymmetric cells decreasing systematically the number of symmetric Ca-O-Ca cells. The number of symmetric Si-O-Si cells remains at a very low level as shown in the left hand side of the figure. The Si-O-Si cell number starts to increase only when all the symmetric cells Ca-O-Ca have disappeared and when all the Ca2+ cations are trapped into Ca-O-Si cells. This abrupt change is directly reflected in the activity of SiO2 displayed in the right hand side of the figure since this activity depends mainly on the number of Si-O-Si cells. Recently, in order to have a more precise SRO description by taking into account interactions between anions, Lehmann et al. (2006) developed the Generalized Central Atom (GCA) model as a generalization of the cell model but with a better approximation of the mixing entropy. In both models, the sulfide capacity description is simply obtained by allowing S/O atoms substitution in the cells and assessing the associated model parameters. It considers cells with sulfur as the central atom (for instance Si-S-Ca) or quadruplets where one or two O atoms can be replaced by S. Figure 3 gives an idea of the precision of the prediction by the cell model.
Desulfurization during secondary metallurgy operations One of the interests to develop precise prediction models is that they provide robust tools to extrapolate experimental data which can then be used to produce useful diagrams for chemically complex systems. Figure 4 compares the desulfurization power of some common oxides used in ferrous pyrometallurgy. It shows why lime is the most common desulfurization agent. MnO and FeO also have good desulfurizing properties, but their concentrations in slag are limited owing to the reducing conditions prevailing during desulfurization operations. In Figure 5 isovalue lines for the quantity “−log(C′S)” calculated with the cell model are drawn at 1500 °C in the Al2O3-CaO-SiO2 system (which is the basic system for secondary steelmaking operations) for composition domains corresponding to completely liquid slags at this temperature. The diagram shows that the maximum values of the sulfide capacity are reached for compositions that are saturated with lime (lower part of the diagram) and that these isovalue lines remain parallel to the lime saturation isotherm for compositions of slags that are poor in silica. Figure 6 has even a more straightforward practical application since it provides information about the equilibrium partitioning of S between slag and metal. In this case, the oxygen activity aO at the slag/metal interface is controlled by the Al content of the steel. Steelmakers refer to this kind of grade as “Al-killed” steels since Al additions are used to deoxidize the steel. For each point on the diagram, the activity of Al2O3 has been calculated with the cell model and the oxygen activity has then been computed through the reaction: 2 Al (metal) + 3 O (metal) → Al2O3 (slag)
(8)
The sulfur activity in the metal has been kept in the expression of the partition coefficient (wt% S)slag/aS = C′S/aO in order to have a diagram applicable to any composition of steel.
Interactions Between Metal & Slag Melts: Steel Desulfurization 22
11
1.8 1.8
0.9 0.9
1.6 1.6
0.8 0.8
Si-O-Si Si-O-Si
1.4 1.4
0.7 0.7
Activity Activity
Rij R ij
1.2 1.2 11
Ca-O-Ca Ca-O-Ca
0.8 0.8
Si-O-Ca Si-O-Ca
0.5 0.5
0.3 0.3 0.2 0.2
0.2 0.2
0.1 0.1
40 40
60 60
wt%SiO wt%SiO22
80 80
SiO SiO 22
0.4 0.4
0.4 0.4
20 20
CaO CaO
0.6 0.6
0.6 0.6
00 00
497
00 00
100 100
20 20
40 40
60 60
wt%SiO22 wt%SiO
80 80
100 100
- log(CS) cell model
binary system CaO-SiO at 1600 °C and resulting figure 2. Distribution of the various types of cells in the 2 activity curves as calculated by CEQCSI using the cell model. Reference states for the activities are respectively pure lime and pure silica in their stable form at 1600 °C. Rij is the mole number of cells i-O-j.
SiO2-Al2O3-MgO-CaO
figure 3. Plot of sulfide capacity calculated with the cell model vs. sulfide capacities from experimental data. Double arrows indicate the CS ranges covered by different oxide systems.
SiO2-Al2O3-MgO-CaO-MnO SiO2-Al2O3-MgO-CaO-MnO-FeOx
-log(CS) exp. -1.0 -1.5 Abraha m a nd Richardson (1960) - CaO-SiO2
-2.0
Carter and Macfarlane (1957) - CaO-SiO2
log(CS)
-2.5
Drakaliysky et al. (1996) - CaO-Al2O3
-3.0
Abraha m a nd Richardson (1960) - MnO-SiO2
-3.5
Nzotta et al. (1998) - MnO-Al2O3
-4.0
Nzotta et al. (1999) - FeO-SiO2
-4.5 -5.0 0.3
0.4
0.5
Nbase
0.6
0.7
0.8
figure 4. Variation of the sulfide capacities with the mole fraction of the basic end-member at 1500 °C for binary oxide systems, composed by a basic oxide (CaO, MnO, MgO) and either an acid (SiO2) or an amphoteric (Al2O3) oxide.
Lehmann & Nadif
498
SiO2
80
20
iO 2 %S
wt %C aO
wt
40
60
80
CaO
20
20
wt%Al 2O 3 60
80
Al2O3
figure 5. Calculated (cell model) iso-“−log(C′S)” lines in Al2O3-CaO-SiO2 melts at 1500 °C. “2” on the diagram means log(C′S) = −2.
wt%CaO SiO wt % 2
wt%Al2O3 figure 6. Sulfur partition coefficients between slag and metal at equilibrium for a metal with a Al activity of 0.03 which corresponds to a low-alloyed steel containing 0.03% Al (Béranger et al. 1994).
Figure 6 has been plotted for slags containing 5 wt% of MgO since this content is not far from the saturation value for very basic slags. Approaching such a value is a consequence of the partial dissolution of the ladle refractory lining or is a result of the deliberate addition of magnesia or dolomite to avoid as much as possible any refractory erosion. Close to the lime saturation line (bottom of the diagram) the partition coefficient can exceed 1000, but decreases very rapidly when the slag gets enriched in Al2O3 and/or in SiO2. As for the sulfide capacity, the isovalue lines of the partition coefficients are nearly parallel to the lime saturation line.
Interactions Between Metal & Slag Melts: Steel Desulfurization
499
Methods of desulfurization Evolution of S-content during liquid steel refining The S-level in liquid steel at the end of the primary steelmaking step depends on the type of refining operation, i.e., whether an EAF or a BOF is used. In the EAF route, liquid steel is produced by melting scraps. The S-content in liquid steel after EAF operations is typically 0.015-0.045 wt% due to the high sulfur content in scraps, and also in coal used as fossil energy and/or for slag foaming. Indeed coal combustion by oxygen can be used to partially replace electrical energy for economical and/or industrial reasons. Achieving low concentrations of S, down to 0.005 wt%, necessitates an effective desulfurization treatment, which is usually carried out by stirring of steel and slag in a ladle furnace. In the BOF route, hot metal coming from the blast furnace is transformed into steel by oxidation of C and P (as well as Mn and Si), which is dissolved in the raw material, in the converter by O2 gas. The S-content in liquid steel is generally 0.01-0.02 wt% after this operation. For the production of ultra-low S-grades, the hot metal is desulfurized prior to BOF charging, leading to a much lower concentration of S in the liquid steel before secondary metallurgy. This is classically performed in dedicated vessels (“torpedo cars”) by adding Mg-bearing materials or CaC2 directly into the metal. In the following we will focus on the processes and the evolution of the sulfur content in the BOF route. Converter. The S-content in steel at the BOF converter tapping is generally higher than the S-content in hot metal and it is in the range of 0.005-0.009 wt% S for ultra-low S-grades (Fig. 7). The cold metallic materials (mainly scraps) that are charged into the BOF converter either as a complement to hot metal or as a coolant for temperature adjustment are selected for ultralow sulfur grades with <0.003 wt% S to minimize the sulfur input. Commercial scrap, often polluted, is generally not allowed, except for steel plants equipped with vacuum tank degasser, a ladle furnace if there is time enough for desulfurization by steel-slag stirring, or steel plants equipped with a lime injection station in the ladle, which permits very efficient desulfurization (Nadif et al. 2008). 100
Hot metal at converter charging Steel at converter tapping
S (wt% x 10000)
80
60
40
20
0
A
B
C
D
E
Figure 7. Comparison of S-content in hot metal at BOF converter charging and in BOF at tapping for ultra-low sulfur grades with <0.003 wt% S (A, B, C, D and E are different Western Flat Carbon Europe ArcelorMittal steelplants and comparison has been done for a one-year production).
Lehmann & Nadif
500
A low S-content of the lime, which is needed for dephosphorization in the converter, is also favorable to minimize the sulfur input in the BOF converter. The use of lime with <0.0350.040 wt% S is highly recommended for the production of ultra-low S-grades. The same precautions are also valid for other non-metallic materials, such as dolomite, used mainly for protecting the refractory material in the converter against wear. A study was carried out on databases from ArcelorMittal steel plant A to identify the S budget in the BOF (Table 1). Sulfur input through hot metal, sulfur output in steel and converter slag were known through an analysis. Sulfur input through scraps, pig iron, iron ore, lime, and dolomite was calculated considering an averaged constant S-content. In the studied example, the greatest part of the sulfur input in the converter remains in the steel tapped to the ladle. About 10% escapes as SO2 during oxygen blowing, and about 30% is dissolved by the slag, mainly during blowing with oxygen (without gas injection the slag traps less then 10% of sulfur input into the BOF converter during metallurgical stirring). Indeed, by using the BOF route the S partition coefficient between slag and steel is generally limited to a value between 5 and 12 due to the high oxygen activity aO at the end of converter refining. Table 1. Sulfur budget in a BOF converter using the example of Western Europe ArcelorMittal steel plant A. Share of sulfur input in converter (%)
S output S in steel before converter tapping SO2 formation S dissolved in slag during O2 blowing S dissolved in slag during metallurgical stirring
61 11 20 8
An example of a converter slag composition is given in Table 2. For grades with <0.003 wt% S, the average lime content is about 50 wt% with an average basicity CaO/SiO2 of 4 and the average slag FeO content is 23 wt%. The silica content in the slag is generally about 12-13 wt%. During BOF converter tapping, it is important to avoid BOF slag carry-over to the ladle because secondary metallurgy is performed under reducing conditions, and FeO and MnO contained in the BOF slag would cause severe reoxidation. Modern converters are generally equipped with infrared cameras to detect the arrival of converter slag associated with a dart or a pneumatic slag stopper to interrupt converter tapping. During tapping, different products are added to the steel; e.g., deoxidizing agents such as aluminum, and oxides such as lime, magnesia, and/or calcium aluminates. The secondary metallurgy slag begins to form, resulting from the reaction of lime and alumina (either added or produced by aluminum oxidation). This slag is mixed mechanically with the steel by the
Table 2. Slag composition of converter slag for desulfurized grades for the production of ultralow S steels using the example of Western Europe ArcelorMittal steel plants. Slag composition (wt%)
SiO2
al2O3
feTot
MnO
MgO
CaO
CaO/SiO2
Average
12.7
1.9
18.0
3.1
5.4
50.2
4.0
Sigma
1.2
0.5
1.4
0.7
2.0
3.2
0.5
<0.003 wt% S
Interactions Between Metal & Slag Melts: Steel Desulfurization
501
tapping stream in the ladle. Desulfurization begins as soon as the deoxidized steel reacts with the lime-rich slag. Thus, the S-content in the ladle after tapping is generally reduced when compared to the converter for low alloyed steels. Gas stirring in the ladle during and/or after converter tapping enhances the desulfurization process. Gas stirring may be achieved by the injection of Ar through porous plugs that are placed at the ladle bottom or by intense argon injection through an immersed lance. There can be some S pick-up during ladle tapping, due to S contained in recarburizing additions and in ferroalloys, which may contain sulfide inclusions, especially Mn-bearing alloys. Secondary metallurgy. Various methods are available for the production of ultralow sulfur steels with <0.003 wt% S in secondary steelmaking. The two main methods are the desulfurization by steel-slag stirring (as schematically represented in Fig. 8) and the desulfurization by lime based powder injection.
figure 8. Schematic representation of gas stirring techniques in the ladle.
Desulfurization by steel-slag stirring can be performed in several reactors: •
Desulfurization station at atmospheric pressure with stirring by Ar injection through a porous plug and/or by using a lance,
•
Ladle furnace with argon stirring through a porous plug (rarely through a lance)
•
Vacuum tank degasser with very strong gas stirring power.
Desulfurization by injection of lime-based powder can be carried out: •
By lance deeply immersed in the ladle at atmospheric pressure,
•
In the RH (Rheinstahl Heraeus) process. Figure 9 shows a scheme of different existing technologies for RH desulfurization.
Figure 10 shows an example of the evolution of the S-content during these steelmaking routes (except RH) for grades with <0.003 wt% S. The specific conditions of the different treatments will be described in the following paragraphs.
Desulfurization of steel by steel-slag stirring Desulfurization by steel-slag stirring mechanisms. As described in the first part of this chapter, desulfurization is an exchange of O and S between steel and slag (see Eqn. 1). For a
Lehmann & Nadif
502
Ar + powder
Powder dispenser RH vacuum vessel
Pumping group
Ar + lime based powder
Powder dispenser RH vacuum vessel
Pumping group
Emerged lance
Downward snorkel
Ar
Pumping group
Pulverized lime based powder
Ar
Ar
Surface slag
Surface slag Immersed lance below upstream snorkel
RH vacuum vessel
Vacuum vessel
Pumping group
Immersed opposed tuyeres
Emerged lance
Pulverized lime based powder
Powder dispenser
Ar + lime based
Ar
powder
Surface slag
Figure 9. Schematic representation of different technologies of lime-based powder injection used in the RH (Rheinstahl Heraeus) process.
Surface slag
100
A-Stir station
B-Stir station
B-VTD
C-VTD
D-LF
E-INJ
S(wt % x 10000)
80 60 40 20 0 Converter
Ladle after tapping
End desulfurization
Tundish / Mold
Figure 10. Evolution of S-content in liquid steel from BOF converter to continuous casting for ultra-low sulfur grades with <0.003 wt% S (example of Western Flat Carbon Europe ArcelorMittal steelplants). VTD: vacuum tank degasser – LF: ladle furnace – INJ: powder injection.
Interactions Between Metal & Slag Melts: Steel Desulfurization
503
given steel composition, the equilibrium partitioning of S between slag and metal increases with an increasing slag sulfide capacity, and a decreasing oxygen activity in steel. As can be seen in Figure 6, a high sulfur partition coefficient DS is obtained for liquid calcium aluminate slags near lime saturation. Desulfurization occurs by reaction of steel with the top slag on one hand, and with slag droplets generated by inert gas stirring and entrapped deeply in the liquid bath. This was evinced by the observation of particles present in steel samples, forming two distinct populations (Lachmund et al. 2003). Therefore, the amount of slag and its composition, and the liquid bath stirring conditions (time, intensity) are important process parameters for steel desulfurization performances. Process of desulfurization by steel-slag stirring. Typical slag-forming additions, slag composition and Ar stirring conditions to achieve efficient steel desulfurization will be described hereafter.
Slag formers additions For desulfurization by steel-slag stirring at atmospheric pressure of grades with <0.003 wt% S, the addition of lime (CaO) during BOF converter tapping is typically between 5 and 6.5 kg/t, depending on the initial S-content of the metal and on the secondary metallurgy equipments (reactors) available in the steel plant. For desulfurization by steel-slag stirring at atmospheric pressure, fluorspar (CaF2) additions are often made to obtain a fluid and reactive desulfurization slag, especially if the steel plants are not equipped with a ladle furnace, which helps in the early formation of liquid slag. Typically, the CaF2/CaO ratio is about 1:5 to 1:4 in the absence of a ladle furnace. For the ladle furnace process, the CaF2/CaO ratio can be reduced (typically to about 1:10) or totally suppressed. In this case, the fluorspar addition at the BOF converter tapping is being replaced by synthetic calcium aluminates that facilitate slag formation, but a larger lime addition is necessary to maintain the desired CaO/Al2O3 ratio necessary for efficient desulfurization. This substitution of CaF2 by Ca-aluminates is not easy in the absence of a ladle furnace, especially because of composition scatter owing to unwanted slag contributions (e.g., BOF slag carry over). Synthetic calcium aluminates also have higher cost. On the contrary, no fluorspar is needed for steel desulfurization by steel-slag stirring in a vacuum tank degasser, because of the strong stirring power conferred by the vacuum to the injected gas. Several steel production sites have MgO addition at the BOF converter tapping, either as magnesia or dolomite, to protect the slag line of the ladle refractory lining which is mainly composed of MgO, against corrosion by desulfurization slag, especially for lower S-grades, as strong stirring increases the kinetics by renewing slag refractory interface.
Slag composition The slag composition results mainly from the addition of minerals performed at the BOF converter tapping, oxides formed during deoxidation and reoxidation of steel, the BOF slag carried over to the ladle and the slag remaining from the previous production performed in the ladle. The BOF slag is a primary source of FeO, MnO and SiO2. Table 3 illustrates a typical secondary metallurgy slag composition for Al-killed steels at the end of steel-slag stirring. Compared to BOF slags (Table 2), note for example that they contain much less iron oxides. For grades with <0.003 wt% S, the average apparent lime content of this secondary metallurgy slag is about 53 wt% with a SiO2 content on the order of 10 wt% and an apparent CaO/Al2O3 ratio of 2.3. Apparent means that the calcium added as CaF2 is not considered.
Stirring conditions For grades with <0.003 wt% S, the steel desulfurization time by steel-slag stirring may
Lehmann & Nadif
504
Table 3. Typical secondary metallurgy desulfurization slag composition at the end of secondary metallurgy for grades with <0.003 wt% S: example of Western Flat Carbon Europe ArcelorMittal steel plants (BOF route). wt%
SiO2
al2O3
feO
MnO
MgO
CaO *
CaO*/ al2O3
Average
10.3
23.1
2.9
1.2
6.6
52.7
2.3
Sigma
3.2
3.7
3.3
1.2
1.5
4.6
0.5
* Apparent lime content, calcium present in CaF2 is not deduced
vary a lot among steel plants. The time for desulfurization by steel-slag stirring may be as short as 5 to 7 min for a very high gas flow rate (1000 L/min STP through a porous plug plus 2000 L/ min STP through a lance; STP = volume at standard temperature (25 °C) and pressure (1 bar)), but such low durations are exceptional. The time is often longer, typically 14 min to 20 min for an Ar flow rate between 500 and 1200 L/min STP injected either through a porous plug or a lance. The longest desulfurization times are encountered in plants which are equipped with a ladle furnace. In this case thermal losses during stirring are compensated by arc heating. Analysis of steel desulfurization performances. The simplified formalism developed by Riboud et al. (1985) can be used to describe desulfurization of liquid metals. To a first approximation, the decrease of the S concentration in steel can be described as: d [ wt% S]metal dt
= − KS ×
( wt% S(t ) )slag A × [ wt% S(t )]metal − V DS
( 9)
with t = treatment time (s), KS = desulfurization kinetic coefficient (m/s), A = surface of the slag / metal interface (m2), V = metal volume (m3), [wt% S(t)]metal = steel S-content at time t, (wt% S(t))slag = slag S-content at time t. Hereafter, the subscript “metal” will be omitted and “wt% S” into square brackets will refer to the S-content of the metal. The integration of the kinetic equation allows one to calculate the desulfurization ratio R, assuming first order kinetics, a constant sulfur partition coefficient DS, no S in the slag at the beginning of desulfurization and no S loss to the gas phase: [ wt% S]final − [ wt% S]0 R= = [ wt% S]0
1 1 − exp − B × (1 + ) λ 1 1+ λ
(10)
where l is a thermodynamic parameter with
l = DS × M
(11)
M is the slag mass (t) per ton of steel, B is a kinetic parameter defined as: B = −KS ×
A ×t V
(12)
The desulfurization kinetics are favored by a high stirring intensity. Assuming a constant DS, the desulfurization ratio R depends on two parameters: •
The thermodynamic parameter l that corresponds to the slag desulfurization potential, taking into account the S equilibrium partition coefficient between slag and steel DS and the specific amount of slag M in ton of slag per ton of steel
Interactions Between Metal & Slag Melts: Steel Desulfurization •
505
The kinetic parameter B that depends on the stirring time and Ar flow rate. This parameter includes the effect of stirring on the increase of interfacial area through the formation of slag droplets entrapped in steel.
Figure 11 illustrates the factors which can limit the performance of this treatment. For instance, if the value of DS·M is too low (insufficient slag mass or inappropriate slag composition), the desulfurization ratio will not exceed a certain value regardless of the intensity of the duration of the stirring. To analyze steel desulfurization performances, the ArcelorMittal CEQCSI model can be used to determine the DS between steel and slag. Slag mass is evaluated by multiple mass balances based on slag components and the consumption of deoxidizers. The B parameter is evaluated based on the stirring and geometrical conditions (Riboud and Vasse 1985). Figure 11 shows an example of positioning for steel-slag stirring at atmospheric pressure (8 min desulfurization stirring with 1500 L/min STP through a porous plug and 2000 L/min STP through a lance). To enhance desulfurization in such a case, it is not necessary to increase the specific amount of slag, or DS, but rather to optimize desulfurization kinetics. Numerical modeling with Computational Fluid Dynamics codes, such as Fluent© (www.ansys.com), are a help for optimizing steel flow in ladles. Note that the DS·M-B desulfurization graph is to be applied on deoxidized steel and for a slag where the MnO and FeO oxides are already reduced. Indeed, the formalism assumes that the sulfur transfer in the metal is the limiting rate and the slag has already at the beginning of the reaction the appropriate composition, i.e. is well reduced, to obtain instantaneously the partition coefficient DS. The time that is required to deoxidize the slag should therefore not be included in the B parameter for already low S-content before steel desulfurization, and as the time needed for slag reduction is not known (lack of samples) and cannot be deduced from desulfurization stirring time, the B parameter in the DS·M-B desulfurization graph of Figure 11 is probably overestimated. Table 4 shows deviation of S from equilibrium for steel compositions with <0.003 wt% S in the metal evaluated using the CEQCSI model, for desulfurization by steel-slag stirring in different reactors of ArcelorMittal Western Flat Carbon Europe BOF steel plants. The
Thermodynamic parameter Ds.M (-)
25
20
15
R Desulfurization rate
10
0.90 0.87 0.85
5
0.80 0.75 0.67 0.50
0
0
2
4 Kinetic parameter B (-)
6
8
figure 11. Desulfurization by steel-slag stirring: Example of positioning in (DS·M, B) plane.
Lehmann & Nadif
506
Table 4. Deviation of S-contents from equilibrium values for grades with <0.003 wt% S, evaluated using the CEQCSI model for desulfurization by steel-slag stirring in different reactors of ArcelorMittal Western Flat Carbon Europe BOF steel plants.
Stirring station at atmospheric pressure Vacuum tank degasser (P < 500 Pa)
Stirring conditions (L/min argon STP)
Steel-slag stirring time (min)
Seq (wt%)
Deviation of S from equilibrium (wt%)
1200 (deep lance)
14
0.00016
0.0019
1500 (porous plug) + 2000 (lance)
6
0.00019
0.0019
500 (porous plug)
20
0.00003
0.0010
deviation from equilibrium is significantly higher for desulfurization in a stirring station at atmospheric pressure than for desulfurization in a vacuum tank degasser. Desulfurization performances depend on the type of reactor: •
Ultra-low S-grades, down to < 0.0010 wt% S, can easily be obtained by steelslag stirring in the vacuum tank degasser, regardless of the initial S-content before vacuum (see for example steel plants B, C in Fig. 10). The low pressure over the bath increases the stirring power by increasing gas bubble dilatation during their rise to the surface. The B parameter in the case of vacuum tank degasser is 26 to 46 times higher than the values calculated at atmospheric pressure,
•
Steel S-content of 0.0020 wt% can be obtained for steels in which the S-content on final products must be less than 0.003 wt% S by steel-slag stirring at atmospheric pressure (see for example steel plants A, B, D in Figure 10). S-levels of 0.0015 wt% are obtainable for steels whose S-content on final products must be less than 0.002 wt% S.
•
Steel grades with < 0.0015 wt% S can be produced at atmospheric pressure by the ladle furnace route by extensive steel-slag stirring with optimized slag composition.
Desulfurization of steel by lime powder injection Mechanisms. Ultra fine lime powder is injected with Ar as a carrier gas for steel desulfurization. There are two desulfurizing mechanisms, which were evinced by the observation of particles present in liquid steel, forming two different populations (Posch et al. 2002; Lachmund et al. 2003): •
A transitory reaction between high desulfurizing flux particles and steel during their ascension through the melt,
•
A permanent reaction of desulfurization by emulsified top slag.
A large part of the obtained desulfurization is due to the transitory reaction. The injection flow rate should not be too high, to keep a high residence time of injected flux particles in the bath. The injected desulfurizing powder must have as fine a grain size as possible, to obtain the largest specific contact area for reaction with liquid steel. The contact area is defined as the product of the number of particles and the surface area of each particle. It is inversely proportional to the particle radii, since the number of particles is inversely proportional to the volume of individual particles, for a given total powder volume in kg per t of liquid steel.
Interactions Between Metal & Slag Melts: Steel Desulfurization
507
It is necessary to have a desulfurizing slag on top of the metal during injection for the floating S-rich lime particles to dissolve. Sulfur reversion (i.e., sulfur going back from the steel to the slag) takes place if the S-rich particles reach an oxidized, silica-rich top slag, decreasing desulfurization potential. Thus, it is necessary to avoid BOF converter slag carry-over to the ladle to limit the FeO+MnO and SiO2 content of the ladle slag. Process of steel desulfurization by injection. Steel desulfurization can be carried out by injection of a desulfurizing agent into a ladle at atmospheric pressure or into the vacuum vessel of a RH (Rhein). Powder injection is performed through a lance immersed deeply in the ladle. The injected powder is usually composed mainly of lime (Gantner 2006; Nadif et al. 2008). For example, in steel plant E on Figure 10, the powder that is injected into the ladle at the injection station is a lime-based powder with 88 wt% CaO, 5 wt% CaF2, 2.5 wt% SiO2, and 2.5 wt% MgO. Its grain size distribution is such that 70% of the grains are < 1 mm, 100% < 2 mm. The powder is transported pneumatically. The amount of injected powder averages close to 3 kg/t. The injection throughput is 90 kg/min for 10 min with 1350 L/min argon STP. The lance is immersed at around 3.2 m depth (0.5 m from the ladle bottom). The Ar flow rate through a porous plug is 350 L/min STP during powder injection. Rinsing after injection is made with 75 L/min Ar STP through the porous plug for 5-10 min, according to the grade. The injection process of the lime-based desulfurizing powder is sensitive to humidity as the grained lime is hygroscopic and can cause lance clogging (this phenomenon is more severe for very small-size powder and the particles should not be smaller than 100 mm). This might be a disadvantage compared to steel-slag stirring. Efficiency of steel desulfurization by powder injection. It is possible to produce ultra-low S-grades, to <0.003 wt% S, even for very high initial S-content, by powder injection in the ladle at atmospheric pressure. Taking the example of ArcelorMittal steel plant E, desulfurization by powder injection in the ladle for 10 min produces ultra-low S grades with <0.003 wt% S for an initial S-content > 0.007 wt% in the ladle at converter tapping. The average steel desulfurization rate is 70%. It is possible to remove up to 0.0120 wt% S (Fig. 12). The injection process is very effective for producing ultra-low S steels, compared to steelslag stirring. For example, both desulfurization processes were compared in ArcelorMittal steel plant E. For equivalent total amount of lime and calcium aluminate additions, and an average initial S-content of 0.008-0.009 wt%, the final S-content in the steel is 0.0007-0.0010 wt% lower with injection, with a comparable N pick-up and without impact on the temperature decrease during secondary metallurgy (Table 5). One drawback is the hygroscopic character of the lime powder, which may cause higher hydrogen pick-up than for steel-slag stirring and, thus, necessitate a post-treatment under vacuum.
SUMMaRy aND fUTURe wORk Ultra-low S-grades with <0.001-0.003 wt% S are required for demanding applications as advanced high strength steels for automotive (for reduction of vehicles weight and CO2 footprint) and HIC (Hydrogen Induced Cracking) grades for plates and long distance pipes. Producing steels with ultra-low sulfur content is a matter of route, and a lot of precautions have to be taken at the different steelmaking steps. In the BOF route, efficient hot-metal-desulfurization preferably by co-injection in the charging ladle, careful raking of hot metal desulfurization slag, use of low-S materials in the BOF (solid pig iron, scraps, lime) and low-S ferroalloys at tapping to avoid S pick-up is necessary. The control of oxidized converter slag carry-over to the ladle is essential, to reduce FeO, MnO and SiO2 contents in the ladle slag.
Lehmann & Nadif
508
S in steel before injection- S in steel after injection (wt-%)
0.016
Average desulfurization rate in secondary metallurgy: 70 %
0.012
0.008
0.004
S < 30 ppm - 96 heats 0.000 0.000
0.004
0.008
0.012
0.016
0.020
S in steel before injection (wt-%)
figure 12. Efficiency of S-removal from steel by lime powder injection: example of ArcelorMittal steel plant E. The dashed line represents the best fit obtained from the industrial results.
Table 5. Comparison of steel desulfurization by lime powder injection and by steel-slag stirring using the example of ArcelorMittal steel plant E. S aim (wt%)
< 0.003
< 0.005
S at tapping (wt%)
S after desulfurization (wt%)
S in tundish (wt%)
S removal (wt%)
S removal / S at tapping (%)
Lime injection
0.0097
0.0028
0.0028
0.0069
72
Steel-slag stirring
0.0094
0.0038
0.0035
0.0059
62
Lime injection
0.0086
0.0023
0.0030
0.0063
72
Steel-slag stirring
0.0082
0.0029
0.0037
0.0053
65
Desulfurization mode
In secondary metallurgy, steel-slag stirring is an efficient, cheap and widely used process for producing ultra-low-S steels. Steels with <0.0020 wt% S can be obtained by steel-slag stirring at atmospheric pressure. Desulfurization kinetics are better in the ladle furnace, which allows the slag to remain liquid, hence more reactive. The vacuum tank degasser is the tool which enables the production steel with <0.001 wt% S and simultaneously very low S and N concentrations (i.e., ≤0.003 wt% N), which is not possible for treatments at atmospheric pressure. Deep powder injection into the ladle combined with the use of pre-melted calcium aluminate or vacuum tank degasser appear as good methods to drastically reduce fluorspar consumption. With the progress of basic knowledge and thermodynamic modeling to describe the partitioning of S between steel and slag, steelmakers have powerful tools to define adequate steel and slag process to achieve good desulfurization from a thermodynamic point of view. To face the increase of ultra-low S product share, it is necessary to still enhance the kinetics of desulfurization by steel-slag stirring at atmospheric pressure. There are several ways to
Interactions Between Metal & Slag Melts: Steel Desulfurization
509
progress. One aspect is the development of coupled thermodynamic and CFD modeling taking into account the fluid flow in the ladle, considering that the liquid steel in the ladle is not any more a chemical homogeneous medium and taking local reactions into account (Andersson et al. 2000, 2002a,b; Hallberg et al. 2005; Al-Harbi et al. 2006, 2007; Lehmann et al. 2008). The implementation of measurements is also important to assess online the gas stirring quality, which dramatically affects the kinetics of desulfurization by steel-slag stirring. The Kettlor system that has been developed by ArcelorMittal R&D (Burty et al. 2006a,b) is presently used to monitor the efficiency of Ar stirring in a ladle by measuring the vibrations generated by the injected gas. With an adapted signal processing system, this method allows one to detect defects in the porous plugs or leaks in the gas supply. Finally, the development of an online desulfurization model would allow one to model predictively the evolution of the S-content in steel as a function of the real situation, such as initial S-content before desulfurization, steel composition in deoxidizers, real stirring intensity, etc. and to define countermeasures to adapt the process when the conditions are different from the nominal situation.
LIST Of SyMBOLS aO, aS
activity in respectively, oxygen and sulfur in liquid steel (reference state 1 wt% diluted solution in liquid Fe)
CS
sulfide capacity of the slag at the treatment temperature (reference states: pure gaseous species)
C′S
sulfide capacity of the slag at the treatment temperature (reference states: 1 wt% diluted solution in liquid iron)
[wt% S]metal
steel sulfur content
[wt% S]
steel sulfur content
(wt% S)slag
slag sulfur content
fS
activity coefficient of sulfur in liquid steel (aS = fS[wt% S])
KS
desulfurization kinetic coefficient (m·s−1)
A
surface of the slag/metal interface (m2)
T
absolute temperature (K)
V
metal volume (m3)
[wt% S(t)]metal
steel sulfur content at time t
(wt% S(t))slag
slag sulfur content at time t
R
desulfurization ratio (-)
l
thermodynamic desulfurization parameter: l = DS × M = sulfur metal-slag equilibrium partition coefficient × slag mass (t per ton of steel)
B
kinetic desulfurization parameter
VTD, LF
ladle metallurgy steelmaking reactors, i.e. vacuum tank degasser, ladle furnace
STP
Standard Temperature and Pressure conditions = 273 K and 105 Pa.
