Water in nominally anhydrous minerals

REVIEWS in MINERALOGY & GEOCHEMISTRY 1855 Volume 62 WATER IN NOMINALLY ANHYDROUS MINERALS EDITORS Hans Keppler and J...

Author: Hans Keppler | J. R. Smyth

348 downloads 1132 Views 7MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

REVIEWS in MINERALOGY & GEOCHEMISTRY

1855

Volume 62

WATER IN NOMINALLY ANHYDROUS MINERALS EDITORS

Hans Keppler and Joseph R. Smyth

GEOCHEMICAL SOCIETY MINERALOGICAL SOCIETY OF AMERICA Series Editor: Jodi J. Rosso 2006

/SSN 1529-6466

Water itl

RiMG Volume 62 Nominally Anhydrous Min erals CONTENTS

1- 28

Anolyllcol Mclhod! for Mcosuring Woltr In Nomlnolly Anh)dro"' Mlntrols G~rwe H. RoJs.rnan

29·52

Th• Slrn
53·66

Slrnclnrnl Slndlt'S of Oil In Nomlnnlly Anhyclrnus Mlnernl< Uolng NMR

Simrm C. Kolin 67·83

AlornJ,ollc Models of Ofl Oef
85· 11 5

llydrngen In lllgh P,...ure SIU<1llt and Oxide Mln•ral Slruclnres

Knte Wright

Jos.ph R. Sm>·th 117· 1$4

Water In Nominally Anhydrous Crustal Mlntral~: Sv«bltlon. ConcentrutJon, and G.oiOjt)c Sl~tnlflcanct Elt
155-161

Woltrln Nalural MantleMintrolsl: P)roxtncs

169-191

Wottr In Nalural Manti• Mlntrols 0 : Olhlnt. Ga,...l and Actt!iSOry Minerals .Anl<m lktnlt and Ewgrn Ubt1Milll}:y

193·230

Thtrmodynamlcs or Wattr Solubility and Partlllonln• lions Krpt'lrrtmd NatlwUtr Boifon·Cosarw''O

231 -24 1

The l'artlllonlng or \Voter llelwt"en Nominally Anhydrnu< Mlnernls and Sllltau~ Melts Simon C. Koh, am/ Kevln J. GroJJt

243·27 1

The Stublllly or llydrnus Mantle rhOS<>

273·289

ll)'di'Ol<S l' hnse.< and

lltnnk Skogby

J)lmltl J. Frosr

Water Trnnsporl in the Subduelln~ Slnb 1(d,w hlko Kawomo1o )tmnfC'k lngti't a/Ill Marc Blllllcharrl

29 1-320

Oltruslnn of llydrogen in Mineral$

321·342

Etrtct of Woler on tht Equation or Slate or Nominally Anhydrous Minerals Sln'en D. )act)h$~n

343-375

Rtmott Stn.•ln& or llydrogen in Earth's Mantle

ShUJt-ichiro K.,oto

377-396

Tb< Role otWattr in Righ·Ttmperature Rock Otrormatlon

Dovid L. KohhtMI

397-420

Tb< I'.ITt<'l ofWaltr on ~lantle Phase TnonsitiOM

421-450

Water In lhellarly Earth

Bcnwrd Marry and Reiko Yokochi

451-473

Wtlttr and Ceodyn.amlcs

Klaus R~gena~r·Litb

I:Jjl Ohtanl and K. D. LitQSIJV

REVIEWS IN MINERALOGY AND GEOCHEMISTRY Volume 62

2006

WATER IN NOMINALLY ANHYDROUS MINERALS EDITORS

Hans Keppler

Joseph R. Smyth

Universität Bayreuth Bayreuth, Germany

University Colorado Boulder, Colorado

COVER PHOTOGRAPH: Thin section of a garnet lherzolite mantle xenolith from

Pali-Aike, Patagonia. The almost colorless grains are olivine, orthopyroxene is brownish-green, clinopyroxene bright green and garnet is red. Grain size is about 1 mm. Photograph courtesy of Sylvie Demouchy.

Series Editor: Jodi J. Rosso GEOCHEMICAL SOCIETY MINERALOGICAL SOCIETY OF AMERICA

SHORT COURSE SERIES DEDICATION Dr. William C. Luth has had a long and distinguished career in research, education and in the government. He was a leader in experimental petrology and in training graduate students at Stanford University. His eﬀorts at Sandia National Laboratory and at the Department of Energy’s headquarters resulted in the initiation and long-term support of many of the cu ing edge research projects whose results form the foundations of these short courses. Bill’s broad interest in understanding fundamental geochemical processes and their applications to national problems is a continuous thread through both his university and government career. He retired in 1996, but his eﬀorts to foster excellent basic research, and to promote the development of advanced analytical capabilities gave a unique focus to the basic research portfolio in Geosciences at the Department of Energy. He has been, and continues to be, a friend and mentor to many of us. It is appropriate to celebrate his career in education and government service with this series of courses.

Reviews in Mineralogy and Geochemistry, Volume 62 Water in Nominally Anhydrous Minerals ISSN 1529-6466 ISBN 0-939950-74-X

Copyright 2006

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.

WATER in NOMINALLY ANHYDROUS MINERALS 62

Reviews in Mineralogy and Geochemistry

62

FROM THE SERIES EDITOR The review chapters in this volume were the basis for a four day short course on Water in Nominally Anhydrous Minerals held in Verbania, Lago Maggiore, Italy (October 1-4, 2006). The editors Hans Keppler and Joe Smyth have done an excellent job organizing this volume and the associated short course. Meeting deadlines (often ahead of schedule!) and keeping track of so many authors can be a thankless job at times but I truly appreciate all their hard work. Hans’ “friendly reminder” e-mails certainly kept us all on task and his eye for detail (small and large) made my job much more enjoyable! I extend my sincere thanks to him for his efforts! Any supplemental material and errata (if any) can be found at the MSA website www.minsocam.org. Jodi J. Rosso, Series Editor West Richland, Washington August 2006

PREFACE Earth is a water planet. Oceans of liquid water dominate the surface processes of the planet. On the surface, water controls weathering as well as transport and deposition of sediments. Liquid water is necessary for life. In the interior, water fluxes melting and controls the solidstate viscosity of the convecting mantle and so controls volcanism and tectonics. Oceans cover more than 70% of the surface but make up only about 0.025% of the planet’s mass. Hydrogen is the most abundant element in the cosmos, but in the bulk Earth, it is one of the most poorly constrained chemical compositional variables. Almost all of the nominally anhydrous minerals that compose the Earth’s crust and mantle can incorporate measurable amounts of hydrogen. Because these are minerals that contain oxygen as the principal anion, the major incorporation mechanism is as hydroxyl, OH−, and the chemical component is equivalent to water, H2O. Although the hydrogen proton can be considered a monovalent cation, it does not occupy same structural position as a typical cation in a mineral structure, but rather forms a hydrogen bond with the oxygens on the edge of the coordination polyhedron. The amount incorporated is thus quite sensitive to pressure and the amount of H that can be incorporated in these phases generally increases with pressure and sometimes with temperature. Hydrogen solubility in nominally anhydrous minerals is thus much more sensitive to temperature and pressure than that of other elements. Because the mass of rock in the mantle is so large relative to ocean mass, the amount that is incorporated the nominally anhydrous phases of the interior may constitute the largest reservoir of water in the planet. Understanding the behavior and chemistry of hydrogen in minerals at the atomic scale is thus central to understanding the geology of the planet. There have been significant recent advances in the detection, measurement, and location of H in the nominally anhydrous silicate and oxide minerals that compose the planet. There have also been advances in experimental methods for measurement of H diffusion and the effects of H on the phase 1529-6466/06/0062-0000$05.00

DOI: 10.2138/rmg.2006.62.0

Water in Nominally Anhydrous Minerals ‒ Preface boundaries and physical properties whereby the presence of H in the interior may be inferred from seismic or other geophysical studies. It is the objective of this volume to consolidate these advances with reviews of recent research in the geochemistry and mineral physics of hydrogen in the principal mineral phases of the Earth’s crust and mantle.

The chapters We begin with a review of analytical methods for measuring and calibrating water contents in nominally anhydrous minerals by George Rossman. While infrared spectroscopy is still the most sensitive and most convenient method for detecting water in minerals, it is not intrinsically quantitative but requires calibration by some other, independent analytical method, such as nuclear reaction analysis, hydrogen manometry, or SIMS. A particular advantage of infrared spectroscopy, however, is the fact that it does not only probe the concentration, but also the structure of hydrous species in a mineral and in many cases the precise location of a proton in a mineral structure can be worked out based on infrared spectra alone. The methods and principles behind this are reviewed by Eugen Libowitzky and Anton Beran, with many illustrative examples. Compared to infrared spectroscopy, NMR is much less used in studying hydrogen in minerals, mostly due to its lower sensitivity, the requirement of samples free of paramagnetic ions such as Fe2+ and because of the more complicated instrumentation required for NMR measurements. However, NMR could be very useful under some circumstances. It could detect any hydrogen species in a sample, including such species as H2 that would be invisible with infrared. Potential applications of NMR to the study of hydrogen in minerals are reviewed by Simon Kohn. While structural models of “water” in minerals have already been deduced from infrared spectra several decades ago, in recent years atomistic modeling has become a powerful tool for predicting potential sites for hydrogen in minerals. The review by Kate Wright gives an overview over both quantum mechanical methods and classical methods based on interatomic potentials. Joseph Smyth then summarizes the crystal chemistry of hydrogen in high-pressure silicate and oxide minerals. As a general rule, the incorporation of hydrogen is not controlled by the size of potential sites in the crystal lattice; rather, the protons will preferentially attach to oxygen atoms that are electrostatically underbonded, such as the non-silicate oxygen atoms in some high-pressure phases. Moreover, heterovalent substitutions, e.g., the substitution of Al3+ for Si4+, can have a major effect on the incorporation of hydrogen. Data on water in natural minerals from crust and mantle are compiled and discussed in three reviews by Elisabeth Johnson, Henrik Skogby and by Anton Beran and Eugen Libowitzky. Among the major mantle minerals, clinopyroxenes usually retain the highest water contents, followed by orthopyroxenes and olivine, while the water contents in garnets are generally low. Most of these water contents need to be considered as minimum values, as many of the mantle xenoliths may have lost water during ascent. However, there are some cases where the correlation between the water contents and other geochemical parameters suggest that the measured water concentrations reflect the true original water content in the mantle. The basic thermodynamics as well as experimental data on water solubility and partitioning are reviewed by Hans Keppler and Nathalie Bolfan Casanova. Water solubility in minerals depends in a complicated way on pressure, temperature, water fugacity and bulk composition. For example, water solubility in the same mineral can increase or decrease with temperature, depending on the pressure of the experiments. Nevertheless, the pressure and temperature dependence of water solubility can be described by a rather simple thermodynamic formalism and for most minerals of the upper mantle, the relevant thermodynamic parameters are known. The highest water solubilities are reached in the minerals wadsleyite and ringwoodite stable in the transition zone, while the minerals of the lower mantle are probably mostly dry. The rather limited experimental data on water partitioning between silicate melts and minerals are reviewed by Simon Kohn and Kevin Grant. One important observation here is that comparing vi

Water in Nominally Anhydrous Minerals ‒ Preface the compatibility of hydrogen with that of some rare earth element is misleading, as such correlations are always limited to a small range of pressure and temperature for a given mineral. The stabilities of hydrous phases in the peridotite mantle and in subducted slabs are reviewed by Daniel Frost and by Tatsuhiko Kawamoto. While most of the water in the mantle is certainly stored in the nominally anhydrous minerals, hydrous phases can be important storage sites of water in certain environments. Amphibole and phlogopite require a significant metasomatic enrichment of Na and K in order to be stabilized in the upper mantle, but serpentine may be an important carrier of water in cold subducted slabs. The diffusion of hydrogen in minerals is reviewed by Jannick Ingrin and Marc Blanchard. An important general observation here is that natural minerals usually do not loose hydrogen as water, but as H2 generated by redox reaction of OH with Fe2+. Moreover, diffusion coefficients of different mantle minerals can vary by orders of magnitude, often with significant anisotropy. While some minerals in a mantle xenolith may therefore have lost virtually all of their water during ascent, other minerals may still preserve the original water content and in general, the apparent partition coefficients of water between the minerals of the same xenolith can be totally out of equilibrium. Accordingly, it would be highly desirable to directly deduce the water content in the mantle from geophysical data. One strategy, based on seismic velocities and therefore ultimately on the effect of water on the equation of state of minerals, is outlined by Steve Jacobsen. The dissolution of water in minerals usually increases the number of cation vacancies, yielding reduced bulk and shear moduli and seismic velocities. Particularly, the effect on shear velocities is strong and probably larger than the effect expected from local temperature variations. Accordingly, the vS / vP ratio could be a sensitive indicator of mantle hydration. A more general approach towards remote sensing of hydrogen in the Earth’s mantle, including effects of seismic anisotropy due to lattice preferred orientation and the use of electrical conductivity data is presented by Shun-ichiro Karato. Probably the most important effect of water on geodynamics is related to the fact that even traces of water dramatically reduce the mechanical strength of rocks during deformation. The physics behind this effect is discussed by David Kohlstedt. Interestingly, it appears that the main mechanism behind “hydrolytic weakening” is related to the effect of water on the concentration and mobility of Si vacancies, rather than to the protons themselves. Water may have major effects on the location of mantle discontinuities, as reviewed by Eiji Ohtani and Konstantin Litasov. Most of these effects can be rationalized as being due to the expansion of the stability fields of those phases (e.g., wadsleyite) that preferentially incorporate water. Together with other geophysical data, the changes in the depths of discontinuities are a promising tool for the remote sensing of water contents in the mantle. The global effects of water on the evolution of our planet are reviewed in the last two chapters by Bernard Marty, Reika Yokochi and Klaus Regenauer-Lieb. By combining hydrogen und nitrogen isotope data, Marty and Yokochi demonstrate convincingly that most of the Earth´s water very likely originated from a chondritic source. Water may have had a profound effect on the early evolution of our planet, since a water-rich dense atmosphere could have favored melting by a thermal blanketing effect. However, Marty and Yokochi also show very clearly that it is pretty much impossible to derive reliable estimates of the Earth´s present-day water content from cosmochemical arguments, since many factors affecting the loss of water during and after accretion are poorly constrained or not constrained at all. In the last chapter, Klaus Regenauer-Lieb investigates the effect of water on the style of global tectonics. He demonstrates that plate tectonics as we know it is only possible if the water content of the mantle is above a threshold value. The different tectonic style observed on Mars and Venus may therefore be directly related to differences in mantle water content. Earth is the water planet – not just because of its oceans, but also because of its tectonic evolution. vii

Water in Nominally Anhydrous Minerals ‒ Preface Acknowledgments This volume and the accompanying short course in Verbania were made possible by generous support from the Mineralogical Society of America, the Geochemical Society, the United State Department of Energy, the German Mineralogical Society and Bayerisches Geoinstitut. The Verbania short course is the first MSA/GS short course ever held in Italy. We are very grateful for the generosity and the international spirit of the supporting institutions, which made this project possible. The preparation of the short course benefited enormously from the permanent advice by Alex Speer. Finally, we would like to thank Jodi Rosso for the efficient and professional handling of the manuscript and for her patience with authors and editors who ignore deadlines. August 2006 Hans Keppler Bayreuth, Germany

Joseph R. Smyth Boulder, Colorado, USA

viii

WATER in NOMINALLY ANHYDROUS MINERALS 62

Reviews in Mineralogy and Geochemistry

62

FROM THE SERIES EDITOR The review chapters in this volume were the basis for a four day short course on Water in Nominally Anhydrous Minerals held in Verbania, Lago Maggiore, Italy (October 1-4, 2006). The editors Hans Keppler and Joe Smyth have done an excellent job organizing this volume and the associated short course. Meeting deadlines (often ahead of schedule!) and keeping track of so many authors can be a thankless job at times but I truly appreciate all their hard work. Hans’ “friendly reminder” e-mails certainly kept us all on task and his eye for detail (small and large) made my job much more enjoyable! I extend my sincere thanks to him for his efforts! Any supplemental material and errata (if any) can be found at the MSA website www.minsocam.org. Jodi J. Rosso, Series Editor West Richland, Washington August 2006

PREFACE Earth is a water planet. Oceans of liquid water dominate the surface processes of the planet. On the surface, water controls weathering as well as transport and deposition of sediments. Liquid water is necessary for life. In the interior, water fluxes melting and controls the solidstate viscosity of the convecting mantle and so controls volcanism and tectonics. Oceans cover more than 70% of the surface but make up only about 0.025% of the planet’s mass. Hydrogen is the most abundant element in the cosmos, but in the bulk Earth, it is one of the most poorly constrained chemical compositional variables. Almost all of the nominally anhydrous minerals that compose the Earth’s crust and mantle can incorporate measurable amounts of hydrogen. Because these are minerals that contain oxygen as the principal anion, the major incorporation mechanism is as hydroxyl, OH−, and the chemical component is equivalent to water, H2O. Although the hydrogen proton can be considered a monovalent cation, it does not occupy same structural position as a typical cation in a mineral structure, but rather forms a hydrogen bond with the oxygens on the edge of the coordination polyhedron. The amount incorporated is thus quite sensitive to pressure and the amount of H that can be incorporated in these phases generally increases with pressure and sometimes with temperature. Hydrogen solubility in nominally anhydrous minerals is thus much more sensitive to temperature and pressure than that of other elements. Because the mass of rock in the mantle is so large relative to ocean mass, the amount that is incorporated the nominally anhydrous phases of the interior may constitute the largest reservoir of water in the planet. Understanding the behavior and chemistry of hydrogen in minerals at the atomic scale is thus central to understanding the geology of the planet. There have been significant recent advances in the detection, measurement, and location of H in the nominally anhydrous silicate and oxide minerals that compose the planet. There have also been advances in experimental methods for measurement of H diffusion and the effects of H on the phase 1529-6466/06/0062-0000$05.00

DOI: 10.2138/rmg.2006.62.0

Water in Nominally Anhydrous Minerals ‒ Preface boundaries and physical properties whereby the presence of H in the interior may be inferred from seismic or other geophysical studies. It is the objective of this volume to consolidate these advances with reviews of recent research in the geochemistry and mineral physics of hydrogen in the principal mineral phases of the Earth’s crust and mantle.

The chapters We begin with a review of analytical methods for measuring and calibrating water contents in nominally anhydrous minerals by George Rossman. While infrared spectroscopy is still the most sensitive and most convenient method for detecting water in minerals, it is not intrinsically quantitative but requires calibration by some other, independent analytical method, such as nuclear reaction analysis, hydrogen manometry, or SIMS. A particular advantage of infrared spectroscopy, however, is the fact that it does not only probe the concentration, but also the structure of hydrous species in a mineral and in many cases the precise location of a proton in a mineral structure can be worked out based on infrared spectra alone. The methods and principles behind this are reviewed by Eugen Libowitzky and Anton Beran, with many illustrative examples. Compared to infrared spectroscopy, NMR is much less used in studying hydrogen in minerals, mostly due to its lower sensitivity, the requirement of samples free of paramagnetic ions such as Fe2+ and because of the more complicated instrumentation required for NMR measurements. However, NMR could be very useful under some circumstances. It could detect any hydrogen species in a sample, including such species as H2 that would be invisible with infrared. Potential applications of NMR to the study of hydrogen in minerals are reviewed by Simon Kohn. While structural models of “water” in minerals have already been deduced from infrared spectra several decades ago, in recent years atomistic modeling has become a powerful tool for predicting potential sites for hydrogen in minerals. The review by Kate Wright gives an overview over both quantum mechanical methods and classical methods based on interatomic potentials. Joseph Smyth then summarizes the crystal chemistry of hydrogen in high-pressure silicate and oxide minerals. As a general rule, the incorporation of hydrogen is not controlled by the size of potential sites in the crystal lattice; rather, the protons will preferentially attach to oxygen atoms that are electrostatically underbonded, such as the non-silicate oxygen atoms in some high-pressure phases. Moreover, heterovalent substitutions, e.g., the substitution of Al3+ for Si4+, can have a major effect on the incorporation of hydrogen. Data on water in natural minerals from crust and mantle are compiled and discussed in three reviews by Elisabeth Johnson, Henrik Skogby and by Anton Beran and Eugen Libowitzky. Among the major mantle minerals, clinopyroxenes usually retain the highest water contents, followed by orthopyroxenes and olivine, while the water contents in garnets are generally low. Most of these water contents need to be considered as minimum values, as many of the mantle xenoliths may have lost water during ascent. However, there are some cases where the correlation between the water contents and other geochemical parameters suggest that the measured water concentrations reflect the true original water content in the mantle. The basic thermodynamics as well as experimental data on water solubility and partitioning are reviewed by Hans Keppler and Nathalie Bolfan Casanova. Water solubility in minerals depends in a complicated way on pressure, temperature, water fugacity and bulk composition. For example, water solubility in the same mineral can increase or decrease with temperature, depending on the pressure of the experiments. Nevertheless, the pressure and temperature dependence of water solubility can be described by a rather simple thermodynamic formalism and for most minerals of the upper mantle, the relevant thermodynamic parameters are known. The highest water solubilities are reached in the minerals wadsleyite and ringwoodite stable in the transition zone, while the minerals of the lower mantle are probably mostly dry. The rather limited experimental data on water partitioning between silicate melts and minerals are reviewed by Simon Kohn and Kevin Grant. One important observation here is that comparing vi

Water in Nominally Anhydrous Minerals ‒ Preface the compatibility of hydrogen with that of some rare earth element is misleading, as such correlations are always limited to a small range of pressure and temperature for a given mineral. The stabilities of hydrous phases in the peridotite mantle and in subducted slabs are reviewed by Daniel Frost and by Tatsuhiko Kawamoto. While most of the water in the mantle is certainly stored in the nominally anhydrous minerals, hydrous phases can be important storage sites of water in certain environments. Amphibole and phlogopite require a significant metasomatic enrichment of Na and K in order to be stabilized in the upper mantle, but serpentine may be an important carrier of water in cold subducted slabs. The diffusion of hydrogen in minerals is reviewed by Jannick Ingrin and Marc Blanchard. An important general observation here is that natural minerals usually do not loose hydrogen as water, but as H2 generated by redox reaction of OH with Fe2+. Moreover, diffusion coefficients of different mantle minerals can vary by orders of magnitude, often with significant anisotropy. While some minerals in a mantle xenolith may therefore have lost virtually all of their water during ascent, other minerals may still preserve the original water content and in general, the apparent partition coefficients of water between the minerals of the same xenolith can be totally out of equilibrium. Accordingly, it would be highly desirable to directly deduce the water content in the mantle from geophysical data. One strategy, based on seismic velocities and therefore ultimately on the effect of water on the equation of state of minerals, is outlined by Steve Jacobsen. The dissolution of water in minerals usually increases the number of cation vacancies, yielding reduced bulk and shear moduli and seismic velocities. Particularly, the effect on shear velocities is strong and probably larger than the effect expected from local temperature variations. Accordingly, the vS / vP ratio could be a sensitive indicator of mantle hydration. A more general approach towards remote sensing of hydrogen in the Earth’s mantle, including effects of seismic anisotropy due to lattice preferred orientation and the use of electrical conductivity data is presented by Shun-ichiro Karato. Probably the most important effect of water on geodynamics is related to the fact that even traces of water dramatically reduce the mechanical strength of rocks during deformation. The physics behind this effect is discussed by David Kohlstedt. Interestingly, it appears that the main mechanism behind “hydrolytic weakening” is related to the effect of water on the concentration and mobility of Si vacancies, rather than to the protons themselves. Water may have major effects on the location of mantle discontinuities, as reviewed by Eiji Ohtani and Konstantin Litasov. Most of these effects can be rationalized as being due to the expansion of the stability fields of those phases (e.g., wadsleyite) that preferentially incorporate water. Together with other geophysical data, the changes in the depths of discontinuities are a promising tool for the remote sensing of water contents in the mantle. The global effects of water on the evolution of our planet are reviewed in the last two chapters by Bernard Marty, Reika Yokochi and Klaus Regenauer-Lieb. By combining hydrogen und nitrogen isotope data, Marty and Yokochi demonstrate convincingly that most of the Earth´s water very likely originated from a chondritic source. Water may have had a profound effect on the early evolution of our planet, since a water-rich dense atmosphere could have favored melting by a thermal blanketing effect. However, Marty and Yokochi also show very clearly that it is pretty much impossible to derive reliable estimates of the Earth´s present-day water content from cosmochemical arguments, since many factors affecting the loss of water during and after accretion are poorly constrained or not constrained at all. In the last chapter, Klaus Regenauer-Lieb investigates the effect of water on the style of global tectonics. He demonstrates that plate tectonics as we know it is only possible if the water content of the mantle is above a threshold value. The different tectonic style observed on Mars and Venus may therefore be directly related to differences in mantle water content. Earth is the water planet – not just because of its oceans, but also because of its tectonic evolution. vii

Water in Nominally Anhydrous Minerals ‒ Preface Acknowledgments This volume and the accompanying short course in Verbania were made possible by generous support from the Mineralogical Society of America, the Geochemical Society, the United State Department of Energy, the German Mineralogical Society and Bayerisches Geoinstitut. The Verbania short course is the first MSA/GS short course ever held in Italy. We are very grateful for the generosity and the international spirit of the supporting institutions, which made this project possible. The preparation of the short course benefited enormously from the permanent advice by Alex Speer. Finally, we would like to thank Jodi Rosso for the efficient and professional handling of the manuscript and for her patience with authors and editors who ignore deadlines. August 2006 Hans Keppler Bayreuth, Germany

Joseph R. Smyth Boulder, Colorado, USA

viii

1

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 1-28, 2006 Copyright © Mineralogical Society of America

Analytical Methods for Measuring Water in Nominally Anhydrous Minerals George R. Rossman Division of Geological and Planetary Sciences California Institute of Technology Pasadena, California, 91125-2500, U.S.A. e-mail: [email protected]

INTRODUCTION Decades of work have shown that trace- to minor-amounts of hydrous components commonly occur in minerals whose chemical formula would be normally written without any hydrogen, namely, the nominally anhydrous minerals (NAMs). When the concentrations of the hydrous components are several tenths of a percent by weight or higher, a variety of analytical methods such as weight loss on heating, X-ray cell parameters, X-ray structure refinement, Karl-Fischer titrations, or even careful electron microprobe analyses can be used to establish their concentrations (e.g., Aines and Rossman 1991). However, for most NAMs, accurate determinations with these common analytical methods prove difficult if not impossible. For this reason, infrared (IR) spectroscopy has become, and remains, the most widely used method to detect and analyze hydrous components (OH or H2O) in minerals and glasses because it is both highly sensitive and can be done rapidly with a commonly available, modestly priced instrument and at dimensions of just a few tens of micrometers. A change in the electric dipole occurs when the OH bond in either water and hydroxyl ions vibrate. This motion has a resonance coupling with electromagnetic radiation generally in the 3500 cm−1 region of the infrared spectrum. In addition, bending motions of the water molecule, and overtones and combination of these motions produce absorption in the infrared. Under favorable conditions, namely a sharp band in a single orientation, just a few nanometers equivalent thickness of a hydroxyl species such as an amphibole can be detected in an otherwise anhydrous mineral such a pyroxene (Skogby et al. 1990). Routinely, detection limits of a few to tens of ppm wt of H2O in a mineral can be detected and often quantitatively determined. The overtone and combination modes of OH and H2O behave in predictable fashion in minerals (Rossman 1975) so that the two species can usually be separated from each other. Infrared spectra, however easily obtained, are not rigorously self-calibrating, so independent methods of analysis have been necessary to calibrate the spectroscopic work. A couple general correlations of IR band intensity with the absorption energy have proven useful, if approximate. Various absolute hydrogen extraction methods have proven highly useful for purpose of rigorous calibration. More recently, nuclear methods that rely upon specific resonant reactions with the hydrogen nucleus or nuclear scattering specific to hydrogen have gained importance and have provided critical absolute calibrations of the infrared spectra. Secondary Ion Mass Spectroscopy (SIMS) for hydrogen is still in the early stages of development but once calibrated, and with established protocols, should play an ever-expanding role in the future. NanoSIMS promises to bring hydrogen analyses to ever finer spatial dimensions but will require significant effort before it can be regarded as an accurate analytical technique for small concentrations of hydrogen. The purpose of this chapter is to review the various methods that have been used to analyze hydrous components in the NAMs. 1529-6466/06/0062-0001$05.00

DOI: 10.2138/rmg.2006.62.1

2

Rossman ANALYTICAL METHODS

Early infrared studies Much of the early interest in OH in minerals came from the study of synthetic minerals used in the electronics industry. Quartz, in particular, was an important phase used for frequency control in telecommunications and radio circuits. Consequently, much effort was directed towards the understanding of factors that influenced the efficiency and cost of these devices. Water in quartz was one of the most important factors. The OH bond is dipolar with a partial negative charge on the oxide ion and a partial positive charge on the hydrogen ion. Thus, the vibrations of the OH bond coupled well to infrared radiation and infrared spectroscopy quickly became the tool of choice to study OH in both natural and synthetic minerals. An important early study was conducted by Kats and Haven (1960) who used deuteration to demonstrate which bands in the complex quartz spectrum in the 3000 to 4000 cm−1 region originated from 1H as opposed to overtone or combination bands of the quartz vibrational spectrum that appeared in the same region. Once the OH vibrations were positively identified, Kats (1962) performed a comprehensive study of OH in quartz and identified which of the sharp band absorptions in the 3000-3600 cm−1 region are due to O-H stretching vibrations. Kats further showed that most of the absorptions are primarily due to the presence of Al3+ substitution for Si4+ with charge compensating cations (such as H+, Li+, Na+) in defects in the crystal. Other studies were taking place at Bell Labs in the United States where elastic properties and dielectric loss in synthetic quartz was related to H defects (King et al. 1960; Dodd and Fraser 1965, 1967). In these studies, the relationship between infrared absorption, and hydroxyl and water defects in quartz was also being established. During these times, Brunner et al. (1961) concluded that H enters defects in clear, natural quartz in the form of OH ions and estimated the amount of H as 1018 per cc (corresponding to about 15 ppm H2O wt). These early estimates showed that small amounts of hydrous components could have a large impact on the physical properties of the host phase. As work on synthetic quartz progressed, studies of quartz also used natural samples and ultimately, the results were reported in the mineralogical literature through the work of Dodd and Fraser (1965). Simultaneously, interest in the low concentrations of water in ring silicate minerals was generated by infrared (Schreyer and Yoder 1964; Wood and Nassau 1967; Farrell and Newnham 1967) and NMR (proton nuclear magnetic resonance) (Pare and Ducros 1964; Sugitani et al. 1966) studies of beryl that demonstrated that water molecules occur in the c-axis channels. The NMR work showed that the water molecules were in motion and the IR studies showed that the water molecule existed in two independent crystallographic orientations in the crystal. In Austria, in the late 1960’s and early 1970’s, Beran and Zemann obtained the IR spectra of a number of minerals such as titanite, kyanite, axinite, titanium oxides, cassiterite (Beran 1970a,b,c,d; Beran and Zemann 1969a,b, 1971) and demonstrated that they had structurally bound, crystallographically oriented OH groups. These studied demonstrated that polarized infrared radiation could establish the orientation of the OH groups in minerals and demonstrated that trace amounts of hydroxyl occur broadly in a number of nominally anhydrous minerals. A couple of significant motivations to develop quantitative understanding of the H-content of nominally anhydrous minerals appeared in the early 1970’s. Martin and Donnay (1972) suggested that hydrogen may be stored as OH groups in minerals in the deep earth, and Wilkins and Sabine (1973) initiated a broad effort to determine the amount of hydrous components in a variety of minerals by combining infrared absorption with independent water analysis (P2O5 electrolytic coulometry). Although we now recognize that many of the analyses of Wilkins and Sabine included alteration products and water in micro-inclusions, they did set the quantitative stage for further detailed studies.

Analytical Methods

3

Another major impetus to the study of water in the nominally anhydrous minerals came from the studies of the rheological properties of, first, quartz (Griggs and Blacic 1965; Kirby and McCormick 1979), and then olivine (Mackwell et al. 1985). To study how water weakens minerals, it was necessary to know both the chemical species of the hydrous components that enter nominally anhydrous minerals, and to know their absolute concentrations.

Quantitative IR methods The determination of the concentration of OH or H2O in an “anhydrous” mineral depends upon accurate measurement of the infrared spectrum and ultimately on an independent calibration. Infrared spectra are intrinsically not self-calibrating. A number attempts have been made to develop generic calibrations. These often may be good as an initial estimate of the water concentration, but, for many systems, have been shown to be inadequate for precise work. Thus, mineral-specific calibrations have been developed. Once such calibrations are established and properly published, they can be used by other labs worldwide, even if an inhouse standard is not available. The well-established Beer-Lambert law is used to determine the concentration of hydrous species in a mineral from the infrared spectra: Absorbance = ε × c × t

(1)

This relates Absorbance (A), the band height in the region of interest (corrected for baseline), c, the concentration of hydrous species expressed in moles of H2O per liter of mineral, and t, the thickness of the path (in cm) through which the measurement is made where ε is a mineral-specific calibration factor. In the classical chemical applications, the sample is in solution, so only one measurement is made. In the case of anisotropic solids, it is necessary to make the measurement in multiple directions (Libowitzky and Rossman 1996). Typically, linearly polarized light would be used and measurements would be made along the three principal extinction directions, X, Y, and Z. In this case, the intensities would be summed so A becomes AX + AY + AZ (where AX is the absorbance obtained with light polarized in the X direction, etc.). This approach tends to work best with phases that have one or a small number of narrow bands in the OH region. It also requires knowledge of the density of the mineral to convert from moles per liter to weight percent (or ppm) water. For most minerals, it is usually more useful to use a modified version of the Beer-Lambert law that uses integrated band areas rather than band heights. Band heights can vary depending on both the quality of the polarizer in the instrument and on the spectroscopic resolution of the instrument whereas band areas are less dependent on these parameters. The band height measured by the Absorbance is replaced by the total integrated area of bands in the region of interest Absorbancetotal (also written as Abstotal or Atotal). The concentration, c, remains expressed as moles of H2O per liter of mineral. In this case, the absorption coefficient, ε, is replaced by the integral molar absorption coefficient, I, in units of L/(mol·cm2). When c is expressed as ppm H2O by weight, the absorption coefficient becomes the integral specific absorption coefficient (I’, ppm−1·cm−2). The absorption coefficient for each species of hydrogen is found by determining the concentration, c, by an independent, absolute method and measuring Abstotal from polarized IR spectra in the three principal optical directions (X, Y, and Z) for the mineral of interest. For an orthorhombic mineral such as olivine: Abstotal =

1 ta

ν2

1

v2

1

v2

∫ Absa dν + tb ∫ Absb dv + tc ∫ Absc dv

v1

v1

(2)

v1

Here, the equation specifies measuring the integrated area of an orthorhombic crystal with light polarized in the E||a, E||b, and E||c directions between the appropriate wavenumber limits of the OH bands, ν1 and ν2. For lower symmetry crystals (monoclinic, and triclinic)

4

Rossman

Abstotal = ∫AbsX + ∫AbsY + ∫AbsZ, and for a uniaxial crystal (hexagonal or tetragonal) Abstotal = 2∫Abs⊥c + ∫Absc (e.g., Libowitzky and Rossman 1996). To be comparable to measurements on lower symmetry crystals, an isotropic crystal would need to have Atotal = 3∫Absa. Paterson’s method. If the absorption frequency and intensity of a unit concentration of OH were a constant, then a single calibration of the OH spectrum would be all that is needed to conduct quantitative analysis with IR spectroscopy. Unfortunately, that is not the case. First of all, while the fundamental stretching vibration of a free (gaseous) hydroxide ion occurs at 3555.59 cm−1 (Lutz 1995), the OH stretching frequency in a mineral commonly can occur over a range of several hundred wavenumbers and can vary by nearly 2000 cm−1. A variety of studies (Nakamoto et al. 1955; Bellamy and Owen 1969; Novak 1974) showed that for a variety of chemical elements, the stretching frequency of an X-H bond in an X-H···Y hydrogen bonded system is a function of the X-Y distance. This includes O-H bonds. These authors derived empirical fits to experimental data that mathematically expressed this relationship. The second observation of interest is that the infrared absorption intensity of a unit concentration of OH in a solid is obviously not constant. Paterson (1982) confirmed that the strength of the OH absorption in the 3600 to 3000 cm−1 region was frequency dependent. From the calibrations available for various substances, he presented a single empirical calibration line that related the OH intensity to band position that could be applied as a first approximation for determining the amount of OH in a variety of substances such as silicate glasses, quartz, and various forms of water. This was the first generic calibration specifically designed for the study of hydrous components in minerals and glasses. Paterson demonstrated that the intensity of an OH band (normalized to a unit concentration of H2O) increases when the band occurs at lower wavenumbers (stronger hydrogen bonding). This trend has been used by a number of authors to estimate the OH content of various minerals. Subsequent work has shown that determinations based on Paterson’s trends are a reasonable first estimate, but that accurate determinations do require mineral-specific calibrations. Paterson’s method first assumes that if a crystal is being measured, it is in a known crystallographic orientation. To determine the concentration of hydroxyl groups in the sample, the integrated absorbance is determined by integration of the infrared spectra over the region dominated by the stretching vibrations due to O-H bonds, typically from approximately 3750 to 3000 cm–1. The integral molar absorption coefficient (I) is scaled to reflect the higher intrinsic intensities of bands at lower wavenumbers (stronger H-bonds) through the equation: I = γ ×150 × (3780 − ν)

(3)

where ν is the wavenumber and gamma (γ) is a factor to take account of the anisotropy of the crystal based on an assumption that O-H bonds are oriented in a single direction. The OH concentration is then calculated from a Beer-Lambert law relationship: ConcentrationOH =

1 A(ν) dν (150 × γ ) ∫ (3780 − ν)

( 4)

assuming that the data are scaled for 1 cm sample thickness. Although uncertainties in this calibration were thought by Paterson to be about 30%, it has been widely adopted, partly in the hope that it would eliminate the need for more involved polarized light observations with multiple crystallographic directions. However, the studies of Libowitzky and Rossman (1997) and Bell et al. (2003) show that it can result in non-systematic underestimates of hydrogen concentrations. Examples of mineral specific calibrations that fall far from the trend are documented, particularly those that involve nominally anhydrous minerals with low concentrations of OH. As examples, the pyrope analyzed by Bell et al. (1995) departs from the Paterson trend by nearly a factor of three, the nuclear reaction analysis

Analytical Methods

5

of olivine by Bell et al. (2003) departs by more than a factor of two (Fig. 1) and the SIMS analysis of both olivine and orthopyroxene (Koga et al. 2003) show that the Paterson trend also underestimates their OH concentrations. Libowitzky and Rossman’s revision. Libowitzky and Rossman (1997) presented an updated version of the correlation of Paterson (1982). They measured polarized IR absorption data from single crystal minerals that contained stoichiometric water contents in the form of either OH or H2O. These data were used to construct a calibration curve for the intensity of the infrared absorption as a function of the band energy. Specifically, integrated molar absorption coefficient, εi (in units of cm−2 per moleH2O/liter), was evaluated as function of the mean wavenumber of the OH stretching band (in units of cm−1). The result in Figure 2 shows that an increase in the hydrogen bonding leads to a decrease in the energy of the OH stretching energy which, in turn, is associated with an increase in the intensity of absorption. The form of the correlation is εi = 246.6 × (3752 − ν)

(5)

where ν is the mean wavenumber of the OH stretching band. The results in Figure 2 show that the revised calibration produces εi values about threequarters of those of Paterson (1982). Measurements of minerals with stoichiometric OH are difficult to obtain. Their OH intensities are so high that crystals must be prepared very thin (perhaps as thin as 2 µm). Such preparations are difficult to near impossible; and when successful, the determination of their thickness to a high degree of accuracy is difficult.

Figure 1. Comparison of the results of the calibration developed by Bell et al. (2003) using Nuclear Reaction Analysis and the OH analysis method of Paterson (1982), as applied to polarized (solid circle) olivine spectra. Modified after Fig. 6 of Bell et al. (2003).

Figure 2. The correlation of the integrated molar absorption coefficient of OH stretching vs. wavenumber. Circles are experimental data points for stoichiometric minerals. The correlation of Paterson (1982) is shown for comparison. This means that if all things are equal, the Paterson trend underestimates the OH content. From Libowitzky and Rossman (1997).

6

Rossman

In a related effort, Libowitzky (1999) evaluated correlations specific to minerals between the frequency of the O-H stretching vibration and the length of the oxygen-oxygen distance and the H···O distances in the O-H···O hydrogen bond. Effectively, the shorter these distances are, the lower becomes the energy of the O-H stretch. Because the intensity of the OH band is related to the energy of the vibration (Libowitzky and Rossman 1997), such correlations provide some degree of a predictive estimate about the intensity of an OH absorption that arises from a particular site in a crystal. Use of unoriented grains. Asimow et al. (2006) present a method that allows multiple, randomly oriented grains of a mineral to be used to determine the total absorbance. In their method, the spectra of oriented sample of the phase of interest must already exist. Then, the spectra of three different randomly oriented crystals are measured, and the orientations of the grains are determined via methods such as electron backscatter diffraction (EBSD) or from the silicate overtone bands in the infrared spectra. They demonstrated that such methods result in angular errors of typically only 6 degrees and provide a surprising good determination of the OH content of the phase. Polarizer considerations. A linear polarizer must be used in the infrared beam of conventional spectrometers to obtain the total absorbance of anisotropic crystals. Commonly, the polarizers are made of a fine, parallel wire grid deposited on an infrared-transparent substrate such as CaF2 or KRS5 (a thallium bromide iodide). These polarizers have wide acceptance angles and are readily available, but have only moderate polarization ratios. Crystal polarizers of a design similar to calcite polarizers used in the visible wavelength region are also available, but often have a narrow range of wavelengths over which they function. Lithium iodate covers a wide wavelength range and has a very high polarization ratio, but is hydroscopic and no longer readily available. Libowitzky and Rossman (1996) discussed the principles of quantitative absorbance measurements of anisotropic crystals and paid particular attention to the influence of the quality of the polarizers upon the results. First, they showed that the use of unpolarized radiation with an anisotropic crystal could not produce quantitatively accurate results. The Beer-Lambert law demands that the height of an absorption band will scale with the thickness of the sample. Figure 3 demonstrates how the spectrum taken with linearly polarized radiation follows the law. It also shows that unpolarized spectra do not scale according to the law. This means that unpolarized spectra should not be used to calibrate the infrared spectrum of OH

Figure 3. Comparison of the intensity of a carbonate overtone band in the calcite spectrum taken with well-polarized and unpolarized radiation. The experiment that used different thickness of calcite to test the Beer Lambert law shows that unpolarized spectra are not appropriate to quantitatively measure anisotropic crystals. From Libowitzky and Rossman (1997).

Analytical Methods

7

in an anisotropic standard, and cannot be used to accurately determine the concentration of OH in an anisotropic unknown. The more highly anisotropic the sample is, the more problematic this issue will become. Libowitzky and Rossman also showed that the intensity of an absorption band of an anisotropic crystal is highly dependent upon the polarization ratio of the polarizers (Fig. 4) which means that if band heights are used to calibrate the infrared spectra, results can vary significantly from lab to lab if the appropriate in-lab standards are not available. Baselines issues. Figure 5 shows that strongly rising, non-linear baselines may be an intrinsic part of the spectrum in the OH region. These baselines commonly arise from Fe2+ and may arise from silicate overtones in thick samples. A major, subjective source of uncertainty in IR measurements of OH in minerals remains the choice of the baseline.

Figure 4. The intensity of absorbance depends on the quality of the polarizer used for the measurement. Here, the spectrum (E ⊥ c) of three bands in the calcite spectrum was obtained with a high efficiency polarizer (LiIO3), a lower efficiency wire-grid polarizer (gold wire on AgBr), and without polarization. Modified after Figure 6 of Libowitzky and Rossman (1996).

Figure 5. Infrared spectra of a clinopyroxene that show the baseline remaining after the crystal is fully dehydrated. Contributions from ferrous iron cause the rising baseline towards the long wavenumber side. From Bell et al. 1995, Figure 1.

8

Rossman

Comments on terminology. The terminology for spectroscopic units has not been consistent in the literature. Chemical terminology, the source of these terms, has evolved, and geoscience has had to modify some of the standard terms for anisotropic materials. Table 1 presents a compendium of terminology taken from the web site of the International Union of Pure and Applied Chemistry. In addition, the currently preferred terminology is compared to other terminology found in the literature.

Mineral specific calibrations While the generic calibrations developed by Paterson (1982) and later refined by Libowitzky and Rossman (1997) are useful first approximations, they are not necessarily accurate. There is no principle of science that demands that the infrared absorption intensity of all OH bonds be the same, or that the intensity of all OH bonds vary smoothly with the O-H···O hydrogen bond distance. Unpublished work by this author has shown that the intensity of other bonds such as C-O (carbonyl) and C-N (cyano) can vary by orders of magnitude. Thus, there is the need for mineral-specific calibrations. A variety of experimental methods, discussed in the following sections, have been used over the years to independently determine the amount of hydrous components in minerals. As is often the case in the history of development of analytical methods for trace components, early attempts suffered from large (and often Table 1. Selected terminology used in quantitative spectroscopy of minerals. Absorbance = log(I0/I) directly measured by the instrument Attenuation coefficient Analogous to the absorption coefficient, but differs from it because it accounts for the diffusion of radiation that includes absorption as well as scattering and luminescence. Formerly, it was called the extinction coefficient, a term that is now discouraged. linear absorption coefficient = Absorbance divided by the optical path length molar absorption coefficient = ε = molar absorptivity in earlier literature = linear absorption coefficient divided by the amount concentration amount concentration = molarity in prior literature commonly expressed in units of moles per liter integral molar absorption coefficient I (in units of cm-2 per molH/liter). εi (in units of cm-2 per molH2O/liter). (note that the mols of H = mols OH) Integral absorbance (not defined by IUPAC) Integrated absorbance Abstot =

1 t1

ν2

1

ν2

1

ν2

∫ Abs1dν + t2 ∫ Abs2 dν + t3 ∫ Abs3dν

ν1

Absorbancetotal ∆ Integrated-Abstot

ν1

ν1

Analytical Methods

9

unrecognized) backgrounds, and the inability to separate the contributions of the hydrogen in the host phase from hydrogen contained in inclusions, cracks, and alteration products.

Thermogravimetric methods Thermogravimetric analysis (TGA) is a commonly used analytical method to determine the amount of mass lost from a sample during heating. It involves simultaneously heating and weighing a sample to produce a weight-loss vs. temperature curve. It is frequently used to determine water of hydration in minerals with more than trace quantities of water. The method has also been applied to water loss from nominally anhydrous minerals but with limited success. Early attempts to determine the H-content of garnets used the TGA method (Aines and Rossman 1984a) and coupled the results of this method with infrared spectra of the same samples. We now recognize that many of the earlier thermogravimetic methods over-estimated the water content of the NAMs due to the inclusion of contaminating water that remained trapped on the surface of the ground samples, even after the sample was “dried” by heating to over 125 °C prior to analysis. TGA was used to determine the water content of nepheline from Bancroft, Ontario (0.36 wt% H2O), and from Mt. Somma, Italy (0.17 wt% H2O) (Beran and Rossman 1989). Because these minerals have comparatively large water contents, the error introduced by the TGA method is small compared to what it may be when minerals with a few hundred ppm or less are analyzed by this method. The results of this method were also used to calibrate the infrared spectra of nepheline. While TGA analyses are conventionally conducted on ground samples, step heating experiments on slabs of single crystals used for infrared experiments demonstrate how difficult it can be to fully dehydrate a sample. Controlled heating experiments that were accompanied with infrared spectra of OH bands indicated that temperatures of about 1400 °C are needed to fully dehydrate slabs of some silicate minerals (sillimanite: Beran et al. 1989). Similar experiments with slabs of single crystal zircon indicated that OH is tightly held. Some OH persists in zircons even after the crystals are heated at 1500 °C (Woodhead et al. 1991). Ilchenko and Korzhinskaya (1993) also conducted step-heating experiments on kimberlitic zircon crystals and found that OH ions were only partially removed after heating to 1300 °C.

P2O5 cell coulometry P2O5 cell coulometry is based on the principle that water released during the thermal decomposition of a sample can react with P2O5, a non-conductor, and turns it into H3PO4, an electrical conductor. The amount of H3PO4 formed can be determined by the amount of electric current (coulombs) necessary to reverse the hydration reaction. One of the more popular commercial models used in mineral analysis was the DuPont moisture evolution analyzer (MEA). It consisted of a thermal decomposition chamber that led to a column containing a pair of closely spaced, P2O5-coated, Pt wire electrodes wound in a helical fashion. A dry nitrogen flow would carry the released water vapor into the electrodes where electrical current would flow between the wires whenever the P2O5 reacted with the water. A known mass of a stoichiometrically hydrated material was used to calibrate the system. The moisture evolution analyzer found use in some of the earlier analyses such as Wilkins and Sabine (1973) study, and the Aines and Rossman (1984) calibration of garnets. In practice, these systems had to be used regularly to prevent the P2O5 columns from going bad, and proved difficult for many users to regenerate once the columns did degrade. Because blanks with this method are typically several tens of micrograms of H2O, samples of at least a few hundred milligrams are required for the analysis of the nominally anhydrous minerals. (Aines and Rossman 1984).

10

Rossman

Hydrogen extraction with uranium reduction methods Hydrogen manometry. Hydrogen manometry has long been a standard and generally reliable method to determine the water content of samples. In this method, several hundred milligrams to gram quantities of samples are weighed into a metal (Mo, or Pt) crucible, and first degassed under vacuum and low heat to drive off the adsorbed moisture. The sample crucibles are then heated with an induction furnace to liberate the bound water while under vacuum. The volatiles (H2 and H2O) are converted to just water and trapped and separated from the condensable and non-condensable gases by distillation in cryogenic traps. The water vapor is next passed over a hot furnace containing uranium metal (Bigeleisen et al. 1952) to reduce the water to molecular hydrogen. Alternatively, zinc has been used to reduce water (Michel and Villemant 2003). The hydrogen is then moved by a mercury-piston Toepler pump into a calibrated chamber in which the volume of hydrogen can be measured at a known pressure. From the PV = nRT relationship, the absolute amount of hydrogen can be determined. The system can be calibrated by known amounts of water, or by dehydration of minerals or compounds with known, stoichiometric water contents. For minerals with very low hydrogen contents such as the nominally anhydrous minerals, significant blank corrections must be applied that correct for degassing from the crucibles (Bell et al. 1995). Errors have been reported to be much less than 1% with this method (Dyar et al. 1996). Additional details of the technique can be found in Holdaway et al. (1986) This method has been used to determine the water content of minerals that are used to calibrate infrared spectra. The advantage of using this approach is that once the sample is destroyed by the hydrogen extraction procedure, its value as a calibrant remains through the calibration of the infrared spectrum which can be used to analyze additional samples of the calibrated phase that have similar spectra. Furthermore, the infrared spectrum allows reevaluation of the calibration because the original spectrum can be compared to the spectrum of other samples re-calibrated by improved methods years later. Early calibration efforts with hydrogen extraction (Aines and Rossman 1984) include a grossular with 0.18 wt% H2O, and a pyrope with 0.08 wt% [that is probably overestimated based on the more recent calibration of Bell et al. (1995) that indicate about 37 ppm H2O]; and perthite feldspar from two pegmatites (Hofmeister and Rossman 1985a,b) that had water in the 0.09 to 0.15 wt% range. More recent calibrations with lower blank contributions (Fig. 6) consist of a pyrope with 56 ppm, an enstatite with 217 ppm, and an augite with 268 ppm (Bell et al. 1995). In these experiments, large quantities of sample had to be carefully prepared, and checked to eliminate inclusions, cracks and other imperfections. The clean material was

Figure 6. Quality of hydrogen extraction determination of water in garnet and two pyroxenes using aliquots of different mass to determine the water content. Figure 4 from Bell et al. (1995).

Analytical Methods

11

then crushed to less than 2 mm particles and the fraction less than 100 µm was discarded to minimize the effects of adsorbed water. Continuous flow mass spectrometry. A more recent variation of the hydrogen extraction technique uses continuous flow mass spectrometry to measure the absolute amount of hydrogen released from minerals by heating (O’Leary et al. 2006). This method is a modification of the method of Eiler and Kitchen (2001) used to determine D/H isotopic ratios of picoliter quantities of hydrogen. It requires about 1/1000 the amount of hydrogen required by conventional hydrogen manometry. Samples in the range of 50 µg to 20 mg of coarsely ground minerals are heated to release hydrous components, which are collected and converted to hydrogen by reaction with uranium (as opposed to carbon in the Eiler and Kitchen paper). The hydrogen is then detected in a mass spectrometer. The system is calibrated with a few hundred micrograms of zoisite grains of known H content. This system has been used to independently calibrate a series of garnets and pyroxenes that have been previously calibrated by conventional hydrogen extraction manometry or by nuclear methods. The linearity and agreement with previous calibrations has been excellent with samples at the few hundred-ppm H2O level and higher (Fig. 7).

Figure 7. Comparison of the water contents determined by the new micro-extraction method compared to conventional methods (O’Leary et al. 2006).

Nuclear methods for hydrogen determination A variety of nuclear reactions can be used to analyze hydrogen in solids (Lanford 1992). Some make use of nuclear reactions and others make use of nuclear scattering. Beams of ions accelerated to high energy can undergo a resonant nuclear reaction with the hydrogen ions in the target sample. Such methods are known either as Nuclear Resonant Reaction Analysis (NRRA), Nuclear Reaction Analysis (NRA) or Nuclear Profile Analysis (NPA) (when the hydrogen concentration is determined as a function of depth in the sample). The 6.42 MeV resonance of 19F with hydrogen and the 6.385 MeV resonance of 15N with hydrogen are the two that are typically used. Additional resonances of 19F at 16.44 MeV and 15N at 13.35 MeV can also be used (Xiong et al. 1987). In each of these reactions, the analysis depends upon the detection of gamma rays emitted from a heavier element that formed from transmutation of the ion beam from its reaction with hydrogen. 19

F Nuclear reaction analysis. Initially, 19F was the ion of choice for analysis of hydrogen in solids. The reaction involves the interaction of 19F with 1H to yield an 16O atom plus an alpha particle and a gamma ray. In the geological sciences, the 16.4 MeV resonance has found use for measuring hydration profiles in glass such as obsidian (Lee et al. 1974) and measurements of the H concentration in synthetic and natural quartz (Clark et al. 1978). Early work on analysis of H in garnets (Rossman 1990) also used 19F, but found that the reproducibility needed improve-

12

Rossman

ment. Because some accelerators can bring the 19F ion to as much as 22 MeV, significant depth profiles are possible. 15 N Nuclear reaction analysis. The most sensitive analyses of hydrogen in minerals have been made by a nuclear resonant reaction using the 15N technique (Lanford, 1978) that is based on the nuclear reaction 1H(15N,αγ)12C. In this method (Fig. 8), the hydrogen ions in the sample (the target) interact with a beam of 15N ions and are transmuted into 16O that immediately decays through alpha decay into 12C in a nuclear excited state. The 12 C has a decay path that emits a gamma ray that is detected in the analysis. The number of 12C gamma rays is proportional to the amount of hydrogen in the sample and does not depend on the chemical species of the hydrous component. A single calibration point is all that is needed to use the method for quantitative analysis of hydrogen.

Figure 8. The nuclear reaction scheme in the 15 N nuclear reaction method. The reaction of 15 N and 1H produce 16O in a nuclear excited state. A decay path of oxygen produces 12C, which comes to the ground nuclear state with the emission of a 4.44 MeV gamma ray that is the analytical signal.

The methods for mineral analysis were initially refined at Caltech and later, when the Caltech accelerator shut down, were transferred to the accelerator laboratory of the Institut für Kernphysik, Frankfurt am Main, where a beam of 15N2+ ions was delivered by a 7-MeV Van de Graaff accelerator onto a sample under high vacuum. At Frankfurt, the apparatus was specially designed and modified for the analysis of low hydrogen concentrations (to 10 ppm wt) in mineral samples. A detailed description of the experimental design can be found in the works of Endisch et al. (1993, 1994). Salient aspects include a Pb-shielded bismuth germanate (BGO) scintillation detector with an anticoincidence counting system for reduction of cosmic ray background, with the sample holder placed in an ultra-high-vacuum (10−10 mbar) chamber. The NPA method for low concentrations of H in minerals has been under development since the late 1970’s. Initially, F-19 was the ion beam of choice, but with the discovery of weak, interfering reactions, the ion beam was changed to N-15. Initially, weak nuclear reactions from carbon contamination were problematic, but improved detection methods, improved instrument vacuum and trapping of carbon compounds in the sample chamber brought them down to a manageable level (Kuhn et al. 1990). Ultimately, the layer of hydrous materials on the surface of the sample became the limiting problem, but high voltage ion sputtering was able to reduce this limitation to low levels (Maldener and Rauch 1997). An additional modification described by Maldener and Rauch allowed accurate sample positioning by Rutherford backscattering. Despite the extensive measures employed to minimize background hydrogen, a finite background or blank level may contribute to the amount of hydrogen measured. Due to the evolving methods of background reduction, the absolute background contribution to each analysis was subject to some degree of variation. One of the key calibrations for olivine was establish using this method (Fig. 9). In the most recent set of procedures, analysis of anhydrous silica glass and a silicon wafer placed the background estimate at 2 ± 2 ppm H2O. In late 2004, the accelerator at Frankfurt was decommissioned and work there on hydrogen in minerals has ceased. During the lifetime of the Frankfurt facility, the nuclear profile analysis method has been applied a variety of minerals including garnets (Rossman et al. 1988; Maldener et al. 2003), olivines (Bell et al. 2003), kyanite (Bell et al. 2004), rutile and cassiterite (Maldener et al. 2001), titanite (Hammer et al. 1996), ortho- and clinopyroxenes and zircon (Rossman et al. in prep.).

Analytical Methods

13

Figure 9. The olivine calibration established with 15N nuclear reaction analysis. This calibration relates the total integrated absorption of the infrared spectrum to the water content determined by NRA. From Bell et al. 2003.

Other workers have used the NPA method for analysis of H in minerals and geological materials. Rauch et al. (1992) used the 15N method to determine the hydration of tektite glass. Semi-quantitative hydrogen concentration depth profiles were obtained on forsterite crystals by Fujimoto et al. (1993). They treated crystals under water at different pH and temperature conditions and found that high surface hydrogen concentrations developed. Under medium to high pH conditions at 25 °C, they found that the hydrogen-rich region extended less than 20 nm into the surface while at low pH conditions; it reached as deep as 200 nm. Elastic recoil detection analysis (ERDA). Methods based on the scattering of nuclei by protons are also used in the analysis of minerals. A particularly promising method is known as Elastic Recoil Detection analysis (Barbour et al. 1995; Sie et al. 1995). This method (Fig. 10) involves using 2 MeV 4He+ ion beam that is focused on the polished surface of the sample at a low angle (15°). Forward scattered 1H+ ions that come from the hydrous component in the mineral (the recoil spectrum) are detected by a silicon ERDA detector. Because the forward scattered protons loose energy as they traverse through the thickness of the sample, their energy at the

Figure 10. A diagram of a typical ERDA sample chamber where a beam of 2 MeV 4He ions are scattered at low angles by protons in the sample. Modified after Fig. 1 of Sweeney et al. 1997.

14

Rossman

detector is a function of the depth of interaction with the 4H+ ion. Sweeney et al. (1997) used a microbeam elastic recoil detection analysis to determine the hydrogen content of minerals. With suitable calibration, a depth profile (Fig. 11) as well as the absolute H-concentration can be obtained, in principle. In ERDA, there is a problem of H-loss due to diffusion away from the He+ beam, but it is quantifiable and the technique is readily applicable to the analysis of H in both hydrous and nominally anhydrous minerals down to the 0.04 wt% (400 ppm wt) level. Sweeney et al. state that this detection limit is potentially improvable with better protected electronics. Proton-proton scattering. Furuno et al. (2003) describe an application of a proton–proton elastic recoil coincidence spectroscopy to hydrogen analysis using a proton microbeam at an energy of 20 MeV. This method provides depth profiles of hydrogen over a thickness of 200 μm of silicate samples in a short time. A typical beam size is as small as 27 × 32 μm. The depth resolution is about 10 μm. The present work proves that the proton–proton elastic recoil coincidence spectroscopy is a promising method for measurements of hydrogen in mineral and rock samples with thickness up to 200 μm. Proton beams at energies of 20 MeV can pass through several hundred micrometers with an energy loss of only a few MeV. Protons passing through a sheet of material experience proton-proton elastic recoil. Measurement of the energy-loss distribution from the protonproton scattering events is specific for H and has a sensitivity in the ppm range (Cohen et al. 1972). A typical detection system (Fig. 12) consists of two detectors that detect scattered protons in coincidence with the recoil protons. If the detectors are the same distance from the sample, both protons arrive at detectors at the same time, but with a 90° separation. Their energy will be less than the incident beam because of energy loss that is a non-linear function of the depth of the reaction below the sample surface (Fig. 13). This method was used by Wegdén et al. (2004) with a 2.8 MeV proton beam at Lund, Sweden. They were able to get strong signals from a synthetic pyroxene with 300 ppm water. Further development of this method demonstrated the analysis of hydrogen at the 100 of ppm H2O concentration level (Fig. 14) and showed that surface hydration could be distinguished from the intrinsic bulk hydrogen content (Wegdén et al. 2005) where depth profiles exceeded 1 micrometer. Reichart et al. (2004) used a similar method to produce a three-dimensional image of the hydrogen distributions in a polycrystalline synthetic CVD diamond film and showed that the hydrogen atoms were concentrated along the grain boundaries.