510
Lehmann & Nadif RefeReNCeS
Abraham KP, Richardson FD (1960) Sulfide capacities of silicate melts part II. J Iron Steel Inst 196:313-317 Al-Harbi M, Atkinson HV, Gao S (2006) Modeling methodology to simulation of molten steel refining in a gas-stirred ladle using coupled CFD and thermodynamic model. In: Modeling of Casting, Welding and Advanced Solidification Processes – XI. Gandin CA, Bellet M (ed) The Minerals, Metals & Materials Society, p 1089-1096 Al-Harbi M, Atkinson HV, Gao S (2007) Sulphide solubility limit in multicomponent slag system and its influence on the desulphurization rate in steelmaking. 7th International Conference Clean Steel, 4-6 June 2007, Balatonfüred, Hungary. p 238-245 Andersson MAT, Jonsson LTI, Jönsson PG (2000) A thermodynamic and kinetic model of reoxidation and desulphurization in the ladle furnace. ISIJ Int 40:1080-1088 Andersson MAT, Hallberg M, Jonsson LTI, Jönsson PG (2002a) Slag-metal reactions during ladle treatment with focus on desulphurisation. Ironmaking Steelmaking 29(3):224-232 Andersson MAT, Jonsson LTI, Jönsson PG (2002b) A study of varying FeO content and temperature on reactions between slag and steel during vacuum degassing. 6th International Conference Clean Steel: June, 10-12, 2002, Balatonfüred, Hungary. p 144-153 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Burty M, Pussé C, Bertoletti C, Wetta P, Carioli E (2006a) Kettlor: Efficient Stirring in Ladle Metallurgy. 5th European oxygen Steelmaking Conference, EOSC 2006 proceedings. p 265-271 Burty M, Pussé C, Wetta P, Sulin F, Bertoletti C, Bornèque Y, Pernet D, Carioli E (2006b) ArcelorMittal patent WO 2006/013239 A2: Method for controlling a metal molten bath bubbling in a metallurgical vessel and a device for carrying out said method. See: http://www.wipo.int/patentscope/search/en/detail.jsf?docId =WO2006013239&recNum=1&docAn=FR2005001460&queryString=WO/2006/013239&maxRec=1 Carter PT, Macfarlane TG (1957) Thermodynamics of slag systems. Part 1. The thermodynamic properties of CaO-Al2O3 slags. J Iron Steel Inst 185:54-62 Drakaliysky E, Sichen Du, Seetharaman S (1997) An experimental study of the sulfide capacities in the system Al2O3-CaO-SiO2. Can Metall Q 36:115-20 Ebel DS (2011) Sulfur in extraterrestrial bodies and the deep earth. Rev Mineral Geochem 73:315-336 Fincham CJB, Richardson FO (1954) The behaviour of sulphur in silicate and aluminate melts. Proc Roy Soc (London) A 223:40-62 Flood H, Förlan T, Grojtheim K (1953) In: The Physical Chemistry of Melts. a symposium on the nature of molten slags and salts held by the Nuffield Research Group in Extraction Metallurgy, London, 20th February, 1952. Institution of Mining and Metallurgy, London, p 46-62 Gantner A, Jungreithmeier A, Felberbauer W, Rössler R (2006) Steel desulfurization by fluidized lime operational and quality results. EOSC 2006:305-311 Gaye H, Lehmann J, Riboud PV, Welfringer J (1989) Thermodynamics of slags: use of a slag model to describe metallurgical reactions. Mem Et Scient Revue Métallurgie 4:237-244 Guggenheim EA (1952) Mixtures. The Theory of the Equilibrium Properties of Some Simple Classes of Mixtures, Solutions and Alloys (International Series of Monographs on Physics). Clarendon Press, Oxford Hallberg M, Jonsson LTI, Jönsson PG, Undvall P (2005) Sulphur and hydrogen refining during vacuum degassing – a new concept for process control. Stahl u Eisen 125(5):39-48 Kang YB, Pelton AD (2009) Thermodynamic model and database for sulfides dissolved in molten oxide slags. Metall Trans B 40B:979-999 Lachmund H, Xie YK, Buhles T, Pluschkell W (2003) Slag emulsification during liquid steel desulphurisation by gas injection into the ladle. Steel Res Int74 (2):77-85 Lehmann J, Gaye H (1992-1993) Modeling of thermodynamic properties of sulphur bearing metallurgical slags. Rev Int Hautes Tempér Réfract Fr 28:81-90 Lehmann J, Bonnet F, Bobadilla M (2006) Thermodynamic description of liquid steels and metallurgical slags by a generalization of the “Central Atoms” model. AIST Transactions 3 (4):115-123 Lehmann J, Bontems N, Simonnet M, Gardin P (2008) Recent advances in the domain of physical-chemistry modeling at ArcelorMittal Maizières Research S.A. Scanmet III:19-28 Béranger G, Henry G, Sanz G (eds) (1994) Le livre de l’acier, Technique et Documentation – Lavoisier ISBN 2-85206-981-4 Nadif M, Suero J, Rodhesly C, Salvadori D, Schadow F, Schutz R, Perrin E, Peeters L (2008) Desulfurization practices in ArcelorMittal Flat Carbon Western Europe. Scanmet III: 569-580 Nzotta MM, Andreasson M, Jönsson P, Seetharaman S (2000) A study on the sulphide capacities of steelmaking slags. Scand J Metall 29:177-184 Nzotta MM, Sichen Du, Seetharaman S (1998) Sulfide capacities in some multi component slag systems. ISIJ Int 38:1170-1179
Interactions Between Metal & Slag Melts: Steel Desulfurization
511
Nzotta MM, Sichen Du, Seetharaman S (1999) A study of the sulfide capacities of iron-oxide containing slags. Metall Mater Trans 30B:909-920 Pelton AD, Eriksson G, Romero-Serrano (1993) Calculation of sulfide capacities of multicomponent slags. Metall Trans B 24B:817-825 Posch V, Miceli P, Pluschkell W, Lachmund H (2002) Desulphurisation of liquid steel with refining top slags. EUR 20474 EN European Communities ISBN 92-894-4279-4 Reddy RG, Blander M (1987) Modeling of sulfide capacities of silicate melts. Metall Trans 18B:591-596 Riboud P, Vasse R (1985) Désulfuration de l’acier en poche: synthèse des résultats théoriques et industriels. CIT de la Revue de Métallurgie 82, 11:801-810 Richardson FO, Fincham CJB (1954) Sulphur in silicate and aluminate slags. J Iron Steel Inst 178:4-14 Rist A, Riboud P, Ancey-Moret MF (1974) Équilibres thermodynamiques en sidérurgie. Techniques de l’Ingénieur M1730-M1733 Sosinsky DJ, Sommerville ID (1986) The composition and temperature dependence of the sulfide capacity of metallurgical slags. Metall Trans B 17B:331-337 Zhang XF, Toguri JM (1987) The equilibrium distribution of sulphur between basic slags and steel. Can Metall Q 26:117-122
16
Reviews in Mineralogy & Geochemistry Vol. 73 pp. 513-578, 2011 Copyright © Mineralogical Society of America
The Role of Magmatic Sulfur in the Formation of Ore Deposits Adam C. Simon Department of Geoscience and High Pressure Science and Engineering Center (HiPSEC), University of Nevada Las Vegas Las Vegas, Nevada 89154, U.S.A. [email protected]
Edward M. Ripley Department of Geological Sciences Indiana University Bloomington, Indiana 47405, U.S.A. INTRODUCTION This chapter focuses on S in porphyry-type ore deposits, layered-mafic-intrusion-hosted ore bodies, and magmatic sulfide deposits. Porphyry-type ore deposits, e.g., Bingham Canyon, Utah, U.S.A. and Grasberg, Irian Jaya, are important hosts of Cu, Mo, Au, and Ag. Ore deposits hosted in layered mafic intrusions, e.g., the Bushveld and Stillwater complexes, contain significant quantities of Ni, Cu, Cr, Au and the platinum group elements (PGE: Pt, Pd, Rh, Re, Ir, Ru). Magmatic sulfide deposits, differentiated from layered mafic intrusions in that the former evince more clearly a role for immiscible sulfide accumulation without the possible presence of an aqueous fluid(s), e.g., Noril’sk and Voisey’s Bay, contain significant quantities of Cu, Ni, and the PGE. Allowing for some variability within any given ore deposit type, each of these types is unique in terms of the range of pressure and temperature of ore deposit formation, tectonic setting, and the compositional type(s) of parental causative magma. However, most porphyry-type ore deposits, layered-mafic-intrusion-hosted ore bodies, and magmatic sulfide deposits share the following features: 1) they are related chemically and physically to silicate magma; 2) they are the byproduct of differentiation of magma; 3) metals are hosted dominantly in sulfide minerals and the deposits can be thought of primarily as S anomalies (e.g., the Butte and Bingham Canyon porphyry ore deposits, located in Montana and Utah, U.S.A., respectively, contain 60 and 100 times, respectively, the ~20 Mt of S emitted during the 1991 eruption of Mt. Pinatubo); and 4) the metal(s) and S in each deposit type, albeit not necessarily the total quantity of S, are together derived from the same magmatic source. It is the connectivity of these commonalities that serves as the basis for this review chapter. The ubiquitous presence of S in magmatic systems is manifested in the commonly observed mass of S associated with volcanic eruptions and the presence of metal-sulfides, e.g., pyrite, in a range of magmatic ore deposits types. The latter suggests that S is fundamentally important in the continuum of events effecting ore metals from their silicate melt (si-mt) source to precipitation and deposition. However, the exact role that S plays during the formation of magmatic ore deposits, i.e., whether or not S is a control on metal deposition or transport or both, remains unclear. The importance of S in scavenging ore metals from si-mt is evinced clearly in magmatic sulfide deposits where petrographic and geochemical data evince a magmatic origin for the magma-hosted metal-sulfides. The importance of S in the formation of 1529-6466/11/0073-0016$10.00
DOI: 10.2138/rmg.2011.73.16
514
Simon and Ripley
layered-mafic-intrusion-hosted ore deposits is more difficult to constrain owing to the potential aqueous-fluid-mediated redistribution of metals in the magma; i.e., the uppers vs. the downers controversy. Sulfur certainly plays an important role in the precipitation and deposition of metals in porphyry-deposits; however, the metals in these deposits have been removed and transported from the parental si-mt and metal precipitation occurs at submagmatic temperatures. Thus, while there are a variety of structural, fluid inclusion, isotopic and experimental data that suggest a genetic connection between si-mt and metals and S in porphyry-deposits (e.g., Candela Table 1. List of abbreviations and symbols. and Piccoli 2005; Seedorff et al. 2005; WilliamsJones and Heinrich 2005), the exact role that S plays mt melt phase in scavenging and transporting ore metals from si-mt silicate melt phase the parental si-mt remains controversial, primarily su-mt sulfide melt phase owing to the competing effects of other metalsu-xtal sufide crystal complexing ligands (e.g., Cl). In this chapter, we aimed to present the state of knowledge on the role that S plays in the formation of S-rich magmatic and magmatic-hydrothermal ore deposits. We first review the basic chemistry of S, the nature of S in magma, the importance of redox conditions on S speciation, the speciation of metals in S-bearing si-mt and aqueous fluids, and the behavior of S in ore forming H-O-S-C-Cl-fluid(s) at magmatic and sub-magmatic conditions, followed by a discussion of individual ore deposit types, their tectonic settings, the nature of mineralization, and ore metal abundances. The data presented here serve to define our current understanding of the role that S plays in the formation of major types of S-rich ore bodies, which are chemically and physically related to magmatism. Abbreviations and terms are defined in Table 1.
iss mss po mag hem v b l fl f S2 fO2 usp Dxi/j
intermediate solid solution monosulfide solid solution pyrrhotite magnetite hematite vapor brine liquid aqueous fluid fugacity of sulfur (in bars) fugacity of oxygen (in bars) ulvöspinel distribution coefficient of a species or a component, x, between two phases i and j
GEOCHEMISTRY OF SULFUR IN MAGMATIC-HYDROTHERMAL SYSTEMS Sulfur basics In geologic environments, S exists in liquid, solid and volatile phases. Sulfur in the liquid state is present as sulfide melt (su-mt) in equilibrium with si-mt that ranges in composition from basalt to rhyolite. Sulfur in the solid state is found as native S and as the base for all sulfide (e.g., pyrite, FeS2) and sulfate (e.g., anhydrite, CaSO4) minerals. Elemental S is abundant near hot springs, in volcanic fumaroles, in salt domes, and in evaporites. Sulfur in the gaseous state exists dominantly as hydrogen sulfide (H2S) and sulfur dioxide (SO2). Common oxidation states in near-surface geologic reservoirs include S0, S2−, and S6+, where S is bonded with oxygen as sulfate (SO4)2−. Sulfur as sulfide and sulfate can coexist in si-mt and in magmatic-hydrothermal fluid(s), and the ratio of sulfide to sulfate is related directly to the oxidation state of the geologic reservoir. Sulfide and sulfate do not coexist in any known natural mineral. The oxidation state of S in magmatic systems is controlled by the oxidation state of silicate magma, hence controlled by the relative fugacities of H2 and O2 in the magmatic environment. Free H2 and O2 are present at vanishingly low concentrations in magmatic systems, but their abundance is the critical control on the stability of Fe-bearing minerals, via the control on the
Magmatic Sulfur and Ore Deposit Formation
515
Fe2+/Fe3+ ratio of the si-mt, and to the oxidation state of S. The relative proportion of H2 and O2 in magmatic systems is represented commonly by the use of the fugacities of H2 and O2, fH2 and fO2, respectively. The values of fO2 and fH2 change with temperature and pressure owing to the fact that the chemical potential of both O2 and H2 are temperature dependant and there is a volume term for each solid. The chemical potential of S in magmatic systems is represented by the use of the fugacity of S, fS2, and may be referenced to an equilibrium such as Fe + 0.5S2 = FeS
(1)
which describes the equilibrium between Fe and troilite, FeS. In geologic systems, the stability of sulfide minerals is intrinsically controlled by the value of fS2. The values for fO2, fH2 and fS2, can be determined by mineral phase assemblages and the proportion of Fe3+ to Fe2+ in rocks or si-mt. Once these values are determined for a particular geologic environment, one can determine values for the fugacities of hydrogen sulfide and sulfur dioxide, fH2S and fSO2, respectively (Clemente et al. 2004), which constrain the relative abundances of reduced to oxidized S in a magmatic system and have important implications for the mobility of ore metals in magmatic environments. The values of fS2, fH2S and fSO2 are related to one another via the equilibria 0.5S2 + H 2 = H 2S
(2)
0.5S2 + O2 = SO2
(3)
At fixed pressure and temperature, the fugacity of any of the S gas species, via the redox equilibria written above, is fixed provided that the values of two additional intensive parameters, e.g., fH2O and fH2, are constrained. The relative abundance of reduced to oxidized S, i.e., the ratio of fH2S to fSO2, in magmatic systems plays a determinant role in the stability of metal-sulfides in the magma and metal-S complexes in a magmatic volatile phase and, as discussed below, is a critical moderator of metal budgets of evolving magmatic-hydrothermal systems. The relative abundance in magmatic systems of sulfide to sulfate controls directly the stability of ore-metal-bearing sulfide minerals. The stability of sulfide minerals has been discussed in detail by Lusk and Bray (2002) and Fleet (2006). Below, and in Figure 1, the stability of sulfides in the Cu-Fe-S system is used to illustrate the relationship between sulfide mineral stability and the fugacities of S species. In the Cu-Fe-S system, the following equilibria, as presented in Lusk and Bray (2002) and Fleet (2006), describe the relationship between these S compounds: sulfur S0 liquid = sulfur S0 vapor pyrite FeS2 + nukundamite (Cu,Fe)4S4 = bornite Cu5FeS4 + sulfur S0 pyrite FeS2 + bornite Cu5FeS4 = chalcopyrite + sulfur S0 pyrite FeS2 + chalcopyrite CuFeS2 = isocubanite CuFe2S3 + sulfur S0 pyrrhotite + chalcopyrite CuFeS2 = isocubanite CuFe2S3 + sulfur S0 pyrite FeS2 + isocubanite CuFe2S3 = pyrrhotite + isocubanite + sulfur S0 pyrite FeS2 + chalcopyrite CuFeS2 = pyrrhotite Fe1–xS + chalcopyrite CuFeS2 + sulfur S0 pyrite FeS2 = pyrrhotite Fe1–xS + sulfur Each of these equilibria defines a unique value for fS2 as shown in Figure 1. The interdependence of these sulfide equilibria and fO2 is shown schematically in Figure 2. It is important to note that that value of fS2 can vary significantly, by several orders of magnitude at a unique fO2, P and T. For example, at 800 °C and fO2 fixed at NNO, the coexistence of po and magnetite fixes the value of fS2 at log fS2 = −3.2 (Fig. 2). The coexistence of po, iss (iss; approximately Cu2Fe3S5;
516
Simon and Ripley
Figure 1. The relationship between fS2 and temperature is shown for reactions in the Fe-S and Cu-Fe-S Figure 1. systems. Note that at a given temperature, the stability of sulfide minerals is strongly controlled by the fS2 which in turn affects the stability of individual sulfide phases; i.e., the partitioning of ore metals from a si-mt to a sulfide phase. [Used by permission of Elsevier, from Lusk and Bray (2002), Chemical Geology, Vol. 192, Fig. 7, p. 244].
iss is the Fe-enriched high temperature form of chalcopyrite) and magnetite fixes the value of fS2 at log fS2 = −0.5. Thus, there is a several order of magnitude variation in the fS2 in a system saturated with po and magnetite vs. a system saturated with po + iss + magnetite. As discussed below, the solubility of ore metals in sulfide minerals is dependent on the fS2, which in turn can be related to variations in the proportion of H2S to SO2, via the equilibria above, which has important consequences for the mobility of ore metals in magmatic systems.
Behavior of sulfur in silicate melts The absolute abundance of S in reduced si-mt parental to magmatic ore deposits is typically on the order of a few tens to a few hundred μg/g dissolved as sulfide, whereas oxidized and water-rich si-mt may contain up to 1.5 wt% S dissolved as sulfate (Jugo et al. 2005a,b; Jugo 2009). The S content of a si-mt, even at low total S concentrations, is critical to the ore-generative
Magmatic Sulfur and Ore Deposit Formation
517
Figure 2. The interdependence of fS2 and fO2, which is shown as described by the equations in the text. The area within the Figure 2. dashed lines represents the stability field of iss. Pyrrhotite transitions to mss with increasing dilution of Fe1−xS by substitution of Cu and Ni for Fe. Figure is modified after Jugo et al. (1999).
ability of many magmatic-hydrothermal ore deposit types owing to the observations from experimental and natural laboratories that S can form metal-sequestering sulfide(s) crystals (su-xtal) and/or immiscible su-mt and exsolve a S-bearing magmatic-hydrothermal fluid(s). Therefore, it is important to constrain the behavior of S in si-mt owing to the recognition that the loss of S to a su-xtal, immiscible su-mt or S-bearing fl has important ramifications for the mobility of ore metals. The behavior of S in si-mt has been the focus of many recent studies (Mavrogenes and O’Neill 1999; Holzheid and Grove 2002; O’Neill and Mavrogenes 2002; Jugo et al. 2005a,b; Li and Ripley 2005; Moretti and Ottonello 2005; Scaillet and Pichavant 2005; Liu et al. 2007; Jugo 2009) and a detailed review of the behavior of S in magmatic systems is presented elsewhere in this volume (see chapters in this volume by Baker and Moretti 2011; Wallace and Edmonds 2011; Webster and Botcharnikov 2011; and Parat et al. 2011). Sulfur speciation is critically important owing to the recognition that magmatic ore deposits depend on the presence of magmatic sulfide. Sulfide and/or sulfate phases that sequester ore metals from the si-mt at a deep level of the magmatic plumbing system may render the ascending silicate magma incapable of generating a magmatic-hydrothermal ore deposit, owing to the potentially strong depletion of the ore metal to the sulfide or sulfate phase. Conversely, as has been proposed for porphyry- and some PGE-layered mafic intrusion-ore deposits, the crystallization of a sulfide phase(s) during the early stage of solidification of a si-mt that ponds in Earth’s upper crust may act to preconcentrate ore metals that subsequently may be made available to a magmatic volatile phase if auto-oxidation, driven by magmatic degassing, leads to sulfide resorption owing to a change in the stable S species in the si-mt (Bell and Simon 2011). Thus, constraining the relationship between the ratio of sulfide to sulfate and the oxidation state of a magma is necessary to more accurately model the evolution of some magmatic ore forming systems. Jugo (2009) presented a model that relates the change in S species, i.e., sulfide vs. sulfate, to the oxidation state of a magma and described how this effects the sulfide capacity of a simt. The S capacity of a si-mt represents the sum of sulfide and sulfate species in the si-mt as
Simon and Ripley
518 described by the equilibrium:
[ST ] = {S2 −} + {S6 +}
( 4)
where [ST] is the total S content of the si-mt, and {S2−} and {S6+} are the sulfide and sulfate species, respectively, dissolved in the si-mt. Several studies have experimentally determined that the measured S capacity of the si-mt is lower when the si-mt is saturated with sulfide relative to sulfate (Carroll and Rutherford 1985, 1987; Luhr 1990; Jugo et al. 2005a,b) and data from natural systems are consistent with the abundance of S being highest in sulfate-saturated si-mt (Metrich and Clocchiatti 1996; Rowe et al. 2007). Jugo (2009) reported data that are consistent with a relationship between fO2 and S speciation in si-mt as described by the equation
(
∆FMQ = 1.29 + 0.45 ln[S6 + ] − ln[S2 − ]
)
(5)
where ΔFMQ is the fO2 relative to the fayalite-magnetite-quartz (FMQ) fO2 reference buffer, and [S2−] and [S6+] are the wt% sulfide and sulfate species dissolved in the si-mt at sulfide and sulfate saturation, respectively. Variations of temperature, pressure and water content of the si-mt exert a measurable effect on the maximum S content of si-mt as a function of fSO2 as displayed in Figure 3. Notably, Figure 3 evinces the order of magnitude increase in the S capacity of a si-mt at sulfide and sulfate saturation as fO2 increases from FMQ to FMQ+4, the range of oxidation states for most magmatic ore deposits, for both anhydrous and hydrous si-mt across a wide range of pressure and temperature; i.e., the S capacity of a si-mt at FMQ is ~0.05 wt% if S is dissolved as sulfide whereas at FMQ+3 the S capacity is ~1.5 wt% if S is dissolved as sulfate. Jugo (2009) pointed out that the S content of the si-mt increases exponentially as fO2 increases from FMQ to ~FMQ+2. This has important consequences not only for the role of S in the formation of magmatic ore deposits, but also for the mass transfer of ore metals from mantle sulfides during partial melting. The model presented in Jugo (2009) indicated that partial melting of the mantle source was not required to exceed twenty-five percent, as had previously been suggested by Hamlyn and Keays (1986) and Keays (1995). Jugo (2009) calculated that 6% partial melting of the mantle wedge, at a fO2 = FMQ+1.7, was sufficient to dissolve completely the mantle wedge sulfides, assuming a total mantle S content of 250 μg/g. These results are consistent with Mungall (2002) who modeled the complete dissolution of sulfides as occurring as at FMQ+2.
Figure 3. Experimentally determined maximum S content of si-mt that is saturated with sulfide or sulfate as a function of fO2 relative to the FMQ buffer. The vertical lines represent the transition from sulfide to sulfate saturation that occurs between ~FMQ+2 to FMQ+2.5. The Carroll and Rutherford (1985, 1987) data are for S capacity of hydrous trachyandesitic si-mt at 1025 °C and 200 MPa. The Jugo et al. (2005a) data are for S capacity of anhydrous basaltic si-mt at 1300 °C and 1 GPa. The curves are fits to equations described in Jugo (2009). [Used by permission of Geological Society of America, from Jugo (2009), Geology, Vol. 37(5), Fig. 1, p. 416].
Figure 3.
Magmatic Sulfur and Ore Deposit Formation
519
The available data for the speciation of S, and the sulfide and sulfate capacity of si-mt, have direct bearing on the ability of a given magmatic system to generate a magmatic-hydrothermal ore deposit. The oxidation state of arc magmas, the magmas parental to most porphyry-type ore deposits, is between ~FMQ and ~FMQ+3 and the oxidation state of the ore deposits themselves are typically between ~FMQ+1 and ~FMQ+4. Over this range of oxidation states, S exists as both sulfide and sulfate and the oxidation state of S changes from dominantly sulfide to dominantly sulfate as fO2 varies from FMQ to FMQ+4; e.g., at 800 °C, 100 MPa, fO2 fixed at FMQ+1, and fH2O fixed at 1000 bars, the ratio of fH2S to fSO2 is ~500, whereas at FMQ+2.5 the ratio of fH2S to fSO2 is ~1 (Burgisser and Scaillet 2007). The implications of the significant S speciation change in porphyry environments are discussed below. A more detailed discussion of the relationship between the S capacity of si-mt, si-mt composition and temperature and pressure is presented in Baker and Moretti (2011, this volume), Wallace and Edmonds (2011, this volume) and Webster and Botcharnikov (2011, this volume)
THE PARTITIONING OF METALS AND SULFUR AMONG MAGMATIC PHASES In this section, we review the available data for the partitioning of metals between su-xtal – si-mt (Dxsu-xtal/si-mt), su-mt – si-mt (Dxsu-mt/si-mt), and fl – si-mt (Dxfl/si-mt) assemblages; all partition coefficients are weight based. Fluid is defined as a mobile phase that is dominated by H-OS-Cl species, and in the presence of geologically relevant abundances of CO2 and Cl, a vapor and saline liquid may coexist with si-mt and su-mt. We acknowledge that there is a growing body of data that support a role in metal transport for CO2-bearing volatile phases (Lowenstern 2001; Tattich et al. 2010); however, there are a dearth of experimental studies on ore metal transport in CO2-systems and, thus, this review is necessarily biased toward H-O-S-Cl fluids. The implications for the genesis of magmatic ore deposits are discussed separately below. We stress that the data reported here have not been filtered by the authors of the current paper. Interested users of any of the data reported here are urged to thoroughly review the original citation and evaluate the experimental and analytical techniques, the approach to equilibrium, whether the experimental data reflect the attainment of steady-state conditions, and whether the data are statistically robust.
The partitioning of sulfur between silicate melt and H-O-S-Cl fluid(s) The S output measured at arc volcanoes (Wallace and Gerlach 1994) demonstrates the volatility of S in magmatic systems. The critical aspect of the partitioning of S in ore deposits is the mass transfer of metal-S species from si-mt to exsolved H-O-S-Cl fluid(s) and/or su-mt owing to the importance of these phases as agents of metal remobilization. Data demonstrate that S partitions strongly into H-O-S-Cl fluid(s) at all P, T, and X (Webster and Botcharnikov 2011, this volume). Experimentally determined DSfl/si-mt values are consistent with the ubiquitous presence of S in volcanic emissions (Westrich and Gerlach 1992; Symonds et al. 1994; Wallace and Gerlach 1994) and demonstrate that exsolved H-O-S-Cl fluid(s) have a significant S-scavenging potential (Keppler 2010). Further details on the magnitude of the partitioning of S among various geologic fluids are discussed in Webster and Botcharnikov (2011, this volume).
Controls on the partitioning of ore metals among silicate melt and crystalline sulfides A large number of sulfide minerals crystallize from a si-mt as the si-mt cools or changes composition during magma mixing. The phase equilibria of magmatic sulfides were reviewed comprehensively by Fleet (2006 and references therein). Here we focus on the sulfide minerals that are known to play a role in controlling the ore metal budget of magmatic systems. The available data for the partitioning of ore metals between si-mt and su-xtal are presented in Table 2.
Simon and Ripley
520
Table 2. Weight % partition coefficients among silicate liquids, sulfide liquid and sulfide crystals. The values in parentheses are the uncertainties reported in the original studies. Pressure and temperature for each experimental study are provided in the text. All data are from experiments except those of Peach et al. (1990), which are derived from measurement of natural sulfide melt and silicate melt in MORB. Fe
Au
Cu
Ni
Pt
Pd
Co
Partition coefficient between su-mt and basalt si-mt 1.2
50
150
3
7 3
1(0.9)×10
9(6)×10
4
1383
1.5-1.9×10
4
9(7)×10
3.5×104
500-900 2×103
2×103
2×104
2×104 3.4×104
4.6(0)×103
5.0(1.8)×103
10(4)×103
17(7)×103
480-696 755-1,303 >4,110->9180
Partition coefficient between mss and su-mt 1.14-1.28
0.006-0.013
0.15-0.22
0.66-0.94
0.12
0.14
0.035-0.052
0.072-0.12 0.053(0.005)0.13(0.001)
Partition coefficient between mss and granite si-mt 554(220) 910(436) 140(40)
2600(300) 174(103)
120±50
≥ 200 174±25 3
4
2.3×10 -1×10
5519(2673)
6×103-3×104
4.4×104-2.2×105
0.5(0.4)2300(4655)
400(525)27,600(33,115)
Partition coefficient between iss and granite si-mt 5700(2200) 1244(862)3(1.8)×105
26-51
Magmatic Sulfur and Ore Deposit Formation
Os
Ir
Ru
Rh
Re
Zn Reference 0.5
5
MacLean & Shimazaki (1976) Stone et al. (1990)
1(0.7)×10
4
3×103
Peach et al. (1990) natural MORE sulfide/glass Fleet et al. (1991) low ΣPGE
3×104
Fleet et al. (1991) high ΣPGE
1.2-1.6×10
3.5×104 3
521
Peach et al. (1990) 3
3
3.7(1.3)×10
3.2(1.1)×10
4.4(2.4)×10
30(6)×103
26(11)×103
6.4(2.1)×103
Fleet et al. (1996): 0-17 mol% NiS and ΣS 100 – 1000 μg/g Fleet et al. (1996): 37 mol% NiS and ΣS 100 – 1000 μg/g Ripley et al. (2002: high fS2 Ripley et al. (2002): low fS2
>30,000>97,000 >413>95,667
1.9(0.3)3.3(0.1)
>26,000>35,000
>8,100>35,000
33,00052,000 22(3)-22,377 (885)
Sattari et al. (2002) Brenan (2008)
10
9
3
Bockrath et al. (2004)
4.55-7.25
8.71-17.43
3.45-5.66
Mungall et al. (2005)
0.99(0.7)2.2(0.4)
1.6(0.3)2.6(0.2)
Brenan (2008)
Lynton et al. (1993): NNO+0.5 Lynton et al. (1993): C-CH4 Jugo et al. (1999) NNO Simon et al. (2006) Simon et al. (2008a,b) Simon et al. (2008a,b) Bell et al. (2009)
Jugo et al. (1999) Bell et al. (2009)
522
Simon and Ripley
The sulfides that most commonly crystallize from basaltic to rhyolitic si-mt include iss, bornite (bn; Cu5FeS4), po (po; Fe1−xS), mss (mss; ternary Fe-Ni-S solid solution where Fe1−xS appears at 1190 °C on the Fe-S join and extends to the Ni-S join at 999 °C; mss commonly contains Cu ranging from a few weight percent to several tens of weight percent (Mungall et al. 2005; Fleet 2006; Bell et al. 2009), and pentlandite ((Fe,Ni)9S8). The partitioning of Cu and Au between si-mt and po has been the focus of several studies (Table 2; Lynton et al. 1993; Jugo et al. 1999; Simon et al. 2006; Bell et al. 2009). The equilibrium mass transfer of Cu between si-mt and po can be represented by: (CuO0.5 )melt + ( FeS)po + 0.5S2 = (CuFeS2 )po + 0.25O2
(6)
where Cu is assumed to be present as a Cu-oxide component in the si-mt. If Cu is present in the si-mt as a Cu-sulfide component, a similar equilibrium can be written as: (CuS0.5 )melt + (FeS)po + 0.25S2 = (CuFeS2 )po
( 7)
These equilibria imply that DCupo/si-mt depends on fO2 and fS2. Similar equilibria can be written to describe DAupo/si-mt , and the reported DAupo/si-mt data are consistent with a direct proportionality between Au-mass transfer and fS2. The published DCupo/si-mt (Table 2), for experiments performed at fO2 ~ NNO, are plotted in Figure 4 as a function of fS2. The data demonstrate that variation in fS2 can affect, by several orders of magnitude, the ability of po to sequester Cu, and by analogy Au, from a si-mt. As discussed in Jugo et al. (1999), the effect of varying fO2 and fS2 on DCupo/si-mt and DAupo/si-mt can be expressed by the equation: po log Dxpo / si − mt = log K ′ + 0.5 log fS2 + log aFeS − 0.25 log fO2
(8)
where K′ =
Dxpo / si − mt × ( fO2 )0.25 po ( fS2 )0.5 × aFeS
( 9)
Figure 4.
and where×represents either Cu or Au. Equation (9) allows one to calculate a model value of
Figure 4. Experimental partitioning data for Cu between po and hydrous granite si-mt are plotted against Figure 4. fS2. The data from Bell et al. (2009), Simon et al. (2006), Jugo et al. (1999) and Lynton et al. (1993) are experimental studies conducted at 150 MPa and 800 °C, 100 MPa and 800 °C, 100 MPa and 850 °C, and 100 MPa and 800 °C, respectively. The Stimac and Hickmont (1994) data are from natural samples of cogenetic mss and silicic glass from volcanic rocks collected at Clear Lake, California.
Magmatic Sulfur and Ore Deposit Formation
523
Dxpo/si-mt at any fO2 and fS2. The value of log aFeSpo approaches zero as the XFeSpo approaches unity, and under these conditions the contribution of aFeSpo to the value of Dxpo/si-mt is eliminated (Jugo et al. 1999). The additive effects of fO2 and fS2 are shown graphically in Figure 5 where contours for model Dxpo/si-mt are plotted as a function of fSO2 and fS2. The model calculations evince the effect that varying fS2 has on the ability of sulfides to sequester ore metals from si-mt.
The partitioning of ore metals among silicate melt, sulfide liquid and sulfide crystals The available data for the partitioning of metals, on a mass basis, among si-mt, su-mt and su-xtal phases are presented in Table 2. It is important to note that the partitioning of elements between magmatic phases, e.g., su-mt and si-mt, is a function of pressure, temperature, bulk composition, fO2 and fS2. Therefore, we present information here for the intensive variables that accompany the experimentally determined partitioning data and caution users of these data to note that Dxi/j is not static in any system. MacLean and Shimazaki (1976) quantified the partitioning of Fe, Co, Ni, Cu and Zn between si-mt and Fe-rich si-mt by performing experiments in the FeS-FeO-SiO2 system at 1150 °C and 1 bar, and fO2 and fS2 internally buffered at low values of Fe and iron-alloy. They reported Dxsu-mt/si-mt as follows: Ni, 150; Cu, 50; Co, 7; Fe, 1.2; Zn, 0.5. The wide range in Dxsu-mt/si-mt indicates that su-mt is capable of efficient fractionation of these metals, e.g., Ni and Cu from Zn, and is related to the strong preference for Ni and Cu, and also the PGE discussed below, to bond covalently with S in the su-mt as opposed to the ionic bonding character of metals in the si-mt. Fleet and MacRae (1987, 1988) investigated the partitioning of Ni between olivine and (Fe,Ni)-su-mt, with the composition of the su-mt varying from 15 to 70 mol% NiS, as a function of temperature, fO2 and fS2 at 900 to 1400 °C at saturated vapor pressure. Their data indicate that the DNiol/su-mt varies from 30 to 35 as olivine varies from Fo97 to Fo0 and log fO2 varies from ~10−8.9 to ~10−8.0 and log fS2 from ~10−1 to ~10−3.5, respectively, as a function of the change in sulfide composition.
Stone et al. (1990) reported experimental data that constrained the partitioning of Pd, Ir, Pt and Au between (Fe-Ni)-monosulfide su-mt and basalt si-mt at 1200 °C and fO2 of 10−9.2 (~WM) and 10−0.9 at saturated vapor pressure. They reported Dxsu-mt/si-mt as follows: Pd, 9 ±7×104 Pt, 9 ±6×103; Ir 1 ±0.7×105; Au, 1 ±0.9×103. The data indicate that the mass transfer of the PGE and
Figure 5.
Figure 5.
Figure 5.
Figure 5. Thermodynamic model for the additive effects of fO2 and fS2 at 100 MPa and 850 °C on the partitioning of Cu between mss and hydrous granite si-mt are shown graphically by using dashed contours for DCumss/si-mt where logD′ = 0.5 log fO2 – 0.25 log fS2. The triangle represents the average value ±1σ for experiments performed at 100 MPa and 850 °C. [Used by permission of Elsevier, from Jugo et al. (1999), Lithos, Vol. 46, Fig. 5, p. 585].
524
Simon and Ripley
Au from basalt si-mt to immiscible Fe-Ni monosulfide su-mt may efficiently fractionate Pd and Ir from Pt and Au; Pd and Ir are approximately 50 and 500 times more soluble in (Fe-Ni)monosulfide su-mt relative to Pt and Au, respectively. Fleet et al. (1991) investigated experimentally the partitioning of Pd, Ir and Pt between (Fe,Ni)-monosulfide su-mt and basalt si-mt at 1300 °C at saturated vapor pressure. The fO2 was fixed at the following buffers: CCO, IQF, IW or WM. They reported that the Dxsu-mt/si-mt for Pd, Pt and Ir varied markedly as a function of total PGE, fO2, total S, total Fe and possibly total Ni, increasing from 2×103, 2×103 and 3×103, to 2×104, 2×104, and 3×104 for Pd, Pt and Ir, respectively, with increasing total PGE. The data indicate that Dxsu-mt/si-mt values are sensitive to variation in the composition of both the su-mt and the basaltic si-mt, and that Dxsu-mt/si-mt correlate with fO2 wherein decreasing fO2, total S and total Fe are associated with an increase in all measured Dxsu-mt/si-mt values. Lynton et al. (1993) reported experimental data that constrained the partitioning of Cu between mss and granite si-mt at 800 °C, 100 MPa, fO2 at NNO+0.5 and C-CH4, and log fS2 values ranging from −1 to −2. The solubility of Cu in mss and granite si-mt at NNO+0.5 is 2170 ± 520 μg/g and 43 ± 14 μg/g, respectively, yielding DCumss/si-mt of 550 ± 220. The solubility of Cu in mss and granite si-mt increases to 2890 ± 820 μg/g and 36 ± 14 μg/g at C-CH4 yielding a DCumss/si-mt of 910 ± 436. The authors pointed out that the two DCumss/si-mt values obtained at NNO + 0.5 and C-CH4 were statistically different at the 88 percent confidence level, and indicated that fS2 had a positive effect on the mass transfer of Cu from granite si-mt to mss. Peach et al. (1994) reported experimentally determined partition coefficients for Ir and Pd between su-mt and basaltic si-mt, DIrsu-mt/si-mt and DPdsu-mt/si-mt respectively, that are 3.5×104 and 3.4×104, respectively, at 800 MPa, 1450 °C. Their experiments were performed at log fO2 of −8.5 to −8.7 and log fS2 of −1.1 to −2.3. They reported that DIrsu-mt/si-mt and DPdsu-mt/si-mt do not depend on fO2/fS2. The data suggest that Ir and Pd should not fractionate from one another if su-mt controls the Ir and Pd budget of evolving magma. They pointed out that their experimental data overlapped DIrsu-mt/si-mt and DPdsu-mt/si-mt data reported by Peach et al. (1990) measured from coexisting, equilibrium MORB sulfide and glass (Table 2). We note that the Peach et al. (1990, 1994) studies quantified PGE in natural su-mt inclusions hosted in MORB glass and by performing experiments that were demonstrated to have achieved equilibrium and, thus likely provide the best constraints on the partitioning of the PGE between su-mt and basaltic si-mt. Fleet et al. (1996) experimentally constrained the partitioning of the PGE between (Fe,Ni)su-mt and vapor-saturated basaltic si-mt at 1200 to 1250 °C at saturated vapor pressure. The log fO2 and log fS2 were controlled at values that ranged from −7.2 to −11.3 and −1.0 to 3.4, respectively. The data indicate that DPGEsu-mt/si-mt values varied widely over the experimental conditions employed in the study, but the variance was correlated with the composition of the su-mt. Average Dxfl/si-mt for the assemblage wherein the su-mt is 37 mol% NiS and ΣS is 1001000 μg/g are: Os, 30 ± 6×103; Ir, 26 ± 11×103; Ru, 6.4 ± 2.1×103; Pt, 10 ± 4×103; Pd, 17 ± 7×103. The Dxsu-mt/si-mt values for the assemblage wherein the su-mt is 0 to 17 mol% NiS and ΣS is <100 μg/g are: Os, 3.7 ± 1.3×103; Ir, 3.2 ± 1.1×103; Ru, 4.4 ± 2.4×103; Pt, 4.6 ± 0×103; Pd, 5.0 ± 1.8×103. The data were consistent with Stone et al. (1990), Fleet et al. (1991) and Crocket et al. (1992) and were interpreted to indicate that the composition of the su-mt played a determining role in the mass transfer of the PGE, Cu and Au from si-mt to su-mt. Jugo et al. (1999) reported experimental data that constrained the partitioning of Au and Cu between mss and granite si-mt, and iss and granite si-mt at 850 °C, 100 MPa, fO2 fixed at ~NNO, and fS2 fixed at log fS2 = −1. They reported concentrations of Cu and Au that yielded DCumss/si-mt of 2600 ± 300 and DAumss/si-mt of 140 ± 40, respectively. They reported concentrations of Au in iss and si-mt that yielded an average DAuiss/si-mt of 5700 ± 2200. Their data indicate that crystallization of iss may deplete a si-mt in Au, and crystallization of mss may deplete a
Magmatic Sulfur and Ore Deposit Formation
525
si-mt in Cu. They noted however, that a si-mt saturated with iss at a fixed bulk composition, temperature, pressure, fO2, and total FeO, will have a fixed concentration of Cu, preventing strong depletion of the si-mt in Cu until iss is fractionated. An important conclusion from their study is that destabilization of both mss and iss relative to magnetite, owing to changes in fO2 and fS2, can result in the release of Cu and Au from the si-xtal to the si-mt and to a magmatichydrothermal fl. Ripley et al. (2002) investigated experimentally the partitioning of Cu and Fe between basalt si-mt and Cu-Fe-su-mt at 1245 °C, and saturated vapor pressure, with fO2 varied systematically from log fO2 of 10−8 to 10−11 bars, corresponding to ~WM to ~3 orders of magnitude below WM. The log fS2 was fixed at approximately −1.0 to −3 bars. The authors reported that the solubility of Cu in the su-mt was correlated with fS2 with the Cu capacity of the su-mt ranging from 594 to 1550 μg/g at log fS2 < −1.65, and from 80 to 768 μg/g at log fS2 > −1.65. The DCusu-mt/si-mt ranged from 480 to 1,303 μg/g at log fS2 < −1.65, and 0.7 to 13.6 at log fS2 > −1.65. The authors reported that the measured variation in DCusu-mt/si-mt for any fS2/fO2 ratio is consistent with non-Raoultian mixing associated with the mass transfer of Cu between basalt si-mt and su-mt. These findings indicate that the interpretation of DCusu-mt/si-mt is meaningless without a proper assessment of fS2/fO2. Sattari et al. (2002) investigated the partitioning of the PGE between (Fe,Ni)-su-mt, basalt si-mt and chromite at 1330 °C and 1 GPa in graphite-lined Pt capsules. The log fO2 and log fS2 ranged from −8.9 ± 0.2 to −9.5 ± 0.2 and −1.6 ± 0.2 to −2.1 ± 0.3, respectively. The measured abundances of Ru, Pd, Re, Os and Ir in run products yielded the following ranges of Dxsu-mt/ si-mt values: Ru, >8,100 to >35,000; Pd, >4,110 to >9180; Re, 33,000 to 52,000; Os, >30,000 to >97,000; Ir, >26,000 to >35,000. The data yielded the following Dxsu-mt/chromite values: Ru, >580 to >1,200; Pd, >130 to >3,700; Re, >85 to >610; Os, >560 to >2,000; Ir, >1,400 to >4,800. The authors suggested that differences between their determined partition coefficient data and the data reported in Stone et al. (1990), Fleet et al. (1991, 1996) and Crocket et al. (1992, 1997) might reflect different analytical techniques. Sattari et al. (2002) used laser-ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) to quantify metal concentrations in quenched sulfide and silicate run products whereas the previous studies used a bulk analytical technique to measure metal abundances in mechanically separated run products. The presence of metal spikes in the LA-ICP-MS transient signals of the run product analyses of Sattari et al. (2002) were interpreted by Sattari et al. (2002) to indicate that previous studies may have inadvertently sampled micro-sulfide inclusions in recovered silicate glass, therefore artificially increasing the apparent measured metal abundances in the silicate glass phase. Despite the variance in absolute metal concentrations among the various studies, the sum of the published data demonstrates unequivocally the significant metal-sequestration potential of su-mt. Bockrath et al. (2004) investigated the partitioning of Pt, Pd, Rh, Ru and Ir among mafic si-mt, (Fe,Ni,Cu)-mss and (Fe,Ni,Cu)-su-mt at 0.5 to 3.0 GPa at temperatures above the dry silicate solidus (the charge contained 60 wt% Fo90 olivine, 20 wt% En90 orthopyroxene, 15 wt% clinopyroxene, and 5 wt% Cr-Al-spinel) in graphite capsules, the latter ensuring low fO2 and, in combination with the sulfide phase assemblage, low fS2. The reported Dxmss/si-mt values are: Ir, 10.0; Ru, 9.0; Rh, 3.0; Pt, 0.12; Pd, 0.14. Their study is significant in that it details the partitioning of the PGE between co-existing sulfide phases and elucidates the fractionation of the PGE during sulfide partial melting. The data indicate that (Fe,Ni,Cu)-mss is enriched in Os, Ir and Ru, whereas the su-mt, in equilibrium with the mafic si-mt and su-xtal, is enriched in Rh, Pt and Pd. Mungall et al. (2005) reported experimentally determined Dxmss/si-mt for PGE, Cu, Ni and Au at 950, 1000 and 1050 °C at saturated vapor pressure. The fO2 and fS2 were fixed by using FMQ and Pt-PtS solid buffers, respectively, with one run having a fO2 of ~IW. The reported Dxmss/si-mt values when fO2 and fS2 were fixed at FMQ and Pt-PtS, respectively, are as follows:
526
Simon and Ripley
Fe, 1.14 to 1.28; Ni, 0.66 to 0.94; S, 1.12 to 1.19; Cu, 0.15 to 0.22; Ru, 8.71 to 17.43; Rh, 3.45 to 5.66; Pd, 0.072 to 0.12; Ir, 4.55 to 7.25; Pt, 0.035 to 0.052; Au, 0.006 to 0.013. The data indicate that DPtmss/si-mt, DPdmss/si-mt, and DAumss/si-mt all increase with decreasing temperature, but that Pt, Pd and Au are highly incompatible in mss in the presence of su-mt. The measured values for DIrmss/si-mt, DRumss/si-mt, and DRhmss/si-mt are all less than unity and decrease slightly with decreasing temperature. Decreasing fO2 from QFM to ~QFM-2 was not correlated with any measurable change in DIrmss/si-mt, DRumss/si-mt, and DRhmss/si-mt; however, decreasing both fO2 and fS2 did yield a measurable decrease in the Dxmss/si-mt for Ir, Ru and Rh. The data indicated that Dxmss/ si-mt values for PGE, Cu, Ni and Au are essentially independent of fO2 when fS2 is fixed at Pt-PtS. Iridium and Rh are highly compatible in mss, whereas Ni behaves incompatibly in mss at the fO2 and fS2 conditions of the study. Copper, Pt, Pd and Au are strongly incompatible in mss while Ru was found to be moderately to strongly compatible in mss. Nickel is correlated weakly to temperature with reported DNimss/si-mt values of 0.66 at 1050 °C and 0.95 at 950 °C. Brenan (2008) reported experimentally determined Dxsu-mt/si-mt and Dxmss/si-mt for Re and Os at 1200 C, 105 Pa wherein gravitational acceleration was used to separate equilibrated su-mt and si-mt, and at 1.5 GPa, and temperatures of 1200 and 1250 °C in a conventional piston cylinder apparatus. The reported Dxsu-mt/si-mt values vary from >20,000 to ~20, depending on the imposed fO2 and fS2 conditions and are constrained according to the imposed fS2 as follows: su − mt / si − mt log DRe = −6.95( ±0.20) + 2.39( ±0.04){0.5 log fS2 − 0.25 log fO2 }
(10)
at 1.5 log units above the FMQ-po buffer, and su − mt / si − mt log DRe = −7.79( ±0.44) + 2.90(±0.11){0.5 log fS2 − 0.25 log fO2 }
(11)
at 1 log unit below the FMQ-po fO2 - fS2 buffer. Brenan (2008) reported Dxmss/si-mt (±1Σ) that range from 1.6 ± 0.3 to 2.6 ± 0.2 for Re, from 1.9 ± 0.3 to 3.3 ± 0.1 for Os, from 0.99 ± 0.7 to 2.2 ± 0.4 for Ir, and from 0.053 ± 0.005 to 0.13 ± 0.001 for Pd. The study reported Dxsu-mt/ si-mt values that range from 22 ± 3 to 22,377 ± 885 for Re, and >413 to >95,667 for Os. Brenan (2008) concluded that DResu-mt/si-mt and DOssu-mt/si-mt, and by analogy other PGE, are sensitive to redox conditions, consistent with previous studies. The partitioning of Cu, Au and Ag between si-mt and magnetite-usp solid solution, and simt and po was the focus of a study by Simon et al. (2008a). They reported data from experiments performed in both hydrous and anhydrous assemblages at temperatures between 800 and 1050 °C, pressures from saturated vapor pressure to 140 MPa, log fO2 varied from NNO-0.25 to NNO and log fS2 varied from −1.5 to −3.0. Values for Dxmag/si-mt (± 1Σ) at 800 °C in a water-saturated assemblage are DCumag/si-mt = 2×10−4 ± 2×10−9, DAumag/si-mt = 0.82 ± 0.69, DCuusp/si-mt = 26 ± 17, DAuusp/si-mt = 50 ± 31, DCumss/si-mt = 174 ± 25. They reported that at 1050 °C, in an anhydrous assemblage, DCupo/si-mt ≥ 200, DAgpo/si-mt (± 1Σ) = 58 ± 8, and DAupo/si-mt (± 1Σ) = 120 ± 50. The values for DAuusp/si-mt and DCuusp/si-mt indicate that dilution of mag by the addition of Ti increases the Au- and Cu-sequestering capacity of mag. They reported Dxmss/si-mt values that indicated that a temperature change from 1050 to 800 °C results in no observable change in the partitioning of Cu. Bell et al. (2009) reported experimental data that elucidate the partitioning of Au, Pt and Pd between mss and granitic si-mt, and iss and granitic si-mt. They performed experiments in a si-mt – su-xtal – su-mt – oxide – supercritical aqueous fluid phase – Pt – Pd – Au assemblage at 800 °C, 150 MPa, fO2 ~NNO, and log fS2 varied from 0 to −5. They reported Dxmss/si-mt (± 1Σ) for Au, Pd and Pt that range from 1244 ± 862 to 3 ± 1.8×105, 0.5 ± 0.4 to 2300 ± 4655, and 400 ± 525 to 27,600 ± 33,115, respectively. These data indicate that mss should strongly fractionate Pd from Pt, and both of these metals from Au and that the mass transfer of Pt and Pd from the si-mt to mss is dependent upon the fS2 of the system; i.e. higher fS2 yields higher metal
Magmatic Sulfur and Ore Deposit Formation
527
solubilities in mss. These data indicate that there is no systematic variation of the Pt, Pd or Au solubility in iss with variation in either fS2 or the Fe/Cu ratio of the recovered iss; however, the data demonstrate that crystallization of iss can fractionate Au and Pd from Pt with Pt being retained more strongly in the granite si-mt.
The partitioning of ore metals among silicate melt and S-bearing aqueous fluid(s) It is well recognized that magmatic-hydrothermal H-O-S-Cl fluids are instrumental in the formation of several types of magmatic-hydrothermal porphyry-type ore deposits. There is growing evidence, e.g., from chemical analyses of natural fluid inclusions and Cl/F ratios of apatite, that aqueous fluids were present during the formation of layered-mafic-instrusion-hosted PGE deposits, and aqueous fluids are postulated to play a role in the subsolidus redistribution of metals in some magmatic sulfide deposits (Boudreau et al. 1986; Boudreau and McCallum 1989; Willmore et al. 2000, 2002; Boudreau and Hoatson 2004; Hanley et al. 2008; Boudreau 2009). The most direct evidence of magmatic H-O-S-Cl-volatile phase transport of metals is found in volcanic fumaroles and their sublimates. Volcanic vapors are comprised of H2O, CO2, SO2, and the proportion of these gases varies among volcanic settings. In arc environments that host porphyry-type ore deposits, volcanic gases are reported to contain on the order of 90 mol% H2O, with lesser amounts of CO2 (up to 10 mol%), SO2 (up to 6 mol%), and HCl (up to 6 mol%) (Williams-Jones and Heinrich 2005; Webster and Mandeville 2007; Métrich and Wallace 2008; Oppenheimer et al. 2011, this volume). Other important volatile components, present typically at concentrations below 1 mol%, include H2, HF, H2S, Ar, Rn, N, Hg Br and CH4 (Symonds et al. 1987; Giggenbach et al. 1993; Giggenbach 1997; Williams-Jones and Rymer 2000; Taran and Giggenbach 2003; Elkins et al. 2006; Oppenheimer et al. 2006; Webster and Mandeville 2007). The ability of magmatic-hydrothermal fluid(s) to dissolve and transport significant quantities of S is well documented (Gerlach 1980; Wallace and Gerlach 1994; Giggenbach 1977, 1997; Wallace and Anderson 1999; Stix and Gaonac’h 2000; Hattori and Keith 2001; Wallace 2001, 2005). Metal abundances in volcanic vapors vary significantly (Table 3), but concentrations can be high enough that fumarole encrustations include sphalerite (ZnS), chalcocite (Cu2S), covellite (CuS), galena (PbS), chalcopyrite (CuFeS2), molybdenite (MoS2), cassiterite (SnO2), wolframite ((FeMn)WO4), scheelite (CaWO4), greenockite (CdS), realgar (As4S4), bismuthinite (Bi2S3) and native Au (Naboko 1964; Stoiber and Rose 1974; Symonds et al. 1987; Quisefit et al. 1989; Taran et al. 2000; Williams-Jones and Heinrich 2005). Korzhinsky et al. (1994) reported the precipitation of rheniite (ReS2) in sublimates from volcanic fumaroles at temperatures of 910 °C, at the Kudriavy volcano in the Kurile arc. Williams-Jones and Heinrich (2005) reported that the variability in the metal composition of volcanic vapors is a function of the composition of the parental si-mt. Volcanic vapors evolved from basaltic si-mt can contain up to several μg/g Cu as opposed to vapors evolved from dacite and rhyolite si-mt that contain ng/g abundances of Cu. Andesite si-mt can evolve magmatic vapors that contain μg/g abundances of Cu, Zn and Pb, but notably these andesite-evolved vapors contain higher concentrations of Mo (3 μg/g from Merapi, Indonesia: Symonds et al. 1987) and Hg (340 μg/g from Kudryavy, Russia: Taran et al. 1995). High concentrations of Sn (7 μg/g at Usu, Japan: Giggenbach and Matsuo 1991) and Mo (1 μg/g at Satsuma Iwojima, Japan: Hedenquist et al. 1994) have also been measured in volcanic vapors. The ability of volcanic vapor to transport S is perhaps best exemplified at Kawah Ijen, Indonesia, where S is mined from an active vent by channeling volcanic vapor through a series of pipes resulting in condensation of molten S that solidifies as it exits the cold end of the pipe (McSween 2010). Practically, the thermal gradient along the length of silica gas-sampling tubes used to sample volcanic vapors allows for the precise determination of the temperature of mineral precipitation, facilitating quantitative constraints on the relationship between temperature and mineral precipitation (Williams-Jones and Heinrich 2005). These data indicate that magnetite,
300-315
0.004-0.005
0.03-0.09
0.002-0.004
1-2
3-11
0.2-0.8
NA
0.3-0.6
0.8-1.0
ND
0.18-0.27
Temp (°C)
Hg
As
Sb
Au
Ag
Cu
Pb
Zn
Sn
W
Mo
0.15-0.5
ND
1.1-2.4
0.4-7.6
1.9-7
0.2-8.4
5-14
1-24
0.003-0.96
0.23-0.49
0.006-0.021
459-770
Motomobo
1
ND
ND
1.4-3.9
0.5-8.6
0.8-5.4
0.1-0.6
6-250
ND
0.003-0.008
ND
0.01-0.03
344-852
Poas
1
2
ND
ND
0.09
9.8
0.6
6
6
4
13
24
ND
1010
Tobalchik
ND
ND
ND
13
12
ND
120
24
0.02
6.1
ND
928
Etna
3
0.09-0.18
NA
NA
5-8
0.08-0.48
0.4-0.9
NA
1.0-5.0
0.04-0.13
0.4-0.5
NA
738-828
Colima
4
0.6-1.6
0.02-0.34
535-940
Kudriavy
0.002-0.271
0.002-0.060
0.1-0.38
0.25-13.5
0.110-9.7
0.03-0.91
NA
NA
0.02-0.510
5
0.08-2.8
0.05-0.13
ND
4-82
0.3-1.6
0.01-1.0
NA
NA
0.16
0.28-0.96
ND
576-796
Merapi
6
0.094
0.003
ND
0.019
0.5
0.05
0.01
0.03
0.008
1.4
0.02
710
Mt. St. Helens
7
0.06
0.003
7
0.04
0.006
0.01
0.5
0.007
0
0.02
0.0001
649
Usu, Japan
8
0.005-1.07
ND
0.2-0.8
0.03-0.24
0.2-1.8
0.006-0.064
ND
0.0015-0.0032
0.03-0.03
1.7-4.6
ND
165-877
Satsuma Iwo Jima
9
NA = not analyzed; ND = not detected. Concentrations reported as μg/g for all elements except for ng/g for Au. All data originally reported as follows: 1) Gemmell (1987); 2) Menyailov and Nikitina (1980); 3) Le Guern (1988); 4) Taran et al. (2000); 5) Taran et al. (1995); 6) Symonds et al. (1987); 7) Bernard (1985); 8) Giggenbach and Matsuo (1991); 9) Hedenquist et al. (1994)
Cerro Negro
Volcano
1
Table 3. Concentrations of metals in subduction-related volcanic vapors as compiled by Williams-Jones and Heinrich (2005).
528 Simon and Ripley
Magmatic Sulfur and Ore Deposit Formation
529
molybdenite and wolframite precipitate at temperatures >500 °C, chalcopyrite, sphalerite, and pyrite precipitate between 500 and 450 °C, and galena, Pb sulfosalts and native As precipitate at <450 °C (Williams-Jones and Heinrich 2005). The observed relationship between temperature and metal condensation is consistent with the expected thermal effect on mineral saturation in the vapor phase; however, Williams-Jones and Heinrich (2005) interpreted data (Korzhinsky et al. 1994; Chaplygin et al. 2005) from Kudriavy volcano in the Kurile arc to indicate that the pressure-temperature ascent path followed by magmatic vapor is instrumental at controlling the metal budget of the ascending vapor phase. Williams-Jones and Heinrich (2005) reported that several fumarole fields located within one-hundred meters of one another precipitated different mineral assemblages: ilsemanite-molybdite-molybdenite-magnetite at 700-870 °C, cosalitelillianite-wurzite-cadmoindite at 400-730 °C, ilsemanite-molybdite-molybdenite-hematitemagnetite-W-powellite at 400-650 °C, and greenockite-rheniite-magnetite at 300-560 °C. The volcanic vapors emitted at all fumaroles in the spatially restricted field at Kudriavy volcano were hypothesized to have evolved from the same magma source, implying that processes affecting the vapors during transit from their source to the atmosphere played a controlling role in determining the characteristic metal assemblage of magmatic vapor. This observation has significant relevance for the evolution of metal ore deposits associated with magmatic degassing. A large number of experimental studies have been performed to elucidate the compositional nature of the volatile phase(s) exsolved from si-mt. This work has been driven primarily by the recognition that volatile phases trapped as fluid inclusions in natural magmatic systems are dominated by H2O, with variable but generally low concentrations of CO2 and S, and chloride salts that range in abundance from <1 wt% NaCl eq. (low-salinity vapor) to pure salt melts (Roedder 1984). Experimental studies indicate that the addition of other halides, e.g., KCl, affects inconsequentially the phase relations as determined in the NaCl-H2O system (Sourirajan and Kennedy 1962; Henley and McNabb 1978; Bodnar et al. 1985; Chou 1982, 1987; Fournier 1987; Vanko et al. 1988; Shinohara et al. 1989; Chou et al. 1992; Anderko and Pitzer 1993a,b; Bodnar 2003). The properties of H-O-S-Cl fluid(s) have been described comprehensively in a number of recent papers (Webster and DeVivo 2002; Williams-Jones and Heinrich 2005; Hack et al. 2007; Heinrich 2007; Webster and Mandeville 2007) and Webster and Botcharnikov (2011, this volume) discuss the partitioning of S between chemically complex aqueous fluid and si-mt. The properties of the NaCl-H2O system influence the role that S plays in the complexation and transport of ore metals during the evolution of magmatic-hydrothermal ore deposits. At high pressure, a si-mt that reaches fluid(s) saturation, caused by decompression and/or crystallization (Burnham 1967, 1979), will exsolve a single-phase aqueous fluid or co-existing vapor plus brine. Natural, chemically complex aqueous fluid(s) will scavenge S, Cl and other halides from the si-mt and is characterized by having a possible composition between pure H2O and a salt melt (Webster and Botcharnikov 2011, this volume). At lower pressure, below the consolute point, the H2O-chloride salt system is characterized by a wide miscibility gap wherein lowsalinity aqueous fluid (i.e., vapor) coexists with high-salinity aqueous fluid (i.e., brine) over the range of P and T associated with magmatic intrusions in Earth’s shallow crust (Sourirajan and Kennedy 1962; Henley and McNabb 1978; Bodnar et al. 1985; Chou 1987; Fournier 1987). For example, at 800 °C and 100 MPa the wt% concentrations of NaCl of vapor and brine are ~2 and ~60, respectively, while at 800 °C and 145 MPa they are ~20 (vapor) and ~35 (brine). Fluid inclusions evince that magmatic-hydrothermal aqueous fluid(s) can complex and transport ore metals both within a magma (e.g., Bushveld: Schiffries 1982; Ballhaus and Stumpfl 1986; Stillwater: Hanley et al. 2008), from the magma to surrounding country rock (Emmons 1927; Eastoe 1978; Ahmad and Rose 1980; Roedder 1984) and from the magma to the atmosphere (Lowenstern et al. 1991). The ability of magmatic-hydrothermal fluid(s) to scavenge ore metals from si-mt, as well as from crystalline and su-mt is a function of the density of the fluid(s)
530
Simon and Ripley
(Williams-Jones et al. 2002; Pokrovski et al. 2005) and the concentration of metal-complexing ligands in the fluid phase(s) (Candela and Piccoli 1995; Candela and Piccoli 2005; WilliamsJones and Heinrich 2005). The ability of a magmatic volatile phase to complex and transport ore metals from a si-mt was demonstrated to be related to the concentration of Cl in the volatile phase. Consequently, many studies were performed to constrain the solubility of ore metals in aqueous fluid(s) as a function of the chlorinity of the fluid (Candela and Holland 1984; Williams et al. 1995; Frank et al. 2002; Simon et al. 2004, 2005, 2006, 2007, 2008a,b, 2009; Hack and Mavrogenes 2006; Ulrich and Mavrogenes 2008). Notably, all of these experimental studies were performed in S-free systems owing to the hypothesis that low-salinity fluid could not transport sufficient metal(s) to form a magmatic-hydrothermal ore deposit. However, recent LA-ICP-MS analyses of coexisting vapor and brine fluid inclusions from mineralized samples collected from magmatic-hydrothermal ore deposits indicate that S, in addition to Cl, can complex with and transport ore metals.
The transport of ore metals in magmatic-hydrothermal fluid(s) Many experimental studies have been performed to investigate the mobility of ore metals in magmatic-hydrothermal fluid(s) over the pressure and temperature range that extends from high-temperature, where a si-mt is present, to low-temperature where only crystals are present. The experiments in the presence of a si-mt have been performed to elucidate the ability of S-free and S-bearing fl, defined as low-density vapor, brine or single phase aqueous fluid, to scavenge ore metals from si-mt, which is hypothesized to be the metal and S source in porphyrytype ore deposits, and also to investigate the potential for magmatic-hydrothermal fluid(s) to mobilize ore metals in layered-mafic-intrusion-hosted ore deposits. Published experimental data for the partitioning of Au, Cu, Ag, Pt, Mo and As between si-mt and fl are presented in Table 4. Experimental studies performed at temperatures <~700 °C, corresponding to the watersaturated granite solidus, have been performed to quantify the ability of S to complex with and transport ore metals in magmatic-hydrothermal fl, which evolves from a causative magma into surrounding country rock where metal precipitation occurs because of changes in PTX. These low-temperature studies were performed in S-free (Henley 1973; Crerar and Barnes 1976; Wood et al. 1987; Zotov et al. 1991; Var’yash 1992; Gammons and Williams-Jones 1995; Stefánsson and Seward 2003a,b) and S-bearing systems (Seward 1973; Hayashi and Ohmoto 1991; Pan and Wood 1994; Benning and Seward 1996; Zotov et al. 1996; Baranova and Zotov 1998; Gibert et al. 1998; Loucks and Mavrogenes 1999; Stefánsson and Seward 2004). The data from many of the experimental studies evince that ore metals are complexed with S, Cl and OH. However, the data indicate that metal-S complexes are more stable at temperatures <600 °C and metal-Cl and/or metal-OH complexes are more stable at >600 °C. The transport of copper and gold in magmatic-hydrothermal fluids. Simon et al. (2006) quantified the effect of S on the partitioning of Cu among si-mt -vapor-brine ± magnetite ± mss at 800 °C, 140 MPa, and fO2 ~NNO. In the S-bearing experiments, log fS2 = −2.99; log fH2S = 1.15 and log fSO2 = −1.51. They reported that (± 2Σ) DCuv/si-mt = 63 ± 31, DCub/si-mt = 240 ± 80, and DCuv/b = 0.27 ± 0.10 in the S-free system. They reported that (± 2Σ) DCuv/si-mt = 316 ± 22, DCub/si-mt = 443 ± 68, and DCuv/b = 0.69 ± 0.16 in the S-bearing system. Their data indicate that the presence of S increases the mass transfer of Cu from si-mt into vapor and brine. The five-fold increase of DCuv/si-mt, from 63 to 316, and nearly three-fold increase of DCuv/b, from 0.27 to 0.69, suggest that Cu may be transported as both a Cl- and S-complex in magmatic-hydrothermal vapor and brine. There are no experimental data to our knowledge that constrain the speciation of Cu, or other metals, in S-bearing (supersolidus) magmatic-hydrothermal fl. However, the ratio of fH2S to fSO2 in their experiments was 530, consistent with the hypothesis that the increase in the abundance of Cu in the lower-Cl vapor reflects the presence of a reduced S–Cu complex. Frank et al. (2011) quantified experimentally the effect of S on the partitioning of Cu and Au among si-mt, su-xtal, vapor, and brine by performing experiments at 800 °C, 100 MPa,
Magmatic Sulfur and Ore Deposit Formation
531
and fO2 fixed at ~NNO, in two separate sulfide-oxide assemblages used to fix the fS2 at two unique values spanning the range of fS2 in natural ore-forming magmatic-hydrothermal systems. Experiments were performed with a vapor-brine- si-mt -iss-mss-magnetite-quartz assemblage and a vapor-brine- si-mt- iss-mss-bornite-magnetite-quartz assemblage where the assemblages fixed log fS2 at −1 and −4, respectively. The reported DCub/si-mt range from 32 to 617 with the highest partition coefficients associated with low fS2. The reported DCuv/si-mt range from 2.6 to 18 with DCuv/si-mt increasing with increasing fS2. The reported DAub/si-mt range from 130 to 240 with the highest DAub/si-mt associated with low fS2. The reported DAuv/si-mt range from 70 to 140 with the highest DAuv/si-mt corresponding to higher fS2. These Div/si-mt and Dib/si-mt indicate that both Cu and Au partition strongly into both vapor and brine, and that on an equivalent mass basis the brine contains a greater absolute quantity of Cu and Au. The DCub/v (±1Σ) range from 7 ± 1 at high fS2 to 160 ± 20 at low fS2, and DAub/v (±1Σ), range 1.7 ± 0.2 at high fS2 to 26 ± 8 at low fS2. These data indicate that the partitioning of both Cu and Au into the vapor increases with increasing fS2 and are consistent with the results of Simon et al. (2006) in demonstrating that Cu and Au at magmatic conditions can be complexed with and transported as a reduced S complex. The mobility of Cu in magmatic-hydrothermally sourced fluid(s) operating at sub-solidus P-T conditions has been studied experimentally by Crerar and Barnes (1976) and Xiao et al. (1998) and Archibald et al. (2002). Crerar and Barnes (1976) quantified the solubilities of chalcopyrite + pyrite + bornite, and Cu + chalcocite in subcritical NaCl and NaHS + H2S aqueous solutions at temperatures of 200 to 350 °C, pH fixed to within 2.5 of neutral in sulfide solutions, and 1.5 of neutral in chloride solutions at 1.52 to 25.9 MPa. Their experiments were designed to elucidate the relative metal-complexing abilities of Cl and S. They reported that Cu is complexed dominantly as Cu(I) chloride at total S from 0.1 to 1 M, and transitions to Cu(I) bisulfide at higher total S. The stability of Cu-bisulfide decreased with increasing temperature owing to the strong decrease in HS−/H2S with increasing temperature and the much-reduced ability of H2S to complex and transport metals; e.g., HS−/H2S decreased from 1.0 at 25 °C to 10−3 at ~375 °C. Their data indicate that >1,000 μg/g each of Fe and Cu can be transported in S-bearing mildly acidic NaCl solutions at >350 °C. Their experimental data evince a strong temperature dependence for both Cu and Fe, indicating that a 100 °C temperature decrease results in precipitation of ~99% of Cu and 90% Fe. Lerchbaumer and Audétat (2009) quantified experimentally the partitioning of Cu between coexisting S-bearing low-salinity aqueous vapor and high-salinity aqueous brine at 600 and 700 °C and 50 and 100 MPa in a bornite + iss + vapor + brine assemblage at a fixed fS2 and fO2. They reported an average DCub/v value of 7.5 ± 2.5 when fS2 is fixed by bn + iss at controlled fO2. Lerchbaumer and Audétat (2009) also performed experiments over the same pressuretemperature range, but at a ΣS content of the two phase aqueous system of ~2.1×104 μg/g (2.1 wt%) and reported an average DCub/v value of 4.1 ± 1.5. The transport of gold and arsenic in magmatic-hydrothermal fluids. Simon et al. (2007) quantified the partitioning of As and Au between rhyolite si-mt and 2 wt% NaCl eq. vapor in a si-mt – vapor – Au metal ± mag ± po assemblage at 800 °C, 120 MPa and fO2 ~NNO. The S species in the S-bearing runs were fixed at log fH2S = 1.1, log fSO2 = −1.5, and log fS2 = −3.0 and the H2S to SO2 ratio was ~400. Reported DAsv/si-mt (±1Σ) in the S-free and S-bearing assemblages are 1.0 ± 0.1 and 2.5 ± 0.3, respectively. The S-free data are consistent with the presence of arsenous acid, As(OH)3 in the vapor phase as described by the equilibrium 0.5As2O3melt + 1.5H 2O = As(OH)3v
(12)
with the corresponding apparent equilibrium constant: ′ = K As
v (CAs(OH) ) 3 si − mt 1 / 2 (CAs ) × ( X Hv2O )3 / 2 2 O3
(13)
Cu
Ag
4
1.0±0.1
18
2.6
140
70
12±0.3
316±22 2.5±0.3
Simon and Pettke (2009); (140 MPa; 800 °C; NNO; H, O, Cl)
1.0±0.2
Frank et al. (2011); (100 MPa; 800 °C; NNO; H, O, S, Cl low fS2)
Frank et al. (2011); (100 MPa; 800 °C; NNO; H, O, S, Cl high fS2)
Simon et al. (2007); (120 MPa; 800 °C; NNO; H, O, S, Cl)
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, S, Cl)
Simon and Pettke (2009); (100 MPa; 800 °C; NNO; H, O, Cl)
Simon et al. (2008b); (140 MPa; 800 °C; NNO; H, O, Cl)
32±10 2.9±1.0
Simon et al. (2008b); (100 MPa; 800 °C; NNO; H, O, Cl)
Simon et al. (2007); (120 MPa; 800 °C; NNO; H, O, S, Cl)
32±30
Partitioning between S-bearing aqueous vapor and si-mt
15±2.5
Simon et al. (2005); (145 MPa; 800 °C; NNO; H, O, Cl)
72±31
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, S, Cl)
Simon et al. (2005; H, O, Cl); (140 MPa; 800 °C; NNO; H, O, Cl)
Simon et al. (2005); (130 MPa; 800 °C; NNO; H, O, Cl)
56±32
Simon et al. (2005); (110 MPa; 800 °C; NNO; H, O, Cl)
4
Hanley et al. (2005); (150 MPa; 800 °C; NNO; 70 wt% NaCl eq.; H, O, Cl)
2
1.0×10 to 2.6×10
Hanley et al. (2005); (150 MPa; 800 °C; NNO; H, O, Cl)
Candela and Holland (1984); (140 MPa;750 °C; NNO; H, O, Cl)
Reference
8±5
63±31
As
Hanley et al. (2005); (150 MPa; 800 °C; NNO; 20 wt% NaCl eq., H, O, Cl)
2.5 ± 1.6
Mo
6.6×103 to 5.2×104
Pt
10±4
Partitioning between S-free aqueous vapor and si-mt
4.2×10 - 1.2×10
3
DCuf/m = 9.1(±2.5)mClv
Partitioning between S-free supercritical aqueous fluid and si-mt
Au
Table 4. Weight % partition coefficients among silicate liquids and aqueous-fluid phase(s). Reported uncertainties are those from the original studies. The presence of H, O, S, C, Cl in the fluid phase(s), P, T and imposed fO2 are provided in parentheses with each reference.