Figure 11. An ERDA profile of a grossular garnet 3-232 compared to a zero-hydrogen synthetic Al2O3 blank. The grossular (GRR 1386) contains 0.17 wt% H2O as was previously determined by NMR, H-extraction and FTIR. Modified after Figure 4 of Sweeney et al. 1997.

Analytical Methods

15

Figure 12. Schematic drawing of the detection system for p-p scattering. Modified after Furuno et al. (2003).

Figure 13. The results of a typical proton-proton scattering experiment on a sample containing a hydrous inclusion. Modified from Figure 4 of Furuno (2003).

Figure 14. Depth profile of the hydrogen concentration (as H) of an orthopyroxene. (25 ppm H = 223 ppm H2O). Modified after Figure 1 of Wegdén et al. (2005).

16

Rossman

In each of these examples of p-p scattering, the potential of the method for geologic samples was clearly demonstrated, but as was the case of the NRA in the early 1980’s, significant effort will be required before it becomes a rigorous, accurate analytical technique. Other approaches have been suggested (Wirth 1997) such as electron energy-loss spectroscopy (EELS), but have not been developed into accurate analytical methods for hydrogen in the nominally anhydrous minerals. Hopefully, geoscientists will remain associated with the nuclear physics community to bring these promising tools into the realm of a routinely useable analytical instrument.

Nuclear magnetic resonance At first, one would think that proton nuclear magnetic resonance (1H-NMR), should be an ideal method for studying H in minerals if the content of iron and other paramagnetic ions is low (less than about 0.4 wt% FeO). Although proton NMR is widely used in the chemical sciences, it has seen comparatively little application to the low concentrations of water in the nominally anhydrous minerals in part because of its low sensitivity for protons. A major challenge to the investigation of nominally anhydrous silicate minerals is overcoming or accommodating the sensitivity limits of the technique. Quantitative NMR measurements becomes difficult at H concentrations less than about 1000 ppm wt because the probe background overwhelms the sample signal unless the background is minimized through the use of pulse sequences or is somehow subtracted from the sample signal. Furthermore, the concentrations of paramagnetic transition elements are sufficiently high in most minerals that they seriously compromise or effectively eliminate the proton signal through inhomogeneous magnetic interactions. Consequently, the small amount of proton NMR conducted on minerals has been largely focused on stoichiometrically hydrous minerals and, in particular, on synthetic ones with a minimal paramagnetic component. Early NMR work on a nominally anhydrous mineral focused on the channel water in beryl where workers found the NMR signal from water in the channels and were able to conclude that the H-H vector was parallel to the c-axis (Pare and Ducros 1964; Sugitani et al. 1966; Zayarzina et al. 1969). Later work by (Carson et al. 1982) was concerned with the water in cordierite and found that the water was undergoing some kind of motion on a time scale faster than one microsecond. However, none of these studies attempted to quantitatively determine the absolute amount of water in the minerals from the NMR spectrum. Subsequent studies of beryl did distinguish between two orientations of water and determined their relative proportions (Charoy et al. 1996; Lodzinski et al. 2005). The first attempt to examine a range of nominally anhydrous minerals (Yesinowski et al. 1988) used a method known as magic angle spinning NMR. NMR spectra of solids are usually very broad due to magnetic anisotropic interactions among components of the crystals. However, high-resolution spectra can often be obtained through a method known as “magic angle” spinning NMR (MAS-NMR). In this experiment, the sample holder is rapidly spun with its axis 54.7° with respect to the applied magnetic field. If the line shape of the non-spinning sample is dominated by inhomogeneous interactions, as it often is for minerals with low hydrogen contents, magic angle spinning produces a sharp central band as well as a set of “spinning sidebands” spaced at integer multiples of the spinning frequency. Paramagnetic metal ions in the sample can complicate the NMR experiment because they introduce additional interactions with their unpaired electron spins. In addition to a number of stoichiometrically hydrous minerals, Yesinowski et al. (1988) examined microcline, quartz, and nepheline and grossular with 1H MAS NMR spectra. Although the found mostly fluid inclusions, they were able to show that different hydrous species could be distinguished but determined only their relative amounts (Fig. 15). Cho and Rossman (1993) further developed the technique for minerals and presented data on OH in grossular crystals with 0.17 to 0.31 wt% H2O. They were able to show that in low

Analytical Methods

17

water-content garnets, the mode of substitution is not dominated by the hydrogarnet substitution (H4O44−), but rather by protons in pairs (Fig. 16). Proton NMR is sensitive to just the hydrogen environment and, and is inherently quantitative. Relative amounts of various species can be determined, and, with suitable calibration, so can the absolute hydrogen content of the sample. To avoid the problem with paramagnetic components in natural samples, Kohn (1996) synthesized synthetic pyroxenes and forsterite and used NMR to study their hydrous components. He reported that they contained 0.02 to 0.24 wt% H2O. This and a subsequent report (Kohn 1998) indicated that the concentrations of hydrous components in these fine-grained materials were much higher than any earlier study suggested. Keppler and Rauch (2000) subsequently showed that polycrystalline materials have much higher water contents than the corresponding single crystal and suggested that the high water contents reported by Kohn (1996) were not representative of the true water content of the crystals. Contributions from hydrous species on grain boundaries, growth defects and submicroscopic fluid (or melt) inclusions are possible sources of these problems. Keppler and Rauch repeated an observation that this author’s group has long recognized: “measurements [of low hydrogen content] on powders are generally not reliable, no matter which analytical method is applied.” A following section discusses observations of elevated concentrations of water in mineral surfaces in more detail. An approximately universal absorption coefficient for the infrared spectra of feldspars was determined from 1H-MAS NMR spectra by Johnson and Rossman (2003). In this study, the spectra were used to determine the H concentration of three alkali feldspars and for the first time, eight plagioclase feldspars. To accurately measure structural H concentrations in

Figure 15. 1H magic angle NMR spectra at 500 MHz of (top) microcline feldspar from Lake George, Colorado and (bottom) microcline from the White Queen Mine, Pala, California. Peak A is an organic contaminant; peak B is water in fluid inclusions; peak C is a structurally bound, isolated H2O group; and peaks C* are the spinning sidebands of the structurally bound H2O group. Modified after Figure 7 of Yesinowski et al. (1988).

Figure 16. Proton MAS-NMR spectra of grossular from the Lelatema Mountains, Tanzania, with 0.17 wt% H2O showing 2 types of hydrogen. Data were obtained with 2000 scans, a 4 second delay between each scan, and a Gaussian line fit. The narrow line signal near zero KHz is from a proton either far removed from other H nuclei or is part of a mobile species within the sample. The broad band arises from pairs of protons in close proximity to each other, rather than a hydrogarnet substitution (Cho and Rossman 1993)

18

Rossman

samples with such low H contents (<1000 ppm H2O) it was necessary to eliminate the signal due to adsorbed water on the coarsely ground (45 to 149 µm particle size) NMR sample through a combination of sample handling protocols and background subtraction. Samples weighed about 150 mg and were spun at 12 kHz in a 500 MHz spectrometer. It was necessary to wait about 100 seconds between scans to allow the spin alignment to recover from the previous scan. They found that their plagioclase samples contained structural OH in the range of 210510 ppm H2O by wt. The microclines contained structural molecular water (1000-1400 ppm H2O) in the microcline and the Eifel sanidine sample contained only structural OH (170 ppm H2O). An approximately linear trend is produced when the total integrated mid-IR absorbance is plotted versus the concentration of structural H determined from NMR (OH and H2O) for both plagioclase and alkali feldspars (Fig. 17). The NMR work of Johnson and Rossman also showed that the pegmatitic and metamorphic albite samples, while transparent, contain variable (40-280 ppm H2O) concentrations of microscopic to sub-microscopic fluid inclusions. Xia et al. (2000) also reported using 1H MAS NMR to calculate the water concentrations of three anorthoclase megacrysts that contained between 365 and 915 ppm H2O. Very little additional quantitative NMR of the nominally anhydrous minerals has been presented. An earlier paper by Kalinichenko et al. (1989) reported quantitative 1H-NMR for andalusite and sillimanite, and concluded that the OH groups are bound to Si ions at concentrations of 2.0 and 1.7 wt% H2O, respectively. In light of other studies, it is unlikely that these values represent intrinsic OH in the phases, but more likely, represent alteration products. In other applications, NMR, in collaboration with other spectroscopic methods, has been used to study the dynamics of water in minerals (Winkler 1996) and for imaging of protons in geological solids (Nakashima et al. 1998).

Secondary ion mass spectrometry (SIMS) SIMS, also commonly known as the ion microprobe, has held promise as an ideal method to analyze hydrogen in minerals. An ion beam sputters ions from the sample and the ions are directed to a mass spectrometer where they are counted. The analytical volume in a conventional

Figure 17. Total polarized integrated band area in the mid-IR per cm thickness versus the concentration of H (in ppm H2O by weight) determined from NMR spectra for feldspars containing structural hydrogen. The slope of the best-fit line through the data is used to obtain the absorption coefficients (I and I’). From Fig. 5a of Johnson et al. (2003).

Analytical Methods

19

SIMS instrument is only a few tens of cubic micrometers, and the sensitivities are potentially in the few ppm range. NanoSIMS instruments have the potential to determine hydrogen in volumes on the order of tens of cubic nanometers. In practice, SIMS microanalysis for trace hydrogen in anhydrous minerals has proven challenging because of the high levels of background signals for hydrogen (Steele 1986; Yurimoto et al. 1989) and the matrix effects (Hervig et al. 1987). Most of the initial attempts to analyze hydrogen by SIMS reported detection limits of hundreds of ppm and few of these studies analyzed the samples by an independent analytical method to confirm the accuracy of low-hydrogen concentration measurements. Early efforts were directed towards the analysis of H in quartz crystals. Yurimoto et al. (1989) used SIMS to analyze hydrogen in quartz crystals and fused silica glasses and found that the hydrogen secondary ion intensities were proportional to the hydrogen contents determined by infrared (IR) absorption over the range of 5 to 3000 ppm-atomic H/Si (down to about 6 ppm H2O). Kurosawa et al. (1992, 1993) reported the successful analysis trace hydrogen in mantle olivines by carefully considering the source of problems and instituting corrective measures. They used the Cameca IMS 3f ion microprobe at the University of Tsukuba with a primary high-purity, mass-filtered 16O− ion beam that was accelerated to 14.5 keV with a beam current of about 100 nA and a spot size of 100 µm in diameter. As the methods were refined, Kurosawa et al. (1997) determined that the hydrogen content in mantle xenolithic olivines ranges from 10 to 60 ppm wt H2O, a concentration range that is consistent with the previous range of hydrogen contents obtained by IR spectra (Miller et al. 1987). However, no single sample was ever run with the two analytical methods as a crosscheck for consistency. A variety of precautions was necessary to obtain this level of sensitivity for hydrogen. Secondary ions, including 1H+, were collected from the central 60 µm region of the sputtered area using a mechanical aperture while the pressure in the sample chamber was maintained at 0.2 µPa. In addition, a cold trap of liquid nitrogen was used to improve the vacuum near the sample. The samples for SIMS measurements were coated with a thin gold film to eliminate electrostatic charging. Hydrogen amounts were determined from a calibration curve. For quantitative hydrogen analysis, the standards were H+-implanted San Carlos olivine. The method provided suitable standard materials for trace hydrogen while simultaneously resolving matrix effect problems. The calibration curve was obtained in the concentration ranges from about 2 to 1600 parts per million (ppm) H2O by weight. Kurosawa et al. (1992a) report that the reproducibility was within 10% in repeat analyses. SIMS determination of water in minerals has been practiced mostly when concentrations of water are in excess of 0.1% wt. A variety of synthetic phases such as silicate perovskite (Murakami et al. 2002) and majoritic garnet (Katayama et al. 2003) have analyzed by this method. The instrument has subsequently been used to study garnets, pyroxenes, and olivines from mantle xenoliths (Kurosawa et al. 1993, 1997). One strategy to improve SIMS analyses for hydrogen is to use 2H (deuterium) rather than H where possible. There are significant advantages with regard to background signals. For example, Pawley et al. (1993) report that background H counts are four orders of magnitude higher than the background D counts. This requires either synthesizing samples with deuterium or conducting deuterium exchange prior to analysis 1

Koga et al. (2003) were the first to report analyses of hydrogen concentrations in both natural mantle minerals and experimentally annealed crystals where the calibration was established with olivine, pyroxene, garnet, amphibole and micas that were previously calibrated by other methods. They employed stringent cleaning and drying procedures to eliminate contamination from water and organic solvents and adhesives used in sample preparation.

20

Rossman

They used a Cs+ primary ion beam rather than an O− beam that gave high hydrogen backgrounds. To minimize hydrogen backgrounds, they took extraordinary precautions. The entire Cameca 6f instrument was baked for at least 24 hours before an analytical session. The electron gun was kept on for 12 hours prior to the analysis at about 7 times the normal analytical current to desorb hydrogen. Organic adhesives were avoided and samples were mounted in indium metal. Through these precautions, they were able to reduce their background to 2 to 4 ppm wt of H on “zero”-hydrogen samples. When they examined the SIMS calibrations for nominally anhydrous minerals, they considered the consistency of the calibration lines and placed a premium on reproducing samples for which OH measurements from nuclear reaction analysis (GRR1012, KLV-23 olivines) and manometry (KBH-1 orthopyroxene, PMR-53 clinopyroxene, MON-9 garnet, hydrous phases) are available. Their method resulted in SIMS calibrations that are less likely to inherit systematic errors from a particular corroborating method. Their results were excellent (Fig. 18). Their success points to the future where SIMS determinations of hydrogen in minerals will be more widely utilized. SIMS offers the advantages of analysis of a smaller volume, and the corresponding ability to obtain finer lateral and depth resolution. Furthermore, it appears not to require orientation of intrinsically anisotropic samples. Further development of the SIMS method with additional calibrations and intercalibration with FTIR standards will be presented by Aubaud et al. (2006). They demonstrate that with careful attention to avoiding contamination and prolonged instrument bakeout, hydrogen background values equivalent to less than 5 ppm by weight H2O in olivine can be obtained. They also observed phase-specific calibration trends for minerals such as olivine, pyroxenes and garnets that varied by up to a factor of four. SIMS will probably always be a complimentary method to infrared spectroscopy because SIMS is, of course, unable to distinguish among hydrous species and cannot distinguish between intrinsic hydrogen and contaminating phases or inclusions. There is no doubt that this application of SIMS will be an area of significant growth in the future.

PREVIOUS REVIEWS OF METHODS An earlier review that covered the use of IR spectroscopy to study hydrous components in minerals was presented by Rossman (1988). Subsequent reviews have focused on various aspects of OH in the nominally anhydrous minerals (Rossman 1996, 1998; Skogby 1999;

Figure 18. SIMS calibration for garnet and a combined olivine-orthopyroxene calibration from Figure 5 of Koga et al. (2003).

Analytical Methods

21

Ingrin and Skogby 2000; Beran and Libowitzky 2003). One review that deals with analytical methods for geological samples is Ihinger et al. (1994) that concentrates on glasses. Several reviews of nuclear reactions used to analyze hydrogen and other light elements have appeared. Among the ones that deal with hydrogen are Lanford (1978, 1992) and Cherniak and Lanford (2001). Reviews focused on mineral applications are few (Ryan 2004).

SURFACE WATER Hydrous components can occur not only within a crystal, but will also occur on its surface. All of the nuclear analysis methods and SIMS show that the surfaces of minerals can contain considerably higher concentrations of hydrogen than is contained in the interior, even when under high-vacuum conditions. Bell et al. (2003) found about 20 times as much water on the surface of olivine KLV-23 as was present in the interior. The NPA analysis of Bell et al. (Fig. 19) shows that outermost 500 nm contains a highly elevated H concentration and that accurate analyses of the bulk hydrogen content should begin 1.5 to 2 µm below the surface. Similar surface concentrations were noted by Clark et al. (1978) and Dersch and Rauch (1999) on quartz samples. In their ERDA experiment with a garnet with 0.17 wt% H2O, Sweeney et al. (1997) detected about 4.5× as much water in the outermost 50 nm of the sample (Fig. 20). Likewise, Katayama et al. (2003) found that a factor of 24× greater water was liberated from the surface of a pyrope when the SIMS experiment began than when a steady state was reached after rastering the surface (Fig. 21). This illustrates why it is common practice to clean the sample by ion-rastering before analyzing the water content of low-hydrogen-content minerals. Clark et al. (1978) obtained F-19 depth profiles of quartz and determined that it exhibited a region of high H concentration near surface region Figure 19. Nuclear reaction analysis depth profile of a (down to a depth of about 200 nm), mantle olivine that shows elevated water content at the surface of the sample even while held in an ultra-high before the concentrations decreased to vacuum chamber. From Bell et al. (2003). the bulk value of the sample. While there is certainly several thousand ppm-wt H2O absorbed water on the surface of minerals while under high vacuum, there is some question of whether the ion beams drive hydrogen atoms below the surface during the analysis, or if a diffusion gradient naturally exists. Obviously, the depth of elevated hydrogen contents will strongly depend upon the quality of the surface and the amount of surface damage experienced by the sample during grinding and polishing. An extreme example of water near the surface of a mineral was demonstrated during the heating experiments on sillimanite crystals (Beran et al. 1989). The experiment consisted of obtaining the infrared spectrum of OH bands after each step in a step-heating experiment. As Figure 22 shows, the weight loss proceeded at a proportionally faster rate than the decrease of

22

Rossman

Figure 20. An ERDA profile of a grossular showing increased water at the surface. Modified after Figure 2b of Sweeney et al. (1997).

Figure 21. A SIMS analysis of pyrope that shows elevated water content at the surface of the sample. From Figure 2 of Katayama et al. (2003).

Figure 22. The results of a step-heating experiment where both the weight loss and the infrared spectrum are measured that shows much “impurity” water is lost from the sample before the intrinsic OH bands begin to decrease. From Figure 5 of Beran and Rossman (1989).

Analytical Methods

23

the OH bands. Beran et al. concluded that much of the water was held as molecular water at the edges of the crystal, probably associated with surface damage and incipient cleavages in a mineral with perfect cleavage perpendicular to the direction in which the infrared light was being transmitted through the crystal. In the infrared experiment, the OH bands were measured only in the center of the crystal, but the weight loss was occurring from both within the center (as OH groups) and from the damaged regions at the edge (mostly as molecular water). Quite likely, a similar problem contributed to the high values of OH in kyanite reported by Beran and Götzinger (1987).

CURRENT STATUS OF CALIBRATIONS A number of minerals have been calibrated sufficiently well that that their infrared spectra can be routinely used in determinations of the OH contents of important mantle phases and an assortment of crustal phases. Because the density of many of these phases is not highly variable as they are commonly encountered, a single calibration constant can provide a useful tool for many routine, practical applications. Those currently available are presented in Table 2. Several of these minerals have significant variation in the general appearance of their infrared spectra and require additional study to determine how the calibration varies with the different types of infrared spectra. It is certain that our work is far from over.

GLASSES Also worth pointing out are the extensive efforts to calibrate the IR spectra of hydrous components in geological glasses (Stolper 1982; Newman et al. 1986; Silver and Stolper 1989). Methods used to analyze volatiles in glasses were reviewed by Ihinger et al. (1994). Since the original infrared calibrations appeared, a number or refinements have appeared involving a variety of calibration methods such as Karl-Fischer titration, nuclear reaction analysis, and SIMS (Ohlhorst et al. 2001; Hauri et al. 2002; Mandeville et al. 2002; Leschik et al. 2004; Okumura and Nakashima 2005). The glass calibrations have made it possible to examine melt inclusions in minerals and to study partitioning of water between crystal and melt. Table 2. Representative calibration formulas for H2O and OH in minerals* Forsterite

H2O (ppm wt) = 0.188 × Abstot (integrated per cm)

Bell et al. (2003)

Kyanite

H2O (ppm wt) = 0.147 × Abstot (integrated per cm)

Bell et al. (2004)

Vesuvianite

H2O (ppm wt) = 0.085 × Abstot (integrated per cm)

Bellatreccia et al. (2005)

Nepheline

H2O (ppm wt) = 0.672 × Abstot (integrated per cm)

Beran and Rossman (1989)

Orthopyroxene

H2O (ppm wt) = 0.067 × Abstot (integrated per cm)

Bell et al. (1995)

Clinopyroxene

H2O (ppm wt) = 0.141 × Abstot (integrated per cm)

Bell et al. (1995)

Pyrope

H2O (ppm wt) = 0.240 × Abstot (integrated per cm)

Bell et al. (1995)

Grossular

H2O (ppm wt) = 0.140 × Abstot (integrated per cm)

Rossman and Aines (1991)

Hydrogrossular

H2O (ppm wt) = 0.264 × Abstot (integrated per cm)

Rossman and Aines (1991)

Spessartine

H2O (ppm wt) = 0.125 × Abstot (integrated per cm)

Rossman (1988)

Feldspars

H2O (ppm wt) = 0.065 × Abstot (integrated per cm)

Johnson and Rossman (2003)

Rutile

H2O (ppm wt) = 0.110 × Abstot (integrated per cm)

Maldener et al. (2001)

Cassiterite

H2O (ppm wt) = 0.039 × Abstot (integrated per cm)

Maldener et al. (2001)

*These formula are representative only for minerals that have spectra close to those of the standard minerals used in the calibrations. In all cases, they represent the sum of the integrated absorption intensities in the OH region for three orientations.

24

Rossman ACKNOWLEDGMENTS

The results in this chapter from the author’s laboratory have been supported for many years by the National Science Foundation (USA), most recently by grant EAR-0337816. The contributions of the author’s students and postdocs, visitors and collaborators have been pivotal in the establishment of quantitative H determinations and are gratefully acknowledged. The collaboration of Prof. Friedel Rauch (Frankfurt, Germany) and his students with nuclear analyses has been invaluable and proved to be the key to quantitative determinations at low concentrations.

REFERENCES Aines RD, Rossman GR (1984) Water content of mantle garnets. Geology 12:720-723 Asimow PD, Stein LC, Mosenfelder JL, Rossman GR (2006) Quantitative polarized infrared analysis of trace OH in populations of randomly oriented mineral grains. Am Mineral 91:278-284 Aubaud C, Withers AC, Hirschmann MM, Guan Y, Leshin LA, Mackwell SJ, Bell DR (2006) Intercallibration of FTIR and SIMS for hydrogen measurements in glasses and nominally anhydrous minerals. Am Mineral. submitted. Barbour JC, Doyle BL (1995) Elastic Recoil Detection: ERD (or Forward Recoil Spectrometry: FRES). In: Handbook of Modern Ion Beam Analysis. Tesmer JR et al. (eds) Materials Research Society, p 83-138 Bell DR, Ihinger PD, Rossman GR (1995) Quantitative analysis of trace OH in garnet and pyroxenes. Am Mineral 80:463-474 Bell DR, Rossman GR, Maldener J, Endish D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Resch 108:ECV 8-1 - 8-9. doi:10.1029/2001JB000679, 2003 Bell DR, Rossman GR, Maldener J, Endish D, Rauch F (2004) Hydroxide in kyanite: A quantitative determination of the absolute amount and calibration of the IR spectrum. Am Mineral 89:998-1003 Bellamy LJ, Owen AJ (1969). A simple relationship between the infrared stretching frequencies and the hydrogen bond distances in crystals. Spectrochim Acta A25:329-333. Beran A, Zemann J (1969a) Messung des Ultrarot-Pleochroismus von Mineralen. XI. Der Pleochroismus der OH Streckfrequenz in Andalusite. Tschermaks Miner Petr Mitt 13:285-292 Beran A, Zemann J (1969b) Ultrarotspektroskopische Untersuchung über den OH-Gehalt von Rutile, Anatas, Brookite, Cassiterite. Österreich Akad Wiss, Sitzung Juni, 27:165-167 Beran A (1970a) Messung des Ultrarot-Pleochroismus von Mineralen. IX. Der Pleochroismus der OHStreckfrequenz in Titanit. Tschermaks Miner Petr Mitt 14:1-5 Beran A (1970b) Ultrarotspektroskopischer Nachweis von OH-Gruppen in den Mineralen der Al2SiO5Modifikationen. Österr Akad Wiss, Math-naturwiss K2, Anzeiger Jg 1970:184-185 Beran A (1970c) Messung des Ultrarot-Pleochroismus von Mineralen. XII. Der Pleochroismus der OHStreckfrequenz in Disthen. Tschermaks Miner Petr Mitt:16:129-135 Beran A (1970d) Messung des Ultrarot-Pleochroismus von Mineralen. XIII. Der Pleochroismus der OHStreckfrequenz in Axinit. Tschermaks Miner Petr Mitt 15:71-80 Beran A, Zemann, J. (1971) Messung des Ultrarot-Pleochroismus von Mineralen. XI. Der Pleochroismus der OH-Streckfrequenz in Rutil, Anatas, Brookit und Cassiterit. Tschermaks Miner Petr Mitt 15:71-80 Beran A, Götzinger MA (1987) The quantitative IR spectroscopic determination of structural OH groups in kyanites. Mineralogy Petrology 36:41-49 Beran A, Rossman GR, Grew ES (1989) The hydrous component of sillimanite. Am Mineral 74:812-817 Beran A, Rossman GR (1989) The water content of nepheline. Mineral Petrology 40:235-240 Beran A and Libowitzky E (2003) IR spectroscopic characterization of OH defects in mineral phases. Phase Transitions 76:1-15 Brunner GO, Wondratschek H, Laves F (1961) Infrared studies on the incorporation of H in natural quartz. Z Elektrochem Angewandte Physik Chemie 65:735-50 Carson DG, Rossman GR, Vaughan RW (1982) Orientation and motion of water molecules in cordierite: A proton nuclear magnetic resonance study. Phys Chem Mineral 8:14-19 Charoy B, de Donato P, Barres O, Pinto-Coelho C (1996) Channel occupancy in an alkali-poor beryl from Serra Branca (Goias, Brazil): spectroscopic characterization. Am Mineral 81:395-403 Cherniak DJ, Lanford WA (2001) Nuclear reaction analysis. In: Non-Destructive Elemental Analysis. Alfassi ZB (ed) Blackwell Science Ltd., p 308-338 Cho H, Rossman GR (1993) Single-crystal NMR studies of low-concentration hydrous species in minerals: Grossular garnet. Am Mineral 78:1149-1164

Analytical Methods

25

Clark GJ, White CW, Allred DD, Appleton BR, Tsong IST (1978) Hydrogen concentration profiles in quartz determined by a nuclear reaction technique. Phys Chem Mineral 3:199-211 Cohen BL, Fink, CL, Degnan JH (1972) Nondestructive analysis for trace amounts of hydrogen. J Appl Phys 43:19-25 Dersch O, Rauch F (1999) Water uptake of quartz investigated by means of ion-beam analysis. Fresenius J Anal Chem 365:114–116 Dodd DM, Fraser DB (1965) The 3000-3900 cm−1 absorption bands and aneleasticity in crystalline α-quartz. Phys Chem Solids 26:673-86 Dodd DM, Fraser DB (1967) Infrared studies of the variation of H-bonded OH in synthetic alpha-quartz. Am Mineral 52:149-160 Dyar MD, Martin SV, Mackwell LSJ, Carpenter S, Grant CA, McGuire AV (1996) Crystal chemistry of Fe3+, H+, and D/H in mantle-derived augite from Dish Hill: Implications for alteration during transport. Mineral Spectroscopy: A Tribute to Roger G. Burns. The Geochemical Society, Special Publication No. 5:289-303. Dyar MD, McCammon C, Schaefer MW (eds) Eiler JM, Kitchen N (2001) Hydrogen-isotope analysis of nanomole (picoliter) quantities of H2O. Geochim Cosmochim Acta 65:4467-4470 Endisch D, Sturm H, Rauch F (1993) Development of a measuring set-up for high-sensitivity analysis for hydrogen by the nitrogen-15 nuclear reaction technique. Fresenius J Anal Chem 346:205-207 Endisch D, Sturm H, Rauch F (1994) Nuclear reaction analysis of hydrogen at levels below 10 at.ppm, Nucl Instrum Methods Phys Res Sect B 84:380–392 Farrell EF, Newnham RE (1967) Electronic and vibrational absorption spectra in cordierite. Am Mineral 52: 380-388 Flörke DW, Köhler-Herbertz B, Langer K, Törges I (1982) Water in microcrystalline quartz of volcanic origin: Agates. Contrib Mineral Petrog 80:329-333 Fujimoto K, Fukutani K, Tsunoda M, Yamashita H, Kobayashi K (1993) Hydrogen depth profiling using 1H(15N, αγ)12C resonant nuclear reaction on water-treated olivine surfaces and characterization of hydrogen species. Geochem J 27:155-162 Furuno K, Komatsubara T, Sasa K, Oshima H, Yamato Y, Ishii S, Kimura H, and Kurosawa M (2003) Measurement of hydrogen concentration in thick mineral or rock samples. Nucl Instr Meth Phys Research Sect B 210:459-463 Griggs DT, Blacic JD (1965) Quartz: anomalous weakness of synthetic crystals. Science 147:292-295. Hammer VMF, Beran A, Endisch D, Rauch F (1996) OH concentrations in natural titanites determined by FTIR spectroscopy and nuclear reaction analysis. Eur J Mineral 8:281–288 Hauri E, Wang JH, 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 Hervig RL, Stanton TR, Williams P (1987) Ion probe microanalyses of hydrogen in glasses and minerals (abstr). EOS 68:441 Hofmeister AM, Rossman GR (1985a) A model for the irradiative coloration of smoky feldspar and the inhibiting influence of water. Phys Chem Mineral 12:324–332 Hofmeister AM, Rossman GR (1985b) A spectroscopic study of irradiation coloring of amazonite: structurally hydrous, Pb-bearing feldspar. Am Mineral 70:794–804 Holdaway MJ, Dutrow BL, Borthwick J, Shore P, Harmon RS, Hinton RW (1986) H content of staurolite as determined by H extraction line and ion microprobe. Am Mineral 71:1135-1141 Ilchenko EA, Korzhinskaya VS (1993) Hydroxyl groups in synthetic and natural zircons. Mineralogicheskii Zhurnal 15:26-39 Ihinger PD, Hervig RL, McMillan PF (1994) Analytical methods for volatiles in glasses. Rev Mineral 30:67121 Ingrin J, Skogby H (2000). Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Johnson EA, Rossman GR (2003) The concentration and speciation of hydrogen in feldspars using FTIR and 1H MAS NMR spectroscopy. Am Mineral 88:901-911 Kalinichenko AM, Katalenets AI, Proshko VY, Pasal’skaya LF (1989) Water in minerals of Al2SiO5 composition (based on the hydrogen-1 and silicon-29 NMR spectra). Geokhimiya 1989:1024-8 (in Russian) Katayama I, Hirose K, Yurimoto H, Nakashima S (2003) Water solubility in majoritic garnet in subducting oceanic crust. Geophys Resch Letts 30:2155: doi:10.1029/2003GL018127, 2003 Kats A, Haven Y (1960) Infrared absorption bands in α-quartz in the 3-µ region. Phys Chem Glasses 1:99-102 Kats A (1962) Hydrogen in α-quartz. Philips Research Reports 17:133-195, 201-279 Keppler H, Rauch M (2000) Water solubility in nominally anhydrous minerals measured by FTIR and 1H MAS NMR: the effect of sample preparation. Phys Chem Mineral 27:371-376 King JC, Wood DL, Dodd DM (1960) Infrared and low-temperature acoustic absorption in synthetic quartz. Phys Rev Lett 4:500-501

26

Rossman

King PL, Vennemann TW, Holloway JR, Hervig RL, Lowenstern JB, Forneris JF (2002) Analytical techniques for volatiles: A case study using intermediate (andesitic) glasses. Am Mineral 87:1077-1089 Kirby SH, McCormick JW (1979) Creep of hydrolytically weakened synthetic quartz crystals oriented to promote {2-1-10}<0001> slip: a brief summary of work to date. Bull Minéral 102:124-137 Koga K, Hauri E, Hirschmann M, Bell D (2003) Hydrogen concentration analyses using SIMS and FTIR: Comparison and calibration for nominally anhydrous minerals. Geochem Geophys Geosyst 4: doi: 10.1029/2002GC000378 Kohn SC (1996) Solubility of H2O in nominally anhydrous mantle minerals using 1H MAS NMR. Am Mineral 81:1523-1526 Kohn SC (1998) 1H MAS NMR studies of water solubilities and dissolution mechanisms in olivine, clinopyroxene and orthopyroxene. Mineral Mag 62A:799-800 Kuhn D, Rauch F, Baumann H (1990) A Low-background detection system using a BGO detector for sensitive hydrogen analysis with the 1H (15N, alpha-gamma)12C reaction. Inst Methods Phys Res B 45:252-255 Kurosawa M, Yurimoto H, Matsumoto K, Sueno S (1992) Hydrogen analysis of mantle olivine by secondary ion mass spectrometry. In: High-Pressure Research in Mineral Phys: Application to Earth and Planetary Sciences. Syono S, Manghnani MH (eds) Terra Pub-Am Geophys Union, p 283–287 Kurosawa M, Yurimoto H, Matsumoto K, Sueno S (1993) Water in Earth’s mantle: hydrogen analysis of mantle olivine, pyroxenes and garnet using the SIMS. Proc Lunar Planet Sci Conf 24th, 839–840 Kurosawa M, Yurimoto H, Sueno S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Mineral 24:385–395 Lanford WA (1978) 15N Hydrogen profiling: Scientific applications, Nucl Instr Meth Phys Res B149:1–8 Lanford WA (1992) Analysis for hydrogen by nuclear reaction and energy recoil detection. Nucl Instrum Methods in Phys Res Sect B 66:65-82 Langer K, Flörke OW (1974) Near infrared absorption spectra (4000-9000 cm−1) of opals and the role of “water” in these SiO2·nH2O minerals. Fortschr Minereral 52:17-51 Lee RR, Leich, DA, Tombrello TA, Ericson JE, Friedman I (1974) Obsidian hydration profile measurements using a nuclear reaction technique. Nature 250:44-7 Leschik M, Heide G, Frischat GH, Behrens H, Wiedenbeck M, Wagner N, Heide K, Geissler H, Reinholz U (2004) Determination of H2O and D2O contents in rhyolitic glasses. Phys Chem Glasses 45:238-251 Libowitzky E, Rossman GR (1996) Principles of quantitative absorbance measurements in anisotropic crystals. Phys Chem Mineral 23:319-327 Libowitzky E, Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82:11111115 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H···O hydrogen bond lengths in minerals. Monatshefte für Chemie 130:1047-1059 Lodzinski M, Sitarz M, Stec K, Kozanecki M, Fojud Z, Jurga, S (2005) ICP, IR, Raman, NMR investigations of beryls from pegmatites of the Sudety Mts. J Mol Struct (2005) 744-747:1005-1015 Lutz HD (1995) Hydroxide ions in condensed materials - correlation of spectroscopic and structural data. Struct Bonding 82:85-103 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90:1319-1333 Maldener J, Rauch F (1997) High energy ion-beam analysis in combination with keV sputtering, in Application of Accelerators in Research and Industry. Duggan JL, Morgan IL (eds) AIP Press, p. 689–692 Maldener J, Rauch F, Gavranic M, Beran A (2001) OH absorption coefficients of rutile and cassiterite deduced from nuclear reaction analysis and FTIR spectroscopy. Mineral Petrol 71:21–2 Maldener J, Hösch A, Langer K, Rauch F (2003) Hydrogen in some natural garnets studied by nuclear reaction analysis and vibrational spectroscopy. Phys Chem Mineral 30:337–344 Mandeville CW, Webster JD, Rutherford MJ, Taylor BE, Timbal A, Faure K (2002) Determination of molar absorptivities for infrared absorption bands of H2O in andesitic glasses. Am Mineral 87:813-821 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554–570 Miller GH, Rossman GR, Harlow GE (1987) The natural occurrence of hydroxide in olivine. Phys Chem Mineral 14: 461–47 Murakami M, Hirose K, Yurimoto H, Nakashima S, Takafuji N (2002) Water in earth’s lower mantle. Science 295:1185-1187 Nakamoto K, Margoshes M, Rundle RE (1955) Stretching frequencies as a function of distances in hydrogen bonds. J Am Chem Soc 77:6480-6486 Nakashima Y, Nakashima S, Gross D, Weiss K, Yamauchi K (1998) NMR imaging of 1H in hydrous minerals. Geothermics 27:43-53 Newman S, Stolper EM, Epstein SR (1986) Measurement of water in rhyolitic glasses: calibration of an infrared spectroscopic technique. Am Mineral 71:1527-1541 Novak A (1974) Hydrogen bonding in solids. Correlation of spectroscopic and crystallographic data. Structure and Bonding 18:177-216

Analytical Methods

27

Ohlhorst S, Behrens H, Holtz F (2001) Compositional dependence of molar absorptivities of near-infrared OHand H2O bands in rhyolitic to basaltic glasses. Chem Geol 174:5-20 Okumura S, Nakashima S (2005) Molar absorptivities of OH and H2O in rhyolitic glass at room temperature and at 400-600 °C. Am Mineral 90:441-447 O’Leary JA, Rossman GR, Eiler J (2006) Hydrogen analysis in minerals by continuous-flow mass spectrometry. Am Mineral. in prep. Pare X, Ducros P (1964) Nuclear magnetic resonance study of beryl water. Bull Soc Franc Mineral Crist 87: 429-33 (in French) Paterson MS (1982). The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bull Min ral 105:20-29 Pawley AR, McMillan PF, Holloway JR (1993) Hydrogen in stishovite, with implications for mantle watercontent. Science 261:1024-1026 Rauch F, Ericson JE, Wagner W, Grimm-Leimsner C, Livi RP, Shi Chengru, Tombrello TA (1992) Hydration of tektite glass. J Non-Cryst Solids 144:224-30 Reichart P, Datzmann G, Hauptner A, Hertenberger R, Wild C, Dollinger G (2004) Three-dimensional hydrogen microscopy in diamond. Science 306:1537-1540 Rossman GR (1975) Joaquinite: the nature of its water content and the question of four-coordinated ferrous iron. Am Mineral 60:435-440 Rossman GR, Rauch F, Livi R, Tombrello TA, Shi CR, Zhou ZY (1988) Nuclear reaction analysis of hydrogen in almandine, pyrope and spessartite garnets. N Jb Miner Mh 1988:172–178 Rossman GR (1988) Vibrational Spectroscopy of Hydrous Components. Rev Mineral 18:193-206 Rossman GR (1990) Hydrogen in “anhydrous” minerals. Nucl Instr Meth Phys Res Sect B 45:41-44 Rossman GR, Aines RG (1991) The hydrous components in garnets: grossular-hydrogrossular. Am Mineral 76: 1153-1164 Rossman GR (1996) Studies of OH in nominally anhydrous minerals. Phys Chem Mineral 23:299-30 Ryan CG (2004) Ion beam microanalysis in geoscience research. Nucl Instr Meth Phys Res Sect B 219-220: 534-549 Schreyer W, Yoder HS (1964) The system Mg cordierite-H2O and related rocks. N Jb Mineralogie Abh 101: 271-342 Sie SH, Suter G, Chekmir A, Green TH (1995) Microbeam recoil detection for the study of hydration of minerals. Nucl Instr Methods Phys Res Sect B 104:261-26 Silver L, Stolper E (1989) Water in albitic glasses. J Petrol 30:667-709 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: Variations in the natural environment. Am Mineral 75:764-774 Skogby H (1999) Water in nominally anhydrous minerals. In: Microscopic Properties and Processes in Minerals. NATO Science Series. Wright K, Catlow R (eds), Kluwer Acad. Publishers, p 509-522 Steele LM (1986) Ion probe determination of hydrogen in geologic samples. N Jb Mineral Mh 1986:193–202 Stolper EM (1982) Water in silicate glasses: an infrared spectroscopic study. Contrib Mineral Petrol 81:1-17 Sugitani Y, Nagashima K, Fujiwara S (1966) NMR (nuclear magnetic resonance) analysis of the water of crystallization in beryl. Bull Chem Soc Japan 39:672-4 Sweeney RJ, Prozesky VM, Springhorn KA (1997) Use of the elastic recoil detection analysis (ERDA) microbeam technique for the quantitative determination of hydrogen in materials and hydrogen partitioning between olivine and melt at high pressures. Geochim Cosmochim Acta 61:101-113 Wegdén M, Kristiansson P, Pastuovic Z, H. Skogby, Skogby H, Auzelyte V, Elfman M, Malmqvist KG, Nilsson C, Pallon J, Shariff A (2004) Hydrogen analysis by p–p scattering in geological material Nucl Instr Meth Phys Res Sect B 219–220:550-554 Wegdén M, Kristiansson P, Skogby H, Auzelyte V, Elfman M, Malmqvist KG, Nilsson C, Pallon J, Shariff A (2005) Hydrogen depth profiling by p-p scattering in nominally anhydrous minerals. Nucl Instr Meth Phys Res Sect B 231:524-529 Wilkins RWT, Sabine W (1973) Water content of some nominally anhydrous silicates. Am Mineral 58: 508–516 Winkler B (1996) The dynamics of H2O in minerals. Phys Chem Mineral 23:310-318 Wirth R (1997) Water in minerals detectable by electron energy-loss spectroscopy EELS. Phys Chem Minerals 24:561-568 Wood DL, Nassau K (1967) Infrared spectra of foreign molecules in beryl. J Chem Phys 47:2220-8. Woodhead JA, Rossman GR, Thomas AP (1991) Hydrous species in zircon. Am Mineral 76:1533-1546 Xia Q, Pan Y, Chen D, Kohn S, Zhi X, Guo L, Cheng H, Wu Y (2000) Structural water in anorthoclase megacrysts from alkalic basalts: FTIR and NMR study. Acta Petrologica Sinica 16:485–491 (in Chinese) Xiong F, Rauch F, Shi C; Zhou Z, Livi RP, Tombrello TA (1987) Hydrogen depth profiling in solids: a comparison of several resonant nuclear reaction techniques. Nucl Instr Meth Phys Res Sect B 27:432-41

28

Rossman

Yesinowski JP, Eckert H, Rossman GR (1988) Characterization of hydrous species in minerals by high-speed 1H MAS-NMR. J Am Chem Soc 110:1367-1375 Yurimoto H, Kurosawa M, Sueno S (1989) Hydrogen analysis in quartz crystals and quartz glasses by secondary ion mass spectrometry. Geochim Cosmochim Acta 53:751-755 Zavarzina NI, Gabuda SP, Bakakin VV, Rylov GM (1969) N.M.R. analysis of water in beryls. Zh Struktur Khimii 10:804-810 (in Russian)

6

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 117-154, 2006 Copyright © Mineralogical Society of America

Water in Nominally Anhydrous Crustal Minerals: Speciation, Concentration, and Geologic Significance Elizabeth A. Johnson* Department of Earth and Space Sciences University of California, Los Angeles Los Angeles, California, 90095, U.S.A. e-mail: [email protected] (*present address: Dept. of Geology & Environmental Sciences, James Madison Univ., Harrisonburg, VA, 22807)

INTRODUCTION Importance of nominally anhydrous minerals in the crust Why should we be interested in trace hydrous species in nominally anhydrous minerals in the Earth’s crust? After all, hydrous minerals dominate the pedosphere and are abundant to fairly common trace minerals in many metamorphic and igneous crustal rocks. On the other hand, the most abundant minerals in the crust—feldspars, quartz, pyroxenes, and garnet—are all nominally anhydrous. They are present even in systems with low total volatiles or fluid contents, or environments with low water activities where hydrous minerals are unstable. These nominally anhydrous minerals provide an opportunity to expand the extent of our knowledge of fluid composition and water activity, as well as the influence of water on physical properties and geochemical signatures of rocks. One advantage to investigations of the crustal component of the lithosphere is that many parts of the crust (especially the continental crust) are available for direct study in outcrops at the surface of the Earth. This allows the nominally anhydrous mineral and its hydrous species to be placed into the context of the hand sample, the outcrop, and even the regional geology.

Scope and goals of this chapter It would be unrealistic to try to cover every water-bearing mineral in the Earth’s crust in this chapter. I have limited my discussion to minerals that do not require hydrous species to complete their stoichiometry, and those for which research has been completed on natural crustal samples. These minerals are: quartz, the feldspars, nepheline, pyroxenes, garnets (except pyrope), kyanite, andalusite, sillimanite, rutile, cassiterite, zircon, titanite, cordierite, and beryl. This selection of minerals restricts the discussion primarily to the continental crust below about 3 km depth. Some references to eclogitic and mantle-wedge minerals are included for completeness. This is a fairly new field of study, and as such, the goal of this chapter is to give an overview of the work that has been done, and more importantly, provide directions for future work. The chapter begins with an overview of the types of hydrous species and the range of absolute concentrations for each mineral or mineral family. The second section provides examples of applications of these measurements to problems of geologic interest. It is assumed that the reader is familiar with the compositions and general structure and crystal chemistry of these minerals. It is also helpful to have a general understanding of absolute OH concentration measurement techniques, and a reading knowledge of polarized infrared spectra of hydrous species in minerals. Overviews of these topics may be found in 1529-6466/06/0062-0006$05.00

DOI: 10.2138/rmg.2006.62.6

118

Johnson

Rossman (1988, 2006); Libowitzky and Beran (2006); Smyth (2006). Hydrogen abundance measurements discussed in this chapter are generally obtained using manometric or infrared spectroscopic methods. A summary of infrared spectroscopic calibrations for common mineral species is given in Table 2 of Rossman (2006). The reader should consult the reference of interest for detailed information about the absorption coefficient used in a particular study.

HYDROUS SPECIES AND CONCENTRATIONS IN CRUSTAL MINERALS Quartz and coesite Quartz, a common crustal mineral, contains structural OH groups, macroscopic fluid inclusions, and nanoscale “fluid inclusions” or water clusters (especially seen in synthetic quartz). Previous studies have compiled detailed summaries of the hydrous species in natural and synthetic quartz, chert, opal, and chalcedony (Aines and Rossman 1984a; Rossman 1988). The infrared spectrum of OH in a natural quartz crystal from Brazil is shown in Figure 1. Diffusion and electrolytic exchange experiments in natural and synthetic α-quartz have established that these sharp bands are due to OH groups associated with other H+ or monovalent cations including Li+, Na+, K+, Cu+, and Ag+, and hydroxyl associated with Al3+ (Kats 1962; Aines and Rossman 1984a; Rovetta et al. 1986; Miyoshi et al. 2005). This structural OH is most commonly found in large, clear, undeformed quartz crystals from high-temperature pegmatites as well as synthetic quartz, although some structural OH bands may occur in spectra of other low-temperature forms of quartz (such as amethyst) (Aines and Rossman 1984a; Kronenberg and Wolf 1990). The OH bands in quartz have been calibrated (Chakraborty and Lehmann 1976) and the reported range of OH concentrations is <1 to ~40 ppm H2O wt. (1-270 H/106 Si; Table 1) (Chakraborty and Lehmann 1976; Rovetta et al. 1986; Kronenberg and Wolf 1990; Grant et al. 2003). The structural OH bands broaden and merge together upon heating quartz to just below the α-β transition temperature (586 °C) (Aines and Rossman 1985). Quartz can also hold up to 8000 ppm H2O wt. (0.8 wt%) in the form of submicroscopic fluid inclusions (Kronenberg and Wolf 1990). Natural quartz always contains water inclusions that behave as fluid- i.e., they freeze to ice at low temperatures (Kronenberg and Wolf 1990). On the other hand, synthetic quartz contains “clusters” of water molecules that do not transform to ice upon freezing (Aines et al. 1984; Aines and Rossman 1984a; Kronenberg and Wolf 1990; Cordier and Doukhan 1991). Although not strictly structurally incorporated water, these fluid inclusions or water clusters have a large effect on the physical properties of quartz (see discussion below and Appendix for a list of studies).

Quartz Brazil 78 K

1.40

Absorbance

1.20 1.00

Figure 1. Polarized (E perpendicular to c) infrared spectrum of OH in a natural Brazilian quartz crystal. Sample is 5 mm thick; spectrum was obtained at 78 K. Data replotted from Aines and Rossman (1984a).

0.80 0.60 0.40 0.20 0.00 3600

3400

3200 -1

Wavenumbers (cm )

3000

Water in Nominally Anhydrous Crustal Minerals

119

Table 1. OH concentrations in nominally anhydrous crustal minerals. Mineral Quartz Coesite Feldspar

Nepheline Clinopyroxene Orthopyroxene Garnet* Kyanite Sillimanite Andalusite† Rutile Cassiterite Zircon Cordierite

Hydrous species OH fluid inclusions OH OH H2O NH4+ sub-µm inclusions H2O OH OH OH OH OH OH OH OH crystalline OH secondary OH H2O

[OH} range (ppm H2O wt.) Typical

Maximum

<1-40 0-8000 0 0-510 135-1350 450-1500 0-2300 500-3500 <5-240 26-117 0-4000 <3-44 0-137 270 310-620 2-120 0-100 0-1000 3000-24000

40 8000 135† 915 1350 buddingtonite 4265 5500 466 350 14400 44 200 270 723 170 100 1000 24000

*Macroscopic garnets. † Only one data point available.

Natural coesite found in eclogite assemblages contains no OH within detection limits (Rossman and Smyth 1990; Mosenfelder et al. 2005). The only exception is coesite inclusions in diamond with a reported OH concentration of about 135 ppm H2O (Koch-Müller et al. 2003).

Feldspars and nepheline The feldspar group contains the widest range of hydrous species of any mineral group. This, in addition to the structural complexity of the feldspars, creates a diverse array of possible structural incorporation mechanisms in these minerals. Feldspars may contain structural OH, structural H2O molecules, ammonium ions (NH4+), and submicroscopic fluid inclusions (Hofmeister and Rossman 1985b; Solomon and Rossman 1988; Beran et al. 1992; Johnson and Rossman 2003; Johnson and Rossman 2004). Representative infrared spectra of these hydrous species in feldspars are plotted in Figure 2. Structural OH is characterized by broad (~600 cm−1) absorption bands, with the maximum intensity in the X optical direction (Johnson and Rossman 2003). The detailed shape and peak position of the OH bands is roughly a function of composition. Plagioclase feldspars have OH bands centered at 3200-3300 cm−1, whereas alkaline feldspars containing OH have bands at ~3060-3450 cm−1 (Johnson and Rossman 2004). Low albite, unlike the other plagioclase feldspars, contains fluid inclusions and very sharp bands assigned to OH, similar to the OH bands for quartz (Wilkins and Sabine 1973; Johnson and Rossman 2003). Beran (1987) proposed that OH substitutes for oxygen on the Ocm site in labradorite. The alkali feldspars microcline and anorthoclase also contain structural H2O molecules, and some anorthoclase and possibly sanidine may contain a combination of structural H2O and OH (Beran 1986; Xia et al. 2000; Johnson and Rossman 2004). There are two types of structural H2O: one with asymmetric and symmetric stretching frequencies at 3450 and 3630 cm−1

120

1.00

A.

Microcline GRR1281 Pala, CA

4.00

Absorbance per mm

Absorbance per mm

5.00

Johnson

H2O 3.00 2.00 1.00

X

Z

B.

Oligoclase GRR1280 Basalt, NV

0.80

OH 0.60 0.40 0.20

X Y Z

Y 0.00 4000 3800 3600 3400 3200 3000 2800 2600

0.00 4000 3800 3600 3400 3200 3000 2800 2600

-1

-1

Wavenumbers (cm )

Wavenumbers (cm ) 3.50

1.50

NH4

D.

C.

CIT19237 Microcline Pegmatite Cañon City, CO

GCS349 Fry Mts Pluton Granodiorite (Unpolarized)

3.00 Absorbance per mm

Absorbance per mm

2.00

+

E||X'

1.00 E||Y'

2.50

Sub-microscopic fluid inclusions

2.00 1.50

77 K 1.00

0.50

298 K 0.50

E||Z'

0.00

0.00 3900

3700

3500

3300

3100

2900

2700

-1

3900

3700

3500

3300

3100

2900

2700

-1

Wavenumbers (cm )

Wavenumbers (cm )

2.00

Absorbance

1.50

E.

Liquid water and ice (unpolarized)

Figure 2. Infrared spectra of hydrous species in feldspars. A. Structural H2O in microcline. B. OH in plagioclase. C. NH4+ in microcline. D. Sub-microscopic fluid inclusions at 298 K and 77 K. E. Infrared bands of water at 298 K and ice at 77 K. Replotted from Johnson and Rossman (2003, 2004).

1.00

0.50

77 K 298 K

0.00 3900

3700

3500

3300

3100

2900

2700

-1

Wavenumbers (cm )

with absorbance greatest in X, and a second type with stretching frequencies at 3285 and 3575 cm−1 with maximum absorbance in Z. It has been hypothesized that one of these two types of structural H2O substitutes into the large cation (K+) site in the structure via a charge-coupled substitution with Ca2+; the second type of H2O may be associated with the large cation site or with defects in the structure along exsolution lamellae boundaries (Hofmeister and Rossman 1985b; Kronenberg and Wolf 1990). Structurally bound ammonium ions have also been reported in microcline from pegmatites and hyalophane (Solomon and Rossman 1988; Beran et al. 1992). Finally, plutonic and metamorphic feldspars contain either fluid inclusions that are <1 µm in size, or water “clusters.” Unlike microscopic fluid inclusions, these inclusions do not freeze to ice at 77 K (Fig. 2) and may contain alteration products including epidote, clays, and sericite (Johnson and Rossman 2004).

Water in Nominally Anhydrous Crustal Minerals

121

Crystal structure does affect the possible speciation of hydrogen in feldspars (Nakano et al. 2001). Figure 3 shows the infrared spectra of cogenetic albite and microcline from a pegmatite vein. The K-rich lamellae of the microcline contain structural H2O, the Na-rich lamellae contain fluid inclusions, and the albite crystal contains structural OH and fluid inclusions. The abundances and species of hydrogen in feldspars are plotted in Figure 4 and listed in the Appendix. All of the hydrous species are present in pegmatite feldspars, but volcanic feldspars contain only structural OH (and melt inclusions). Absolute OH concentrations in volcanic feldspars are not dependent upon major-element composition (Johnson and Rossman 2004). Feldspars from intrusive bodies other than pegmatites contain only sub-micrometer

3.00

GRR2066 Pegmatite Myanmar

Microcline

Absorbance per mm

2.50

K-rich transparent areas: H2O (E||X')

2.00

1.50

Na-rich turbid areas: fluid inclusions (unpolarized)

Figure 3. Infrared spectra of hydrous species in coexisting perthitic microcline and albite in a pegmatite. Replotted from Johnson and Rossman (2004).

1.00

Albite 0.50

OH and fluid inclusions (unpolarized)

0.00 3900

3700

3500

3300

3100

2900

2700

-1

Wavenumbers (cm )

Pegmatitic

H2O NH4

Volcanic

+

Plutonic

OH Fluid Inclusions Alteration

OH Melt Inclusions

OH Fluid Inclusions Alteration

0

1000

2000

3000

4000 +

Hydrogen Concentration (ppm H2O/NH4 )

Figure 4. Concentrations of hydrous species in feldspars plotted by rock type. Data from Johnson and Rossman (2004).

122

Johnson

fluid inclusions or “water clusters” and secondary alteration minerals. The highest (likely) OH concentration recorded for feldspar is 915 ppm in a volcanic phenocryst (Xia et al. 2000). Structural H2O concentrations in feldspars range from 80-1350 ppm H2O (Table 1). Nepheline, (Na,K)AlSiO4, is a nominally anhydrous feldspathoid mineral that contains structural H2O molecules (Beran 1974). There are two types of structurally distinct H2O sites in nepheline, and a third type develops upon heating to 300 °C (Beran and Rossman 1989). All three types of water have their H-H vectors oriented perpendicular to the c-axis. Absolute water concentrations in nepheline range from 0.05 to 0.55 wt% H2O (Beran and Rossman 1989; Balassone and Beran 1995) (Table 1).

Pyroxenes Pyroxenes contain structural OH groups. An overview of structure of OH in pyroxenes and crystal chemical relationships is given in Skogby (2006) and Smyth (2006). Polarized infrared spectra of OH in diopside and enstatite are shown in Figure 5. Diopside and enstatite typically have four bands in the 3350-3645 cm−1 region, whereas omphacite and augite have only one to two distinct OH bands. Amphibole lamellae and disordered pyribole and jimthompsonite

1.50

A.

Diopside 95AK8f Adirondacks, NY

Absorbance per mm

1.25 1.00 0.75

X

Y

0.50 0.25

Z

0.00 4000

3800

3600

3400

3200

3000

-1

Wavenumbers (cm ) 1.50

B.

Enstatite HS-37 Bonin Islands, Japan

Absorbance per mm

1.25 1.00 0.75 0.50

X Y

0.25 Z 0.00 4000

3800

3600

3400

3200

3000

-1

Wavenumbers (cm )

Figure 5. Polarized single-crystal infrared spectra of A) diopside and B) enstatite. Data from Skogby et al. (1990); Johnson et al. (2002).

Water in Nominally Anhydrous Crustal Minerals

123

produce very narrow bands (~10 cm−1) at energies of 3650 cm−1 and higher (Skogby et al. 1990). Spodumene spectra have three bands at 3395, 3410, and 3434 cm−1, and some samples have a second set of bands in the 3474-3490 cm−1 region (Filip et al. 2006). Concentrations of OH in natural pyroxenes are plotted in Figure 6 according to general rock type or geologic origin (Skogby et al. 1990; Bell et al. 1995; Johnson et al. 2002; Peslier et al. 2002). In general, clinopyroxenes have higher OH concentrations than orthopyroxenes that formed under similar conditions. Mantle pyroxenes and those from basaltic xenoliths have higher OH concentrations than volcanic phenocrysts and pyroxenes from high-grade metamorphic regimes. Crustal pyroxenes with the highest OH concentrations (up to 466 ppm H2O) are found in pegmatites and authigenic environments. Table 1 lists the range of OH concentrations for crustal and select mantle samples.

Garnets Obtaining quantitative analyses of OH in garnets is more straightforward than for many other minerals due to their cubic symmetry. However, the OH band patterns for garnet are some of the most complicated and difficult to interpret. The garnet group includes extensive solid solution between at least seven end-member compositions, including the hydrogrossular (Ca3Al2(OH)12) end-member and the intermediate hydrated garnets hibschite and katoite. Here I concentrate on the trace to minor OH concentrations in the common crustal garnets with compositions in the grossular-andradite and spessartine-almandine fields. Pyrope is principally a mantle mineral, and is reviewed in Beran and Libowitzky (2006). The hydrogarnet substitution ((OH)4 ↔ SiO4) is an important mechanism for incorporation of water (hydroxyl) into the garnet structure, and the structure of the (OH)4 clusters has been evaluated with neutron diffraction (Bartl 1967; Lager et al. 1987a; Lager et al. 1989). Although an attempt to locate the hydrogen in a low water content garnet using neutron diffraction was not successful (Lager et al. 1987b), a proton nuclear magnetic resonance study of grossular garnets found evidence for clusters of two OH groups as well as the hydrogrossular

Pyroxenes

Authigenic

Opx = Pegmatites/Igneous Intrusions

Cpx =

Metamorphic High-grade metamorphic Volcanic Basaltic xenolith /xenocryst Mantle wedge xenolith Eclogite or mantle xenocryst/xenolith

0

200

400

600

800

[OH] ppm H2O wt.

Figure 6. OH concentrations of crustal and mantle pyroxenes plotted according to rock type. Data from Skogby et al. (1990); Bell et al. (1995); Johnson et al. (2002); Peslier et al. (2002).

124

Johnson

substitution (Cho and Rossman 1993). The major element composition of garnets has an effect on average OH band position, width, and complexity (Fig. 7), as well as the infrared absorption coefficient used for quantitative analysis (Rossman et al. 1988; Rossman 2006). The origin of the fine structure of the OH bands is not well understood, but may be due to crystal chemical substitutions of cations in sites adjacent to the OH in the structure (Aines and Rossman 1984b; Andrut and Wildner 2001). Some optically birefringent garnets, especially Ca-rich garnets from hydrothermal or metasomatic deposits, have anisotropic OH bands in the infrared (Rossman and Aines 1986). The origin of this pleochroism is unknown, but may be associated with defects along twinned growth sectors (Allen and Buseck 1988; Hofmeister et al. 1998). A study of birefringent natural uvarovite garnets concluded that SiO3(OH) tetrahedral groups are an important mechanism of OH defects in garnets with low water contents (Andrut et al 2002). Figure 8 shows garnet OH concentrations according to geologic provenance (Aines and Rossman 1984b; Rossman and Aines 1991; Locock et al. 1995; Amthauer and Rossman 1998; Arredondo et al. 2001; Johnson 2003). The typical OH concentration in mantle pyrope is much lower than concentrations reported in Aines and Rossman (1984b) (Figure 8; Beran and Libowitzky 2006). The maximum reported OH concentration for natural microscopic hydrogrossular garnets is 20 wt% H2O (O’Neill et al. 1993), but in most macroscopic garnets concentrations range from below detection limits up to 0.4 wt% (4000 ppm H2O) (Table 1).

Al2SiO5 polymorphs The Al2SiO5 polymorphs (kyanite, sillimanite, and andalusite) are found in a variety of metamorphic crustal and mantle rocks including amphibolites, granulites, and eclogites. Representative polarized infrared spectra of these minerals are plotted in Figures 9, 10, and 11 (Beran et al. 1989; Bell et al. 2004a; Burt et al. 2006). The absorbance bands in these spectra are assigned to structural hydroxyl groups. It can be seen that even though these

Pyrope

Garnets -196°C

5.0

Absorbance

4.0

3.0

Almandine

Figure 7. Typical single-crystal infrared spectra of pyrope, almandine, spessartine, and grossular obtained at −196 °C. Data from Aines and Rossman (1984b).