532 Simon and Ripley
32
130
0.13±0.05
0.69±0.16
Simon et al. (2008b); (100 MPa; 800 °C; NNO; H, O, S, Cl) Simon et al. (2008b); (140 MPa; 800 °C; NNO; H, O, S, Cl)
0.06±0.03
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, S, Cl)
Simon and Pettke (2009); (140 MPa; 800 °C; NNO; H, O, S, Cl)
0.026±0.004
Partition coefficients between vapor and brine in S-bearing system
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, Cl)
0.27±0.20
Simon and Pettke (2009); (100 MPa; 800 °C; NNO; H, O, S, Cl)
Williams et al. (1995); (50 MPa; 800 °C; NNO; H, O, Cl) 0.15±0.05
Williams et al. (1995); (100 MPa; 800 °C; NNO; H, O, Cl)
200
Frank et al. (2011); (100 MPa; 800 °C; NNO; H, O, S, Cl; low fS2)
Frank et al. (2011); (100 MPa; 800 °C; NNO; H, O, S, Cl; high fS2)
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, S, Cl)
Simon and Pettke (2009); (140 MPa; 800 °C; NNO; H, O, Cl)
120
Partition coefficients between vapor and brine in S-free system
617
240
443±68
Partitioning between S-bearing aqueous brine and si-mt
Simon and Pettke (2009); (100 MPa; 800 °C; NNO; H, O, Cl)
67±27
Simon et al. (2008b); (140 MPa; 800 °C; NNO; H, O, Cl)
413±172 6.8±2.4
Simon et al. (2008b); (100 MPa; 800 °C; NNO; H, O, Cl)
1151±238
Simon et al. (2006); (100 MPa; 800 °C; NNO; H, O, Cl)
Frank et al. (2002); (100 MPa; 800 °C; NNO; ; H, O, Cl; CHClbrine <1.1×104 μg/g)
log CAubrine = 0.05[log CHClbrine] + 1.4
240±80
Frank et al. (2002); (100 MPa; 800 °C; NNO; H, O, Cl; CHClbrine > 1.1×104 μg/g)
logCAubrine = [2.2 × log CHClbrine] – 7.2
Partitioning between S-free aqueous brine and si-mt
Magmatic Sulfur and Ore Deposit Formation 533
534
Simon and Ripley
where CAs(OH)3v and CAs2O3m are the concentrations of As in the vapor and si-mt , respectively. Values for log KAs′ (±1Σ) are −1.3 (0.1) and −1.1 (0.1) for the S-free and S-bearing runs, respectively. The increase in the value of KAs′ with the addition of S is consistent with the presence of both arsenous acid and a As-S complex in S-bearing magmatic vapor. These results suggest that S has a small, but statistically meaningful, effect on the mass transfer of As between si-mt and low-salinity vapor. The reported DAuv/si-mt (±1Σ) in S-free and S-bearing assemblages are 15 ± 2.5 and 12 ± 0.3, respectively. The partitioning of Au between si-mt and vapor can be described by the apparent equilibrium constant: ′ = K Au
v × ( fH2 O )0.5 CAu si − mt v )2 × (CHCl CAu
(14)
where Civ and Cim are the concentrations of component i in the vapor and si-mt, respectively. The average value of log KAu′ (±1Σ) is −4.4 (0.1) in the S-free system. Data for the S-bearing system yield a mean value of log KAu′ (±1Σ) equal to –4.2 (0.2). These results indicate that S has a measurable effect on the mass transfer of Au from si-mt to a low-salinity vapor. Wood et al. (1987) quantified the solubility of Au and Ag in 0 to 5 m NaCl-bearing aqueous fluid at 200 to 350 °C, pCO2 ranged from 0.69 to 172 bars and fO2 and fS2 was fixed by the assemblage pyrite-po-mag. The authors highlight that hard-soft acid-base chemistry rules (cf. Pearson 1963) essentially prohibit anything but a minimal contribution to the solubility of Au and Ag by carbonate or bicarbonate complexes. Notably, hard-soft chemistry rules predict that Au should be quite stable as a Au-bisulfide complex, as demonstrated by Weissberg (1970) and Seward (1973). Seward (1973) experimentally determined Au solubility in S-bearing solutions and identified the Au species Au(HS)0 and Au(HS)2− in acidic solutions, and Au2(HS)2S−2 in basic solutions as dominating Au solubility. Seward (1984) proposed that Au solubility in mildly acidic solutions is controlled by HAu(HS)20 rather than Au(HS)0. Wood et al. (1987) reported that Au(HS)2− was not sufficient to account for the measured Au solubility values in their experiments. Wood et al. (1987) concluded that HAu(HS)20 was the dominant form of Au in acidic solutions at pH < 5 at temperatures <350 °C, but that Au(HS)2− may become increasingly important as a Au complex at ≥350 °C. They proposed that Au solubility may be controlled by the equilibrium Au( s ) + 2H 2S( aq ) = 3HAu(HS)02 + 0.5H 2 ( g )
(15)
and reported experimental data that indicated a temperature dependence on the value of log K, supporting the solubility of Au being dominated by HAu(HS)20 in acidic solutions at pH < 5, consistent with the results of Seward (1984). Dadze et al. (2001) confirmed the existence of stable Au-bisulfide complexes by performing experiments at 300 °C, 30.4 MPa, pH from 4.0 to 4.5, and variable H2S concentrations from 5×10−2 to 0.7 m. These authors reported that the solubility of Au was correlated with increasing concentration of H2S. Dadze et al. (2001) reported that the solubility data are consistent with the following equilibrium 2 Au( s ) + 3H 2S( aq ) + HS− = Au(HS)2− + HAu(HS)02 + H 2 ( g )
(16)
The solubility of Au in S-bearing magmatic-hydrothermally sourced fluid(s) operating at sub-solidus P-T conditions was quantified by Stefánsson and Seward (2004). They performed experiments at 50 MPa and 100 to 500 °C and measured the stability and stoichiometry of Au(I)sulfide complexes. They reported that the solubility of Au ranged from 3.6×10−8 to 6.65×10−4 molar in solutions over a pH range of 1.5 to 9.8, corresponding to ΣS2− between 0.0164 and 0.133 molar, ΣCl between 0.000 and 0.240 molar, ΣNa between 0.000 and 0.200 molar, and ΣH2 between 1.63×10−5 and 5.43×10−4 molar. The dissolution of Au in sulfide-bearing aqueous fluid(s) was described by the following equilibria:
Magmatic Sulfur and Ore Deposit Formation
535
Au( s ) + H 2S( aq ) = AuHS( aq ) + 0.5H 2 ( g )
(17)
Au( s ) + H 2S( aq ) + HS− = Au(HS)2− + 0.5H 2 ( g )
(18)
They reported that Au(HS)(aq) and Au(HS)2− were the dominant Au(I)sulfide species in acidic solutions at near-neutral pH. At temperatures >400 °C their data indicate that Au(HS)(aq) and Au(I)-chloride complexes become important, consistent with Frank et al. (2002), Hanley et al. (2005) and Simon et al. (2005). The equilibria presented above that describe the dissolution of Au in S- and Cl-bearing aqueous fluid(s) indicate that, in addition to temperature as noted in the studies by Stefánsson and Seward (2003a,b, 2004), the dissolution of Au is sensitive to changes in the redox state of the system as well as pH. Zezin et al. (2007) investigated experimentally the solubility of Au in H2S gas at temperatures of 300, 350 and 400 °C and pressures up to 23 MPa. They reported Au solubility values of 0.41.4 ng/g at 300 °C, 1-8 ng/g at 350 °C and 8.6-95 ng/g at 400 °C. The data indicated a positive dependence of log fAu on log fH2S and the authors proposed that the solubility of Au may be controlled by solvated gaseous sulfide or bisulfide complexes through the following equilibria: Au( s ) + ( n + 1)H 2S( g ) = AuS(H 2S)n ( g ) + H 2 ( g )
(19)
or Au( g ) + ( n + 1)H 2S( g ) = AuHS × (H 2S)n ( g ) + 0.5H 2 ( g )
(20)
where n is a statistical solvation number. The authors noted that both equilibria depend on the value of fH2 which has an important control on the redox equilibria for S species. In the experiments reported by Zezin et al. (2007), the S redox potential was buffered by S/H2S, yielding the possible reactions: Au( s ) + n × H 2S( g ) + S( l ) = AuS × (H 2S)n ( g )
(21)
Au( s ) + ( n + 0.5) × H 2S( g ) + 0.5S( l ) = AuS × (H 2S)n ( g )
(22)
Reported values of n for these latter reactions are 1.8 and 1.2 at 300 °C, 1.7 and 1.0 at 350 °C, 2.2 and 1.7 at 400 °C, respectively. The equilibrium constants for the latter two reactions increase with temperature and have values of log K = −11.1 ± 0.2 at 300 °C (both reactions), −10.7 ± 0.3 and −10.5 ± 0.3 at 350 C, and −10.6 ± 0.2 and −10.4 ± 0.2 at 400 C, respectively. The data indicate that the fH2S plays a role in the mobilization and transport of Au in magmatichydrothermal fluid(s). The transport of molybdenum in magmatic-hydrothermal fluids. Available data for the role of S in complexing and transporting Mo in high temperature aqueous fluid indicate that Mo solubility correlates negatively with S abundance. Tingle and Fenn (1984) quantified experimentally at 1000 °C and 100 MPa the effects of F and S on the mass transfer of Mo between granite si-mt and aqueous fluid and the solubility and transport of Mo in the fluid phase. They reported that the presence of F and S are not correlated with DMofl/si-mt. The authors concluded that while the presence of S was inconsequential to the Mo-scavenging potential of the fluid phase in their experiments, S is obviously required for deposition of Mo as MoS2, the dominant Mo-bearing mineral in magmatic ore deposits. Wood et al. (1987) quantified experimentally the solubility and speciation of Mo in 0 to 5 m NaCl-bearing aqueous fluid at 200 to 350 °C, pCO2 ranging from 0.69 to 172 bars, and fO2 and fS2 fixed by the assemblage pyrite-po-mag. They reported that molybdenite solubility was independent of total Cl and proposed that Mo was present in the fluid as molybdic acid, consistent with experimental data at higher temperatures. The capacity of the fluid to transport Mo is <1 μg/g over their range of experimental conditions. The authors cited thermodynamic
536
Simon and Ripley
calculations by Smith et al. (1980) that indicate that Mo is transported as a molybdate species in aqueous fluid and that molybdenite solubility is strongly dependent on temperature. The data in Wood et al. (1987) are inconsistent with the model results presented in Smith et al. (1980); i.e., there does not appear to be a strong temperature effect on Mo solubility in aqueous fluid. Wood et al. (1987) also discussed the potential existence of a molybdenite-carbonate complex and concluded that the presence of such a complex in aqueous fluid at temperatures up to several hundred degrees is unlikely. Rempel et al. (2006) studied experimentally the solubility of Mo trioxide in liquidundersaturated H2O-vapor at 300, 320 and 360 °C and 4 to 15 MPa to explore the ability of pure water to transport Mo at elevated temperature. Previous work on the volatility of Mo was reported by Millner and Neugenbauer (1949), Belton and Jordan (1965) and Glemser and von Haeseler (1962). Rempel et al. (2006) reported that the solubility of Mo trioxide ranges from 1 to 39 μg/g and was likely controlled by the equilibrium MoO3( g ) + nH 2 O = MoO3 × nH 2 O
(23)
where n, the hydration number, is 2.0 ± 1.0 at 300 °C, 2.4 ± 0.6 at 320 °C and 3.1 ± 0.3 at 360 °C. Values for the equilibrium constant for the above reaction are 18 ± 5 at 300 °C, 16 ± 3 at 320 °C and 12 ± 1 at 360 °C. The value of log MoO3⋅nH2O varies non-linearly with log fH2O. The authors proposed that the solubility of Mo in pure-H2O vapor may be controlled by the equilibrium MoO3( g ) + H 2O( g ) + 2H 2S( g ) = MoS2 ( s ) + 3H 2O( g ) + 0.5O2 ( g )
(24)
which yields a Mo solubility of 0.5 μg/g, at an fO2 and fS2 fixed by the assemblage pyritehematite-mag, at 600 °C and 50 MPa, conditions within the range of those determined for porphyry-type ore deposits. The transport of silver in magmatic-hydrothermal fluids. There are no data to our knowledge that constrain the effect of S on the partitioning of Ag between si-mt and aqueous fluid(s). However, in light of the data discussed below for the solubility of Ag in S- and Clbearing aqueous fluid at lower temperatures, it is likely that significant quantities of Ag are transported as an Ag-bisulfide complex. This is also consistent with acid-base predictions (cf. Pearson 1963). It is likely that both S and Cl are important ligands for Ag and future experimental research should aim to constrain the effect of S on complexing with and transporting Ag in high-temperature magmatic fluids. The mobility of Ag in magmatic-hydrothermally sourced fluid(s) operating at sub-solidus P-T conditions has been the focus of several experimental studies (Seward 1976; Kozlov et al. 1983; Webster 1986; Wood et al. 1987; Gammons and Barnes 1989; Gammons and WilliamsJones 1995; Spilsbury 1995; Gammons and Yu 1997; Migdisov et al. 1999; Akinfiev and Zotov 2001; Stefánsson and Seward 2003c). The sum of the available data elucidating the behavior of Ag in low-temperature (i.e., <500 °C) aqueous fluid(s) suggests that, depending on the pH and fHCl/fH2S ratio of the aqueous fluid, Ag solubility is dominated by either Ag-S complexes or Ag-Cl complexes (Akinfiev and Zotov 2001). Seward (1976), based on experiments in a S-free assemblage, reported that AgCln1−n describes the stoichiometry of Ag in S-free NaCl-H2O vapor and that the ligand number, n, increased with temperature up to at least 353 °C, at saturated vapor pressure. Further, the coefficient n increases as total chlorinity of the volatile phase increases. Seward (1976) reported that the species AgCl2− dominated Ag speciation in aqueous vapor at 277 to 353 °C. Stefánsson and Seward (2003c) reported, based on experimental data at 25 to 400 °C at saturated water vapor pressure and 50 MPa, that Ag solubility in sulfide-bearing aqueous solutions can be described by the following equilibria: 0.5Ag2S( s ) + 0.5H 2S( g ) = AgHS( aq )
(25)
Magmatic Sulfur and Ore Deposit Formation
537
0.5Ag2S( s ) + 0.5H 2S( g ) + HS− = Ag(HS)2−
(26)
Ag2S( s ) + 2HS− = Ag2S(HS)22 −
( 2 7)
The data from their study indicate that the solubility of Ag is dominated by AgHS(aq) in acidic solutions, by Ag(HS)2− in near-neutral solutions, and Ag2(HS)22− in alkaline solutions. The measured solubilities are between 0.01 and 14 μg/g in solutions with 0.007 to 0.176 m total reduced S. Other studies have examined the properties of Ag in more complex NaCl-KCl-HCl-H2O fluids (Gammons and Williams-Jones 1995; Akinfiev and Zotov 2001). Data from these studies are consistent with AgCln1−n as the dominant Ag-species in S-free fluids at temperatures up to 500 °C. The acid-base rules (cf. Pearson 1963) predict that Ag, being a soft metal, should complex with bisulfide in a S-bearing volatile phase. However, Wood et al. (1987) quantified experimentally the solubility of argentite (Ag2S) in NaCl-H2O-CO2 aqueous solutions (0.5 to 5.0 m NaCl; pCO2 ranging from 0.69 to 172 bars and fO2 and fS2 fixed by the assemblage pyritepo-mag) from 200 to 350 °C and reported that Ag0 and AgCl2− dominate Ag-speciation in S-bearing NaCl-H2O-CO2 aqueous solutions. The available experimental data for Ag solubility in Cl- and S-bearing aqueous fluid(s) at temperatures of a few hundred degrees and higher, and geologically reasonable redox conditions, indicate that Ag is transported as Ag-bisulfide, Agchloride and Ag-oxy-chloride. The transport of platinum group elements in magmatic-hydrothermal fluids. The measured high abundances of the PGE in high-salinity fluid inclusions from several porphyrytype (cf. Economou-Eliopoulos 2005) and layered-mafic-intrusion-hosted PGE deposits (e.g., Bushveld: Schiffries 1982; Ballhaus and Stumpfl 1986; Stillwater: Hanley et al. 2008;) and high Cl/F ratios in magmatic apatite (indicative of apatite growth in the presence of magmatichydrothermal fluid; Boudreau et al. 1986; Boudreau and McCallum 1989, 1992; Meurer and Boudreau 1996; Meurer et al. 1998; Boudreau and Meurer 1999; Boudreau and Simon 2007) interpreted as contemporaneous with PGE mineralization, indicate that aqueous transport of the PGE may be an important ore-forming process. There have been only a few studies of PGE solubility in magmatic-hydrothermal fluid(s) at high temperature and the data vary by at least one order of magnitude among studies. We present data from S-bearing studies, but caution the reader to carefully read all original studies to evaluate the experimental data. Fleet and Wu (1993) quantified experimentally the volatility of the PGE in S + Cl ± NaClbearing fluid at saturated vapor pressure in sealed silica tubes at 1000 °C by equilibrating either pure metals or PGE-bearing Fe-Ni-Cu sulfides and alloys with basalt si-mt; fS2 and ΣNaCl varied systematically. The data suggest that for individual metals, volatile transport occurs in the following decreasing order: Fe > Cu, > Ni > Pd. They reported the following metal concentrations in dry vapor at 1000 °C: 250 ng Pt, 100 ng Pd, 25 ng Au, 10 ng Rh, 250 ng Ru, 5 ng Ir, and 50 ng Os. The measured abundances of Pt and Pd in the volatile phase increased with increasing fS2, and the authors concluded that the measured metal abundances at high fS2 were inconsistent with the presence of only metal-chloride species. They concluded that metal transport is likely via a combination of both Cl- and S-metal complexes. Fleet and Wu (1995) studied experimentally the volatility of the PGE and Au at 1000 °C in the assemblage Fe-Ni-PGE-Au-mss ± S ± FeCl2 ± NaCl ± C at saturated vapor pressure in sealed silica tubes, with fCl2 and fS2fixed at geologically reasonable values. The data suggest that the volatility of the PGE does not depend on the fCl2. The presence of FeCl2, in addition to S and NaCl, enhanced minimally the volatility of the PGE and, notably, the data were interpreted to indicate that the presence of FeCl2, in the absence of NaCl, resulted in reduced volatility of Au. The concentration of Pd correlated with Pt (r = 0.89) and the concentration of Rh with Ir (r = 0.86), but Au was only weakly correlated with Pt (r = 0.67) and Pd (r = 0.43), as was Os with Ir
538
Simon and Ripley
(r = 0.57). The transport of the PGE and Au correlated with the composition of the base-metal sulfide; i.e., the presence of FeS resulted in measured fractionation of the PGE and Au with enhanced transport of Pt, Pd, and Au, whereas the presence of NiS-FeS resulted in minimal fractionation. Ballhaus et al. (1994) experimentally quantified the partitioning of Pt, Au, Fe, Ni and Cu in a NaCl-SiO2-C-H-O-S-volatile phase + (Fe,Ni,Cu)1–x-mss + metal assemblage at 900 °C and 1 GPa. The values of fO2 (~FMQ, ~MH, ~CCO, ~WM, ~IW) and fS2 were controlled at conditions appropriate for the evolution of PGE-rich layered mafic intrusions; H2S was the dominant S species (i.e., a high H2S to SO2 ratio). The fluid composition was dominantly H2O with minor CO2, CH4 and H2S, and all fluids were NaCl- and SiO2-saturated. The authors reported the following metal abundances in the fl in equilibrium with mss: Fe from 0.1 to 13.6 wt%; Cu from 600 to 2,000 μg/g; Ni was close to or below detection in all runs products; Au from 150 to 300 μg/g; Pt varied from below detection to several wt%. Interestingly, the two experiments from which Pt could be measured in run-product fluid were at quite different fO2 values; one was mildly reduced in the presence of mss and the other was highly oxidized at or above the hemmag buffer. The authors pointed out that the standard deviation of the Pt solubilities rendered it impossible to properly interpret the volatility of Pt in their experiments. Peregoedova et al. (2006) investigated PGE, Au, Ni and Cu in the Fe–Ni–Cu sulfide system at saturated vapor pressure in silica tubes at 1000 °C. They measured the mass transfer of the PGE, Cu and Au via a S-vapor from a PGE donor [S-rich (Fe,Ni,Cu)1−x S doped with about 2000 μg/g each of PGE and Au] to a S-poor PGE-free mss. The data indicate that the originally PGE-free mss acquired significant Ni, Cu, Au, Pt and Pd, but little Ir, Ru and Rh, during the course of the experiment. The authors reported that fS2 and the phase assemblage present in the donor, which was controlled by temperature, S, Ni and Cu content of the system, were the most important factors influencing the concentrations of Ni, Cu, Au and the PGE in the vapor. Copper was much more mobile in the vapor phase relative to Ni in experiments containing only Au-alloy and mss in the donor. The mobility of Au was an order of magnitude greater those of Pd and Pt, which in turn were an order of magnitude more mobile than Rh, Ru and Ir. The mass transfer of Pd was more efficient than that of Pt in experiments at the lowest S contents, when Pt-alloys formed in the donor system. The mass transfer of the PGE was quite efficient in experiments containing mss and su-mt; however, the efficiency of volatility was consistently Pd-Pt > Rh-Ru-Ir. The mass transfer of Ni was enhanced in the mss – alloy assemblage, whereas the mass transfer of Au was diminished in the presence of su-mt. In all cases, the mass transfer of the PGE and Ni was found to increase with increasing fS2, consistent with the enhanced stability of these metals as metal-sulfide complexes. Interestingly, the data indicated that the mobility of Cu and Au were not correlated with fS2, and the authors interpreted this as indicating a lack of metal-S complexing. Experiments to constrain the mobility of the PGE in magmatic-hydrothermally sourced aqueous H-O-S-Cl-fluid(s) operating at temperatures up to 300 °C demonstrate that Pt and Pd associate with HS− and Cl− ligands and that the solubilities of PGE are maximized in highlysaline, acidic aqueous fluids at highly oxidized (i.e., in the hematite stability field) conditions (Gammons et al. 1992; Wood 2002). Platinum and Pd dissolved as PtCl3− and PdCl42− in NaClH2O vapor (0.01 to 3.0 molal NaCl) at 300 °C and saturated water vapor pressure with total Pt and Pd concentrations of ≥ 1 ng/g each in the fluid phase (Gammons et al. 1992). Pt and Pd dissolved as Pt(HS)2 and Pd(HS)2 in H2S-rich aqueous solutions over a temperature range of 30 to 300 °C with the solubilities of Pt an Pd both < 0.1 ng/g (Gammons and Bloom 1993). This latter study reported that metal-Cl solubilities were much lower if H2S dominated the aqueous solution. These data demonstrate that Pt and Pd can be transported as both Cl- and S-complexes. Theoretical modeling of the mobility of the PGE using these low-temperature data results in estimates of PGE solubility at magmatic conditions that disagree with extrapolations
Magmatic Sulfur and Ore Deposit Formation
539
of experimental data (Sassani and Shock 1990, 1998; Wood 2002). Xiong and Wood (2000) determined experimentally that Pd solubility in 0.01m KCl supercritical aqueous fluid was 40 ng/g at 400 °C and 15 ng/g at 500 °C. These experimental solubility values are six and four orders of magnitude higher at 400 °C and 500 °C, respectively, than theoretical values reported by Sassani and Shock (1990, 1998). The large inconsistency demonstrates that extrapolating room temperature solubility data to high temperatures may result in underestimation of actual solubility values; hence, accurate models of PGE mobility in magmatic-hydrothermal fluids require more careful experimental work at high temperature.
The partitioning of ore metals between vapor and brine at temperatures below the water-saturated granite solidus Pokrovski et al. (2005; Fig. 6) experimentally quantified the partitioning of Fe, Zn, As, Sb, Ag, Au and Cu between low-density aqueous vapor and high-density aqueous brine at 350 to 450 °C along the vapor-liquid equilibrium curve in the model H2O-NaCl ± KCl ± HCl system. The salinity of the fluid phase was varied from pure H2O to pure NaCl. They reported that the Dxv/l for each metal was positively correlated with fluid density. All metals partitioned preferentially into the liquid relative to the vapor, however the Dxv/l trended toward unity as the critical point was approached. Note that at the critical point the densities of each fluid phase become identical and there is only a single one-phase fluid. Pokrovski et al. (2008) furthered their earlier effort by experimentally constraining the partitioning of the same suite of metals between vapor and brine in a S-bearing system. They conducted experiments at 350, 450 and 500 °C along the vapor-liquid equilibrium curve in the model H2O-NaCl ± KCl ± NaOH system over a range of fluid pH from 2 to 10 and ΣS from 0 to 0.7 m. They buffered fO2 to values close to the mag-hem
Figure 6. The experimentally determined data for the partitioning of Au and Cu in a H-O-S fluid assemblage, expressed as Km where Km = (molality of element in the vapor)/(molality of element in the liquid), published in Pokrovski et al. (2008) for Au and Cu between aqueous vapor and aqueous brine are plotted as a function of temperature (a) and pressure (b). In these plots, any datum that plots below the horizontal reference line of 1 indicates that the metal partitions preferentially into brine relative to coexisting vapor, and any datum that plots above the horizontal reference line of 1 indicates that the metal preferentially partitions into the vapor.
Figure 6.
540
Simon and Ripley
buffer; fS2 was self-buffered by S hydrolysis products and the use of NaOH. They reported that H2S was the dominant sulfide species with SO2 present at 10 to 30% of the total S. The authors reported that at acidic-to-neutral pH, the measured Dxv/l for As and Sb ranged from 0.1 to 1.0, and for Zn, Fe, Pb and Ag ranged from 0.001 to 0.1. These data indicate that Fe, Zn, As, Sb, and Ag preferentially partitioned into high-salinity aqueous fluid relative to low-salinity aqueous vapor even in the presence of geologically plausible fO2 and fS2 conditions and concentrations of reduced S. The Au and Cu data demonstrate that in the presence of ~1 wt% ΣS, with H2S being the dominant (>70% of ΣS) S species, Au partition preferentially into the low-salinity vapor relative to the high-salinity aqueous liquid, and that Cu displays an equal preference for the vapor and liquid. These results indicate that Au and Cu are likely both transported as both metal-bisulfide complexes and, in light of the experimental data for S-free systems (Table 4), metal-Cl complexes in aqueous vapor at both super- and sub-solidus conditions.
PORPHYRY-TYPE ORE DEPOSITS Porphyry basics Porphyry-type ore deposits are large-tonnage, low-grade (e.g., Cu, Mo and Au typically <1 wt%, <0.1 wt%, and <1 g/t, respectively), ore deposits associated spatiotemporally with intrusive rocks located stratigraphically above mid- to shallow-crustal plutons (Sillitoe 2010). The majority of primary Cu, Au and Mo mined from this deposit type is in metal-sulfides, e.g., chalcopyrite (CuFeS2), bornite (Cu5FeS4) and molybdenite (MoS2), hence requiring the presence of S for ore-metal precipitation. The absolute quantity of S in porphyry-type ore deposits can exceed 1 billion tons ΣS (Gustafson and Hunt 1975; Gustafson 1979). A number of recent review papers elucidate the details of porphyry deposits (Sillitoe 1997, 2000; Kerrich et al. 2000, Hattori and Keith 2001; Tosdal and Richards 2001; Richards 2003; Candela and Piccoli 2005; Cooke et al. 2005; Seedorff et al. 2005). Porphyry deposits account for approximately 65% of all mined Cu, 20% of all mined Au, 99% of all mined Mo, and 14% of all mined Ag (Singer and Cox 1986; Singer 1995; Kirkham and Sinclair 1996) and also contain high concentrations of other notable metals including Re, Fe, As, Bi, B, Sr, Zn, Pb, Co, V, and the PGE (Candela 2003). In terms of absolute metal abundance, there are twenty-five porphyry deposits containing >10 million tons Cu and >250 tons Au, and eleven porphyry deposits containing >0.5 million tons Mo (Cooke et al. 2005). Significant progress has been made in the past few decades to elucidate the chemical link between spatially and temporally associated porphyry-type and high-sulfidation-type ore deposits. This evidence is based on the observation of magmatic signatures for oxygen, hydrogen and S isotopes in most sampled porphyry-type and high-sulfidation ore deposits. We do not focus in this chapter on the connection between these deposit types. Interested readers are referred to key publications (White and Hedenquist 1990; Henley 1991; White 1991; Sillitoe 1993; Hedenquist and Lowenstern 1994; Arribas 1995; Hedenquist et al. 1998; Muntean and Einaudi 2000, 2001; Simmons et al. 2005; Heinrich et al. 2009)
Tectonic setting and associated magma composition of porphyry deposits All porphyry ore deposits share the commonality that they are related spatially, temporally, and chemically to silicate magma in Earth’s upper crust. Porphyry ore deposits are typically found at paleodepths ranging from 1 to 8 km, and there is an observed relationship between the depth of formation and the dominant metals in the ore body. Allowing for some overlap, porphyry-Cu-Au-(Mo) deposits are typically found at depths ranging from 1 to 4 km, and porphyry Mo-(Cu, Au) deposits at depths from 4 to 8 km. While there is an observed relationship between the formative tectonic environment and the principal metal endowment of some subclasses of porphyry deposits, i.e., an apparent distinct difference in the tectonic
Magmatic Sulfur and Ore Deposit Formation
541
position of porphyry-Cu-Au vs. porphyry-Mo deposits, there is no unique tectonic environment for porphyry deposit formation. Porphyry-Mo deposits are associated commonly with A-type granites emplaced in continental extensional settings, although porphyry-Mo deposits are also found in purely arc environments. Porphyry-Mo deposits in rift-related environments are referred to as Climax-type Mo deposits and those in subduction environments as Endako-type Mo deposits (Sillitoe 1980; Sinclair 2007). The distinction between these deposit types also correlates with metal tenor; i.e., subduction-related Mo deposits typically contain <0.15% Mo and extensional-related Mo deposits typically contain >0.15% Mo (Kirkham and Sinclair 1995; Sillitoe 1980; Sinclair 2007). Climax-type Mo deposits are associated commonly with F-rich (>0.1 wt% F) high silica granitic (s.l.) rocks whereas Endako-type Mo deposits are thought to evolve from F-poor (<0.1 wt%) magmas that range from granodiorite to granite (Mutschler et al. 1981; Theodore and Menzie 1984; Linnen et al. 1995; Selby et al. 2000; Seedorff et al. 2005; Sinclair 2007). Porphyry-Cu-Au deposits evolve from magmas that range in composition from low-silica (~45 wt%) to calc-alkaline granites (s.l) (Seedorff et al. 2005). Porphyry-Cu deposits are associated with intermediate to felsic, calc-alkaline intrusive rocks ranging in composition from granodiorite to granite (Kesler et al. 1975, 1977; Titley and Beane 1981). Most porphyry-Cu-Au deposits that formed in the root zones of andesitic stratovolcanoes are associated with Mesozoic to Cenozoic orogenic belts in the western Americas, the western Pacific margin and the Tethyan orogenic belt in eastern Europe and southern Asia (Sillitoe 1972, 1973, 1997; Kerrich et al. 2000; Sinclair 2007). However, there is no fundamental requirement that porphyry deposits be overlain by a contemporaneous volcanic system. Some of the largest tonnage porphyry-Mo (e.g., Bingham Canyon) deposits formed several hundred km inland from active subduction, and some of the largest porphyry-Au-Cu deposits (e.g., Grasberg) formed distal from steadystate subduction (Sinclair 2007). Pasteris (1996) suggests that volcanic activity may eliminate the ability of a magmatic system to produce a porphyry deposit owing to the loss of metals to a volcanic gas phase that vented to the atmosphere. The near-surface ponding of magma, without extensive contemporaneous volcanism, leads to temporally protracted differentiation of the si-mt and this, in turn, may allow for the ponding of exsolved metal-rich volatiles in the apical portions of the magma chamber (Candela 1991). An oblique compressional environment (Richards 2003) minimizes the development of unique apophyses that may form from the primary magma chamber and, in turn, facilitates more focused fluid flow as the exsolved volatiles shatter the solidification front (Burnham 1967; Candela 1991; Sillitoe 1998). Nearinstantaneous “eruption” of the magmatic volatile phase(s) is followed by rapid fluid ascent into the surrounding rock, which can be wallrock or part of the carapace of the original intrusion, where metal deposition occurs.
Source of sulfur in porphyry environments Sulfur has four naturally occurring stable isotopes: 32S, 33S, 34S and 36S (Marini et al. 2011, this volume). Knowledge of the intra-element fractionation of S during geological processes allows the measured abundances of S isotopes to be used to identify the source reservoir for S in sulfides and sulfates found in magmatic-hydrothermal ore deposits (Ohmoto and Rye 1979; Seal 2006). A large number of studies have quantified the δ34S of sulfide minerals that host ore mineralization in porphyry-type ore deposits (Field 1966; Field and Gustafson 1976; Eastoe 1978; Ohmoto and Rye 1979; Frei 1995; Hezarkhani and Williams-Jones 1998; Spry et al. 1996; Zhang et al. 1996; Imai 2000; Hattori and Keith 2001). Published δ34S data from a large number of porphyry deposits are presented in Figure 7. The value of δ34S for S in magmatic systems is 0% and, allowing for some scatter that is explained by minor assimilation during ore formation, evinces that the majority of S in porphyry ore deposits was sourced from a magmatic reservoir.
Simon and Ripley
542
Figure 7.
Figure 7. The published δ34S data from a large number of porphyry deposits are consistent with a magmatic source for S. Originally published by Hattori and Keith (2001) with data from Eastoe (1983), Field (1966), Field and Gustafson (1976), Ohmoto and Rye (1979), Frei (1995), Hezarkhani and WilliamsJones (1998), Spry et al. (1996), Zhang et al. (1996) and Imai (2000). A δ34S value of zero is interpreted to represent magmatically-sourced S and, thus, the clustering of ore-mineralized sulfides from porphyry ore deposits are consistent with S having been sourced dominantly from a magmatic source. The +10 to −10% deviation in the measured values of δ34S for some deposits are hypothesized in the original studies to reflect post-mineralization alteration. Abbreviations: Bg, Bingham Canyon, Utah; Bt, Butte, Montana; ElSv, El Salvador, Chile, Ch, Chino, New Mexico; Yr, Yerington, Nevada, Bis, Bisbee, Arizona; Lp, Lepanto Far Southeast, Philippines, Aj, Ajo, Arizona, Sg, Sungun, Iran, Fr, Frieda River, Papua New Guinea,VC, Valley Copper, British Columbia, Pang, Pangua, Papua New Guinea, Tn, Tintic, Utah; GC, Galore Creek, British Columbia, GM, Globe-Miami, Arizona; Cm, Craigmont, British Columbia, Cm, Craigmont, British Columbia; Gs, Gaspé Copper, Québec; CV, Cerro Verde-Santa Rosa, Peru; MP, Mineral Park, Arizona; SK, Skouries, Greece, Hb, Hillsboro, New Mexico; GS, Golden Sunlight, Montana.
Source of ore fluids in porphyry environments The δ34S and δ65Cu (Fig. 8) suggest that S and Cu in porphyry deposits are sourced from magma (Maréchal et al. 1999; Zhu et al. 2000; Larson et al. 2003; Albarède 2004; Ehrlich 2004; Markl et al. 2006; Asael et al. 2007; Mathur et al. 2009). Further constraints can be placed on the source of fluid components in porphyry deposits by investigating the source of H and O isotopes in minerals that are contemporaneous with ore-metal-sulfide deposition. The use of measured δD and δ18O values to assign a fluid source to a specific geologic reservoir is based on data that evince a lack of variation of δD and δ18O in geologic reservoirs over much of Earth’s history (Criss 1999; Mojzsis et al. 2001). Hedenquist and Lowenstern (1994) discussed the δD and δ18O characteristics of porphyry ore deposits in comparison with the following crustal fluids: subduction-related volcanic vapor, arc magmas, crustal felsic magmas, degassed si-mt, midocean ridge basalt, and meteoric (i.e., atmospheric) water. Hedenquist and Lowenstern (1994) reported that minerals associated spatiotemporally with ore metal precipitation in porphyry environments consistently yielded δ18O values between +6 to +9%. These data were interpreted by Hedenquist and Lowenstern (1994) to indicate that ore fluid evolved from silicate magma. The published δD data indicate that δD ranges from −35 to −75%, which is consistent with open-system degassing of the causative magma that drives δD to more negative values relative
Magmatic Sulfur and Ore Deposit Formation
543
Figure 8. Figure 8. Published δ65 Cu data from Mathur et al. (2009), Marechal et al. (1999); Zhu et al. (2000); Larson et al. (2003); Ehrlich (2004); Markl et al. (2006); Asael et al. (2007) for the primary ore-metal bearing Cu-sulfide minerals chalcopyrite and bornite from nine porphyry ore deposits (Butte, Montana, U.S.A.; Silver Bell, Arizona, U.S.A.; Collahuasi, Chile; Chuquicamata, Chile; Bisbee, Arizona, U.S.A.; Escondida, Chile; El Salvador, Chile; Sisir-100, Turkey). A δ65Cu value of zero is interpreted to represent magmatically sourced S. The data indicate that the Cu in primary mineralized Cu-sulfides from these ore deposits is magmatically source. The deviation in the measured values of δ65Cu for some deposits reflects postmineralization alteration. Secondary-enrichment and leach cap minerals data are from minerals derived from secondary mineralization (chalcocite, Fe-oxides, Cu-oxides). Figure is modified from Mathur et al. (2009).
to the range (i..e, δD = −20 to −45%) expected for pristine magmatic water. The available δD and δ18O data suggest that the primary ore-forming fluid(s) in the porphyry environment are sourced from degassing si-mt in magma chambers beneath the porphyry environment.
Constraints on the composition of porphyry-ore forming fluids There is no debate on the role of magmatic-hydrothermal fluids in the formation of porphyry-type ore deposits. As discussed above, magmatic volatiles in volcanic settings evince that H-O-S-Cl-fluid(s) evolved from near surface magma chambers may efficiently scavenge and transport metals. Metals in porphyry-type ore deposits are accepted to have precipitated from magmatic-hydrothermal H-O-S-Cl fluid(s) evolved from magma (Burnham 1979; Holland 1972; Whitney 1975; Candela and Holland 1984; Dilles 1987; Hedenquist and Lowenstern 1994; Cline and Bodnar 1994; Candela and Piccoli 1995, 2005; Cline 1995; Shinohara and Hedenquist 1997; Heinrich 2005; Williams-Jones and Heinrich 2005). This genetic link is based on numerous studies that have described the close association between ore-stage mineralization and H-O-S-Cl fluid inclusions, interpreted to have been trapped at temperatures close to the water-saturated granite solidus, based on homogenization temperatures up to 700 °C, to temperatures coincident with metal precipitation, from 300 to 450 °C (Roedder 1972, 1984; Eastoe 1978; Henley and McNabb 1978; Hedenquist and Lowenstern 1994; Bodnar 1995; Candela and Piccoli 2005). Fluid inclusion assemblages associated with porphyry ore mineralization are comprised characteristically of both low-salinity aqueous vapor and higher-salinity aqueous brine, represented commonly by the H2O-NaCl system (cf. Bodnar et al. 1985). Natural fluid inclusions also contain S, but the ability to quantify S in fluid inclusions has been available only recently (Seo et al. 2009; Ripley et al. 2011, this volume). The fluid inclusion solute data that include S suggest that the abundances of S and metals such as Cu and Au correlate positively in some magmatic-hydrothermal ore deposits (Seo et al. 2009), and much more work needs to be performed on natural and experimental inclusion assemblages to investigate the importance of S as a metal-transporting ligand. In the porphyry environment, the H-O-S-Cl volatiles are preserved as fluid inclusions, trapped contemporaneously in ore-stage gangue minerals during
544
Simon and Ripley
mineralization. Fluid inclusions evince absolutely that aqueous fluid(s) exsolved from si-mt are the primary agent of metal mass transfer from the causative magma to the porphyry environment. The fluid inclusion record constrains the pre-, syn- and post-ore stage compositional evolution of fluids that are produced during protracted fractionation of shallow-level si-mt. A large number of fluid inclusion studies in porphyry-ore environments have been performed to elucidate the nature of magmatic-hydrothermal fluids and the details are summarized in several comprehensive reviews (Roedder 1984; Williams-Jones and Heinrich 2005). Data from natural fluid inclusion boiling assemblages are consistent with a non-Cl ligand playing a role in ore metal-complexation and transport in magmatic-hydrothermal fluid(s) (Fig. 9) (Heinrich et al. 1992; Audétat et al. 1998, 2000a,b, 2008; Heinrich et al. 1999; Ulrich et al. 1999; Baker et al. 2004; Seo et al. 2009). The data from natural inclusion assemblages indicate that, in some samples from some porphyry-type ore deposits, Cu, Au, and As display a preference for low-density vapor relative to cogenetic higher-density brine. Charge balance demonstrates that the average Cl concentration in vapor fluid inclusions is insufficient to balance the heavy-metal budget of vapor fluid inclusions (Heinrich et al. 1992). Thus, the strong enrichment of metals in the vapor phase cannot result solely from volatile metal-Cl species, e.g., CuCl0. Metal complexation must be related to aqueous components other than just Cl. Sulfur is a prime candidate owing to its high concentration in volcanic gases, as discussed above (Hedenquist and Lowenstern 1994; Williams-Jones and Heinrich 2005), and strong fractionation into vapor relative to coexisting brine (Drummond and Ohmoto 1985). Heinrich et al. (1999) cited the presence of up to 1 wt% S in vapor inclusions associated with mineralization in the Yankee Lode as evidence that a S ligand (e.g., H2S, HS− or SO2) may have been responsible for the observed fractionation of Cu into the vapor phase. It is important to note that vapor-brine boiling assemblages in many porphyry-deposits are consistent with the paradigm requiring Cl-rich aqueous fluid to be the metal transporting agent in porphyry environments. The available data from natural inclusions for ore metal transport in H-O-S-Cl fluids indicate that ore metals sometimes fractionate into the brine and sometimes into the vapor. Numerous studies have demonstrated that both Cu (Nagaseki and Hayashi 2008; Pokrovski et al. 2008) and Au (Benning and Seward 1996; Gibert et al. 1998; Loucks and Mavrogenes 1999; Pokrovski et al. 2008) may be present as metal-S species in magmatichydrothermal H-O-S-Cl fluids at the subsolidus conditions which prevail in the porphyry depositional environment and recent data from natural coexisting co-existing H-O-S-Cl vapor and brine provide important clues as to the role of S in ore metal transport. Seo et al. (2009) reported data that quantified the concentrations of S, Cu, Au, As, Mo, Cs, Pb, Na, K and Fe in coexisting, cogenetic vapor and brine fluid inclusions trapped during mineralization in the following environments: the Bingham Canyon, U.S.A., and Bajo de la Alumbrera, Argentina porphyry deposits; Mole Granite veins; Zinnwald, Germany granite related Sn-W veins; and, Rito del Medio, U.S.A. In all of these ore systems, the analyzed inclusions were trapped at temperatures from 320 to 490 °C and pressures from 9 to 48 MPa. The authors quantified S and metal abundances by using LA-ICP-MS and reported that NaCl, KCl, FeCl2, Cu and S comprise >97% of the fluid inclusions solute load of both vapor and brine. The measured concentrations of S and Cu are similar to one another, and ΣS exceeded the concentrations of all other chalcophile metals in both vapor and brine. Their Cu and S data (Fig. 9) indicate that the measured concentration of Cu was greater in the brine relative to coexisting vapor in samples from Rita del Medio, Bingham Canyon and Bajo de la Alumbrera. In samples from the other locations, the measured concentration of Cu was greater in the vapor relative to coexisting brine. Their data (Fig. 9) were presented as Cu/Na and S/Na ratios owing to the low salinity of the vapor phase fluid inclusions, preventing the determination of microthermometric salinities for use as an internal standard for data reduction. The ratios of Cu/Na vs. S/Na indicated that the molar S/Cu ratio is ~2, implying that the fluids contained
Magmatic Sulfur and Ore Deposit Formation
545
Figure 9. The concentrations of S and Cu in natural co-existing, cogenetic vapor and brine magmatichydrothermal fluid inclusions; elemental ratios are molar based. The data plotted in (a) and (b) indicate that both vapor and brine contain S concentrations equal to Cu or contain an excess of S over Cu. The data are plotted in (c) as ratios of Cu to Na and S to Na to eliminate uncertainties introduced by analytical calibration (cf. Heinrich et al. 2003). The data demonstrate that these fluid inclusions contain sufficient S to complex and transport Cu. [Used by permission of Elsevier, from Seo et al. (2009), Earth and Planetary Science Letters, Vol. 282, Fig. 2, p. 327].
Figure 9.
sufficient ΣS to precipitate Cu without depleting the fluid in S, the fluid maintaining high enough additional S to precipitate additional metal sulfide phases (e.g., MoS2, FeS2). The authors postulated that the enhanced partitioning of Cu into the vapor, relative to brine, in samples from the Mole Granite and Zinnwald veins may reflect the oxidation state of the source magma. As discussed above, the oxidation state of the system controls the speciation of S, and the stability of possible metal-S complexes in H-O-S-Cl fluid(s). In reduced granitic systems (FMQ+2) Rita del Medio, Bingham Canyon and Bajo de la Alumbrera systems, where Cu preferentially partitions into the brine, metals such as Cu
546
Simon and Ripley
and Au may be transported as a charged S-metal species, e.g., Cu(HS)2−, which is more stable in the high-density aqueous liquid relative to the low density aqueous vapor, or as a metal-Cl complex as discussed above.
Harmonizing fluid transport data from nature and experiments The sum of the available experimental and natural data that elucidate the transport of S and ore metals in H-O-S-Cl fluid(s) demonstrates that a number of factors, including pH, fO2, fS2, total Cl, total S, pressure and temperature, are critical controls on the metal-transport capacity of the fluid phase and the partitioning of metals between immiscible H-O-S-Cl fluids (i.e., vapor and brine) and si-mt. The extant data indicate that while ore metals (e.g., Cu) display no consistent behavior with respect to the partitioning of ore metals between low-density and high-density H-O-S-Cl fluids (i.e., the available data do not facilitate the development of a quantitative predictive model of metal behavior), the presence of S enhances the metaltransporting capacity of all sampled magmatic-hydrothermal H-O-S-Cl fluids (i.e., vapor, brine, and supercritical single-phase fluids). To explore the role of S in transporting metals in immiscible aqueous fluid assemblages, we will use Cu owing to the larger available partitioning data set for this metal. The available experimental data that elucidate the partitioning of Cu are plotted in Figure 10 (top) and a compilation of partitioning data for Cu from experiments and several ore deposits are plotted in Figure 10 (bottom). The data are plotted as DCub/v against fluid temperature. Figure 10 is annotated with pressure and relative fS2 is denoted for all experiments; all natural inclusions contain S. The experimental DCub/v data (Fig. 10 top) indicate that Cu partitions preferentially into brine over the range of geologically plausible fS2 values and temperatures from magmatic to precipitation temperatures common for porphyry environments. The addition of natural fluid inclusion DCub/v data in Figure 10 (bottom) allows the following observations to be made: 1) there is seemingly no relationship between pressure and the partitioning of Cu between brine and vapor; 2) as discussed above, metals such as Cu partition sometimes into vapor and sometimes into brine; and, 3) higher values of fS2, within the range of values measured for porphyry environments, correlate positively with an increase of the Cu-capacity of the vapor, but increasing fS2 at geologically plausible values does not result in a higher Cu content of the vapor. The data (Fig. 10) are seemingly consistent with Seo et al. (2009) who postulated a relationship between magma oxidation state and the ability of S to scavenge and transport ore metals in porphyry environments. The exception to this is the data set from Bajo de la Alumbrera (Halter et al. 2002, 2005) where the data indicate a potential relationship at high temperature between DCub/v and pressure, but at lower temperature the reported DCub/v values are > 1. Clearly, more experimental work is required to constrain the partitioning of metals in H-O-S-Cl fluid assemblages.
Oxidation state of causative magmas: the role of sulfide vs. sulfate The oxidation state of source magmas for porphyry deposits ranges from ~ FMQ<1 to FMQ+3 (Rowins 2000; Chambefort et al. 2008; Richards 2009). As noted for several reduced ore deposits (i.e., the Mole Granite, Zinnwald and Mogilata) and oxidized ore deposits (i.e., Grasberg and Bajo de la Alumbrera), there might be a relationship between magma oxidation state and the nature of the partitioning of metals among coexisting H-O-S-Cl magmatichydrothermal fluids. Metals in Cu-, Au- and Mo-porphyry ore deposits are mined as metalsulfide and this is consistent with experimental and field data that evince the significant role for sulfide to scavenge metals from si-mt and transport the metals into the porphyry environment. Halter et al. (2002, 2005) present data for samples from Bajo de la Alumbrera that elucidate the role of immiscible su-mt, preserved as su-mt inclusions, during the formation of this worldclass porphyry deposit. This ore deposit was formed by fluids evolved from calc-alkaline magma at an oxidation state that stabilized sulfide and not sulfate; i.e., fO2
Magmatic Sulfur and Ore Deposit Formation
547
Figure 10.
Figure 10. Published wt% based experimental partitioning data for Cu are plotted (top) and a compilation of metal partitioning data from experiments and several ore deposits are plotted (bottom). Pressure in MPa is indicated in parentheses next to each datum, and relative fS2 is denoted for all experiments. The data are plotted as DCub/v against fluid temperature. In these plots, any datum that plots above the horizontal reference line of 1 indicates that the metal preferentially partitions into brine relative to coexisting vapor, and any datum that plots below the horizontal reference line of 1 indicates that the metal preferentially partitions into the vapor. Deposits locations: Mole Granite, Australia (Audétat et al. 2000a,b); Zinnwald, Germany (Heinrich et al. 1999); Grasberg, Irian Jaya (Ulrich et al. 1999); Bingham, Utah (Redmond et al. 2004); Mogilata, Bulgaria (Kostova et al. 2004); Bajo de la Alumbrera, Argentina (Halter et al. 2002, 2005).
stage minerals that formed prior to the si-mt reaching aqueous volatile phase saturation. They reported that su-mt inclusions contain up to 9 wt% Cu and several μg/g Au. The concentrations of Cu and Au in the su-mt inclusions are orders of magnitude greater than in the coexisting simt, consistent with the data for the partitioning of Cu and Au presented above. The Cu/Au ratio of su-mt inclusions overlaps the Cu/Au ratio of H-O-S-Cl-brine inclusions that were trapped contemporaneously with the earliest stage of mineralization. Notably, the authors documented the absence of sulfide in the volcanic groundmass and in silica-enriched si-mt inclusions, the latter tracking the silica enrichment during differentiation of the source magma. The authors postulated that early formed immiscible su-mt efficiently sequestered Cu and Au from the simt and that this sulfide was subsequently resorbed by the si mt during degassing of the si-mt, and the latter process generated the Cu- and Au-mineralizing brine. Resorption is consistent
548
Simon and Ripley
with the absence of sulfide in the volcanic groundmass. While these data do not imply that all porphyry ore deposits formed via early sulfide immiscibility and degassing-driven resorption, the data are consistent with the observations reported by Seo et al. (2009) and demonstrate the significant metal-transporting role of reduced S during porphyry evolution. The resorption of sulfides during magma evolution of layered mafic intrusions has been discussed by Boudreau and Meurer (1999) who emphasized that sulfide crystals and/ or immiscible su-mt formed prior to volatile saturation can act as efficient pre-concentrating agents that may promote the mineralization process. Boudreau and McCallum (1992) suggested that sulfide resorption from the early cumulus assemblage of magma bodies may increase the ore-forming potential of a given magmatic system. Resorption of the sulfides is driven by the exsolution of a volatile phase which causes self-oxidation of the magma, resulting in an increase in the sulfate/sulfide ratio of the si-mt (Bell and Simon 2011). Aqueous phase volatile saturation of si-mt and concomitant sulfide destabilization have also been proposed as the mechanisms responsible for ore mineralization in the Bingham porphyry-Cu, -Mo deposit, Utah (Keith et al. 1997; Stavast et al. 2006). These studies document the strong enrichment of Cu and Ag in magmatic sulfides present in latites that are co-genetic with the ore-related intrusions. Early sulfides in the ore-related magma would also be expected to contain high concentrations of ore metals. The exsolution of an aqueous volatile phase may have caused large-scale resorption or dissolution of magmatic sulfides which would supply significant quantities of Cu to the evolving magmatic volatile phase(s) (Keith et al. 1997). Hattori (1993) postulated the potential resorption of sulfides owing to an influx of SO2-rich fluids that can oxidize sulfides during magma recharge. The sulfide-resorption hypothesis calls upon the introduction of significant quantities of SO2 into shallow-level magma, resulting in magma oxidization and sulfide resorption. The data on S capacity of si-mt indicate that self-oxidation need only increase the fO2 of typical arc magmas by one to two orders of magnitude, e.g., from FMQ+2 to FMQ+3, to reduce the stability of sulfide relative to sulfate. Sulfate in the form of anhydrite (CaSO4) is a seemingly ubiquitous phase in many porphyry ore deposits, and thermodynamic data for the solubility of anhydrite (Newton and Manning 2005) and for ore metal transport in H-O-S-Cl-fluids are certainly not inconsistent with a role for oxidized S in the evolution of porphyry ore deposits (Webster et al. 2009). Magmatic anhydrite has been documented as a primary pre-eruptive phase in many water-rich, oxidized (>FMQ+2) arc volcanic systems, e.g., El Chicón, Mexico (Luhr et al. 1984; Carroll and Rutherford 1987; Luhr 1990), Pinatubo, Philippines (Bernard et al. 1991; Scaillet and Evans 1999), Lascar, Chile (Matthews et al. 1994, 1999) and Eagle Mountain, U.S.A. (Parat et al. 2002), and in magmatichydrothermal ore deposits at e.g., Yerington (Streck and Dilles 1998), Santa Rita (Audétat et al. 2004) and El Teniente (Stern et al. 2007). The pre-eruptive origin of anhydrite is evinced by the presence of anhydrite inclusions within silicate phenocrysts, e.g., hornblende and apatite, indicative of a hydrous si-mt (Luhr 2008). Magmatic anhydrite has also been documented in plutonic rocks. For example, Barth and Dorais (2000) documented the presence of anhydrite hosted within poikilitic hornblende and by using phase equilibria they determined that the anhydrite precipitated from si-mt at temperatures between 700 and 800 °C at 600 MPa from a hydrous, oxidized sulfate-saturated andesitic to dacitic magma. Barth and Dorais (2000) noted that comagmatic apatite is sulfate rich and suggested that analysis of the sulfate/sulfide ratio of apatite may serve as a proxy for the oxidation state of the magma. Chambefort et al. (2008) reported data from the world-class Yanacocha, Peru, ore deposit, an ore system that includes Cu-Au porphyry and high-sulfidation epithermal mineralization, and discussed the role for magmatic sulfate during the development of magmatic-hydrothermal ore deposits. They reported that the causative magma at Yanacocha contained >1000 μg/g, and S was present as both sulfide and sulfate in the si-mt and also as the discrete sulfate phase anhydrite. The presence of magmatic anhydrite constrains the oxidation state of the magma to
Magmatic Sulfur and Ore Deposit Formation
549
greater than ~FMQ+2.5, based on the stability of sulfate and sulfide in silicic magmas (Jugo 2009). Chambefort et al. (2008) reported rare earth element abundances of magmatic anhydrite that are consistent with the anhydrite having nucleated and grown at equilibrium from the host andesite and dacite si-mt (Chambefort et al. 2008). Co-magmatic apatite at Yanacocha is S-rich, consistent with the findings of Streck and Dilles (1998) who documented the transition from sulfate-rich cores to sulfate-poor rims in apatite as evidence for the crystallization of magmatic apatite from si-mt in the causative magma in the Yerington porphyry system, Nevada, U.S.A. Chambefort et al. (2008) documented the presence of amoeboidal and wormy anhydrite trapped as inclusions within hornblende. The authors postulated that the texture of anhydrite is consistent with the presence of immiscible CaSO4-H2O melt. This is consistent with Jugo et al. (2005a,b) who reported the existence of sulfate melt in equilibrium with basalt si-mt at 1300 °C, 1 GPA, fO2 > FMQ+3, and ΣS = 1.5 ± 0.2 wt%. The findings reported by Chambefort et al. (2008) evince that world-class magmatic ore deposits may evolve from oxidized magma (i.e., >FMQ+3) and that magmatic sulfate, i.e., not sulfate produced by devolatilization-driven selfoxidation, may be a potentially critical ingredient to the porphyry recipe in some environments.
Causative magma sources: normal or enriched? There is no debate regarding the magmatic source of S and metals in porphyry ore deposits. The available data from experiments and natural assemblages evince the ability of exsolved H-O-S-Cl-fluid(s) to scavenge ore metals from si-mt, su-mt and su-xtals and transport the metals into the porphyry environment seem conclusive with respect to the ability of “normal” magmas to generate metal anomalies such as porphyry deposits. However, the role of “normal” vs. metal-enriched source magmas in the development of porphyry deposits remains a focus of discussion. Core et al. (2006) discussed this topic with specific relevance to the role of magmatic S. They presented data from the Bingham Canyon, Utah, U.S.A., Cu-Mo-porphyry deposit, which they interpreted as consistent with a role for a S- and metal-rich source magma. The presence of such an enriched source may obviate the need for large magma volumes typically required to mass balance the metal endowment of porphyry deposits. The authors documented the presence of bornite- and chalcopyrite-rich mafic enclaves associated cogenetically with mineralization in the Last Chance stock, an intrusion parental to the Bingham porphyry deposit. The mafic enclaves were demonstrated to have crystallized from the Cu-rich causative magma. Core et al. (2006) concluded that the source magma may have involved melting of a lowercrustal source with an anomalously high Fe and Cu content, owing to the requirement that early crystallization of bornite and chalcopyrite require such a source. They hypothesized that the source magma may have entrained sulfide cumulates that were resorbed during ascent, and they cited evidence of sulfide cumulates in mafic xenoliths of deep crustal origin from other studies (Dromgoole and Pasteris 1987; Fleet and Stone 1991; Guo et al. 1999; McInnes et al. 2001). A second possibility calls upon assimilation of Cu-Fe-sulfides present in high-grade metamorphic rocks that contain Cu-sulfide ore deposits. Such rocks have been reported in the Okiep area of South Africa and the Curaca Valley of Brazil (Clifford et al. 1981; Raith and Prochaska 1995; Kisters et al. 1998; Maier and Barnes 1999; Maier 2000). Partial melting and assimilation of Cu-Fe-sulfide-rich source rock into a lower crustal magma that ascends through the crust may ultimately result in resorption of the sulfides, as demonstrated experimentally by Tomkins and Mavrogenes (2003), who provided evidence that such entrained sulfide may be fully resorbed. Certainly, these studies do not necessitate the need for metal-rich sources; however, they illustrate the amplification potential of involving a S- and metal-rich source and should serve to stimulate future studies of the nature of the magma source. A model for the oxidation state and composition of the magma source for porphyry deposits was developed by Mungall (2002) who discussed the necessity to oxidize the mantle wedge and resorb sulfides, the latter controlling fundamentally the ore-generative capacity of an arc magmatic system. Mungall (2002) presented a thermodynamic argument for the necessity
550
Simon and Ripley
of a component of slab si-mt at the base of arc volcanic plumbing systems as a critical agent for the oxidation of the mantle wedge to FMQ+2, the critical boundary between sulfide and sulfate stability (Jugo 2009). Mungall (2002) pointed out that most arc volcanism is not associated with porphyry deposits and, thus, perhaps there is something unique about the nature of melting in the source that generates high ore-potential magmas well below the depths of final resting in the upper crust. This hypothesis does not eliminate the need for efficient metal scavenging by H-O-S-Cl fluid(s) in the near-surface environment, nor the potential anatexis and assimilation of metal-sulfide rich protoliths, nor tectonic oblique convergence (cf. Richards 2003). Rather, the model is based on the observation that ore potential is maximized when the sub-arc mantle is oxidized, hence ore-metals bound as sulfides are made available to the evolving and ascending arc magma. If the sub-arc mantle remains completely in the sulfide stability field (Jugo 2009), ore metals will remain as restite and this seemingly eliminates ore potential at higher stratigraphic levels of the arc-crust system, unless the ascending magma assimilates metal-sulfide rich domains (e.g., sulfide ore deposits; Tomkins and Mavrogenes 2003; Core et al. 2006). We reiterate here the conclusion presented in Jugo (2009) that 6% partial melting of the mantle wedge, at a fO2 = FMQ+1.7, is sufficient to dissolve completely the sulfide budget of the sub-arc mantle wedge, assuming a total mantle S content of 250 μg/g. The key to the Mungall (2002) model is that it calls upon Fe3+ as the oxidant of the mantle wedge owing to the recognition that slab si-mt is the only agent capable of carrying a high-redox potential into the mantle wedge. The presence of a slab si-mt in the sub-arc plumbing system, based on petrologic evidence, is manifest in arcs with adakitic, sodic-alkaline and potassic-ultrapotassic affinities. Mungall (2002) pointed out that sulfide resorption may be driven by the oxidation of sulfide via equilibria such as FeSliq + 2O2fluid = FeO melt + SO3melt
(28)
that can be combined with the stability of clinopyroxene and orthopyroxene FeSsulfide + CaFeSi 2 O6cpx + 2O2fluid = CaSOanhydrite + 2 FeSiO3opx 4
(29)
to describe the relationship between oxidation, sulfide resorption and sulfate stability. The coexistence of su-mt and anhydrite establish an oxygen buffer, which Mungall (2002) termed the sulfide-sulfur oxide (SO3) buffer (SSO). The value of fO2 at SSO and 800 °C is log fO2 = −11, which is approximately FMQ+2. Mungall (2002) concluded that partial melting of the mantle wedge, induced by the infiltration of slab si-mt and/or a H2O-silicate single-phase fluid, will promote the oxidative resorption of metal-rich sulfide in the wedge provided that the ambient oxidation state is >SSO. The resulting arc magma should contain evidence for the presence of slab si-mt or hydrous-silicate single-phase fluid, and also should contain an elevated metal ratio (i.e., high Cu, Au) relative to arc magma that is produced at fO2 < SSO. Thus, the nature of the oxidation state of S at the base of arc plumbing systems is as critical to the metallogenic potential as is S in the near-surface environment. It is interesting to note the increased “discovery” of magmatic sulfate in porphyry ore deposits (cf. Streck and Dilles 1998; Chambefort et al. 2008). We highlight that we do not intend to diminish the importance of sulfide; however, it is tempting to think that the finding of oxidized S in the causative magma may force us to reevaluate a possible sulfide paradigm, similar to that described above for the role of S vs. Cl in ore-metal transport in C-O-H-S fluids (Heinrich et al. 1999; Williams-Jones and Heinrich 2005). We also point out that there is no unique oxidation state for arc environments. The oxidation state of the subducted slab reflects the original oxidation state of the basaltic lithosphere, fO2 ~FMQ, and reaction with seawater during transit from the mid-ocean ridge to the subduction zone, and the subduction of terriginous and pelagic sediments (Wood et al. 1990; Mungall 2002; Kelley and Cottrell 2009; Rowe et al. 2009; Lee et al. 2010). As such, the ambient fO2 of the subducting slab will vary both across and within arcs and, thus, the ability of the downgoing material to effect redox changes in the mantle wedge must vary. Perhaps, it is the confluence
Magmatic Sulfur and Ore Deposit Formation
551
of this expected and observed variability, coupled with all processes related to the evolution of magmatic-hydrothermal systems, ranging from oxidation and partial melting of sulfides in the mantle wedge and lower crust, to anatexis and assimilation of metal-sulfides in the lower crust and subcontinental lithospheric mantle, to the evolution of fluid focusing pathways in the nearsurface magma, to an efficient precipitation mechanism(s), that leads to the selective formation of metal-sulfide anomalies in only a few locations in any one arc environment. For example, if the 20 Mt of S erupted at Mt. Pinatubo had not been vented to the atmosphere, but rather had precipitated as metal-sulfides in the sub-volcanic environment, this might have generated an ore deposit with several million tonnes Cu. Deposition of metal-sulfides in the porphyry environment The ability of a magmatic-hydrothermal system to form a porphyry ore deposit depends first on conditions in the magma chamber wherein efficient mass transfer of metals occurs from the si-mt to an exsolved aqueous fluid and second, on volumetrically concentrated and efficient mechanisms of precipitation in the overlying, sub-magmatic porphyry environment. Inefficient precipitation of metals from an aqueous fluid will result either in direct venting of the magmatically-sourced metals to the atmosphere or diffuse, hence uneconomic, precipitation in the magma plumbing system. Efficient metal precipitation can be driven by a combination of hydrolysis of SO2 in the fluid (Burnham 1967, 1979), throttling or rapid changes in pressure, i.e., from lithostatic to hydrostatic, that drives fluid unmixing (e.g., Butte; Rusk et al. 2008), and variation in the pressure-temperature-density paths experienced by ascending ore-forming fluid (e.g., Bingham; Redmond et al. 2004; Landtwing et al. 2010). An ascending ore fluid will contain both SO2 and H2S, the ratio of these two gas species being controlled by redox conditions in the generative magma. As the fluid ascends, SO2 in the volatile phase disproportionates to yield H2S and H2SO4 via the equilibrium 4SO2 + 4H 2O = H 2S + 3H 2SO 4
(30)
This hydrolysis reaction, driven to the right both by cooling and decompression, results in the production of H2SO4 and H2S that are present at a ratio of H2SO4/H2S ~3. The increase in the fH2S with continued disproportionation causes metal-sulfide precipitation and, importantly, increasing the fH2SO4 in conjunction with an increase in the activity of CaCl2 in the fluid, aCaCl2, via the reaction of the fluid with wall rock as described by the equilibria (CaAl2Si 2O8 )an + 2(KCl) fluid + 4(SiO2 )qtz = 2(KAlSi3O8 )or + (CaCl 2 ) fluid
(31)
(CaAl2Si 2O8 )an + (NaAlSi3O8 )ab + 2(HCl) fluid + 2(KCl) fluid
(32)
= (KAl3Si3O10 (OH 2 ))ms + 2(SiO2 )qtz + (CaCl 2 ) fluid + (NaCl) fluid
results in the precipitation of anhydrite via the equilibrium (CaCl2 ) fluid + (H 2SO 4 ) fluid = (CaSO 4 ) + 2(HCl) fluid
(33)
These equilibria offer one explanation for the ubiquitous presence of both metal-sulfides, precipitated owing to increases in the fH2S and continual reaction of H2S and metals contained in the fluid(s), and anhydrite in some ore-stage porphyry environments. Studies of porphyry ore deposits provide evidence that cooling, decompression, and fluid unmixing promote metal precipitation from magmatic-hydrothermal fluid(s). Rusk et al. (2008) documented the strong effect that decompression had on the above equilibria that control metal-sulfide precipitation. These authors performed a detailed study of fluid inclusions in samples from the Butte porphyry-Cu-Mo deposit and concluded that rapid decompression of a low salinity single-phase aqueous fluid (i.e., above the critical pressure in the NaClH2O system) promoted unmixing of the single-phase fluid into coexisting vapor and brine that, in combination with fluid-rock reaction and thermal decline, resulted in significant, and
552
Simon and Ripley
volumetrically-concentrated, precipitation of both Cu and Mo from the ore fluid. Redmond et al. (2004) and Landtwing et al. (2010) reported data that constrained ore-fluid evolution in the Bingham Canyon porphyry-Cu-Mo deposit. These latter two studies reported that the parental ore-fluid was a single-phase fluid that unmixed during decompression and cooling at depths of several hundred meters below the ore body. The newly formed immiscible vapor and brine both ascended and metals transport was dominated by the vapor phase, owing to a combination of a high vapor/brine ratio and efficient partitioning of ore metals into the vapor phase. The measured abundances of ore metals in all three fluid types throughout the vertical extent of the deposit and underlying pluton are consistent with fluid density (cf. Pokrovski et al. 2005, 2008; Williams-Jones and Heinrich 2005) having played a determinant role in controlling the depth of mineralization.
Ni-Cu-(PGE) DEPOSITS Magmatic Ni-Cu-(PGE) deposits occur within or adjacent to mafic to ultramafic igneous rocks in a number of tectonic settings. Magmatic PGE deposits are characterized by the presence of only disseminated sulfide minerals with the world’s richest deposits occurring in relatively thin layers or “reefs” within large layered intrusions. The characteristics and genesis of magmatic PGE deposits are reviewed in a separate section below. Excellent books on the subject of magmatic sulfide ores are available authored by Naldrett (1989, 2004), as well as a recent review edited by Li and Ripley (2009c). Journal articles focusing on reviews of Cu-Ni sulfide deposits include those by Lesher and Keays (2002), Naldrett (2005), Arndt et al. (2003, 2005), Barnes and Lightfoot (2005) and Eckstrand and Hulbert (2007).
Characteristics and classification of magmatic Cu-Ni-(PGE) deposits Magmatic Ni-Cu-(PGE) deposits are sources of metals mentioned in their name, along with by-product Co and Au. However, Ni is the primary economic metal in most of these deposits and its enrichment as sulfide is an important process in the understanding of the genesis of the deposits. Sulfur is particularly important because the principal ore minerals are sulfides (e.g., po, pentlandite, chalcopyrite) and the formation of immiscible su-mt is perhaps the key factor in the formation of the deposits. For this reason the solubility of S in mafic si-mt, discussed above, is an important variable; su-mt saturation must be attained in order for dense metal-bearing sumt to accumulate in favorable locations within a magmatic conduit/storage system. Texturally, the ores range from massive sulfide through semi-massive (net-textured) and disseminated (Fig. 11). Classification schemes that are based on the form and composition of host rocks normally include classes such as (e.g., Naldrett 2009) meteorite impact-related, flood basalts, high-MgO basalts, anorthosites-troctolites, komatiite, ferropicrite and Alaskan. In the case of the worldfamous deposits associated with the Sudbury Igneous Complex, a crustal si-mt produced as a result of bolide impact was the host magma from which immiscible su-mt segregated (e.g., Lightfoot et al. 2001). In all other cases mantle-derived magmas served as the host. Examples of komatiite-associated deposits include those at Kambalda and Mt. Keith, Australia (e.g., Lesher and Keays 2002), and those of the Raglan area in Quebec (e.g., Lesher 2007) and Thompson belt of Manitoba (Paktune 1984). Flood basalt-associated deposits include those of Noril’sk, Siberia (e.g., Naldrett et al. 1995; Czamanske et al. 1995), and those in the Midcontinent Rift System of Minnesota (e.g., Duluth Complex) and Michigan (e.g., Eagle) (see Hauck et al. 1997; Ripley et al. 2007; Ding et al. 2010). The ores of the Voisey’s Bay deposit in Labrador (e.g., Ryan 2000; Li and Naldrett 2000) are the best examples of those in the anorthosite-related class. Jinchuan in China (e.g., Chai and Naldrett 1992; Lehmann et al. 2007) represents ores associated with highMgO basalts whose tectonic setting is still debated. Both Jinchuan and Voisey’s Bay appear to be associated with either continental suturing or continental breakup, and hence may be
Magmatic Sulfur and Ore Deposit Formation
553
Figure 11. Textures of magmatic Ni-CuPGE deposits. (A) Massive sulfide with po, chalcopyrite (cp) and pentlandite (pn). (B) Semi-massive, net-textured sulfide. (C) Disseminated sulfide in peridotite. Samples are from the Eagle deposit, northern Michigan.
strongly related to rifting. Pechanga (e.g., Barnes et al. 2001) is the primary example of Ni-Cu Figure 11. sulfide mineralization associated with ferropicrites; the tectonic environment may be that of a rifted continental margin. Alaskan or Ural-Alaska type intrusions form in convergent settings above the mantle wedge. Most of these are sulfide-poor but may be the source of Pt found in nearby placer occurrences (more common in the Urals Pt-belt). The low sulfide abundance is generally attributed to elevated fO2 conditions in the source magma; however, with appropriate conditions (reviewed below) magmatic sulfide mineralization may form as witnessed by that present in the Turnagain (e.g., Nixon 1998; Scheel et al. 2009) and Duke Island (Thakurta et al. 2008a) complexes. Other deposits may also be related to convergent zone processes, such as those of Aquablanca, Spain (Pina et al. 2006), Phoenix and Selkirk, Botswana (Maier et al. 2008), St. Stephen and Moxie in the Appalachians (Thompson 1984; Paktune 1989) and Vammala in Finland (Peltonen 2003). Whether or not these deposits actually formed in the convergent zone environment or have been transported there remains a contentious issue. What is clear is that the world’s largest deposits are related to tectonic regimes (generally rift-related) where large volumes of mafic to ultramafic magma have been produced.