2.0 Spessartine

1.0 Grossular 0.0 3800

3700

3600

3500 -1

Wavenumbers (cm )

3400

Water in Nominally Anhydrous Crustal Minerals

125

Pegmatite

Serpentinite or altered ultramafic body Skarn

Magmatic and igneous intrusives

Garnet group

Gneiss

0

0.5

1

1.5

2

2.5

[OH] wt% H2O

Garnet group

Mantle xenoliths/xenocrysts

0

0.1

0.2

0.3

0.4

0.5

[OH] wt% H2O

Figure 8. The OH concentration in crustal garnets plotted according to rock type. Representative mantle garnets are also plotted. Data are from Aines and Rossman (1984b); Locock et al. (1995); Amthauer and Rossman (1998); Arredondo et al. (2001); Johnson (2003). 0.50

Absorbance per mm

Sillimanite GRR273

0.40 0.30 0.20

X

Figure 9. Polarized single-crystal infrared spectra of structural OH bands obtained on sillimanite from Reinbolt Hills, Antarctica (GRR 273). Data replotted from Beran et al. (1989).

Y

0.10 Z 0.00 4000

3500

3000

2500

-1

Wavenumbers (cm )

2.50

Figure 10. Polarized single-crystal infrared spectra of structural OH bands obtained on kyanite from a kyanite- and corundum-bearing eclogite xenolith from the Frank Smith kimberlite, South Africa. Data replotted from Bell et al. (2004a).

Absorbance per mm

Kyanite FSM-15

2.00 1.50 1.00 0.50 0.00

Y X Z 3600

3400

3200 -1

Wavenumbers (cm )

3000

126

Johnson 2.00 Andalusite GRR 278 Brazil

1.80

Figure 11. Polarized single-crystal infrared spectra of structural OH bands of andalusite from Minas Gerais, Brazil. Data from Burt et al. (2006).

Absorbance

1.60 1.40 1.20 1.00

Z, E||a

0.80 0.60 0.40

Y, E||b

0.20

X, E||c 0.00 3800

3600

3400

3200

-1

Wavenumbers (cm )

minerals have identical compositions, their OH band patterns and therefore the local hydrogen bonding environments are different (Libowitzky 1999). The structural OH does not seem to be associated with trace constituents such as Fe or B in sillimanite (Beran et al. 1989). This, together with the limited crystal chemistry of the Al2SiO5 polymorphs, suggests that hydrogen is incorporated into these minerals via the exchange mechanism: 3(OH)− + vacancy ↔ Al3+ + 3O2−

(1)

The OH concentrations in these minerals should therefore reflect the water activity during peak metamorphism (Beran et al. 1989). The absolute concentrations of OH in the Al2SiO5 polymorphs are low compared to many other nominally anhydrous minerals (Table 1; Appendix). The total range of OH concentrations observed in kyanite (Fig. 12) is 3-230 ppm H2O, with the highest concentrations in kyanite from eclogite xenoliths (Beran and Götzinger 1987; Beran et al. 1993; Bell et al. 2004a). The maximum OH concentration in kyanite calculated from the calibration of Bell et al. (2004a) is much smaller than previously estimated using conductometry (Beran and Götzinger 1987). The reported range of OH in sillimanite is 0-200 ppm H2O. Figure 13 shows OH concentration in sillimanite plotted by metamorphic facies, in order of inferred water activity and temperature range. The maximum OH concentration increases with increased expected water activity in the system. Only one OH concentration has been reported for andalusite: a sample from Brazil contains 270 ppm H2O (Rossman 1996). One reason there are not more analyses of aluminosilicates may be the difficulty in avoiding abundant fluid inclusions and alteration minerals trapped within cleavage planes. Beran et al. (1989) found that weight loss (assumed to be water) from sillimanite measured over the range 23 °C to 500 °C did not affect the intensities of the structural OH modes in the sample.

Rutile and cassiterite Polarized infrared spectroscopic studies have determined that hydrogen is incorporated into the rutile (TiO2) structure as hydroxyl groups and that the OH vector orientation is perpendicular to the c crystallographic axis (Rossman and Smyth 1990; Swope et al. 1995). This strong directional anisotropy of OH in the rutile structure results in pronounced absorbance features in spectra obtained with the electric vector (E) of incident infrared light perpendicular to c, and small or undetectable absorption in the spectrum obtained when E is parallel to the c crystallographic axis (Fig. 14). There is one prominent OH band present at 3280 cm−1 in spectra of both natural and synthetic rutile (Rossman and Smyth 1990; Vlassopoulos et al. 1993;

[OH] in kyanite (ppm H2O wt.)

Water in Nominally Anhydrous Crustal Minerals

127

120 Kyanite 100

Granulites, gneiss, greenstone belt

80

Eclogites

Figure 12. The OH concentration in kyanite as a function of peak metamorphic temperature for crustal rocks and mantle xenoliths. Data are from Beran and Götzinger (1987), and have been recalculated according to the most recent infrared calibration of kyanite (Bell et al. 2004a).

60 40 20 0 0

200

400

600

800

1000

Xenolith

Sillimanite Increasing Temperature

Figure 13. The approximate OH concentration in sillimanite plotted according to rock type and in order of inferred peak metamorphic temperature and water activity. Data are from Beran et al. (1989), with approximate OH concentrations estimated with the calibration given in that study.

Increasing Water Activity

Temperature (ºC)

Pyroxene granulite facies Hornblende granulite facies Upper amphibolite facies Unknown or alluvial granulite

0

50

100

150

200

250

Approximate [OH] (ppm H2O wt.) 20

Absorbance per mm

Rutile 15

10

R-6 E||a

Magnetite-apatitechlorite schist

R-1 E||a

Kyanite quartzite

Figure 14. Polarized singlecrystal infrared spectra (with the E-vector oriented parallel to the a- and c-crystallographic axes) of structural OH bands of rutile from a magnetite-apatitechlorite schist (R-6) and a kyanite quartzite (R-1). Data replotted from Vlassopoulos et al. (1993).

5

0 4000

E||c 3800

3600

3400

3200 -1

Wavenumbers (cm )

3000

128

Johnson

Bromiley et al. 2004). Some infrared spectra of natural samples also have one additional, minor band present at 3320 cm−1 or 3360 cm−1 (Vlassopoulos et al. 1993). The limited number of OH bands and therefore distinctly different OH sites is presumably related to the fairly simple, high-symmetry structure and limited solid solution chemistry for rutile. Several previous studies (Rossman and Smyth 1990; Hammer and Beran 1991; Vlassopoulos et al. 1993; Maldener et al. 2001) reported infrared band areas or OH concentrations of natural rutile crystals. The OH concentration data for natural rutile have not been compiled previously in the literature, at least in part because the reported OH concentrations were determined using different values for the integrated absorption coefficient (Johnson et al. 1973; Hammer 1988; Maldener et al. 2001). The OH concentrations from the four studies were recalculated (or calculated) if necessary, using the absorption coefficient from Maldener et al. (2001). These studies report only a general rock type of origin for many of the rutile crystals, allowing the rutile OH concentrations to be categorized very roughly in terms of increasing temperature and pressure (Fig. 15). The maximum OH concentration generally increases with increasing inferred temperature and pressure. Minimum reported OH concentrations are 70-100 ppm H2O by weight. The maximum OH concentrations are found in mantle samples, up to a maximum of about 2770 ppm (Appendix). Cassiterite, SnO2, has a structure identical to that of rutile. The infrared spectra and polarization behavior of OH in cassiterite are therefore not surprisingly very similar to those of rutile; peaks occur at ~3350 cm−1 and ~3250 cm−1, with the latter sometimes split into two bands centered on that energy (Losos and Beran 2004). The range of OH concentrations found in natural cassiterite (Appendix) is 32-170 ppm H2O (Maldener et al. 2001; Losos and Beran 2004).

Zircon and titanite Zircon, ZrSiO4, is a common accessory mineral in igneous and metamorphic rocks, and is especially useful to geochronologists because trace U and Th incorporated into the structure (and their daughter products) can be used for radiometric dating. Although fluid inclusions and alteration minerals such as clays may contribute bands to the mid-infrared spectra of turbid zircons, most zircons contain only structural OH (Woodhead et al. 1991a). The nature of the structural OH and resulting band pattern in the infrared spectra (Fig. 16) depend upon the degree of metamictization of the crystal (Woodhead et al. 1991a,b). Crystalline zircon contains well-ordered OH that is anisotropic and is manifested by sharp bands in the infrared spectra. As the degree of radiation damage increases, broad, isotropic bands increasingly dominate the spectra. These bands are also assigned to structural OH, because of a lack of absorbance in the water combination mode region (5250 cm−1) (Woodhead et al. 1991a,b). Broader OH bands are indicative of a wider range of OH hydrogen bonding distances (Libowitzky 1999) brought on by structural damage. The overall concentrations of OH in zircon (Table 1) are not well established, because of a lack of published extinction coefficients for zircon (Bell et al. 2004b). Estimated OH concentrations in zircon are given based on the assumption that the absorption coefficient for OH in zircon is similar to that of grossular garnet (Rossman and Aines 1991). There is limited data available for both crustal and mantle zircons, but the maximum estimated “crystalline” OH concentration, ~0.01 wt% H2O (100 ppm), was found in a zircon megacryst from a kimberlite (Woodhead et al. 1991a). Maximum “primary” OH concentration is estimated at 0.05 wt% H2O for a partially metamict zircon from a hydrothermal vug. The maximum metamict OH concentration (primary or secondary hydration) in zircon is about 0.1 wt% H2O (Woodhead et al. 1991a). A study of Sri Lankan zircons from alluvial deposits that have undergone various degrees of radiation damage, from very little to a large degree of metamictization, found that the crystalline zircons from this area contain very little to no OH, but the metamict ones have a great deal of the broad-band, isotropic OH (Woodhead

Water in Nominally Anhydrous Crustal Minerals

129

Increasing P and T

Rutile Figure 15. The OH concentration in rutile plotted according to rock type and in order of approximate increasing pressure and temperature. Data are from Rossman and Smyth (1990); Hammer and Beran (1991); Vlassopoulos et al. (1993); Maldener et al. (2001) and are recalculated using the calibration of Maldener et al. (2001).

Eclogites and ultramafic xenoliths Blueschist and kyanite + rutile Amphibolites Mica schists Pegmatites

0

1000 2000 [OH] (ppm H2O wt.)

3000

1.6

Figure 16. Representative polarized infrared spectra of OH bands obtained on partially metamict (Gneiss, Madagascar) and crystalline (Kaalvallei, South Africa) zircons. Data are from Woodhead et al. (1991a); Bell and Rossman (1992).

Absorbance per mm

1.4 1.2

Zircon

1.0

Gneiss, Ampanobe Madagascar Partially metamict

0.8

GRR1225b E||c

0.6

E||a

Kaalvallei Kimberlite, South Africa Crystalline

0.4 0.2

KLV-22 E||c E|c

0.0 4000

3800

3600

3400

3200

3000

-1

Wavenumbers (cm )

et al. 1991b). This shows that this type of broad-band OH is taken up by the mineral after structural damage occurs, and is not an original part of the mineral (Nasdala et al. 2001). Titanite, CaTiSiO5, also takes up trace U and Th into its structure (Hawthorne et al. 1991), and therefore also suffers radiation damage and metamictization. There is a single crystalline OH band at 3486 cm−1, with maximum absorbance in the X direction (Beran 1970). This band broadens, especially on the low energy side <3486 cm−1, with increasing metamictization (Zhang et al. 2001).

Cordierite and beryl Cordierite, (Mg,Fe)2Al4Si5O18 and beryl, Be3Al2Si6O18, can incorporate cations and small molecules within the channels running parallel to the c-axis of these minerals. A summary of the structure and types of channel water in cordierite and beryl can be found in Rossman (1988). A brief description of the structure of H2O and CO2 in cordierite, and a summary of studies that have used volatiles in cordierite to attempt to evaluate metamorphic fluid compositions are provided here. There are two types of structurally distinct H2O molecules in cordierite (Fig. 17). Type I H2O has its H-H vector oriented parallel to the c-axis (and channels), while type II H2O is oriented with its H-H vector perpendicular to the c-axis (Farrell and Newnham 1967; Aines

130

Johnson 0.20 H2O region

Absorbance

0.15

Cordierite Powder in KBr pellet CO2

type I

Figure 17. Unpolarized infrared spectrum of a KBr pellet of powdered cordierite. Bands are assigned to type I and II H2O and CO2 in the channels within the cordierite structure. Data from Vry et al. (1990).

0.10 type II 0.05

0.00 3800

3400

3000

2600

2200

-1

Wavenumbers (cm )

and Rossman 1984c; Rossman 1988; Kolesov and Geiger 2000). Type II H2O is associated with large cations such as Na, K, and Ca that are also incorporated into the channels. Many natural cordierites also contain CO2 in their channels, with the molecular axis oriented parallel to the a-axis (Aines and Rossman 1984c; Kolesov and Geiger 2000). At high temperatures, both type I and II water are dynamically disordered, and dehydration of water occurs by 800 °C; loss of CO2 is complete by 900 °C (Aines and Rossman 1984c). Two studies have reported hydrocarbons in cordierite (Mottana et al. 1983; Khomenko and Langer 1999). End-member Mg cordierite can theoretically hold up to 2.99 wt% H2O or 6.99 wt% CO2 (Vry et al. 1990). Typical concentrations of H2O in cordierite range from 0.3 to 2.04 wt% and CO2 concentrations range from 0.1 to 2.2 wt% (Armbruster et al. 1982; Vry et al. 1990; Swamy et al. 1992; Visser et al. 1994; Kalt 2000). Although it may seem to be the perfect monitor of fluid composition, crystal chemistry (particularly cations in the channels) and grain size can bias uptake of water over CO2 during metamorphism (Vry et al. 1990). Post-metamorphic re-equilibration may also affect the resulting estimated water activity of the system (Visser et al. 1994). A complete list of studies involving cordierite as an indicator of metamorphic fluid history is given in the Appendix.

UNDERSTANDING GEOLOGIC SYSTEMS As shown above, structural hydrous species are incorporated into essentially all major nominally anhydrous crustal minerals. These studies have established the range of hydrous species concentrations in various crustal minerals as well as the crystallographic orientations of these species. The plots in the previous section of water concentration as a function of general geologic provenance or rock type hint at the type of information that could be derived from these measurements. This section provides a summary of geologic issues that could be informed by such hydrogen concentration measurements. These issues can be divided into two categories: evaluation of thermodynamic properties of a system, and evaluation of the effect of water on physical properties of common crustal minerals.

Thermodynamic properties Water concentrations in nominally anhydrous minerals can be used as indicators of water activity or the oxygen fugacity of the geologic environment. This is analogous to using

Water in Nominally Anhydrous Crustal Minerals

131

water contents and Fe3+/Fe2+ ratios in hydrous minerals to evaluate water activity and oxygen fugacity (e.g., Lamb and Valley 1988; King et al. 2000). With the exception of cordierite, most minerals hydrous or nominally anhydrous do not incorporate measurable CO2 into their structures. This means that we can derive the water activity of the system, but not the complete fluid composition or total fluid pressure, without knowing the pressure and temperature of the system. This method of evaluating fluids in the geologic environment is therefore best used where the fluid is predominantly H2O, or in environments where water is the particular fluid phase of interest. Details of the solubility of water in nominally anhydrous minerals are given in Keppler and Bolfan-Casanova (2006). The studies described below qualitatively and quantitatively investigate water activity and oxygen fugacity in natural systems using nominally anhydrous minerals. Garnets as indicators of fluid evolution. Conductive heat loss models indicate that pegmatite veins should crystallize from the outer walls of the vein (wall zone) inwards to the center (core). This progressively concentrates incompatible elements and water in the remaining melt, so that the largest and most beautiful minerals crystallize in the core area of the vein (Lumpkin 1998). Arredondo et al. (2001) measured the OH concentrations of spessartine in transects across three pegmatite bodies (the Rutherford No.2 pegmatite, VA, and the Himalaya and George Ashley Block pegmatites in CA) to establish if pegmatite garnets record the evolution of fluids in these systems. The OH concentrations and the major (Mn) and minor (Fe, Ca) element compositions of the spessartine record a general increase in water content of the melt from wall to core zone for the Rutherford No.2 (Fig. 18) and Himalaya pegmatites, and record a magma reinjection event in the George Ashley Block pegmatite (Arredondo et al. 2001). Johnson (2003) investigated the zonation of OH and minor elements in cm-sized grossularandradite garnets from an irregular epidote-grossular-calcite skarn. This skarn is less than 10 meters wide, and formed in a contact metamorphic zone between lower Cambrian sedimentary rocks and the granitic intrusives of the Inyo batholith, located at the foot of Birch Creek on the northwest side of Deep Springs Valley, Inyo County, California (Nash 1962; Shieh and Taylor 1969). In all three of the large, optically zoned garnets investigated, the OH concentration monotonically decreases from core (0.43-0.47 wt% H2O) to rim (0.14-0.20 wt% H2O). Although Fe exhibits oscillatory zonation, the Ti concentration in the zoned garnets is highly correlated with the OH concentration, and H and Ti may participate in charge-coupled substitution into the garnet crystal structure. This association of H and Ti in grossular from Birch Creek is not

0.2 Spessartine garnets in pegmatite Rutherford No.2 Mine

Wt% H2O

0.15

Figure 18. The OH concentration in spessartine garnets from the Rutherford No.2 Pegmatite plotted from the outer wall zone to the center (core) of the pegmatite. Figure modified from Arredondo et al. (2001).

0.1

0.05

0 0 1 2 Wall zone

3

4

5

6

7

Sample position

8

9

10 11 Core

132

Johnson

observed in other grossular samples (Rossman and Aines 1991). The small skarn at Birch Creek likely formed during infiltration of magmatic fluid from the late-stage intrusive dike into the surrounding country rock. The solubility of Ti in the fluid phase is expected to decrease sharply with decreasing temperature (Van Baalen 1993). After initial infiltration of magmatic fluid into the carbonate sediments, the system cooled and Ti solubility in the fluid quickly dropped, resulting in the observed OH and Ti zonation in the Birch Creek garnets. These two studies show that OH concentration measurements can be used to qualitatively investigate the fluid history of igneous and metamorphic rocks. Volatile composition of melts. The OH concentrations in volcanic phenocrysts could be used to quantitatively evaluate magmatic water content in volcanic systems if the partitioning behavior of hydrogen between nominally anhydrous minerals and melts is known. A few studies have experimentally (Aubaud et al. 2005) and empirically (Dobson et al. 1995; Johnson 2005; Seaman et al. 2006) determined partition coefficients for hydrogen in such systems. The experimental study found the following partition coefficients (D) between coexisting peridotite minerals and a basaltic melt: D(ol/melt) = 0.0017, D(opx/melt) = 0.019, and D(cpx/melt) = 0.023. These D values are not affected by changes in pressure or total water content over the experimental range (1-3 GPa; melt water contents of 3.1-8.8 wt% H2O). An empirical study (Dobson et al. 1995) determined the partition coefficient of water between orthopyroxene phenocrysts and coexisting boninite glass from the Bonin Islands, Japan, to be 0.003-0.004. Johnson (2005) used hydroxyl concentrations of feldspar phenocrysts and water concentrations of their melt inclusions from the 1980-1981 series of eruptions of Mount St. Helens to come up with D of 0.004. On the other hand, Seaman et al. (2006) recently estimated a partition coefficient for water of 0.1 between anorthoclase crystals and melt inclusions from Mount Erebus, Antarctica. Further work should be done to determine partition coefficient values so that OH measurements may be used to study volatile histories in volcanoes. Diffusion of hydrous species in minerals plays a potentially important role in preservation of OH abundances in phenocrysts during and immediately after an eruption, especially those that undergo oxidation-dehydration reactions. For example, Fe-bearing minerals such as amphiboles, micas, pyroxenes, and olivine are known to lose structural hydrogen when Fe is oxidized (e.g. Addison et al. 1962; Vedder and Wilkins 1969; Skogby and Rossman 1989): Fe2+ + OH−(structure) = Fe3+ + ½H2 (g) + O2−(structure)

(2)

Diffusion of hydrogen in minerals is generally several orders of magnitude faster than diffusion of larger cations and oxygen, and is especially rapid when controlled by redox reactions such as Equation (2) (Ingrin and Blanchard 2006). In general, the time elapsed during eruption of a single block of pumice or lava flow is short compared to the time scale of significant diffusive loss of hydrogen from a phenocryst (Johnson 2005), but magma and coexisting phenocrysts may be transported and stored within the Earth before eruption. If transport and storage takes place over time scales longer than hours to weeks, the OH concentration in the nominally anhydrous phenocryst has time to re-equilibrate with the surrounding volatile conditions. It is therefore likely that OH in phenocrysts record volatile contents from the last storage event. Hydrothermal alteration. The sub-micrometer fluid inclusion concentrations in a suite of feldspars from the epizonal Fry Mountains Pluton, CA, were determined in Johnson and Rossman (2004). These samples represent varying degrees of hydrothermal exchange between meteoric water and the granite during cooling (Solomon and Taylor 1991). The concentration of sub-micrometer fluid inclusions in these feldspars generally increases with decreasing δ18O of the whole rock, a measure of the degree of hydrothermal exchange (Fig. 19). The sub-micrometer fluid inclusions appear to form during the process of oxygen isotope exchange between the feldspar and the meteoric fluids (O’Neil and Taylor 1967; Yund and Anderson 1978; David and Walker 1990; David et al. 1995), and their concentrations could be used as a proxy for the amount of water that has interacted with a given volume of rock in a hydrothermal system.

Fluid inclusion concentration (ppm H 2O)

Water in Nominally Anhydrous Crustal Minerals

133

4500 Fry Mountains Pluton San Bernardino Co., CA

4000 3500 3000

Figure 19. Sub-micrometer fluid inclusion concentration versus whole rock δ18O for feldspars from the Fry Mountain Pluton, CA. Data from Johnson and Rossman (2004).

2500 2000 1500 1000

Increasing degree of hydrothermal exchange with meteoric water

500 0 0

5

10

18

G O whole rock (‰)

Water activity and oxidation state in the lower crust. The fluid conditions in granulitefacies calc-silicate rocks from the Adirondacks Mountains, NY, were investigated by Johnson et al. (2002). The concentration of structural OH in diopside was determined for four different localities in the region for which water fugacities (~0.15-0.8 kbar) had been previously estimated using oxygen isotope systematics (Edwards and Valley 1998). The total diopside OH content ranges from 55 to 138 ppm H2O by weight, and the OH concentration in diopside increases monotonically with increasing water fugacity (Fig. 20) and linearly with the OH concentrations assigned to bands that increase in intensity during hydrothermal exchange experiments (Skogby and Rossman 1989). There is no significant variation in OH content within a single diopside grain or among diopside grains from the same hand sample. Charge-coupled substitution with M3+ and Ti4+ in the crystal structure may have allowed retention of OH in the diopside structure during and after peak metamorphism (~750 ºC, 7-8 kbar). The Cascade Slide diopside has an Fe3+/Fe2+ ratio of 0.98, compared to 0-0.05 for the other samples, implying that some loss of hydrogen through oxidation of Fe (via Eqn. 2) was possible in this sample. This study shows that OH concentrations in nominally anhydrous minerals containing Fe can be indicators of

160 140

OH (ppm H 2O)

120 100

Figure 20. Total OH concentration of diopside (diamonds) and “hydrothermal” OH (circles) plotted versus water fugacity, as predicted from oxygen isotope measurements (Edwards and Valley 1998). Figure modified from Johnson et al. (2002).

80 60 40 Total OH content

20 0 0.00

3645 and 3450 cm

0.20

0.40

0.60

0.80

-1

1.00

ƒH2O (kbar) 18

18

18

18

[' O(m eas) - ' O(dry)] / [' O(w et) - ' O(dry)]

1.20

134

Johnson

water fugacity in the crust for well-characterized rocks, although the redox chemistry of Fe in the crystal structure may interfere with this signal. Oxidation state of the mantle wedge. Although the Peslier et al. (2002) study of pyroxenes from mantle wedge xenoliths does not deal directly with crustal rocks, it is informative to crustal studies in terms of the relationship it finds between partial melting, oxidation state, and OH concentration in Fe-bearing nominally anhydrous minerals. This study investigated the OH concentrations of orthopyroxene and clinopyroxene from spinel peridotite xenoliths from Mexico and the northwest United States. The OH concentrations of the pyroxenes are correlated with whole-rock Al2O3 content, implying that mantle geochemical signatures are preserved in these samples. The ratio of OH concentrations in coexisting clinopyroxene and orthopyroxene is consistently ~2 to 1, and absolute OH concentrations range from 140-528 ppm H2O in clinopyroxene and 39-265 ppm H2O in orthopyroxene. The OH concentrations in the pyroxenes are negatively correlated with calculated oxidation state of the peridotite, expressed as deviations above and below the fayalitemagnetite-quartz (FMQ) oxidation buffer (Fig. 21). This decrease in OH concentration with increased oxidation is thought to be caused by metasomatism of the sub-arc mantle wedge peridotites by oxidized fluids or melts derived from the subducting slab (Peslier et al. 2002). Increased oxidation state drives the redox reaction (2) and dehydration of the pyroxenes. This process of oxidation by melt or fluid infiltration could also affect the retention of OH in Fe-bearing crustal minerals. It is also interesting to note that the OH concentration in Fe- or Ti-bearing minerals may also be used as an indicator of oxygen fugacity, if the geologic system is well characterized (Johnson et al. 2005).

Physical properties Water affects the physical properties of minerals and rocks, including deformation rates and shear strength, phase transitions, stabilization of radiation damage, and diffusion rates of major elements such as Al, Si, and O which can in turn affect exsolution processes and geochemical signatures. Hydrolytic weakening and deformation rates. An overview of the influence of water on rock deformation is given in Kohlstedt (2006). Many experimental studies have examined the influence water has on lowering the temperature of the brittle-ductile transition of quartz (Griggs 1967; Tullis and Yund 1980; Kekulawala et al. 1981; Kronenberg et al. 1986; Ord and Hobbs 1986; Rovetta et al. 1986; Cordier et al. 1988; Gerretsen et al. 1989; Kronenberg and Wolf 1990; Kronenberg 1994; Post and Tullis 1998). The structural OH in natural quartz has a negligible effect on deformation behavior; instead, it is the fluid inclusions that play a major role in structural weakening (Kekulawala et al. 1981; Kronenberg et al. 1986; Gerretsen et al. 1989; Post and Tullis 1998). Several studies have investigated fluid inclusions in natural quartz samples to better understand the role of fluid inclusions in deformation and hydrothermal alteration of quartzbearing rocks (Kronenberg and Wolf 1990; Berzina et al. 1991; Nakano et al. 2001; Ito and Nakashima 2002; Niimi 2002). A study of the sheared quartz near the Moine thrust zone in Scotland (Kronenberg and Wolf 1990) reported a general increase in fluid inclusion content of quartz from 4200 ppm H2O at 110 m from the contact to 7400 ppm H2O within several centimeters of the contact, although there were large variations in water content among grains from a single hand sample. Another study (Nakashima et al. 1995) of a mylonite zone adjacent to the Median Tectonic Line (MTL) in Japan found that quartz fluid inclusion concentrations increased from 300 ppm H2O at 800 m from the MTL to 2500 ppm H2O 100 m away from the MTL (Fig. 22). This study also found a decrease in the fluid inclusion concentration of metacherts with increasing metamorphic grade (Nakashima et al. 1995).

Water in Nominally Anhydrous Crustal Minerals

135

600

[OH] ppm H2O wt.

500 400

Figure 21. OH concentrations of clinopyroxene from mantle wedge peridotite xenoliths plotted as a function of calculated oxygen fugacity. Data replotted from Peslier et al. (2002).

300 200 100

Clinopyroxene mantle wedge peridotites

0 -2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

'FMQ

Fluid inclusions in quartz from deformed granites in the Kamimura-Yanazawa area near the Median Tectonic Line (MTL), Japan

3000

Kamimura

H2O in qtz (ppm)

2500

Yanazawa

2000 1500 1000 500 0 0

200

400

600

800

1000

1200

1400

Distance from MTL (m)

Figure 22. Fluid inclusion H2O concentrations of quartz from deformed granitic rocks versus distance from the Median Tectonic Line (MTL) mylonite zone. Data replotted from Nakashima et al. (1995).

Phase transformations and radiation damage. There is evidence that hydrous species may promote or stabilize physical transformations in the crystal structure of a nominally anhydrous mineral. For example, coesite in ultrahigh-pressure rocks is preferentially preserved in areas with little or late, low-temperature infiltration of retrograde fluids (Mosenfelder et al. 2005). Experiments have determined that the transformation rate of coesite to quartz is more than an order of magnitude higher for coesite containing structural OH than for “dry” coesite (Lathe et al. 2005). As discussed previously, “anisotropic” OH in partially or completely metamict zircon and titanite is associated with radiation damage (Hawthorne et al. 1991; Woodhead et al. 1991a,b). This secondary OH likely stabilizes the metamict state (Hofmeister and Rossman 1985b; Aines and Rossman 1986). In some minerals including brown topaz, blue plagioclase, and amazonite, hydrous species help to charge-balance the structure and are key to stabilization

136

Johnson

of radiation-damage associated color centers (Aines and Rossman 1986; Hofmeister and Rossman 1986). On the other hand, water molecules inhibit coloration in amethyst and smoky sanidine (Hofmeister and Rossman 1985a; Aines and Rossman 1986). Diffusion rates of Al, Si, and O. The presence of water dramatically increases the Al-Si interdiffusion rate (Goldsmith 1986, 1987, 1988) as well as the oxygen diffusion rate in many silicate and oxide minerals (e.g., Yund 1983; Farver and Yund 1990; Farver and Yund 1991; Farver 1994). Faster oxygen diffusion rates lower the effective mineral closure temperature and affect the oxygen isotope systematics used to constrain the thermal histories of igneous and metamorphic rocks. The increased rate of Al and Si interdiffusion in feldspars under hydrous conditions may lead to larger exsolution lamellae and a coarser perthitic texture in alkali feldspars (Yund and Ackermand 1979). It has been suggested that the structural hydrous species such as OH play a role in the Al-Si interdiffusion process in feldspars (Goldsmith 1986; Goldsmith 1988). A comparison of perthitic feldspars from pegmatites showed that those with larger exsolution lamellae have higher hydrogen concentrations (as structural hydrous species and fluid inclusions) than those with smaller exsolution lamellae (Johnson and Rossman 2004).

The water budget of the Earth Structural hydrous species in nominally anhydrous crustal minerals do not significantly contribute to the overall water budget of the Earth, but sub-micrometer-sized fluid inclusions in feldspars and fluid inclusions in quartz are significant due to their high abundances and the ubiquity of feldspars and quartz in the continental crust. The total mass of water in feldspar inclusions in the upper continental crust is estimated to be about 1 × 1019 kg, assuming that the crust contains 30% feldspar and an average feldspar fluid inclusion concentration of 2000 ppm H2O (Johnson and Rossman 2004). If quartz (~15% of the upper crust) (McLennan and Taylor 1999) is included in the calculation, again assuming an average fluid inclusion content of 2000 ppm H2O, the total mass of water residing in nominally anhydrous minerals in the upper crust is ~2 × 1019 kg. This is roughly equivalent to the amount of water stored in the hydrous minerals in the upper crust (6 × 1019 kg), and a few percent of the ~4 × 1020 kg of water estimated to be stored in nominally anhydrous minerals within the upper mantle (Ingrin and Skogby 2000).

SUMMARY AND FUTURE POSSIBILITIES The hydrous species in quartz, feldspars, pyroxenes, garnets, kyanite, andalusite, sillimanite, rutile, cassiterite, zircon, titanite, and cordierite include hydroxyl (OH), structural water (H2O), the ammonium ion (NH4+), and sub-micrometer fluid inclusions. Cordierite also commonly includes CO2 in its channels. Plotting concentrations of hydrous species for a given mineral by rock type or metamorphic grade reveals trends broadly related to the expected water activities for each rock type. The concentrations of these species in each mineral can also be affected by crystal chemistry or structure. Most of the work involving nominally anhydrous crustal minerals to date has been exploratory. It consists primarily of detailed crystallographic studies of a few mineral samples and surveys of hydrous species and concentrations from exceptional mineral samples from various known and unknown geologic provenances. The handful of studies that have used nominally anhydrous minerals to investigate water activity, oxygen fugacity, and deformation behavior in specific locations or geologic provenances have had encouraging results. This chapter provides guidelines for future studies in this area of research.

Water in Nominally Anhydrous Crustal Minerals

137

ACKNOWLEDGMENTS The author would like to thank Anton Beran, Eugen Libowitzky, Henrik Skogby, and an anonymous reviewer for careful and constructive criticism that improved this chapter. Jason Burt, Satoru Nakashima, and Julie Vry graciously provided data for figures. I greatly appreciate Hans Keppler’s editorial guidance and efforts in organizing this project. Special thanks go to George Rossman, for providing access to old data files, digitizing software, and the Caltech spectroscopic facilities. This work was supported by NSF grant EAR-0409883.

REFERENCES Addison CC, Addison WE, Neal GH, Sharp JH (1962) Amphiboles: Part I. The oxidation of crocidolite. J Chem Soc 1962:1468-1471 Aines RD, Kirby SH, Rossman GR (1984) Hydrogen speciation in synthetic quartz. Phys Chem Minerals 11: 204-212 Aines RD, Rossman GR (1984a) Water in minerals? A peak in the infrared. J Geophys Res 89:4059-4071 Aines RD, Rossman GR (1984b) The hydrous component in garnets: pyralspites. Am Mineral 69:1116-1126 Aines RD, Rossman GR (1984c) The high temperature behavior of water and carbon dioxide in cordierite and beryl. Am Mineral 69:319-327 Aines RD, Rossman GR (1985) The high temperature behavior of trace hydrous components in silicate minerals. Am Mineral 70:1169-1179 Aines RD, Rossman GR (1986) Relationships between radiation damage and trace water in zircon, quartz, and topaz. Am Mineral 71:1186-1193 Allen FM, Buseck PR (1988) XRD, FTIR, and TEM studies of optically anisotropic grossular garnets. Am Mineral 73:568-584 Amthauer G, Rossman GR (1998) The hydrous component in andradite garnet. Am Mineral 83:835-840 Andrut M, Wildner M (2001) The crystal chemistry of birefringent natural uvarovites. Part I. Optical investigations and UV-VIS-IR absorption spectroscopy. Am Mineral 86:1219-1230 Andrut M, Wildner M, Beran A (2002) The crystal chemistry of birefringent natural uvarovites. Part IV. OH defect incorporation mechanisms in non-cubic garnets derived from polarized IR spectroscopy. Eur J Mineral 14:1019-1026 Armbruster T, Schreyer W, Hoefs J (1982) Very high CO2 cordierite from Norwegian Lapland: Mineralogy, petrology and carbon isotopes. Contrib Mineral Petrol 81:262-267 Arredondo EH, Rossman GR, Lumpkin GR (2001) Hydrogen in spessartine-almandine garnets as a tracer of granitic pegmatite evolution. Am Mineral 86:485-490 Aubaud C, Withers AC, Hirschmann MM, Guan Y, Leshin LA, Mackwell SJ, Bell DR (2005) A new calibration of H measurements by SIMS in glasses and nominally anhydrous minerals: application to experimental determinations of H partitioning. Eos Trans AGU 86(52), Fall Meet. Suppl. Abstract V41A-1419 Balassone G, Beran A (1995) Variable water content of nepheline from Somma-Vesuvio, Italy. Mineral Petrol 52:75-83 Bartl H (1967) Röntgen-Einkristalluntersuchungen an 3CaO·Al2O3 und an 12CaO·7Al2O3·H2O; neuer Vorschlag zur 12CaO·7Al2O3-Struktur. N Jb Mineral Mh 2-3:33-42 Bell DR, Ihinger PD, Rossman GR (1995) Quantitative analysis of trace OH in garnet and pyroxenes. Am Mineral 80:465-474 Bell DR, Rossman GR (1992) Water in Earth’s mantle; the role of nominally anhydrous minerals. Science 255: 1391-1397 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2004a) Hydroxide in kyanite: A quantitative determination of the absolute amount and calibration of the IR spectrum. Am Mineral 89:998-1003 Bell DR, Rossman GR, Moore RO (2004b) Abundance and partitioning of OH in a high-pressure magmatic system: Megacrysts from the Monastery Kimberlite, South Africa. J Petrol 45:1539-1564 Beran A (1970) Messung des Ultrarot-Pleochroismus von Mineralen. IX. Der Pleochroismus der OHStreckfrequenz in Titanit. Tscherm Mineral Petr Mitt 14:1-5 Beran A (1974) UR-spektroskopischer Nachweis von H2O in Nephelin. Tschermaks Mineralogische und Petrographische Mitteilungen 21:299-304 Beran A (1986) A model of water allocation in alkali feldspar, derived from infrared-spectroscopic investigations. Phys Chem Minerals 13:306-310 Beran A (1987) OH− groups in nominally anhydrous framework structures: an infrared spectroscopic investigation of danburite and labradorite. Phys Chem Minerals 14:441-445 Beran A, Armstrong JT, Rossman GR (1992) Infrared and electron microprobe analysis of ammonium ions in hyalophane feldspar. Eur J Mineral 4:847-850

138

Johnson

Beran A, Götzinger MA (1987) The quantitative IR spectroscopic determination of structural OH groups in kyanites. Mineral Petrol 36:41-49 Beran A, Langer K, Andrut M (1993) Single crystal infrared spectra in the range of OH fundamentals of paragenetic garnet, omphacite and kyanite in an eklogitic mantle xenolith. Mineral Petrol 48:257-268 Beran A, Libowitzky E (2006) Water in natural mantle minerals II: olivine, garnet and accessory minerals. Rev Mineral Geochem 62:169-191 Beran A, Rossman GR (1989) The water content of nepheline. Mineral Petrol 40:235-240 Beran A, Rossman GR, Grew ES (1989) The hydrous component of sillimanite. Am Mineral 74:812-817 Berzina AP, Sotnikov VI, Kovaleva LT, Gimon VO (1991) Water in quartz of Cu-Mo deposits. Soviet Geol Geophys 32:103-109 Bromiley G, Hilairet N, McCammon C (2004) Solubility of hydrogen and ferric iron in rutile and TiO2(II): implications for phase assemblages during ultrahigh-pressure metamorphism and for the stability of silica polymorphs in the lower mantle. Geophys Res Lett 31:doi:10.1029/2004GL019430 Burt JB, Ross NL, Gibbs GB, Rossman GR, Rosso KM (2006) Potential protonation sites in Al2SiO5 polymorphs based on polarized FTIR spectroscopy and properties of the electron density distribution. (in prep) Chakraborty D, Lehmann G (1976) Distribution of OH in synthetic and natural quartz crystals. J Solid State Chem 17:305-311 Cho H, Rossman GR (1993) Single-crystal NMR studies of low-concentration hydrous species in minerals: Grossular garnet. Am Mineral 78:1149-1164 Cordier P, Boulogne B, Doukhan JC (1988) Water precipitation and diffusion in wet quartz and wet berlinite AlPO4. Bull Minéral 111:113-137 Cordier P, Doukhan JC (1991) Water speciation in quartz: a near infrared study. Am Mineral 76:361-369 David F, Walker L (1990) Ion microprobe study of intragrain micropermeability in alkali feldspars. Contrib Mineral Petrol 106:124-128 David F, Walker L, Lee MR, Parsons I (1995) Micropores and micropermeable texture in alkali feldspars: geochemical and geophysical implications. Mineral Mag 59:505-534 Dobson PF, Skogby H, Rossman GR (1995) Water in boninite glass and coexisting orthopyroxene; concentration and partitioning. Contrib Mineral Petrol 118:414-419 Edwards KJ, Valley JW (1998) Oxygen isotope diffusion and zoning in diopside; the importance of water fugacity during cooling. Geochim Cosmochim Acta 62:2265-2277 Farrell EF, Newnham RE (1967) Electronic and vibrational absorption spectra in cordierite. Am Mineral 52: 380-388 Farver JR (1994) Oxygen self-diffusion in calcite: Dependence on temperature and water fugacity. Earth Planet Sci Lett 121:575-587 Farver JR, Yund RA (1990) The effect of hydrogen, oxygen, and water fugacity on oxygen diffusion in alkali feldspar. Geochim Cosmochim Acta 54:2953-2964 Farver JR, Yund RA (1991) Oxygen diffusion in quartz: dependence on temperature and water fugacity. Chem Geol 90:55-70 Filip J, Novák M, Beran A, Zbo il R (2006) Crystal chemistry and OH defect concentrations in spodumene from different granitic pegmatites. Phys Chem Minerals 32:733-746 Gerretsen J, Paterson MS, McLaren AC (1989) The uptake and solubility of water in quartz at elevated pressure and temperature. Phys Chem Minerals 16:334-342 Goldsmith JR (1986) The role of hydrogen in promoting Al-Si interdiffusion in albite (NaAlSi3O8) at high pressures. Earth Planet Sci Lett 80:135-138 Goldsmith JR (1987) Al/Si interdiffusion in albite: effect of pressure and the role of hydrogen. Contrib Mineral Petrol 95:311-321 Goldsmith JR (1988) Enhanced Al/Si diffusion in KAlSi3O8 at high pressures: the effect of hydrogen. J Geol 96:109-124 Grant K, Gleeson SA, Roberts S (2003) The high-temperature behavior of defect hydrogen species in quartz: Implications for hydrogen isotope studies. Am Mineral 88:262-270 Griggs D (1967) Hydrolytic weakening of quartz and other silicates. Geophys J Royal Astron Soc 14:19-31 Hammer VMF (1988) Quantitative IR-spectroscopic determination of structural OH-groups in natural rutiles of various occurrences. Z Kristallogr 185:631 Hammer VMF, Beran A (1991) Variations in the OH concentration of rutiles from different geological environments. Mineral Petrol 45:1-9 Hawthorne FC, Groat LA, Raudsepp M, Ball NA, Kimata M, Spike FD, Gaba R, Halden NM, Lumpkin GR, Ewing RC, Greegor RB, Lytle FW, Ercit TS, Rossman GR, Wicks FJ, Ramik RA, Sherriff BL, Fleet ME, McCammon C (1991) Alpha-decay damage in titanite. Am Mineral 76:370-396 Hofmeister AM, Rossman GR (1985a) A model for the irradiative coloration of smoky feldspar and the inhibiting influence of water. Phys Chem Minerals 12:324-332 Hofmeister AM, Rossman GR (1985b) A spectroscopic study of irradiation coloring of amazonite: structurally hydrous, Pb-bearing feldspar. Am Mineral 70:794-804

Water in Nominally Anhydrous Crustal Minerals

139

Hofmeister AM, Rossman GR (1986) A spectroscopic study of blue radiation coloring in plagioclase. Am Mineral 71:95-98 Hofmeister AM, Schaal RB, Campbell KR, Berry SL, Fagan TJ (1998) Prevalence and origin of birefringence in 48 garnets from the pyrope-almandine-grossularite-spessartine quaternary. Am Mineral 83:1293-1301 Ingrin J, Blanchard M (2006) Diffusion of hydrogen in minerals. Rev Mineral Geochem 62:291-320 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Ito Y, Nakashima S (2002) Water distribution in low-grade siliceous metamorphic rocks by micro-FTIR and its relation to grain size: A case from the Kanto Mountain region, Japan. Chem Geol 189:1-18 Johnson EA (2003) Hydrogen in nominally anhydrous crustal minerals. PhD Dissertation, California Institute of Technology, Pasadena, California Johnson EA (2005) Magmatic water contents recorded by hydroxyl concentrations in plagioclase phenocrysts from Mount St. Helens, 1980-1981. Geochim Cosmochim Acta 69:A743 Johnson EA, Manning CE, Antignano A, Tropper P (2005) OH in rutile: an oxygen and water barometer. Eos Trans AGU 86(52), Fall Meet. Suppl.:Abstract V41A-1420 Johnson EA, Rossman GR (2003) The concentration and speciation of hydrogen in feldspars using FTIR and 1H MAS NMR spectroscopy. Am Mineral 88:901-911 Johnson EA, Rossman GR (2004) A survey of hydrous species and concentrations in igneous feldspars. Am Mineral 89:586-600 Johnson EA, Rossman GR, Dyar MD, Valley JW (2002) Correlation between OH concentration and oxygen isotope diffusion rate in diopsides from the Adirondack Mountains, New York. Am Mineral 87:899-908 Johnson OW, DeFord J, Shaner JW (1973) Experimental technique for the precise determination of H and D concentration in rutile (TiO2). J Appl Phys 44:3008-3012 Kalt A (2000) Cordierite channel volatiles as evidence for dehydration melting: an example from hightemperature metapelites of the Bayerische Wald (Variscan belt, Germany). Eur J Mineral 12:987-998 Kats A (1962) Hydrogen in alpha quartz. Phillips Research Report 17:133-279 Kekulawala KRSS, Paterson MS, Boland JN (1981) An experimental study of the role of water in quartz deformation. Geophysical Monograph 24:49-60 Keppler H, Bolfan-Casanova N (2006) Thermodynamics of water solubility and partitioning. Rev Mineral Geochem 62:193-230 Khomenko VM, Langer K (1999) Aliphatic hydrocarbons in structural channels of cordierite: a first evidence from polarized single-crystal IR-absorption spectroscopy. Am Mineral 84:1181-1185 King PL, Hervig RL, Holloway JR, Delaney JS, Dyar MD (2000) Partitioning of Fe3+/Fetotal between amphibole and basanite melt as a function of oxygen fugacity. Earth Planet Sci Lett 178:97-112 Koch-Müller M, Dera P, Fei Y, Reno B, Sobolev N, Hauri E, Wysoczanski R (2003) OH- in synthetic and natural coesite. Am Mineral 88:1436-1445 Kohlstedt DL (2006) The role of water in high-temperature rock deformation. Rev Mineral Geochem 62: 377-396 Kolesov BA, Geiger CA (2000) Cordierite II: The role of CO2 and H2O. Am Mineral 85:1265-1274 Kronenberg (1994) Hydrogen speciation and chemical weakening of quartz. Rev Mineral 29:123-176 Kronenberg AK, Kirby SH, Aines RD, Rossman GR (1986) Solubility and diffusional uptake of hydrogen in quartz at high water pressures: Implications for hydrolytic weakening. J Geophys Res 91:12723-12744 Kronenberg AK, Wolf GH (1990) Fourier transform infrared spectroscopy determinations of intragranular water content in quartz-bearing rocks: implications for hydrolytic weakening in the laboratory and within the earth. Tectonophysics 172:255-271 Lager GA, Armbruster T, Faber J (1987a) Neutron and X-ray diffraction study of hydrogarnet Ca3Al2(O4H4)3. Am Mineral 72:756-765 Lager GA, Armbruster T, Rotella FJ, Rossman GR (1989) OH substitution in garnets: X-ray and neutron diffraction, infrared, and geometric modeling studies. Am Mineral 74:840-851 Lager GA, Rossman GR, Rotella FJ, Schultz AJ (1987b) Neutron-diffraction structure of a low-water grossular at 20 K. Am Mineral 72:766-768 Lamb WM, Valley JW (1988) Granulite facies amphibole and biotite equilibria, and calculated peakmetamorphic water activities. Contrib Mineral Petrol 100:349-360 Lathe C, Koch-Müller M, Wirth R, van Westrenen W, Mueller H-J, Schilling F, Lauterjung J (2005) The influence of OH in coesite on the kinetics of the coesite-quartz phase transition. Am Mineral 90:36-43 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H...O hydrogen bond lengths in minerals. Mh Chemie 130:1047-1059 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52 Locock A, Luth RW, Cavell RG, Smith DGW, Duke MJM (1995) Spectroscopy of the cation distribution in the schorlomite species of garnet. Am Mineral 80:27-38 Losos Z, Beran A (2004) OH defects in cassiterite. Mineral Petrol 81:219-234

140

Johnson

Lumpkin GR (1998) Rare-element mineralogy and internal evolution of the Rutherford no.2 pegmatite, Amelia County, Virginia: A classic locality revisited. Can Mineral 36:339-353 Maldener J, Rauch F, Gavranic M, Beran A (2001) OH absorption coefficients of rutile and cassiterite deduced from nuclear reaction analysis and FTIR spectroscopy. Mineral Petrol 71:21-29 Maldener J, Hösch A, Langer K, Rauch F (2003) Hydrogen in some natural garnets studied by nuclear reaction analysis and vibrational spectroscopy. Phys Chem Minerals 30:337-344 McLennan SM, Taylor SR (1999) Earth’s Continental Crust. In: Encyclopedia of Geochemistry. Marshall CP, Fairbridge RW (eds) Kluwer Academic Publishers, p 712 Miyoshi N, Yamaguchi Y, Makino K (2005) Successive zoning of Al and H in hydrothermal vein quartz. Am Mineral 90:310-315 Mosenfelder JL, Schertl HP, Smyth JR, Liou JG (2005) Factors in the preservation of coesite: The importance of fluid filtration. Am Mineral 90:779-789 Mottana A, Fusi A, Bianchi Potenza B, Crespi R, Liborio G (1983) Hydrocarbon-bearing cordierite from the Dervio-Colico road tunnel (Como, Italy). N Jb Mineral Abh 148:181-199 Nakano S, Makino K, Eriguchi T (2001) Microtexture and water content of alkali feldspar by Fourier-transform infrared microscopy. Mineral Mag 65:675-683 Nakashima S, Matayoshi H, Yuko T, Michibayashi K, Masuda T, Kuroki N, Yamagishi H, Ito Y, Nakamura A (1995) Infrared microspectroscopy analysis of water distribution in deformed and metamorphosed rocks. Tectonophysics 245:263-276 Nasdala L, Beran A, Libowitzky E, Wolf D (2001) The incorporation of hydroxyl groups and molecular water in natural zircon (ZrSiO4). Am J Sci 301:831-857 Nash DB (1962) Contact metamorphism at Birch Creek, Blanco Mountain Quadrangle, Inyo County, California. MS Dissertation, University of California, Berkeley, California Niimi N (2002) Static recrystallization of the deformed quartz in the granite from Mt. Takamiyama. J Geosci 45:89-100 O’Neil JR, Taylor Jr. HP (1967) The oxygen isotope and cation exchange chemistry of feldspars. Am Mineral 52:1414-1437 O’Neill B, Bass JD, Rossman GR (1993) Elastic properties of hydrogrossular garnet and implications for water in the upper mantle. J Geophys Res 98 (B11):20,031-20,037 Ord A, Hobbs BE (1986) Experimental control of the water-weakening effect in quartz. Geophysical Monograph 36:51-72 Peslier AH, Luhr J, Post JE (2002) Low water contents in pyroxenes from spinel-peridotites of the oxidized, sub-arc mantle wedge. Earth Planet Sci Lett 201:69-86 Post A, Tullis J (1998) The rate of water penetration in experimentally deformed quartzite: Implications for hydrolytic weakening. Tectonophysics 295:117-137 Rossman GR (1988) Vibrational spectroscopy of hydrous species. Rev Mineral 18:193-206 Rossman GR (1996) Studies of OH in nominally anhydrous minerals. Phys Chem Minerals 23:299-304 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Rossman GR, Aines RD (1986) Spectroscopy of a birefringent grossular from Asbestos, Quebec, Canada. Am Mineral 71:779-780 Rossman GR, Aines RD (1991) The hydrous components in garnets: Grossular-hydrogrossular. Am Mineral 76: 1153-1164 Rossman GR, Rauch F, Livi R, Tombrello TA, Shi CR, Zhou ZY (1988) Nuclear reaction analysis of hydrogen in almandine, pyrope, and spessartite garnets. N Jb Mineral Mh 4:172-178 Rossman GR, Smyth JR (1990) Hydroxyl contents of accessory minerals in mantle eclogites and related rocks. Am Mineral 75:775-780 Rovetta MR, Holloway JD, Blacic JD (1986) Solubility of hydroxyl in natural quartz annealed in water at 900°C and 1.5GPa. Geophys Res Lett 13:145-148 Seaman SJ, Dyar MD, Marinkovic N, Dunbar NW (2006) An FTIR study of hydrogen in anorthoclase and associated melt inclusions. Am Mineral 91:12-20 Shieh YN, Taylor Jr. HP (1969) Oxygen and carbon isotope studies of contact metamorphism of carbonate rocks. J Petrol 10:307-331 Skogby H (2006) Water in natural mantle minerals I: pyroxenes. Rev Mineral Geochem 62:155-167 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: variations in the natural environment. Am Mineral 75:764-774 Skogby H, Rossman GR (1989) OH− in pyroxene; an experimental study of incorporation mechanisms and stability. Am Mineral 74:1059-1069 Smyth JR (2006) Hydrogen in high pressure silicate and oxide mineral structures. Rev Mineral Geochem 62: 85-115 Solomon GC, Rossman GR (1988) NH4+ in pegmatitic feldspars from the southern Black Hills, South Dakota. Am Mineral 73:818-821

Water in Nominally Anhydrous Crustal Minerals

141

Solomon GC, Taylor HP Jr (1991) Oxygen isotope studies of Jurassic fossil hydrothermal systems, Mojave Desert, southeastern California. In: Stable Isotope Geochemistry: A Tribute to Samuel Epstein. Taylor HP Jr, O’Neil JR, Kaplan IR (eds), The Geochemical Society, Special Publication Vol 3, p 449-462 Swamy V, Godhavari KS, Menon AG, Anantha Iyer GV (1992) Channel volatiles of south Indian cordierites as indicators of metamorphic fluid composition. N Jb Mineral Mh 8:359-375 Swope RJ, Smyth JR, Larson AC (1995) H in rutile-type compounds: I. Single-crystal neutron and X-ray diffraction study of H in rutile. Am Mineral 80:448-453 Tullis J, Yund RA (1980) Hydrolytic weakening of experimentally deformed Westerly granite and Hale albite rock. J Struct Geol 2:439-451 Van Baalen MR (1993) Titanium mobility in metamorphic systems: a review. Chem Geol 110:233-249 Vedder W, Wilkins RWT (1969) Dehydroxylation and rehydroxylation, oxidation and reduction of micas. Am Mineral 54:482-509 Visser D, Kloprogge JT, Maijer C (1994) An infrared spectroscopic (IR) and light element (Li, Ba, Na) study of cordierites from the Bamble Sector, South Norway. Lithos 32:95-107 Vlassopoulos D, Rossman GR, Haggerty SE (1993) Coupled substitution of H and minor elements in rutile and the implications of high OH contents in Nb- and Cr-rich rutile from the upper mantle. Am Mineral 78: 1181-1191 Vry JK, Brown PE, Valley JW (1990) Cordierite volatile content and the role of CO2 in high-grade metamorphism. Am Mineral 75:71-88 Wilkins RWT, Sabine W (1973) Water content of some nominally anhydrous silicates. Am Mineral 58:508-516 Woodhead JA, Rossman GR, Thomas AP (1991a) Hydrous species in zircon. Am Mineral 76:1533-1546 Woodhead JA, Rossman GR, Silver LT (1991b) The metamictization of zircon: Radiation dose-dependent structural characteristics. Am Mineral 76:74-82 Xia Q, Pan Y, Chen D, Kohn S, Zhi X, Guo L, Cheng H, Wu Y (2000) Structural water in anorthoclase megacrysts from alkalic basalts: FTIR and NMR study. Acta Petrol Sinica 16:485-491 Yund RA (1983) Diffusion in feldspars. Rev Mineral 2:203-222 Yund RA, Ackermand D (1979) Development of perthite microstructures in the Storm King Granite, N.Y. Contrib Mineral Petrol 70:273-280 Yund RA, Anderson TF (1978) The effect of fluid pressure on oxygen isotope exchange between feldspar and water. Geochim Cosmochim Acta 42:235-239 Zhang M, Groat L, Salje EKH, Beran A (2001) Hydrous species in crystalline and metamict zircons. Am Mineral 86:904-909

142

Johnson

APPENDIX DATA TABLES OF HYDROUS SPECIES CONCENTRATIONS IN NOMINALLY ANHYDROUS CRUSTAL MINERALS The following data tables summarize the quantitative measurements of hydrous species discussed or graphed in this chapter. Published works containing infrared spectra of quartz and cordierite are listed according to date; individual measurements are listed under each referenced work for feldspars, pyroxenes, garnets, kyanite, sillimanite, rutile, and cassiterite. Page 143

Studies of hydrogen in quartz.

Page 143

Geological studies of H2O and CO2 in cordierite.

Page 144

Hydrous species in feldspars.

Page 146

Structural hydroxyl concentrations in crustal and mantle pyroxenes.

Page 148

Structural hydroxyl concentrations in crustal garnets.

Page 152

Structural hydroxyl concentrations in kyanite.

Page 152

Structural hydroxyl concentrations in sillimanite.

Page 153

Structural hydroxyl concentration in andalusite.

Page 153

Structural hydroxyl concentrations in rutile.

Page 154

Structural hydroxyl concentrations in cassiterite.

Water in Nominally Anhydrous Crustal Minerals

143

Studies of hydrogen in quartz. Hydrolytic Weakening and Deformation Studies Reference

Title

Griggs (1967)

Hydrolytic weakening of quartz and other silicates

Kekulawala et al. (1981)

An experimental study of the role of water in quartz deformation

Kronenberg et al. (1986)

Solubility and diffusional uptake of hydrogen in quartz at high water pressures: Implications for hydrolytic weakening

Rovetta et al. (1986)

Solubility of hydroxyl in natural quartz annealed in water at 900 °C and 1.5 GPa

Ord and Hobbs (1986)

Experimental control of the water-weakening effect in quartz

Cordier et al. (1988)

Water precipitation and diffusion in wet quartz and wet berlinite AlPO4

Gerretsen et al. (1989)

The uptake and solubility of water in quartz at elevated pressure and temperature

Kronenberg and Wolf (1990)

Fourier transform infrared spectroscopy determinations of intragranular water content in quartz-bearing rocks: Implications for hydrolytic weakening in the laboratory and within the earth

Post and Tullis (1998)

The rate of water penetration in experimentally deformed quartzite: Implications for hydrolytic weakening

Geologic Surveys Reference

Title

Berzina et al. (1991)

Water in quartz of Cu-Mo deposits

Nakashima et al. (1995)

Infrared microspectroscopy analysis of water distribution in deformed and metamorphosed rocks

Ito and Nakashima (2002)

Water distribution in low-grade siliceous metamorphic rocks by micro-FTIR and its relation to grain size: A case from the Kanto Mountain region, Japan

Niimi (2002)

Static recrystallization of the deformed quartz in the granite from Mt. Takamiyama

Geological studies of H2O and CO2 in cordierite. Reference

Title

Armbruster et al. (1982)

Very high CO2 cordierite from Norwegian Lapland: Mineralogy, petrology and carbon isotopes

Mottana et al. (1983)

Hydrocarbon-bearing cordierite from the Dervio-Colico road tunnel (Como, Italy)

Vry et al. (1990)

Cordierite volatile content and the role of CO2 in high-grade metamorphism

Swamy et al. (1992)

Channel volatiles of south Indian cordierites as indicators of metamorphic fluid composition

Visser et al. (1994)

An infrared spectroscopic (IR) and light element (Li, Be, Na) study of cordierites from the Bamble Sector, South Norway

Kalt (2000)

Cordierite channel volatiles as evidence for dehydration melting: an example from high-temperature metapelites of the Bayerische Wald (Variscan belt, Germany)

144

Johnson

Hydrous species in feldspars. Sample Number

Feldspar composition/ structure

Occurrence

15020 15020 15021

Data from Hofmeister (1985a) amazonite; deep blue pegmatite amazonite; green pegmatite amazonite pegmatite

NSC20 PSSC3 HNBC4

anorthoclase anorthoclase anorthoclase

968 1281 1533 2066 CIT19237 CIT19249 146 1275 1618 CIT10774 2063 680 901 904 1605 1608 1609 1610 2066 580 681 HPT 1532 638 JB1 JV1 1554 1276a 1277 1280 1276b 1389 1604 15 25 145 289 1613 1615 1679 1920 1606

Data from Xia et al. (2000) alkali basalt alkali basalt alkali basalt

Species*

ppm H2O wt.

H2O H2O H2O

650 900 1100

OH OH, H2O? OH

405 915 365

Data from Johnson and Rossman and references therein (2004) microcline pegmatite H2O microcline pegmatite H2O microcline pegmatite NH4+ microcline pegmatite H2O fi microcline pegmatite fi NH4+ microcline pegmatite fi orthoclase pegmatite OH orthoclase pegmatite OH orthoclase (adularia) pegmatite OH, H2O orthoclase pegmatite alteration fi sanidine pegmatite OH albite pegmatite OH albite pegmatite fi OH albite pegmatite OH albite pegmatite fi OH albite pegmatite OH fi albite pegmatite OH fi albite pegmatite fi albite pegmatite fi OH oligoclase pegmatite OH oligoclase pegmatite OH anorthite pegmatite alteration fi hyalophane pegmatite NH4+ sanidine volcanic OH sanidine volcanic OH sanidine volcanic OH anorthoclase volcanic OH anorthoclase volcanic OH anorthoclase volcanic OH oligoclase volcanic OH andesine volcanic OH andesine volcanic OH andesine volcanic OH labradorite volcanic OH labradorite volcanic OH labradorite volcanic nd labradorite volcanic OH labradorite volcanic OH labradorite volcanic OH labradorite volcanic OH labradorite volcanic nd bytownite volcanic OH

1000 1350 542 135 196 674 454 878 4 5 80 1949 1351 1 192 248 247 39 114 37 13 518 290 21 1401 574 1500 170 93 14 238 322 270 230 249 510 4 0 125 0 18 21 4 80 0 46

Water in Nominally Anhydrous Crustal Minerals

145

Hydrous species in feldspars (continued) Sample Number

Feldspar composition/ structure

Occurrence

Species*

Data from Johnson and Rossman and references therein (2004) anorthite volcanic OH anorthite volcanic OH anorthite volcanic OH anorthite volcanic mi plagioclase volcanic OH plagioclase volcanic mi plagioclase volcanic OH feldspar volcanic mi alkali feldspar plutonic fi alteration HPT alkali feldspar plutonic fi alteration S-12 HPT oligoclase plutonic inclusions CIT19224 andesine plutonic fi alteration 1550 andesine plutonic fi alteration CIT19234 labradorite plutonic fi alteration CIT19232 bytownite plutonic inclusions CIT19230 anorthite plutonic fi alteration CIT19231 plagioclase plutonic fi alteration B46c HPT plagioclase plutonic fi alteration 81-205 HPT plagioclase plutonic alteration D81-75 HPT plagioclase plutonic fi alteration 81-SR3 plagioclase plutonic nd 83-AUS-11 plagioclase plutonic nd RT3#1 plagioclase plutonic inclusions GCS026 (PUP11) feldspar plutonic fi PI 210 HPT feldspar plutonic alteration Sky-9 HPT feldspar plutonic fi Plutonic Series: Fry Mountains Pluton GCS349 plagioclase plutonic fi GCS348 plagioclase plutonic fi GCS284 orthoclase plutonic fi GCS283 plagioclase plutonic fi GCS280 plagioclase plutonic fi Plutonic Series: Skaergaard Intrusion KG-109 HPT plagioclase plutonic OH alteration SK-15 HPT plagioclase plutonic fi alteration KG-236 HPT plagioclase plutonic fi alteration Plutonic Series: Southern Idaho Batholith RH 68 HPT feldspar plutonic fi RB 163 HPT alkali feldspar plutonic fi 1597 1884 1968 M-1 144 97KC02 URA5 ARG4b B45 HPT

Mt. Erebus

Data from Seaman et al. (2006) anorthoclase volcanic

*nd = not determined; fi = fluid inclusions; mi = melt inclusions

OH

ppm H2O wt.