Resource and grade characteristics The resource potential of the magmatic sulfide deposits for Ni is variable; currently about one half of the world’s supply is produced from magmatic deposits and the remainder is contributed from Ni-rich laterites. According to Eckstrand and Hulbert (2007) there are 142 Ni-Cu-PGE deposits in the world that contain more than 100,000 tons of resources and/or production. There are only 51 Ni-Cu deposits or districts with greater than 10 million metric tons and 13 with greater than 100 million metric tons (Fig. 12). Grades of deposits generally vary between 0.7 and 3 percent Ni. Copper grades generally range between 0.2 and 2 percent. Ore tonnages range from a few hundred thousand to a few tens of millions (Fig. 12). Sudbury
554
Simon and Ripley
Figure 12. Map showing the distribution of some of the large magmatic Ni-Cu-PGE deposits worldwide. Figure is modified from Eckstrand and Hulbert (2007). Figure 12.
is characterized by 1645 million metric tons and Noril’sk by 1903 million metric tons of sulfide ore. The Duluth Complex of Minnesota contains ~8,000 million metric tons of potential sulfide-rich ore, but the low-grade (0.2 wt% Ni) has not been favorable for Ni-mining. Other current major Ni producers include Mt. Keith/Kambalda (Australia), Voisey’s Bay, Raglan and Thompson (Canada), Pechanga (Russia) and Jinchuan (China).
A general model for magmatic Ni-Cu ore genesis Although details from deposit to deposit may vary, there is general consensus with respect to the important processes that are necessary for the formation of magmatic Cu-Ni sulfide deposits. The ore-forming processes begin with the generation of a mantle-derived si-mt (except in the special case of Sudbury). As the magma ascends through the crust, interaction with a variety of rock types is possible. Prior to such interaction the mantle-derived magma is unlikely to have been saturated in sulfide or sulfate. For example, 25% partial melting of a mantle source with 250 ppm S will produce a magma with ~1000 ppm S. This amount is insufficient to cause the saturation of basaltic (or more mafic) magmas at temperatures above 1200 to 1400 °C (see above). In order for mafic magmas to reach saturation prior to a temperature where large amounts of olivine have crystallized and sequestered Ni, interaction with crust is necessary. Experimental studies have shown that the addition of siliceous material can cause a basaltic si mt to become saturated in sulfide (see summary in Li and Ripley 2009c). Oxidation of magma via assimilation of carbonate or H2O- rich country rocks can also lead to sulfide saturation. Cotectic amounts of sulfide are produced in this process and large amounts of magma are required for the generation of large sulfide ore bodies. It has been proposed for systems such as Jinchuan and Nebo-Babel where S-rich country rocks are rare, that sulfide saturation was promoted by such processes. The assimilation of S from country rocks is a process through with supersaturation of sulfide can occur; the partial assimilation of sulfidic country rocks has been a key process in the formation of many magmatic Ni-Cu deposits (see a recent evaluation by Keays and Lightfoot 2010). Figure 13 is based on the MELTS algorithm of Ghiorso and Sack (1995) and the equation for the S content of a basaltic si-mt at sulfide saturation by Li and Ripley (2009b). Fractional crystallization can drive a basaltic si mt to sulfide saturation as discussed by Mavrogenes and O’Neill (1999), but the depletion of Ni by olivine crystallization severely limits the potential development of Ni-rich sulfide ores. Assimilation of country rock sulfide will lead to an earlier saturation of the basaltic magma in sulfide. Another important
Magmatic Sulfur and Ore Deposit Formation
555
Figure 13
Figure 13. Sulfur content of the si-mt at sulfide saturation (SCSS) for a starting picritic magma appropriate for the Eagle deposit (Ding et al. 2010) as a function of magma crystallization. Phase boundaries are from Ghioroso and Sack (1995) and the SCSS curve is calculated from the equation of Li and Ripley (2009b). Path C represents fractional crystallization and the attainment of sulfide saturation at a time when Ni has been strongly sequestered by olivine. Path A refers to the early addition of sulfide leading to super saturation and the potential accumulation of large volumes of sulfide liquid. Path B represents the addition of externally derived sulfide, but in insufficient amounts to promote sulfide saturation. Continued fractional crystallization leads to early saturation of Ni-rich liquid but only cotectic proportions of sulfide are produced.
point centers on whether or not oversaturation in sulfide is promoted. Such a process could be caused by either dissolution of sulfide in country rock xenoliths or by melting of sulfides in country rocks. Should sulfide oversaturation occur, large amounts of su-mt may be produced. If this does not occur (e.g., path B on Fig, 13) cotectic proportions of Ni-rich su-mt are generated. Some disseminated-sulfide ore bodies associated with komatiites are proposed to have formed in such a manner (see Barnes, 2006, for a discussion). Massive sulfides could accumulate in embayments and zones of decreased magma velocity as a result of gravitational segregation of sulfide droplets and magma flowage. The huge Ni-Cu-(PGE) deposits at Noril’sk illustrate another key point in terms of S-assimilation. Stable isotopic data (e.g., Grinenko 1985; Li et al. 2003) clearly indicate that crustal S has been involved in the genesis of the ore bodies in the Noril’sk-Talnakh area. However, the elevated δ34S values could reflect the incorporation of reduced S from pelitic sediments and sour gas (Grinenko 1985), or the incorporation of S derived from evaporites in the surrounding country rock. If sulfate were assimilated then a reductant would be required. Both radiogenic isotope data (Arndt et al. 2003) and the presence of magmatic anhydrite in olivinebearing picrites (Li et al. 2009; Ripley et al. 2010) strongly suggest that evaporite assimilation (via dissolution mechanisms rather than melting) occurred. Jugo and Lesher (2005) proposed that carbonaceous country rocks served as the reductant. Ripley et al. (2010) pointed out the observation of locally elevated spinel concentrations in the picrites and the oxidation of Fe2+ coupled to the reduction of S6+. Carbonaceous sedimentary rock xenoliths or reduced volatile phases have been proposed as reductants for oxidized S in Alaskan-type complexes such as Duke Island (Thakurta et al. 2008a) and Turnagain (Nixon et al. 1998). Metals are enriched in su-mt due to their chalcophile nature and the Dxsu-mt/si-mt for Ni and Cu are in the 500-1000 range, whereas DPGEsu-mt/si-mt are considerably higher (Table 2;
556
Simon and Ripley
cf. Barnes and Lightfoot 2005). Because Ni is compatible in olivine, the importance of early sulfide saturation for the generation of Ni-rich sulfides is clear. Because Cu and PGE do not readily substitute in early crystallizing silicate minerals, later-stage saturation in sulfide could potentially lead to Cu- and PGE-rich ores. Although a few deposits exist that may be examples of such a process, in general they are rare. This supports the premise that basaltic magmas are able to assimilate country rock S only in early stages of crystallization when they possess sufficient heat to drive the process. Fractional crystallization of basaltic magmas that have not assimilated country rocks and contain only mantle-derived S can lead to the development of PGE- and Cu-rich layers (see below for PGE deposits), but the likelihood of large tonnage NiCu deposits being produced by this mechanism is small. In the case of Sudbury, the si-mt sheet was characterized by a temperature in excess of 1900 °C (Zieg and Marsh 2005), well above the silicate liquidus. For any reasonable amount of S in the si-mt sheet, sulfide saturation would have been attained well before the onset of pyroxene crystallization (Li and Ripley 2005). Because of this set of circumstances, immiscible su-mt had a much longer time to sequester metals and gravitationally accumulate in basal embayments than would be possible in most other cases where super-liquidus temperatures were not attained. The si-mt sheet must have contained crustal S as well as mantle-derived S from Proterozoic intrusive rocks in the target area, but assimilation of country rock S per se was not required. Ore bodies such as those at Noril’sk are exceptionally metal-rich and upgrading processes occurred in the conduit environment. Brugmann et al. (1993) proposed that uncontaminated, metal-bearing magma passing through the conduit interacted with accumulated sulfide. Because of the high Dxsu-mt/si-mt values for chalcophile elements, the sulfide sequestered metal via exchange reactions involving Fe2+ and became enriched. For this to occur, the new si mt must have been very near sulfide saturation. Kerr and Leitch (2005) showed that upgrading could occur in a conduit very effectively if the new si-mt was sulfide under-saturated. The passing si-mt would dissolve some of the accumulated sulfide but again due to the high Dxsu-mt/si-mt values, the remaining sulfide would become metal enriched. Li et al. (2009) extended this mass-balance treatment to the composition of the si-mt into which the sulfide was resorbed. Continued passage of S-undersaturated si mt and sulfide resorption results in a steadily increasing concentration of Cu, Ni and PGEs in the si-mt. PGE-rich sulfides may then be produced as the metal-rich, but still sulfide-undersaturated, si mt interacts with crustal rocks and attains sulfide saturation.
Source magmas for Ni-Cu deposits A question frequently asked regarding magmatic sulfide-rich Ni-Cu deposits is “what magmas are most favorable”. Arndt et al. (2005) provide a comprehensive review of answers to this question. There appear to be two principal factors. One is that most mafic magmas, no matter what the extent of mantle partial melting may have been, contain a sufficient amount of Ni and Cu to form economic ore bodies if sulfide saturation is attained early in the crystallization history. More important factors relate to the ability of the magma to interact with crustal rocks, and include temperature, viscosity and volatile content. Arndt et al. (2005) suggest that hightemperature and low-viscosity komatiites and picrites are most likely to assimilate S-bearing country rocks, and therefore have the greatest potential to produce large Cu-Ni-(PGE) sulfide deposits. Although higher degree partial melts of the mantle may have higher Ni contents due to the melting of olivine, the strongly chalcophile nature of both Ni and Cu assures that virtually any mantle-derived magma could produce a Ni-Cu sulfide accumulation.
Transport of sulfide melt The attainment of sulfide saturation in relatively deep-staging chambers implies that dense su-mt must be transported within its mafic host magma. Immiscible su-mt droplets in MORBs conclusively illustrate that basaltic magmas do erupt carrying immiscible su-mt droplets. Experimental studies by de Bremond d’Ars et al. (2001) have shown that immiscible sulfide
Magmatic Sulfur and Ore Deposit Formation
557
droplets can be carried by magma flowing at typical velocities. Accumulation occurs when magma velocity decreases as when conduits enlarge or when flow changes from vertical to horizontal (as is observed in cases such as Voisey’s Bay and Eagle). At Sudbury, the downward accumulation of immiscible su-mt occurred in a large si-mt sheet, but little flow was apparently involved. This was possible because of the crystal-free nature of the Sudbury si-mt sheet. Mungall and Su (2005) and Barnes et al. (2009) have shown that su-mt does not wet olivine crystals where si-mt is also present and su-mt is unlikely to migrate through pore spaces of olivine-rich cumulates. These findings also have ramifications with respect to the fate of sulfide minerals during mantle melting. Sulfide melts are most likely to form isolated droplets that would be difficult to entrain in partial melts. A sufficient volume of si-mt is required to dissolve all of the sulfide and generate S-undersaturated, relatively metal-rich si-mt (e.g., Keays 1995). This work also suggests that the presence of immiscible su-mt droplets entrained in mafic magmas indicates that interaction with crust resulted in the attainment of sulfide saturation in staging chambers.
PGE DEPOSITS IN LAYERED INTRUSIONS There are several types of magmatic and hydrothermal PGE deposits, but only a few are currently being mined. In general these deposits differ from magmatic Ni-Cu-(PGE) deposits in being much poorer in sulfide content. The world’s largest deposits are all located in Southern Africa (Fig. 12), with two layers of the Bushveld Complex (Merensky Reef and UG2 Chromitite) in South Africa in production, as well as deposits in the Great Dyke of Zimbabwe. The J-M Reef of the Stillwater Complex, USA, is another low-sulfide but high-PGE deposit currently being mined. Significant tonnages of PGEs are produced as by-products from high-sulfide NiCu mining at Noril’sk and also from the Sudbury Complex. The overwhelming predominance of production coming from the Bushveld Complex is clear. Several excellent reviews of PGE deposits and concentration mechanisms have been published in recent years, including a book on geology, geochemistry, mineralogy and beneficiation of PGEs edited by Cabri (2002). Additional review articles include those by Maier (2005), Cawthorn et al. (2005), Maier and Barnes (2005) Eckstrand and Hulbert (2007), and Naldrett et al. (2008).
Characteristics and classification of PGE deposits The PGE deposits in the Bushveld and Stillwater Complexes, as well as those in the Great Dyke, occur as relatively thin layers known as “reefs.” Although the reefs are typically low in sulfide abundance (<~3 vol%, Fig. 14), the PGEs are strongly associated with sulfide minerals and may occur as PGE-sulfide minerals such as braggite ((Pt,Pd)S) and cooperite (PtS), or as spatially associated tellurides, bismuthinides, arsenides or alloys. In some cases Pd may occur primarily in solid solution in pentlandite. For most PGE deposits S is thought to have played a primary role in the concentration and localization of the PGE. This is principally due to the chalcophile nature of the PGE and their strong affinity for sulfide at hydrothermal and magmatic temperatures. The DPGEsu-mt/si-mt are very high, typically in the range of 1,000 to 100,000 (Table 2). The reasons why PGE often combine with As, Sb and Te are currently under investigation (e.g., Helmy et al. 2010). It remains unclear if the PGE complex with these elements in su-mt or if low-temperature transformations control the final PGE mineral assemblages. Low-sulfide assemblages with Pt-Fe alloys may also be the result of desulfidation reactions (e.g., Naldrett and Lehmann 1988; Li and Ripley 2006) at subsolidus temperatures. Hydrothermal fluids may also play an important role in the final distribution of PGE and platinum-group minerals (Table 4) as further discussed below. Naldrett et al. (2008) have described several processes that may lead to the enrichment of PGE in a variety of rock types. Hydrothermal deposits such as those associated with the dunite
Simon and Ripley
558
Figure 14
Figure 14. Hand samples of the Merensky Reef (A) and the J-M Reef (B) illustrating the low percentages of sulfide in the ores.
pipes in the Bushveld Complex (e.g., Schiffries 1982) are not well understood and appear to be related to PGE transport via Cl-rich aqueous fluids. The PGE may also be enriched in organic rich sedimentary rocks either by hydrothermal processes (e.g., Sukhoi Log deposit in Siberia; Distler et al. 1996) or via low-temperature biologic mediation (e.g., Pasava 1991; Ripley et al. 2001). However, most of the world’s inventory of the PGE comes from deposits in mafic to ultramafic igneous rocks where magmatic (including potential late-state hydrothermal fluid activity) processes of PGE concentration have been operative. The classification schemes of Maier (2005), Maier and Barnes (2009), Naldrett et al. (2008) and Cawthorn et al. (2005) differ in approach, but all clearly emphasize the magmatic processes involved in the concentration of the PGE. One of the types of PGE occurrences where S has been an essential ingredient for PGE concentration is enrichment in Ni-Cu deposits. Deposits such as those at Noril’sk and Sudbury contain large quantities of the PGE that are recovered as by-products and that were collected by immiscible su-mt during magma evolution. These constitute what may be referred to as “sulfide-rich” occurrences where Ni and Cu are also strongly concentrated. The sulfide-rich PGE occurrences obviously significantly differ in their form when compared to the low-sulfide, but PGE-rich, occurrences found in layered intrusions. It is important to note that PGE enrichment may occur in a number of different rock types. For example, the sequence of rocks that defines
Magmatic Sulfur and Ore Deposit Formation
559
the Merensky Reef (anorthosite – chromitite - pegmatoidal feldspathic pyroxenite/norite - upper chromitite) is quite distinct from the package of rocks that define the J-M Reef of the Stillwater Complex (e.g., troctolite, gabbronorite, norite). The PGE-enriched sulfide mineralization of the J-M Reef varies in location with respect to the base of the host sequence, and is laterally transgressive (e.g., Zientek et al. 2002). The Merensky Reef package is also laterally variable, with the gap between the lower and upper chromitites increasing, or locally merging (Cawthorn et al. 2005). In contrast, the Main and Lower Sulfide Zones of the Great Dyke (e.g., Oberthur 2002; Wilson 2001) occur in pyroxenites where only the presence of sulfide distinguishes the PGE-enriched layers from surrounding rocks. The UG-2 chromitite of the Bushveld represents yet another relatively common host rock where the PGE are often concentrated.
Models for the genesis of PGE deposits in layered mafic intrusions There are two general models that have evolved for the generation of PGE-enriched horizons in layered intrusions. These have been developed for large deposits such as those in the Bushveld Complex, Great Dyke, and Stillwater Complex, and may not be appropriate for some of the smaller PGE occurrences. Because of the differences in the rock types that host the PGE occurrences in the large layered intrusions, how universal these models may be continues to spark debate. In brief, one of the models focuses on the importance of magma mixing and the formation of immiscible su-mt in concentrating the PGE. The other model involves the resorption of trace amounts of PGE-bearing su-xtal by late-stage magmatic fluids, rise of the fluid through a cumulate pile and concentration above the underlying PGE-depleted rocks. Additional details for each of these models are presented below. Before discussing the models for PGE accumulation in large layered intrusions, we will first briefly review how fractional crystallization has lead to the development of a PGE-enriched zone in a relatively small layered intrusion. The role of an immiscible su-mt is at least in theory very clear in this example, and with this background the differences exemplified by the PGEenriched layers in the large layered intrusions are more strongly elucidated. The Sonju Lake intrusion is part of the Beaver Bay Complex, a multiple intrusive suite of hypabyssal intrusions associated with the North American Midcontinent Rift System and associated volcanic rocks. The Sonju Lake intrusion is a sheet-like body with a thickness of ~1200 m and a strike length of 20 km. Miller and Ripley (1996) described the internal stratigraphy of the intrusion (Fig. 15), and concluded that crystallization occurred under nearly closed-system conditions with mineral accumulation at the floor of the magma chamber. Miller (1999) described the presence of a PGE-enriched layer near the center of the intrusion, within the oxide-rich gabbro (Fig. 15). Concentration peaks for Pd, Pt and Au coincide, but maximum values for Cu and S are found in overlying units (Fig. 15). Miller (1999) suggested that the PGE-enriched zone of the Sonju Lake intrusion formed as a result of fractional crystallization and the attainment of sulfide saturation in the magma chamber when the oxide-rich gabbro was crystallizing. Because of the strong partitioning of the PGE into su-mt, the cotectic amounts of sulfide would be enriched. Li and Ripley (2005) showed that fractional crystallization of a likely parental magma for the Sonju Lake intrusion would lead to sulfide saturation at a time when magnetite was on the liquidus. The capacity of the si-mt to dissolve sulfide would have been much reduced due to the crystallization of magnetite and depletion of Fe2+ in the si-mt (Fig. 16). The process of sulfide saturation and PGE enrichment should be expected for any mafic intrusion undergoing closed-system fractional crystallization. Park et al. (2004) concluded that the offset in both Cu and S peaks from those of Pt, Pd and Au was related to late-stage hydrothermal alteration, and preferential mobilization of Cu and S. In addition to the much larger nature of the Stillwater and Bushveld Complexes relative to the Sonju Lake intrusion, these intrusions show strong evidence for open-system evolution.
Figure 15.
Figure 15. Internal stratigraphy and Pt, Pd, Au, Cu and S distribution in the mafic rocks of the Sonju Lake intrusion, North America. Figure is modified from Miller (1999).
560 Simon and Ripley
Figure 16.
Magmatic Sulfur and Ore Deposit Formation
561
Figure 16. The S content of si-mt at sulfide saturation (SCSS) versus Zn concentration and % crystallization for the estimated parent tholeiitic mafic magma of the Sonju Lake intrusion (Table 5 of Miller and Ripley 1996). The crystallization sequence was computed by using MELTS (Ghioroso and Sack 1995) and the S content of the si-mt at SCSS by using equation 9 of Li and Ripley (2005.) For any reasonable estimate of the initial S content of the Sonju Lake parent magma, sulfide saturation will be reached after ~60% crystallization, when magnetite has joined plagioclase and clinopyroxene as a primocryst mineral. This prediction corresponds well to the observed peaks in Pt and Pd in the oxide-rich gabbro unit of the Sonju Lake intrusion shown in Figure 9. Figure is modified from Li and Ripley (2005).
The J-M Reef of the Stillwater Complex, and both the Merensky Reef and UG-2 chromitite of the Bushveld Complex occur above ultramafic sequences and at a level where plagioclase is 16 found as a cumulus mineral (Fig. 14). The development of both intrusions is often linked to the emplacement of two distinct magmas, one that gave rise to the lower, more ultramafic portions of the intrusions and another which produced the mafic rock types of overlying zones (Eales et al. 1990; Kruger 1994; Lambert et al. 1994; Todd et al. 1982; Irvine et al. 1983). The possible involvement of two magma types in the development of the Bushveld and Stillwater Complexes lead to the postulation that magma mixing caused the attainment of sulfide saturation in the mixed magma, and that turbulence in the chamber allowed the su-mt droplets to scavenge large amounts of the PGE (Campbell et al. 1983). Li and Ripley (2005) verified that mixing of more primitive and more evolved magmas could indeed lead to sulfide saturation (Fig. 17), with the potential for PGE concentration via an immiscible su-mt. Cawthorn (2005) proposed that pressure fluctuations, rather than magma mixing, lead to the attainment of sulfide saturation, and also the crystallization of chromite. Both of these models emphasize the important role of su-mt in the initial concentration of the PGE. Mondal and Mathez (2007) proposed that the UG-2 chromitite formed by the intrusion of new batches of magma carrying suspended chromite grains. The chromite-rich slurry would have also contained dispersed PGE-rich sulfides; the crystallization of chromite and sulfide saturation was thought to have occurred in a deeper staging chamber, perhaps in part related to contamination of the parent magma by country rocks. A second model for the genesis of PGE-enriched horizons in layered intrusions focuses on the role of magmatic-hydrothermal fluids in dissolving trace amounts of PGE-bearing sulfides.
Simon and Ripley
562
Figure 17.
Figure 17. (A) Sulfide content of mafic si-mt at sulfide saturation as a function of percent crystallization for a potential parent tholeiitic mafic magma involved in the formation of the Merensky Reef. The crystallization sequence was determined by using MELTS (Ghioroso and Sack 1995) at 2 kbar, QFM, and 1 wt% H2O in the initial magma. Path E-F shows the attainment of sulfide saturation at less than 10% crystallization. Point C represents the evolved si-mt where the S content has17 dropped to just above 400 ppm. This is the magma that is proposed to undergo mixing with introduced magma as illustrated in (B) below. (B) The SCSS of hybrid magma produced by mixing of basaltic magmas, one sulfide saturated with only ~400 ppm S (point A) and others less evolved with higher S contents. The fraction of tholeiitic basalt added is shown on the X-axis. Of importance is the fact that sulfide oversaturation may be produced as a result of mixing a magma at sulfide saturation with a S-under-saturated magma. In this example, which may be applicable to the Merensky and J-M Reefs, the addition of a sulfide-under-saturated magma containing ~900 ppm S in the amount of 10 to 50% could lead to a mixed magma that is sulfide-oversaturated. Figure is modified from Li and Ripley (2005).
Magmatic Sulfur and Ore Deposit Formation
563
The presence of saline fluids in both the Bushveld and Stillwater Complexes has been verified by the presence of fluid inclusions in the interstitial quartz of the Merensky Reef (Ballhaus and Stumpfl 1986) and in pegmatoidal minerals in the Stillwater Complex (Hanley et al. 2008). Boudreau (2009 and references therein) has proposed a sulfide resorption and chromatographic model to explain the formation of PGE reefs in layered intrusions. The model is based on the attainment of fluid saturation in the interstices of mafic cumulates and the resorption of trace amounts of sulfide minerals by this fluid. As fluid rises, additional quantities of the PGE and S are transported upwards. When the PGE- and S-rich fluid reaches the top of the cumulate pile, an immiscible su-mt forms that is PGE enriched. Although the role of S in the final concentration of the PGE is clear in both models, the obvious distinction is that the PGE are sequestered from underlying cumulates in the hydromagmatic model and from overlying and convecting magma in the magmatic model.
Source magmas for PGE deposits The example of the PGE-enrichment in the Sonju Lake intrusion illustrates that most mafic magmas contain sufficient amounts of the PGE to form a potentially economic ore body. For most mantle-derived magmas, the concentration of each of the PGE is less than 15ppt (Arndt et al. 2005). Concentration mechanisms, most of which involve S, are the principal keys for deposit formation. It has been proposed that high-degree mantle si-mt (e.g., komatiites, picrites) are favorable for PGE deposits because all of the sulfide in the mantle can be dissolved. Because the solubility of S in mantle si-mt at relatively high pressures is in the range of 500 to 1000 ppm (e.g., Mavrogenes and O’Neill 1999), and because the mantle contains ~200 to 250 ppm S, low-degree and reduced mantle si-mt cannot dissolve all of the S present (Keays 1995). For this reason, PGE-rich sulfides will remain in the mantle residue and produced si-mt should have less potential to form PGE deposits than high-degree mantle si-mt. Mungall (2002), Thakurta et al. (2008b), and Jugo (2009) have stressed that mantle si-mt which form under oxidized conditions where sulfate is stable may be enriched in the PGE at relatively low degrees of mantle melting. Jugo (2009) has shown that only ~6% mantle melting is required to eliminate PGE-bearing mantle sulfide at fO2 of FMQ + 1.7 (appropriate for many arc and back-arc environments). In the cases of the Bushveld Complex, Great Dyke and Stillwater Complexes, at least one high-MgO and PGE-enriched si-mt is assumed to have been involved in their genesis. This may be of more significance if mixing of magmas promoted sulfide saturation and was the principal process involved in PGE concentration. If hydrothermal processes were the main concentration mechanisms, then the nature of the parent in terms of PGE concentration may be of less significance.
FUTURE RESEARCH: WHAT DO WE NEED? Much progress has been made to constrain the behavior of S and associated ore metals in S-rich magmatic and magmatic-hydrothermal ore deposits. High-spatial resolution analytical techniques such as EPMA, LA-ICP-MS and SIMS now allow us to quantify S and ore metal abundances down to concentrations in the single ng/g range in natural fl, si-mt, and su-mt inclusions associated with ore stage mineralization. We know that S can complex with ore metals in su-mt and su-xtals and there is evidence that S also plays a role in controlling the solubility of ore metals in si-mt. We know that S complexes with and transports ore metals in H-O-S-Cl-bearing magmatic-hydrothermal fluid(s), and there is evidence that both sulfate(e.g., Yanacocha) and sulfide-saturated si-mt exsolve ore-metal-rich H-O-C-S fluid(s). In order to interpret the natural data, it is imperative to have experimental data at fixed P, T, and X over the full range of ore deposit formation. The currently available experimental data are simply too few. In some case the experiments were poorly constrained or may not have achieved equilibrium, and in other cases they were performed in systems that do not accurately reproduce
564
Simon and Ripley
natural assemblages. For example, in the case of porphyry deposits, there has been a small amount of experimental work performed at the magmatic conditions attending the evolution of the parental ore fluid; however, most experimental data are derived from felsic si-mt systems and not the intermediate to mafic magmas also demonstrated to play a role in the evolution of these deposits. We need experimental data from mafic to intermediate assemblages where the fluid(s) chemistry reflects accurately the complexity of natural fluids. Many groups (cf. Webster and Botcharnikov 2011, this volume) are performing experiments in the right direction to constrain the behavior of S during the evolution of magmatic-hydrothermal systems. These types of experiments are technically challenging, but in light of the significant progress in analytical techniques developed for the determination of S (Ripley et al. 2011, this volume), the challenge is now to the experimentalists to investigate, at PTX appropriate for high-temperature magmatic and magmatic-hydrothermal ore deposits, the isolated and synergistic effects of S, Cl, and CO2, on the behavior of ore metals.
ACKNOWLEDGMENTS We sincerely appreciate the detailed reviews by Steve Kesler, Ed Mathez, Jim Webster and Harald Behrens. Some research cited in this work was supported by NSF EAR 0609550 to ACS, and additional funding to A.C.S from the UNLV High Pressure Science and Engineering Center (HiPSEC) via the U.S. Department of Energy Cooperative Agreement Nos. DE-FC0801NV14049 and DE-FC8806NA27684.