180 115 210 56 1 46 25 11 700 362 1299 607 2092 1374 1008 287 350 65 6 77 369 338 322 307 1992 3426 1346 831 421 0 0 263 382 3900 359 812 1511 2321 1375 4265 82 121 409 3786 1078 666 123 2161 126

146

Johnson

Structural hydroxyl concentrations in crustal and mantle pyroxenes. Sample Number

Pyx. composition

47 20 30 13 17 16 7 32 48 33 50 40 1 3 9 46 6 37 5 39 18 12 8 10 51 19 28 2 38 21 31 4 36 35 41 14 29 25 34

omphacite augite aegirine diopside augite diopside diopside enstatite enstatite enstatite diopside diopside spodumene diopside augite augite diopside enstatite diopside hedenbergite enstatite diopside aegirine-augite diopside augite enstatite diopside diopside enstatite diopside enstatite diopside enstatite esseneite diopside diopside aegirine aegirine fassaite

PMR-53 KBH-1

augite enstatite

Data from Bell et al. (1995) megacryst, kimberlite megacryst, kimberlite

268 186

95ADK1A 95AK8f 95AK24 95AK6

diopside diopside diopside diopside

Data from Johnson et al. (2002) granulite-facies marble granulite-facies marble granulite-facies marble granulite-facies marble

138 134 81 55

SLP-402 SLP-402 SLP-400 SLP-403 SLP-403 SLP-101 SLP-101 SLP-114 SLP-114 SLP-405 SLP-405 SLP-142

orthopyroxene clinopyroxene orthopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene

Data from Peslier et al. (2002) mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith

189 413 140 243 528 155 398 107 222 171 373 280

Occurrence Data from Skogby et al. (1990) eclogite xenolith in kimberlite xenolith in basalt authigenic high-grade pyroxeneite xenocryst mantle diatreme xenocryst mantle diatreme megacryst in basalt metamorphic megacryst in kimberlite metamorphic metamorphic? xenolith in basalt granitic pegmatite metamorphic volcanic volcanic? metamorphic boninite lava metamorphic metamorphic lower crustal metamorphic neph. sy. pegmatite metamorphic gabroic xenolith, basalt basalt xenolith rhyolitic pumice metamorphic low-P metamorphic oxidized alkali picrite megacryst calcite vein megacryst, anorthosite buchite metamorphic metamorphic limestone nepheline syenite neph. sy. pegmatite meteorite

wt% OH

[OH] ppm H2O wt.*

0.12 0.1 0.088 0.085 0.073 0.073 0.066 0.066 0.05 0.05 0.045 0.037 0.041 0.028 0.027 0.026 0.025 0.022 0.018 0.018 0.016 0.016 0.015 0.014 0.014 0.012 0.011 0.009 0.008 0.007 0.006 0.006 0.005 0.004 0.004 0.003 0.002 0.001 0

636 530 466 450 387 387 350 350 265 265 238 196 217 148 143 138 132 117 95 95 85 85 79 74 74 64 58 48 42 37 32 32 26 21 21 16 11 5 0

Water in Nominally Anhydrous Crustal Minerals

147

Structural hydroxyl concentrations in crustal and mantal pyroxenes (continued). Sample Number

Pyx. composition

SLP-142 DGO-166 DGO-166 DGO-160 DGO-160 SIN-3 BCN-200D BCN-200D BCN-130 BCN-130 BCN-201B BCN-201B BCN-203 SIM-9c SIM-9c SIM-24 SIM-24 SIM-3 SIM-3

clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene orthopyroxene clinopyroxene

Sp14 Sp16 Sp26 SpKor Sp01 Sp03 Sp04 Sp05 Sp06 Sp12 Sp13 Sp17 Sp25 Sp15 Sp20 Sp21 Sp22 Sp23 Sp02 Sp11 Sp18 Sp28 Sp07 Sp08 Sp10 Sp19 Sp24 Sp27a Sp27b Sp29 SpPak

spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene spodumene

Occurrence

wt% OH

[OH] ppm H2O wt.*

Data from Peslier et al. (2002) mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith mantle wedge xenolith

514 166 387 86 256 288 190 477 128 313 203 342 39 71 205 115 158 109 140

Data from Filip et al. (2006) granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite granitic pegmatite

1.4 0.89 0.82 1.73 3.78 3.43 1.21 0.44 1.09 0.61 2.49 0.63 1.81 0.35 0.41 0.42 0.72 1.26 0.53 0.57 0.64 0.49 0.4 0.32 0.24 0.25 0.39 0.21 0.29 0.37 0.13

* [OH] calculated using Bell et al. (1995) calibrations, except for samples from Skogby et al. (1990), which use calibration from that study (same within error of Bell et al. (1995) calibration), and Filip et al. (2006), which use the calibration from Libowitzky and Rossman (1997). †Maximum reported concentration for each sample.

148

Johnson

Structural hydroxyl concentrations in crustal garnets. Sample Number

Garnet composition

Occurrence

wt% H2O

2 4 5 10 12 13 35 49 104 113 114

Data from Aines and Rossman (1984b) Spessartine Pegmatite Spessartine Pegmatite Almandine-Spessartine Pegmatite Spessartine-Almandine ? Spessartine-Almandine Pegmatite? Grossular Altered ultramafic body Almandine-Pyrope Gneiss Pyrope Diatreme megacryst Almandine-Spessartine Pegmatite Pyrope Diatreme Pyrope Diatreme

0.084 0.25 0.36 0.059 0.17 0.18 0.06 0.08 0.27 0.15 0.22

42 1359 1360 1409 1411

Data from Rossman and Aines (1991) Grossular Skarn Grossular Metarodingite Grossular Metarodingite Grossular Skarn Grossular Skarn

0.1 0.28 0.85 0.185 0.08

Ice River

Data from Locock et al. (1995) Schorlomite Alkaline complex

0.036

GRR1669 GA33 GA34 GRR48 rim GRR134 GRR169 GRR684 GRR1263 GRR1328 GRR1765 GA32 GA35 GRR54 GRR149 GRR1015 GRR1137a2 GRR1447 GRR1448 GA24 GA36 GRR1446 CITH3110

Data from Amthauer and Rossman (1998) Hydroandradite Basalt vug Andradite Serpentinite Andradite Serpentinite Andradite Serpentinite Andradite Serpentinite Andradite Serpentinite Melanite Serpentinite Melanite Serpentinite Melanite Serpentinite Demantoid Serpentinite Andradite Skarn Andradite Skarn Andradite Skarn Andradite Skarn Andradite Skarn Andradite Skarn Andradite Skarn Andradite Skarn Melanite Phonolite Melanite Magmatic Melanite Magmatic Melanite Magmatic

5.92 0.01 0.07 0.1 0.09 0.15 1.17 0.04 2.45 0.1 0.15 0.05 0.38 0.04 1.44 0.4 0.03 0.01 0.01 0.03 0.04 0.02

Ru. 1 Ru. 2 Ru. 3 Ru. 4 Ru. 5 Ru. 6 Ru. 7 Ru. 8 Ru. 9 Ru. 10 Him. 1 Him. 2 Him. 3 Him. 4 GAB 0 GAB 30 GAB 60

Data from Arredondo et al. (2001) Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite

0.023 0.022 0.083 0.127 0.128 0.134 0.13 0.158 0.147 0.121 0.007 0.009 0.019 0.034 0.0026 0.0034 0.004

Water in Nominally Anhydrous Crustal Minerals

149

Structural hydroxyl concentrations in crustal garnets (continued). Occurrence

wt% H2O

Sample Number

Garnet composition

GAB 90 GAB 120 GAB 150 GAB 180 GAB 240 GAB 300 GAB 330 GAB 420 GAB 480 GAB 540 GAB 600 GAB 660 GAB 720 GAB 780 GAB 810

Data from Arredondo et al. (2001) Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite Spessartine Pegmatite

0.006 0.001 0.0025 0.0031 0.0047 0.0042 0.0055 0.0035 0.0009 0.0004 0.0077 0.0025 0.0015 0.0035 0.0073

BC2B1 BC2B3 BC2B4 BC-11 W77-102 99W-2 99W-2 W50-419 99W-1 99W-4 99W-4 W81-15 148′2″ W81-15 148′2″ W81-15 149′9″ W81-15 149′9″ W81-15 150′9″ W81-15 150′9″ W81-15 150′9″ W81-15 321′ W81-15 321′ W81-15 321′ W81-15 269′4″ W81-15 269′10″ W81-15 269′10″ W81-15 270′5″ W81-15 273′ W81-15 276′2″ W81-15 276′2″ W81-15 276′5″ W81-15 278′5″ W81-15 282′5″ W81-15 282′5″ W81-15 285′ W81-15 285′ W81-15 299′6″ W81-15 299′6″ W81-15 301′8″ W81-15 301′8″ W81-15 304′ W81-15 304′ W81-15 307′ W81-15 311′ W81-15 313′6″ W81-15 313′6″ W81-15 313′6″ W81-15 314′ W81-15 314′ W81-15 318′7″ W81-15 318′7″

Data from Johnson (2003) Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn

0.331 0.311 0.311 0.182 0.0592 0.1862 0.1842 0.0277 0.0466 0.0342 0.0211 0.0239 0.0282 0.0281 0.0247 0.0238 0.0171 0.0268 0.0524 0.0487 0.0363 0.0097 0.0161 0.0164 0.0494 0.0489 0.1264 0.1619 0.1117 0.0893 0.0865 0.0839 0.0177 0.0255 0.0922 0.1037 0.0751 0.0801 0.1453 0.1163 0.1165 0.0271 0.0333 0.0336 0.0304 0.0033 0.0040 0.0381 0.0417

150

Johnson

Structural hydroxyl concentrations in crustal garnets (continued). Sample Number

Garnet composition

Occurrence

W81-15 328′6″ W81-15 347′7″ W81-15 347′7″ W81-15 370′ W81-15 396′8″ W81-15 396′8″ W81-15 404′3″ W81-15 404′3″ W81-15 407′2″ W81-15 407′2″ W81-15 411′ W81-15 411′ W81-15 33_′0″ W81-15 33_′0″ 103′3″ 103′3″ 113′6″ 113′6″ 134′4″ 136′8″ And. 136′8″ Gros. 136′8″ Gros. 138′10″ 139′9″ gt 139′9″ gt 139′9″ 139′9″ 140′7″ 140′7″ 140′7″ 144′2″ 144′2″ 144′6″ And. 148′3″ 148′3″ 154′5″ 154′5″ 154′5″ 158′2″ 158′5″ 162′6″ 162′6″ 167′8″ 167′8″ 171′6″ 174′9″ 174′9″ 177′ 177′ 190′ 194′6″ 194′6″ 200′3″ 205′9″ 206′2″ 210′6″ 215′8″ 218′8″ 219′8″ 219′8″ 221′ 226′ 226′ 238′3″ 238′3″

Data from Johnson (2003) Grossular-andradite Skarn Almandine Anorthosite Almandine Anorthosite Almandine Anorthosite Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Grossular-andradite Skarn

wt% H2O 0.0039 0.0000 0.0000 0.0039 0.0294 0.0410 0.1333 0.1353 0.1653 0.1659 0.0383 0.0597 0.0474 0.0712 0.0485 0.0703 0.0159 0.0180 0.0669 0.0778 0.0619 0.0814 0.0751 0.0798 0.0877 0.0707 0.0780 0.0439 0.0720 0.0783 0.1402 0.1480 0.1175 0.1081 0.1075 0.0070 0.0112 0.0108 0.0118 0.0236 0.0466 0.0822 0.0293 0.0285 0.0725 0.0751 0.0795 0.0155 0.0142 0.0565 0.0323 0.0349 0.0524 0.0829 0.1016 0.0832 0.1087 0.0912 0.1077 0.0970 0.0789 0.0857 0.1163 0.1451 0.1385

Water in Nominally Anhydrous Crustal Minerals

151

Structural hydroxyl concentrations in crustal garnets (continued). Occurrence

wt% H2O

Sample Number

Garnet composition

238′8″ 239′10″ 239′10″ 245′6″ 245′6″ 247′10″ 247′10″ 248′2″ 248′2″ 248′2″ 251′ 253′ 254′6″ 254′6″ 259′8″ 259′11″ 259′11″ 261′1″ 261′1″ 267′ 267′ 271′9″ 271′9″ 278′6″ 278′6″ 282′3″ 282′3″ 283′4″ 285′6″ 285′6″ 287′2″ 289′3″ 289′3″

Data from Johnson (2003) Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Skarn Grossular-andradite Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn Grossular-andradite Skarn

0.1185 0.0221 0.0354 0.1210 0.1217 0.0950 0.1144 0.0505 0.1111 0.1143 0.0610 0.0692 0.1014 0.1121 0.1021 0.1105 0.1221 0.0914 0.1131 0.0626 0.1088 0.1483 0.1490 0.0806 0.0977 0.0874 0.0909 0.0845 0.0787 0.1022 0.0393 0.0250 0.0460

RHOTAN1 PYALTAN RHOTAN2 HAI RAJA GTALX SPESSOR HESSI(thick) HESSI(thin) TSAV MALI GRMALI

Data from Maldener et al. (2003) Pyrope-almandine Pyrope-almandine Pyrope-almandine Almandine-pyrope Almandine-pyrope Almandine-pyrope Spessartine Grossular Grossular Grossular Grossular-andradite Grossular-andradite

0.0019 0.0017 0.0018 0.0018 0.0014 0.0048 0.0025 0.0950 0.0870 0.0480 0.0170 0.0190

152

Johnson

Structural hydroxyl concentrations in kyanite. Sample Rock Type Number

P (kbar)

T (ºC)

[OH] (ppm H2O [OH] ppm wt%) using Bell H2O wt% et al. (2004)

1 2 3 4 5 6 7 8 9 10

Data from Beran and Götzinger (1987) and references within Granulite 5-10 530-710 50 Granulite 5-10 530-710 50 Gt-ky-st-mica schist ? ? 50 Cor-fuchsite-ky greenstone belt 5-7 550-650 100 Eclogite ? ? 450 Eclogite 6-10 330-550 500 Eclogite 5.5-9 500-660 750 Eclogite 6-10 330-550 800 Gneiss (ky-st-qtz-fsp) 4-8 500-600 800 Eclogite 28-32 750-850 1800

G1/Ky3

Eclogite

H105650 H85943 LTL-3 FSM-15

Eclogite Eclogite? Eclogite Eclogite

3 3 3 6 25 28 42 44 44 100

Data from Beran et al. (1993) 410

23

Data from Bell et al. (2004a) 4 22 27 230

Structural hydroxyl concentrations in sillimanite. Sample Number

Rock Type*

GRR273 GRR439 GRR385 GRR380 GRR1585 GRR384 GRR547 GRR383 GRR450 GRR382 GRR608 GRR386 GRR311 GRR509 GRR306 GRR560 GRR1485 AB-Gho AB-Sm1 GRR799 GRR802

HG HG X HG UA UA UA UA UA ? PG UA X AG X PG AG UA AG HG PG

Total OH band absorbance (cm−1)

Data from Beran et al. (1989) 0.408 0.472 0.096 0.547 0.056 0 0.744 0.452 0.791 1.263 0.193 0.908 0 1.14 0.351 0 0.085 1.129 1.848 0.924 0

Est. [OH] ppm H2O wt%† 44 51 10 59 6 0 81 49 86 137 21 98 0 123 38 0 9 122 200 100 0

* AG = alluvial deposit, presumably from granulite facies; HG = hornblende granulite facies; PG = pyroxene granulite facies; UA = upper amphibolite grade; X = xenolithic. † OH concentration estimated from thermal weight loss data in Beran et al. (1989).

Water in Nominally Anhydrous Crustal Minerals

153

Structural hydroxyl concentration in andalusite. Sample Number

Locality

[OH] ppm H2O wt%*

Data from Rossman (1996) GRR273

Minas Gerais, Brazil

270

Structural hydroxyl concentrations in rutile. Sample Number

Rock Type

[OH] ppm H2O wt%*

[OH] (ppm H2O wt%) using Maldener et al. (2001)

EJ-86b

eclogite

Data from Rossman and Smyth (1990) 240

2257

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Data from Hammer and Beran (1991) pegmatite 900 pegmatite 1300 pegmatite 1000 unknown 900 pegmatite 900 pegmatite 1600 pegmatite 900 xenolith? 1600 phyllite 1000 amphibolite 1800 mica schist 1800 amphibolite 1400 chl-mica schist 1500 chl-mica schist 1100 amphibolite 1900 amphibolite 2100 amphibolite 1600 eclogite 800 blueschist (pyrope incl.) 400

310 447 344 310 310 551 310 551 344 620 620 482 516 379 654 723 551 275 138

R-1 R-2 R-4 R-5 R-6 R-7 R-8 JAG83-30-1 JAG83-30-2 JAG83-30-3 JAG85-2

Data from Vlassopoulos et al. (1993) Ky+rut+lazulite 2900 Ky+rut+lazulite 3100 carbonatite 1200 unknown 1000 mt+ap+chl schist 2800 hydrothermal vein 1200 hydrothermal vein 4800 rt-ilm xenolith 5400 rt-ilm xenolith 6000 rt-ilm xenolith 8000 rt-ilm xenolith 2500

1001 1074 416 333 962 401 1666 1887 2084 2768 868

R5 R5a R6 R7a

pegmatite pegmatite pegmatite pegmatite

Data from Maldener et al. (2001) 70 330 820 270

* Determined using the integrated absorption coefficient from Hammer and Beran (1991).

154

Johnson

Structural hydroxyl concentrations in cassiterite. Sample Number

Locality

[OH] ppm H2O wt%

CC ZZ

Data from Maldener et al. (2001) Cornwall Zinnwald

170 160

4 6 10 11 12 13 14 15 16 17 18

Data from Losos and Beran (2004) Greisen Greisen Greisen Hydrothermal vein ? Cassiterite-sulphide deposit Tin melt (synthetic) Tin melt (synthetic) Greisen/quartz vein ? Cut gemstone

49 110 93 120 32 170 2 8 90 32 82

7

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 155-167, 2006 Copyright © Mineralogical Society of America

Water in Natural Mantle Minerals I: Pyroxenes Henrik Skogby Department of Mineralogy Swedish Museum of Natural History Box 50007, SE-10405 Stockholm, Sweden e-mail: [email protected]

INTRODUCTION A large number of studies of water in pyroxenes have confirmed that essentially all pyroxenes of terrestrial origin as a rule contain substantial amounts of OH, varying as a function of chemical composition and geological occurrence. The highest OH-concentrations are observed in mantle-derived samples, and it is now well established that both clino- and orthopyroxene can be considered as major host minerals for water in the mantle (e.g., Bell and Rossman 1992; Bell et al. 2004; Bolfan-Casanova 2005). The first firm observations of OH in pyroxenes were made more than three decades ago. Even earlier, the “excess” water recorded in many wet-chemical analyses of pyroxenes and other nominally anhydrous minerals had been taken as indications of structurally bonded water, although it was difficult to confirm whether this “excess” water was really due to intrinsic OH, or if it was caused by fluid or solid inclusions, or contaminants on cracks and surfaces. A water incorporation model based on OH ions replacing O at their normal structural positions in pyroxenes was proposed by Martin and Donnay (1972), who also suggested that OH-containing nominally anhydrous minerals could provide the main storage mechanism for water in the mantle. Somewhat later Wilkins and Sabine (1973) published IR spectra of several nominally anhydrous minerals, including a diopside from crustal occurrence (Fig. 1). The spectra contained sharp OH absorption bands which were verified by deuterium exchange, and the water contents were also analyzed by an electrolytic technique, which showed considerably lower concentrations than what had earlier been indicated from wet-chemical analyses. Also Runciman et al. (1973) observed absorption bands in the OH range in a spectroscopic study in the full infrared to UV-range of an enstatite from Kimberley, South Africa, which they suspected to be caused by traces of OH. A more detailed study based on polarized IR spectra

Figure 1. IR transmission spectrum of diopside from Rotenkopf, Tyrol. Modified after Wilkins and Sabine (1973).

1529-6466/06/0062-0007$05.00

DOI: 10.2138/rmg.2006.62.7

156

Skogby

of diopside was published by Beran (1976), who presented an incorporation model based on crystal-chemical arguments and the pleochroic behavior of the absorption bands. The study was followed up by an investigation of additional pyroxenes (Beran 1981), some of which were of high-pressure origin. An incorporation model for OH also in enstatite was presented by Beran and Zemann (1986) based on polarized IR spectroscopy. Later studies addressed the stability relations of OH in the pyroxene structures by experimental methods; Ingrin et al. (1989) and Skogby and Rossman (1989) showed that hydrogen could be diffused out from the structure at elevated temperatures, but also restored under reducing conditions, with a thermal stability in the same order as OH in amphiboles and micas. The generality of the occurrence of OH in pyroxenes was surveyed by Skogby et al. (1990). The study comprised over 50 samples from various types of geological environments, and showed that OH occurs in terrestrial pyroxenes as a rule, with the highest concentrations recorded in samples from mantle origin. Much of the more recent interest has been focused on samples of mantle origin, due to the perspectives of pyroxenes forming a mantle reservoir for water, together with other NAMs (nominally anhydrous minerals), and the imposed effects on mantle properties (e.g., Smyth et al. 1991; Bell and Rossman 1992; Rossman 1996; Peslier et al. 2002; Bell et al. 2004; KochMüller et al. 2004). However, also crustal pyroxenes have been considered in some studies, (e.g., Johnson et al. 2002), as well as pyroxenes occurring in subducted crustal rocks (e.g., Katayama et al. 2003, 2005).

OH ABSORPTION BANDS IN IR SPECTRA The vast majority of studies of water species in mantle pyroxenes have been performed using FTIR spectroscopy. Apart from estimates of water concentration, IR spectra provide valuable information regarding the local OH dipole environment, as orientation and possible O-H..O distances (cf. Libowitzky and Beran 2006). The different types of OH bands observed in IR spectra of pyroxenes are discussed in the following section. As the OH bands in samples from mantle and crustal origin are similar, both types of occurrences are considered. A more thorough overview of the occurrence of water in crustal pyroxenes is given in Johnson (2006).

Diopside Infra-red spectra of diopside normally contain four major OH absorption bands, which show two types of pleochroic behavior (Fig. 2). One band occurs around 3640 cm−1 and is strong in the α and β directions, but weak or absent in the γ direction. Three bands occurring close to 3535, 3460 and 3355 cm−1 have a different pleochroism with γ as the strongest direction, and weaker absorption parallel α and β. The band at 3355 cm−1 is sometimes absent in mantle diopside (Peslier et al. 2002). The pleochroic schemes observed for different pyroxene minerals are summarized in Table 1.

Augite Augite spectra are normally rather similar to diopside spectra, with one band

Figure 2. Polarized IR spectra of diopside from a basalt megacryst, Maui, Hawaii, normalized to 1 mm thickness. Sample no. 7 from Skogby et al. (1990).

Water in Natural Mantle Minerals I: Pyroxenes

157

Table 1. Pleochroism of OH bands in clino- and orthopyroxene spectra. Wavenumber (cm−1) Mineral 3630 - 3640

3530 - 3540

3460

3355

Diopside Augite Omphacite

α = β, γ = 0 α = β, γ = 0 α = β, γ = 0 *

γ>α=β γ>α=β γ>α=β

γ>α=β γ>α=β γ>α=β

γ>α=β* — —

3600 - 3610

3410 - 3560

3060 - 3300

Orthopyroxene

α = β, γ = 0 *

γ>α≈β

γ>α>β

* Band not always observed.

close to 3630 cm−1 polarized in the α and β directions, and two bands close to 3530 and 3460 with the pleochroism γ > α = β (Fig. 3). Compared to diopside, the 3355 cm−1 band is absent, and the bands are normally somewhat broader and slightly shifted to lower wavenumbers.

Omphacite IR spectra of omphacite are similar to those of augite, with one α− and β–polarized band around 3630 cm−1 and two bands around 3530 and 3460 cm−1 with their strongest absorption in the γ direction (Fig. 4). However, omphacite spectra differ from augite spectra in that the 3460 cm−1 band in omphacite is normally much stronger than the other bands, and that the 3630 cm−1 bands is sometimes absent (Koch-Müller et al. 2004; Katayama et al. 2005).

Figure 3. Polarized IR spectra of augite from mantle diatreme xenolith, Hopi Butte, Arizona, normalized to 1 mm thickness. Modified after Skogby et al. (1990).

Orthopyroxene OH absorption bands in orthopyroxene spectra are quite different compared to those in clinopyroxene spectra, and show also substantial variation. One group of bands occurs at 3600-3610 cm−1 and is polarized in the α and β directions (e.g., Bell et al. 1995; Rauch and Keppler 2002), but is not always present in orthopyroxene spectra (Skogby et al. 1990). A second group of bands with the pleochroic behavior γ > α ≈ β occur at 3560, 3510 and 3410 cm−1 (Fig. 5). A third group occur at the relatively low wavenumbers 3300, 3210 and 3060 cm−1, with the pleochroic scheme γ > α > β (e.g., Bell et al. 1995; Peslier et al. 2002).

Figure 4. Polarized IR spectra of omphacite from an eclogite xenolith in kimberlite, Roberts Victor Mine, South Africa, normalized to 1 mm thickness. Modified after Skogby et al. (1990).

158

Skogby

The OH bands in orthopyroxene spectra are often sharper than those observed in clinopyroxene spectra, except the bands in the third group that follow the general trend with increasing band width towards lower wavenumbers. Besides normal “pyroxene” OH bands, absorption features from hydrous phases included in the pyroxene structures are sometimes encountered. Absorption bands caused by fluid and glass inclusions are considerably broader than pyroxene bands, and can normally be distinguished from them by the H2O bending (1630 cm−1) and combination (5200 cm−1) modes, and their isotropic behavior in polarized measurements. However, they may cause severe distortions of the spectral baseline that make quantification procedures more problematic. Amphibole lamellae frequently occur in pyroxenes, as documented by HRTEM studies (e.g., Veblen and Buseck 1981; Ingrin et al. 1989). Amphiboles normally have sharp and well-defined OH absorption bands, which are sometimes observed in pyroxene spectra. As they occur at somewhat higher frequencies, they can normally be easily distinguished in pyroxene spectra (Fig. 6.). Amphibole bands appear to be more frequent in spectra of crustal pyroxenes than mantle pyroxenes. Since the molar absorption coefficients of amphibole OH bands are very high (Skogby and Rossman 1991), IR spectroscopy provides a very efficient means to detect low amounts (ppm level) of submicroscopic amphibole lamellae in pyroxenes (Skogby et al. 1990; Andrut et al 2003).

Orthopyroxene

Absorbance

Absorption from inclusions

1.0

0.8

α

0.6

β

0.4

0.2

γ 0.0 3800

3600 3400

3200

3000

2800

Wavenumber (cm-1) Figure 5. Polarized IR spectra of orthopyroxene from a spinel-peridotite xenolith, Simcoe, WA, normalized to 1 mm thickness. Modified after Peslier et al. (2002).

Figure 6. IR spectra for diopside crystals in γ-polarization normalized to 1 mm thickness. Sample occurrence: a) xenolith in mantle diatreme, Hopi Butte, AZ; b) metamorphic, Outokumpu, Finland; c) metamorphic, Binntal Vallis, Switzerland; d) metamorphic, Mount Bity, Madagascar; e) metamorphic, Sinnidal, Norway. Note sharp bands around 3670 cm−1 that are due to amphibole inclusions. Modified after Skogby et al. (1990).

Also sheet silicates have been shown to occur in mantle pyroxenes and cause absorption in the OH range. In a study based on FTIR spectroscopy and HRTEM characterization, Koch-Müller et al. (2004) identified nm-sized inclusions of clinochlore, amesite and biotite in omphacite samples. They interpreted unusually strong absorption bands observed at 3600-3624 cm−1 to be caused by these hydrous sheet silicate inclusions (Fig. 7). Similar strong absorption bands were observed by Katayama and Nakashima (2003) in spectra of clinopyroxene from an eclogite.

Water in Natural Mantle Minerals I: Pyroxenes

159

Figure 7. Polarized IR spectra of omphacite and omphacitic clinopyroxene from eclogite and granulite occurrences in Yakutia, Russia. Bands around 3600-3625 cm−1 are related to hydrous nano-inclusions. Spectra are polarized along α and normalized to 1 cm thickness. Modified after Koch-Müller et al. (2004).

The relatively weak α- and β-polarized absorption bands normally observed in this region in omphacite spectra (Table 1) have been interpreted to be due to intrinsic OH groups. However, the possibility that also these bands are due to sheet silicate inclusions can at present not be ruled out, and there is hence a strong need for further studies to investigate how widespread these nanoscale inclusions are in omphacitic pyroxenes.

CORRELATIONS OF OH AND SAMPLE CHEMISTRY The incorporation of hydrogen in the pyroxene structure needs to be accompanied by other substitutions, or formation of vacancies, to maintain charge-balance. Such coupled hydrogen incorporation mechanisms are expected to lead to correlations between OH concentrations and sample chemistry. Several studies have searched for such correlations (Skogby et al. 1990; Peslier et al. 2002; Bell et al. 2004; Koch-Müller et al. 2004), taking both trace, minor and major elements into account. However, well-defined correlations have only rarely been found. As the amounts of OH are relatively high in mantle derived pyroxenes with concentrations ranging from hundreds up to a few thousands wt-ppm H2O, trace elements are generally not present in sufficient amounts to account for the hydrogen incorporation mechanisms. Minor elements (e.g., Al, Cr, Fe3+, Na) and vacancies are more likely candidates, and have also been shown to be at least weakly correlated with OH concentration or absorbance of individual bands in IR spectra. For clinopyroxenes, Skogby et al. (1990) observed a correlation between absorption coefficients of the high-wavenumber (3640 cm−1) bands in spectra of diopside and augite samples and the amount of trivalent ions (Fig. 8). Positive correlations of OH absorbance and Al contents in a suite of clinopyroxene were also found by Peslier et al. (2002), but in this case with the absorption band at 3540 cm−1 (Fig. 9). Furthermore, Andrut et al. (2003) noted a correlation of OH and Al in a zoned gem-quality diopside crystal of metamorphic origin. Aluminum appears to be involved also in hydrogen incorporation in omphacite, as shown by Koch-Müller et al. (2004) who observed a correlation between aluminum in the tetrahedral position and the absorbance of bands in the 3500-3540 cm−1 region (Fig. 10a). In addition, they noted a continuous shift of the position of the main absorption band in this region with the amount of tetrahedrally coordinated Al (Fig. 10b). Other elements that have shown to be correlated with OH concentration include Ti and K, as reported by Bell et al. (2004) in a study of megacrysts from the Monastery kimberlite. They found that H and Ti were correlated with an atomic ratio of 1:1 (Fig. 11), but concluded that this does not necessarily mean that

160

Skogby

Absorption coefficient (3640 cm-1)

6

Clinopyroxenes

Figure 8. Summed absorption coefficients for OH bands around 3640 cm−1 plotted versus Al + Cr + Fe3+, in atoms per formula unit. Data represent diopside-hedenbergite, augite and aegirine-augite samples of crustal and mantle origin. Data from Skogby et al. (1990).

4

2

0 0.0

0.1

0.3 0.2 0.4 Me3+ (apfu)

0.5

0.6

Clinopyroxenes

Figure 9. Correlation between absorption coefficients for the 3540 cm−1 band in clinopyroxene spectra and Al content. Data obtained on spinelperidotite xenolith samples from Mexico and Simcoe, WA. Data from Peslier et al. (2002).

Al (apfu)

0.30

0.20

0.10 1.0

3.0 4.0 2.0 Absorption coefficient (cm-1)

3550

18

b

a

12 9 6 3 0 0.0

Clinopyroxenes

3540

Wavenumber (cm-1)

Absorption coefficient (cm-1)

Clinopyroxenes 15

5.0

3530 3520 3510 3500

0.1 IVAl3+

0.2

(apfu)

0.3

3490 0.0

0.1 IVAl3+

0.2

0.3

(apfu)

Figure 10. Plots of (a) linear absorbance for γ-polarized OH bands at 3500-3540 cm−1 in omphacitic clinopyroxene, and (b) main wavenumber, versus tetrahedrally coordinated Al (apfu). Samples from eclogite and granulite occurrences in Yakutia, Russia. Modified after Koch-Müller et al. (2004).

Water in Natural Mantle Minerals I: Pyroxenes

161

a 1:1 substitutional mechanism can be inferred, as also other elements (e.g., K) were found to be correlated with H, but at strongly different atomic ratios.

Apart from correlations with minor elements, hydrogen concentrations have also been shown to follow trends with major elements. This is exemplified by the studies of Peslier et al. (2002), which displayed negative correlations with the major elements Si, Mg and Ca, and Bell et al. (2004), which showed that OH concentrations follow differentiation trends coupled to Ca-number (Ca/ [Ca+Mg] ratio) for clinopyroxenes from the Monastery kimberlite.

Figure 11. Correlation of H with (a) Ti, and (b) K, in clinopyroxene from mantle-derived megacrysts from the Monastery kimberlite, South Africa. Modified after Bell et al. (2004). 8

Absorbance (3640) / (3525 + 3450)

In general, the correlations of OH concentration and sample chemistry that have been observed in natural samples have been relatively weak. An explanation for the absence of stronger correlations can be that the variable degrees of OH saturation caused by equilibration in different geological environments may obscure possible correlations, when samples from different geological environments are considered. A possible way to decrease the effect of various degrees of OH saturation is to search for correlations between ratios of OH band intensities and sample chemistry. Using this approach, Skogby et al. (1990) observed a pronounced correlation between the ratio of the intensities of high- and lowwavenumber bands in clinopyroxene spectra and the ratio of the concentration of trivalent ions and Fe2+ (Fig. 12).

Clinopyroxenes 6

4

2

0

0

2

4

6

8

10

12

Me3+ / Fe2+

Figure 12. Ratio of summed absorbances for OH bands

at 3640 and 3450-3525 cm−1 plotted vs. the atomic For omphacite, and also sodic ratio (Al + Cr + Fe3+)/Fe2+. Data represent diopsideclinopyroxenes, several studies have hedenbergite, augite and aegirine-augite samples of shown that OH absorbances correlate crustal and mantle origin with Fe > 0.10 apfu. Data with the amount of cation vacancies at from Skogby et al. (1990). the M2 site. This was first observed by Smyth et al. (1991) who studied a suite of sodic clinopyroxenes from South African mantle eclogites. The presence of vacancies at the M2 site in mantle eclogite has been confirmed by crystal-chemical methods (McCormick 1986), and can be seen as a Ca-Eskola (Ca0.50.5AlSi2O6) component. The solubility of the CaEskola component in clinopyroxenes under P and T conditions representative for eclogite facies has been verified experimentally (Gasparik 1986). Smyth et al. (1991) found that the intensity of the band near 3460 cm−1 correlates with the amount of M2 cation vacancies, and that samples with high Ca-Eskola components contained the highest OH concentrations observed in pyrox-

Skogby

enes. Similar relations with pronounced correlations of bands in the 3460 cm−1 area and cation vacancies (Fig. 13) have also been found for other series of mantle eclogite clinopyroxenes (Koch-Müller et al. 2004), as well as crustal eclogite clinopyroxenes (Katayama et al. 2003, 2005). Moreover, experimental studies (Bromiley and Keppler 2004) have demonstrated that small amounts of Ca-Eskola components considerably increase the amount of OH incorporation in synthetic jadeite. Some of the studied clinopyroxenes characterized by relatively high Ca-Eskola components also contain exsolved garnet and kyanite, indicating that the precursor pyroxene phase has contained even higher amounts of M2 vacancies. By extrapolation of the observed trends between OH absorbance and vacancies, OH concentrations ranging up to several thousands ppm H2O have been estimated (Smyth et al. 1991).

30

Clinopyroxenes

Absorption coefficient (cm-1)

162

20

10

0 -0.03

-0.01

0.03 0.01 Vacancies

0.05

0.07

Figure 13. Plot of linear absorbance for γ-polarized OH bands at 3500-3540 cm−1 in omphacitic clinopyroxene as a function of vacancy concentration, calculated as 4 minus total cations per 6 oxygen atoms. Samples from eclogite and granulite occurrences in Yakutia, Russia. Data from Koch-Müller et al. (2004).

Mantle orthopyroxenes generally show less chemical variability than mantle clinopyroxenes, and correlations between OH absorbances and compositions have been observed only in a few cases. However, in the study of Monastery kimberlite samples, Bell et al. (2004) found OH concentrations to follow a trend of decreasing OH with Fe enrichment, and also Al following closely the same trend. A weak positive correlation of Al and OH in mantle orthopyroxene was also observed by Peslier et al. (2002), as well as for OH and wholerock Al contents. These positive correlations observed for OH and Al in mantle orthopyroxene is further supported by work on synthetic orthopyroxene (Rauch and Keppler 2002; Stalder 2004) which have shown that the OH solubility increases strongly with Al content, and that the bands in the higher wavenumber range (>3400 cm−1) are caused by OH associated with Al.

WATER CONCENTRATION IN MANTLE PYROXENES Ortho-and clinopyroxenes have been shown to carry the largest amounts of water among the major upper mantle minerals, and it is evident that pyroxenes play important roles both in providing a repository for water in the upper mantle and in mantle water recycling processes. A fairly large number of studies have addressed the specific amounts of OH in pyroxenes from different mantle occurrences, including kimberlites (e.g., Bell et al. 1992, 2004), peridotites of different types (e.g., Skogby et al. 1990; Peslier et al. 2002) as well as eclogites from both mantle and crustal environments (Smyth et al. 1991; Katayama and Nakashima 2003; Koch-Müller at al. 2004; Katayama et al. 2005). Most studies have relied on IR spectroscopy for quantification of water concentrations. The calculation procedures involved to translate spectral absorption parameters to absolute water concentration data require calibration by independent water analysis methods, which have continuously been improved over time. An overview of suitable hydrogen analysis methods and their application in calibrations of OH absorbances in IR spectra is given in Rossman (2006). As also the IR measurements can be performed in different ways, for instance concerning polarization, background subtraction, and use of linear or integrated intensities, published concentration data are not always directly

Water in Natural Mantle Minerals I: Pyroxenes

163

comparable. A summary of published data on contents in mantle pyroxenes are given in Table 2, including information on analytical procedures. The OH concentrations vary as a function of mineral species and composition, and in some cases also in relation to the type of mantle environment. However, strong variations are also observed for samples within each mineral group from similar environments. Calcic clinopyroxenes (diopside and augite) from peridotites vary in concentration from 140 to 740 ppm H2O, whereas orthopyroxene show lower concentrations in the range 40 to 530 ppm H2O. In general, orthopyroxenes hold about half the amount of OH being present in co-existing clinopyroxenes. The calcic clinopyroxenes do not appear to show significant correlation with type of the environment in general. However, Demouchy (2004) observed an increasing trend in water concentrations in diopside when going from spinel lherzolite via garnet/spinel lherzolite to garnet lherzolite. Similarly, Bell et al. (1992) observed the highest OH concentration for orthopyroxene in coarse-grained samples from garnet peridotites, but noted that these samples may have been affected by metasomatic reactions in the mantle. The highest levels of water concentrations among pyroxenes have been recorded for omphacite and sodic clinopyroxene (Smyth et al. 1991; Katayama and Nakashima 2003; Koch-Müller at al. 2004; Katayama et al. 2005). Reported concentrations (Table 2) range Table 2. Summary of observed OH concentrations in mantle pyroxenes. Geological occurrence

Mineral

# of samples

kimberlite and alkali basalt xenoliths, mantle eclogite

calcic cpx omphacite opx

5 1 3

mantle eclogite

omphacite

kimberlite and alkali basalt xenoliths, mantle eclogite

calcic cpx omphacite opx

basalt xenoliths

Analysis OH conc. (wt-ppm H2O) method*

Ref.

200 - 530 640 60 – 260

1

[1]

11

130 - 970

1

[2]

7 2 10

150 – 590 470 - 1080 50 – 460

2

[3]

cpx opx

3 3

388 – 492 174 – 212

2

[4]

spinel peridotite xenoliths

cpx opx

15 16

140 – 528 39 – 265

3

[5]

crustal eclogite

omphacite

6

230 - 870

4

[6][7]

kimberlite megacrysts

cpx opx

9 3

195 – 620 215 – 263

2

[8]

peridotite xenoliths

diopside opx

4 5

150 - 420 70 - 310

5

[9]

mantle and crustal eclogite

cpx

8

31 – 514 61 - 872

6 2

[10]

* Analysis based on: 1) IR, based on a linear molar absorption coefficient εΟΗ = 150 L/(mol.cm), Skogby et al. (1990); 2) IR, calibrations of Bell et al. 1995, with integral molar absorption coefficients Icpx= 38300, Iopx= 80600 L/(mol.cm2); 3) IR, calibrations of Bell et al. 1995, using “specific” integral absorption coefficients I’cpx= 7.09, I’opx= 14.84 ppm−1cm−2); 4) SIMS data, unpolarized IR data using calibration of Bell et al. 1995 (see above) indicate that the OH concentrations are twice as high as those listed here; 5) Unpolarized IR data, based on the calibration of Paterson (1982), with the general integral molar absorption coefficient defined as IH = 150.(3780 – ν); 6) IR, calibration of Libowitzky and Rossman (1997), with the general integral molar absorption coefficient defined as IH2O = 246.6.(3753 – ν). References: [1] Skogby et al. (1990); [2] Smyth et al. (1991); [3] Bell et al. (1992); [4] Ingrin and Beran (2001); [5] Peslier et al. (2002); [6] Katayama and Nakashima (2003); [7] Katayama et al. (2005); [8] Bell et al. (2004); [9] Demouchy (2004); [10] Koch-Müller et al. (2004)

164

Skogby

from 30 to 1080 ppm H20, although it should be noted that a mineral specific IR calibration for omphacite is lacking, and the reported concentrations based on IR data should be viewed with some caution. This is exemplified by recent SIMS data on omphacite from a crustal eclogite occurrence (Katayama et al. 2005) that indicated that the concentration based on IR absorbances was overestimated by a factor of two. Additional uncertainties regarding the high water concentrations observed in omphacite are the observations of nano-inclusions of hydrous phases related to high-wavenumber bands observed by Koch-Müller et al. (2004). A Figure 14. Pressure dependence of water contents in strong dependence of water concentration omphacite in eclogites from the Kokchetav massif, and crystallization depth was observed by Kazachstan. Data from Katayama et al. (2005). Katayama et al. 2005 (Fig. 14), who related the pressure-dependence to increased stability of the Ca-Eskola component. However, Koch-Müller et al. (2004) observed a reversed situation with unusually low concentrations (31 ppm) in diamond-bearing eclogite xenoliths and the highest concentrations (437-514 ppm) in lower-pressure grospydites and granulites. They interpreted the unusually low water concentrations for the samples from the high-pressure occurrence as being caused by low water activity during crystallization, or alternatively by hydrogen loss during uplift.

IMPLICATIONS FOR WATER IN THE UPPER MANTLE A fundamental question regarding OH contents in mantle pyroxenes, as well as in other mantle-derived NAMs, is whether the water concentrations recorded in xenolith samples are representative for the conditions in the upper mantle, or if the original hydrogen contents have been reset during different types of ascent processes. The substantial amounts of kinetic data for hydrogen diffusion in pyroxenes that now are available (e.g., Hercule and Ingrin 1999; Carpenter Woods et al. 2000; Stalder and Skogby 2003; Ingrin and Blanchard 2006) indicate that major resetting is indeed possible, also during relatively fast ascent processes (Ingrin and Skogby 2000). On the other hand, several lines of evidence indicate that OH concentrations representative for mantle conditions to large extents are preserved in pyroxenes and other mantle NAMs. A major argument for preservation of mantle OH concentrations comes from the observations of systematic correlations between sample chemistry and OH content. In a study of OH concentration of pyroxenes from spinel peridotites from the sub-arc mantle wedge, Peslier et al. (2002) found that the OH contents were correlated with sample composition, but also with the chemistry of associated spinels and whole-rock xenolith data. These parameters can be expected to be completely independent of xenolith transport processes, which indicate that the OH contents were neither reset to large extents. An important parameter for OH incorporation in peridotite pyroxenes appears to be the redox conditions. This was demonstrated in the study by Peslier et al. (2002) who observed a clear negative correlation between OH concentration and oxygen fugacities estimated from spinel compositions in a series of samples collected from different regions of the sub-arc

Water in Natural Mantle Minerals I: Pyroxenes

165

mantle wedge, which is known to be more oxidized than other parts of the upper mantle. The low OH concentrations observed in the more oxidized environments were interpreted to be caused by redox dehydration reactions associated with metasomatism and partial melting, which led to a loss of more than half of the initial pyroxene water contents. Estimates of the water budget in subduction zones indicated, however, that the water released from NAMs in the sub-arc mantle wedge only account for a minor proportion of ca 5% of the total water in this environment. Negative correlations between oxygen fugacity and OH in pyroxene have also been observed in experimental studies (e.g., Skogby 1994). Support for preservation of mantle OH concentrations in pyroxenes is put forward by the study of Bell et al. (2004) on different NAMs from a suite of megacrysts from the Monastery kimberlite. They observed that the OH contents in both clino- and orthopyroxenes followed trends with other elements in Fe-enrichment and Ca content, reflecting igneous differentiation. In accordance with the study of Peslier et al. (2002), they concluded that such systematic behavior would unlikely be observed if OH contents were fully reset at crustal environments at later stages of ascent processes. Furthermore, they noted that the relatively high OH concentrations recorded for the Monastery olivines (54-262 ppm H2O) according to experimental studies require equilibration pressures corresponding to mantle conditions, and are not compatible to resetting in crustal environments. Even if arguments exist for preservation of original mantle water contents in pyroxenes found in xenoliths, the available kinetic data regarding dehydrogenation reactions (cf. Ingrin and Blanchard 2006) indicate that hydrogen loss may be significant. The fastest reaction in this respect involves concomitant oxidation of Fe2+ according to the redox reaction: Fe2+ + OH− = Fe3+ + O2− + ½ H2 This reaction has been shown to be considerable faster than dehydrogenation reactions involving more rigorous structural changes (e.g., cation diffusion, resetting of defect chemistry). Significant dehydrogenation following this reaction will lead to enhanced Fe3+/ Fe2+ ratios, similar to what have been observed for some mantle-derived amphiboles (e.g., Dyar et al. 1992). The ferric iron contents of mantle pyroxenes can hence be used to estimate the maximum amounts of hydrogen that may have been lost via the redox reaction. By adopting this approach, Ingrin and Skogby (2000) found that the ferric iron levels observed in mantle pyroxenes correspond to a maximum loss of 900 ppm H2O for clinopyroxene and 570 ppm H2O for orthopyroxene. Apart from providing a host phase for mineral-bound water in the upper mantle, pyroxenes appear to play an important role for water recycling related to subduction zones. The high OH concentrations recorded in omphacitic pyroxenes from both crustal and mantle occurrences indicate that they provide an efficient means for water transport down to the deeper levels of the upper mantle, beyond the stability fields of the hydrous minerals. In a recent study, Katayama et al. (2005) showed that the water contents of omphacite from the Kokchetav massif, which has been subducted to a depth of 180 km, contain up to 870 ppm H2O and that water contents systematically increase with pressure (Fig. 14). They suggested that omphacite, together with garnet and rutile, in subducted crust may carry water down towards the transition zone at depths around 400 km, where the eclogite minerals are expected to transform to majoritic garnet and stishovite.

ACKNOWLEDGMENTS E. Libowitzky, A. Beran and H. Keppler are thanked for providing constructive reviews of this manuscript.

166

Skogby REFERENCES

Andrut M, Brandstätter F, Beran A (2003) Trace hydrogen zoning in diopside. Mineral Petrol 78:231-241 Bell DR, Ihinger PD, Rossman GR (1995) Quantitative analysis of trace OH in garnet and pyroxenes. Am Mineral 80:465-474 Bell DR, Rossman GR (1992) Water in Earth’s mantle: The role of nominally anhydrous minerals. Science 255:1391-1397 Bell DR, Rossman GR, Moore RO (2004) Abundance and partitioning of OH in a high-pressure magmatic system: Megacrysts from the Monastery kimberlite, South Africa. J Petrol 11:1539-1564 Beran A (1976) Messung des Ultrarot-Pleochroismus von Mineralen. XIV. Der Pleochroismus der OHStreckfrequenz in Diopsid. Tscherm Min Petr Mitt 23:79-85 Beran A (1981) Infrarotspektroskopische Untersuchungen über den OH-Gehalt von Pyroxenen und Cyaniten aus Gesteinen. Forts Mineral 59:16-18 Beran A, Zemann J (1986) The pleochroism of a gem-quality enstatite in the region of the OH stretching frequency, with a stereochemical interpretation. Tscherm Min Petr Mitt 35:19-25 Bolfan-Casanova N (2005) Water in the Earth’s mantle. Mineral Mag 69:229-257 Bromiley GD, Keppler H (2004) An experimental investigation of hydroxyl solubility in jadeite and Na-rich clinopyroxenes. Contrib Mineral Petrol 147:189-200 Carpenter Woods S, Mackwell S, Dyar MD (2000) Hydrogen in diopside: Diffusion profiles. Am Mineral 85: 480-487 Demouchy S (2004) Water in the Earth’s interior: Thermodynamics and kinetics of hydrogen incorporation in olivine and wadsleyite. PhD Dissertation, University of Bayreuth, Germany Dyar MD, McGuire AV, Mackwell SJ (1992) Fe3+/H+ and D/H in kaersutites – Misleading indicators of mantle source fugacities. Geology 20:565-568 Gasparik T (1986) Experimental study of subsolidus phase relations and mixing properties of clinopyroxene in the silica-saturated system CaO-MgO-Al2O3-SiO2. Am Mineral 71:686-693 Hercule S, Ingrin J (1999) Hydrogen in diopside: diffusion, extraction-incorporation, and solubility. Am Mineral 84:1577-1588 Ingrin J, Beran A (2001) Hydrogen content in spinel lherzolite xenoliths coming from different depths (Kilbourne Hole, NM). Terra abstract, X, 458 Ingrin J, Blanchard M (2006) Diffusion of hydrogen in minerals. Rev Mineral Geochem 62:291-320 Ingrin J, Latrous K, Doukhan JC, Doukhan N (1989) Water in diopside: an electron microscopy and infrared spectroscopy study. Eur J Mineral 1:327-341 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentrations levels and implications. Eur J Mineral 12:543-570 Johnson EA (2006) Water in nominally anhydrous crustal minerals: speciation, concentration, and geologic significance. Rev Mineral Geochem 62:117-154 Johnson EA, Rossman GR, Dyar MD, and Valley JW (2002) Correlation between OH concentration and oxygen isotope diffusion rate in diopsides from the Adirondack Mountains, New York. Am Mineral 87: 899-908 Katayama I, Nakashima S (2003) Hydroxyl incorporation from deep subducted crust: Evidence for H2O transport into the mantle. Am Mineral 88:229-234 Katayama I, Nakashima S, Yurimoto H (2005) Water content in natural eclogite and implication for water transport into the deep upper mantle. Lithos 86:245-259 Koch-Müller M, Matsyuk SS, Wirth R (2004) Hydroxyl in omphacites and omphacitic clinopyroxenes of upper mantle to lower crustal origin beneath the Siberian platform. Am Mineral 89:921-931 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52 Libowitzky E, Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82:11111115 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554-570 McCormick T (1986) Crystal-chemical aspects of nonstoichiometric pyroxenes. Am Mineral 71:1434-1440 Paterson MS (1982) The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bull Minéral 105:20-29 Peslier AH, Luhr JF, Post J (2002) Low water contents in pyroxenes from spinel-peridotites of the oxidized, sub-arc mantle wedge. Earth Plan Sc Lett 201:69-86 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Rossman GR (1996) Studies of OH in nominally anhydrous minerals. Phys Chem Minerals 23: 299-304 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Runciman WA, Sengupta D, Marshall M (1973) The polarized spectra of iron in silicates. I. Enstatite. Am Mineral 58:444-450

Water in Natural Mantle Minerals I: Pyroxenes

167

Skogby H (1994) OH incorporation in synthetic clinopyroxene. Am Mineral 79:240-249 Skogby H, Rossman GR (1989) OH− in pyroxene: An experimental study of incorporation mechanisms and stability. Am Mineral 74:1059-1069 Skogby H, Rossman GR (1991) The intensity of amphibole OH bands in the infrared absorption spectrum. Phys Chem Min 18:64-68 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: Variations in the natural environment. Am Mineral 75:764-774 Stalder R (2004) Influence of Fe, Cr and Al on hydrogen incorporation in orthopyroxene. Eur J Mineral 16: 703-711 Stalder R, Skogby H (2003) Hydrogen diffusion in natural and synthetic orthopyroxene. Phys Chem Minerals 30:12-19 Smyth JR, Bell DR, Rossman GR (1991) Incorporation of hydroxyl in upper-mantle clinopyroxenes. Nature 351:732-735 Veblen DR, Buseck PR (1981) Hydrous pyriboles and sheet silicates in pyroxenes and uralites: Intergrown microstructures and reaction mechanisms. Am Mineral 66:1107-1134 Wilkins RW, Sabine W (1973) Water content of some nominally anhydrous silicates. Am Mineral 58:508-516

8

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 169-191, 2006 Copyright © Mineralogical Society of America

Water in Natural Mantle Minerals II: Olivine, Garnet and Accessory Minerals Anton Beran and Eugen Libowitzky Institut für Mineralogie und Kristallographie Universität Wien - Geozentrum Althanstraße 14, A-1090 Wien, Austria [email protected]

[email protected]

INTRODUCTION Hydrogen traces change the physical properties of mantle minerals to an extent that is far out of proportion to its low concentration. These properties include mechanical strength, melting behavior, diffusion rate, electrical conductivity, viscosity and rheology. Besides minerals of the pyroxene group (as discussed by Skogby 2006, this volume), the close-packed mineral structures of olivine, garnet and some accessory minerals offer important storage sites for hydrogen traces in the Earth’s mantle. However, the water content stored in olivine and mantle garnet is quite low compared to that in pyroxenes. Based on chemical considerations, but also on information obtained from infrared (IR) data, Martin and Donnay (1972) proposed the existence of hydroxyl in nominally anhydrous minerals (NAMS) occurring in the upper mantle, especially in pyroxene and olivine. The review articles of Bell and Rossman (1992a), Skogby (1999), and Ingrin and Skogby (2000) reveal a wide range of water contents for mantle-derived pyroxenes, olivines and garnets, which are derived from IR spectroscopic data. Fourier transform infrared (FTIR) spectroscopy provides an extremely sensitive method for detecting trace hydrogen bonded to oxygen in the structures of various NAMS (Beran 1999; Beran and Libowitzky 2003; Libowitzky and Beran 2004). As this method is not self-calibrating, attempts have been made to calibrate the IR spectra with independent absolute methods. These methods include hydrogen manometry and measurement of thermally released water in an electrolytic cell or by Karl Fischer titration. 1H Magic-Angle-Spinning Nuclear Magnetic Resonance (MAS NMR), Secondary Ion Mass Spectrometry (SIMS) and Nuclear Reaction Analysis (NRA) are encouraging but experimentally demanding and expensive methods (as discussed by Rossman 2006, this volume). To overcome these problems, approximations such as that proposed by Paterson (1982) and refined by Libowitzky and Rossman (1997), attempt to provide a way to deal with the general dependence of the molar absorption coefficient on OH band positions in a more accurate way. The basis of the quantitative determination of the water content is the Beer-Lambert’s law. IR absorbances A (A = log I0/I) are directly related by the molar absorption coefficient ε to the concentration c of OH groups and to the thickness t of the sample: A = ε · c · t. In optically anisotropic crystals normally the sum of absorbances measured with polarized radiation in principal axis directions of the optical indicatrix in oriented crystal sections results in accurate values which can be used for the determination of the water content (Libowitzky and Rossman 1996). In optically isotropic (cubic) crystals, e.g., garnets, the absorbance values from (un)polarized spectra must be multiplied by 3 to account for all three spatial directions (compare with Paterson’s 1982 orientation factor γ = 1/3). Concentration values may be best obtained from integral absorbances Ai (in cm−1). The integral absorption coefficient αi = Ai/t (in cm−2; t measured in cm) is then expressed by αi = εi·c, where εi is the integral molar absorption coefficient in L·mol−1·cm−2. When εi is determined for a specific structural matrix, e.g., 1529-6466/06/0062-0008$05.00

DOI: 10.2138/rmg.2006.62.8

170

Beran & Libowitzky

for olivine or garnet of specific composition, then for all other olivines or garnets of the same composition and same type of OH defect the water content can be quantitatively determined by the relation c (in wt% H2O) = (αi · 1.8)/(εi · D), where D is the density of the mineral in g·cm−3 (Beran et al. 1993; Libowitzky and Beran 2004). Asimow et al. (2006) reported a method to derive accurate water contents from polarized measurements of randomly oriented grains.

OLIVINE Basic structure and possible sites of hydrogen incorporation The structure of olivine is best described as an approximately hexagonal close-packing of oxygen atoms with one half of the distorted octahedral interstices occupied by (Mg,Fe) atoms and one eighth of the tetrahedral interstices occupied by Si. One formula unit (Mg,Fe)2SiO4 contains two crystallographically different (Mg,Fe) sites, i.e., M1 on a center of symmetry, M2 in a mirror plane. Si is also placed in that mirror plane. Two of the three different oxygen positions, O1 and O2, are localized in a mirror plane, while the third, O3, occupies a general position. All oxygen atoms are coordinated by three (Mg,Fe) and one Si atom in a distorted tetrahedron. Thus, any possible structural OH defect is part of a full/vacant coordination tetrahedron and octahedron. A first model for OH positions in olivine based on polarized IR spectroscopic measurements was proposed by Beran and Putnis (1983) for gem-quality crystals of hydrothermal origin from Zabargad, Egypt. This olivine is characterized by pleochroic absorption bands at ~3590, 3570, 3520, and 3230 cm−1. It was suggested that [O(OH)3] and [O2(OH)2] tetrahedra with a specific combination of hydrogen positions occur as structural elements, assuming that vacancies are on Si sites. If M2 site vacancies were assumed, [SiO3(OH)] and [SiO2(OH)2] tetrahedra occur as structural elements. Libowitzky and Beran (1995) presented a polarized IR study of a colorless near-endmember forsterite, revealing OH stretching bands (Fig. 1) predominantly in the high-energy wavenumber

1.4

group I

group II

Absorbance

1.2 Pamir, 1.1 mm

1.0 0.8 0.6 0.4

a c

0.2

b

0.0

3700

3600

3500

3400

3300

3200

-1

Wavenumber (cm ) Figure 1. Polarized IR absorption spectra of an oriented forsterite crystal from a skarn deposit in Pamir, Tadzikistan, representing a crustal occurrence (Libowitzky and Beran 1995). The sharp, strongly pleochroic OH absorption bands are restricted to the high-energy region of band group I, comprising the range 36503450 cm−1. Band group II covers the 3450-3200 cm−1 region (according to Bai and Kohlstedt 1993).

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals region (band group I according to Bai and Kohlstedt 1993). Sharp, strongly pleochroic band doublets centered at 3674/3624, 3647/3598, 3640/3592 cm−1 are assigned to OH dipoles oriented parallel to [100]. An OH band doublet at 3570/3535 cm−1 shows both, a strong absorption parallel to [100] and a strong component parallel to [001]. Under the assumption of vacancies at Si and (Mg,Fe) sites, the O1 site represents the most favorable position for OH defects pointing to a vacant Si site. O3 is proposed as donor oxygen of OH dipoles lying near the O3-O1 tetrahedral edge or roughly pointing to a vacant M2 site. In this model also O2 can act as donor oxygen of an OH group oriented along the O2-O3 edge of a vacant M1 octahedron (Fig. 2).

171

[M2] O3

M1

O3 [Si] O2

M2

H O1

c M1

O3

a Figure 2. Schematic diagram of a part of the olivine structure with a Si vacancy showing possible OH orientations derived from the pleochroic behavior of the Pamir forsterite spectra presented in Figure 1 (modified after Libowitzky and Beran 1995).

In a recent polarized FTIR spectroscopic study of synthetic pure forsterite by Lemaire et al. (2004) the proposed OH incorporation model assuming vacant Si, M1, and M2 sites is essentially confirmed. OH bands at 3613, 3580, 3566, 3555, and 3480 cm−1 are assigned to OH groups compensating Si vacancies. Bands at 3600, 3220, and 3160 cm−1 are enhanced in samples with higher silica activity, suggesting that these bands are related to M1 (3160 cm−1) and M2 vacancies (3600 and 3220 cm−1). From the crystal chemical approach it is interesting to note that the predominant alignment of OH groups parallel to [100] was also observed by IR spectroscopic studies of Bauerhansl and Beran (1997) in the olivine-type mineral chrysoberyl, Al2BeO4. An intense discussion of the cation vacancy type, i.e., Mg vs. Si, related to hydrogen solubility was initiated on the basis of experimental results by Bai and Kohlstedt (1992, 1993). The authors carried out annealing experiments on olivine crystals from San Carlos, Arizona, with samples left unbuffered, samples buffered with orthopyroxene, and samples buffered with magnesiowustite. IR spectra from the annealed samples revealed two distinct groups of OH bands, group I bands occurred in the 3650-3450 cm−1 region, group II bands in the 34503200 cm−1 range. The origin of these band groups—Si vacancy and/or M1, M2 vacancies related—and their relation to P, T, silica activity, and oxygen fugacity, is still a matter of debate (Kohlstedt et al. 1996; Matveev et al. 2001, 2005; Lemaire et al. 2004; Zhao et al. 2004; Berry et al. 2005; Mosenfelder et al. 2006, and contributions in this volume, e.g., Keppler and BolfanCasanova 2006).

Defect types in mantle-related olivines from different localities Based on TEM observations and IR spectroscopic investigations, Kitamura et al. (1987) described planar OH bearing defects in olivine from a kimberlite in Buell Park, Arizona. The structure of the defects resembles that of an OH bearing monolayer within the olivine structure as it exists in humite group minerals. Bands present at 3571, 3524, 3402, and possibly 3319 cm−1 can be assigned to titanian clinohumite. Polarized IR spectra of an olivine crystal from this locality, which contains about 50 wt ppm H2O, have been reported by Mosenfelder et al. (2006). The strongest bands are at 3613, 3598, 3579, and 3567 cm−1. The spectra of this olivine sample differ from that reported by Kitamura et al. (1987) by lack of bands due to planar titanian clinohumite defects. Two modes of hydrogen incorporation in mantle olivine from Yakutia, Siberia, were suggested by Khisina et al. (2001): Intrinsic hydrogen in form of ordered OH bearing point defects in “hydrous olivine” and extrinsic hydrogen contained in exsolutions of hydrous minerals in form of “large” inclusions. Bands observed in the olivine

172

Beran & Libowitzky

spectra at 3704 and 3683 cm−1 can be referred to serpentine and a band at 3677 cm−1 to talc. The bands observed at 3591 and 3660 cm−1 match those of the “10 Å-phase.” Lamellar and hexagon-like inclusions of several ten nm in size of Mg-vacant hydrous olivine have been described in a combined FTIR/TEM study by Khisina and Wirth (2002). Typical OH band positions of mantle-derived olivines, including the band assignment for possible planar OH bearing defects are summarized in Table 1. Considering especially the Ti-clinohumite defects, synthetic olivines, crystallized experimentally under upper mantle conditions, that reproduce the common and intense OH bands at 3572 and 3525 cm−1 were reported by Berry et al. (2005). According to these authors the bands arise from OH point defects associated with traces of Ti. It should also be noted that bands at 3355 and 3325 cm−1 are assigned to Fe3+ related OH defects. A polarized IR study of naturally occurring olivines from 17 different localities by Miller Table 1. OH band positions (in cm−1) and band assignments for possible planar OH-bearing defects for mantle olivines from different localities and occurrences. Locality 1

2

3675 3637

3610 3598 3571

3524

3623 3615 3602 3592 3576 3565 3542 3527

3 3683 3676 3660 3637 3630 3624 3613 3597 3591 3572 3540 3526

3499 3481 3455

3500 3482 3458

3413 3400 3375

3414 3401 3374 3355 3330

4 3704 3688 3677 3660 3640 3623 3599 3591 3572 3562 3540 3526

5

3639 3630 3624 3612 3597 3591 3572 3567 3541 3525 3512

6

7

Band assignment serpentine serpentine talc 10 Å-phase serpentine

3599 3572 3562 3545 3525

3481 3458

3573

10 Å-phase Ti-clinohumite

3525

Ti-clinohumite

3486 3451

3402

Ti-clinohumite 3369 3331

3354 3331

3319 3230

Ti-clinohumite 3298 3225

Wavenumber values with deviations of ± 2.5 cm−1 are listed within one line. 1 – kimberlite xenolith, Buell Park, Arizona (Kitamura et al. 1987) 2 – kimberlite xenolith, Monastery, South Africa (Miller et al. 1987) 3 – kimberlite xenolith, Obnazennaya, Yakutia (Kishina et al. 2001) 4 – kimberlite xenolith, Udachnaya, Yakutia (Kishina et al. 2001) 5 – kimberlite xenolith, Udachnaya, Yakutia (Koch-Müller et al. 2006) 6 – spinel peridotite, Ichinomegata, Japan (Kurosawa et al. 1997) 7 – garnet peridotite, Wesselton, South Africa (Kurosawa et al. 1997)

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

173

et al. (1987) clearly demonstrated that olivines from kimberlite occurrences contain the highest hydrogen contents at a concentration level of 37-138 wt ppm H2O (if a factor of 2.3 is applied to adjust Paterson’s 1982 approximation to the recent calibration of Bell et al. 2003 - see the following chapter). The IR spectra are essentially characterized by bands in the 3600-3500 cm−1 high-energy and 3400-3300 cm−1 low-energy region. The authors also noted that over 30 distinct OH absorption bands have been identified in olivine from Monastery kimberlite, South Africa, and that the majority of these bands are inclined towards [100]. Representative polarized IR absorption spectra of olivines from South African occurrences, including olivine from Monastery, are shown in Figure 3. The presence of serpentine and talc has been determined by their characteristic OH absorption bands at 3685 and 3678 cm−1, respectively (see above). Prominent OH bands at 3572 and 3525 cm−1 are attributed to humite group minerals (see above). Relatively uniform spectra of olivines from the Monastery kimberlite have been

4.0 Monastery, 10 mm

3.5

3.0

Absorbance

a

2.5

c b

2.0 Kimberley, 1 mm

1.5 a

1.0

c b

Kaalvallei, 1 mm

0.5 a c b

0.0 3800

3700

3600

3500

3400

3300

3200

3100

-1

Wavenumber (cm ) Figure 3. Representative polarized IR absorption spectra in the OH stretching frequency region of olivines from kimberlitic xenoliths of South African occurrences, indicating a preferred orientation of the OH defects parallel to a (modified after Miller et al. 1987 and Bell et al. 2003).

174

Beran & Libowitzky

reported by Bell et al. (2004a). The olivine megacrysts represent the crystallization product of a kimberlite-like magma at pressures of about 5 GPa and temperatures of 1400-1100 °C. Spectra of olivines from the magnesian “main silicate trend” group (~13 wt% FeO) differ slightly from spectra of high-Fe olivines (17-19 wt% FeO). Both groups show a main band centered at 3572 cm−1 with strong polarization parallel to [100]. The high-Fe olivines are characterized by an enhanced intensity of bands at the high-energy side of the main band, and by a reduced intensity of the 3526 cm−1 band, relative to the main band. With respect to the OH defect assignment it is important to note that the Ti content of high-Fe olivines is significantly lower than that of low-Fe olivines. The olivines display H2O contents in the range 45-262 wt ppm. Olivine and clinopyroxene water contents appear to increase with differentiation of the host magma, consistent with an enrichment of water in the residual melt during fractional crystallization. Inter-mineral distribution coefficients for OH between olivine and clinopyroxene are thus constant. However, the presence of strong, titanium related OH defect bands is evident. Matsyuk and Langer (2004) published a comprehensive IR study of Yakutian upper mantle material and proposed a new nomenclature for the hydrous component in olivines. Selected IR absorption spectra, also illustrating the presence of group II bands, are shown in Figure 4. Hydroxyl groups in the form of non-intrinsic separate inclusions (NSI) were discerned from isolated local defects (ILD) or condensed extended defects (CED) intrinsic to the olivine

25

E // a

Linear absorption coefficient (cm-1)

20

15 Mir

10

Udachnaya

5 Bazovaya-3

Udachnaya 0 3700

3600

3500

3400

3300

3200

-1

Wavenumber (cm ) Figure 4. Selected IR absorption spectra, polarized parallel to a, of olivines from kimberlitic xenoliths of occurrences from the Siberian Platform, Yakutia, showing OH bands in both group I and group II regions (modified after Matsyuk and Langer 2004).

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals 10

Udachnaya

a b c

5

-1

Linear absorption coefficient (cm )

structure. As the two latter types cannot be simply distinguished by IR spectroscopy and as they are presumably interconnected by condensation reactions, it was proposed to symbolize the intrinsic defects as ILD/CED. NSI frequently comprise serpentine and talc with OH bands in the 3704-3657 cm−1 range (see above); Mg-edenite and Mg-pargasite occur rarely, showing bands at 3711-3709 cm−1. OH stretching bands strongly polarized along [100] are a significant feature of ILD/CED. Bands in the 3570-3510 cm−1 region are intensity-correlated and are assigned to Sidepleted “Ti-clinohumite-like” defects (for OHclinohumite and OH-chondrodite see Liu et al. 2003). Bands in the 3500-3300 cm−1 low-energy region and in the 3640-3580 cm−1 high-energy region are suggested to originate from OH in different types of (Mg,Fe)-depleted defects. The complex nature of the strongly polarized OH bands in the group I region is demonstrated in Figure 5. The study of Matsyuk and Langer (2004) is based on a total of 335 olivine crystal grains extracted from 174 different specimen of Yakutian upper mantle material representing all rock types occurring in kimberlites of the Siberian platform. Though there are indications that the occurrence of individual defect types is related to the genetic peculiarities of their host rocks, straight-forward and simple correlations do not exist. It is important to note that, according to Matsyuk and Langer (2004), olivine included in diamond does not contain water, detectable by IR spectroscopy, neither as NSI nor as ILD/CED. Among the rock-forming olivines, those of the ilmenite bearing specimens are highest in total water content, except for olivines from peridotites with primary phlogopite. Values of the absolute water content range from 4 to about 350 wt ppm H2O with an average around 140 wt ppm.

175

0 Slyudyanka

10

5

0 Udachnaya

10

5

0

3700

3600

3500

Wavenumber (cm-1) Figure 5. Polarized IR absorption spectra of olivines from kimberlitic xenoliths of Yakutian occurrences, showing the complex nature of the OH bands around 3570 cm−1. The preferred orientation of the OH defects parallel to a is clearly indicated (modified after Matsyuk and Langer 2004).

Kurosawa et al. (1997) determined the concentration of hydrogen and other trace elements in olivines from mantle xenoliths by a combined SIMS and IR spectroscopic study. The H2O contents are in the range of 13 to 60 wt ppm. In contrast to the observations made by Matsyuk and Langer (2004), the SIMS analyses of olivine included in diamond yielded similar hydrogen concentrations. Therefore, the authors concluded that the hydrogen content of xenolithic olivines does not equilibrate with water in the host magma during transport from the mantle to the surface. Comparing olivines from spinel peridotites with those from garnet peridotites, the presence of additional absorption bands in the band group II region (3354, 3331 cm−1) is a significant feature of spinel peridotitic olivines. In garnet peridotitic olivines a positive correlation of hydrogen with the trivalent cation content (Al+Cr) was observed by

176

Beran & Libowitzky

Kurosawa et al. (1997), thus indicating the incorporation of hydrogen into mantle olivines by a coupled substitution mechanism. Light elements for a suite of ten mantle-derived olivine crystals have been measured by EMP, SIMS and FTIR spectroscopy (Kent and Rossman 2002). Li, B and H2O concentrations are in the range of 0.9-7.8, 0.01-67, and 0.8-61 wt ppm, respectively. Although Li, B and H2O contents vary substantially, their cation proportions are not strongly correlated, arguing against coupled substitutions. More than 20 strongly polarized OH bands in the 3730-3330 cm−1 range have been reported by Koch-Müller et al. (2006) for olivine crystals from the Udachnaya kimberlite. Bands in the 3730-3670 cm−1 region were assigned to inclusions of serpentine, talc and “10 Å-phase.” All other bands were believed to be intrinsic to the olivine structure and it was proposed that the corresponding OH point defects are associated with vacant Si and vacant M1 sites, i.e., O1-H defects, aligned strongly parallel to [100]. The absolute water contents range from 49 to 392 wt ppm.

Calibration approaches and summary of hydrogen contents Though there exist older approaches to calibrate the hydrogen content in olivine, e.g., by the general diagram of Paterson (1982), only the two most recent calibration studies that end up with comparable values are reported here. The hydrogen contents of three natural olivines determined by 15N nuclear reaction analysis (NRA), i.e., 140, 220, and 16 wt ppm H2O, were used by Bell et al. (2003) to calibrate the IR spectroscopic data for the quantitative hydrogen analysis of olivines. The OH defect concentration expressed as wt ppm H2O is 0.188 times the total integral absorbance of the fundamental OH stretching bands, rigorously applicable to samples dominated by OH absorptions in the high-wavenumber range 3650-3450 cm−1. This equals a value of the integral molar absorption coefficient εi = 28450 ± 1830 L·mol−1·cm−2. A comparison of the OH concentrations determined with the Bell et al. (2003) calibration and those derived from the general Paterson (1982) trend indicates that by using polarized radiation, the Bell et al. (2003) calibration yields OH concentrations that are higher by a factor of 2.3. Based on SIMS analyses of four olivine samples from the Udachnaya kimberlite, KochMüller et al. (2006) calculated the integral molar absorption coefficient to 37500 ± 5000 L·mol−1·cm−2. This value is by a factor of about 1.3 slightly higher than that determined by Bell et al. (2003). Table 2 summarizes the OH concentration values of mantle olivines from different studies and geological occurrences, based on the calibration of Bell et al. (2003). In general, the scatter of data is relatively limited and ranges from a few wt ppm to maximum values of about 400 wt ppm H2O. The mean values are roughly in the region between 100-200 wt ppm H2O. Considering the wide variation of localities the rather homogeneous hydrogen contents of olivines are amazing. In comparison to water contents of further important mantle minerals the sequence pyroxene > olivine > garnet can be observed. On the other hand, the water contents of these natural mantle minerals are comparatively low in the light of high-P/T phases, such as wadsleyite and ringwoodite.

GARNET Structural and spectral features The structure of garnet group minerals is built up by alternating corner-sharing Me3+O6 octahedra and SiO4 tetrahedra, forming chains parallel to the three axes of the cubic unit cell. The resulting framework contains pseudo-cubic cavities which incorporate the larger Me2+ cations

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

177

Table 2. Observed OH defect concentrations in mantle olivines. Mean values in parentheses. Geological occurrence; Locality

No. of samples

OH concentration (in wt ppm H2O)

Kimberlite, xenoliths; South Africa

3

37 - 138 (71)

Kimberlite, xenoliths; South Africa

2

140 - 220 (180)

Bell et al. (2003)

Kimberlite, megacrysts; Monastery, South Africa

29

45 - 262 (159)

Bell et al. (2004a)

Kimberlite, xenoliths; Siberian platform

36

4 - 350 (140)

Matsyuk and Langer (2004)

Kimberlite, xenoliths; Udachnaya, Yakutia

9

49 - 392 (239)

Koch-Müller et al. (2006)

Kimberlite; Buell Park, Arizona

1

~ 50

Mosenfelder et al. (2006)

References Miller et al. (1987)

Concentration values are derived from IR spectroscopic data, calibrated against H2O values obtained from 15N Nuclear Reaction Analysis (Bell et al. 2003).

in eight-fold dodecahedral coordination. The SiO4 tetrahedra are distorted by an amount that depends on the size of the Me2+ cation in the distorted pseudo-cubes, which share two opposite edges with two tetrahedra. However, the relatively rigid SiO4 tetrahedra can accommodate to varying Me2+ cation sizes by a rotation which increases the size of the Me2+ sites and therefore the shared Me3+O6 octahedral edges as well. The oxygen atoms, representing possible docking sites of hydrogen, occupy only one general crystallographic site and are coordinated by one Si, one Me3+ and two Me2+ cations in the form of an almost ideal tetrahedron. One of the well-established OH defect types is the hydrogarnet substitution (see also Libowitzky and Beran 2006, this volume), where (SiO4) is (partially) replaced by (OH)4. This substitution mode is generally observed in OH rich samples of the grossular (more often with an andradite component)-hydrogrossular series. As originally observed by Cohen-Addad et al. (1967) and by Kobayashi and Shoji (1983) the IR spectroscopic characteristics of hydrogrossulars with more than five wt% H2O are two broad overlapping absorption bands centered around 3600 and 3660 cm−1 (Rossman and Aines 1991). However, these spectroscopic characteristics were generally not observed in grossular containing less than 0.3 wt% H2O. The great variability of the IR spectra of grossular with low H2O content suggests that the hydrogarnet substitution is not the only means of incorporating OH groups. In addition to tetrahedral sites, OH defects apparently exist in multiple other environments. Because of the optically isotropic behavior of garnets and their widely varying OH stretching frequencies, usually in the 3700-3500 cm−1 region, possible sites of hydrogen incorporation can only scarcely be assigned. The anisotropic OH stretching vibrational behavior of non-cubic natural garnet crystals with compositions close to the uvarovite-grossular binary was investigated by Andrut et al. (2002). The IR absorption behavior complies with orthorhombic, monoclinic, and triclinic crystal symmetry, respectively. According to the individual pleochroic behavior of ten nonisotropic bands, six different pleochroic patterns, i.e., four band doublets and two single bands, are distinguished. For band doublets at 3559/3540, 3572/3565, and 3595/3588 cm−1 as well as for a single band at 3618 cm−1, models for a structural OH incorporation based on the classical (OH)4 hydrogarnet substitution are proposed. In contrast, for the band doublet at 3652/3602 cm−1 and the single band at 3640 cm−1, OH defect incorporation is explained by assuming the

178

Beran & Libowitzky

presence of vacancies on octahedral and dodecahedral cation positions, leading to [SiO3(OH)] tetrahedral groups. It is concluded that in garnets containing only H traces the [SiO3(OH)] substitution mode plays an essential role as OH defect incorporation mechanism. Anisotropy of OH bands for a birefringent grossular from Asbestos, Quebec, has also been reported by Rossman and Aines (1986). Based on the presence of absorption bands around 3685, 3570, and 3530 cm−1 of hydrothermally grown Ti-bearing pyropes, an [(OH)3O] substitution was proposed by Khomenko et al. (1994) to compensate for the higher valence of Ti4+ at the Al3+ site. Distance-least-squares calculations were used by Lager et al. (1989) to simulate the effect of the hydrogarnet substitution on the grossular structure. Those garnets in which the shared octahedral edge is longer than the unshared can incorporate more OH. The application of these observations to other garnet compositions suggests that mantle garnets, rich in pyrope component, may contain only very limited amounts of water. Due to the presence of a relatively sharp absorption band near 3600 cm−1 in synthetic pyrope, Ackermann et al. (1983) proposed the hydrogarnet substitution as a possible location for water in the mantle. According to Geiger et al. (1991) OH defects in synthetic pyropes, grown from oxides are also incorporated into the structure as a hydrogarnet component, showing a characteristic absorption band at 3629 cm−1. The IR spectra of pyropes, grown from a gel starting material display several absorption bands, indicating that OH substitution is not governed solely by the hydrogarnet substitution. The splitting of the 3629 cm−1 band at 79 K into two narrow bands centered around 3636 and 3618 cm−1 has been confirmed by Geiger et al. (2000) for pyrope single-crystals doped with transition elements (Co, Cr, Ni, Ti, V). The spectrum of the Ti-bearing pyrope essentially corresponds to that observed by Khomenko et al. (1994). Geiger et al. (2000) suggested that due to the presence of four OH stretching bands in Ti-bearing pyrope (3686, 3630, 3568, and 3527 cm−1), additional mechanisms of OH substitution occur. In the IR spectrum of a V4+-bearing pyrope the same number of bands is observed, suggesting that higher charged cations cause additional OH substitutions and increased OH concentrations in garnet. Withers et al. (1998) found that under identical conditions of high pressure and temperature the OH content of pyrope is similar to that of grossular; at P = 3 GPa and T = 1000 °C the H2O values amount to 0.04 wt% for pyrope and to 0.02 wt% for grossular. Both garnets, pyrope and grossular, are characterized by a single band centered at 3630 and 3622 cm−1, respectively, being much sharper in grossular than in pyrope. The IR spectra of some natural pyropes appear to be different from those of synthetic samples. Observed OH band positions of garnets from mantle occurrences are summarized in Table 3. The OH spectrum of a nearly endmember natural pyrope from high-grade blueschists of Dora Maira, Western Alps, was originally described by Rossman et al. (1989). The spectrum does not resemble that of any other natural pyrope. As shown in Figure 6, the spectrum consists of four narrow bands at 3661, 3651, 3641, and 3602 cm−1, forming a triplet and a single band system. If the calibration of Bell et al. (1995) is applied (see below), the estimated H2O content is about 58 wt ppm. From high-temperature and high-pressure IR spectra, Lu and Keppler (1997) concluded that the absorption features arise from almost free OH groups in sites with different compressibility and thermal expansivity. The intensity of the high-energy triplet increases with increasing pressure (up to 10 GPa), while the intensity of the single band at 3602 cm−1 decreases significantly. However, Dora Maira pyrope may not be fully representative for mantle garnets. Representative IR absorption spectra in the OH stretching frequency region of garnets from kimberlitic xenoliths are presented in Figure 7. The IR spectra of garnets rich in pyrope component from mantle-derived xenoliths of the Colorado Plateau show a broad absorption band centered around 3570 cm−1 with an additional weak but broad absorption around 3650 cm−1 (Aines and Rossman 1984a,b). The presence of broad OH bands centered around 3570 and 3670 cm−1 for megacrysts of pyrope from ultramafic diatremes of the Colorado Plateau containing 22-112 wt ppm H2O was reported by Wang et al. (1996). The authors stated that pyrope crystals from the mantle may dehydrogenate during ascent and that caution should be exercised in using the OH content of natural pyrope to infer conditions of the source region.

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

179

Table 3. OH band positions (in cm−1) of mantle garnets from different localities and occurrences. Locality 1

2

3

3661 3651 3641

3650

4

5

6

Remarks

7

3650

usually weak 3630

3630

3602 3590 3570

3570 3512

3570 3512

usually strong Ti-related band

Wavenumber values with deviations of ± 4 cm−1 are listed within one line. 1 – blueschist, Dora Maira, Italy (Lu and Keppler 1997) 2 – kimberlite, megacryst, Kaalvallei, South Africa (Bell and Rossman 1992b) 3 – kimberlite, megacryst, Lace, South Africa (Bell and Rossman 1992b) 4 – kimberlite, megacryst, Monastery, South Africa (Bell et al. 2004a) 5 – kimberlite, xenolith, Udachnaya, Yakutia (Snyder et al. 1995) 6 – eclogite, Rietfontein, South Africa (Bell and Rossman 1992b) 7 – grospydite, Zagadochnaya, Yakutia (Beran et al. 1993)

Linear absorption coefficient (cm-1)

1273 K 1073 K

8

873 K 6

673 K

4

473 K

2

293 K

3700

3650

3600

3550

3500

-1

Wavenumber (cm ) Figure 6. Room- and high-temperature IR spectra of the OH stretching vibrational region of a pyrope single-crystal from high-grade blueschists of Dora Maira, Italy (modified after Lu and Keppler 1997).

Two broad bands centered in the same wavenumber region were reported by Bell and Rossman (1992b) for pyrope-rich garnet samples from different localities of the subcontinental mantle of southern Africa (Fig. 7). A usually weak absorption centered around 3510 cm−1 is described as a typical feature of Ti-bearing mantle garnets (Table 3). A significant absorption at 3570 cm−1 for pyrope-rich garnet from the Liaoning-50 kimberlite, NE China was

180

Beran & Libowitzky 6

6

Megacryst garnet 5

5

Louwrensia

-1

Linear absorption coefficient (cm )

Peridotite garnet

4

Monastery

3

Lace

Kimberley

4

3

Eclogite garnet

Monastery 2

2 Monastery

Rietfontein 1

1

Roberts Victor

Kaalvallei

4000

Rietfontein 0

0

3800

3600

3400

3200 -1

Wavenumber (cm )

3000 4000

3800

3600

3400

3200

3000

-1

Wavenumber (cm )

Figure 7. Representative IR absorption spectra in the OH stretching frequency region of garnets from kimberlitic xenoliths of occurrences in southern Africa. The broad absorption rising towards higher wavenumbers is due to an electronic transition of Fe2+ (modified after Bell and Rossman 1992b).

reported by Langer et al. (1993). Beran et al. (1993) reported a single broad absorption centered around 3630 cm−1 for a garnet of an eclogitic mantle xenolith from the Zagadochnaya kimberlite, Yakutia. Absorptions near 3590 and 3650 cm−1 are dominant in the spectra of pyrope-rich garnets from the Udachnaya kimberlite, Yakutia (Snyder et al. 1995; Table 3). A suite of 200 garnet single-crystals extracted from 150 mantle xenoliths from kimberlites of the Siberian platform was studied in the OH vibrational range by Matsyuk et al. (1998). Representative OH spectra show either one or a combination of two major bands in the 3660-3645, 3585-3560, and 3525-3515 cm−1 range. Bands in the latter region occur only in Ti-rich (>0.4 wt% TiO2) garnets. Bands at 3570 and 3512 have also been reported for pyrope garnets from the Monastery kimberlite, South Africa, by Bell et al. (2004a) (Figure 7). The presence of amphibole exsolution lamellae discovered by TEM in garnets from peridotites from northern Tibet (Song et al. 2005) confirms the role of garnet as an important reservoir of water in the mantle.

Calibration and hydrogen content Several attempts have been made to calibrate the amount of the hydrous component of garnets derived from IR spectroscopic investigations. Water contents were calculated by Aines and Rossman (1984b) by calibrating the integrated IR absorbances against the water content measured on the basis of P2O5 cell coulometry. 15N Nuclear Reaction Analysis (NRA) has first been performed by Rossman et al. (1988) for the determination of the hydrogen content of a series of almandine, pyrope, and spessartine garnets. It was stated that pyropes contain so little hydrogen (<0.02 wt% H2O) that satisfactory calibrations have not yet been obtained. NRA for

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

181

pyrope-almandine garnets reported by Rossman (1990) display a general positive correlation with IR peak intensity, but with considerable scatter. Vacuum extraction and H2 manometry has been used by Bell et al. (1995) to measure the hydrogen contents of spectroscopically characterized garnet samples of high purity and to calibrate the IR spectra for quantitative analysis. The H2O content in the pure separates of pyrope from the Monastery kimberlite has been determined to 56 ± 6 wt ppm, resulting in an integral molar absorption coefficient of 6700 ± 670 L·mol−1·cm−2 (the linear molar absorption coefficient amounts to 96.9 ± 9.6 L·mol−1·cm−1). A suite of 11 gem-quality garnet crystals with a broad variety of compositions were analyzed for trace amounts of hydrogen by Maldener et al. (2003) using 15N NRA and FTIR microspectroscopic methods. The integral molar absorption coefficients for three pyroperich garnet samples with H2O contents of 19, 17, and 18 wt ppm have been determined to 5000, 1960, and 3440 L·mol−1·cm−2, respectively, which would result in an average value of about 3470 L·mol−1·cm−2. A value of 3630 ± 1580 L·mol−1· cm−2 is proposed for the use in routine water determinations of compositionally different garnets, except for garnets near to endmember grossular. This value proposed by Maldener et al. (2003) is by a factor of 1.85 lower than that determined by Bell et al. (1995). Observed OH concentrations, calibrated against the H2O values obtained from hydrogen manometry by Bell et al. (1995) are presented in Table 4. Pyrope-rich garnets from Colorado Plateau diatremes (Green Knobs, Garnet Ridge) reported by Aines and Rossman (1984a) contain up to 260 wt ppm H2O, garnets from the Wesselton kimberlite, South Africa, up to 70 wt ppm. OH abundances of garnets from a wide variety of rock types occurring as xenoliths in kimberlites from southern Africa (Bell and Rossman 1992b), range from less than 1 up to 135 wt ppm H2O. A pyrope-rich garnet from the Liaoning-50 kimberlite, NE China, was reported by Langer et al. (1993) to contain about 95 wt ppm H2O (calibrated after Maldener et al. 2003).

Table 4. Observed OH defect concentrations in mantle garnets. Mean values in parentheses. Geological occurrence; No. of samples Locality

OH concentration (in wt ppm H2O)

References

High-grade blueschist; Dora Maira, Italy

1

58

Lu and Keppler (1997)

Ultramafic diatremes; Colorado Plateau

11

22 - 112 (59)

Wang et al. (1996)

Kimberlite, xenolith; Southern Africa

166

<1 - 135 (29)

Bell and Rossman (1992b)

Kimberlite, xenolith; Udachnaya, Yakutia

11

<1 - 72 (20)

Snyder et al. (1995)

Kimberlite, xenolith; Siberian Platform

85

<1 - 290 (70)

Matsyuk et al. (1998)

Kimberlite, megacrysts; Monastery, South Africa

19

15 - 74 (44)

Bell et al. (2004a)

Eclogite; Kokchetav, Kazakhstan

6

50 - 150 (90)

Katayama et al. (2006)

Grospydite; Zagadochnaya, Yakutia

1

620

Beran et al. (1993)

Concentration values are derived from IR spectroscopic data, calibrated against H2O values obtained from hydrogen manometry (Bell et al. 1995).

182

Beran & Libowitzky

The calibration of Bell et al. (1995) was used by Matsyuk et al. (1998) for the study of a suite of 200 garnet crystals from kimberlites of the Siberian platform, where H2O values from below the detection limit up to maximum values of 290 wt ppm have been reported. The highest measured H2O content of pyrope included in diamond is 36 wt ppm. The water content of most pyrope samples yields values about 70 wt ppm. According to Snyder et al. (1995) the absolute H2O concentrations of eclogitic pyrope garnets from the Udachnaya kimberlite, Yakutia, generally cluster in the range from near 0 to 22 wt ppm, although samples occur with concentrations exceeding 70 wt ppm (Table 4). The extremely high water content of 620 wt ppm for a mantlederived garnet from the Zagadochnaya kimberlite, Yakutia (grospydite Z 13) reported by Beran et al. (1993) is evidently related to its high grossular component (gross54pyr26alm20). The water contents of pyrope-rich garnet megacrysts reported by Bell et al. (2004a) from the Monastery kimberlite are in the range of 15-74 wt ppm H2O and are positively correlated with Fe-enrichment—the FeO content varies between 9.5 and 14 wt%. An influence of temperature on the OH content of garnet is also suggested. OH partitioning between olivine (plus clinopyroxene) and garnet shows a factor 4-10 variation. FTIR spectroscopy and SIMS have been used by Katayama et al. (2006) to quantify trace amounts of water in pyrope-rich garnets of eclogites from the Kokchetav massif, Kazakhstan, which have been subducted to ~180 km depths. These garnets contain up to 150 wt ppm H2O and are characterized by a single broad band with its maximum ranging from 3630 to 3580 cm−1.

ACCESSORY MINERALS Upper mantle rocks, as typically represented by peridotite and eclogite xenoliths in kimberlites, contain a range of cogenetic nominally anhydrous accessory minerals. Many of the eclogites can be argued to be crustal in origin. However, their accessories can be classified as mantle minerals since they have been subducted to mantle depths. Kyanite and rutile are probably the most abundant phases but coesite, spinel, and zircon are also included in xenoliths and present potential storage sites for hydrogen in the Earth’s mantle (Rossman and Smyth 1990; Beran 1999; Libowitzky and Beran 2004).

Kyanite The triclinic crystal structure of kyanite can be described on the basis of a distorted cubic close-packed oxygen arrangement. The aluminum cations fill 40% of the octahedral interstices in such a way that half of the occupied octahedra form single-chains parallel to the c-axis. The silicon atoms fill 10% of the tetrahedrally coordinated structural sites. There are two distinct silicon (Si1, Si2), four aluminum (Al1-Al4), and ten oxygen atoms (OA-OH, OK and OM) in the unit cell. Eight oxygen atoms are coordinated by one Si and two Al, two oxygen atoms (OB and OF), not bound to Si, are coordinated by four Al. The recognition of OH defects in kyanite by Beran (1971) was initially based on the observation of OH absorption bands in crystals from the classical location Alpe Sponda, Switzerland. In a detailed IR spectroscopic study based on quantitative data, Beran and Götzinger (1987) established the presence of OH groups in kyanites from different geological environments, including a sample from an eclogitic xenolith in kimberlite from the Roberts Victor Mine, South Africa. OH groups in kyanites from the Roberts Victor Mine were also reported by Rossman and Smyth (1990). Beran et al. (1993) detected OH in kyanite from an eclogitic xenolith from the Zagadochnaya kimberlite, Yakutia. Two kyanites extracted by Bell et al. (2004b) from the Frank Smith kimberlite, South Africa, and from an eclogite xenolith in kimberlite from Letlhakane, Botswana, show characteristic absorption features in the OH stretching vibrational region. According to Wieczorek et al. (2004) two main spectral types can be discerned. One type is represented by the mantle-derived kyanite from the Roberts Victor mine, consisting of a band

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

183

triplet with individual bands at 3386, 3410, and 3440 cm−1 of approximately equal intensities and a relatively broad low-energy band with a distinct maximum around 3275 cm−1. Band deconvolution revealed a satellite band at 3260 cm−1. This kyanite contains a significantly enhanced amount of Ti (0.13 wt% TiO2). In contrast, kyanite samples from crustal rocks (and also that from the blueschists of Dora Maira) show a dominating single band at 3386 cm−1 and only very low intensities of the bands at 3410 and 3440 cm−1. The low-energy band maxima at 3275 and 3260 cm−1 are clearly separated and show almost equal intensities. Both spectral types are presented in Figure 8. The pleochroic scheme of the 3386 cm−1 band indicates a preferred orientation of the OH dipole in the nβ′ direction of the (100) cleavage plane. This orientation is in agreement with an incorporation model where OB (not bound to Si) acts as donor oxygen of an OH group pointing directly to the centrosymmetric OB′ atom. The pleochroic behavior of the doublet band at 3275/3260 cm−1 is in agreement with an OH group where OF (not bound to Si) acts as donor and OA as acceptor oxygen (Beran 1971; Wieczorek et al. 2004).

Linear absorption coefficient (cm-1)

The integral molar absorption coefficient for OH in kyanite determined by Bell et al. (2004b) on the basis of 15N NRA of hydrogen amounts to 32900 L·mol−1·cm−2. The recalculated H2O contents of kyanites from Roberts Victor mine, South Africa, range between 100 and 58 wt ppm (Beran and Götzinger 1987; Rossman and Smyth 1990; Wieczorek et al. 2004). The H2O content determined by Bell et al. (2004b) for kyanite from the Letlhakane mine, Botswana, amounts to 27 wt ppm and that from the Frank Smith mine, South Africa, to 230 wt ppm. The H2O content for the kyanite from the Zagadochnaya kimberlite, Yakutia (Beran et al. 1993), amounts to 23 wt ppm (Table 5). However, comparing the H2O values of mantle-derived kyanites with those of kyanites from crustal occurrences (Beran and Götzinger 1987; Bell et al. 2004b; Wieczorek et al. 2004) it is evident that mantle-derived kyanites show enhanced H2O contents. High water contents are apparently related to kyanites formed under high-P,T conditions. Finally, it should be noted that according to Wieczorek et al. (2004) kyanite from a pyrope-rutile-kyanite nodule of the blueschists from Dora Maira, Western Alps, contains 41 wt ppm H2O.

6 4

Dora Maira

2 0 4 3 Roberts Victor

2 1 0 3700

3600

3500

3400

3300

3200

Wavenumber (cm-1) Figure 8. Polarized IR absorption spectra of two types of kyanites measured parallel to the nβ′ direction in (100) sections (modified after Wieczorek et al. 2004).

184

Beran & Libowitzky Table 5. Observed OH concentrations in the accessory mantle minerals kyanite and rutile. Mean values in parentheses.

Geological occurrence; Locality

No. of samples

OH concentration (in wt ppm H2O)

High-grade blueschist; Dora Maira, Italy

1

41

Eclogitic xenolith; Roberts Victor, S.Africa

3

58 - 100 (73)

Kimberlite, xenolith; Letlhakane, Botswana

1

27

Bell et al. (2004b)

Kimberlite, xenolith; Frank Smith, S.Africa

1

230

Bell et al. (2004b)

Grospydite; Zagadochnaya, Yakutia

1

23

Beran et al. (1993)

High-grade blueschist; Dora Maira, Italy

1

270

Hammer and Beran (1991)

Kimberlite, xenolith; Jagersfontein, S.Africa

4

1710 - 5470 (3740)

Vlassopoulos et al. (1993)

Eclogite Kokchetav, Kazakhstan

3

330 - 740 (560)

References

Kyanite Wieczorek et al. (2004) Beran and Götzinger (1987); Rossman and Smyth (1990); Wieczorek et al. (2004)

Rutile

Katayama et al. (2006)

Concentration values are derived from IR spectroscopic data, calibrated against H2O values obtained from Nuclear Reaction Analysis (for kyanite see Bell et al. 2004b, for rutile Maldener et al. 2001).

15

N

Rutile The presence of OH defects in rutile from natural occurrences was established by Beran and Zemann (1971) on the basis of strongly pleochroic absorption bands in the 3300 cm−1 region. Rutile as inclusion in Dora Maira pyrope, reported by Hammer and Beran (1991) is characterized by absorption bands at 3320 and 3280 cm−1. A doublet of sharp bands at 3320 and 3300 cm−1 has been reported by Rossman and Smyth (1990) in the IR spectra of rutiles from eclogites of Roberts Victor mine, South Africa. According to Vlassopoulos et al. (1993) mantle-derived Nb- and Cr-rich rutiles from kimberlites of Jagersfontein, South Africa, are characterized by a single band centered at 3290 cm−1. A single OH band at 3280 cm−1 has also been observed by Katayama et al. (2006) in rutile of mantle eclogites from the Kokchetav massif with corresponding H2O contents ranging up to 740 wt ppm. Polarized spectra of this rutile sample are shown in Figure 9. From studies on synthetic material Bromiley and Hilairet (2005) concluded that the absorption band at 3279 cm−1, present in the spectra of many natural samples, corresponds to OH groups that occur independently of various cation substitutions. The OH absorption bands in rutile are strongly pleochroic, with maximum absorption perpendicular to the tetragonal c axis. Based on the pleochroic scheme, Beran and Zemann (1971) proposed that OH groups at the oxygen sites are oriented approximately perpendicular to the plane of the three coordinating Ti atoms. This orientation is also confirmed by the excellent agreement between expected hydrogen bond lengths calculated from band energies (~3300 → ~2.75-2.80 Å according to the d(O···O)–ν correlation of Libowitzky 1999) and

185

-1

Linear absorption coefficient (cm )

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

15

E^c E // c

Kokchetav

10

5

0 3500

3400

3300

3200

3100

Wavenumber (cm-1) Figure 9. Polarized IR absorption spectra of accessory rutile from a mantle eclogite occurrence in Kazakhstan (modified after Katayama et al. 2006).

actually observed O···O distances of 2.78 Å. A strong deviation from this reasonable orientation has been reported by Swope et al. (1995) on the basis of neutron diffraction data obtained from an OH-rich mantle-derived rutile from Roberts Victor mine. Their proposed position of the H atom, located by examining the negative residuals in difference Fourier maps, is near to the shared edge (d(O···O) only 2.53 Å!) of the TiO6 octahedron. Using unpolarized radiation, but crystal plates cut parallel to c, Hammer and Beran (1991) determined the integral molar absorption coefficient for OH in rutile to 6540 L·mol−1·cm−2. If polarized radiation is used, Paterson’s (1982) orientation factor γ = ¼ has to be considered, resulting in a value which amounts to 26160 L·mol−1·cm−2. This value is based on a coulometric water determination. 15N NRA has been used by Maldener et al. (2001) for the calibration of the water content in rutile from crustal occurrences, resulting in an integral molar absorption coefficient of 38000 ± 4000 L·mol−1·cm−2. Using this recent calibration, the H2O content of the Dora Maira rutile (Hammer and Beran 1991) amounts to 270 wt ppm. The H2O contents of mantle-derived rutiles from Jagersfontein, South Africa, reported by Vlassopoulos et al. (1993) range from about 1700 to 5300 wt ppm (Table 5). For comparison, maximum water contents of crustal-derived rutiles range up to about 2000 wt ppm (Hammer and Beran 1991; Vlassopoulos et al. 1993; Maldener et al. 2001), thus supporting the idea of enhanced H2O contents in mantle-derived materials. Structural OH groups, strongly polarized perpendicular to c, have also been reported for a number of isostructural minerals, i.e., cassiterite, SnO2, from various localities (Beran and Zemann 1971; Maldener et al. 2001; Losos and Beran 2004), sellaite, MgF2, from Brazil (Beran and Zemann 1985) and synthetic stishovite, the highest-pressure polymorph of SiO2 (Pawley et al. 1993; Bolfan-Casanova et al. 2000).

Coesite In coesite, the SiO4 tetrahedra are connected into four-membered rings that are linked to a monoclinic framework structure which is similar to that of feldspar group minerals. The solubility of hydrogen in coesite was experimentally studied by Mosenfelder (2000). IR spectra of coesite crystals show distinct bands at 3573, 3523, 3459, and 3298 cm−1. Based on Paterson’s (1982) calibration, H2O values of coesites synthesized at 1200 °C and 5-10 GPa range from 43 to 212 wt ppm. These results show that the solubility of OH in coesite is comparable to that of

186

Beran & Libowitzky

pyrope-rich garnets. However, IR investigations on a variety of ultra high-pressure metamorphic rocks have failed in all cases to detect the presence of hydrogen (Mosenfelder 2000). Koch-Müller et al. (2001) confirmed the presence of hydrogen in coesite synthesized at 3.1-7.5 GPa and 700, 800, and 1100 °C. IR spectra are characterized by three dominating bands at 3575, 3516, and 3459 cm−1, two additional weak bands are centered at 3296 and 3210 cm−1. Based on detailed polarized investigations, under the assumption of a vacant Si2 site the authors proposed a hydrogarnet substitution model, where O2, O3, O4, and O5 oxygen atoms are replaced by OH. Ion microprobe measurements revealed H2O concentrations ranging from 4 to 200 wt ppm. The derived molar absorption coefficient for OH in coesite amounts to 190000 ± 30000 L·mol−1·cm−2. This value is by a factor of about two higher than that deduced from Paterson’s (1982) diagram. Koch-Müller et al. (2003) investigated the incorporation of hydrogen into the coesite structure at pressures ranging from 4-9 GPa and temperatures from 750-1300 °C, confirming the presence of OH bands in the 3580-3450 cm−1 range. At 8.5 GPa and 1200 °C the hydrogen incorporation mechanism changes significantly; four bands appear at 3460, 3422, 3407, and 3379 cm−1 and the 3580-3450 cm−1 bands disappear. Coesite from grospydites of the Roberts Victor mine, South Africa, has been described by Rossman and Smyth (1990) to contain traces of molecular H2O, present as hydrous material residing in fractures. Based on measurements with synchrotron IR radiation, OH absorption bands in the 3600-3350 cm−1 region have been reported by Koch-Müller et al. (2003) in natural coesite occurring as an inclusion in diamond. The calculated H2O content, calibrated against ion microprobe data, amounts to about 135 wt ppm. In contrast, no OH was detected in coesite by Mosenfelder et al. (2005) in pyropes from Dora Maira, though the silica phases quartz, chalcedony, and opal which surround the coesite inclusions, show considerable amounts of H2O. Similar signatures have been observed for coesite in a grospydite xenolith from the Roberts Victor mine (Mosenfelder et al. 2005).

Spinel Ringwoodite, the spinel polymorph of (Mg,Fe)2SiO4 likely to occur in the transition zone of the Earth’s mantle can incorporate major amounts of water in form of OH groups (Bolfan-Casanova 2005; Ohtani 2005). The corresponding IR spectra are characterized by broad absorption bands centered in the 3700-3100 cm−1 region (Kohlstedt et al. 1996; BolfanCasanova et al. 2000; Kudoh et al. 2000; Ohtani et al. 2000; Smyth et al. 2003). Singlecrystals of γ-Mg2GeO4 spinel were synthesized by Hertweck and Ingrin (2005) under hydrous conditions at 1.9 GPa and 1000 °C revealing a relatively sharp OH absorption band at 3531 cm−1. The estimated H2O content is in the order of 5-10 wt ppm. Preliminary results have been reported by Halmer et al. (2003) from highly disordered non-stoichiometric Verneuil-grown MgAl spinels. Two significant sharp bands centered at 3510 and 3355 cm−1 with varying band intensities are assigned to weakly hydrogen-bonded OH defects. No indications for the presence of OH have been observed in natural aluminate (MgAl2O4rich) spinels of mantle eclogites (Rossman and Smyth 1990). OH groups in spinels have neither been found in samples from the upper Earth’s mantle nor from crustal occurrences. One may speculate whether disorder is a basic requirement for the incorporation of OH defects in the spinel structure. An additional complication results from the fact that in natural spinel samples the OH stretching vibrational region is usually overlapped by extremely strong and broad absorptions due to d-d transitions of tetrahedrally coordinated Fe2+ (Skogby and Halenius 2003), thus making the detection of trace OH a difficult task.

Zircon Incorporation of hydrous species in natural zircon is a widely observed phenomenon. Observed water contents may range from almost zero in crystalline samples up to more than 15 wt% H2O in radiation-damaged metamict zircons. The spectroscopy of zircon has recently

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

187

been reviewed by Nasdala et al. (2003). Whereas well-crystallized samples are characterized by sharp and strongly pleochroic OH absorption bands, metamict zircons usually display additional broad non-pleochroic IR absorption bands. The sharp bands indicate incorporation of structural OH groups and the broad bands are interpreted in terms of H2O molecules. Narrow IR absorption bands in crystalline zircons have been described by Nasdala et al. (2001) at 3180, 3420, and 3385 cm−1; the latter band is polarized perpendicular to the c axis, the former two are polarized parallel to c (Fig. 10). The 3385 cm−1 band is assigned to an OH group coordinated by one Si and two Zr sites, probably substituted by REE3+ (ZrZrSi). Assuming cation vacancies (), the 3180 cm−1 band is assigned to OH coordinated by one Si and one Zr (ZrSi), and the 3420 cm−1 band to OH coordinated by two Zr atoms (ZrZr) (Fig. 11). OH defects in a mantle-derived zircon sample from Kimberley, South Africa, were reported by Woodhead et al. (1991). The estimated H2O content amounts to about 70 wt ppm (based on the garnet calibration published by Bell et al. 1995). The authors stated that this value presents the maximum OH content observed in crystalline zircon in comparison with a suite of 30 zircons of crustal occurrences, suggesting high activity of H2O in the mantle. The corresponding IR spectrum is characterized by sharp and relatively strong absorption bands at 3417 and 3384 cm−1 and by weak bands at 3510 and 3180 cm−1. Very similar spectral features are evident in the unpolarized IR spectrum of a well-ordered zircon from a Yakutian kimberlite pipe reported by Nasdala et al. (2001), showing a strong absorption band at 3420 and a weak band at 3180 cm−1 (Fig. 10). Bell and Rossman (1992a) reported H2O concentrations of zircons, formed by crystallization of a magma at high pressure, which range from about 50 to 100 wt ppm. Water contents determined by Bell et al. (2004a) on zircons associated with high-Fe olivines, ilmenite and phlogopite from the Monastery kimberlite, South Africa, are in the 28-34 wt ppm range. The spectra of these zircons are characterized by two sharp pleochroic bands centered around 3420 and 3380 cm−1, superimposed on a small amount of broad, featureless absorption. However, the rarity of zircon in mantle rocks renders it of minor importance for storing essential amounts of hydrogen.

Linear absorption coefficient (cm-1)

8

Kaalvallei, RSA

6

E // c

4 E^c Mir, Yakutia

2 unpolarized

0 3800

3700

3600

3500

3400

3300

3200

3100

3000

Wavenumber (cm-1) Figure 10. Characteristic IR absorption spectra of well-ordered (non-metamict) zircons from kimberlitic xenoliths (modified after Bell and Rossman 1992a, Nasdala et al. 2001).

188

Beran & Libowitzky

(a)

Zr O Si

(b) [Zr]

(c)

[Si]

Figure 11. Potential locations of OH defects in crystalline zircon from kimberlitic xenoliths, coordinated by (a) Zr, Zr, Si, (b) Zr, Zr, Si, (c) Zr, Zr, Si (modified after Nasdala et al. 2001).

ACKNOWLEDGMENTS The authors wish to thank H. Keppler and J. Smyth for the invitation to contribute to the present MSA volume. H. Keppler, J.L. Mosenfelder and an anonymous referee helped to improve the quality of the manuscript. The topics of this paper were partly sponsored by the European Commission, Human Potential-Research Training Network, HPRN-CT-2000-0056.

REFERENCES Ackermann L, Cemic L, Langer K (1983) Hydrogarnet substitution in pyrope: a possible location for “water” in the mantle. Earth Planet Sci Lett 62:208-214 Aines RD, Rossman GR (1984a) Water content of mantle garnets. Geology 12:720-723 Aines RD, Rossman GR (1984b) The hydrous component in garnets: pyralspites. Am Mineral 69:1116-1126 Andrut M, Wildner M, Beran A (2002) The crystal chemistry of birefringent natural uvarovites. Part IV. OH defect incorporation mechanisms in non-cubic garnets derived from polarized IR spectroscopy. Eur J Mineral 14:1019-1026 Asimow PD, Stein LC, Mosenfelder JL, Rossman GR (2006) Quantitative polarized infrared analysis of trace OH in populations of randomly oriented mineral grains. Am Mineral 91:278-284 Bai Q, Kohlstedt DL (1992) Substantial hydrogen solubility in olivine and implications for water storage in the mantle. Nature 357:672-674

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

189

Bai Q, Kohlstedt DL (1993) Effects of chemical environment on the solubility and incorporation mechanism for hydrogen in olivine. Phys Chem Minerals 19:460-471 Bauerhansl P, Beran A (1997) Trace hydrogen in the olivine-type minerals chrysoberyl, Al2BeO4 and sinhalite, MgAlBO4 - a polarized FTIR spectroscopic study. Schweiz Min Petr Mitt 77:131-136 Bell DR, Rossman GR (1992a) Water in Earth’s mantle: The role of nominally anhydrous minerals. Science 255:1391-1397 Bell DR, Rossman GR (1992b) The distribution of hydroxyl in garnets from the subcontinental mantle of southern Africa. Contrib Mineral Petrol 111:161-178 Bell DR, Ihiniger PD, Rossman GR (1995) Quantitative analysis of trace OH in garnet and pyroxenes. Am Mineral 80:465-474 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res 108, B2:21052113 Bell DR, Rossman GR, Moore RO (2004a) Abundance and partitioning of OH in a high-pressure magmatic system: Megacrysts from the Monastery kimberlite, South Africa. J Petrol 45:1539-1564 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2004b) Hydroxide in kyanite: A quantitative determination of the absolute amount and calibration of the IR spectrum. Am Mineral 89:998-1003 Beran A (1971) Messung des Ultrarot-Pleochroismus von Mineralen XII. Der Pleochroismus der OHStreckfrequenz in Disthen. Tschermaks Min Petr Mitt 16:129-135 Beran A (1999) Contribution of IR spectroscopy to the problem of water in the Earth’s mantle. In: Microscopic Properties and Processes in Minerals. Wright K, Catlow R (eds) NATO Science Series. Kluwer Acad. Publishers, p 523-538 Beran A, Götzinger MA (1987) The quantitative IR spectroscopic determination of structural OH groups in kyanites. Mineral Petrol 36:41-49 Beran A, Libowitzky E (2003) IR spectroscopic characterization of OH defects in mineral phases. Phase Trans 76:1-15 Beran A, Putnis A (1983) A model of the OH positions in olivine, derived from infrared-spectroscopic investigations. Phys Chem Minerals 9:57-60 Beran A, Zemann J (1971) Messung des Ultrarot-Pleochroismus von Mineralen. XI. Der Pleochroismus der OHStreckfrequenz in Rutil, Anatas, Brookit und Cassiterit. Tschermaks Min Petr Mitt 15:71-80 Beran A, Zemann J (1985) Polarized absorption spectra of sellaite from the Brumado mine, Brazil, in the near infrared. Bull Geol Soc Finlande 57:113-118 Beran A, Langer K, Andrut M (1993) Single crystal infrared spectra in the range of OH fundamentals of paragenetic garnet, omphacite and kyanite in an eclogitic mantle xenolith. Mineral Petrol 48:257-268 Berry AJ, Hermann J, O’Neill HSC, Foran GJ (2005) Fingerprinting the water site in mantle olivine. Geology 33:869-872 Bolfan-Casanova N (2005) Water in the Earth’s mantle. Mineral Mag 69:229-257 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett 182:209-221 Bromiley GD, Hilairet N (2005) Hydrogen and minor element incorporation in synthetic rutile. Mineral Mag 69:345-358 Cohen-Addad C, Ducros P, Bertaut EF (1967) Étude de la substitution du groupement SiO4 par (OH)4 dans les composés Al2Ca3(OH)12 et Al2Ca3(SiO4)2,16(OH)3,36 de type grenat. Acta Crystallogr 23:220-230 Geiger CA, Langer K, Bell DR, Rossman GR, Winkler B (1991) The hydroxide component in synthetic pyrope. Am Mineral 76:49-59 Geiger CA, Stahl A, Rossman GR (2000) Single-crystal IR- and UV/VIS- spectroscopic measurements on transition-metal-bearing pyrope: the incorporation of hydroxide in garnet. Eur J Mineral 12:259-271 Halmer MM, Libowitzky E, Beran A (2003) IR spectroscopic determination of OH defects in spinel group minerals. Geophys Res Abstr 5:06742 Hammer VMF, Beran A (1991) Variations in the OH concentration of rutiles from different geological environments. Mineral Petrol 45:1-9 Hertweck B, Ingrin J (2005) Hydrogen incorporation in a ringwoodite analogue: Mg2GeO4 spinel. Mineral Mag 69:337-343 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Katayama I, Nakashima S, Yurimoto H (2006) Water content in natural eclogite and implication for water transport into the deep upper mantle. Lithos 86:245-259 Kent AJR, Rossman GR (2002) Hydrogen, lithium, and boron in mantle-derived olivine: The role of coupled substitutions. Am Mineral 87:1432-1436 Keppler H, Bolfan-Casanova N (2006) Thermodynamics of water solubility and partitioning. Rev Mineral Geochem 62:193-230

190

Beran & Libowitzky

Khisina NR, Wirth R (2002) Hydrous olivine (Mg1-yFe2+y)2-xvxSiO4H2x – a new DHMS phase of variable composition observed as nanometer-sized precipitations in mantle olivine. Phys Chem Minerals 29:98111 Khisina NR, Wirth R, Andrut M, Ukhanov AV (2001) Extrinsic and intrinsic mode of hydrogen occurrence in natural olivines: FTIR and TEM investigation. Phys Chem Minerals 28:291-301 Khomenko VM, Langer K, Beran A, Koch-Müller M, Fehr T (1994) Titanium substitution and OH-bearing defects in hydrothermally grown pyrope crystals. Phys Chem Minerals 20:483-488 Kitamura M, Kondoh S, Morimoto N, Miller GH, Rossman GR, Putnis A (1987) Planar OH-bearing defects in mantle olivine. Nature 328:143-145 Kobayashi S, Shoji T (1983) Infrared analysis of the grossular-hydrogrossular series. Mineral J 11:331-343 Koch-Müller M, Fei Y, Hauri E, Liu Z (2001) Location and quantitative analysis of OH in coesite. Phys Chem Minerals 28:693-705 Koch-Müller M, Dera P, Fei Y, Reno B, Sobolev N, Hauri E, Wysoczanski R (2003) OH− in synthetic and natural coesite. Am Mineral 88:1436-1445 Koch-Müller M, Matsyuk SS, Rhede D, Wirth R, Khisina N (2006) Hydroxyl in mantle olivine xenocrysts from the Udachnaya kimberlite pipe. Phys Chem Minerals 33:276-287 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kudoh Y, Kuribayashi T, Mizobata H, Ohtani E (2000) Structure and cation disorder of hydrous ringwoodite, γ-Mg1.89Si0.98H0.30O4. Phys Chem Minerals 27:474-479 Kurosawa M, Yurimoto H, Sueno S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Minerals 24:385-395 Lager GA, Armbruster T, Rotella FJ, Rossman GR (1989) OH substitution in garnets: X-ray and neutron diffraction, infrared, and geometric-modeling studies. Am Mineral 74:840-851 Langer K, Robarick E, Sobolev NV, Shatsky VS, Wang W (1993) Single-crystal spectra of garnets from diamondiferous high-pressure metamorphic rocks from Kazakhstan: indications for OH−, H2O, and FeTi charge transfer. Eur J Mineral 5:1091-1100 Lemaire C, Kohn SC, Brooker RA (2004) The effect of silica activity on the incorporation mechanisms of water in synthetic forsterite: a polarised infrared spectroscopic study. Contrib Mineral Petrol 147:48-57 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H···O hydrogen bond lengths in minerals. Mh Chem 130:1047-1059 Libowitzky E, Beran A (1995) OH defects in forsterite. Phys Chem Minerals 22:387-392 Libowitzky E, Beran A (2004) IR spectroscopic characterisation of hydrous species in minerals. In: Spectroscopic Methods in Mineralogy. Beran A, Libowitzky E (eds), EMU Notes Mineral 6, Eötvös Univ Press, p 227-279 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52 Libowitzky E, Rossman GR (1996) Principles of quantitative absorbance measurements in anisotropic crystals. Phys Chem Minerals 23:319-327 Libowitzky E, Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82:11111115 Liu Z, Lager GA, Hemley RJ, Ross NL (2003) Synchrotron infrared spectroscopy of OH-chondrodite and OHclinohumite at high pressure. Am Mineral 88:1412-1415 Losos Z, Beran A (2004) OH defects in cassiterite. Mineral Petrol 81:219-234 Lu R, Keppler H (1997) Water solubility in pyrope to 100 kbar. Contrib Mineral Petrol 129:35-42 Maldener J, Rauch F, Gavranic M, Beran A (2001) OH absorption coefficients of rutile and cassiterite deduced from nuclear reaction analysis and FTIR spectroscopy. Mineral Petrol 71:21-29 Maldener J, Hösch A, Langer K, Rauch F (2003) Hydrogen in some natural garnets studied by nuclear reaction analysis and vibrational spectroscopy. Phys Chem Minerals 30:337-344 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554-570 Matsyuk SS, Langer K (2004) Hydroxyl in olivines from mantle xenoliths in kimberlites of the Siberian platform. Contr Mineral Petrol 147:413-437 Matsyuk SS, Langer K, Hösch A (1998) Hydroxyl defects in garnets from mantle xenoliths in kimberlites of the Siberian platform. Contr Mineral Petrol 132:163-179 Matveev S, O’Neill HSC, Ballhaus C, Taylor WR, Green DH (2001) Effect of silica activity on OH− IR spectra of olivine: Implication for low-α-SiO2 mantle metasomatism. J Petrol 42:721-729 Matveev S, Portnyagin M, Ballhaus C, Brooker R, Geiger CA (2005) FTIR spectrum of phenocryst olivine as an indicator of silica saturation in magmas. J Petrol 46:603-614 Miller GH, Rossman GR, Harlow GE (1987) The natural occurrence of hydroxide in olivine. Phys Chem Minerals 14:461-472 Mosenfelder JL (2000) Pressure dependence of hydroxyl solubility in coesite. Phys Chem Minerals 27:610617

Water in Mantle Minerals: Olivine, Garnet, & Accessory Minerals

191

Mosenfelder JL, Schertl H-P, Smyth JR, Liou JG (2005) Factors in the preservation of coesite: The importance of fluid infiltration. Am Mineral 90:779-789 Mosenfelder JL, Deligne NI, Asimow PD, Rossman GR (2006) Hydrogen incorporation in olivine from 2-12 GPa. Am Mineral 91:285-294 Nasdala L, Beran A, Libowitzky E, Wolf D (2001) The incorporation of hydroxyl groups and molecular water in natural zircon (ZrSiO4). Am J Sci 301:831-857 Nasdala L, Zhang M, Kempe U, Panczer G, Gaft M, Andrut M, Plötze M (2003) Spectroscopic methods applied to zircon. Rev Mineral Geochem 53:427-467 Ohtani E (2005) Water in the mantle. Elements 1:25-30 Ohtani E, Mizobata H, Yurimoto H (2000) Stability of dense hydrous magnesium silicate phases in the systems Mg2SiO4-H2O and MgSiO3-H2O at pressures up to 27 GPa. Phys Chem Minerals 27:533-544 Paterson MS (1982) The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bull Minéral 105:20-29 Pawley AR, McMillan PF, Holloway JR (1993) Hydrogen in stishovite, with implications for mantle water content. Science 261:1024-1026 Rossman GR (1990) Hydrogen in “anhydrous” minerals. Nucl Instr Meth Phys Res B45:41-44 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Rossman GR, Aines RD (1986) Spectroscopy of a birefringent grossular from Asbestos, Quebec, Canada. Am Mineral 71:779-780 Rossman GR, Aines RD (1991) The hydrous components in garnets: Grossular-hydrogrossular. Am Mineral 76: 1153-1164 Rossman GR, Smyth JR (1990) Hydroxyl contents of accessory minerals in mantle eclogites and related rocks. Am Mineral 75: 775-780 Rossman GR, Beran A, Langer K (1989) The hydrous component of pyrope from the Dora Maira Massif, Western Alps. Eur J Mineral 1:151-154 Rossman GR, Rauch F, Livi R, Tombrello TA, Shi CR, Zhou ZY (1988) Nuclear reaction analysis of hydrogen in almandine, pyrope, and spessartite garnets. N Jb Miner Mh 1988:172-178 Skogby H (1999) Water in nominally anhydrous minerals. In: Microscopic Properties and Processes in Minerals. Wright K, Catlow R (eds) NATO Science Series, Kluwer Acad Publishers, p 509-522 Skogby H (2006) Water in natural mantle minerals I: pyroxenes. Rev Mineral Geochem 62:155-167 Skogby H, Halenius U (2003) An FTIR study of tetrahedrally coordinated ferrous iron in the spinel-hercynite solid solution. Am Mineral 88:489-492 Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Langenhorst F, McCammon CA (2003) Structural systematics of hydrous ringwoodite and water in Earth’s interior. Am Mineral 88:1402-1407 Snyder GA, Taylor LA, Jerde EA, Clayton RN, Mayeda TK, Deines P, Rossman GR, Sobolev NV (1995) Archean mantle heterogeneity and the origin of diamondiferous eclogites, Siberia: Evidence from stable isotopes and hydroxyl in garnet. Am Mineral 80:799-809 Song S, Zhang L, Chen J, Liou JG, Niu Y (2005) Sodic amphibole exsolutions in garnet from garnet-peridotite, North Qaidam UHPM belt, NW China: Implications for ultradeep-origin and hydroxyl defects in mantle garnets. Am Mineral 90:814-820 Swope RJ, Smyth JR, Larson AC (1995) H in rutile-type compounds: I. Single-crystal neutron and X-ray diffraction study of H in rutile. Am Mineral 80:448-453 Vlassopoulos D, Rossman GR, Haggerty SE (1993) Coupled substitution of H and minor elements in rutile and the implications of high OH contents in Nb- and Cr-rich rutile from the upper mantle. Am Mineral 78: 1181-1191 Wang L, Zhang Y, Essene EJ (1996) Diffusion of the hydrous component in pyrope. Am Mineral 81:706-718 Wieczorek A, Libowitzky E, Beran A (2004) A model for the OH defect incorporation in kyanite based on polarised IR spectroscopic investigations. Schweiz Min Petr Mitt 84:333-343 Withers AC, Wood BJ, Carroll MR (1998) The OH content of pyrope at high pressure. Chem Geol 147:161171. Woodhead JA, Rossman GR, Thomas AP (1991) Hydrous species in zircon. Am Mineral 76:1533-1546 Zhao Y-H, Ginsberg SB, Kohlstedt DL (2004) Solubility of hydrogen in olivine: dependence on temperature and iron content. Contrib Mineral Petrol 147:155-161

9

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 193-230, 2006 Copyright © Mineralogical Society of America

Thermodynamics of Water Solubility and Partitioning Hans Keppler Bayerisches Geoinstitut Universität Bayreuth 95440 Bayreuth, Germany e-mail: [email protected]

Nathalie Bolfan-Casanova Laboratoire Magmas et Volcans - CNRS UMR 6524 63038 Clermont-Ferrand cedex, France e-mail: [email protected]

INTRODUCTION When it was realized that nominally anhydrous minerals could be a major storage site of water in the mantle (Smyth 1987; Bell and Rossman 1992), it also became rather obvious that many mantle samples probably have lost most of their water during ascent (e.g., Ingrin and Skogby 2000). While analyses of natural samples therefore in many circumstances only provide a lower limit of the actual water content in the mantle, measurements of water solubility give an upper limit of the amount of water that might be stored in a mineral. Experimental measurements of water solubility have therefore naturally evolved in the last decades as a major tool for understanding the water storage capacity of the mantle. Closely related to studies of water solubility are studies of water partitioning among minerals. If the water solubility in two minerals is known under given conditions, the partition coefficient is just the ratio of the water solubilities under the same conditions, provided that the composition of the fluid coexisting with the two minerals is the same. Since the pioneering work of Bai and Kohlstedt (1992) and of Kohlstedt et al. (1996), the water solubility in all major upper mantle minerals has been studied as well as the partitioning of water between the minerals of the lower mantle and the transition zone. Solubility studies by themselves do not directly give the water content in the mantle. They constrain the storage capacity of the mantle, i.e., the maximum amount of water that may be stored, if the mantle were water-saturated. However, as most of the mantle is very likely waterundersaturated, the actual water content is probably far below the storage capacity. Estimates of the actual abundance and distribution of water in the mantle may be obtained by combining solubility and partitioning data with direct observations of water contents in the upper mantle. If one assumes chemical equilibrium throughout the mantle one can model the water abundance in the transition zone and the lower mantle using experimentally derived water partition coefficients. This was first demonstrated by Bolfan-Casanova et al. (2000) and there is evidence from direct geophysical observations that this model is not too far away from reality (Huang et al. 2005; see also Hirschmann 2006). Whether the water distribution in the Earth’s interior really approaches chemical equilibrium is, however, uncertain, although the relatively high diffusion coefficients of hydrogen in minerals (Ingrin and Blanchard 2006; this volume) and the high mobility of aqueous fluids would favor an equilibrium distribution. Ultimately, the actual abundance of water in the mantle can only be established by direct analyses of undegassed mantle samples and by geophysical observations. 1529-6466/06/0062-0009$05.00

DOI: 10.2138/rmg.2006.62.9

194

Keppler & Bolfan-Casanova

Finally, measurements of water solubility can help to constrain the recycling of water into the mantle in subduction zone, because they determine how much water can be retained in nominally anhydrous minerals when hydrous phases break down and release an aqueous fluid.