REFERENCES Ahmad SN, Rose AW (1980) Fluid inclusions in porphyry and skarn ore at Santa Rita, New Mexico. Econ Geol 75(2):229-250 Akinfiev NN, Zotov AV (2001) Thermodynamic description of chloride, hydrosulfide, and hydroxo complexes of Ag(I), Cu(I), and Au(I) at temperatures of 25-500 °C and pressures of 1-2000 bar. Geochem Int 39:990106 Albarède F (2004) The stable isotope geochemistry of copper and zinc. Rev Mineral Geochem 55(1):409-427 Anderko A, Pitzer KS (1993a) Equation-of-state representation of phase equilibria and volumetric properties of the system NaCl-H2O above 573 K. Geochim Cosmochim Acta 57:1657-1680 Anderko A, Pitzer KS (1993b) Phase equilibria and volumetric properties of the systems KCl-H2O and NaClKCl-H2O above 573 K: equation of state representation. Geochim Cosmochim Acta 57:4885-4897 Archibald SM, Migdisov AA, Williams-Jones AE (2002) An experimental study of the stability of copper chloride complexes in water vapor at elevated temperatures and pressures. Geochim Cosmochim Acta 66(9):1611-1619 Arndt NT, Czamanske GK, Walker RJ, Chauvel C, Fedorenko VA (2003) Geochemistry and origin of the intrusive hosts of the Noril’sk and Talnakh intrusions, Siberia: implications for ore-forming processes in dynamic conduits. Econ Geol 98:495-515 Arndt NT, Lesher CM, Czamanske GK (2005) Mantle-derived magmas and magmatic Ni-Cu-(PGE) deposits. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 5-23 Arribas A Jr. (1995) Characteristics of high-sulfidation epithermal deposits, and their relation to magmatic fluid. In: Magmas, Fluids, and Ore Deposits. Thompson, JFH (ed) Min Assoc Canada, Short Course Vol 23:419454 Asael D, Matthews A, Bar-Matthews M, Halicz L (2007) Copper isotope fractionation in sedimentary copper mineralization (Timna Valley, Israel). Chem Geol 243(3-4):238-254 Audétat A, Günther D, Heinrich CA (1998) Formation of a magmatic hydrothermal ore deposit; insights in with LA-ICP-MS analysis of fluid inclusions. Science 279:2091-2094 Audétat A, Günther D, Heinrich CA (2000a) Magmatic-hydrothermal evolution in a fractionating granite: A microchemical study of the Sn-W-F-mineralized Mole Granite (Australia). Geochim Cosmochim Acta 64:3373-3393 Audétat A, Günther D, Heinrich CA (2000b) Causes for large-scale metal zonation around mineralized plutons: Fluid inclusion LA-ICP-MS evidence from the Mole Granite, Australia. Econ Geol 95:1563-1581
Magmatic Sulfur and Ore Deposit Formation
565
Audétat A, Pettke T, Dolejs D (2004) Magmatic anhydrite and calcite in the ore-forming quartz-monzodiorite magma at Santa Rita, New Mexico (USA): genetic constraints on porphyry-Cu mineralisation. Lithos 72:147-161 Audétat A, Pettke T, Heinrich C, Bodnar RJ (2008) The composition of magmatic-hydrothermal fluids in barren and mineralized intrusions. Econ Geol 103(5):877-908 Baker DR, Moretti R (2011) Modeling the solubility of sulfur in magmas: a 50-year old geochemical challenge. Rev Mineral Geochem 73:167-213 Baker T, Van Achterberg E, Ryan CG, Lang JR (2004) Composition and evolution of ore fluids in a magmatichydrothermal skarn deposit. Geology 32(2):117-120 Ballhaus C, Ryan CG, Mernagh TP, Green DH (1994) The partitioning of Fe, Ni, Cu, Pt, and Au between sulfide, metal and fluid phases: A pilot study. Geochim Cosmochim Acta 58:811-826 Ballhaus CG, Stumpfl EF (1986) Sulfide and platinum mineralization in the Merensky Reef: evidence from hydrous silicates and fluid inclusions. Contrib Mineral Petrol 94:193-204 Baranova, NN, Zotov AV (1998) Stability of gold sulphide species (AuHS0(aq)) and Au(HS)−2 at 300, 350 °C and 500 bar: Experimental study. Mineral Mag 62A:116-117 Barnes SJ (2006) Cotectic precipitation of olivine and sulfide liquid from komatiite magma and the origin of komatiite-hosted disseminated nickel sulfide mineralization at Mount Keith and Yakabindie, Western Australia. Econ Geol 102:299-304 Barnes SJ, Fiorentini ML, Austin P, Gessner K, Hough R, Squelch A (2009) Three-dimensional morphology of magmatic sulfides sheds light on ore formation and sulfide melt migration. Geology 36:655-658 Barnes S-J, Lightfoot PC (2005) Formation of magmatic nickel sulfide deposits and processes affecting their copper and platinum group elements contents. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 179-213 Barnes S-J, Melchik V, Sokolov SV (2001) The composition and mode of formation of the Pechanga nickel deposits, Kola Peninsula, northwestern Russia. Can Mineral 39:447-472 Barth AP, Dorais MJ (2000) Magmatic anhydrite in granitic rocks: First occurrence and potential consequences. Am Mineral 85(3-4):430-435 Bell A, Simon A (2011) Evidence for the alteration of the Fe3+/ΣFe of silicate melt caused by the degassing of chlorine-bearing aqueous volatiles. Geology 39:499-502 Bell A, Simon A, Guillong M (2009) Experimental constraints on Pt, Pd, and Au partitioning in silicate meltsulfide-oxide-aqueous fluid systems at 800 °C, 150 MPa, and variable sulfur fugacity. Geochim Cosmochim Acta 73(19):5778-5792 Belton GR, Jordan AS (1965) The volatilization of molybdenum in the presence of water vapor. J Phys Chem 69:2065–2071 Benning LG, Seward TM (1996) Hydrosulphide complexing of Au(I) in hydrothermal solutions from 150 to 400 C and 500 to 1500 bars. Geochim Cosmochim Acta 60:1849-1871 Bernard A (1985) Les mécanismes de condensation des gaz volcaniques – chimie minéralogic et équilibre des phases condenses majeures et mineures: Unpublished PhD thesis, Belgium, University of Brussels, 195 p Bernard A, Demaiffe D, Mattielli N, Punongbayan RS (1991) Anhydrite-bearing pumices from Mount Pinatubo – Further evidence for the existence of sulfur-rich silicic magmas. Nature 354:139-140 Bockrath C, Ballhaus C, Holzheid A (2004) Fractionationof the platinum-group elements during mantle melting. Science 304(5692):1951-1953 Bodnar RJ (1995) Fluid inclusion evidence for a magmatic source for metals in porphyry copper deposits. In: Magmas, Fluids and Ore Deposits. Thompson JFH (ed) Mineralogical Association of Canada Short Course, 23:139-152 Bodnar RJ (2003) Introduction to aqueous fluid systems. In: Fluid Inclusions: Analysis and Interpretation. Samson I, Anderson A, Marshall D (eds) Mineralogical Association of Canada Short Course, 32:81-99 Bodnar RJ, Burnham CW, Sterner SM (1985) Synthetic fluid inclusions in natural quartz. III. Determination of phase equilibrium properties in the system H2O-NaCl to 1000 C and 1500 bars. Geochim Cosmochim Acta 49:1861-1873 Boudreau A E, McCallum IS (1989) Investigations of the Stillwater Complex. Part V. Apatite as indicators of evolving fluid composition. Contrib Mineral Petrol 102:138-153 Boudreau A, Simon AC (2007) Crystallization and degassing in the basement sill, McMurdo Dry Valleys, Antarctica. J Petrol 48(7):1369-1386 Boudreau AE (2009) Transport of the platinum-group elements by igneous fluids in layered intrusions. In: New developments in magmatic Ni-Cu and PGE deposits. Li C, Ripley EM (eds) Geological Publishing Company, Beijing, p 229-249 Boudreau AE, Hoatson DM (2004) Halogen variations in the paleoproterozoic layered mafic-ultramafic intrusions of the East Kimberly, Western Australia: Implications for platinum-group element mineralization. Econ Geol 99:1015-1026 Boudreau AE, Mathez EA, McCallum IS (1986) Halogen geochemistry of the Stillwater and Bushveld Complexes: Evidence for transport of the platinum-group-elements by Cl-rich fluids. J Petrol 27:967-978
566
Simon and Ripley
Boudreau AE, McCallum IS (1992) Concentration of platinum-group elements by magmatic fluids in layered intrusions. Econ Geol 87:1830-1848 Boudreau AE, Meurer WP (1999) Chromatographic separation of the platinum-group elements, gold, base metals and sulfur during degassing of a compacting and solidifying crystal pile. Contrib Mineral Petrol134:174-185 Brenan JM (2008) Re-Os fractionation by sulfide-silicate partitioning: A new spin. Chem Geol 248:140-165 Brügmann GE, Naldrett AJ, Asif M, Lightfoot PC, Gorbachev NS, Fedorenko VA (1993) Siderophile and chalcophile metals as tracers of the evolution of the Siberian trap in the Noril’sk region, Russia. Geochim Cosmochim Acta 57:2001-2018 Burgisser A, Scaillet B (2007) Redox evolution of a degassing magma rising to the surface. Nature 445:194-197 Burnham CW (1967) Hydrothermal fluids at the magmatic stage. In: Geochemistry of Hydrothermal Ore Deposits. Barnes HL (ed) Holt, Rinehart and Winston, New York, p 34-76 Burnham CW (1979) Magmas and hydrothermal fluids. In: Geochemistry of hydrothermal ore Deposits. Barnes HL (ed), New York, Wiley, p 71-136 Cabri LJ (2002) The geology, geochemistry, mineralogy and mineral beneficiation of platinum-group elements. Can Inst Mining Metall Petrol Special Volume 54, 852 pp Campbell IH, Naldrett AJ, Barnes SJ (1983) A model for the origin of platinum-rich sulfide horizons in the Bushveld and Stillwater Complexes. J Petrol 24:133-165 Candela PA (1991) Physics of aqueous phase exsolution in plutonic environments. Am Mineral 76:1081-1091 Candela PA (2003) Ore in the Earth’s crust. In: Treatise on Geochemistry. Volume 3: The Crust. Rudnick R (ed) Elsevier Science, p 411-431 Candela PA, Holland HD (1984) The partitioning of copper and molybdenum between silicate melts and aqueous fluids. Geochim Cosmochim Acta 48:373-380 Candela PA, Piccoli PM (1995) Model ore-metal partitioning from melts into vapor and vapor/brine mixtures. In: Magmas, Fluids, and Ore Deposits. Thompson JFH (ed) Mineralogical Association of Canada Short Course, 23:101-128 Candela PA, Piccoli PM (2005) Magmatic Processes in the Development of Porphyry Type Ore Systems. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 25-38 Carroll MR, Rutherford MJ (1985) Sulfide and sulfate saturation in hydrous silicate melts. J Geophys Res 90:C601-C612 Carroll MR, Rutherford MJ (1987) The stability of igneous anhydrite – Experimental results and implications for sulfur behavior in the 1982 El Chicon trachyandesite and other evolved magmas. J Petrol 28:781-801 Cawthorn RG (2005) Pressure fluctuations and the formation of the PGE-rich Merensky and chromitite reefs, Bushveld Complex. Miner Deposita 40:231-235 Cawthorn RG, Barnes SJ, Ballhaus C, Malitch KN (2005) Platinum group element, chromium, and vanadium deposits in mafic and ultramafic rocks. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 215-249 Chai G, Naldrett AJ (1992) Characteristics of Ni-Cu-PGE mineralization and genesis of the Jinchuan deposit, northwest China. Econ Geol 87:1475-1495 Chambefort I, Dilles JH, Kent AJR (2008) Anhydrite-bearing andesite and dacite as a source for sulfur in magmatic-hydrothermal mineral deposits. Geology 36(9):719-722 Chaplgyin IV, Mozgova NN, Magazina LO, Kuznetsova OY, Safonov YG, Bryzgalov IA, Makovicky E, BaličŽunič T (2005) Kudriavite, (Cd,PbBi2S4), a new mineral species from Kudriavy Volcano, Iturup Island, Kurile Arc, Russia. Can Mineral 43(2):695-701 Chou IC (1987) Phase relations in the system NaCl-KCl-H2O: III, Solubilities of halite in vapor-saturated liquids above 445 degrees C and redetermination of phase equilibrium properties in the system NaCl-H2O to 1000 degrees C and 1500 bars. Geochim Cosmochim Acta 51:1965-1975 Chou IM (1982) Phase Relations in the system NaCl-KCl-H2O. Part I: Differential thermal analysis of the NaClKCl liquidus at 1 atmosphere and 500, 1000, 1500, and 2000 bars. Geochim Cosmochim Acta 46:19571962 Chou I-M, Sterner SM, Pitzer KS (1992) Phase relations in the system NaCl-KCl-H2O: IV. Differential thermal analysis of the sylvite liquidus in the KCl-H2O binary, the liquidus in the NaCl-KCl-H2O ternary, and the solidus in the NaCl-KCl binary to 2 kb pressure, and a summary of experimental data for thermodynamicPTX analysis of solid-liquid equilibria at elevated P-T conditions. Geochim Cosmochim Acta 56:22812293 Clemente B, Scaillet B, Pichavant M (2004). The solubility of sulphur in rhyolitic melts. J Petrol 45:2171-2196 Clifford TN, Stumpfl EF, Burger AJ, MacCarthy TS, Rex DC (1981) Mineral-chemical and isotopic studies of Namaqualand granulies, South Africa: A Grenville analogue. Contrib Mineral Petrol 77:225-250
Magmatic Sulfur and Ore Deposit Formation
567
Cline JS (1995) Genesis of porphyry copper deposits: The behavior of water, chloride, and copper in crystallizing melts. In: Porphyry copper deposits of the American Cordillera, Pierce FW, Bolm JG (eds) Arizona Geological Society Digest 20:69-82 Cline JS, Bodnar RJ (1994) Direct evolution of a brine from a crystallizing silicic melt at the Questa, New Mexico, molybdenum deposit. Econ Geol 89:1780-1802 Cooke DR, Hollings P (2005) Giant porphyry deposits: Characteristics, distribution, and tectonic controls. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 801-818 Core DP, Kesler SE, Essene EJ (2006) Unusually Cu-rich magmas associated with giant porphyry copper deposits: Evidence from Bingham, Utah. Geology 34(1):41-44 Crerar D, Barnes H (1976) Ore solution chemistry V. Solubilities of chalcopyrite and chalcocite assemblages in hydrothermal solutions at 200 °C to 350 °C. Econ Geol 71:772-794 Criss RE (1999) Principles of Stable Isotope Distribution. Oxford University Press Inc, New York Crocket JH, Fleet ME, Stone WE (1992) Experimental partitioning of osmium, iridium and gold between basalt melt and sulphide liquid at 1300 °C. Aust J Earth Sci 39:427-432 Crocket JH, Fleet ME, Stone WE (1997) Implications of composition for experimental partitioning of platinumgroup elements and gold between sulfide liquid and basaltic melt: the significance of nickel content. Geochim Cosmochim Acta 61:4139-4149 Czamanske GK, Zenko GK, Fedorenko RE, Calk VA, Budahn JR, Bullock JH, Fries TL, King BW, Siems DF (1995) Petrographic and geochemical characterization of ore-bearing intrusions of the Noril’sk type, Siberia: with discussion of their origin. Resour Geol Special Issue 18:1-48 Dadze TP, Akhmedzhanova GM, Kashirtseva GA, Orlov RY (2001) Solubility of gold in sulfide-containing aqueous solutions at T = 300 °C. J Mol Liq 91:99-102 de Bremond d’Ars J, Arndt NT, Hallot E (2001) Analog experimental insights into the formation of magmatic sulfide deposits. Earth Planet Sci Lett 186:371-381 Dilles JH (1987) Petrology of the Yerington Batholith, Nevada; evidence for evolution of porphyry copper ore fluids. Econ Geol 82(7):1750-1789 Ding X, Li C, Ripley EM, Rossell D, Kamo S (2010) The Eagle and East Eagle sulfide ore-bearing maficultramafic intrusions in the Midcontinent Rift System, upper Michigan: Geochronology and petrologic evolution. Geochem Geophys Geosyst 11: Q03003, doi:10.1029/2009GC002546 Distler VV, Mitrofanov GL, Nemerov VK, Kovalenker AV, Mokhov AV, Semeikina LK, Yudovskaya MA (1996) Mode of occurrence of the platinum group elements and their origin in the Sukhoi Log Gold Deposits, Russia. Geol Ore Deposits 38:413-328 Dromgoole EL, Pasteris JD (1987) Interpretation of sulfide assemblages in a suite of xenoliths from Kilbourne Hole, New Mexico. Spec Paper Geol Soc Am 215:25-26 Drummond SE, Ohmoto H (1985) Chemical evolution and mineral deposition in boiling hydrothermal systems. Econ Geol 80:126-147 Eales HV, deKlerk WJ, Teigler B (1990) Evidence for magma mixing processes within the Critical and Lower Zones of the Northwestern Bushveld Complex, South Africa. Chem Geol 88:261-278 Eastoe CJ (1978) A fluid inclusion study of the Panguna porphyry copper deposit, Bougainville, Papua New Guinea. Econ Geol 73:721-748 Eastoe CJ (1983) Sulfur isotope data and the nature of the hydrothermal systems at Panguna and Frieda porphyry copper deposits, Papua New Guinea. Econ Geol 78:201-213 Eckstrand OR, Hulbert LJ (2007) Magmatic nickel-copper-platinum group element deposits. In: Mineral Deposits of Canada: A Synthesis of Major Deposit Types, District Metallogeny, the Evolution of Geological Provinces and Exploration Methods. Goodfellow WD (ed) Geological Association of Canada, Mineral Deposits Division Special Publication, 5:205-222 Economou-Eliopoulos M (2005) Platinum group element potential of porphyry deposits. In: Exploration for Platinum-group Element Deposits. Mungall J (ed) Mineral Assoc Canada Short Course Series 35:203-246 Ehrlich S, Butler I, Halicz L, Rickard D, Oldroyd A, Matthews A (2004) Experimental study of the copper isotope fractionation between aqueous Cu (II) and covellite, CuS. Chem Geol 209:259-269 Elkins LJ, Fischer TP, Hilton DR, Sharp ZD, McKnight S, Walker J (2006) Tracing nitrogen in volcanic and geothermal volatiles from the Nicaraguan volcanic front. Geochim Cosmochim Acta 70:5215-5235 Emmons WH (1927) Relations of disseminated copper ores in porphyry to igneous intrusions. Am Inst Mining Metall Engineers Trans 75:797-809 Field CW (1966) Sulfur isotope abundance data, Bingham district, Utah. Econ Geol 61:850-871 Field CW, Gustafson LB (1976) Sulfur isotopes in the porphyry copper deposit at El Salvador, Chile. Econ Geol 71:1533-1548 Fleet ME (2006) Phase equilibria at high temperatures. Rev Mineral Geochem 61:365-419 Fleet ME, Crocket JH, Stone WE (1996) Partitioning of platinum-group elements (Os, Ir, Ru, Pt, Pd) and gold between sulfide liquid and basalt melt. Geochim Cosmochim Acta 60:2397-2412
568
Simon and Ripley
Fleet ME, MacRae ND (1987) Partition of Ni between olivine and sulfide: equilibria with the effect of temperature, fO2 and fS2. Contrib Mineral Petrol 95:336-342 Fleet ME, MacRae ND (1988) Partition of Ni between olivine and sulfide: equilibria with sulfide-oxide liquids. Contrib Mineral Petrol 100:462-469 Fleet ME, Stone WE (1991) Partitioning of platinum-group elements in the Fe-Ni-S system and their fractionation in nature. Geochim Cosmochim Acta 55(1):245-253 Fleet ME, Stone WE, Crocket JH (1991) Partitioning of palladium, iridium, and platinum between sulfide liquid and basalt melt. Effects of melt composition, concentration, and oxygen fugacity. Geochim Cosmochim Acta 55(9):2545-2554 Fleet ME, Wu TW (1993) Volatile transport of platinum-group elements in sulfide chloride assemblages at 1000 °C. Geochim Cosmochim Acta 57:3519-3531 Fleet ME, Wu TW (1995) Volatile transport of precious metals at 1000 °C: Speciation, fractionation, and effect of base-metal sulfide. Geochim Cosmochim Acta 59:487-495 Fournier RO (1987) Conceptual models of brine evolution in magmatic-hydrothermal systems: U.S. Geol Surv Prof Paper 1350:1487-1506 Frank MR, Candela PA, Piccoli PM, Glascock MD (2002) Gold solubility, speciation, and partitioning as a function of HCl in the brine-silicate melt-metallic gold system at 800 °C and 100 MPa. Geochim Cosmochim Acta 66:3719-3732 Frank MR, Simon AC, Pettke T, Candela PA, Piccoli PM (2011) Gold and copper partitioning in magmatichydrothermal systems at 800 °C and 100 MPa. Geochim Cosmochim Acta, doi:10.1016/j.gca.2011.02.012 Frei R (1995) Evolution of mineralizing fluid in the porphyry copper system of the Skouries deposit, northease Chalkidiki (Greece): evidence from combined Rb-Sr and stable isotope data. Econ Geol 90:746-762 Gammons CH, Barnes HL (1989) The solubility of Ag2S in near-neutral aqueous sulfide solutions at 25 to 300 °C. Geochim Cosmochim Acta 53:279-290 Gammons CH, Bloom MS (1993) Experimental investigation of the hydrothermal geochemistry of platinum and palladium: II. The solubility of PtS and PdS in aqueous ssulfide solutions to 300 °C. Geochim Cosmochim Acta 57:2451-2467 Gammons CH, Bloom MS, Yu Y (1992) Experimental investigation of the hydrothermal geochemistry of platinum and palladium: I. Solubility of platinum and palladium sulfide minerals in NaCl/H2SO4 solutions at 300 °C. Geochim Cosmochim Acta 56:3881-3894 Gammons CH, Williams-Jones AE (1995) Hydrothermal chemistry of electrum; thermodynamic constraints. Econ Geol 90(2):420-432 Gammons CH, Yu Y (1997) The stability of silver bromide and iodide complexes at 25 to 300 °C: Experiments, theory and geologic applications. Chem Geol 137:155-173 Gerlach TM (1980) Evaluation of volcanic gas analyses from Kilauea Volcano. J Volcanol Geotherm Res 7:295317 Ghiorso MS, Sack RO (1995) Chemical mass transfer in magmatic processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid-solid equilibrium in magmatic systems at elevated temperatures and pressures. Contrib Mineral Petrol 119:197-212 Gibert F, Pascal ML, Pichavant M (1998) Gold solubility and speciation in hydrothermal solutions: Experimental study of the stability of hydrosulphide complex of gold (AuHS0) at 350 to 450 °C and 500 bars. Geochim Cosmochim Acta 62:2931-2947 Giggenbach WF (1977) The isotopic composition of sulfur in sedimentary rocks bordering the Taupo Volcanic Zone: geochemistry. NZ Dept Sci Indus Research Bull 218:57-64 Giggenbach WF (1997) The origin and evolution of fluids in magmatic-hydrothermal systems: In: Geochemistry of Hydrothermal Ore Deposits, 3rd Edition. Barnes HL (ed) New York, John Wiley and Sons, p 737-796 Giggenbach WF, Matsuo S (1991) Evaluation of results from second and third IAVCEI field workshops on volcanic gases; Mt. Usu, Japan and White Island, New Zealand. Appl Geochem 6:125-141 Giggenbach WF, Sano Y, Wakita H (1993) Isotopic composition of helium, and CO2 and CH4 contents in gases produced along the New Zealand part of a convergent plate boundary: Geochim Cosmochim Acta 57:3427-3455 Glemser O, von Haeseler R (1962) Gaseous hydroxides. IV: gaseous hydroxides of molybdenum and tungsten. Z Anorg Allgem Chem 316:168-181 Grinenko LN (1985) Sources of sulfur of the nickeliferous and barren gabbro-dolerite intrusions of the northwest Siberian platform. Int Geol Rev 28:695-708 Guo J, Griffin WL, O’Reilly SY (1999) Geochemistry and origin of sulfide minerals in mantle xenoliths: Qilin, southeastern China. J Petrol 40:1125-1149 Gustafson LB (1979) Porphyry copper deposits and calc-alkaline volcanism, In: The Earth: Its Origin, Structure and Evolution. McElhinny MW (ed) London, Academic Press, p 427-468 Gustafson LB, Hunt JP (1975) The porphyry copper deposit at El Salvador, Chile. Econ Geol 70:857-912
Magmatic Sulfur and Ore Deposit Formation
569
Hack AC, Mavrogenes JA (2006) A synthetic fluid inclusion study of copper solubility in hydrothermal brines from 525 to 725 °C and 0.3 to 1.7 GPa. Geochim Cosmochim Acta 70:3970-3958 Hack AC, Thompson AB, Aerts M (2007) Phase relations involving hydrous silicate melts, aqueous fluids, and minerals. Rev Mineral Geochem 65:129-185 Halter WE, Heinrich CA, Pettke T (2005) Magma evolution and the formation of porphyry Cu-Au ore fluids: evidence from silicate and sulfide melt inclusions. Miner Deposita 39:845-863 Halter WE, Pettke T, Heinrich CA (2002) The origin of metal ratios in porphyry-type ore deposits. Science 296:1844-1846 Hamlyn PR, Keays RR (1986) Sulfur saturation and second stage melts: applications to the Bushveld platinum metal deposits. Econ Geol 81:1431-1445 Hanley JJ, Mungall JE, Pettke T, Spooner ETC, Bray CJ (2005) Ore metal redistribution by hydrocarbon-brine and hydrocarbon-halide melt phases, North Range footwall of the Sudbury Igneous Complex, Ontario, Canada. Miner Deposita 40:237-256 Hanley JJ, Mungall JE, Pettke T, Spooner ETC, Bray CJ (2008) Fluid and halide melt inclusions of magmatic origin in the Ultramafic and Lower Banded Series, Stillwater Complex, Montana, USA. J Petrol 49:11331160 Hattori K (1993) High-sulfur magma, a product of fluid discharge from underlying mafic magma: Evidence from Mount Pinatubo, Philippines. Geology 21:1083-1086 Hattori KH, Keith JD (2001) Contribution of mafic melt to porphyry copper mineralization: evidence from Mount Pinatubo, Philippines, and Bingham Canyon, Utah, USA. Miner Deposita 36:799-806 Hauck SA, Severson M, Zanko L, Morton P, Barnes SJ, Alminas H, Foord E, Dahlberg H (1997) An overview of the geology and oxide, sulfide, and platinum-group element mineralization along the western and northern contacts of the Duluth Complex. In: Middle Proterozoic to Cambrian rifting, central North America. Ojakangas RW, Dickas AB, Green JC (eds) Geol Soc Am Special Paper 312:137-186 Hayashi KI, Ohomoto H (1991) Solubility of gold in NaCl- and H2S-bearing aqueous solutions at 250-350 °C. Geochim Cosmochim Acta 55:2111-2126 Hedenquist JW, Aoki M, Shinohara H (1994) Flux of volatiles and ore-forming metals from the magmatichydrothermal system of Satsuma Iwojima volcano. Geology 22:585-588 Hedenquist JW, Arribas A, Reynolds TJ (1998) Evolution of an intrusion-centered hydrothermal system; Far Southeast-Lepanto porphyry and epithemal Cu-Au deposits, Philippines. Econ Geol 93(4):373-404 Hedenquist JW, Lowenstern JB (1994) The role of magmas in the formation of hydrothermal ore deposits. Nature 370:519-527 Heinrich CA (2005) The physical and chemical evolution of low-salinity magmatic fluids at the porphyry to epithermal transition: a thermodynamic study. Miner Deposita 39:864-889 Heinrich CA (2007) Fluid – fluid interactions in hydrothermal ore formation. Rev Mineral Geochem 65:363-387 Heinrich CA, Pettke T, Halter WE, Aigner-Torres M, Audétat A, Günther D, Hattendorf D, Bleiner D, Guillong M, Horn I (2003) Quantitative multi-element analysis of minerals, fluid and melt inclusions by LaserAblation Inductively-Coupled-Plasma Mass Spectrometry. Geochim Cosmochim Acta 67:3473-3497 Heinrich CA, Günther D, Audétat A, Ulrich T, Frischknecht R (1999) Metal fractionation between magmatic brine and vapor, determined by microanalysis of fluid inclusions. Geology 27:755-758 Heinrich CA, Guillong M, Pettke T, Pudack, Seo JH (2009) Low-salinity fluids at the porphyry-to-epithermal transition: from magmatic vapour to aqueous liquid. In: Proceedings of the Tenth Biennial SGA Meeting: Smart Science for Exploration and Mining (Society for Geology Applied to Mineral Deposits), p 14-16 Heinrich CA, Ryan CG, Mernagh TP, Eadington PJ (1992) Segregation of ore metals between magmatic brine and vapor. Econ Geol 87:1566-1583 Helmy HM, Ballhaus C, Wohlgemuth-Ueberwasser C, Fonseca ROC, Laurenz V (2010) Partitioning of Se, As, Sb, Te and Bi between monosulfide solid solution and sulfide melt – Application to magmatic sulfide deposits. Geochim Cosmochim Acta 21(1):6174-6179 Henley RW (1973) Solubility of gold in hydrothermal chloride solutions. Chem Geol 11:73–87 Henley RW (1991) Epithermal Gold Deposits in Volcanic Terranes. In: Gold Metallogeny and Exporation. Foster RP (ed) Blackie and Sons Ltd, Glasgow, p 133-164 Henley RW, McNabb A (1978) Magmatic vapor plumes and ground-water interaction in porphyry copper emplacement. Econ Geol 73:1-20 Hezarkhani A, Wiliams-Jones AE (1998) Controls of alteration and mineralization in the Sungun porphyry copper deposit, Iran: Evidence from fluid inclusions and stable isotopes. Econ Geol 93:651-670 Holland HD (1972) Granites, solutions, and base metal deposits. Econ Geol 67:281-301 Holzheid A, Grove TL (2002) Sulfide saturation limits in silicate melts and their implications to core formation scenarios for terrestrial planets. Am Mineral 87:227-237 Imai A (2000) Mineral paragenesis, fluid inclusions and sulfur isotope systematics of the Lepanto Far Southeast porphyry Cu-Au deposit, Mankayan, Philippines. Resour Geol 50:151-168
570
Simon and Ripley
Irvine TN, Keith DW, Todd SG (1983) The J-M platinum-palladium reef of the Stillwater Complex, Montana: II. Origin by double diffusive convective magma mixing and implications for the Bushveld Complex. Econ Geol 78:1287-1334 Jugo PJ (2009) Sulfur content at sulfide saturation in oxidized magmas. Geology 37:415-418 Jugo PJ, Candela PA, Piccoli PM (1999) Magmatic sulfides and Au:Cu ratios in porphyry deposits: an experimental study of copper and gold partitioning at 850 °C, 100 MPa in a haplogranitic melt-pyrrhotiteintermediate solid solution-gold metal assemblage, at gas saturation. In: Granites; Crustal Evolution and Associated Mineralization. Sial NA, Stephens WE, Ferreira VP (eds) Lithos 46:573-589 Jugo PJ, Lesher CM (2005) Redox changes caused by evaporite and carbon assimilation at Noril’sk and their role in sulfide precipitation. Geol Soc Am Abs Prog 39:360 Jugo PJ, Luth RW, Richards JP (2005a) An experimental study of the sulfur content in basaltic melts saturated with immiscible sulfide or sulfate liquids at 1300 °C and 1.0 GPa. J Petrol 46:783-798 Jugo PJ, Luth RW, Richards JP (2005b) Experimental data on the speciation of sulfur as a function of oxygen fugacity in basaltic melts. Geochim Cosmochim Acta 69:497-503 Keays RR (1995) The role of komatiitic and picritic magmatism and S-saturation in the formation of ore deposits. Lithos 34:1-18 Keays RR, Lightfoot PC (2010) Crustal sulfur is required to form magmatic Ni-Cu sulfide deposits: evidence from chalcophile element signatures of Siberian and Deccan Trap basalts. Miner Deposita 45:241-257 Keith JD, Whitney JA, Hattori K, Ballantyne GH, Christiansen EH, Barr DL, Cannan TM, Hook CJ (1997) The role of magmatic sulfides and mafic alkaline magmas in the Bingham and Tintic mining districts, Utah. J Petrol 38(12):1679-1690 Kelley KA, Cottrell E (2009) Water and the oxidation state of subduction zone magmas. Science 325:605-607 Keppler H (2010) The distribution of sulfur between haplogranitic melts and aqueous fluids. Geochim Cosmochim Acta 74:645-660 Kerr A, Leitch AM (2005) Self-destructive sulfide segregation systems and the formation of high-grade magmatic ore deposits. Econ Geol 100:311-332 Kerrich R, Goldfarb R, Groves DM, Garwin S (2000) The characteristics, origins, and geodynamic settings of supergiant gold metallogenic provinces. Sci China Ser D Earth Sci 43:1-68 Kesler SE, Jones LM, Walker RL (1975) Intrusive rocks associated with porphyry copper mineralization in island arc areas. Econ Geol 70:515-526 Kesler SE, Sutter JF, Issigonis MJ, Jones LM, Walker, RL (1977) Evolution of porphyry copper mineralization in an oceanic island arc, Panama. Econ Geol 72:1142-1153 Kirkham RV, Sinclair WD (1995) Porphyry copper, gold, molybdenum, tungsten, tin, silver. In: Geology of Canadian mineral deposit types, Eckstrand OR, Sinclair WD, Thorpe RI (eds) Geol Survey Canada, Geology of Canada No. 8:421-446 Kirkham RV, Sinclair WD (1996) Porphyry copper, gold, molybdenum, tungsten, tin, silver. In: Geology of Canadian mineral deposit types. Eckstrand OR, Sinclair WD, Thorpe RI (eds) Geological Survey of Canada, Geology of Canada No. 8, p 421-446 Kisters AFM, Gibson RL, Charlesworth EG, Anhaeusser CR (1998) The role of strain localization in the segregation and ascent of anatectic melts, Namaqualand, South Africa. J Struct Geol 20:229-242 Korzhinsky MA, Tkachenko SI, Shmulovich KI, Taran YA, Steinberg GS (1994) Discovery of a pure rhenium mineral at Kudriavy volcano. Nature 369:51-52 Kostova B, Pettke T, Driesner T, Petrov P, Heinrich CA (2004) LA-ICPMS study of fluid inclusions in quartz from the Yuzhna Petrovitsa deposit, Madan ore field, Bulgaria. Schweiz Mineral Petrograph Mitteilungen 84:25-36 Kozlov VK, Kuznetsov VN, Khodakovskiy IL (1983) The thermodynamic parameters of Ag2O(c) and silver(I) hydroxy complexes in aqueous solution at elevated temperatures. Geochem Int 20:137–149 Kruger FJ (1994) The Sr-isotope stratigraphy of the Western Bushveld Complex, South Africa. J Earth Sci 97:393-398 Lambert DD, Walker RJ, Morgan JW, Shirey SB, Carlson RW, Zientek ML, Lipin BR, Koski MS, Cooper RL. (1994) Re-Os and Sm-Nd isotope geochemistry of the Stillwater Complex, Montana: Implications for the petrogenesis of the J-M Reef. J Petrol 35:1717-1753 Landtwing MR, Furrer C, Redmond PB, Pettke T, Guillong M, Heinrich CA (2010) The Bingham Canyon Porphyry Cu-Mo-Au deposit. III. Zoned copper-gold ore deposition by magmatic vapor expansion. Econ Geol 105:91-118 Larson PB, Maher K, Ramos FC, Chang Z, Gaspar M, Meinert LD (2003) Copper isotope ratios in magmatic and hydrothermal ore-forming environments. Chem Geol 201(3-4):337-350 Le Guern F (1988) Ecoulements gazeux reactifs à hautes temperatures, mesures et modelisation. Unpublished PhD thesis, University of Paris, 314 p Lee CTA, Luffi P, Le Roux V, Dasgupta R, Albarede F, Leeman WP (2010) The redox state of arc mantle using Zn/Fe systematics. Nature 468:681-685
Magmatic Sulfur and Ore Deposit Formation
571