BASIC THERMODYNAMICS OF WATER SOLUBILITY AND PARTITIONING The meaning of the term “water solubility” At relatively low pressures, water solubility in a mineral can be conveniently defined as the equilibrium water content of the respective mineral coexisting with an aqueous fluid. This definition is often meant to imply that the water activity in the fluid is close to unity. However, at high pressures and temperatures, the solubility of silicates in aqueous fluids becomes appreciable. In this situation, water activity in the fluid can be greatly reduced (e.g., Mibe et al. 2002). Naturally, the reduction of water activity will also lead to reduced equilibrium water contents in the coexisting minerals. This can be a problem if a thermodynamic model of water solubility in a mineral is calibrated at low P and T, where water activity in the fluid is close to unity and if this model is then extrapolated to much higher pressures and temperatures. In this case, the equilibrium water contents in minerals will be significantly overestimated, because the model predicts water solubilities in equilibrium with a hypothetical fluid consisting of essentially pure water, while in reality, water activity in the fluid may be significantly depressed due to dissolved silicate components. Probably in all silicate-water systems, silicate melt and aqueous fluid become completely miscible at very high pressures and temperatures (Shen and Keppler 1997; Bureau and Keppler 1999; Kessel et al. 2005). The critical curve, which defines the pressure and temperature conditions where the solvus between fluid and melt closes, has been mapped out for some compositions. Some terms that are meaningful below the critical curve cannot be applied anymore at the “supercritical” pressures and temperatures beyond the critical curve. For example, one cannot define a water-saturated melt anymore, because this would imply a silicate melt coexisting with a separate aqueous fluid phase, which is impossible under conditions where fluid and melt are completely miscible. For the same reason, one cannot define the water solubility in a mineral in a simple way anymore, if the mineral coexists with a silicate melt, because in most cases, the water content of the melt and therefore the water activity imposed by the melt will depend on the bulk composition of the system studied (e.g., they will depend on the amount of water loaded into a sample capsule). There are some circumstances, however, where one can define the term “water solubility” in a thermodynamically meaningful way even at pressures and temperatures beyond the critical curve. This is possible if the phase assemblage is invariant at given P und T and therefore buffers water activity and water solubility in all coexisting phases. Demouchy et al. (2005) studied the water content in Mg2SiO4 wadsleyite coexisting with MgSiO3 clinopyroxene and a hydrous silicate melt in the system MgO-SiO2-H2O. According to the phase rule, the coexistence of three phases in a three-component system leaves two degrees of freedom. If pressure and temperature are fixed, these two degrees of freedom are used up and the system is invariant, i.e., the composition and therefore the water content of all phases will be a function of pressure and temperature only. Increasing the bulk water content in the experimental charge at constant P and T will simply increase the melt fraction without changing melt composition. Under these circumstances, it is justified to call the measured water content of wasleyite the water solubility in wadsleyite in equilibrium with a hydrous melt and clinopyroxene. It should be noted, however, that this is a definition of water solubility which is very different from the one generally used at lower pressures.

Thermodynamics of Water Solubility and Partitioning

195

In numerous experimental studies, water contents of run products were measured and very often, these water contents were called “water solubility” without reference to any proper definition of this term. In many cases the term “water solubility” was even used in experiments were the phases investigated neither coexisted with a hydrous fluid nor with a hydrous silicate melt. This should be strictly avoided. Throughout this article, we will use the term water solubility according to the following definitions: Definition 1: Water solubility is the equilibrium water content of a mineral coexisting with an aqueous fluid phase. Definition 2: Water solubility is the equilibrium water content of a mineral coexisting with a hydrous melt in a phase assemblage that buffers the compositions of all coexisting phases in such a way that the composition of each phase only depends on pressure and temperature. Throughout the discussion of water solubility in upper mantle minerals, we will usually refer to definition 1. Definition 2 will only be used in a few cases at pressures and temperatures of the deepest upper mantle, the transition zone or the lower mantle, i.e., at conditions that are likely to be beyond the critical curve in the systems considered.

Thermodynamics of water solubility “Water” or “hydrogen” is dissolved in nominally anhydrous minerals as OH groups or rarely as molecular water. Molecular hydrogen (H2) is probably an important diffusing species in these minerals, particularly during dehydration processes that involve the oxidation of ferrous iron or during hydration processes involving reduction of ferric iron. However, it has never been demonstrated that molecular hydrogen would contribute significantly to the storage of hydrogen in these minerals, although this may be possible under very reducing conditions. One can therefore rather safely assume that under normal circumstances, all hydrogen in a nominally anhydrous mineral will be present as OH groups or as molecular water. Since the chemical composition of silicates is usually expressed in the form of oxides, the terms “hydrogen solubility” and “water solubility” essentially have the same meaning. The thermodynamics of water solubility is primarily controlled by the type of OH defect that forms. These may be isolated OH groups, pairs of OH groups or the “hydrogarnet defect” which is a cluster of four OH groups (Table 1). Isolated OH groups may form by the direct substitution of a proton for a univalent cation such as Na+ or Li+, by the substitution of a trivalent cation and a proton for a tetravalent cation (e.g., Al3+ + H+ for Si4+) or by the substitution of a trivalent cation and a proton for two divalent cations (e.g., Al3+ + H+ for 2 Mg2+). The latter two substitution mechanisms are very important for the dissolution of water in aluminous orthopyroxene (Rauch and Keppler 2002; Mierdel 2006). The reaction between water and the mineral can be written as H2Ofluid + Omineral = 2 OHmineral

(1)

where O is some unprotonated oxygen atom in the structure of the mineral. The equilibrium constant K1 of this reaction is K1 =

2 aOH f H 2 O aO

(2)

where fH2O is the water fugacity. If the concentration of OH, i.e., the “water solubility” is proportional to the activity of OH and if the activity of the unprotonated oxygen atoms is constant, which is very likely at the low OH concentrations usually involved, this equation implies that water solubility should be proportional to the square root of water fugacity: cwater ∼ fH02.5O

(3)

196

Keppler & Bolfan-Casanova Table 1. Hydrogen-bearing defects in nominally anhydrous minerals. Substitution mechanism H+ + Al3+ ↔ 2 Mg2+ H+ + Al3+ ↔ Si4+ H+ + B3+ ↔ Al3+ H+ + Li+ ↔ Mg2+ 2 H+ ↔ Mg2+

2 H+ + Mg2+ ↔ Si4+ interstitial H2O 4 H+ ↔ Si4+

Mineral

Reference

------------- Isolated protons ------------orthopyroxene Mierdel (2006) orthopyroxene Rauch and Keppler (2002) B-rich olivine Sykes et al. (1994) possibly in pyrope Lu and Keppler (1997) --------------- Proton pairs --------------olivine Smyth, pers. comm. enstatite Rauch and Keppler (2002) wadsleyite Smyth (1987) ringwoodite Smyth et al. (2003) possibly MgSiO3 perovskite Ross et al. (2003) possibly ringwoodite Kudoh et al. (2000) feldspars Johnson and Rossman (2004) --------- Cluster of four protons ---------garnet Ackermann et al. (1983) Lager et al. (2005)

OH-pairs, i.e., two OH groups that cannot dissociate away from each other, typically form when two protons substitute for divalent cations, such as Mg2+. This is believed to be an important substitution mechanism in olivine and in Al-free enstatite. Alternatively, OH pairs may also be generated by the substitution of two protons and a divalent cation for a tetravalent cation. Such a substitution would for example be the replacement of Si4+ by Mg2+ plus two protons, although such a substitution mechanism appears to be hardly considered in most studies. Reactions leading to OH pairs may be written as H2O + O = (OH)2 where (OH)2 is the pair of OH groups. The equilibrium constant K2 of this reaction is a(OH)2 K2 = f H 2 O aO

(4)

(5)

Assuming that the concentration of OH is proportional to the activity of OH, this equation implies that water solubility should be directly proportional to the water fugacity: cwater ∼ fH 2 O

(6)

The same kind of relationship would be obtained if water were dissolved as molecular H2O. Finally, four protons may substitute for a tetravalent cation, such as Si4+ in the hydrogarnet defect (OH)4. This may be written as 2 H2O + 2 O = (OH)4

(7)

The equilibrium constant K3 of this reaction is K3 =

a(OH)4 fH22 O aO2

(8)

Thermodynamics of Water Solubility and Partitioning

197

This equation implies that water solubility should be proportional to the square of water fugacity: cwater ∼ fH22 O

(9)

Different substitution types—isolated OH groups, OH pairs and the hydrogarnet defect—can therefore be distinguished by differences in the dependence of water solubility on water fugacity. Equilibrium constants are related to the Gibbs free energy ∆G of the respective reaction by: −RTlnK = ∆G = ∆H1bar – T∆S1bar + ∆Vsolid (P-1 bar)

(10)

In this equation, the effect of pressure on the thermodynamic properties of the fluid is contained in the water fugacity. ∆H1bar is the reaction enthalpy at 1 bar, ∆S1bar is the reaction entropy at 1 bar and ∆Vsolid is the volume change of the solids. R is the gas constant, T is temperature and P is pressure. By inserting the expressions for the equilibrium constants (Eqns. 2, 5, or 8) in this equation and by assuming that (1) the activity of the dissolved OH species is proportional to the concentration of dissolved water and that (2) the activity of the unprotonated oxygen atoms remains essentially constant, one obtains: ⎛ ∆H 1bar + ∆V solid P ⎞ cwater = AfHn2 O exp ⎜⎜ − ⎟⎟ RT ⎝ ⎠

(11)

where A is a constant which essentially contains the entropy of reaction and n is an exponent related to the dissolution mechanism of OH: n = 0.5 for isolated OH groups, n = 1 for OH pairs (or molecular water), n = 2 for the hydrogarnet defect. Experimentally derived parameters for Equation (11) applied to the water solubility in a variety of nominally anhydrous minerals are compiled in Table 2. The significance of the term ∆Vsolid in Equation (11) requires some further discussion. The reactions leading to the dissolution of water in minerals were described in a somewhat Table 2. Thermodynamic models for water solubility in minerals.

n

∆Vsolid

∆H1bar

(ppm/barn)

(cm3/mol)

(kJ/mol)

0.0066 0.0147 * 0.54 0.0135 0.042 7.144 2.15 0.679 0.0004

1 1 1 1 0.5 0.5 0.5 0.5 0.5

10.6 10.2 10.0 12.1 11.3 8.02 7.43 5.71 4.0

— — 50 −4.56 −79.7 — — — —

A Mineral Olivine

MgSiO3 enstatite Aluminous enstatite# Jadeite Cr-diopside§ Pyrope Ferropericlase

Reference Kohlstedt et al. (1996) Mosenfelder et al. (2006) Zhao et al. (2004)** Mierdel and Keppler (2004) Mierdel (2006) Bromiley and Keppler (2004) Bromiley et al. (2004) Lu and Keppler (1997) Bolfan-Casanova et al. (2002)

Tabulated parameters refer to Equation (11). Where no value for ∆H1bar is given, the enthalpy term is missing because the temperature dependence of water solubility was not calibrated and the equations are strictly valid only at the temperatures they were calibrated. Notes: * Recalculated from a value of A = 2.45 H/106 Si/MPa which in the original publication is probably misprinted as A = 2.45 H/106 Si/GPa; ** The equation by Zhao et al. (2004) also includes a term exp(αxFa/RT), where a is 97 kJ/ mol and xFa is the molar fraction of fayalite.; # This equation gives the water solubility coupled to Al of an Al-saturated enstatite. In order to get the total water solubility in Al-saturated enstatite, the water solubility in pure MgSiO3 according to Mierdel and Keppler (2004) has to be added. § These data may reflect metastable equilibria.

198

Keppler & Bolfan-Casanova

simplified way by the Equations (1), (4), and (7). In reality, however, one has to consider that the conversion of an unprotonated oxygen atom to an OH group in the structure requires charge balancing by cation vacancies. Alternatively, this may also be described as a direct substitution of protons for cations. In pure enstatite, for example (Rauch and Keppler 2002) the main substitution mechanism is probably the direct replacement of Mg2+ by two protons. This may be written as H2O + MgSiO3 = H2SiO3 + MgO

(12)

where H2SiO3 is the hydrous component in enstatite. The volume change of the solids during reaction is then ∆V solid = VH 2 SiO3 + VMgO − VMgSiO3

(13)

The physical meaning of ∆Vsolid then depends on the behavior of the MgO formed during the reaction. If the MgO remains in solid solution in the pyroxene lattice during the reaction, then ∆Vsolid is simply the volume change of the pyroxene structure upon hydration. However, if the MgO diffuses out of the crystal, ∆Vsolid is the sum of the molar volume of MgO and the volume change of enstatite upon water dissolution. Interestingly, ∆Vsolid for enstatite and also for olivine (Table 2) is quite close to the molar volume of MgO (11.25 cm3/mol) and high-precision measurements of lattice constants of hydrous olivine (JR Smyth, pers. comm.) suggest that the incorporation of water causes a change of the unit cell volume comparable to ∆Vsolid derived from fitting solubility data to Equation (11). This may perhaps imply that the MgO produced by reaction does indeed remain dissolved in the crystal structure and that the volume change of the lattice is largely due to the dissolution of this component. It should be obvious from reactions such as the one shown in Equation (12), that water solubility will be sensitive to the activities of the oxide components involved. A substitution of protons for Mg2+ will proceed more easily at low MgO activity, while a substitution of protons for Si4+ will be favored by low silica activity. Eventually, it may even be possible to change the dissolution mechanism of water by changing the activities of MgO and SiO2 and there is evidence that this may be possible for olivine at relatively low pressures (Lemaire et al. 2004). However, the most important chemical parameter influencing water solubility, is probably the activity of components that allow for coupled heterovalent substitutions, such as Al3+ + H+ for Si4+ or Al3+ + H+ for 2 Mg2+. On the other hand, varying activities of different cations of the same valence, such as Mg2+ and Fe2+ should have a much smaller effect on water solubilities, if any. Although Equation (11), which describes water solubility in minerals may look simple, the actual dependence of water solubility on pressure and temperature can be surprisingly complex. The water fugacity term in Equation (11) suggests that water solubility should increase with increasing water pressure. However, the term ∆Vsolid, which is always positive, acts in the opposite direction. As a result, water solubility at a given temperature often first increases with pressure, reaches a maximum and then decreases at higher pressure. A physical explanation for this is that fluids are more compressible than solids and beyond a certain pressure the molar volume of water in the fluid may be smaller than the partial molar volume in the solid (e.g., Withers et al. 1998). A further increase in pressure then reduces water solubility. Similarly, the terms ∆H1bar and ∆VsolidP, which determine the temperature dependence at given pressure, may have opposite sign. At 1 bar, water solubility may decrease with temperature, while it increases with temperature at higher pressures.

Relationship between water solubility and partitioning If two minerals α and β are in equilibrium with an aqueous fluid of the same composition, the solubility of water in each mineral is given by Equation (11) with the parameters relevant

Thermodynamics of Water Solubility and Partitioning

199

for the respective mineral. The partition coefficient of water is then simply the ratio of the water solubilities in the two minerals:

α /β Dwater =

⎛ ∆H α1bar + ∆Vαsolid P ⎞ α cwater = Aα fHn2αO exp ⎜⎜ − ⎟⎟ RT ⎝ ⎠

(14)

⎛ ∆H β1bar + ∆Vβsolid P ⎞ n β cwater = Aβ fH 2βO exp ⎜ − ⎟ ⎜ ⎟ RT ⎝ ⎠

(15)

α ⎛ ( ∆H α1bar − ∆H β1bar ) + ( ∆Vαsolid − ∆Vβsolid )P ⎞ cwater Aα nα − nβ = f exp ⎜− ⎟ H O β ⎜ ⎟ Aβ 2 RT cwater ⎝ ⎠

(16)

However, it is very important to note that at given P and T, the partition coefficient of water will only be independent of water activity, if the exponent n is the same for both minerals, i.e., if water is dissolved in both minerals as isolated OH groups or as OH pairs or as the hydrogarnet defect. If, however, the dissolution mechanism of water in the two minerals is different, the partition coefficient will depend on water activity. For example, if in mineral α water is α /β dissolved as OH pairs, while in mineral β water is dissolved as isolated OH groups, Dwater will increase with the square root of water activity at constant pressure and temperature. A similar effect of water activity on partition coefficients may occur for mineral/melt partitioning.

EXPERIMENTAL STRATEGIES FOR MEASURING WATER SOLUBILITY AND WATER PARTITION COEFFICIENTS Annealing experiments Annealing experiments are probably the most simple procedure for measuring water solubility in a mineral. In a typical annealing experiment, a preexisting crystal (natural or synthetic) is exposed to an aqueous fluid under controlled conditions of pressure, temperature and possibly oxygen fugacity. After a run duration which is deemed sufficient to reach equilibrium the charge is quenched and the water content of the crystal is measured. Most of the early studies on water solubility in minerals were carried out in this way (e.g., Bai and Kohlstedt 1992, 1993; Kohlstedt et al. 1996; Lu and Keppler 1997). The main advantage of annealing experiments is the fact that one can easily obtain large and oriented single crystals. There are potential problems due to intense cracking of the original single crystal during compression or decompression of the experimental charges in the piston cylinder press or multi-anvil apparatus, but these problems can be solved by using low-friction assemblies (e.g., Bromiley et al. 2004). Reaching chemical equilibrium can be a major problem in annealing experiments. In most silicate minerals, the dissolution of water is probably coupled to cation vacancies. Unless the crystal already contains a high population of vacancies, which only need to be “decorated” by protons, this means that equilibrium water solubilities can only be achieved if cations are able to diffuse out of the crystal during the run. In particular for cations with very low diffusion coefficients such as Si4+, this is nearly impossible for mm-sized crystals during reasonable run durations. Accordingly, annealing experiments do not necessarily reach equilibrium, even if no evidence for diffusion profiles is seen in the samples. However, if overgrowth rims form during annealing experiments, the water contents observed in these rims may represent true equilibrium solubility. Mosenfelder et al. (2006) demonstrated that identical water contents and virtually identical infrared spectra are obtained for olivine of natural composition, no matter whether

200

Keppler & Bolfan-Casanova

single crystals were annealed in a hydrous fluid or whether crystals were grown under watersaturated conditions. On the other hand, the same authors noted that water concentrations obtained by hydrous annealing of pure forsterite single crystals are one order of magnitude lower than the water concentration in forsterite crystals grown under nearly identical conditions. This is probably due to the slower diffusion kinetics in forsterite, which contains less mobile defects than olivine. During annealing of a chromian diopside, Bromiley et al. (2004) observed clear evidence for disequilibrium. In particular, short run durations first yielded anomalously high water contents, which later decreased during the course of the experiment (Fig. 1). Probably, the initial incorporation of water in these and many other experiments is due to the local reduction of Fe3+ and the incorporation of H+ close to the Fe2+ produced by the reduction process. Upon longer annealing times, the defects can then migrate to a more stable environment. Whether a real stable equilibrium or only a metastable state is ultimately reached in these experiments probably depends on the available populations of cation vacancies and their diffusivities. In any case, the question of attainment of equilibrium needs to be carefully addressed in any study of water solubility in minerals by annealing experiments. Even if annealing experiments do not reequilibrate all defects, they can be useful for special applications. Mantle xenoliths that have lost most of their water during ascent may still contain defects (e.g., metal vacancies) that were related to the dissolution of water. It may then be possible to decorate these defects with hydrogen by annealing the sample in water under conditions where the diffusivity of the metal vacancies is low. In this way, it may ultimately be possible to reconstruct the original water content of a mantle mineral, even if it has lost all water during ascent to the surface (Kohlstedt and Mackwell 1998).

Crystallization experiments In crystallization experiments, crystals of the desired mineral are directly grown under water-saturated conditions. In these experiments, attainment of equilibrium should be much

Figure 1. Infrared spectra of natural chromian diopside annealed in water at 15 kbar and 1100 °C under Ni-NiO buffer conditions for different run durations. Bulk water contents are given in ppm H2O by weight. Initially, very high water contents are produced after short run durations, probably due to local reduction of Fe3+ to Fe2+. Diffusion of defects towards a more stable environment leads to a reduction of water contents after longer run durations. From Bromiley et al. (2004).

Thermodynamics of Water Solubility and Partitioning

201

less of a problem than with annealing experiments. However, often it is quite difficult to produce clear and inclusion-free crystals that are sufficiently large to be oriented for polarized FTIR measurements. One way to overcome this problem is to grow the crystals from a melt by slow cooling (e.g., Stalder 2002; Lemaire et al. 2004). However, the water content of such crystals will only represent true water solubility if it can be demonstrated that the melt was water-saturated, i.e., in equilibrium with a separate aqueous fluid phase throughout the entire experiment. Moreover, the method can only be applied to temperatures above the watersaturated solidus and if water solubility is a function of temperature, it is difficult to interpret the water solubility, as it is not precisely known to which temperature they correspond. An alternative approach to generate large single crystals of silicates at constant temperature was introduced by Mierdel and Keppler (2004). In these experiments, alternating layers of starting mixtures with different compositions (e.g., with silica excess and deficient in silica) are introduced into the sample capsule, so that the phase of interest can only nucleate at the boundary between the layers. With this method, it was possible to grow mm-sized water-saturated crystals of enstatite at relatively low temperatures. However, because these experiments necessarily involve a gradient in activities (e.g., silica activity), it is not possible to buffer all activities during the growth of the crystal.

WATER IN UPPER MANTLE MINERALS Water solubility in and the Al content of orthopyroxenes as “geohygrometer” Although orthopyroxene is less abundant in the upper mantle than olivine, it is probably the most important host of water, particularly in the shallow part of the upper mantle. Moreover, water solubility in orthopyroxene is particularly well understood and calibrated. Water solubility in pure MgSiO3 enstatite was measured by Rauch (2000), Rauch and Keppler (2002) and Mierdel and Keppler (2004). At 1100 °C, water solubility increases with pressure to a maximum of 867 ppm by weight at 75 kbar and then decreases at higher pressures to 714 ppm at 100 kbar in the stability field of high-clinoenstatite. While water solubility at room pressure decreases with temperature, it significantly increases with temperature at pressures above about 10 kbar (Fig. 2). The entire available data set can be described by one single equation of the type of Equation (11) with the fit parameters given in Table 2. Stalder and Skogby (2002) report a water content in an Al-free enstatite crystallized from a water-saturated melt at 25 kbar and 1400-1150 °C of 29.5 ppm H which corresponds to 280 ppm of water, in reasonable agreement with Figure 2. At much higher pressures, Inoue et al. (1995) and Yamada et al. (2004) report water contents between 0.3 and 0.8 wt% in high-clinoenstatite coexisting with a hydrous melt between 1200 and 1500 °C and 130-155 kbar. Since these water contents were measured by SIMS, it is not certain whether they are representative for the chemically dissolved water in the sample or whether some of the water may come from hydrous inclusions. The infrared spectra of synthetic hydrous enstatite crystals are simple (Fig. 3). They consist of a strong narrow band at 3363 cm−1 and a weak broader band at 3064 cm−1, which are both strongly polarized parallel c. In samples synthesized below 1000 °C, the strong band may be split into two components, possibly reflecting a very subtle change in the pyroxene structure. In addition, there are some very weak features at higher frequencies. All observations are consistent with a direct substitution of Mg2+ by two protons. Silica activity (buffering by excess SiO2 or by forsterite) appears to have very little influence both on water solubility and on the type of infrared spectra observed. Aluminum has a drastic effect on water solubility in enstatite (Rauch 2000; Rauch and Keppler 2002; Stalder and Skogby 2002; Mierdel 2006; Mierdel et al. in prep.). The water solubility in enstatite saturated with aluminum (i.e., in equilibrium with MgAl2O4 spinel or

202

Keppler & Bolfan-Casanova

Figure 2. Water solubility in pure MgSiO3 enstatite according to the thermodynamic model of Mierdel and Keppler (2004).

Figure 3. Infrared spectra of water-saturated pure MgSiO3 enstatite and of Al-doped enstatite, synthesized at 15 kbar and 1100 °C. After Rauch and Keppler (2002).

with pyrope) may be more than hundred times higher than the water solubility in Al-free enstatite under the same conditions (Mierdel 2006). In the presence of aluminum, new bands appear in the infrared spectrum of hydrous enstatite (Fig. 3), in addition to the bands seen in Al-free enstatite. At very high aluminum contents, the bands tend to broaden and to overlap, so that most of the fine structure is lost. In the Earth’s mantle, orthopyroxene (enstatite) coexists with olivine, clinopyroxene and an aluminum-rich phase (MgAl2O4-rich spinel in the spinel peridotite stability field and pyrope-rich garnet in the garnet peridotite field). Mierdel (2006) and Mierdel et al. (in prep.) calibrated the water solubility in enstatite coexisting with olivine and MgAl2O4 spinel or pyrope as a function of pressure and temperature. Water solubility in Al-saturated enstatite

Thermodynamics of Water Solubility and Partitioning

203

reaches values close to 1 wt% at 15 kbar and 800 °C. Water solubility decreases rapidly both with increasing pressure and temperature. A thermodynamic model calibrated by Mierdel (2006) and Mierdel et al. (in prep) describes water solubility in aluminum-saturated enstatite as the sum of two terms, one term being equal to the water solubility in Al-free enstatite and one term describing the water solubility related to the incorporation of aluminum. Both terms are described by equations of the type of Equation (11), with the respective fit parameters given in Table 2. A notable difference between the two equations used is in the exponent of the water fugacity. While the term for the water solubility in the absence of Al has an exponent of 1, implying incorporation of OH pairs charge balanced by a magnesium vacancy, the term describing the water solubility related to aluminum has an exponent of ½, implying the dissolution of water as isolated OH groups. Water solubilities in Al-saturated enstatite predicted by the thermodynamic model are shown in Figure 4.

Figure 4. Water solubility in Al-saturated enstatite as a function of pressure and temperature. The term “Alsaturated” refers to enstatite in equilibrium with forsterite and the stable aluminous phase, i.e., MgAl2O4 spinel or pyrope, depending on pressure. After the thermodynamic model by Mierdel (2006) and Mierdel et al. (in prep.).

204

Keppler & Bolfan-Casanova

Rauch (2000) and Rauch and Keppler (2002) originally assumed that aluminous pyroxenes dissolve water primarily as Mg2(AlH)SiO6 component, i.e., by a coupled substitution of tetrahedral Al3+ + H+ for Si4+. This would imply that the ratio of tetrahedral over octahedral Al in an orthopyroxene should increase with the water content, as in the absence of water, Al is dissolved as Tschermak component MgAlAlSiO6 with a ratio of tetrahedral to octahedral Al of 1:1. The ratio of tetrahedral over octahedral Al in an orthopyroxene could therefore be a potential “geohygrometer,” which may be a measure of the original water content of a pyroxene and therefore a measure of water fugacity in the mantle. Because of the much slower diffusion of Al as compared to water, this geohygrometer may still be applicable even if a mantle pyroxene has lost all of its water during ascent to the surface. However, Kohn et al. (2005) presented 27Al NMR evidence suggesting that water has little effect on the intracrystalline partitioning of Al in enstatite and that even water-rich aluminous pyroxenes still contain tetrahedral and octahedral Al in an approximate 1:1 ratio. This is also consistent with results by Stalder and Skogby (2002) and Stalder (2004) who suggested that a large fraction of Al in hydrous orthopyroxene is dissolved as Tschermak component. Mierdel (2006) carried out extensive electron microprobe and infrared analyses of aluminous enstatite crystals synthesized under water-saturated conditions at pressures between 15 and 35 kbar and at temperatures between 800 and 1100 °C. According to these data and single crystal structure refinements (Mierdel et al. in prep.; Smyth et al. in prep.), aluminum is incorporated in enstatite by three substitution mechanisms, corresponding to the following three end members: Mg Al Al Si O6

(Tschermak component)

Mg Mg (AlH) Si O6

(“Rauch” component; Rauch 2000)

H Al Si Si O6

(“Jadeite-like” component)

In most crystals analyzed, the Rauch component and the jadeite-like component appear to occur at a molar ratio of 1:1, implying an equal ratio of tetrahedral over octahedral aluminum, equivalent to a substitution by a “hydro-Tschermak-component” H2AlAlSiO6 This may perhaps be related to a minimization of distortions of the pyroxene structure, because the Rauch substitution would tend to increase the volume of the tetrahedral sites, while the jadeite-like substitution would reduce the volume of the octahedral sites. In two experiments, however, microprobe analyses suggest that virtually all Al is dissolved as Rauch component, as originally proposed by Rauch (2000) and Rauch and Keppler (2002). The reason for this is unclear. Interestingly, the aluminum contents of water-saturated enstatite crystals coexisting with MgAl2O4 spinel or pyrope are significantly higher than predicted by previous experimental studies and models of aluminum solubility in enstatite (MacGregor 1974; Dankwerth and Newton 1978; Perkins et al. 1981; Gasparik 2003), particularly at low temperatures and pressures, where water solubility is high (Fig. 5). Although in many previous studies on aluminum solubility in enstatite, water was added to the charge, these experiments were probably not water-saturated, because in the absence of special precautions, water is easily lost from the sample capsules in piston cylinder experiments (Truckenbrodt and Johannes 1999; Freda et al. 2001). Moreover, most of these previous experiments were carried out at high temperatures, where water solubility in the aluminous enstatite is relatively low. The available thermodynamic models of aluminum solubility in enstatite therefore probably only describe the Tschermak-type solubility in the absence of water. Since both the Rauch component and the jadeite-like component imply a molar 1:1 ratio of aluminum to hydrogen, one would therefore expect that the “excess” of aluminum found in the hydrous pyroxenes (compare Fig. 5) correlates with the water solubility coupled to aluminum on a molar 1:1 ratio. This was observed by Mierdel (2006): within error, the molar fraction of water dissolved in the aluminous enstatite minus the water present in the Al-free enstatite under the same

Thermodynamics of Water Solubility and Partitioning

205

Figure 5. Aluminum content of enstatite in equilibrium with olivine, MgAl2O4 spinel and water at 15 kbar after Mierdel (2006). For comparison, the Al-solubility in orthopyroxene after Gasparik (2003) is shown (dashed line). The model of Gasparik probably only reflects aluminum solubility at low water activities, where the Tschermak-substitution dominates. In the presence of water, coupled substitutions of Al and H are possible at low temperatures which lead to elevated aluminum solubilities that were not included in existing thermodynamic models of aluminum in orthopyroxene. The “excess” aluminum content, i.e., the difference between the two curves in the figure at a given temperature correlates in a molar 1:1 ratio with the water content in the pyroxene due to aluminum. Above 1100 °C, the aluminum solubility in watersaturated orthopyroxene approaches the curve for anhydrous conditions.

conditions equals the molar fraction of excess aluminum found in these crystals, where excess aluminum means the difference between the measured aluminum content and that predicted by the model of Gasparik (2003). In this sense, Al in orthopyroxene is indeed a geohygrometer. In principle, it should be possible to determine the original water content of an orthopyroxene from the mantle simply by looking at its aluminum content. This can be done if the pressure and temperature conditions of the formation of this orthopyroxene are known. If the aluminum content in the pyroxene is higher than predicted based on calibrations of the Tschermak-type solubility, this suggests that the crystal originally contained water. The molar fraction of the water coupled to Al simply equals the molar fraction of excess Al observed. The observed water content can then be compared to the water content predicted by Equation (11) with the fit parameters in Table 2 as a function of P, T and water fugacity. If P and T are known with sufficient accuracy, the water fugacity in the mantle source of the pyroxene can be calculated. Stalder (2004) and Stalder et al. (2005) showed that in addition to Al3+, other trivalent ions, particular Fe3+ and Cr3+ also enhance water solubility in enstatite. However, in natural mantle orthopyroxenes Cr3+ and Fe3+ are much less abundant than Al3+. Therefore, the effect of chromium and trivalent iron on water solubility in orthopyroxene is probably small. Similarly, some annealing experiments on natural orthopyroxenes reported by Rauch and Keppler (2002) suggest that limited substitution of Fe2+ for Mg2+ has little effect on water solubility. The main compositional control of water solubility in orthopyroxene certainly is aluminum activity, as first observed by Rauch (2000).

Water solubility in olivine Olivine is compositionally simpler than orthopyroxene, with the Mg/Fe ratio being the

206

Keppler & Bolfan-Casanova

only important variable. This compositional simplicity contrasts with the diversity of infrared bands seen in natural olivines from mantle xenoliths and other sources (e.g., Miller et al. 1987; Matsyuk and Langer 2004). The first experimental calibration of water solubility in olivine under mantle conditions was carried out by Kohlstedt et al. (1996). In this study, oriented single crystals of natural San Carlos olivine were annealed with excess water at 1100 °C and 25 to 130 kbar for run durations between 3 hours and more than 2 days. Oxygen fugacity was controlled by the Ni-NiO-buffer. At the beginning of the experiments, the olivine crystals were surrounded by a talc-brucite mixture, which reacted during the experiment to a mixture of forsterite and enstatite, i.e., silica activity in these experiments was buffered by the presence of enstatite. Water contents of run products were measured by FTIR using extinction coefficients after Paterson (1982). Water solubility was found to increase with pressure up to a maximum of 1510 ppm by weight at 12 GPa (Fig. 6; Table 2). The infrared spectra of the run product olivines were found to be dominated by strong bands at 3613 cm−1, 3598 cm−1 and 3579 cm−1. Since the experiments by Kohlstedt et al. (1996) were carried out by annealing natural crystals, it may be questionable whether all defects were really reequilibrated in these experiments and whether the water contents measured really reflect equilibrium water solubilities (e.g., Matveev et al. 2001). This was tested by Mosenfelder et al. (2006). They carried out experiments with starting materials made of San Carlos olivine either as single crystals or as fine-grained powders. While in the first type of experiment, water diffuses into the crystals as in the study by Kohlstedt et al. (1996), in the second type of experiment large crystals grow at the expense of the fine-grained starting material and incorporate water during growth. Both types of experiments were found to yield similar water contents and similar infrared spectra at otherwise identical conditions (Fig. 7). The infrared spectra are also similar to those reported by Kohlstedt et al (1996). However, Mosenfelder et al. (2006) calculated water contents by applying a newly calibrated infrared extinction coefficient for water in

Figure 6. Water solubility in olivine. Shown are experimental results on San Carlos olivine by Kohlstedt et al. (1996; open circles) and by Mosenfelder et al. (2006; black squares). Both studies were mostly carried out at 1100 °C, but the data by Mosenfelder et al. (2006) include some experiments at 1000-1300 °C. Note that the difference between the two studies mostly results from the use of different infrared extinction coefficients for quantifying the water content of the run products. Fit parameters of the solubility laws resulting from the two studies are given in Table 2.

Thermodynamics of Water Solubility and Partitioning

207

Figure 7. Infrared spectra (polarized parallel a) of water-saturated olivine from high-pressure experiments Numbers in brackets are water fugacities in GPa. From Mosenfelder et al. (2006).

olivine (Bell et al. 2003). This extinction coefficient generally yields water contents by a factor of 2-4 higher than those derived from the calibration of Paterson (1982) and therefore, the water solubilities in olivine obtained by Mosenfelder et al. are higher than those by Kohlstedt et al (Fig. 6). However, these differences are mostly due to the different extinction coefficients used. In all other aspects, both studies agree quite well; for example, they yield nearly the same value for ∆Vsolid, the volume change of the solids during the dissolution of water (Table 2). Moreover, Mosenfelder et al. (2006) were also able to show that the infrared spectra of olivines hydrated at high pressure are similar, dominated by high-frequency bands around 3600 cm−1, irrespective of the composition of the starting material (Fig. 8). The infrared spectra of the experimental samples are quite similar to those of some natural olivines from garnet peridotites (Fig. 9). Smyth et al (J. Smyth, pers. comm.) recently studied water solubility in pure forsterite at 120 kbar and 1100 to 1600 °C. They also found that the infrared spectra of their samples were dominated by high-frequency bands around 3600 cm−1, as in the studies by Kohlstedt et al. (1996) and Mosenfelder et al. (2006). The type of infrared spectrum observed by Smyth et al. appears to be nearly independent of silica activity; spectra of forsterite coexisting with melt and enstatite are similar to the spectra of forsterite coexisting with melt and clinohumite. According to single crystal structure refinements by Smyth et al (J. Smyth, pers. comm.) the dominant dissolution mechanism of water in olivine appears to be the replacement of Mg2+ by 2 H+, as suggested by the direct observation of cation vacancies on the M1 and M2 sites of Mg. This is entirely consistent with the exponent of 1 in the water fugacity term of Equation (11) (Table 2) describing water solubility in olivine as found by Kohlstedt et al (1996) and Mosenfelder et al. (2006). Moreover, the fact that Smyth et al. observe infrared spectra in chemically pure hydrous forsterite that are similar to those observed by Kohlstedt et al. (1996) and Mosenfelder et al. (2006) demonstrates that the dominant mechanism for the dissolution of water in olivine cannot be related to minor chemical impurities. Zhao et al. (2004) observed at relatively low pressures (3 kbar) that water solubility in olivine increases significantly with temperature and also with Fe/Mg ratio (Table 2). Smyth et al. (J Smyth, pers. comm.) also observed that water solubility in pure forsterite approximately doubles from 1100 °C to 1250 °C.

208

Keppler & Bolfan-Casanova

Figure 8. Infrared spectra (polarized parallel a) of water-saturated olivine grown from three different starting materials. Clearly, the main absorption bands seen in these samples are independent of the starting material. From Mosenfelder et al. (2006).

Figure 9. Polarized infrared spectra of a water-saturated olivine from an experiment at 8 GPa and 1150 °C (black) and from a natural crystal of olivine from a garnet peridotite (Bull peak diatreme, Arizona). The similarity of the spectra demonstrates that the experiments faithfully reproduce the defects found in natural olivines from the garnet peridotite field. From Mosenfelder et al. (2006).

Bai and Kohlstedt (1993) already noted that olivines may show infrared bands which behave differently in response to changes in the chemical environment. They distinguished “group I bands” above 3450 cm−1 from “group II bands” below 3450 cm−1. The relative intensity of the group II bands was very much enhanced at oxidizing conditions, while they pretty much disappeared under reducing conditions.

Thermodynamics of Water Solubility and Partitioning

209

Matveev et al. (2001) studied the water solubility in olivine buffered by orthopyroxene (high silica activity) or magnesiowüstite (low silica activity). The experiments were carried out at 1300 °C, Re-ReO2 buffer conditions and mostly at 20 kbar. They observed that silica had a major effect on the infrared spectra observed in the run products. While the olivines in equilibrium with magnesiowüstite (low silica activity) showed infrared bands at high frequencies around 3600 cm−1 (group I of Bai and Kohlstedt), these bands disappeared in the runs buffered by orthopyroxene (high silica activity) while new bands appeared at low frequency in the range between 3400 and 3200 cm−1 (group II of Bai and Kohlstedt). Lemaire et al. (2004) observed a similar effect of silica activity on the infrared spectra of pure hydrous forsterite grown at 20 kbar and about 1500 °C. However, in their experiments, the bands at high frequency (group I) never disappeared completely. From their study, Matveev et al. (2001) concluded: (1) The ratio of group I over group II bands is an indicator of silica activity. (2) Group I bands are due to tetrahedral Si vacancies, because they are enhanced at low silica activity. (3) Group II bands are due to octahedral Mg vacancies, because they are enhanced at high silica / low MgO activity. (4) Olivines from natural mantle xenoliths, which show the high-frequency group I bands are out of equilibrium with the surrounding mantle, because they appear to reflect a very low silica activity environment. This may reflect metasomatism by a very silica-poor medium, such as a carbonatite melt. The work by Matveev et al (2001) has very much stimulated research into the significance of individual bands in the infrared spectrum of olivine. The dependence of the intensity ratio of group I over group II bands on silica activity is certainly real, at least at the low pressures (20 kbar) and high oxygen fugacities (Re-ReO2-buffer) studied by Matveev et al. However, there are major problems with some of the interpretations by these authors: (1) The assignment of group I and group II bands to silicon and magnesium vacancies does not explain why the intensity of the group II bands is greatly enhanced at high oxygen fugacity and why they tend to disappear at low oxygen fugacity (Bai and Kohlstedt 1993). (2) The studies by Kohlstedt et al (1996), Mosenfelder (2006) and by Smyth et al. (J. Smyth, pers. comm.) all show that at high pressures above 20 kbar and under reasonable conditions of oxygen fugacity, hydrous olivines always exclusively show the high-frequency group I bands, independent of silica activity. Moreover, single crystal structure refinements suggest that these bands are primarily related to Mg2+ vacancies (J. Smyth, pers. comm.). In order to reconcile all of the observations outlined above, we suggest the following model: (1) Silicon vacancies are not important for dissolving water in olivine. This is consistent with experimental evidence that the hydrogarnet substitution in pyrope becomes unfavorable at high pressure, because of the large volume expansion of the tetrahedral site associated with this kind of defect (Withers et al. 1998). (2) Throughout most of the mantel, in particular at high pressure and under reasonably reducing conditions, the main dissolution mechanism for water in olivine is the substitution of Mg2+ by proton pairs. This substitution mechanism is responsible for the prominent group I bands around 3600 cm−1. (3) Under conditions of high silica activity (which necessarily means low MgO activity)

210

Keppler & Bolfan-Casanova and only at relatively low pressures, additional substitution mechanisms become important which lead to group II bands. At low MgO activity, the population of Mg vacancies will be particularly high. This charge deficiency on the Mg site may be compensated in the crystal by the oxidation of Fe2+ to Fe3+ and by simultaneous protonation. The group II bands may therefore be due to protonated vacancies close to an Fe3+ site. Therefore, it is not surprising that the intensity of group II bands, particularly those around 3300 cm−1 (Matveev et al. 2001; Bai and Kohlstedt 1993) increases both with oxygen fugacity and with silica activity. In the absence of iron or other trivalent ions, low MgO activity may lead to associated defects (clusters of Mg vacancies). The broad bands seen at about 3150-3200 cm−1 in pure hydrous forsterite (Lemaire et al. 2004) are probably related to such associated defects. This is consistent with the observation by Berry et al. (2005) that these bands disappear completely when titanium is added to the system. If Ti4+ enters the octahedral site of olivine, it can charge compensate not only the Mg2+ vacancy which it occupies, but also a neighboring vacancy. Therefore, no hydrogen can be stored in such vacancy pairs anymore once a suitable tetravalent cation is introduced into the system.

When comparing infrared spectra of olivines from high-pressure experiments and natural samples, the following things should be kept in mind: (1) Most natural olivines have probably lost a large fraction of their initial water content during ascent (Ingrin and Skogby 2000). Probably, not all species of water will diffuse out of the crystal with the same rate. Water coupled to some cations with low diffusivity (such as Ti4+) as well as water in planar defects may be particularly immobile. This may explain the prominence and great diversity of bands related to such species in the spectra of natural samples. For example, Berry et al. (2005) showed that the 3572 cm−1 band as well as the 3525 cm−1 band seen in many natural olivines are probably related to a Ti4+-bearing defect cluster. However, the low abundance of Ti4+ in natural olivine makes it unlikely that this defect would contribute significantly to the bulk solubility of water in olivine at high water fugacities. (2) Water loss during ascent of a xenolith is usually accompanied by oxidation, i.e., “water” is not lost as H2O, but as H2, produced by the simultaneous oxidation of Fe2+ to Fe3+(Ingrin and Skogby 2000). As the intensity ratio of group I to group II bands is sensitive to oxidation state as well as to silica activity, one should be cautious in interpreting this intensity ratio as observed in natural samples in terms of silica activity alone (Matveev et al. 2005). Moreover, as olivines synthesized at high pressures only show group I bands even in equilibrium with enstatite, the predominance of these bands in natural samples does not imply that these samples are out of equilibrium with orthopyroxene. (3) Experiments carried out under quite oxidizing conditions, such as the Re-ReO2 buffer may generate enhanced intensities of group II bands that are not realistic for any natural sample. (4) Differences in band positions by a few cm−1 relate to extremely subtle changes in chemical environment (essentially oxygen-oxygen distances), which should not be interpreted in terms of radically different substitution mechanisms. From the discussion above, it is obvious that many details of the dissolution mechanism of water in olivine still need further investigation. The most serious problem, however, is related to the very different water solubilities in olivine obtained by applying either the Paterson (1982) or the Bell et al. (2003) calibrations of infrared extinction coefficients to the same samples. This difference has serious geological consequences as will be pointed out below. Further independent calibrations of infrared extinction coefficients and water contents in olivine would therefore be highly desirable.

Thermodynamics of Water Solubility and Partitioning

211

Water solubility in garnet Garnets have attracted a lot of attention as potential host of water because of the natural occurrence of katoite or “hydrogarnet” Ca3Al2(H4O4)3. Katoite is a mineral with garnet structure, which can be considered as a grossular Ca3Al2(SiO4)3 with each of the Si4+ cations replaced by four protons. The structure of katoite is well studied. The four protons are located outside of the tetrahedral site, above the edges of the tetrahedron (Lager et al. 2005 and references therein). The infrared spectrum of hydrogarnet shows essentially one broad band at about 3600 cm−1 with a shoulder at 3660 cm−1 (Rossman and Aines 1991). The first experimental evidence that the hydrogarnet defect may occur in pyrope-rich garnets relevant to the upper mantle was provided by Ackermann et al. (1983). They observed an absorption band close to 3600 cm−1 in the infrared spectrum of pyrope synthesized in the presence of excess water at 25 kbar and 1000 °C. A systematic study of water solubility in pyrope at 1000 °C and up to 130 kbar was carried out by Withers et al. (1998). They synthesized pyrope coexisting with excess SiO2 from oxide mixes and observed the typical broad hydrogarnet band at 3600 cm−1 in the run products. Water solubility in garnet was found to increase with pressure up to about 1000 ppm at 40 kbar and then to decrease to nearly zero at pressures above 80 kbar. The decrease of water solubility at high pressures is probably due to the large volume expansion of the solids upon dissolution of water. Lu and Keppler (1997) studied water solubility in pyrope-rich garnet at 1000 °C and up to 100 kbar by annealing a natural garnet sample from Dora Maira in water. Water solubility in this garnet was found to increase continuously to about 200 ppm at 100 kbar. The parameters obtained from a fit of the solubility data to an equation similar to Equation (11) suggest the dissolution of water as isolated OH groups in this sample (Table 2; note exponent ½ of water fugacity term). This is consistent with the infrared spectrum of the sample which is quite different from that observed by Withers et al (1998) for pure pyrope. In addition to a rather sharp band at 3600 cm−1, there are three more bands around 3650 cm−1. Possibly, the dissolution of water in this sample is coupled to chemical impurities. Although there is still a need for more systematic studies of water solubility in garnets particularly as a function of bulk composition, the available data suggest that garnet is probably not an important host for water in the upper mantle, both because of its limited storage capacity for water and its low modal abundance.

Water solubility in clinopyroxene Clinopyroxenes, particularly omphacites are among the most water-rich samples from mantle xenoliths (Skogby 2006). Due to its low modal abundance, clinopyroxene probably does not contribute very much to the bulk water storage capacity of the upper mantle. However, omphacites are likely to be important carriers for transporting water down into the mantle in subducted slabs. Bromiley and Keppler (2004) investigated water solubility in pure jadeite at 600 °C and pressures up to 100 kbar. Infrared spectra of the synthetic hydrous jadeites show two prominent sharp peaks at 3373 cm−1 and 3613 cm−1 together with several weaker features. Band positions and polarizations suggest a water dissolution mechanism involving vacancies on the M2 site, consistent with observations on natural omphacites by Smyth et al. (1991), although the infrared spectra of natural omphacites are quite different from the synthetic pure jadeites (Skogby 2006 and references therein). Water solubility reaches a maximum of about 450 ppm by weight at 20 kbar and slowly decreases with increasing pressure to about 100 ppm at 100 kbar. A fit of the experimental data to an equation similar to Equation (11) suggest that water solubility increases with the square root of water fugacity, consistent with the incorporation of hydrogen as isolated OH groups (Table 2).

212

Keppler & Bolfan-Casanova

The most important result of the study by Bromiley and Keppler (2004) is probably that chemical composition has a major effect on water solubility in clinopyroxene. In particular, water solubility is strongly enhanced by the presence of some Ca-Eskola component Ca0.5 0.5AlSi2O6. However, a systematic study of water solubility in omphacite throughout the compositional space of natural samples is still lacking. Considering the likely importance of omphacite for recycling water back into the mantle, this may be the most important unsolved problem in calibrating the water solubility in the nominally anhydrous minerals of the upper mantle. Bromiley et al. (2004) carried out some hydrous annealing experiments of a natural chromian diopside at 1000-1100 °C and 5 to 40 kbar. Water solubility was found to increase with pressure to 229 ppm at 40 kbar, but there is evidence that the solubility law derived from this study (Table 2) reflects some metastable equilibrium dictated by the defect populations inherited in the starting material. An interesting observation of the study by Bromiley et al. was, however, that the ratio of the two bands at 3646 cm−1 and at 3434 cm−1 is independent of water fugacity, but dependent on oxygen fugacity (Fig. 10). Since water loss from a xenolith during ascent usually involves oxidation of Fe2+ to Fe3+, the ratio of these bands therefore could potentially indicate whether a mantle clinopyroxene still preserves its original water content or not.

Water partitioning among upper mantle minerals Olivine and orthopyroxene are the two main constituents of the upper mantle. The partition coefficient of water between these two phases at water saturation is simply the ratio of the water solubilities in the two phases. Figure 11 shows the calculated partition coefficient for water between olivine and aluminum-saturated orthopyroxene as a function of pressure and temperature. These partition coefficients are based on the solubility law of Kohlstedt et al. (1996) for olivine without any correction for the effect of temperature and on the solubility

Figure 10. Averaged polarized infrared spectra of chromian diopside annealed in water at 15 kbar and 1100°C under different oxygen fugacities, corresponding to the Ni-NiO and the Fe-FeO-buffer. Clearly, the band at 3434 cm−1 is very much enhanced under reducing conditions. The relative intensity of this band can therefore be used as a measure of the oxygen fugacity recorded by the sample. Since the loss of water upon ascent of a xenolith involves oxidation of Fe2+ to Fe3+, the relative intensities of the infrared bands could also be used as an indicator of water loss from a mantle clinopyroxene. From Bromiley et al. (2004).

Thermodynamics of Water Solubility and Partitioning

213

Figure 11. Calculated partition coefficients of water between aluminous orthopyroxene and olivine in equilibrium with an aluminous phase (spinel or garnet, depending on pressure) and under the conditions of water saturation. The calculation is based on the model for water solubility in aluminous orthopyroxene of Mierdel (2006; see Fig. 4) and on the water solubility in olivine as measured by Kohlstedt et al. (1996). After Mierdel (2006).

law of Mierdel (2006) and Mierdel et al. (in prep.) for aluminum-saturated orthopyroxene, which explicitly includes the effect of temperature (see Table 2 for parameters). While water partitions slightly more into olivine at pressures above 90 kbar, it strongly partitions into the aluminous orthopyroxene at lower pressures. Partition coefficients along a typical oceanic and continental geotherm are shown in Figure 12. In the uppermost mantle, water partitions very strongly into the aluminous pyroxenes, particularly along the relatively cold continental geotherm. This effect will be even enhanced if the mantle is far below water saturation, as one would expect. While the solubility of water in olivine is proportional to water fugacity, the part of the solubility of water in pyroxenes which is coupled to Al only increases with the square root of water fugacity. Therefore, at a water activity of 0.1, water solubility in olivine would be reduced by a factor of 10, while water solubility in aluminous orthopyroxene would only be reduced by a factor of about 3. Therefore, water would partition into the aluminous orthopyroxene even three times stronger than at water saturation. In the light of these data, it should not be surprising that pyroxenes from mantle xenoliths often contain much more water than the coexisting olivines. The calculated partition coefficients in Figure 11 agree quite well with the measured olivine/orthopyroxene partition coefficient of 0.11 ± 0.01 (4 measurements) by Aubaud et al. (2004), measured at 10-20 kbar and 1230-1380 °C.

214

Keppler & Bolfan-Casanova

Figure 12. Calculated partition coefficient of water between olivine and aluminous orthopyroxene in equilibrium with spinel or garnet under conditions of water saturation. Shown is the partition coefficient as a function of depth for a continental and an oceanic geotherm. After Mierdel (2006).

If olivine/orthopyroxene partition coefficients were calculated based on the calibration of water solubility in olivine by Mosenfelder et al. (2006), they would increase about three times. This would not, however, change the fact that in the uppermost mantle, most of the water partitions into the aluminous orthopyroxene.

Water storage capacity of the upper mantle and the origin of the Earth´s asthenosphere Figure 13 shows the water solubility in a model mantle consisting of 60% olivine and 40% aluminum-saturated orthopyroxene as a function of depth for a continental geotherm (Mierdel 2006; Mierdel et al. in prep.). Water solubility in mantle minerals has a pronounced minimum at a depth interval, which coincides with the location of the seismic low-velocity zone. The minimum is due to the sharp decrease of water solubility in aluminum-saturated orthopyroxene with depth, while the water solubility in olivine continuously increases. Figure 13 suggests that in the seismic low-velocity zone, about 1000 ppm of water would be required for a free aqueous fluid to coexist with water-saturated mantle minerals. However, since the temperatures at these depths are already above the water-saturated solidus, the presence of excess water will trigger the formation of a hydrous melt. In this melt, water activity will be significantly reduced, probably to values between 0.1 to 0.3 (Mierdel et al. in prep.). At these reduced water activities, however, the solubility of water in olivine and pyroxene will also be reduced. Therefore, Figure 13 suggests that in depth interval of the seismic low-velocity zone, a few hundred ppm of water are probably sufficient to trigger the formation of a hydrous melt. Partial melting in the low-velocity zone is therefore not related to volatile enrichment, but to a minimum in the solubility of water in mantle minerals. As the water solubility in minerals below and above the low-velocity zone increases, water increasingly partitions into the solid minerals and the melt solidifies. The sharp increase in water solubility above the low-velocity

Thermodynamics of Water Solubility and Partitioning

215

Figure 13. Water solubility along a continental geotherm in a model mantle consisting of a 60:40 mixture of olivine and Al-saturated enstatite. The depth interval of the seismic low velocity zone below continents is shaded. The low velocity zone corresponds to a minimum in water solubility in minerals, suggesting that the excess water forms a partial melt. Water solubility in olivine is after Kohlstedt et al. (1996), water solubility in aluminous enstatite after Mierdel (2006). Diagram after Mierdel (2006) and Mierdel et al. (in prep.).

zone therefore explains why this upper boundary is usually well resolved in seismic data, while on the other hand the gradual increase in water solubility below the low-velocity zone corresponds to the diffuse nature of this seismic boundary. For a hotter, oceanic geotherm, the upper boundary of the low-velocity zone is lifted upwards because of the decrease of water solubility in aluminous orthopyroxene with temperature, again consistent with seismic observations (Mierdel et al. in prep.). The seismic low-velocity zone is often identified with the asthenosphere, the ductile layer below the lithosphere, which allows the sliding of lithospheric plates required by plate tectonics. The model outlined here suggests that plate tectonics can only exist in a planet with a water-bearing mantle, because without water, no asthenosphere would exist.

Water recycling by subducted slabs Water recycling into the mantle was already discussed by Ito et al. (1983). Rüpke et al. (2004) recently estimated that most of the water presently residing inside the mantle was actually recycled by subduction. Moreover, they estimated that the global sea level dropped by several hundred meters due to subduction of water since the Cambrian. However, all of these estimates are based on the assumption of widespread serpentinization of the suboceanic mantle. Whether such a serpentinization does indeed occur is uncertain. It may very well be that serpentinization of the suboceanic mantle is only a local phenomenon and therefore, serpentine is not a significant carrier for recycling of water into the mantle. Moreover, even if serpentine is present initially, it will decompose unless the temperature inside the subducted slab remains quite low as during fast subduction of old slabs.

216

Keppler & Bolfan-Casanova

The potential importance of nominally anhydrous minerals in transporting water back into the mantle was probably first pointed out by Lu and Keppler (1997). Even if all hydrous minerals become unstable in the subducted slab at relatively shallow depth, not all water will be lost because the nominally anhydrous minerals will absorb at least part of the water released. According to Lu and Keppler (1997), the amount of water recycled by subducted oceanic crust can be estimated by multiplying the total length of subduction zones on Earth (presently 42 000 km) with the average speed of subduction (presently 5 cm/year) the thickness of the oceanic crust (about 6 km) with the density of the oceanic crust (about 3 g/cm3) and its water content. For a water content of 1000 ppm, this yields the subduction of 3.78 × 1016 tons or 2.7% of the total ocean mass (1.4 × 1018 tons) over one billion years. If this loss of water were not balanced by outgasing, it would translate into a reduction of global sea level by 100 meters. The basaltic layer of the subducted oceanic crust consists of eclogite, with pyrope-rich garnet and omphacite being the main constituent minerals. While garnet is unlikely to carry major amounts of water, natural omphacite samples are among the most water-rich nominally anhydrous minerals. A water content of 1000 ppm by weight in the subducted slab would be consistent with observed water contents in natural omphacites of several thousand ppm (Skogby 2006). The precise amount of water subducted in the oceanic crust will depend on the trajectory of the subducted slab in a pressure temperature diagram, which determines the depth of decomposition of hydrous minerals. The water solubility in nominally anhydrous minerals, particularly omphacite, will then determine how much water can be carried down into the mantle. Unfortunately, as noted above, the water solubility in omphacite as a function of pressure, temperature and bulk composition is not yet appropriately calibrated. It is likely, however, that water recycling will vary significantly with the trajectory of the subducted slabs in pressure temperature space, which in turn depends on such parameters such as the age of the slab. It is therefore conceivable that these parameters ultimately contribute to slow variations in the volume of the hydrosphere. In the considerations above, it was assumed that only the oceanic crust itself would contribute to water recycling into the mantle. However, advective flow of the mantle inside the mantle wedge above the subducted slab may contribute significantly to water recycling, as water released by the subducted slab will hydrate orthopyroxene and olivine in the mantle immediately above the slab. As shown in Figure 13, the water solubility in a bulk mantle consisting of 60% olivine and 40% orthopyroxene will never be much below 1000 ppm. If one assumes that for example, a 60 km thick layer of mantle above the subducted slab were dragged down together with the slab, the amount of water subducted may increase by one order of magnitude as compared to the value calculated above. Clearly, any kind of accurate estimate of water recycling over geologic time needs to integrate mantle convection models, thermal models of subduction zones, thermodynamic models of the stability of hydrous phases with models of water solubility in minerals such as omphacite. The development of such an integrated model of the Earth’s internal water cycle is certainly a major challenge for the coming decade.

WATER IN TRANSITION ZONE MINERALS Water solubility in wadsleyite and water partitioning between wadsleyite and olivine Wadsleyite has an unusual structure with a Si2O7 group and an oxygen atom that is not attached to a silicate tetrahedon. Smyth (1987) first suggested that this O1 oxygen atom is strongly underbonded and may therefore be a suitable site for protonation. Full protonation of this oxygen atom, charge compensated by magnesium vacancies, would lead to a water content of 3.3 wt% and would therefore make wadsleyite a major storage site for water in the Earth’s interior.

Thermodynamics of Water Solubility and Partitioning

217

Infrared spectroscopic evidence for water in synthetic wadsleyite was first detected by McMillan et al. (1991; about 0.06 wt%) and by Young et al. (1993; 0.06-0.39 wt%). The samples in these early studies, however, were probably not water-saturated. Much higher water contents were later reported by Inoue et al. (1995; 3.1 wt% measured by SIMS), by Kohlstedt et al (1996; up to 2.4 wt% measured by FTIR) and by others. Jacobsen et al. (2005) reported the first polarized FTIR spectra of hydrous wadsleyite (Fig. 14). They appear to be completely consistent with a protonation of the O1 site only, as originally proposed by Smyth (1987), while Kohn et al. (2002) had suggested some disordering of protons over different oxygen atoms. In virtually all studies of hydrous wadsleyite, high water contents were found to correlate with reduced Mg/Si ratios, implying charge balancing of two protons by an Mg2+-vacancy. Under the pressure and temperature conditions of the transition zone, the meaning of the term water solubility has to be carefully considered. Under these conditions, hydrous fluids and silicate melts are likely to be completely miscible and at realistic transition zone temperatures, solid silicates will never coexist with a water-rich fluid, but with a hydrous silicate melt. Therefore, water solubility can only be defined as the equilibrium water content of a mineral coexisting with a hydrous melt in a phase assemblage that buffers the compositions of all coexisting phases in such a way that the composition of each phase only depends on pressure and temperature. Demouchy et al. (2005) reported the water contents in wadsleyite coexisting with hydrous melt and clinoenstatite in the system MgO-SiO2-H2O at 130-180 kbar and 9001400 °C. In this system, the composition of all phases is buffered, i.e., increasing the amount of water in the charge would only increase the melt fraction, without changing the composition or water content of any of the phases. Therefore, the experimental results from this study can be considered to represent true water solubilities. Figure 15 shows the dependence of water solubility in wadsleyite on temperature. Clearly, water solubility is nearly constant between 900 and 1100 °C, but then drops drastically at higher temperatures. This effect is entirely due to the decrease of water activity in the melt with increasing temperature, because the partition

Figure 14. Polarized infrared spectrum of hydrous wadsleyite with 3200 ppm water. From Jacobsen et al. (2005).

218

Keppler & Bolfan-Casanova

Figure 15. SIMS derived water contents as a function of temperature for pure Mg2SiO4 wadsleyite (Inoue et al. 1995: open squares; Demouchy et al. 2005: open diamonds); (Mg,Fe)2SiO4 wadsleyite in the KLB-1 system (Kawamoto et al. 1996: open circles); and pure Mg2SiO4 ringwoodite (Ohtani et al. 2000: filled circles). The decrease in water solubility above 1200 °C is due to a reduced water activity in the melt. Curves are a guide for the eye only. wadsleyite / melt coefficient of water between wadsleyite and melt remains nearly unchanged (Dwater = 0.08). On the other hand, pressure was found to have a negligible effect on water solubility in wadsleyite. A similar effect of temperature on water solubility in wadsleyite was observed by Bolfan-Casanova (2005) and Hirschmann et al. (2005) based on a compilation of published analyses from various previous studies (Gasparik 1993; Inoue et al. 1995; Kawamoto et al. 1996; Kohlstedt et al. 1996; Smyth and Kawamoto 1997; Smyth et al. 1997; Chen et al. 2002; Kohn et al. 2002; Litasov and Ohtani 2003; Jacobsen et al. 2005; Demouchy et al. 2005; see Fig. 15). Bolfan-Casanova (2005) and Hirschmann et al. (2005) also noted that water solubility in ironbearing (Mg,Fe)2SiO4 wadsleyite may be somewhat higher than in pure Mg2SiO4 wadsleyite.

The value of the partition coefficient of water between wadsleyite and olivine primarily depends on the choice of the infrared extinction coefficient used to quantify water contents in olivine. As noted above, the experimental results on water solubility in olivine by Kohlstedt et al. (1996) and by Mosenfelder et al. (2006) are in many aspects comparable, but they yield water contents differing by a factor of about 3, because the former study used extinction coefficients by Paterson (1982) while the latter used the recently determined extinction coefficients by Bell et al. (2003). Water solubility in wadsleyite, on the other hand, is much higher and therefore, it has been quantified by several methods such as SIMS that are independent of infrared measurements. At 120 kbar and 1100 °C, Kohlstedt et al. (1996) report 1510 ppm of water in olivine, while Mosenfelder et al. (2006) find 6399 ppm using the new infrared extinction coefficients. This compares to 2.4 wt% of water in pure Mg2SiO4 wadsleyite measured by Demouchy et al. (2005) at 150 kbar and 1100 °C. Since water solubility in wadsleyite appears wadsleyite / olivine to be rather insensitive to pressure, these numbers give at 1100 °C Dwater of 16 and 3.8, respectively. Since both in wadsleyite and in olivine, proton pairs substitute for Mg2+, water solubility in both phases should be directly proportional to water fugacity. Accordingly,

Thermodynamics of Water Solubility and Partitioning

219

this partition coefficient should remain constant even if the water activity is reduced at higher temperatures due to a reduced water content of the residual melt. However, the partition coefficient may be further reduced if the water solubility in olivine intrinsically increases with temperature even at constant water fugacity, as suggested by Zhao et al (2004) and Smyth et al (Smyth, pers. comm.), while the intrinsic water solubility in wadsleyite is independent of temperature (Demouchy et al. 2005). Direct SIMS measurements of water contents in wadsleyite / olivine coexisting olivine and wasleyite by Chen et al (2002) yielded Dwater = 5. The preferential partitioning of water into wadsleyite implies that with increasing water fugacity, the stability field of wadsleyite should grow at the expense of the stability field of olivine, consistent with experimental observations (Chen et al. 2002; Smyth and Frost 2002). The 410 km discontinuity between the upper mantle and the transition zone should therefore be lifted upwards in regions of high water content. In addition, water may broaden the width of the seismic discontinuity (Wood 1995; Smyth and Frost 2002). Bercovici and Karato (2003) proposed an intriguing model of the chemical evolution of the Earth’s mantle based on the different solubility of water in wadsleyite and olivine. They suggested that material upwelling from the transition zone may contain that much water that it cannot be completely accommodated in upper mantle minerals, particularly olivine. Therefore, a partial melt will form on top of the transition zone. This partial melt may extract most incompatible elements out of the upwelling mantle peridotite and may recycle it back into the transition zone or lower mantle. This model is very attractive, because it could reconcile a chemically stratified mantle with whole-mantle convection. The essential requirement of this model is that the water content present in the upper transition zone cannot be dissolved in olivine and the other upper mantle minerals. In the light of the discussion above, this again ultimately depends on the value of the infrared extinction coefficient of water in olivine. If the calibration by Paterson (1982) is accepted, the resulting water solubilities in olivine (Kohlstedt et al. 1996) are sufficiently low to make the Bercovici-Karato model quite plausible. On the other hand, the infrared extinction coefficient by Bell et al. (2003) yields such high values for the water solubility in olivine in the deepest part of the upper mantle (Mosenfelder et al. 2006) that the Bercovici-Karato model would require implausibly high water contents in the transition zone in order to work.

Partitioning of water between wadsleyite and ringwoodite Mg2SiO4 ringwoodite has the spinel structure. Since ordinary spinels are generally among the most water-poor minerals, it was believed for a long time that the water solubility in ringwoodite should be very low. This changed with the work by Kohlstedt et al. (1996) who observed water contents in (Mg,Fe)2SiO4 ringwoodite up to 2.6 wt% at 19.5 GPa and 1100 °C. Since then, many studies confirmed that ringwoodite may dissolve more than 2 wt% of water (e.g., Kudoh et al. 2000; Ohtani et al. 2000; Smyth et al. 2003), although the mechanism of water incorporation is not completely understood. Infrared spectra of hydrous ringwoodite with water contents up to about 1 wt% show a rather broad absorption band at about 3100 cm−1 (Fig. 16), while at higher water contents, additional bands appear at 3345 cm−1 and 3645 cm−1 (Kohlstedt et al. 1996). The position of the bands at low frequency would be consistent with the incorporation of protons on edges of the tetrahedral site, while the band at 3645 cm−1 would be more consistent with protons located on the edge of an octahedral site. Both mechanisms are possible, because X-ray crystal structure refinements of hydrous ringwoodite show some vacancies on the Mg2+ site (Smyth et al. 2003), but there is also evidence for Mg2+Si4+-disordering (Kudoh et al. 2000), i.e., for some Mg2+ entering the tetrahedral site, which may involve local charge compensation by two protons. As with wadsleyite, the observed water contents in ringwoodite cannot be easily interpreted in terms of “solubility,” as the coexisting phase is usually a hydrous silicate

220

Keppler & Bolfan-Casanova

Figure 16. Infrared spectra of hydrous ringwoodite (Fo88-90) with up to 1 wt% water. The spectra shown here are characteristic for ringwoodite with low to moderate water content. At water contents of 2 wt% or above, there is a prominent band at 3645 cm−1 which is only present as a shoulder in the spectra shown here. From Smyth et al. (2003).

melt, not an aqueous fluid. Ohtani et al. (2000) observed that the partition coefficient of ringwoodite / melt water between ringwoodite and silicate melt (Dwater ) increases from 0.021-0.024 at 1300 °C to 0.04-0.044 at 1450 °C. Together with the data on the partitioning of water between wadsleyite and silicate melt by Demouchy et al. (2005), this would imply that water partitioning between wadsleyite and ringwoodite is strongly temperature dependent, with wadsleyite / ringwoodite Dwater decreasing from about 4 at 1300 °C to about 2 at 1450 °C and to about 1 at 1600 °C. This would imply that the sharpness of the 520 km discontinuity, if it is related to the wadsleyite-ringwoodite transition, may be a strong function of temperature. Moreover, some repartitioning of water from the lower transition zone to the upper transition zone may have occurred during the cooling of the Earth since the Archean (Demouchy et al. 2005).

Partition coefficients of water between other high-pressure phases Table 3 shows some partition coefficients of water between various high-pressure phases in the system MgO-SiO2-H2O derived from infrared measurements of coexisting phases in high-pressure experiments (after Bolfan-Casanova et al. 2000). Note that while pure SiO2 stishovite dissolves very little water, water solubility increases very much in the presence of aluminum (Pawley et al. 1993). Also, in stishovite the OH dipole vector is oriented parallel to the c-axis, which causes a strong anisotropy of infrared absorption. Water partition coefficients derived from unpolarized infrared measurements are therefore subject to large errors. The infrared spectrum of hydrous majorite shows only one broad band at 3550 cm−1. The infrared spectrum of hydrous akimotoite is shown in Figure 17. A detailed structural model for water incorporation in akimotoite was derived by Bolfan-Casanova et al. (2002a). The 3390 cm−1 band is due to an OH group pointing into a tetrahedral void, while the band at 3320 cm−1 is due to an OH group located on the triangular face of a vacant octahedron. Bolfan-Casanova (2000) report up to 680 and 445 ppm wt of water in majorite and akimotoite, respectively, which is far below the water content found in wadsleyite and ringwoodite under comparable conditions. Thus the high-pressure polymorphs of MgSiO3 probably do not contribute much to the water storage capacity of the transition zone.

Thermodynamics of Water Solubility and Partitioning

Table 3. Experimentally determined partition coefficients of water among minerals of the lower mantle and the transition zone. P

T

(GPa)

(°C)

15 15 17.5 19 19 21 24 24

1300 1500 1500 1200 1300 1500 1600 1500

24 24

Phase assemblage

Partition coefficient

-------------System MgO – SiO2 – H2O------------Wads + Cen + melt DWadsleyite/Clinoenstatite∼ 3.8 Cen + Stish + melt DClinoenstatite/Stishovite ∼ 8.2 Maj + Stish + melt DMajorite/Stishovite ∼ 270 Ringw + Stish DRingwoodite/Stishovite ∼ 521 Ringw + Akim + Stish + melt DRingwoodite/Akimotoite ∼ 21 Akim + Stish + melt DAkimotoite/Stishovite ∼ 18 Perov + Akim + Stish + melt DAkimotoite/Perovskite >>1 Pe + Perov + melt DPericlase/Perovskite >1

-------------System MgO – FeO – SiO2 – H2O------------1400 Ringw + Perov DRingwoodite/Perovskite ∼ 1050 1400 Ringw + Perov + Fe-Pe + melt DRingwoodite/Perovskite ∼ 1400 DFerropericlase/Perovskite ∼ 60

Wads = wadsleyite, Cen = clinoenstatite, Stish = stishovite, Maj = majorite, Ring = ringwoodite, Akim = akimotoite, Perov = MgSiO3 perovskite, Fe-Pe = ferropericlase. After Bolfan-Casanova et al. (2000) and Bolfan-Casanova et al. (2003).

Figure 17. Polarized infrared spectra of MgSiO3 akimotoite containing about 350 ppm of water. From Bolfan-Casanova et al. (2002).

221

222

Keppler & Bolfan-Casanova WATER IN MINERALS OF THE LOWER MANTLE

Water in ferropericlase A peridotitic lower mantle is composed of 80% by volume of orthorhombic magnesium silicate perovskite (Mg,Fe)(Si,Al)O3, 15% of (Mg,Fe)O and 5% of cubic calcium silicate perovskite CaSiO3 (Wood 2000). Unlike the perovskite phases, ferropericlase is stable over a wide range of pressures and temperatures. Moreover, in a simple system MgO-H2O, the composition of coexisting periclase and water is a function of pressure and temperature only and therefore, equilibrium water contents measured in periclase represent true water solubilities. Although the addition of iron to the system introduces an additional degree of freedom, the system (Mg,Fe)O – H2O is still invariant, if pressure, temperature, oxygen fugacity and bulk molar Fe/Mg ratio are fixed. The solubility of water in (Mg0.93Fe0.07)O was studied by Bolfan-Casanova et al. (2002b), as a function of pressure from 1.2 to 25 GPa at 1200°C and at Re-ReO2 buffer conditions. The hydrous ferropericlase samples showed infrared bands at 3320 and 3480 cm−1 (Fig. 18), together with a broad band at 3400 cm−1 and one peak at ~3700 cm−1. The intensity of the latter feature increases dramatically in areas containing inclusions and is therefore attributed to brucite precipitated upon quenching of the MgO-rich fluid (see also Gonzalez et al. 1982). Whether the broad band at 3400 cm−1 is due to structural OH is uncertain, but since the intensity of this band decreases with increasing pressure and becomes very weak at the conditions where ferropericlase is stable in the mantle, it was not included in the thermodynamic model of water solubility (Table 2). The concentration of OH increases as a function of pressure to 20 ppm H2O by weight (100 H/106Me, where Me = Mg and Fe) at 25 GPa. On the other hand, the ferric iron content decreases, as indicated by Mössbauer spectra, even if oxidizing conditions

Figure 18. Infrared spectra showing the OH stretching region of periclase with different compositions hydrothermally annealed at 25 GPa and 1200°C. The effect of iron content from XFe = 0 to 13% p.f.u. is shown as well as the effect of adding trivalent cations to MgO, such as Al3+ and Cr3+.

Thermodynamics of Water Solubility and Partitioning

223

are maintained in the capsule by using the Re-ReO2 buffer. Thus the Fe3+ and OH contents in ferropericlase show opposite trends with pressure. Water solubility increases with the square root of water fugacity (Table 2), indicating the incorporation of water as isolated OH groups. The change in volume of the solid upon hydration is very low (3.95 cm3/mol). The effect of iron content on water solubility in ferropericlase was studied by BolfanCasanova et al. (2006). They showed that varying the iron content from 1 to 13% has little effect on the solubility of water in ferropericlase at 25 GPa, and the incorporation mechanism appears to be the same, judging from the similarity in the IR spectra (Fig. 18). The main effect of iron is a shift of the infrared absorption bands to higher frequencies. Fe3+/Fetot decreases with pressure of hydrothermal annealing at 1200 °C for the three compositions studied (XFe = 0.01, 0.07, 0.13), and the higher the iron content, the higher is the Fe3+ content for similar (P, T, fO2) conditions. To the contrary, increasing temperature from 1200 °C to 1600 °C induces an oxidation of the samples, whereas the OH content decreases. Thus the solubilities of Fe3+ and OH are again anticorrelated as a function of increasing temperature. The infrared spectrum of MgO hydrothermally annealed at 25 GPa and 1200 °C is also shown in Figure 18. The spectrum is mainly composed of three bands at 3372, 3309 and 3296 cm−1. When comparing the OH absorption bands in pure MgO with those of MgO doped with Al3+ or Cr3+ at 1200 °C, one observes that the band at 3372 cm−1 is intrinsic to pure MgO. This band is also present in the spectra of Al-, Cr-, and Fe-doped MgO but with much lower intensity. Again, in Cr-doped MgO, a broad band centered around 3400 cm−1 is observed, the origin of which is still not understood. Thus, the effect of doping with trivalent cations is to increase the intensity of the bands at 3309 and 3296 cm−1, showing that one important mechanism of H incorporation in MgO is via coupling with trivalent cations such as Al3+ and Cr3+, while in (Mg,Fe)O the concentrations of Fe3+ and H+ are anticorrelated. At room pressure, under relatively oxidizing conditions, the most stable ionic defect species in anhydrous (Mg,Fe)O are cation vacancies and ferric iron in octahedral coordination (Gourdin and Kingery 1979). Tetrahedrally coordinated ferric iron is stabilized in Fe-rich compositions (Jacobsen et al. 2002). At low iron contents or low ferric iron contents, the point defect concentrations will be dominated by impurities, particularly trivalent cations such as Al3+ in octahedral coordination (see Mackwell et al. 2005). In the dry case, the incorporation of such trivalent cations is charged balanced by the creation of magnesium vacancies. H+ probably enters periclase coupled to trivalent cations in the octahedral site according to the substitution mechanism 2 Mg2+ ↔ M3+ + H+, in agreement with the observed exponent ½ in the water fugacity term. The fact that this mechanism is not very efficient for ferric iron might be due to the low fraction of Fe3+/Fetot in ferropericlase at high pressures.

Water in magnesium silicate perovskite Evidence for structurally bound H in MgSiO3 perovskite was first reported by Meade et al. (1994) based on IR studies of single crystals quenched from an H2O-rich melt. Two pleochroic bands at 3483 and 3423 cm−1 were observed consistent with 60 ppm H2O by weight (700 ± 170 H/106 Si). Bolfan-Casanova et al. (2000) did not detect any OH peak in MgSiO3 perovskite in an extensive experimental study performed under various conditions in the MgO-SiO2-H2O system, even when the perovskite coexisted with other minerals that clearly contained water. For example, while MgSiO3 akimotoite coexisting with MgSiO3 perovskite in a sample synthesized at 24 GPa and 1600°C, contained up to 425 ppm wt H2O, the perovskite was essentially dry. Experiments where MgO coexisted with perovskite at 24 GPa and 1500 °C, yield the same result (Table 3). The spectrum of periclase displayed two very sharp bands located at 3295 and 3372 cm−1, corresponding to ~2 ppm wt H2O. Bolfan-Casanova et al. (2000) argued that the H content observed by Meade et al. (1994) was not in equilibrium

224

Keppler & Bolfan-Casanova

because the sample was allowed to equilibrate only for a few minutes. Litasov et al. (2003) also studied MgSiO3 perovskite at 25 GPa and 1300°C and observed the bands at 3423 and 3482 cm−1, together with a more intense band at 3448 cm−1 and a weaker one at 3397 cm−1. They calculated a water content of about 40 ppm wt H2O. It is now probably generally agreed that MgSiO3 perovskite in the simple MgO-SiO2-H2O system dissolves only very little water, although the available experimental results are difficult to interpret in terms of a thermodynamically well-defined solubility. The somewhat higher water contents observed by Litasov et al.(2003) as compared to Bolfan-Casanova et al. (2000) could reflect a decrease of water solubility with temperature. Alternatively, they could be due to differences water activity or MgO activity. Ross et al. (2003) studied the electron density distribution in a number of high-pressure silicates and suggested that MgSiO3 perovskite has only one potential docking site for hydrogen. The OH vector is in the (110) plane and hydrogen incorporation probably requires a vacant Mg site, implying that water solubility may be reduced under conditions of high MgO activity, i.e., when perovskite coexists with MgO as in some of the experiments by Bolfan-Casanova et al. (2000). In the system MgO-FeO-SiO2-H2O, the IR spectra of (Mg,Fe)SiO3 perovskite synthesized at 24 GPa and 1400°C display only one very weak peak, at 3388 cm−1 (Bolfan-Casanova et al. 2003), yielding 2 ppm wt water (Fig. 19). This study also determined the partition coefficient of water between (Mg,Fe)2SiO4 ringwoodite and (Mg,Fe)SiO3 perovskite to be ~1050. The effect of aluminum on water solubility in MgSiO3 perovskite is subject of a major debate. It has been suggested that the incorporation of trivalent ions, especially Al3+, will enhance the solubility of hydrogen in magnesium silicate perovskite (Navrotsky 1999). This idea is based on the observation that in ceramic perovskites the trivalent cation occupies the B site in the ABO3 structure, resulting in the creation of oxygen vacancies. These oxygen vacancies may then be filled by OH groups. Ultimately, this would be equivalent to a substitution where Al3+ and H+ replace Si4+. However, the picture may not be that simple in magnesium silicate perovskite as Al3+ can also enter the A site. Also, the coupling of Al3+

Figure 19. Infrared spectra showing the OH stretching region of perovskite (Mg0.98Fe0.02)SiO3 after BolfanCasanova et al. (2003) and of (Mg,Si,Al)O3 perovskite containing 4.4 wt% Al2O3 after Litasov et al. (2003). Samples were synthesized at 25 GPa and 1400 and 1200 °C, respectively. The origin of the broad band is not fully understood, it may be related to inclusions of a separate phase.

Thermodynamics of Water Solubility and Partitioning

225

and Fe3+ plays an important role in the point defect chemistry of perovskite (Lauterbach et al. 2000). Litasov et al. (2003) find that in (Al,Fe)-perovskite, coupling of Al3+ with Fe3+ (up to 60% of the total iron being ferric) controls the amount of oxygen vacancies. In perovskite grown from a starting material of MORB composition, the Fe contents are high enough to charge compensate the Al3+ defects on the B site by Fe3+ defects on the A site. Litasov et al. (2003) analysed Al-perovskites containing 2, 4.5 and 7.2 wt% Al2O3, synthesized at temperatures of 1200-1400 °C and pressures of 25-26 GPa. They report OH peaks located at 3397, 3404 and 3448 cm−1, superimposed on a very broad band centered around 3400-3450 cm−1, the intensity of which increases with Al2O3 content in the perovskite (Fig. 19). The integrated water contents are 100, 1100 and 1400 ppm H2O by weight in the order of increasing aluminum content. However, it seems that the intensity of the sharp peaks decreases with aluminum. If only these peaks are considered as structural water in perovskite, the water contents in Al-perovskite would accordingly decrease with Al. Litasov et al. (2003) also studied Al-Fe bearing perovskites in the MORB and peridotite systems. In the MORB system, perovskites synthesized at pressures of 25-26 GPa and temperatures of 1000, 1200 and 1300 °C display IR peaks at 3397, 3423 and 3448 cm−1. The water content associated with the most intense band, at 3397 cm−1, is 40-110 ppm wt H2O, decreasing with increasing temperature of synthesis. In perovskites synthesized at 25 GPa and temperatures of 1400 and 1600°C in the peridotite system, the IR bands are more intense than in MORB related perovskite and yield 1400-1800 ppm wt H2O, with water contents decreasing slightly with increasing temperature. These perovskites generally display very broad bands in the IR spectra. Murakami et al. (2002) reported about 0.2 wt% water in MgSiO3 perovskite and up to 0.4 wt% water in amorphized CaSiO3 perovskite grown in a natural peridotitic composition at ~25 GPa and 1600-1650°C. The IR absorption features are very broad with a major broad band centered on 3400 cm−1 and a sharp peak at 3690 cm−1. The latter band resembles the band of brucite Mg(OH)2 at 3698 cm−1 (brucite microinclusions have also been identified in (Mg,Fe)O hydrothermally annealed at high pressures and temperatures, see above). The infrared spectra measured on different crystals display different intensities, and the SIMS measurements also show a large scatter in water contents (from 0.1 to 0.36 wt% H2O). These heterogeneities either mean that the samples are not homogeneous in water content, which means that they are not in equilibrium, or that the signal arises from an impurity phases. Whether aluminous perovskite is indeed able to dissolve much more water than pure MgSiO3 perovskite entirely depends on the interpretation of the IR spectra. The high water contents reported for experimental samples are always due to very broad infrared bands. Although it is possible that such broad bands may be caused by OH point defects, a number of observations suggest that these bands may be due to inclusions of a second phase: (1) Murakami et al. (2002) found this band both in coexisting ferropericlase and in perovskite and observed a significant pleochroism in both phases, despite the fact that ferropericlase is cubic. (2) Bolfan-Casanova et al. (2003) demonstrated that the broad bands are strong only in those parts of perovskite crystals that appear milky under the microscope, while they are not found in the infrared spectra taken on perfectly clear spots. (3) The same authors also demonstrated that some of the milky parts of synthetic aluminous perovskite show Raman peaks of superhydrous phase B and that the infrared peaks of this phase coincide with the major OH peaks reported for aluminous perovskite. Based on these observations, we tentatively conclude that water solubility in aluminous perovskite is generally very low and the high water contents reported in the literature are probably due to impurities.

The distribution of water at the 660 km discontinuity Based on the discussion above, it appears that both ferropericlase and particularly magnesium silicate perovskite dissolve much less water than ringwoodite under comparable conditions. No experimental data are available on water in calcium silicate perovskite.

226

Keppler & Bolfan-Casanova

However, Ross et al. (2003) could not locate any suitable docking site for hydrogen in the structure of calcium silicate perovskite based on its electron density distribution. This would suggest that this phase is not a significant storage site for water in the mantle. The data on water solubility in ferropericlase together with the data on water partitioning between ringwoodite and perovskite and between perovskite and ferropericlase (Table 3) suggest a bulk partition ringwoodite /( ferropericlase + perovskite ) coefficient between ringwoodite and perovskite + ferropericlase Dwater of about 710. This implies that the 660 km discontinuity should be broadened and shifted to greater depth in the presence of water. Compared to the transition zone, the bulk storage capacity of the lower mantle is probably negligible.

THE EQUILIBRIUM DISTRIBUTION OF WATER IN THE EARTH’S INTERIOR Measurements of water solubility in minerals cannot directly yield the actual water content of the mantle. They only constrain the maximum water storage capacity of the mantle and the partitioning of water between various phases. However, if one assumes chemical equilibrium throughout the mantle, partitioning data can be used to calculate the water distribution in the mantle based on independent estimates of the actual water content in one of the major mantle reservoirs. Whether the water distribution in the mantle is indeed close to equilibrium, is uncertain, although the relatively high diffusion coefficients of hydrogen in minerals (Ingrin and Blanchard 2006, this volume) and the high mobility of aqueous fluids could help to establish equilibrium over geologic timescales. Richard et al. (2002) have investigated the distribution of water in the Earth’s interior using convection models together with data on the diffusion of water in minerals. From their calculations, it appears that an equilibrium distribution of water in the mantle is only possible if aqueous fluids, hydrous silicate melts or hydrous carbonatite melts are involved in the transport of water. One model of equilibrium water distribution in the mantle is given in Table 4, after Bolfan-Casanova (2000). Key input parameters in this model are a bulk water content of 250 ppm in the upper mantle (consistent with estimates from both mantle xenoliths and analyses of MORB glasses), a partition coefficient of water between wadsleyite and olivine of 20 and a very low water solubility in lower mantle phases. The predicted water concentration in the transition zone of 1400 ppm is consistent with recent estimates based on electrical conductivity (Huang et al. 2005; see also Hirschmann 2006). On the other hand, one may assume a water content of 1500 ppm in wadsleyite of the upper transition zone from electrical conductivity measurements (Huang et al. 2005) as a starting point for the calculation. Together with a water partition coefficient of 3.8 between wadsleyite and olivine, which may result from the new calibration of infrared extinction coefficients in olivine according to Bell et al. (2003), this yields a water content in upper mantle olivine of 395 ppm. While this appears significantly higher than most estimates for water in the MORB

Table 4. A model for the equilibrium distribution of water in the mantle. Mass fraction of the Earth

Water content

Mass of water

(ppm by weight)

(1020 kg H2O)

% of ocean mass

Upper mantle

0.103

∼ 250

1.54

11

Transition zone

0.075

∼ 1400

6.27

45

Lower mantle

0.492

∼ 10

0.29

2

Reservoir

From Bolfan-Casanova (2000); mass of the Earth = 5.97 × 1023 kg, mass of the oceans = 1.4 × 1021 kg

Thermodynamics of Water Solubility and Partitioning

227

source, it is within the range of water contents inferred for some enriched mantle sources (up to 700 ppm, see review by Bolfan-Casanova 2005 and references therein).

ACKNOWLEDGMENTS Much of the work presented here has been supported by German Science Foundation (DFG; Gerhard Hess Award and Leibniz Award to HK) and by the EU Hydrospec Network organized by Jannick Ingrin. A review by Marc Hirschmann helped to improve the manuscript.

REFERENCES Ackermann L, Cemic L, Langer K (1983) Hydrogarnet substitution in pyrope: a possible location for “water” in the mantle. Earth Planet Sci Lett 62:208-214 Aubaud C, Hauri EH, Hirschmann MM (2004) Hydrogen partition coefficients between nominally anhydrous minerals and basaltic melts. Geophys Res Lett 31:L20611, doi:10.1029/2004GL021341 Bai Q, Kohlstedt DL (1992) Substantial hydrogen solubility in olivine and implications for water storage in the mantle. Nature 357:672-674 Bai Q, Kohlstedt DL (1993) Effects of chemical environment on the solubility and incorporation mechanism for hydrogen in olivine. Phys Chem Minerals 19:460-471 Bell DR, Rossman GR (1992) Water in Earth’s mantle: The role of nominally anhydrous minerals. Science 255: 1391-1397 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res 108 (B2):2105, doi:10.1029/2001JB000679 Bercovici D, Karato S-I (2003) Whole-mantle convection and the transition-zone water filter. Nature 425:3944 Berry AJ, Herrmann J, O’Neill H St C, Foran GJ (2005) Fingerprinting the water site in mantle olivine. Geology 33:869-872 Bolfan-Casanova N (2000) The distribution of water in the Earth’s mantle: An experimental and infrared spectroscopic study. Ph. D. thesis, University of Bayreuth. Bolfan-Casanova N (2005) Water in the Earth’s mantle. Mineral Mag 69:229-257 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: Implications for the distribution of water in the Earth´s mantle. Earth Planet Sci Lett 182:209-221 Bolfan-Casanova N, Keppler H, Rubie DC (2002a) Hydroxyl in MgSiO3 akimotoite: A polarized and highpressure IR study. Am Mineral 87:603-608 Bolfan-Casanova N, Mackwell S, Keppler H, McCammon CA, Rubie DC (2002b) Pressure dependence of H solubility in magnesiowustite up to 25 GPa: Implications for the storage of water in the Earth’s lower mantle. Geophys Res Lett 29:1029-1032 Bolfan-Casanova N, Keppler H, Rubie DC (2003) Water partitioning at the 660 km discontinuity and evidence for very low water solubility in magnesium silicate perovskite. Geophys Res Lett 30(17):1905 Bolfan-Casanova N, McCammon C, Mackwell S (in press) Water in transition zone and lower mantle minerals In: The Earth’s deep water cycle. Jacobsen S, Marone F (eds) American Geophysical Union Monograph, in press Bromiley GD, Keppler H (2004) An experimental investigation of hydroxyl solubility in jadeite and Na-rich clinopyroxenes. Contrib Mineral Petrol 147:189-200 Bromiley GD, Keppler H, McCammon C, Bromiley FA, Jacobsen SD (2004) Hydrogen solubility and speciation in natural, gem-quality chromian diopside. Am Mineral 89:941-949 Bureau H, Keppler H (1999) Complete miscibility between silicate melts and hydrous fluids in the upper mantle: experimental evidence and geochemical implications. Earth Planet Sci Lett 165:187-196. Chen J, Inoue T, Yurimoto H, Weidner DJ (2002) Effect of water on olivine-wadsleyite phase boundary in the (Mg,Fe)2SiO4 system. Geophys Res Lett 29:1875, doi:10.1029/2001GL014429 Dankwerth PA, Newton RC (1978) Experimental determination of the spinel to garnet peridotite reaction in the system MgO-Al2O3-SiO2 in the range 900 °C-1100 °C and Al2O3 isopleths of enstatite in the spinel field. Contrib Mineral Petrol 66:189-201 Demouchy S, Deloule E, Frost DJ, Keppler H (2005) Pressure and temperature dependence of water solubility in Fe-free wadsleyite. Am Mineral 90:1084-1091

228

Keppler & Bolfan-Casanova

Freda C, Baker DR, Ottolini L (2001) Reduction of water loss from gold-palladium capsules during pistoncylinder experiments by use of pyrophyllite powder. Am Mineral 86:234-237 Gasparik T (1993) The role of volatiles in the transition zone. J Geophys Res 98:4287-4299 Gasparik T (2003) Phase Diagrams for Geoscientists: An Atlas of the Earth Interior. Springer Gonzalez R, Chen Y, Tsang KL (1982) Diffusion of deuterium and hydrogen in doped and undoped MgO crystals Phys Rev B 26:4637-4645 Gourdin WH, Kingery WD (1979) The defect structure of MgO containing trivalent defect: solutes shell model calculations. J Material Sci 14:2053-2073 Hirschmann MM (2006) A wet mantle conductor? Nature 439:E3, doi:10.1038/nature04528 Hirschmann MM, Aubaud C, Withers AC (2005) Storage capacity of H2O in nominally anhydrous minerals in the upper mantle. Earth Planet Sci Lett 236:167-181 Huang XG, Xu YS, Karato SI (2005) Water content in the transition zone from electrical conductivity of wadsleyite and ringwoodite. Nature 434:746-749 Ingrin J, Blanchard M (2006) Diffusion of hydrogen in minerals. Rev Mineral Geochem 62:291-320 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Inoue T, Yurimoto H, Kudoh Y (1995) Hydrous modified spinel Mg1.75SiH0.5O4 – A new water reservoir in the mantle transition region. Geophys Res Lett 22:117-120 Ito E, Harris DM, Anderson AT (1983) Alteration of oceanic crust and geologic cycling of chlorine and water. Geochim Cosmochim Acta 47:1613-1624 Jacobsen SD, Reichmann HJ, Spetzler HA, Mackwell SJ, Smyth JR, Angel RJ, McCammon CA (2002) Structure and elasticity of single-crystal (Mg,Fe)O and a new method of generating shear waves for gigahertz ultrasonic interferometry. J Geophys Res 107 B: Art. No. 2037 Jacobsen SD, Demouchy S, Frost D, Boffa Ballaran T, Kung J (2005) A systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single-crystal X-ray diffraction: Oxygen sites for hydrogen storage in Earth’s interior. Am Mineral 90:61-70 Johnson EA, Rossman GR (2004) A survey of hydrous species and concentrations in igneous feldspars. Am Mineral. 89:586-600 Kawamoto T, Hervig RL, Holloway JR (1996) Experimental evidence for a hydrous transition zone in the early Earth’s mantle. Earth Planet Sci Lett 142:587-592 Kessel R, Schmidt MW, Ulmer P, Pettke T (2005) Trace element signature of subduction-zone fluids, melts and supercritical liquids at 120-180 km depth. Nature 437:724-727 Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ (2002) Ordering of hydroxyl defects in hydrous wadsleyite (β-Mg2SiO4). Am Mineral 87:293-301 Kohn SC, Roome BM, Smith ME, Howes AP (2005) Testing a potential mantle geohygrometer; the effect of water on the intracrystalline partitioning of Al in orthopyroxene. Earth Planet Sci Lett 238:342-350 Kudoh Y, Kuribayashi T, Mizobata H, Ohtani E (2000) Structure and cation disorder of hydrous ringwoodite, γ-Mg1.89Si0.98H0.30O4. Phys Chem Minerals 27:474-479 Lager GA, Marshall WG, Liu Z, Downs RT (2005) Re-examination of the hydrogarnet structure at high pressure using neutron powder diffraction and infrared spectroscopy. Am Mineral 90:639-644 Lauterbach S, McCammon C, van Aken P, Langenhorst F, Seifert F (2000) Mössbauer and ELNES spectroscopy of (Mg,Fe)(Si,Al)O3 perovskite: a highly oxidized component of the lower mantle. Contrib Mineral Petrol 138:17-26 Lemaire C, Kohn SC, Brooker RA (2004) The effect of silica activity on the incorporation mechanism of water in synthetic forsterite: a polarized infrared spectroscopic study. Contrib Mineral Petrol 147:48-57 Litasov K, Ohtani E (2003) Stability of various hydrous phases in CMAS pyrolite-H2O system up to 25 GPa. Phys Chem Minerals 30:147-156 Litasov K, Ohtani E, Langenhorst F, Yurimoto H, Kubo T, Kondo T (2003) Water solubility in Mg-perovskites and water storage capacity in the lower mantle. Earth Planet Sci Lett 211:189-203 Lu R, Keppler H (1997) Water solubility in pyrope to 100 kbar. Contrib Mineral Petrol 129:35-42 MacGregor ID (1974) The system MgO-Al2O3-SiO2: Solubility of Al2O3 in enstatite for spinel and garnet peridotite compositions. Am Mineral 59:110-119 Mackwell S, Bystricky M, Sproni C (2005) Fe-Mg Interdiffusion in (Mg,Fe)O. Phys Chem Mineral 32:418425 Matsyuk SS, Langer K (2004) Hydroxyl in olivines from mantle xenoliths in kimberlites of the Siberian platform. Contrib Mineral Petrol 147:413-437 Matveev S, O´Neill H St C, Ballhaus C, Taylor WR, Green DH (2001) Effect of silica activity on OH− IR spectra of olivine: Implications for low-α−SiO2 mantle metasomatism. J Petrol 42:721-729

Thermodynamics of Water Solubility and Partitioning

229

Matveev S, Portnyagin M, Ballhaus C, Brooker R, Geiger CA (2005) FTIR spectrum of phenocryst olivine as an indicator of silica saturation in magmas. J Petrol 46:603-614 McMillan PF, Akaogi M, Sato RK, Poe B, Foley J (1991) Hydroxyl groups in β-Mg2SiO4. Am Mineral 76: 354-360 Meade C, Reffner JA, Ito E (1994) Synchrotron infrared absorbance measurements of hydrogen in MgSiO3 perovskite. Science 264:1558-1560 Mibe K, Fujii T, Yasuda A (2002) Composition of aqueous fluids coexisting with mantle minerals at high pressure and its bearing on the differentiation of the Earth’s upper mantle. Geochim Cosmochim Acta 66: 2273-2285 Mierdel K (2006) Wasserlöslichkeit in Enstatit. PhD thesis, University of Tübingen Mierdel K, Keppler H (2004) The temperature dependence of water solubility in enstatite. Contrib Mineral Petrol 148:305-311 Miller GH, Rossman GR, Harlow GE (1987) The natural occurrence of hydroxide in olivine. Phys Chem Minerals 14:461-472 Mosenfelder JL, Deligne NI, Asimow PD, Rossman GR (2005) Hydrogen incorporation in olivine from 2-12 GPa. Am Mineral 91:285-294 Murakami M, Hirose K, Yurimoto H, Nakashima S, Takafuji N (2002) Water in the Earth’s lower mantle. Science 295:1885-1887 Navrotsky A (1999) A lesson from ceramics. Science 284:1788-1789 Ohtani E, Mizobata H, Yurimoto H (2000) Stability of dense hydrous magnesium silicate phases in the systems Mg2SiO4-H2O and MgSiO3-H2O at pressure up to 27 GPa. Phys Chem Minerals 27:533-544 Paterson MS (1982) The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bull Minéral 15:20-29 Pawley AR, McMillan PF, Holloway JR (1993) Hydrogen in stishovite, with implications for mantle water content. Science 261:1024-1026 Perkins D, Holland TJB, Newton RC (1981) The Al2O3 content of enstatite in equilibrium with garnet in the system MgO-Al2O3-SiO2 at 15-40 kbar and 900-1600 °C. Contrib Mineral Petrol 78:99-109 Rauch M (2000) Der Einbau von Wasser in Pyroxene. PhD thesis, University of Bayreuth Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Richard G, Monnereau M, Ingrin J (2002) Is the transition zone an empty water reservoir? Inferences from numerical model of mantle dynamics. Earth Planet Sci Lett 205:37-51 Ross NL, Gibbs GV, Rosso K (2003) Potential docking sites and positions of hydrogen in high-pressure silicates. Am Mineral 88:1452-1459 Rossman GR, Aines RD (1991) The hydrous components in garnets: Grossular-hydrogrossular. Am Mineral 76: 1153-1164 Rüpke LH, Phillips Morgan JP, Hort M, Connolly JAD (2004) Serpentine and the subduction zone water cycle. Earth Planet Sci Letters 223:17-34 Shen A, Keppler H (1997) Direct observation of complete miscibility in the albite-H2O system. Nature 385: 710-712 Skogby H (2006) Water in natural mantle minerals I: pyroxenes. Rev Mineral Geochem 62:155-167 Smyth JR (1987) β-Mg2SiO4: A potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR, Frost DJ (2002) The effect of water on the 410-km discontinuity: an experimental study. Geophys Res Lett 29:1485, doi:10.1029/2001GL014418 Smyth JR, Kawamoto T (1997) Wadsleyite II, a new high pressure hydrous phase in the peridotite-H2O system. Earth Planet Sci Lett 146:E9-E16 Smyth J, Bell D, Rossman G (1991) Incorporation of hydroxyl in upper-mantle clinopyroxenes. Nature 351: 732-735 Smyth JR, Kawamoto T, Jacobsen SD, Swope RJ, Hervig RL, Holloway JR (1997) Crystal structure of monoclinic hydrous wadsleyite, β-(Mg,Fe)2SiO4. Am Mineral 82:270-275 Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Langenhorst F, McCammon CA (2003) Structural systematics of hydrous ringwoodite and water in Earth´s interior. Am Mineral 88:1402-1407 Stalder R (2002) Synthesis of enstatite single crystals at high pressure. Eur J Mineral 14:637-640 Stalder R (2004) Influence of Fe, Cr and Al on hydrogen incorporation in orthopyroxene. Eur J Mineral 16: 703-711 Stalder R, Skogby H (2002) Hydrogen incorporation in enstatite. Eur J Mineral 14:1139-1144 Stalder R, Klemme S, Ludwig T, Skogby H (2005) Hydrogen incorporation in orthopyroxene: interaction of different trivalent cations. Contrib Mineral Petrol 150:473-485 Sykes D, Rossman GR, Veblen DR, Grew ES (1994) Enhanced H and F incorporation in borian olivine. Am Mineral 79:904-908 Truckenbrodt J, Johannes W (1999) H2O loss during piston-cylinder experiments. Am Mineral 84:1333-1335 Withers AC, Wood BJ, Carroll MR (1998) The OH content of pyrope at high pressure. Chemical Geology 147: 161-171

230

Keppler & Bolfan-Casanova

Wood BJ (1995) The effect of H2O on the 410-kilometer seismic discontinuity. Science 268:74-76 Wood BJ (2000) Phase transformations and partitioning relations in peridotite under lower mantle conditions. Earth Planet Sci Lett 174:341-354 Yamada A, Inoue T, Irifune T (2004) Melting of enstatite from 13 to 18 GPa under hydrous conditions. Phys Earth Planet Int 147:45-56 Young TE, Green HW, Hofmeister AM, Walker D (1993) Infrared spectroscopic investigation of hydroxyl in β-(Mg,Fe)2SiO4 and coexisting olivine: Implications for mantle evolution and dynamics. Phys Chem Minerals 19:409-422 Zhao YH, Ginsberg SB, Kohlstedt DL (2004) Solubility of hydrogen in olivine: dependence on temperature and iron content. Contrib Mineral Petrol 147:155-161

10

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 231-241, 2006 Copyright © Mineralogical Society of America

The Partitioning of Water Between Nominally Anhydrous Minerals and Silicate Melts Simon C. Kohn and Kevin J. Grant1 Department of Earth Sciences University of Bristol Bristol, BS8 1RJ, United Kingdom e-mail: [email protected] (1present address: Dept. of Earth & Planetary Sci., Macquarie Univ., Sydney, NSW 2109, Australia)

INTRODUCTION Many chapters in this volume emphasize that water plays a crucial role in determining the properties and behavior of the mantle. One of the most important examples of the effect of water on mantle behavior is its dramatic effect in reducing melting temperatures, and, in the case of decompression melting, increasing the depth of initiation of melting. Even the small amount of water which could be contained within nominally anhydrous minerals (NAMs) is sufficient to play an important role in generation of MORBs (Hirth and Kohlstedt 1996; Asimow and Langmuir 2003; Asimow et al. 2004). Figure 1 illustrates how, according to the model of Aubaud et al. (2004), small amounts of water affect the depth of initiation of melting in upwelling mantle. Furthermore, in regions of the mantle where large concentrations of water could be stored (e.g., the transition zone), phase changes could release water and generate melt in previously unexpected environments (Bercovici and Karato 2003; Hirschmann 2006). The effect of water on mantle melting is one of many reasons why knowledge of the spatial and temporal distribution of water in the mantle is crucial to understanding the evolution of our planet.

PPM PPM PPM PPM

1

PPM

7ATERCONCENTRATION

bat Plume adia

3

50

5

7 1000

s lidu So Dry

6

150

1500 1750 1250 Temperature (°C)

1529-6466/06/0062-0010$05.00

200

Depth (km)

100

4

bat Ridge adia

Pressure (GPa)

2

Figure 1. Model for the effect of small concentrations of water in the mantle on the depth of initiation of mantle melting (after Aubaud et al. 2004). Note that even 200 ppm water reduces the depth of melting by at least 20 km and that the effect is more pronounced for a high mantle potential temperature.

2000

DOI: 10.2138/rmg.2006.62.10

232

Kohn & Grant

It is also possible to look at this problem from the other direction, because the products of partial melting of the mantle provide an opportunity to assess the water concentrations in the source regions of these melts. The amount of water in different regions of the mantle can, in principle, be calculated from the water concentrations of mantle-derived melts. Using this approach it has been suggested that the source of mid-ocean ridge basalts contains 50-200 ppm H2O (e.g., Dixon et al. 1988; Michael 1988; Michael 1995; Sobolev and Chaussidon 1996; Danyushevsky et al. 2000; Saal et al. 2002) and the source of ocean island basalts contains 500-1000 ppm H2O (e.g., Bureau et al 1998; Wallace 1998; Nichols et al. 1999; Dixon and Clague 2001; Dixon et al. 2002; Simons et al 2002). Several parameters are needed for these calculations, namely: i)

The degree of degassing (if any) of the melt at the point of sampling.

ii) The amount of shallow level crystallization. iii) The amount and nature of contamination (if any) by non-mantle-derived fluids. iv) The degree and nature of melting in the source region. v)

Partition coefficients to describe the distribution of hydrogen between all the individual phases and the melt, at appropriate conditions of pressure, temperature, oxygen fugacity and chemical composition.

In practice, the effect of degassing can be minimized by using glasses from the rims of pillow lavas erupted under pressure from deep-water environments (although corrections may be required, e.g., Pineau et al 2004; Aubaud et al. 2005) and by studying melt inclusions. The degree of melting can be estimated from the major element composition and, if the melt fraction is sufficiently high, there is no significant difference between the aggregated products of fractional or batch melting. As water is highly incompatible, the exact partition coefficients are unimportant if the melt fraction is sufficiently high. However, interpreting the water concentrations of any magma which results from small degrees of melting relies heavily on a detailed knowledge of the relevant partition coefficients. The water concentrations in mantle-derived magmas have been measured over at least 20 years (for examples see reviews in Michael 1988, 1995; Jambon 1994; Johnson et al. 1994), and water has long been known to be incompatible in mantle melting processes. In recent years the amount of water in primary mantle melts has been studied in much more detail because of the availability of SIMS to analyze melt inclusions. It has been observed that in undegassed MORB, the ratio of H2O/Ce is approximately constant and has a value of 150-300 (depending on the type of MORB). Within this range there are correlations with the type of MORB, and it has been argued that the variations are related to the chemistry of the source rather than fractionation during melting. These concepts have been extended to other mantle-derived melts and there are now quite a large number of publications dealing with the geochemistry of mantle-derived melts and the significance of the water concentrations and ratios to incompatible elements such as H2O/Ce, H2O/K or H2O/La (e.g., Wallace 1998, 2002; Danyushevsky et al. 2000; Dixon and Clague 2001; Workman et al. 2006). While these geochemical studies strongly suggest that water is highly incompatible during mantle melting the details are not yet well understood. The incompatibility of water may be very different in mid ocean ridge melting, subduction zone melting or melting in and around the transition zone for example. Careful consideration of the abundance of trace elements and isotopes can provide a wealth of information on the source regions of mantle-derived magmas, but a complete understanding requires independent measurements of the partitioning of water and other incompatible elements between minerals and melts as a function of pressure, temperature and composition. There is now a large literature on experimental measurements of trace element partition coefficients, but few studies of water partitioning have been published. The aim of this paper is therefore to critically review the existing data on water partitioning and

Water Partioning Between NAM’s and Silicate Melts

233

briefly examine how these data impact on our understanding of mantle processes. Throughout the paper the term “water” will be used for H dissolved in both minerals and melts although H is thought to be stored mainly as hydroxyl species in nominally anhydrous minerals and both hydroxyl and molecular water in silicate melts. The partition coefficients DHNAM/melt are calculated as wt% H expressed as H2O in a NAM divided by wt% H2O in the coexisting melt.

PARTITIONING OF WATER BETWEEN NAMS AND MELTS; METHODOLOGY AND APPROACH There are several ways in which the partition coefficients, DHNAM/melt, can be determined. The only rigorous method is to experimentally equilibrate one or more NAMs with a hydrous silicate liquid, then quench the experiment and measure the water concentrations in all the phases. Alternatively the water concentrations in naturally occurring minerals and glasses can be measured, although in this case there may be uncertainties in the origin and history of the sample. Finally, data from a range of studies can be combined to generate partition coefficients. The determination of water concentrations in NAMs has been discussed elsewhere in this volume, and it is clear that such measurements are far from straightforward. It is relatively easy to obtain an OH stretching spectrum, using FTIR, for most NAMs, but converting the absorption to concentrations requires an appropriate extinction coefficient. Some studies have used mineralspecific calibrations (Bell et al. 1995, 2003), while others have used a calibration which varies according to the frequency of the OH vibration (Paterson 1982; Libowitzky and Rossman 1997). The mineral-specific calibrations are not necessarily the best choice, as the FTIR spectra of an individual NAM can be quite variable (e.g., Skogby et al. 1990; Lemaire et al. 2004; Matsyuk and Langer 2004), and one would not expect extinction coefficients to be constant for a given mineral. As different groups have used different infrared calibrations, great care has to be taken in comparing partition coefficients from different studies even if FTIR has been used as the method of analysis. This problem is particularly acute for olivine (Bell et al. 2003). Alternative methods of H analysis in NAMs include SIMS, ERDA and nuclear techniques (see below). Analysis of the water concentration in the melt phase involves fewer difficulties than for NAMs, but even with FTIR or SIMS some uncertainties in calibration remain.

Experimental studies of water partitioning between NAMs and melts The partitioning of trace elements between minerals and melts during igneous processes is a major tool in geochemistry. Data on the trace element concentrations of rocks are interpreted using experimentally determined partition coefficients, and a large number of publications are concerned with determining these coefficients (e.g., Green 1994; Blundy and Wood 2003). Typically the experimental method is to partially crystallize a bulk composition to obtain coexisting minerals and silicate melt which are in equilibrium with each other. The concentrations of the trace elements of interest are then analyzed using a microprobe technique such as electron probe microanalysis (EPMA), secondary ion mass spectrometry (SIMS) (also known as ion microprobe in the geological literature) or laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS). A similar procedure can, in principle, be applied to the partitioning of H between NAMs and melts. The main difference from other trace elements is the limitations of the analytical methods. EPMA is not suitable for light elements, the background for LA-ICPMS cannot be lowered sufficiently and SIMS suffers from a number of problems, include significant H background signals, and matrix dependent calibrations. Recently the problems with SIMS have been addressed (Koga et al. 2003) and if SIMS can be used routinely, it will improve the prospects for obtaining much more data in the future. SIMS analyses can usually be performed with a spot size of 5-10 µm, although for the measurements using a Cs primary ion beam the spot size is generally around 30 µm (Cyril Aubaud, pers.

234

Kohn & Grant

comm.). The beam size for SIMS is thus larger than for EPMA, but is small enough for many experimentally grown samples. Fourier transform infrared spectroscopy (FTIR) is sensitive to very low concentrations of OH, but the minimum practical spot size is about 50 µm for a laboratory source and around 5 µm, for a synchrotron source (Dumas and Miller 2003), and a homogeneous, inclusion-free slice of several 10s of µm is typically required. There are also complications arising from the pleochroism of the IR absorption (Libowitzky and Rossman 1996) and (as mentioned previously) calibration of the absorption. Alternative analytical techniques such as elastic recoil detection analysis (ERDA) and proton-proton scattering (Sweeney et al. 1997; Wegden et al. 2004, 2005) have a high spatial resolution, but these techniques have not yet been widely applied. The partition coefficients, DHNAM/melt, are expected to be small, of the order 10−2 or less, so one major analytical issue is to ensure that the area of NAM to be analyzed is entirely free of melt or fluid inclusions, or indeed of any hydrous phases. In this respect FTIR has an advantage over SIMS, because contamination by other phases may be detectable from the shape of the absorption spectrum. Orthopyroxene/melt partitioning. One value of DHen/melt can be estimated from the MgOSiO2-H2O system in the study of Grant et al. (2006b). Most of the samples in this study contained forsterite±enstatite in equilibrium with an aqueous fluid and did not contain melt, but one sample synthesized at 1 GPa and 1420 °C contained enstatite and hydrous melt. Enstatite water concentrations measured using FTIR were either 73 ppm (using the calibration by Libowitzky and Rossman 1997), or 90 ppm (using the calibration by Bell et al. 1995). The water concentration of the melt phase was not known precisely, but was estimated to be around 18 wt%, giving a value of DHen/melt = 4 × 10−4. The effect of additional components in the system can be assessed using the data of Grant et al. (2006c) in the system Fo-Alb-H2O. Six samples in this study yielded values for DHen/melt at pressures of 1.5 to 2.5 GPa, temperatures of 1295-1400 °C and melt water concentrations in the range 1.4-9.4 wt%. Amongst the samples synthesized at 1.5 GPa, DHen/melt varies from 3 × 10−3 to 6 × 10−3 (with the enstatite water contents based on the calibration by Libowitzky and Rossman 1997). One sample, synthesized at 2.5 GPa and 1330 °C gave a value of DHen/melt = 0.016. More data is required to unpick in detail the dependence of the partition coefficient on temperature, pressure, water concentration and chemical composition, but appears that a major control is the aluminum concentration of the orthopyroxene. This is not unexpected, since it is known that addition of Al enhances the solubility of H in orthopyroxene (Rauch and Keppler 2002; Stalder and Skogby 2002). More detail on the controls on DHopx/melt can be obtained by considering results from more complex compositions. Koga et al. (2003) described methods to improve the analysis of small concentrations of hydrogen in nominally anhydrous minerals and included analyses of one melting experiment at 1.8 GPa and 1380 °C, which contained coexisting olivine, orthopyroxene and silicate melt. The bulk composition contained Fe, Ca, Al, Na and Ti in addition to Mg and Si, and the experiment employed a Ni capsule to buffer the oxygen fugacity. As a result, both olivine and orthopyroxene had more complex compositions than those of Grant et al. (2006b,c) and were very Ni-rich. The value of DHopx/melt = 0.025 is an order of magnitude higher than the values reported by Grant et al. (2006c) for runs at 1.5 GPa, but there are major differences in mineral composition and analytical methodology. The study of Koga et al (2003) was followed by a more detailed SIMS study of water partitioning between orthopyroxene, clinopyroxene, olivine and water-undersaturated melt (Aubaud et al. 2004). In this study the pressure and temperature range was 1-2 GPa and 12301380 °C respectively and no oxygen buffer was used; as a result the mineral compositions match more closely those of mantle samples. The water concentrations of the synthetic NAMs were measured by SIMS and selected glasses were also analyzed by FTIR. Eight separate samples contained orthopyroxene-melt pairs and DHopx/melt varied from 0.013-0.027.

Water Partioning Between NAM’s and Silicate Melts

235

Recently, Aubaud et al. (2006) have refined their SIMS calibration, and give a revised values of DHopx/melt = 8 × 10−3 for a sample which previously was reported to have DHopx/melt = 19 × 10−3. The strong correlation between DHopx/melt and the Al concentration in orthopyroxene is illustrated in Figure 2. This figure includes data from four experimental studies, one study of a phenocryst/ melt pair and one study of xenoliths where DHopx/melt has been deduced indirectly. The slope of the best fit through all the data should be taken as provisional, as it would be substantially altered if all the partition coefficients of Aubaud et al. (2004) were to be corrected by a factor of 8/19. Furthermore, in natural samples the effect of Al could be reduced by interactions between trivalent cations (Stalder et al. 2005). Additional data is available from the study of Hauri et al. (2004). In this study, 23 experimentally produced assemblages were analyzed for water using SIMS, and partition coefficients were calculated. They report that the partition coefficients for water partitioning between NAMs and melt are independent of pressure, temperature and total water concentration over the range of conditions covered (1-4 GPa, 1000-1380 °C and melt water contents of 1-22 wt% H2O). The partition coefficient reported for orthopyroxene is DHopx/melt = 0.015, but the aluminum concentration of the orthopyroxene was not given. Olivine/melt partitioning. The early and influential work of Hirth and Kohlstedt (1996) combined the experimental water solubility data for olivine and basaltic melts which were available at that time, and deduced a partition coefficient of DHolivine/melt = 4 × 10−4 for a pressure of 300 MPa. They noted that the solubility of water in olivine and in silicate melts are different functions of water fugacity, and argued that at the depth of MORB melting DHolivine/melt would change to 3 × 10−3. Note that these figures are based on the IR extinction coefficient of Paterson (1982) and would be revised upwards according to Bell et al. (2003). The dataset of Grant et al. (2006b) does not include a direct measurement of DHfo/melt for the MgO-SiO2-H2O system. However, DHfo/melt can be estimated by combining the value of

0.03

Grant et al. (2006c) Grant et al. (2006b) Aubaud et al. (2004) Koga et al. (2003) Bell et al. (2004) Dobson et al. (1995)

0.025

DHopx/melt

0.02 0.015 0.01 0.005 0

0

2

4

6

8

10

wt % Al2O3 in enstatite

Figure 2. The effect of Al concentration in orthopyroxene on DHopx/melt (after Grant et al. 2006c). The filled symbols represent data from mineral/melt partitioning experiments, the open symbols are data from natural samples. Various analytical methods and calibrations were used in the different studies; only the Dobson et al. (1995) data point has been recalibrated from the original publication. This point uses the same calibration as the Bell et al. (2004) data. The dotted line is the best linear fit through the experimental data only and has the equation y = 0.000477 + 0.00315x.