Lehmann J, Arndt N, Windley B, Zhou M-F, Wang CY, Harris C (2007) Field relationships and geochemical constraints on the emplacement of the Jinchuan intrusion and its Ni-Cu-PGE sulfide deposit, Gansu, China. Econ Geol 102:75-94 Lerchbaumer L, Audétat A (2009) Partitioning of Cu between vapor and brine – An experimental study based on LA-ICP-MS analysis of synthetic fluid inclusions. Goldschmidt Conference, Cologne, Germany, A744 Lesher CM (2007) Deposits in the Raglan Area, Cape Smith Belt, New Quebec. In: Mineral Deposits of Canada: A synthesis of Major Deposit-Types, District Metallogeny, the Evolution of Geological Provinces, and Exploration Methods. Goodfellow WD (ed) Geological Society of Canada Special Publication 5:351-386 Lesher CM, Keays RR (2002) Komatiite-associated Ni-Cu-PGE deposits: geology, mineralogy, geochemistry and genesis. In: The Geology, Geochemistry, Mineralogy and Mineral Beneficiation of Platinum Group Elements. Cabri LJ (ed) Can Inst Mining Metallurgy Petrol Special Vol 54:579-617 Li C, Naldrett AJ (2000) Melting reactions of gneissic inclusions with enclosing magma at Voisey’s Bay, Labrador, Canada: Implications with respect to ore genesis. Econ Geol 95:801-814 Li C, Ripley EM (2005) Empirical equations to predict the sulfur content of mafic magmas at sulfide saturation and applications to magmatic sulfide deposits. Miner Deposita 40:218-230 Li C, Ripley EM (2006) Formation of Pt-Fe alloy by desulfurization of Pt-Pd sulfide in the J-M Reef of the Stillwater Complex, Montana. Can Mineral 44:895-903 Li C, Ripley EM (2009a) A new genetic model for the giant Ni-Cu-PGE sulfide deposits associated with the Siberian flood basalts. Econ Geol 104:291-301 Li C, Ripley EM (2009b) Sulfur contents at sulfide-liquid and anhydrite saturation in silicate melts: Empirical equations and example applications. Econ Geol 104:405-412 Li C, Ripley EM (eds) (2009c) New Developments in Magmatic Ni-Cu and PGE deposits. Geological Publishing House, Beijing Li C, Ripley EM, Naldrett AJ (2003) Compositional variations of olivine and sulfur isotopes in the Noril’sk and Talnakh intrusions, Siberia: implications for ore-forming processes in dynamic conduits. Econ Geol 98:69-86 Li C, Ripley EM, Naldrett AJ, Schmitt AK, Moore CH (2009) Magmatic anhydrite-sulfide assemblages in the plumbing system of the Siberian Traps. Geology 37:259-262 Lightfoot PC, Keays RR, Doherty W (2001) Chemical evolution and origin of nickel sulfide mineralization in the Sudbury Igneous Complex, Ontario, Canada. Econ Geol 96:1855-1875 Linnen RL, Williams-Jones AE, Leitch CHB, Macauley TN (1995) Molybdenum mineralization in a fluorinepoor system: The Trout Lake stockwork deposit, southeastern British Columbia. In: Porphyry Deposits of the Northwestern Cordillera of North America, Schroeter TG (ed) Can Inst Mining Metall Petrol Special Publication 46:771-780 Liu Y, Samaha NT, Baker DR (2007) Sulfur concentration at sulfide saturation (SCSS) in magmatic silicate melts. Geochim Cosmochim Acta 71:1783-1799 Loucks RR, Mavrogenes JA (1999) Gold solubility in supercritcal hydrothermal brines measured in synthetic fluid inclusions. Science 284:2159-2163 Lowenstern JB (2001) Carbon dioxide in magmas and implications for hydrothermal systems. Miner Deposita 36:490-502 Lowentern JB, Mahood GA, Rivers ML, Sutton SR (1991) Evidence for extreme partitioning of copper into a magmatic vapor phase. Science 252:1405-1409 Luhr JF (1990) Experimental phase relations of water- and sulfur-saturated arc magmas and the 1982 eruptions of El Chichón Volcano. J Petrol 31:1071-1114 Luhr JF (2008) Primary igneous anhydrite: Progress since its recognition in the 1982 El Chichón trachyandesite. J Volcanol Geotherm Res 175:394-407 Luhr JF, Carmichael ISE, Varekamp JC (1984) The 1982 eruptions of El Chichon Volcano, Chiapas, Mexico; mineralogy and petrology of the anhydrite-bearing pumices. J Volc Geotherm Res 23:69-108 Lusk J, Bray DM (2002) Phase relations and the electrochemical determination of sulfur fugacity for selected reactions in the Cu-Fe-S and Fe-S systems at 1 bar and temperatures between 185 and 460 °C. Chem Geol 192:227-248 Lynton SJ, Candela PA, Piccoli PM (1993) An experimental study of the partitioning of copper between pyrrhotite and a high silicate rhyolitic melt. Econ Geol 88(4):901-915 Maclean WH, Shimazaki H (1976) The partition of Co, Ni, Cu, and Zn between sulfide and silicate liquids. Econ Geol 71:1049-1057 Maier WD (2000) Platinum-group elements in Cu-sulphide ores at Carolusberg and East Okiep, Namaqualand, South Africa. Miner Deposita 35:422-429 Maier WD (2005) Platinum-group element (PGE) deposits and occurrences: mineralization styles, genetic concepts, and exploration criteria. J African Earth Sci 41:165-191 Maier WD, Barnes SJ (1999) The origin of Cu sulphide deposits in the Curaca Valley, Bahia, Brazil: Evidence from Cu, Ni, Se and platinum-group element concentrations. Econ Geol 94:165-183
572
Simon and Ripley
Maier WD, Barnes S-J (2005) Formation of PGE deposits in layered intrusions. In: New Developments in Magmatic Ni-Cu and PGE Deposits. Li C, Ripley EM (eds) Geological Publishing Company, Beijing, p 250-276 Maier WD, Peltonen P, Livesey T (2008) The composition of magmatic Ni-Cu-(PGE) sulfide deposits in the Tati and Selebi-Phikwe belts of eastern Botswana. Miner Deposita 43:37-60 Maréchal C, Télouk P, Albarède F (1999) Precise analysis of copper and zinc isotopic compositions by plasmasource mass spectrometry. Chem Geol 156:251-273 Marini L, Moretti R, Accornero M (2011) Sulfur isotopes in magmatic-hydrothermal systems, melts, and magmas. Rev Mineral Geochem 73:423-492 Markl G, Lahaye Y, Schwinn G (2006) Copper isotopes as monitors of redox processes in hydrothermal mineralization. Geochim Cosmochim Acta 70(16):4215-4228 Mathur R, Ruiz J, Titley S, Liermann L, Buss H, Brantley S (2005) Cu isotopic fractionation in the supergene environment with and without bacteria. Geochim Cosmochim Acta 69(22):5233-5246 Mathur R, Titley S, Barra F, Brantley S, Wilson M, Phillips A, Munizaga F, Maksaev V, Vervoort J, Hart G (2009) Exploration potential of Cu isotope fractionation on porphyry copper deposits. J Geochem Explor 102:1-6 Matthews SJ, Jones AP, Gardeweb MC (1994) Lascar Volcano, Northern Chile; Evidence for steady-state disequilibrium. J Petrol 35(2):401-432 Matthews SJ, Sparks RSJ, Gardeweg MC (1999) The Piedras Grandes-Soncor eruptions, Lascar Volcano, Chile; evolution of a zoned magma chamber in the Central Andean upper crust. J Petrol 40(12):1891-1919 Mavrogenes JA, O’Neill HSt.C (1999) The relative effects of pressure, temperature and oxygen fugacity on the solubility of sulfide in mafic magmas. Geochim Cosmochim Acta 63:1173-1180 McInnes BIA, Gregoire M, Binns RA, Herzig PM, Hannington MD (2001) Hydrous metasomatism of oceanic sub-arc mantle, Lihir, Papua New Guinea: Petrology and geochemistry of fluid-metasomatized mantle wedge xenoliths. Earth Planet Sci Lett 188:169-183 McSween H (2010) Hell on Earth. Elements 6(2):67 Métrich N, Clocchiatti R (1996) Sulfur abundance and its speciation in oxidized alkaline melts. Geochim Cosmochim Acta 60:4151-4160 Métrich N, Wallace PJ (2008) Volatile abundances in basaltic magmas and their degassing paths tracked by melt inclusions. Rev Mineral Geochem 69:363-402 Meurer WP, Boudreau AE (1996) The petrology and mineral compositions of the Middle Banded series of the Stillwater Complex, Montana. J Petrol 37:583-607 Meurer WP, Willmore CC, Boudreau AE (1998) Metal redistribution during fluid exsolution and migration in the Middle Banded series of the Stillwater complex, Montana. Lithos 47:143-156 Migdisov Art A, Williams-Jones AE, Suleimenov OM (1999) Solubility of chlorargyrite (AgCl) in dry water vapor at elevated temperatures and pressures. Geochim Cosmochim Acta 63:3817-3827 Miller JD Jr (1999) Geochemical evaluation of platinum group element (PGE) mineralization in the Sonju Lake Intrusion, Finland, Minnesota. Minnesota Geol Surv Circ 44:1-32 Miller JD Jr, Ripley EM (1996) Layered intrusions of the Duluth Complex, Minnesota USA. In: Layered Intrusions. Cawthorn RG (ed) Elsevier p 257-301 Millner T, Neugenbauer J (1949) Volatility of the oxides of tungsten and molybdenum in the presence of water vapour. Nature 163:601-602 Mojzsis SJ, Harrison TM, Pidgeon RT (2001) Oxygen isotope evidence from ancient zircons for liquid water at Earth’s surface 4300 M year ago. Nature 409:178-181 Mondal SK, Mathez EA (2007) Origin of the UG2 chromitite layer, Bushveld Complex. J Petrol 48:495-510 Moretti R, Ottonello G (2005) Solubility and speciation of sulfur in silicate melts: The conjugaged Toop-SamisFlood-Grjotheim (CTSFG) model. Geochim Cosmochim Acta 69:801-823 Mungall JE (2002) Roasting the mantle: Slab melting and the genesis of major Au and Au-rich Cu deposits. Geol 30:915-918 Mungall JE, Su S (2005) Interfacial tension between magmatic sulfide and silicate liquids: Constraints on kinetics of sulfide liquation and sulfide migration through silicate rocks. Earth Plan Sci Lett 234:135-149 Mungall JE, Andrews DR, Cabri LJ, Sylvester PJ, Tubrett M (2005) Partitioning of Cu, Ni, Au, and platinumgroup elements between monosulfide solid solution and sulfide melt under controlled oxygen and sulfur fugacities. Geochim Cosmochim Acta 69:4349-4360 Muntean JL, Einaudi MT (2000) Porphyry gold deposits of the Refugio district, Maricunga belt, northern Chile: Econ Geol 95:1445-1472 Muntean JL, Einaudi MT (2001) Porphyry-epithermal transition: Maricunga Belt, Northern Chile. Econ Geol 96(4):743-772 Mutschler FE, Wright EG, Ludington S, Abbott JT (1981) Granite molybdenite systems. Econ Geol 76:874-897 Naboko SI (1964) Contemporary volcanoes and gas hydrothermal activity. Geologiya SSSR 31:323-387
Magmatic Sulfur and Ore Deposit Formation
573
Nagaseki H, Hayashi K-I (2008) Experimental study of the behavior of copper and zinc in a boiling hydrothermal system. Geology 36:27-30 Naldrett AJ (1989) Magmatic Sulfide Deposits. Clarendon Press – Oxford University Press, New York Naldrett AJ (2004) Magmatic Sulfide Deposits: Geology, Geochemistry and Exploration. Springer-Verlag, Heidelberg Naldrett AJ (2005) A history of our understanding of magmatic Ni-Cu sulfide deposits. Can Mineral 43:20692098 Naldrett AJ (2009) Fundamentals of magmatic sulfide deposits. In: New Developments in Magmatic Ni-Cu and PGE deposits. Li C, Ripley EM (eds) Geological Publishing Company, Beijing 1-26 Naldrett AJ, Fedorenko VA, Lightfoot PC, Kunilov NS, Gorbacher NS, Doherty W, Johan Z (1995) Ni-Cu-PGE deposits of Noril’sk region, Siberia: Their formation in conduits for flood basalt volcanism. Trans Inst Min Metal Section B 104:18-36 Naldrett AJ, Kinnaird J, Wilson A, Chunnett G (2008) Concentration of PGE in the Earth’s crust with special reference to the Bushveld Complex. Earth Sci Frontiers 15:264-297 Naldrett AJ, Lehmann J (1988) Spinel non-stoichiometry as the explanation for Ni-, Cu-, and PGE-enriched sulphides in chromitites. In: Geoplatinum ’87. Prichard H, Potts P, Bowles J (eds) London Elsevier p 93110 Newton RC, Manning CE (2005) Solubility of anhydrite, CaSO4, in NaCl-H2O solutions at high pressures and temperatures: applications to fluid-rock interaction. J Petrol 46:701-716 Nixon GT (1998) Ni-Cu sulfide mineralization in the Turnagain Alaskan-type Complex: A unique magmatic environment. British Columbia Geological Survey Fieldwork Report 1998-1, p 18-1–18-12. http://www. empr.gov.bc.ca/Mining/Geoscience/PublicationsCatalogue/Fieldwork/Documents/1997/nixon.pdf O’Neill H, Mavrogenes JA (2002) The sulphide capacity and the sulfur content at sulfide saturation of silicate melts at 1400 °C and 1 bar. J Petrol 43(6):1049-1087 Oberthur T (2002) Platinum-group element mineralization of the Great Dyke, Zimbabwe. In: The geology, geochemistry, mineralogy and mineral beneficiation of platinum-group elements. Cabri LJ (ed) Can Inst Min Metall Petrol Spec Vol 54:483-506 Ohmoto H, Rye RO (1979) Isotopes of sulfur and carbon. In: Geochemistry of Hydrothermal Ore Deposits, Second Edition. Barnes HL (ed) John Wiley & Sons, p 509-567 Oppenheimer C, Scaillet B, Martin RS (2011) Sulfur degassing from volcanoes: source conditions, surveillance, plume chemistry and earth system impacts. Rev Mineral Geochem 73:363-421 Oppenheimer C, Tsanev VI, Braban CF, Cox RA, Adams JW, Aiuppa A, Bobrowski N, Delmelle P, Barclay, J, McGonigle AJS (2006) BrO formation in volcanic plumes. Geochim Cosmochim Acta 70:2935-2941 Paktune AD (1984) Petrogenesis of ultramafic and mafic rocks of Thompson Nickel Belt, Manitoba. Contrib Mineral Petrol 88:348-353 Paktune AD (1989) Petrology of the St. Steven intrusion and the genesis of related nickel-copper sulfide deposits. Econ Geol 84:817-840 Pan P, Wood SA (1994) Solubility of Pt, Pd sulfides and Au in aqueous bisulfide solutions. I. Results at 200 °C to 350 °C and saturated vapor pressure. Miner Deposita 29:373-390 Parat F, Dungan MA, Streck MJ (2002) Anhydrite, pyrrhotite, and sulfur-rich apatite: Tracing the sulfur evolution of an Oligocene andesite (Eagle Mountain, CO, USA). Lithos 64:63-67 Parat F, Holtz F, Streck MJ (2011) Sulfur-bearing magmatic accessory minerals. Rev Mineral Geochem 73:285314 Park Y-R, Ripley EM, Miller JD Jr, Li C, Mariga J, Shafer P (2004) Stable isotopic constraints on fluid-rock interaction and Cu-PGE-S redistribution in the Sonju Lake Intrusion, Minnesota. Econ Geol 99:325-338 Pasava J (1991) Comparison between the distribution of PGE in black shales from the Bohemian Masif (CSFR) and other black shale occurrences. Miner Deposita 26:99-103 Pasteris JD (1996) Mount Pinatubo volcano and “negative” porphyry copper deposits. Geology 24(12):10751078 Peach CL, Mathez EA, Keays RR (1990 Sulfide melt-silicate melt distribution coefficients for noble metals and other chalcophile as deduced from MORB: Implicaations or partial melting. Geochim Cosmochim Acta 54:3379-3389 Peach CL, Mathez EA, Keays, RR, Reeves SL (1994) Experimentally determined sulfide melt – silicate melt partition coeffients for iridium and palladium. Chem Geol 117:361-377 Pearson RG (1963) Hard and Soft Acids and Bases. J Am Chem Soc 85:3533-3539 Peltonen P (2003) Svecofennian mafic-ultramafic intrusions. In: The Precambrian Bedrock of Finland – Key to the Evolution of the Fennoscadian Shield. Lehtinen M, Nurmi PA, Ramo OT (eds) Elsevier – Amsterdam, p 411-446 Peregoedova A, Barnes S-J, Baker DR (2006) An experimental study of mass transfer of platinum-group elements, gold, nickel and copper in sulfur-dominated vapor at magmatic temperatures. Chem Geol 235:59-75
574
Simon and Ripley
Pina RR, Lunar R, Ortega L, Gervilla F, Alapietti T, Martinez C (2006) Petrology and geochemistry of maficultramafic fragments from the Aquablanca Ni-Cu ore breccia, southwest Spain. Econ Geol 101:865-881 Pokrovski GS, Borisova AY, Harrichoury J-C (2008) The effect of sulfur on vapor-liquid fractionation of metals in hydrothermal systems. Earth Planet Sci Lett 266:345-362 Pokrovski GS, Roux J, Harrichoury J-C (2005) Fluid density control on vapor-liquid partitioning of metals in hydrothermal systems: Geology 33:657-660 Quisefit JP, Toutain JP, Bergametti G, Javoy M, Cheyet B, Person A (1989) Evolution versus cooling of gaseous volcanic emissions from Momotombo volcano, Nicaragua: Thermochemical model and observations. Geochim Cosmochim Acta 52:2591-2608 Raith JG, Prochaska W (1995) Tungsten deposits in the Wolfram Schist, Namaqualand, South Africa: Stratabound versus granite-related genetic concepts. Econ Geol 90:1934-1959 Redmond PB, Einaudi MT, Inan EE, Landtwing MR, Heinrich CA (2004) Copper deposition by fluid cooling in intrusion-centered systems; new insights from the Bingham porphyry ore deposit, Utah. Geology 32:217220 Rempel KU, Migdisov AA, Williams-Jones AE (2006) The solubility and speciation of molybdenum in water vapor at elevated temperatures and pressures: Implications for ore genesis. Geochim Cosmochim Acta 70:687-696 Richards (2003) Tectono-magmatic precursors for porphyry Cu-(Mo-Au) deposit formation. Econ Geol 98:1515-1533 Richards (2009) Postsubduction porphyry Cu-Au and epithermal Au deposits: products of remelting of subduction-modified lithosphere. Geology 37:247-250 Ripley EM, Brophy JG, Li CS (2002) Copper solubility in a basaltic melt and sulfide liquid/silicate melt partition coefficients of Cu and Fe. Geochim Cosmochim Acta 66:2791-2800 Ripley EM, Li C, Moore CH, Elswick ER, Maynard JB, Paul RL, Sylvester P, Seo JH, Shimizu N (2011) Analytical methods for sulfur determination in glasses, rocks, minerals and fluid inclusions. Rev Mineral Geochem 73:9-39 Ripley EM, Li C, Moore CH, Schmitt AK (2010) Micro-scale S isotope studies of the Kharaelakh intrusion, Noril’sk region, Siberia: Constraints on the genesis of coexisting anhydrite and sulfide minerals. Geochim Cosmochim Acta 74:634-644 Ripley EM, Park YR, Lambert DD, Frick LR (2001) Re-Os isotopic composition and PGE contents of Proterozoic carbonaceous argillites, Virginia Formation, northeastern Minnesota. Org Geochem 32:857-866 Ripley EM, Taib NI, Li C, Moore CH (2007) Chemical and mineralogical heterogeneity in the basal zone of the Partridge River Intrusion: implications for the origin of Cu-Ni sulfide mineralization in the Duluth Complex, Midcontinent Rift System. Contrib Mineral Petrol 154:35-54 Roedder (1972) The composition of fluid inclusions. In: Data of Geochemistry [6th ed.]. Fleisher M (ed) US Geol. Survey Prof Paper, pp. 440-J J, 164 p Roedder E (ed) (1984) Fluid Inclusions. Rev Mineral Geochem 12:1-644 Rowe MC, Kent AJR, Nielsen RL (2007) Determination of sulfur speciation and oxidation state of olivine hosted melt inclusions. Chem Geol 236:303-322 Rowe MC, Kent AJR, Nielsen RL (2009) Subduction influence on oxygen fugacity and trace and volatile elements in basalts across the Cascade Volcanic Arc. J Petrol 50(1):61-91 Rowins SM (2000) Reduced porphyry copper-gold deposits: A new variation on an old theme. Geology 28(6):491-494 Rusk B, Dilles J, Reed M (2008) Fluid inclusion evidence for magmatic-hydrothermal fluid evolution in the porphyry copper-molybdenum deposit at Butte, Montana. Econ Geol 103:307-334 Ryan B (2000) The Nain-Churchill boundary and the Nain Plutonic Suite: A regional perspective on the geologic setting of the Voisey’s Bay Ni-Cu-Co deposit. Econ Geol 95:703-724 Sassani DC, Shock EL (1990) Speciation and solubility of palladium in aqueous magmatic-hydrothermal solutions. Geology 18:925-928 Sassani DC, Shock EL (1998) Solubility and transport of platinum-group elements in supercritical fluids: Summary and estimates of thermodynamic properties for ruthenium, rhodium, palladium, and platinum solids, aqueous ions, and complexes to 1000 °C and 5 kbar. Geochim Cosmochim Acta 62:2643-2671 Sattari P, Brenan JM, Horn I, McDonough WF (2002) Experimental constraints on the sulfide- and chromitesilicate melt partitioning behavior of rhenium and platinum-group elements. Econ Geol 92:385-398 Scaillet B, Evans BW (1999) The June 15, 1991 eruption of Mount Pinatubo. I. Phase equilibria and preeruption P–T–fO2–fH2O conditions of the dacite magma. J Petrol 40:381-411 Scaillet B, Pichavant M (2005) A model of sulphur solubility for hydrous mafic melts: application to the determination of magmatic fluid compositions of Italian volcanoes. Ann Geophys 48:671-698 Scheel JE, Scoates JS (2009) Chromian spinel in the Turnagain Alaskan-type ultramafic intrusion, northern British Columbia, Canada. Can Mineral 47:63-80
Magmatic Sulfur and Ore Deposit Formation
575
Schiffries CM (1982) The petrogenesis of a platiniferous dunite pipe in the Bushveld Complex: infiltration metasomatism by a chloride solution. Econ Geol 77:1439-1453 Seal RR II (2006) Sulfur isotope geochemistry of sulfide minerals. Rev Mineral Geochem 61:633-677 Seedorff E, Dilles JH, Proffett JM Jr., Einaudi MT, Zurcher L, Stavast WJA, Johnson, DA, Barton MD (2005) Porphyry deposits: Characteristics and origin of hypogene features. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 251-298 Selby D, Nesbitt BE, Muehlenbachs K, Prochaska W (2000) Hydrothermal alteration and fluid chemistry of the Endako porphyry molybdenum deposit, British Columbia. Econ Geol 95:183-202 Seo JH, Guillong M, Heinrich CA (2009) The role of sulfur in the formation of magmatic-hydrothermal coppergold deposits. Earth Planet Sci Lett 282:323-328 Seward TM (1973) Thio complexes and the transport of gold in hydrothermal ore solutions, Geochim Cosmochim Acta 37:379-399 Seward TM (1976) The stability of chloride complexes of silver in hydrothermal solutions up to 350 C. Geochim Cosmochim Acta 40:1329-1341 Seward TM (1984) The transport and deposition of gold in hydrothermal systems. In: Gold’82. Foster RP (ed) Rotterdam, A. A. Balkema Pub p 165-181 Shinohara H, Hedenquist JW (1997) Constraints on magma degassing beneath the Far Southeast porphyry CuAu deposit, Philippines. J Petrol 38(12):1741-1752 Shinohara H, Iiyama JT, Matsuo S (1989) Partition of chlorine compounds between silicate melt and hydrothermal solutions: I. Partition of NaCl–KCl. Geoch Cosmochim Acta 53:2617-2630 Sillitoe RH (1972) A plate tectonic model for the origin of porphyry copper deposits. Econ Geol 67(2):184-197 Sillitoe RH (1973) The tops and bottoms of porphyry copper deposits. Econ Geol 68:799-815 Sillitoe RH (1980) Cauldron subsidence as a possible inhibitor of porphyry copper formation. Mining Geol Special Issue 8:85-93 Sillitoe RH (1993) Epithermal models: genetic types, geometric controls and shallow features. Geol Assoc Canada Spec Vol 40:403-417 Sillitoe RH (1997) Characteristics and controls of the largest porphyry copper-gold and epithermal gold deposits in the circum-Pacific region. Austral J Earth Sci 44:373-388 Sillitoe RH (1998) Major regional factors favoring large size, high hypogene grade, elevated gold content and supergene oxidation and enrichment of porphyry copper deposits. In: Porphyry and hydrothermal copper and gold deposis: A global perspective. Porter TM (ed) Australian Mineral Foundation, p 21-34 Sillitoe RH (2000) Gold-rich porphyry deposits: Descriptive and genetic models and their role in exploration and discovery. Rev Econ Geol 13:315-345 Sillitoe RH (2010) Porphyry copper systems. Econ Geol 105:3-41 Simmons SF, White NC, John DA (2005) Geological characteristics of epithermal precious metal and base metal deposits. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 451-484 Simon AC, Candela PA, Piccoli PM, Englander L (2008a) The effect of crystal – melt partitioning on the budgets of copper, gold, and silver. Am Mineral 93:1437-1448 Simon AC, Pettke T (2009) Platinum solubility and partitioning in a felsic melt – vapor – brine assemblage. Geochim Cosmochim Acta 73(12):438-454 Simon AC, Pettke T, Candela PA, Piccoli PM (2008b) The partitioning behavior of silver in a vapor – brine – rhyolite melt assemblage. Geochim Cosmochim Acta 72(6):1638-1659 Simon AC, Pettke T, Candela PA, Piccoli PM, Heinrich CA (2004) Magnetite solubility and iron transport in magmatic-hydrothermal environments. Geochim Cosmochim Acta 68:4905-4914 Simon AC, Pettke T, Candela PA, Piccoli PM, Heinrich CA (2005) Gold partitioning in melt-vapor-brine systems. Geochim Cosmochim Acta 69:3321-3335 Simon AC, Pettke T, Candela PA, Piccoli PM, Heinrich CA (2006) Copper partitioning in sulfur bearing magmatic systems. Geochim Cosmochim Acta 70:5583-5600 Simon AC, Pettke T, Candela PA, Piccoli PM, Heinrich CA (2007) The partitioning behavior of As and Au in a haplogranite - vapor at magmatic conditions in sulfur-free and sulfur-bearing systems. Geochim Cosmochim Acta 71:1764-1782 Sinclair WD (2007) Porphyry deposits. In: Mineral Deposits of Canada: A Synthesis of Major Deposit-Types, District Metallogeny, the Evolution of Geological Provinces, and Exploration Methods. Goodfellow WD (ed) Geol Assoc Can Mineral Dep Div, Spec Pub No. 5:223-243 Singer DA (1995) World class base and precious metal deposits – a quantitative analysis. Econ Geol 90:88-104 Singer DA, Cox DP (1986) Grade and tonnage model of porphyry Cu-Au. U.S. Geol Survey Bull 1693:110-114 Smith RW, Norman DI, Popp CJ (1980) Calculated solubility of molybddnite in hydrothermal solutions. Geological Society of America Abstracts with Programs 12:525 Sourirajan S, Kennedy GC (1962) The system NaCl-H2O at elevated temperatures and pressures. Am J Sci 260:115-141
576
Simon and Ripley
Spilsbury TW (1995) The Schaft creek copper–molybdenum–gold–silver porphyry deposit, northwestern British Columbia. In: Porphyry Deposits of the Northwestern Cordillera of North America. Can Inst Min Metall Petrol Spec 46:239-246 Spry PG, Paredes MM, Foster F, Truckle JS, Chadwick TH (1996) Evidence for a genetic link between goldsilver telluide and porphyry molybdenum mineralization at the Golden Sunlight deposit, Whitehall, Montana: fluid inclusion and stable isotope studies. Econ Geol 91:501-526 Stavast WJA, Keith JD, Christiansen EH, Dorais MJ, Tingey D, Larocque A, Evans N (2006) The fate of magmatic sulfides during intrusion or eruption, Bingham and Tintic Districts, Utah. Econ Geol 101:329345 Stefánsson A, Seward TM (2003a) The hydrolysis of gold(I) in aqueous solutions to 600 °C and 1500 bar. Geochim Cosmochim Acta 67:1677-1688 Stefánsson A, Seward TM (2003b) Stability of chloridogold(I) complexes in aqueous solutions from 300 to 600 °C and from 500 to 1800 bar. Geochim Cosmochim Acta 67:4559-4576 Stefánsson A, Seward TM (2003c) Experimental determination of the stability and stoichiometery of sulphide complexes of silver(I) in hydrothermal solutions to 400 C and 500 bar. Geochim Cosmochim Acta 67:1395-1413 Stefánsson A, Seward TM (2004) Gold(I) complexing in aqueous sulphide solutions to 500 °C and 500 bar. Geochim Cosmochim Acta 68:4121-4143 Stern CR, Funk JA, Skews MA (2007) Magmatic anhydrite in plutonic rocks at the El Teniente Cu-Mo deposit, Chile, and the role of sulfur- and copper-rich magmas in its formation. Econ Geol 102(7):1335-1344 Stimac J, Hickmott D (1994) Trace element partition coefficients for ilmenite, orthopyroxene, and pyrrhotite in rhyolite determined by micro-PIXE analysis. Chem Geol 117:313-330 Stix J, Gaonac’h H (2000) Gas, plume and thermal monitoring. In: Encyclopedia of Volcanoes. Sigurdsson H, Houghton B, Rymer H, Stix J, McNutt S (eds) Academic Press p 1141-1163 Stoiber RE, Rose WI (1974) Fumarole incrustations at active Central American volcanoes. Geoch Cosmochim Acta 38:495-510 Stone WE, Crocket JH, Fleet ME (1990) Partitioning of palladium, iridium, platinum, and gold between sulfide liquid and basalt melt at 1200 °C. Geochim Cosmochim Acta 54:2341-2344 Streck MJ, Dilles J (1998) Sulfur evolution of oxidized arc magmas recorded in apatite: Evidence from the Yerington porphyry copper batholith, Nevada. Geology 26:523-524 Symonds RB, Rose WI, Bluth GJ, Gerlach TM (1994) Volcanic-gas studies: Methods, results, and applications. Rev Mineral 30:1-66 Symonds RB, Rose WI, Reed MH, Lichte FE, Finnegan DL (1987) Volatilization, transport and sublimation of metallic and non-metallic elements in high temperature gases at Merapi Volcano, Indonesia. Geochim Cosmochim Acta 51:2083-2101 Taran YA, Bernard A, Gavilanes J-C, Africano F (2000) Native gold in mineral precpitates from high-temperature volcanic gases of Colima volcano, Mexico. App Geochem 15:337-346 Taran YA, Giggenbach WF (2003) Geochemistry of light hydrocarbons in subduction-related volcanic and hydrothermal fluids. In: Volcanic, Geothermal, and Ore-Forming Fluids: Rulers and Witnesses of Processes Within the Earth. Simmons SF, Graham IJ (eds) Soc Econ Geol Spec Pub, Littleton, Colorado 10:61-74 Taran YA, Hedenquist JF, Korzhinsky MA, Tkachenko SI, Shmulovich KI (1995) Geochemistry of magmatic gases from Kudryavy volcano, Iturup, Kurile islands. Geochim Cosmochim Acta 59:1749-1761 Tattich B, Candela P, Piccoli P, Bodnar RJ, Fedele L (2010) The effect of CO2 on copper partitioning in a felsic melt-vapor-brine assemblage. Geol Soc Am Abstracts with Programs, Vol. 42, No. 5, p. 47 Thakurta J, Ripley EM, Li C (2008a) Geochemical constraints on the origin of sulfide mineralization in the Duke Island Complex, southeastern Alaska. Geochem Geophys Geosys 9: doi 10.1029/2008GC001982 Thakurta J, Ripley EM, Li C (2008b) Pre-requisites for sulphide-poor PGE and sulphide-rich Cu-Ni-PGE mineralization in Alaskan-type complexes. J Geol Soc India 72:611-622 Theodore TG, Menzie WD (1984) Fluorine deficient porphyry molybdenum deposits in the Western North American cordillera. In: Proc. 6th Quad. IAGOD Symp, Tbilisi. Schweiz, Janelidze TV, Tvalchrelidze TV (eds) Stuttgart p 463-470 Thompson JFH (1984) Acadian synorogenic mafic intrusions in the Maine Appalachians. Am J Sci 284:462-483 Tingle T, Fenn P (1984) Transport and concentration of molybdenum in granite molybdenite systems: Effects of fluorine and sulfur. Geology 12:156-158 Titley SR, Beane RE (1981) Porphyry copper deposits: Park I. Geology settings, petrology, and tectogenesis. In: Economic Geology 75th Anniversary Volume p 214-269 Todd SG, Keith DW, LeRoy LW, Mann EL, Irvine TN (1982) The J-M platinum-palladium reef of the Stillwater Complex, Montana: I. stratigraphy and petrology. Econ Geol 77:1454-1480 Tomkins AG, Mavrogenes JA (2003) Generation of metal-rich felsic melts during crustal anatexis. Geology 31:765-768
Magmatic Sulfur and Ore Deposit Formation
577
Tosdal RM, Richards JP (2001) Magmatic and structural controls on the development of porphyry Cu±Mo±Au deposits. Revn Econ Geol 14:157-181 Ulrich T, Günther D, Heinrich CA (1999) Gold concentrations of magmatic brines and the metal budget of porphyry copper deposits. Nature 399:676-679 Ulrich T, Mavrogenes J (2008)An experimental study of the solubility of molybdenum in H2O and KCl-H(2)O solutions from 500 degrees C to 800 degrees C, and 150 to 300 MPa. Geochim Cosmochim Acta 72:23162330 Vanko DA, Bodnar RJ, Sterner SM (1988) Synthetic fluid inclusions: VIII. Vapor saturated halite solubility in part of the system NaCl-CaCl2-H2O, with applications to fluid inclusions from oceanic hydrothermal systems. Geochim Cosmochim Acta 52:2451-2456 Var’yash LN (1992) Cu(I) complexing in NaCl solutions at 300 and 350 °C. Geochem Int 29:84-92 Wallace P (2001) Volcanic SO2 emissions and the abundance and distribution of exsolved gas in magmas. In: Journal of Volcanology and Geothermal Research Special Issue, Magma Degassing Through Volcanoes: A Tribute to Werner Giggenbach. Allard P (ed) 108:85-106 Wallace P (2005) Volatiles in subduction zone magmas: concentrations and fluxes based on melt inclusion and volcanic gas data. J Volcan Geotherm Res 140:217-240 Wallace P, Anderson AT (1999) Volatiles in Magmas. In: Encyclopedia of Volcanoes. Sigurdsson H, Houghton B, Rymer H, Stix J, McNutt S (eds) Academic Press 149-170 Wallace PJ, Edmonds M (2011) The sulfur budget in magmas: evidence from melt inclusions, submarine glasses, and volcanic gas emissions. Rev Mineral Geochem 73:215-246 Wallace PJ, Gerlach TM (1994) Magmatic vapor source for sulfur dioxide released during volcanic eruptions: evidence from Mount Pinatubo. Science 265:497-499 Webster JD, Botcharnikov RE (2011) Distribution of sulfur between melt and fluid in S-O-H-C-Cl-bearing magmatic systems at shallow crustal pressures and temperatures. Rev Mineral Geochem 73:247-283 Webster JD, DeVivo B (2002) Experimental and modeled solubilities of chlorine in aluminosilicate melts, consequences of magma evolution, and implications for exsolution of hydrous chloride melt at Mt. Somma-Vesuvius, Italy. Am Mineral 87:1046-1061 Webster JD, Mandeville C (2007) Fluid immiscibility in volcanic environment. Rev Mineral Geochem 65:313362 Webster JD, Sintoni MF, De Vivo B (2009) The partitioning behavior of Cl and S in aqueous fluid- and salineliquid saturated phonolitic and trachytic melts at 200 MPa. Chem Geol 263:19-36 Webster JG (1986) The solubility of gold and silver in the system Au-Ag-S-O2-H2O at 25 °C and 1 atm. Geochim Cosmochim Acta 50:1837-1845 Weissberg BG (1970) Solubility of gold in hydrothermal alkaline sulphide solutions. Econ Geol 65:551-556 Westrich HR, Gerlach TM (1992) Magmatic gas source for the stratospheric SO2 cloud from the June 15, 1991 eruption of Mount Pinatubo. Geology 20:867-870 White NC (1991) High sulfidation epithermal gold deposits: characteristics and a Model for their origin; in hightemperature acid fluids and associated alteration and mineralization. Geol Surv Japan Report No. 277:9-20 White NC, Hedenquist JW (1990) Epithermal Environments and Styles of Mineralization: Variations and their Causes, and Guidelines for Exploration. In: Epithermal Gold Mineralization of the Circum-Pacific: Geology, Geochemistry, Origin and Exploration, II. Hedenquist JW, White NC, Siddeley G (eds) J Expl Geochem 36:445-474 Whitney JA (1975) Vapour generation in a quartz monzonite magma: a synthetic model with application to porphyry copper deposits. Econ Geol 70:346-358 Williams TJ, Candela PA, Piccoli PM (1995) The partitioning of copper between silicate melts and two-phase aqueous fluids: An experimental investigation at kbar, 800 °C and 0.5 kbar, 850 °C. Contrib Mineral Petrol 121:388-399 Williams-Jones AE, Heinrich CA (2005) Vapor transport of metals and the formation of magmatic-hydrothermal ore deposits. In: Economic Geology 100th Anniversary Volume. Hedenquist JW, Thompson JFH, Goldfarb RJ, Richards JP (eds) p 1287-1312 Williams-Jones AE, Migdisov AA, Archibald SM, Xiao ZF (2002) Vapor-transport of ore metals. Geochem Soc Spec Pub 7:279:305 Williams-Jones G, Rymer H (2000) Hazards of volcanic gases. In: Encyclopedia of Volcanoes. Sigurdsson H, Houghton B, Rymer H, Stix J, McNutt S (eds) Academic Press, p 997-1004 Willmore CC, Boudreau AE, Kruger FJ (2000) The halogen geochemistry of the Bushveld Complex, Republic of South Africa: Implications for chalcophile element distribution in the Lower and Critical zones. J Petrol 41:1517-1539 Willmore CC, Boudreau, AE, Spivack A, Kruger FJ (2002) Halogens of the Bushveld Complex, South Africa: d37Cl and Cl/F evidence for hydration melting of the source region in a back-arc setting. Chem Geol 182:503-511
578
Simon and Ripley
Wilson AH (2001) Compositional and lithologic controls on the PGE-bearing sulphide zones in the Selukwe Subchamber, Great Dyke: A combined equilibrium-Rayleigh fractionalization model. J Petrol 42:18451867 Wood BJ, Bryndzia LT, Johnson KE (1990) Mantle oxidation state and its relationship to tectonic environment and fluid speciation. Science 248:337-345 Wood SA (2002) The aqueous geochemistry of the platinum-group elements with applications to ore deposits. In: The Geology, Geochemistry, Mineralogy and Mineral Beneficiation of Platinum-Group Elements. Cabri LJ (ed) Can Inst Miner Metall Petrol Spec Vol 54:211-249 Wood SA, Crerar DA, Borcsik MP (1987) Solubility of the assemblage pyrite-pyrrhotite-magnetite-sphaleritegalena-gold-stibnite-bismuthinite-argentite-molybdenite in H2O-NaCl-CO2 solutions from 200 to 350 °C. Econ Geol 82:1867-1867 Xiao Z, Gammons CH, Williams-Jones AE (1998) Experimental study of copper(I) chloride complexing in hydrothermal solutions at 40 to 300 °C and saturated water vapor pressure. Geochim Cosmochim Acta 2:2949-2964 Xiong Y, Wood SA (2000) Experimental quantification of hydrothermal solubility of platinum-group elements with special reference to porphyry environments. Mineral Petrol 68(1-3):1-28 Zezin DY, Migdisov AA, Williams-Jones AE (2007) The solubility of gold in hydrogen sulfide gas: An experimental Study. Geochim Cosmochim Acta 71(12):3070-3081 Zhang L, Dilles JH, Field CW, Reed MH (1996) Initial stable isotopic results from the pre-main stage porphyry copper mineralization at Butte, Montana. Trans Am Geophys Union 77:F775 Zhu XX, O’Nions RK, Guo Y, Belshaw NS, Rickard D (2000) Determination of natural Cu-isotope variation by plasma-source mass spectrometry; implications for use as geochemical tracers. Chem Geol 163(1-4):139149 Zieg MH, Marsh BD (2005) The Sudbury Igneous Complex: Viscous emulsion differentiation of a superheated impact melt sheet. GSA Bulletin 117:1427-1450 Zientek ML, Cooper RW, Corson SR, Geraghty EP (2002) Platinum-group element mineralization in the Stillwater Complex Montana. In: The Geology, Geochemistry, Mineralogy, and Beneficiation of PlatinumGroup Elements. Cabri LJ (ed) Can Inst Miner Metall Petrol Spec Vol 54:459-481 Zotov AV, Baranova NN, Bannykh LN (1996) Solubility of the gold sulfides Au2S and AuAgS in solutions containing hydrogen sulfide at 25-80 °C and pressures of 1 and 500 bar. Geochem Intl 34:216-221 Zotov AV, Baranova NN, Dar’yina TG, Bannykh LM (1991) The solubility of gold in aqueous chloride fluids at 350-550 °C and 500-1500 atm: Thermodynamic parameters of AuCl2-(aq) up to 750 °C and 5000 atm. Geochem Intl 28:63-71