236

Kohn & Grant

DHen/melt = 4 × 10−4 (at 1GPa and 1420 °C ) with DHen/fo = 3, a typical value found for other samples in the study of Grant et al. (2006b), to give DHfo/melt = 1.3 × 10−4. The effect of Al on DHfo/melt can be assessed from the data of Grant et al. (2006c). DHfo/melt varies from 2.7 × 10−4 to 1.2 × 10−4 at 1-2 GPa in Al-bearing samples from the Fo-Alb-H2O system. These values are close to that for the Al-free system despite the fact that the FTIR spectrum (and hence dissolution mechanism) depends strongly on the presence or absence of Al in the structure (Grant et al. 2006c). An olivine/melt partition coefficient was reported for a nickel-rich olivine by Koga et al. (2003). The value of DHolivine/melt = 2 × 10−3 is an order of magnitude higher than those reported by Grant et al. (2006b,c). The range of values from the study of Aubaud et al. (2004) is DHolivine/melt = 1.1 × 10−3 – 2.9 × 10−3, but use of the revised calibration of Aubaud et al. (2006) gives a value of DHolivine/melt = 6 × 10−4 for a sample which had previously been reported with DHolivine/melt = 16 × 10−4. The data of Hauri et al. (2004) suggest DHolivine/melt = 1.3 × 10−3. Sweeney et al. (1997) used ERDA as the analytical method in a study of olivine-melt partitioning of water. They reported a strongly pressure dependent partition coefficient, varying from DHolivine/melt = 0.04 at 1.5 GPa and 1400 °C , to DHolivine/melt = 0.13 at 6 GPa and 1740 °C and DHolivine/melt = 0.12 at 10 GPa and 2000 °C. The melt compositions described in this study are very alkali-rich and also vary with pressure, but the partition coefficients are extremely high when compared those determined using more familiar analytical techniques, and should perhaps be treated with caution until additional work has been done to confirm them. In summary there are major discrepancies in the values of DHolivine/melt reported in the literature and much more work on the effect of chemical composition (e.g., the concentration of Fe, Ti and Cr components and oxidation state) is required. In our opinion, however, crosstechnique calibration will be essential in resolving discrepancies in the data. Clinopyroxene/melt partitioning. There have been fewer experimental studies of water partitioning between clinopyroxene and melt, but the expectation is that DHcpx/melt will be higher than either DHopx/melt or DHolivine/melt, because many studies of natural and synthetic systems have shown that DHcpx/opx > 1 (e.g., Bell and Rossman 1992; Ingrin and Skogby 2000; Peslier et al. 2002). The partition coefficients reported in SIMS studies are DHcpx/melt = 0.019 – 0.026 (Aubaud et al. 2004), which was revised to 0.013 (Aubaud et al. 2006), and DHcpx/melt = 0.014 (Hauri et al. 2004). Garnet/melt partitioning. The only experimental measurement of water partitioning between garnet and melt was reported by Hauri et al. (2004) to be DHgarnet/melt = 3.2 × 10−3. No details are given in this abstract, but this value is surprisingly high, since most studies of coexisting olivine and garnet in mantle xenoliths suggest DHolivine/garnet is greater than 1. For example Bell et al. (2004) report DHolivine/garnet in the range 3-14 and observed a strong dependence of the OH partitioning in garnet on its Mg-number (100Mg/(Mg+Fe)). DHolivine/garnet is even higher in South African garnet lherzolites for garnets with higher Mg-numbers (Grant et al. 2006a). Wadsleyite/melt and ringwoodite/melt partitioning. All the examples of mineral melt partitioning considered so far in this review have been for pressure and temperature conditions where hydrous silicate melts and aqueous fluids (albeit with dissolved silicate components) are distinct entities. However, it is now well established that under high pressure and temperature conditions, silicate melts and aqueous fluids become completely miscible, at the 2nd critical end point (Shen and Keppler 1997). The pressure at which this point is reached varies with composition (Bureau and Keppler 1999), and for mantle compositions is probably somewhere in the deep upper mantle (Stalder et al. 2001). Transition zone minerals cannot therefore be equilibrated with aqueous fluids at high temperatures, and all solubility measurements for wadsleyite and ringwoodite are in fact measurements of the water concentration of the mineral in equilibrium with a supercritical fluid, whether such a fluid is recovered and

Water Partioning Between NAM’s and Silicate Melts

237

analyzed or not. Recently, Demouchy et al. (2005) performed a careful study of the effect of temperature on the water concentrations of the Mg-end member of wadsleyite. They combined the water concentrations in the minerals with the water concentration in the coexisting melt (supercritical fluid), calculated from mass balance, and found that the value of DHwads/melt = 0.08 was independent of temperature over the range 1100 °C to 1400 °C at 15 GPa. This partition coefficient is similar to those reported by Kawamoto et al. (1996) for a hydrous peridotite system. In that study DHwads/melt was estimated to be 0.10 ± 0.04, in good agreement with the data of Demouchy et al. (2005) despite the difference in chemistry between the two studies. Kawamoto et al. (1996) also reported a value for DHring/melt = 0.04 for ringwoodite coexisting with wadsleyite and melt in one sample at 1300 °C and 15.5 GPa. This can be compared with DHring/melt = 0.02–0.05 for ringwoodite in the MSH system at 1300 °C to 1450 °C and 20-23 GPa (Ohtani et al. 2000). It is possible that the change in partition coefficient is correlated with differences in either pressure or temperature of the runs, but only three samples are used to derive partition coefficients, and more work would be required to confirm this. In general terms all the data for wadsleyite and ringwoodite are quite consistent despite the differences in composition, pressure, temperature and methods for determining the water concentrations of the coexisting melt; although a strong pressure or temperature dependence of DHwads/ring would have interesting consequences for the structure for the transition zone (Demouchy et al. 2005), there is not yet any strong evidence for such behavior. Water partitioning observed in naturally coexisting mineral-melt pairs. An alternative approach to determining water partition coefficients for NAMs is to analyze natural mineralmelt pairs. Such pairs could be phenocrysts from a porphyritic volcanic rock and a glassy groundmass, or crystals of NAMs containing melt inclusions. If the crystals are large enough, they can be analyzed using FTIR, otherwise SIMS or synchrotron FTIR can be used. Dobson et al. (1995) used IR spectroscopy to measure the water concentrations in orthopyroxene, boninite glass melt inclusions, and quenched boninite glass from pillow lava rims. Water concentrations in the two types of glass were similar with only minor degassing from the pillow lava rims. It was therefore concluded that the measured partitioning coefficient of DHopx/melt = 3 × 10−3 to 4 × 10−3 represented equilibrium. These partition coefficients are comparable with experimental data, even though the pressure was probably much lower in the natural boninite samples. However, it should be noted that the water concentration in the orthopyroxene was quantified using the calibration of Skogby et al. (1990); if the calibration of Bell et al. (1995) is used instead, DHopx/melt is closer to 1 × 10−3. Kurosawa et al. (1997) used SIMS to measure the water and trace element concentrations in mantle olivine crystals from a range of environments. In addition they compared the water concentrations in phenocrysts and xenocrysts in basalts from Oki-dogo and Ichinomegata with the water concentrations of melt inclusions in the phenocrysts. They found that there was a good correlation between the water concentrations in olivine phenocrysts and melt inclusions, independent of total water concentration, and concluded that these pairs were in equilibrium. The water concentrations of the xenocrysts were lower, and were interpreted as not having equilibrated with the melt. The partition coefficients implied by the phenocryst/melt inclusion pairs are in the range DHolivine/melt = 0.013 to 0.022. These values are much higher than most of the experimental measurements and should be treated with some caution. Matveev et al. (2005) also measured water partitioning between olivine and melt by analyzing the water concentration in melt inclusions and their olivine phenocryst hosts. SIMS was used to measure the water concentration of the inclusions and FTIR (using the calibration of Libowitzky and Rossman 1997) was used for the olivine host. Combining the results gave a partition coefficient of DHolivine/melt = 1 × 10−4. This value is 2 orders of magnitude lower than the data of Kurosawa et al. (1997) which were also for phenocrysts in basaltic magmas, and presumably both sets of olivines crystallized at relatively low pressures. The reasons for the

238

Kohn & Grant

discrepancy are not yet clear, and it is possible that calibration problems make some of the data unreliable. A detailed study of megacrysts in the Monastery kimberlite was published by Bell et al. (2004). Very consistent patterns of partitioning between the NAMs were recorded using FTIR, but no melt was directly available for analysis. Bell et al. (2004) estimated the melt water concentration using the concentrations of other incompatible elements and generated the following partition coefficients (with estimated errors of ± 40%): DHcpx/melt = 0.013-0.016; DHopx/melt = 6 × 10−3 – 9 × 10−3; DHolivine/melt = 4.6 × 10−3 – 5.3 × 10−3; DHgarnet/melt = 3 × 10−4 – 1.4 × 10−3. They combined these partition coefficients to give a bulk garnet lherzolite/melt partition coefficient of DHgarnet lhezolite/melt = 0.0051-0.0063. This is not significantly different from the commonly cited value of DHmantle/melt = 0.01 for MORB melting, despite the difference in depth of melting and melt chemistry. A recent study of water in anorthoclase megacrysts and melt inclusions (Seaman et al. 2006) reports an average water concentrations of 130 ppm in the feldspar and 1500 ppm in the melt inclusions. This would imply a partition coefficient of DHanorthoclase/melt = 0.09, although the authors emphasize that the partition coefficient could be much lower for higher water concentrations.

SUMMARY, IMPLICATIONS AND FUTURE RESEARCH The water partitioning data reviewed above illustrate that there are major discrepancies between different studies and that the problem of water partitioning between mantle phases and melts can in no way be considered to be a closed question. Some of the differences have been shown to be related to mineral composition, and others are likely to be related to different temperatures, pressures and analytical techniques. The calibrations for FTIR of NAMs are still uncertain (Bell et al. 1995 2003; Libowitzky and Rossman 1997) and there are even major sources of error in the treatment of backgrounds and baselines in FTIR spectra of NAMs (Grant et al. 2006b,c). The methodology and calibration of SIMS are improving (Aubaud et al. 2006), but all non-spectroscopic techniques for analyzing H in NAMs are subject to contamination by inclusions or lamellae of hydrous phases, which would be hard to detect. NMR is a spectroscopic technique that could provide an additional method for cross calibration (Kohn 1996, 2006), but lacks spatial resolution, so would not be directly applicable to partitioning studies. Development of alternative methods for analysis of H in NAMs would be very valuable. Until these analytical difficulties are resolved it is quite hard to make sense of the variety of partition coefficients that have been reported. Currently there is very little evidence for temperature or pressure effects on partition coefficients, even though Hirth and Kohlstedt (1996) predicted that DHolivine/melt would vary strongly with pressure. More experimental measurements at pressures of 0.01–0.1 GPa and 4–13 GPa are required to test this hypothesis. Table 1 summarizes the most likely values of DHmineral/melt for the shallow mantle. The table also Table 1. Plausible partition coefficients under mantle melting conditions. DHmineral/melt

DCemineral/melt

Olivine

2 × 10−4 – 2 × 10−3

1 × 10−5 – 1 × 10−4

Orthopyroxene Clinopyroxene Garnet

3 × 10 – 0.02 0.01 – 0.02 < 1 × 10−3

1 × 10−3 – 0.02 0.01 – 0.3 0.01 – 0.03

−3

Water Partioning Between NAM’s and Silicate Melts

239

includes values of DCemineral/melt taken from the literature (Green 1994; Kelemen et al. 1993; Salters and Longhi 1999; McDade et al. 2003; Johnston and Schwab 2004), to make a very crude comparison between the individual partition coefficients for H and Ce for each mineral. There is also a considerable range in published partition coefficients for Ce. Within the uncertainties it is quite possible that under specific conditions of pressure, temperature and major element composition (and hence proportions of the different phases) that the bulk partition coefficients for water and Ce could be the same, and equal to 0.01. However, there are many differences in the individual partition coefficients of H2O and Ce; for example under most conditions it is likely that DCecpx/melt> DHcpx/melt, whereas DCeolivine/melt < DHolivine/melt. Furthermore we know of no reason to expect that the pressure, temperature and bulk compositional dependence of DCe and DH would be identical. Caution should therefore be exercised in assuming that Ce and H2O cannot be fractionated during melting and that all variations in H2O/Ce are related to differences in the source. A more detailed exploration of possible fractionation of water from other incompatible trace elements will not be warranted until more, well constrained, partition coefficients for water are established, ideally with K, Ce, La and H2O partitioning measured on the same samples.

ACKNOWLEDGMENTS We would like to thank our colleagues Richard Brooker, Jannick Ingrin, Celine Lemaire and Jens Najorka for many useful discussions on NAMs, Cyril Aubaud and an anonymous reviewer for very helpful comments on the manuscript and Hans Keppler, both for the invitation to participate in this volume and his patience in waiting for the manuscript.

REFERENCES Asimow PD, Dixon JE, Langmuir CH (2004) A hydrous melting and fractionation model for mid-ocean ridge basalts: Application to the Mid-Atlantic Ridge near the Azores. Geochem Geophys Geosys 5 Q01E16 doi: 10.1029/2003GC000568 Asimow PD, Langmuir CH (2003) The importance of water to ocean ridge melting regimes. Nature 421:815820 Aubaud C, Hauri EH, Hirschmann MM (2004) Hydrogen partition coefficients between nominally anhydrous minerals and basaltic melts. Geophy Res Lett 31 L20611, doi:10.1029/2004GL021341 Aubaud C, Pineau F, Hekinian R, Javoy M (2005) Degassing of CO2 and H2O in submarine lavas from the Society hotspot. Earth Planet Sci Lett 235:511-527 Aubuad C, Withers AC, Hirschmann MM, Guan Y, Leshin LA, Mackwell SJ, Bell DR (2006) Intercalibration of FTIR and SIMS for hydrogen measurements in glasses and nominally anhydrous minerals. Am Mineral (submitted) Bell DR, Rossman GR (1992) Water in Earths mantle: The role of nominally anhydrous minerals. Science 255: 1391-1397 Bell DR, Ihinger PD, Rossman GR (1995) Quantitative analysis of trace OH in garnet and pyroxenes. Am Mineral 80:465-474 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res-Solid Earth, 108(B2). 2105 doi 10.1029/2001JB000679 Bell DR, Rossman GR, Moore RO (2004) Abundance and partitioning of OH in a high-pressure magmatic system: Megacrysts from the Monastery kimberlite, South Africa. J Petrol 45:1539-1564 Bercovici D, Karato S (2003) Whole-mantle convection and the transition-zone water filter. Nature 425:39-44 Blundy J, Wood B (2003) Mineral-melt partitioning of uranium, thorium and their daughters. Rev Mineral Geochem 52:59-123 Bureau H, Keppler H (1999) Complete miscibility between silicate melts and hydrous fluids in the upper mantle: experimental evidence and geochemical implications. Earth Planet Sci Lett 165:187-196 Bureau H, Pineau F, Metrich N, Semet MP, Javoy M (1998) A melt and fluid inclusion study of the gas phase at Piton de la Fournaise volcano (Reunion Island). Chem Geol 147:115-130

240

Kohn & Grant

Danyushevsky LV, Eggins SM, Falloon TJ, Christie DM (2000) H2O abundance in depleted to moderately enriched mid-ocean ridge magmas; Part I: Incompatible behaviour, implications for mantle storage, and origin of regional variations. J Petrol 41:1329-1364 Demouchy S, Deloule E, Frost DJ, Keppler H (2005) Pressure and temperature-dependence of water solubility in Fe-free wadsleyite. Am Mineral 90:1084-1091 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 J, Leist L, Langmuir CH, Schilling J-G (2002) Recycled dehydrated lithosphere observed in plumeinfluenced mid-ocean ridge basalt. Nature 420:385-389 Dixon JE, Stolper EM, Delaney JR (1988) Infrared spectroscopic measurements of CO2 and H2O in Juan de Fuca Ridge basaltic glasses. Earth Planet Sci Lett 90:87-104 Dobson PF, Skogby H, Rossman GR (1995) Water in boninite glass and coexisting orthopyroxene— Concentration and partitioning. Contrib Mineral Petrol 118:414-419 Dumas P, Miller L (2003) The use of synchrotron microspectroscopy in biological and biomedical investigations. Vib Spectr 32:3-21 Green TH (1994) Experimental studies of trace-element partitioning applicable to igneous petrogenesis— Sedona 16 years later. Chem Geol 117:1-36 Grant KJ, Ingrin J, Lorand JP, Dumas P (2006a) Hydrogen partitioning between mantle minerals from peridotite zenoliths. Contrib Mineral Petrol (submitted) Grant KJ, Kohn SC, Brooker RA (2006b) Solubility and partitioning of water in synthetic forsterite and enstatite in the system MgO-SiO2-H2O-Al2O3. Contrib Mineral Petrol 151(6):651-664 Grant KJ, Kohn SC, Brooker RA (2006c) Hydrogen solubility and partitioning between olivine, orthopyroxene and melt synthesized in the system albite-forsterite-H2O. Earth Planet Sci Lett (submitted) Hauri EH, Gaetani GA, Green TH (2004) Partitioning of H2O between mantle minerals and silicate melts. Geochim Cosmochim Acta 68:A33-A33 Hirschmann MM (2006) Water, melting and the deep earth H2O cycle. Ann Rev Earth Planet Sci 34:629-653 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: Implications for rheology, melt extraction and the evolution of the lithosphere Earth Planet Sci Lett 144:93-108 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Jambon A (1994) Earth degassing and large-scale geochemical cycling of volatile elements. Rev Mineral 30: 479-509 Johnson MC, Anderson ATJ, Rutherford MJ (1994) Pre-eruptive volatile contents of magmas. Rev Mineral 30: 281-330 Johnston AD, Schwab BE (2004) Constraints on clinopyroxene/melt partitioning of REE, Rb, Sr, Ti, Cr, Zr, and Nb during mantle melting: First insights from direct peridotite melting experiments at 1.0 GPa. Geochim Cosmochim Acta 68:4949-4962 Kawamoto T, Hervig RL, Holloway JR (1996) Experimental evidence for a hydrous transition zone in the early Earth’s mantle. Earth Planet Sci Lett 142:587-592 Kelemen PB, Shimizu N, Dunn T (1993) Relative depletion of niobium in some arc magmas and the continental crust: partitioning of K, Nb, La and Ce during melt/rock reaction in the upper mantle. Earth Planet Sci Lett 120:111-134 Koga K, Hauri E, Hirschmann MM, Bell D (2003) Hydrogen concentration analyses using SIMS and FTIR: Comparison and calibration for nominally anhydrous minerals. Geochem Geophys Geosys 4:1019, doi 101029/2002GC000378 Kohn SC (1996) The solubility of water in nominally anhydrous mantle minerals using 1H MAS NMR. Am Mineral 81:1523-1526 Kohn SC (2006) Structural studies of OH in nominally anhydrous minerals using NMR. Rev Mineral Geochem 62:53-66 Kurosawa M, Yurimoto H, Sueno, S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Minerals 24: 385-395 Lemaire C, Kohn SC, Brooker RA (2004) The effect of silica activity on the incorporation mechanisms of water in synthetic forsterite: a polarised infrared spectroscopic study. Contrib Mineral Petrol 147:48-57 Libowitzky E, Rossman GR (1996) Principles of quantitative absorbance measurements in anisotropic crystals. Phys Chem Minerals 23:319-327 Libowitzky E, Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82:11111115 Matsyuk SS, Langer K (2004) Hydroxyl in olivines from mantle xenoliths in kimberlites of the Siberian platform. Contrib Mineral Petrol 147:413-437 Matveev S, Portnyagin M, Ballhaus C, Brooker R, Geiger CA (2005) FTIR spectrum of phenocryst olivine as an indicator of silica saturation in magmas. J Petrol 46:603-614

Water Partioning Between NAM’s and Silicate Melts

241

McDade P, Blundy JD, Wood BJ (2003) Trace element partitioning on the Tinaquillo Lherzolite solidus at 1.5 GPa. Phys Earth Planet Int 139:129-147 Michael PJ (1988) The concentration, behavior and storage of H2O in the suboceanic upper mantle: implications for mantle metasomatism. Geochim Cosmochim Acta 52:555-566 Michael PJ (1995) Regionally distinctive sources of depleted MORB. Evidence from trace elements and H2O. Earth Planet Sci Lett 131:301-320 Nichols ARL, Carroll MR, Hökuldsson Á (1999) Is the Iceland hotspot also wet? Evidence from the water contents of undegassed submarine and subglacial pillow basalts. Earth Planet Sci Lett 202:77-87 Ohtani E, Mizobata H, Yurimoto H (2000) Stability of dense hydrous magnesium silicate phases in the systems Mg2SiO4-H2O and MgSiO3-H2O at pressures up to 27 GPa. Phys Chem Minerals 27:533-544 Paterson MS (1982) The determination of hydroxyl by infrared-absorption in quartz, silicate glasses and similar materials. Bull Mineral 105:20-29 Peslier AH, Luhr JF, Post J (2002) Low water contents in pyroxenes from spinel peridotites of the oxidized, subarc mantle wedge. Earth Planet Sci Lett 201:69-86 Pineau F, Shilobreeva S, Hekinian R, Bideau D, Javoy M (2004) Deep-sea explosive activity on the Mid-Atlantic Ridge near 34 degrees 50' N: a stable isotope (C, H, O) study. Chem Geol 211:159-175 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Saal A, Hauri EH Langmuir CH (2002) Vapor undersaturation in primitive mid-ocean ridge basalt and the volatile content of the Earth’s upper mantle. Nature 419:451-455 Salters VJM, Longhi J (1999) Trace element partitioning during the initial stages of melting beneath mid-ocean ridges. Earth Planet Sci Lett 166:15-30 Seaman SJ, Dyar MD, Marinkovic N, Dunbar NW (2006) An FTIR study of hydrogen in anorthoclase and associated melt inclusions. Am Mineral 91:12-20 Shen AH, Keppler H (1997) Direct observation of complete miscibility in the albite-H2O system. Nature 385: 710-712 Simons K, Dixon J, Schilling JG, Kingsley R, Poreda R (2002) Volatiles in basaltic glasses from the EasterSalas y Gomez Seamount Chain and Easter Microplate: Implications for geochemical cycling of volatile elements. Geochem Geophys Geosys 3:1039, doi:10.1029/2001GC000173. Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene – variations in the natural environment. Am Mineral 75:764-774 Sobolev AV, Chaussidon M (1996) H2O concentrations in primary melts from supra-subduction zones and midocean ridges: implications for H2O storage and recycling in the mantle. Earth Planet Sci Lett 137:45-55 Stalder R, Skogby H (2002) Hydrogen incorporation in enstatite. Eur J Mineral 14:1139-1144 Stalder R, Ulmer P, Thompson AB, Gunther D (2001) High pressure fluids in the system MgO-SiO2-H2O under upper mantle conditions. Contrib Mineral Petrol 140: 607-618 Stalder R, Klemme S, Ludwig T, Skogby H (2005) Hydrogen incorporation in orthopyroxene: interaction of different trivalent cations. Contrib Mineral Petrol 150:473-485 Sweeney RJ, Prozesky VM, Springhorn KA (1997) Use of the elastic recoil detection analysis (ERDA) microbeam technique for the quantitative determination of hydrogen in materials and hydrogen partitioning between olivine and melt at high pressures. Geochim Cosmochim Acta 61:101-113 Wallace PJ (1998) Water and partial melting in mantle plumes: Inferences from the dissolved H2O concentrations of Hawaiian basaltic magmas. Geophys Res Lett 25:3639-3642 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 Wegden M, Kristiansson P, Pastuovic Z, Skogby H, Auzelyte V, Elfman M, Malmqvist KG, Nilsson C, Pallon J, and Shariff A (2004) Hydrogen analysis by p-p scattering in geological material. Nucl Inst Methods Phys Res B 219-20:550-554 Wegden M, Kristiansson P, Skogby H, Auzelyte V, Elfman M, Malmqvist KG, Nilsson C, Pallon J, Shariff A (2005) Hydrogen depth profiling by p-p scattering in nominally anhydrous minerals Nucl Inst Methods Phys Res B 231:524-529 Workman RK, Hauri E, Hart SR, Wang J, Blusztajn J (2006) Volatile and trace elements in basaltic glasses from Samoa: Implications for water distribution in the mantle. Earth Planet Sci Lett 241:932-951

11

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 243-271, 2006 Copyright © Mineralogical Society of America

The Stability of Hydrous Mantle Phases Daniel J. Frost Bayerisches Geoinstitut University of Bayreuth D-95440 Bayreuth, Germany e-mail: [email protected]

INTRODUCTION It is widely recognized that hydrous minerals are involved in a number of geochemical processes in the Earth’s mantle (Michael 1988; Thompson 1992; Schmidt and Poli 1998). Their presence affects the onset of melting (Lambert and Wyllie 1968; Mysen and Boettcher 1975) and can control the partitioning of trace elements during partial melting (Adam et al. 1993; Ionov and Hoffmann 1995; La Tourrette et al. 1995; Tiepolo et al. 2000). They have accordingly been implicated in the source regions of many types of magmas including alkaline basalts and highly-potassic lavas (Edgar and Vukadinovic 1992; Foley 1992; Halliday et al. 1995; Yang et al. 2003; Elkins-Tanton and Grove 2003; Conceicao and Green 2004). As well as being intimately linked with the occurrence of mantle metasomatism (Bailey 1982; Roden and Murthy 1985), hydrous phases can also buffer fluid compositions in the mantle (Eggler 1978; Wyllie 1978) and consequently dictate the style of metasomatism. The dehydration or melting of hydrous minerals in subducting lithosphere and associated infiltration of hydrous solutions (fluids or melts) into the overlying mantle wedge are important steps in the production of island arc magmatism (Tatsumi et al. 1986; Kushiro 1987; Ulmer 2001). Conversely, the persistence of some hydrous minerals in cold regions of subduction zones may result in the transport of hydrogen into the deep mantle (Bose and Ganguly 1995; Kawamoto et al. 1995). It is therefore important to ascertain the conditions at which hydrous minerals are stable in the mantle and to recognize the situations where they breakdown and release H2O. The presence of H2O in the upper mantle can lead to the production of melts that may either rise to the surface or migrate, accumulate and possibly crystallize at depth. In either case this results in chemical fractionation within the mantle. Hydrous minerals such as amphiboles and micas are found in mantle nodules brought to the surface mainly by alkaline basalt and kimberlite lavas (Frey and Prinz 1978; Erlank et al. 1987). Most of these nodules are pieces of the subcontinental lithospheric mantle (Wilshire and Shervais 1975; Witt-Eickschen et al. 1993) although hydrous minerals are also found in nodules from the suboceanic lithosphere (Hauri et al 1993; Gregoire et al. 2000). Hydrous minerals form as a result of the interaction between the lithosphere and incoming H2O-bearing melts or fluids principally from the underlying asthenosphere (Menzies et al. 1987). High pressure and temperature laboratory experiments are crucial for understanding the conditions at which such fluids or melts form, interact and crystallize in the lithosphere. Hydrous minerals don’t seem to be recovered from the underlying asthenospheric mantle, which may have two reasons. 1) Melts generally arise in the top of the asthenosphere and therefore only bring samples from the overlying lithosphere. 2) Temperatures in the asthenosphere are mostly outside of the stability fields of hydrous minerals. Laboratory experiments, however, indicate that dehydration and melting temperatures of some dense hydrous minerals increase with pressure and the results of these experiments can be used to examine the likely role of hydrous minerals in H2O storage throughout the mantle. 1529-6466/06/0062-0011$05.00

DOI: 10.2138/rmg.2006.62.11

244

Frost

In this chapter I will address the stability of nominally hydrous minerals in the lithospheric and underlying asthenospheric/convecting mantle. The metasomatic processes that lead to the formation of hydrous minerals will be discussed, followed by a review of the types of hydrous minerals found in mantle samples from nodules and alpine peridotite massifs. The stability fields of known mantle hydrous minerals as determined by high pressure and temperature experiments will then form a foundation for a discussion of the likely stability of hydrous minerals at pressures beyond those from where xenoliths originate. These results will be used to examine the stability of hydrous minerals throughout the mantle with an emphasis on the lithosphere and what might be termed the ambient convecting mantle, as opposed to subduction zones, which are covered specifically in this volume by Kawamoto (2006).

MANTLE METASOMATISM Peridotite nodules from the continental Achaean lithosphere brought to the surface by kimberlites provide good evidence for the action of mantle metasomatism (Jones et al. 1982; Erlank et al. 1987). Such ultramafic xenoliths are often strongly depleted in basalt-forming major elements as a result of the removal of high temperature partial melts (O’Hara and Mercy 1963; Boyd and Mertzman 1987). They have been variably enriched, however, in some of the most incompatible trace elements, such as light rare earths, that would have been strongly depleted by melt extraction (Shimizu 1975; Hoal et al 1994). This incongruous behavior of major and trace elements is explained as the action of metasomatism, whereby a region is enriched in incompatible elements by the influx of fluids or melts (Bailey 1982; Roden and Murthy 1985; Menzies et al. 1987). The formation of new, frequently hydrous minerals as a result of the metasomatic influx is termed modal (Harte 1983) or patent (Dawson 1984) mantle metasomatism. Anhydrous minerals such as clinopyroxene, magnetite, sphene or sulfides may be also added but modal metasomatism is generally characterized by hydrous minerals. On the other hand, cryptic metasomatism is often used to describe rocks that are clearly trace element enriched but contain no obviously new metasomatic minerals. Evidence for metasomatism is found in xenoliths from the oceanic and continental lithosphere and in a range of different volcanic settings. Dating of the enrichment processes, which can be performed using isotopic systems such as Sm/Nd and U-Pb, has shown that some events occurred in the subcontinental lithosphere a very long time ago, i.e., >1 Ga (Hawkesworth et al. 1983; Cohen et al. 1984; Kinny et al 1989; 1994). There are also examples where the timing of metasomatism in xenoliths carried by alkali basalts cannot be separated from the magmatic event that brought the samples to the surface (Menzies and Murthy 1980). The fluids or melts that caused the chemical changes are only rarely found as crystallized melts or glasses in mineral inclusions (Schrauder and Navon 1994). In some composite xenoliths and alpine massifs metasomatism can be observed in the wall rocks adjacent to veins and dikes (Jones et al. 1982; Boyd 1990; Woodland et al. 1996). Small degree melts that crystallize in the mantle and expel fluids into the wall rocks are implicated in many metasomatic events. Some alpine peridotite massifs, however, seem to have experienced phases of pervasive metasomatism over large regions, with no apparent relationship to veining (Zanetti et al. 1999; Scambelluri et al. 2006) Metasomatism can occasionally be directly attributed to subduction zones processes, where the slab provides both fluids and a source of incompatible elements (Brandon and Draper 1996; Zanetti et al 1999; McInnes et al. 2001; Scambelluri et al. 2006). Although many metasomatized regions of the subcontinental lithosphere have not been at convergent margins for over a billion years, it is highly likely that components added by ancient subduction could be remobilized at a later date as a result of heating or decompression. In many instances metasomatism is more directly attributed to the infiltration and crystallization of small degree melts that migrate from the asthenosphere as a result of plume activity or decompression due to rifting. Menzies et al

Stability of Hydrous Mantle Phases

245

(1987) proposed that in general terms the chemical changes that are induced in the lithosphere occur because it forms a mechanical barrier between asthenospheric melts and the surface. A number of geochemical and laboratory based studies have addressed the nature of metasomatic agents in the mantle (Ryabchikov and Boettcher 1980; Schneider and Eggler 1986; McNeil and Edgar 1987; Bodinier et al. 1988; Gregoire et al. 2003) but it is often difficult to categorically attribute natural metasomatic assemblages to specific fluid or melt compositions. H2O-rich fluids are often implicated as metasomatic agents but the conditions where they can exist in the mantle are constrained to relatively low temperatures (e.g., <900 °C between 2-3 GPa) by the H2O-saturated peridotite solidus (Mysen and Boettcher 1975; Kawamoto 2004). Experiments performed at approximately 2 GPa show that H2O-rich fluids in equilibrium with peridotite are poor in mafic components such as Mg, Fe and Ti and rich in Si, Na and K, with Na+KAl) (Schneider and Eggler 1986). H2O-CO2 fluid compositions are likely to be buffered by the presence of hydrous and carbonate minerals in the mantle (Eggler 1978; Wyllie 1978). At pressures above 2 GPa the carbonates magnesite and dolomite are stable in the mantle and coexisting H2O-CO2 fluids will become H2O-rich. Below 2 GPa carbonates breakdown and as amphiboles are stable, fluids are likely to be more CO2-rich. The association of mantle metasomatism with either fluids or silicate melts may become indistinct at high pressures because there may be no sharp division between silicate-rich fluids at lower temperature and volatile-rich silicate melts at higher temperatures (Shen and Keppler 1997; Bureau and Keppler 1999). Never the less, higher temperature hydrous silicate melts are clearly more effective metasomatic agents as they have higher solubilities of major and trace elements (Schneider and Eggler 1986; Adam et al. 1997). As will be seen in the next section there is abundant field evidence for the passage and crystallization of small degree hydrous silicate melts in the mantle. Carbonatite melts may also be effective metasomatic agents (Wallace and Green 1988; Thibault et al. 1992; Hauri et al. 1993; Yaxley et al. 1998). Such melts that can form at pressures above 2 GPa are enriched in Na, K, Rb, Sr and P and very low in Si and high field strength elements like Ti (Wallace and Green 1988). Migrating melts would decarbonate in a reaction with enstatite at pressures below 2 GPa, releasing CO2 vapor and producing clinopyroxene and olivine. The metasomatized assemblage would have higher Ca/Al and Na/Ca but as carbonatite melts have Mg/Fe ratios similar to mantle minerals the assemblage is not richer in Fe, which is counter to the effects of silicate melt infiltration as observed in some xenoliths (Yaxley et al. 1998).

246

Frost

Natural samples frequently show evidence for multiple generations of different metasomatic events (e.g., Woodland et al. 1996; Witt-Eickschen and Kramm 1998). From assessment of the temperature at which metasomatic events took place it is in some cases possible to make reasonable assumptions as to the types of agents involved. Scambelluri et al. (2006) for example identified two phases of metasomatism of the Ulten Zone (Eastern Italian Alps) peridotite, which likely formed part of the mantle wedge above a subduction zone. Cryptic metasomatism of spinel peridotite causing large ion lithophile and light rare earth enrichment is attributed to silicate melt percolation because it occurred at temperatures above the wet peridotite solidus. This was followed, however, by a later stage of enrichment that must have occurred at higher pressure and approximately 850 °C because hydrous minerals and garnet were formed. As this is below the wet solidus the agent was most likely an H2O-rich fluid. There is a final aspect in the definition of metasomatism that becomes very important when addressing the stability of hydrous minerals but is less important and generally ignored when the term metasomatism is used to explain source enrichment in many geochemical studies. If mantle metasomatism is considered to result from a chemical reaction between a rock, e.g., peridotite, and an incoming agent, e.g., H2O-rich fluid, then the bulk composition in which a hydrous mineral forms, although the system is open, is presumably closer to that of the rock. Therefore in an experimental study on hydrous mineral stability under these conditions, a peridotite plus H2O bulk composition may be more appropriate, perhaps with the recognition that concentrations of elements such as Na, K or Ti may be raised (e.g., Mengel and Green 1989). However, it would seem that many geochemical signatures in erupted lavas that are attributed to source metasomatism i.e., that require trace element enrichment and/or residual hydrous minerals in the source (e.g., Halliday et al. 1995), could originate from partial melting of crystallized hydrous silicate melt veins or dykes in the mantle (Foley 1992). Metasomatism, in the form of wall rock reaction, is often observed around such crystallized melt veins but in some instances it is volumetrically very small or entirely absent (e.g., Woodland et al. 1996). When considering hydrous mineral stability in the veins themselves the bulk composition to be considered is that of the silicate melt. This presents a serious problem for experimental studies on the formation conditions because such veins probably do not represent crystallization from a single liquid composition but are formed instead by accumulation as melts differentiate in the mantle (Foley 1992). If the vein crystallizes without reacting significantly with wall rocks this cannot, in the petrological sense or on the scale of the vein itself, be considered to be a metasomatic event. However, when considered at the broader scale of a mantle-melting event, the crystallization of such veins without wall rock interaction is at least still metasomatism in the sense that mass transfer has occurred into the source region of melting. In this chapter there is a focus on hydrous mineral stability in peridotitic assemblages as this is the dominant rock type in the mantle. The stability of a hydrous mineral as it crystallizes from a silicate melt, however, can be quite different and several such examples of known recurrent vein compositions are discussed, such as alkaline silicate melt veins or MARID rocks. It must be appreciated, however, that a significant amount of water may be concentrated in such veins particularly in the lithospheric mantle and it is quite possible that hydrous minerals exist in deeper regions of the mantle within unknown melt vein compositions.

EVIDENCE FROM MANTLE XENOLITHS Hydrous mantle minerals occur in spinel peridotite xenoliths that are generally found in alkaline basalts and in garnet peridotite nodules that are more typical of kimberlite lavas. Hydrous phases are also found in ultramafic rocks from alpine type peridotite massifs. As the compositions and types of hydrous minerals vary between spinel and garnet peridotites it is helpful to consider these rock types separately.

Stability of Hydrous Mantle Phases

247

Peridotite massifs and xenoliths from alkaline basalts Two groups of ultramafic nodules have been identified from alkaline basalts, and while this classification is not all encompassing, it is useful in this context because different hydrous mineral occurrences are associated with each group. Termed group I by Frey and Prinz (1978) or Cr-diopside by Wilshire and Shervais (1975) these rocks are typical mantle peridotites, dominated by olivine with Cr-rich spinels and clinopyroxenes. These rocks have Mg# > 85 and are generally Ti-poor. They occur with a wide range of fertility from harzburgite to lherzolite and detailed studies often reveal complex histories involving melt extraction, metasomatism and reaction with the host magma during transport to the surface. Where hydrous minerals occur they are generally the calcic amphibole pargasite, (often replacing spinel), and phlogopite mica. Group II (Frey and Prinz 1978) or Al-augite rocks (Wilshire and Shervais 1975) occur as clinopyroxene dominated veins or layers in xenoliths. They are Ti-rich, Cr-poor and have Mg# < 85. They frequently contain the Ti-rich calcic amphibole kaersutite and more occasionally phlogopite. Kaersutite of similar composition can occur as monomineralic veins and is also found as megacrysts in alkaline lavas that have been interpreted as disaggregated mantle veins (Wilkinson and Le Maitre 1987). Similar relations can be seen in alpine peridotites such as the lherz massif. Woodland et al. (1996) for example describe pyroxenite and hornblendite dikes similar to the group II/Al-augite series cross cutting spinel lherzolite. Ti-rich pargasite and kaersutite form within the dykes whereas metasomatism of the wall rocks is revealed by growth of pargasite with lesser amounts of phlogopite and a general increase in Fe content. This type of wall rock interaction that is prevalent in spinel peridotites has been termed Fe-Ti metasomatism (Menzies et al. 1987) and has been attributed to late stage crystallization of dykes and veins containing alkaline silicate melt similar to basanite. Chemical gradients develop as a response to fluid infiltration of the wall rock with zones of modal metasomatism, giving way to wider zones of trace element enrichment. As previously discussed, however, wall rock metasomatism does not always occur around such veins and in some instances the contacts with the wall rocks are sharp (Woodland et al. 1996). These relatively simple relations for spinel peridotites are not, however, without exceptions. O’Reilly and Griffin (1988) for example report Ti-poor and Ti-rich pargasite forming in both veins and wall rocks in xenoliths from Victoria, Australia. They attribute metasomatism to CO2-rich fluids. The occurrence of apatite Ca5(PO4)3(OH,F,Cl) in mantle rocks is also frequently attributed to CO2 rich fluids or carbonatite melts (O’Reilly and Griffin 1988; Chazot et al. 1996; Woodland et al 1996). O’Reilly and Griffin (2000) argue that apatite is more widespread in lithospheric xenoliths than generally accepted as it is often overlooked or removed by acids during sample preparation. Apatite is a major host for trace elements such as Sr, Th, U and rare earths as well as P and F. O’Reilly and Griffin (2000) report that apatite found in group I type Cr diopsite lherzolites (which they term type A apatites) have significant CO2 substitution (0.7-1.7 wt%) and generally higher Cl contents than F. Whereas in veins, megacryts and group II type Alaugite rocks more F-rich hydroxyl-fluor apatites (termed B-type) are found with undetectable CO2 contents.

Xenoliths from kimberlites Metasomatism in garnet peridotite xenoliths found in kimberlites is frequently attributed to K and H2O-rich fluids as more K-rich hydrous minerals generally occur. The main hydrous mineral found is phlogopite, which is often Mg-rich and Ti-poor in comparison to its occurrence in spinel peridotites. In addition the K-rich and Al-poor amphibole K-richterite can occur and pargasitic amphiboles occur in some xenoliths. Erlank et al. (1987) recognized a series of assemblages in xenolith samples from the Kimberley cluster of kimberlites in South Africa, which they reasoned reflected increasing degrees of metasomatism of a garnet

248

Frost

peridotite precursor. Progressive metasomatism is documented by the replacement of garnet by phlogopite and eventually growth of K-richterite. K is introduced while Al decreases. Garnet peridotite rocks (GP) are thus succeeded by garnet phlogopite peridotites (GPP), phlogopite peridotites (PP) and ultimately phlogopite K-richterite peridotites (PKP). PKP rocks also exhibit TiO2 and Fe2O3 enrichments. Similar sequences have been reported for rocks from other kimberlite localities (van Achterbergh et al. 2001). Erlank et al. (1987) estimate that garnet phlogopite peridotites (GPP), that contain ca. 1% phlogopite or less, comprise 50% of the sampled peridotite xenoliths, whereas phlogopite peridotites (PP) with over 1% phlogopite comprise 30%. Hydrous minerals also occur in veins in kimberlite peridotite xenoliths. Veins are generally dominated by phlogopite and diopside or K-richterite, phlogopite and diopside (Jones et al 1982; Erlank et al. 1987). Another group of xenoliths that occur in kimberlite lavas worldwide are the MARID (mica-amphibole-rutile-ilmenite-diopside) suite of rocks (Dawson and Smith 1977; Wagner et al 1996). These rocks are dominated by phlogopite but compared with peridotite xenoliths, phlogopites and K-richterite amphiboles are generally higher in Fe and lower in Cr. An igneous cumulate origin for MARID rocks is often argued (Sweeney et al. 1993; Konzett et al. 1997) and several studies point to the similarity between MARID rocks and group II (orangeite) micaceous kimberlite lavas (Jones 1989; Ulmer and Sweeney 2002). It has also been proposed that the aforementioned garnet peridotite metasomatic suite described by Erlank et al. (1987) could have been formed by hydrous-fluids expelled by crystallizing MARID rocks (Jones et al. 1982; Menzies et al. 1987; Jones 1989), creating a similar relationship to group I and II xenoliths from alkaline basalts.

Mantle amphibole mineralogy Even in ultramafic systems amphiboles are complex solid solutions, which contain all major elements in significant proportions. There are a number of detailed reviews of amphibole chemistry, structure and nomenclature (Thompson et al. 1981; Robinson et al. 1982; Leake et al. 1997). The standard amphibole formula is A0−1B2C5T8O22(OH)2, although in the absence of OH measurements formulae are normally reported on the basis of 23 oxygens. The important substitutions with respect to mantle amphiboles can be considered relative to tremolite Ca2Mg5Si8(OH)2 where the A site is vacant, Ca fills B, Mg is in C and Si in T. Replacing Mg and Si for 2Al constitutes the tschermakite substitution, which can be described by the exchange vector Al2Mg−1Si−1 (Thompson et al. 1981). The edenite substitution of Na or K onto the previously vacant A site is accompanied by Al replacing Si (NaAl−1Si−1), whereas replacement of Ca by Na in the B site results in the NaAlCa−1Mg−1 glaucophane or NaNa−1Ca−1 richterite substitutions. The resulting basic amphibole classification is shown in Figure 1a with shaded boxes indicating regions that contain most mantle amphibole compositions i.e., pargasites and K-richterites. Figure 1b shows this in more detail for mantle samples in terms of (Na+K) A site occupancy and Si content. K-richterites aside, compositions from most spinel peridotite localities and alkaline basalt xenoliths cluster between pargasite and edenite but generally have a tschermakite component and occupancy of the B site by Na can be up to 40%. In addition to K-richterite, pargasite-edenite amphiboles also occur in some kimberlite xenoliths. Amphiboles in peridotite xenoliths from Jagersfontein kimberlites have been reported with compositions that span the region between pargasite and K-richterite, with some amphiboles even approaching Mg-kataphorite (Field et al. 1989). This broad compositional range may be inherited from chemical variations in the original peridotite. In Figure 1b amphibole compositions from the Finero (Zanetti et al. 1999) and Nonsberg (Obata and Morten 1987) complexes in the Alps, which may exhibit subduction related metasomatism, show the widest scatter in compositions, whereas kaersutites have a relatively narrow compositional range. The difference between pargasitic amphiboles associated with group I (Cr-diopside) and group II (Al-augite) kaersutites, which are defined as having

Stability of Hydrous Mantle Phases

249

B

A

NaNa2Mg4AlSi8O22(OH)2 Eckermannite

a (N

+K

)A

Ri 1

Na2Mg3Al2Si8O22(OH)2 Glaucophane

2

Ed

Pa

NaCaNaMg5Si8O22(OH)2 NaCaNaMg4AlSi7AlO22(OH)2 NaCaNaMg3Al2Si6Al2O22(OH)2 Taramite Richterite Katophorite

1

CaNaMg4AlSi8O22(OH)2 Winchite

1 NaB

K a+ (N

CaNaMg3Al2Si7AlO22(OH)2 Barroisite

NaCa2Mg5Si7AlO22(OH)2 Edenite

Spinel peridotites Alpine peridotites Kaersutites Finero-Nonsberg Oceanic K-Richterites

NaCa2Mg4AlSi6Al2O22(OH)2 Pargasite

)A

8 Si Si Ca2Mg5Si8O22(OH)2 Tremolite

(Na+K)

A

NaB

7 Si Ca2Mg4AlSi7AlO22(OH)2 Hornblende

6 Si

Hb

Tr

0

Ca2Mg3Al2Si6Al2O22(OH)2 Tschermakite

8

C

Ts

7 Si

6

D 1.5

1.2

Pargasites Kaersutites 1

1

K

Fe

0.8

0.5

Spinel Peridotites Alpine Peridotites Kaersutites Finero-Nonsberg Oceanic K-Richterites

0.6

0.4

0.2

0

0

0

0.1

0.2

0.3

Cr

0.4

0.5

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Na

Figure 1. (A) Amphibole compositional space in the NCMASH system. The coordinate axes are Si formula units (which follows the tschermakite exchange vector Al2Mg−1Si−1), Na (and K) occupancy of the A site (NaMgAl−1) and Na in the B site (NaAlCa−1Mg−1). The compositions of mantle amphiboles generally cluster in the grey-boxed regions. (B) Mantle amphibole compositions plotted as Si formula units (23 oxygens) versus Na+K in the A site, which is a projection onto the basal plane of Figure 1A. In the legend spinel peridotites refer to xenolith samples from alkaline basalts. Finero and Nonsberg refer to peridotite bodies from the Italian Alps that may have been sections of mantle wedge. (C) Formula units of Fe versus Cr for kaersutite amphiboles from group II type xenoliths compared with low Ti pargasitic amphiboles from all localities shown in Figure 1B. (D) K versus Na contents, in formula units, of amphiboles from various localities. Amphibole compositions for all figures are from Menzies et al. (1987), Dawson and Smith (1982), Wilkinson and Le Maitre (1987), Yaxley et al. (1998), O’Reilly and Griffin (1988), Ionov (1998), Obata (1980), Seyler and Mattson (1989), Woodland et al. (1996), Fabries et al. (2001), Frey and Prinz (1978), Zanetti et al. (1999), Obata and Morten (1987), McInnes et al. (2001), Agrinier et al. (1993), Arai et al. (1997), Gregoire et al. (2000), Gregoire et al. (2003), Dawson and Smith (1977), Waters (1987), Jones et al. (1982) and Erlank et al. (1987).

250

Frost

more than 0.5 formula units of Ti but are otherwise similar to pargasites, can be seen in Figure 1c. In addition to higher Ti, kaersutites have generally low Cr and high Fe, which is consistent with their proposed origin as crystallizing from low degree partial melt of alkaline basalt composition (e.g., basanite). Most other amphiboles have Cr contents that are likely inherited from the protolith. Amphibole Na and K contents are shown in Figure 1d. Pargasitic amphiboles show clear differences in Na/K depending on locality whereas K-richterite compositions scatter around the general formula KCaNaMg5Si8O22(OH)2. Studies have shown (Dyar et al. 1993; Popp et al. 1995; King et al. 1999) that Ti and Fe3+ bearing mantle amphiboles typical of Group II type xenoliths and megacrysts are hydrogen poor as a result of oxy-amphibole substitutions such as, Al3+ + OH− ↔ Ti 4+ +O2− + 0.5H2 Fe2+ + OH− ↔ Fe3+ + O2− + 0.5H2

(1) (2)

The substitution of higher valence cations onto the C site is charge balanced by the loss of hydrogen. A number of competing substitutions and the effects of closure, as discussed by Young et al. (1997), whereby two elements show correlation simply because site occupancy must add up to a constant sum, can make substitutions like (1) and (2) difficult to separate using compositional variations.

Mantle mica mineralogy The biotite micas, (general formula X2Y6Z8O20[OH]4), that occur in ultramafic xenoliths are dominated by the K2Mg6[Si6Al2]O20(OH)4 phlogopite end member but have up to 30% K2Mg4Al2[Si4Al4]O20(OH)4 eastonite substitution, which is the mica Tschermak’s component (Al2Mg−1Si−1). Formulas are generally reported for 22 (or 11) oxygens in the absence OH measurements. Some phlogopite compositions clearly have Al < 2 which requires a highSi end member that could be either the tetrasilicic mica montdorite K2Mg5[Si8O20](OH)4 (Konzett and Ulmer 1999) or talc Mg6[Si8O20](OH)4. Phlogopites from spinel peridotite nodules show generally a greater degree of eastonite substitution than garnet peridotites. For phlogopites from garnet peridotites the eastonite component decreases with increasing degree of metasomatism (Fig. 2a) reflecting progressive Al depletion of the host rock (Erlank et al. 1987). Although in Figure 2a MARID phlogopites appear similar to those in garnet peridotites, as shown in Figure 2b phlogopites in MARIDs actually have lower Mg# (=100·Mg/[Mg+Fe]) than those from garnet peridotites. In xenoliths from alkaline basalts high Ti and low Mg# phlogopites reflect group II type vein and megacryst occurrences, whereas Mg# ca. 90 and low Ti contents are typical of group I spinel-lherzolite xenoliths. These variations can be viewed as evidence that MARID and group II rocks likely crystallized from silicate melts. The cation sum for the octahedral (Y) site in biotites (i.e., Y = Mg + Fe + AlVI + Ti where AlVI = AlTotal − AlIV and AlIV = 8 − Si) is often less than the ideal 6, which has been attributed to vacancies in the octahedral site that balance divalent cation substitution by higher valence cations (see Fleet 2003). Al for example may substitute for Mg in the Y site with charge balance provided by the creation of a vacancy (i.e., Al2Mg−3). For mantle phlogopites shown in Figure 2c the octahedral cation deficit shows some correlation with Ti content, which may result from an octahedral Ti-vacancy substitution i.e., TiMg−2 for which Trønnes et al. (1985) found evidence in experiments on Ti-rich systems. There is, in addition, evidence that mantle phlogopites range in OH concentration to values both slightly higher and significantly lower than the ideal 4. Matson et al. (1986) analyzed H2O contents of phlogopites from kimberlite xenoliths and found an excess of OH cations of over 0.5 that correlated with cation deficiency in octahedral and tetrahedral sites. This implies a similar OH substitution mechanism to some nominally anhydrous minerals. On the other hand Righter et al. (2002) report that most mantle phlogopites have a deficiency in OH that implies a relatively large oxy-component via the involvement of substitutions like Equations (1) and (2).

Stability of Hydrous Mantle Phases 5

0.7

A

B

0.6

251

Spinel Peridotites Garnet Peridotites MARIDs

4 0.5

3

Group II

Ti

Al

0.4

GPP PP PKP

2

0.3 0.2

1

MARIDs Garnet peridotites Spinel peridotites Erlank et al. (1987)

0.1 0

0 4

4.5

5

5.5

MARIDs

6

6.5

7

75

Si

0.7

C

0.6

Spinel Peridotites Garnet Peridotites MARIDs

0.5

Ti

0.4 0.3 0.2 0.1 0

5

5.5

6

6.5

7

80

85 90 95 Mg# (100.Mg/[Mg+Fe])

100

Figure 2. (A) Natural phlogopite compositions plotted in terms of Si versus total Al in formula units. The eastonite substitution (Al2Mg−1Si−1) is shown by the diagonal curve. Average mica compositions for garnet phlogopite peridotite (GPP), phlogopite peridotite (PP) and phlogopite-K-richterite peridotite (PKP) from Erlank et al. (1987) are shown. (B) Phlogopites from MARID rocks and group II type veins and megacrysts from alkaline basalt xenoliths have higher Ti contents and lower Mg# than micas from garnet peridotites or group I type spinel lherzolite rocks. (C) Total cations in the phlogopite octahedral site frequently sum to less than the ideal 6. This deficit shows some correlation with Ti content. Phlogopite compositions for all figures are from Shaw and Eyzaguirre (2000), Menzies et al. (1987), Dawson and Smith (1977), Wagner et al. (1996), Jones et al. (1982) and Erlank et al. (1987).

EXPERIMENTAL STUDIES ON THE STABILITY OF KNOWN MANTLE HYDROUS MINERALS Pargasitic amphiboles There have been a significant number of experimental studies on the stability of amphiboles in simple and complex systems and over a range of water activities (see Gilbert et al. 1982). Although the experimental study of natural amphibole compositions is important, in order to identify the main factors that control amphibole stability it is often necessary to reduce the number of variables by considering simplified systems and end-members. The simplest amphibole end member relevant to ultramafic assemblages is tremolite, which has an upper thermal stability in the presence of forsterite that is controlled by the reaction: Ca2Mg5Si8O22(OH)2 + Mg2SiO4 = 2.5Mg2Si2O6 + 2CaMgSi2O6 + H2O amphibole

olivine

orthopyroxene

clinopyroxene

(3)

fluid

The equilibrium curve for this reaction is shown in Figure 3 along with stability relations of other calcic amphiboles. In comparison to tremolite, the major amphibole substitutions all lead to larger P-T stability fields. Jenkins (1983) showed that adding 20-30 mol% of the tschermakite component (Ca2Mg3Al2Si6Al2O22(OH)2) expands the stability field by a modest amount. The coupled pargasite substitution of Na and Al, however, expands the stability field

252

Frost 2.5

Pressure (GPa)

2.0

1.5 Ti-rich Pargasite (Kaersutite)

A

1.0

G

B Trem + Fo

0.5

0

H

AlTrem + Fo

E

700

800

F

Parg + Opx

Parg

C

D

900 1000 1100 Temperature (oC)

1300

Figure 3. Calcic-amphibole stability relations. The curve labeled (A) is the dehydration of tremolite coexisting with forsterite via the reaction tremolite + forsterite = orthopyroxene + clinopyroxene + fluid (Jenkins 1983). (B) Dehydration of tremolite with 20-30 mol% of the tschermakite component in equilibrium with forsterite (Jenkins 1983). (C) The high temperature breakdown of pure pargasite (NaCa2Mg4Al(Al2Si6)O22(OH)2) by the melting reaction pargasite = clinopyroxene + forsterite + spinel + melt (Holloway 1973). (D) Low pressure breakdown of pure pargasite through the reaction pargasite = clinopyroxene + forsterite + spinel + nepheline + anorthite + fluid (Boyd 1959). (E) The breakdown of pargasite in the presence of orthopyroxene by the reaction pargasite + orthopyroxene = clinopyroxene + forsterite + plagioclase + fluid and by (F) the higher pressure reaction pargasite + orthopyroxene = clinopyroxene + forsterite + melt (Lykins and Jenkins 1992). (G) the stability of Ti-rich pargasite (magnesio-hastingsite) and (H) kaersutite (Huckenholtz et al. 1992; Merrill and Wyllie 1975). Both curves are from experiments performed on natural complex bulk compositions (K, Fe3+ and Ti-rich magnesio-hastingsite and a kaersutite megacryst respectively). In both compositions the amphiboles breakdown above the solidus via melting reactions that involve olivine, clinopyroxene and spinel.

by over 200 as determined by Holloway (1973). Pure pargasite has a higher thermal stability than most other amphiboles and consequently its breakdown reaction at pressures above a few kilobars produces a silicate melt rather than a fluid phase. In ultramafic systems Lykins and Jenkins (1992) proposed that pargasite would breakdown in a reaction with coexisting orthopyroxene: NaCa2Mg4Al3Si6O22(OH)2 + 2.5Mg2Si2O6 pargasite

orthopyroxene

= 4Mg2SiO4 + CaMgSi2O6 + 2(Na0.5Ca0.5)(Al1.5Si2.5)O8 + H2O olivine

clinopyroxene

plagioclase

(4)

fluid

The bracketed equilibrium curve for this reaction is shown in Figure 3. The thermal stability of pargasitic amphiboles is further increased by the addition of Ti to produce kaersutite, which has the ideal formula NaCa2Mg4TiSi6Al2(O+OH)24. Experiments on kaersutite have generally been performed in complex systems, such as natural kaersutite megacrysts or alkaline basalts, where Ti and Fe3+ may both act to increase the thermal stability

Stability of Hydrous Mantle Phases

253

(Huckenholz et al. 1992; Merrill and Wyllie 1975). Kaersutites breakdown via melting reactions that generally involve olivine, clinopyroxene and spinel. The significant hydrogen deficiency caused by the oxy-substitution involving the high field strength Ti and Fe3+ ions in natural kaersutites (hence the sum of O+OH to 24 in the formula) can also help to explain their high thermal stability. Popp et al. (1995) performed experiments on a natural kaersutitic amphibole under controlled P, T and hydrogen fugacity (ƒH2) up to 1 GPa and 1200 °C and determined the equilibrium constant (K) for amphiboles undergoing dehydrogenation by Equation (2). Using the expression for K(P,T) and measurements of the OH content and ferric/ferrous ratio it is therefore possible to make a calculation of the ƒH2 under which a natural sample equilibrated. If the equilibrium oxygen fugacity can be also estimated using an oxy-thermobarometer (e.g., Bryndzia and Wood 1990) then the equilibrium water activity (aH2O) can be determined. Mysen et al. (1998) used a kaersutite inclusion in an SNC meteorite of likely Martian origin to estimate that it crystallized from a melt with only 100-1000 ppm H2O. A recurrent question, however, is whether low H contents reflect crystallization conditions or if the dehydrogenation reaction, Equation (2), occurs mainly during ascent and cooling of the xenolith. Dyar et al. (1993) argued that hydrogen diffusivity would be too slow for significant dehydrogenation to occur during ascent, while Popp et al. (1995) reasoned that a rock or melt containing amphibole would be more likely to reduce during closed system cooling and would, therefore, not undergo dehydrogenation. King et al. (1999) and Miyagi et al. (1998), on the other hand, recognized a correlation between amphibole H-content and xenolith ascent rate, whereby amphiboles from rapidly cooled rocks have higher H-contents, but slower cooled rocks have lower and more scattered values. Some natural amphiboles have such low H contents (Dyar et al. 1993; King et al. 1999) that it is difficult to understand why these amphiboles were stable with respect to an anhydrous assemblage if these compositions reflect crystallization conditions. In addition to the chemical composition, the activity of H2O plays a key role in the stability of amphiboles, as with any hydrous mineral. At the relatively low temperatures at which an amphibole such as tremolite dehydrates the resulting fluid will be relatively pure H2O. Lowering the activity of H2O in the fluid, by adding CO2 for example, will decrease the high temperature stability of tremolite, which is, therefore, at a maximum in the presence of a pure H2O fluid. Pargasitic and kaersutitic mantle amphiboles, however, are stable at temperatures above the H2O-saturated silicate solidus and the melting reactions that control their stability are, therefore, displaced to lower temperatures with increasing H2O activity. Contrary to the tremolite thermal stability range, the maximum thermal stability of these amphiboles occurs under fluid-absent conditions (i.e., PH2O < Ptotal). For the same reason experiments saturated in H2O-CO2 fluids below 1 GPa (Holloway 1973) show that pargasite thermal stability increases as H2O activity in the vapor phase decreases (Fig. 4). This is because the H2O activity in the coexisting melt phase is also decreasing, with the maximum thermal stability occurring where the H2O activity of the silicate melt is similar to that which occurs at the onset of vapor free melting. Whereas increasing CO2/H2O ratio of the fluid expands the pargasite stability field in the low-pressure, fluid-saturated experiments of Holloway (1973), the increased CO2 solubility in silicate melts at pressures above 2 GPa lowers the melting temperature with increasing CO2/H2O ratio at 2-3 GPa (Wallace and Green 1988), and thus destabilizes pargasite. Under more reducing conditions, however, increasing the CH4 content of a C-O-H fluid phase raises the solidus temperature, as a result of lowering the melt H2O activity, causing the amphibole stability field to expand relative to pure H2O saturated conditions (Taylor and Green 1988). A number of studies have examined amphibole stability under vapor absent and vapor present conditions using different peridotite bulk compositions (Kushiro 1970; Green 1973; Millhollen et al. 1974; Mysen and Boettcher 1975; Mengel and Green 1989; Wallace and Green 1991; Niida and Green 1999). Figure 5 shows the amphibole stability field for 4 peridotite compositions at vapor absent conditions with the stability field for Hawaiian Pyrolite (HPY-0.2%; where 0.2% is the bulk H2O content by weight) also indicated at water saturated

254

Frost 5.2 1090 Temperature (oC)

Figure 4. The melting of pargasite amphibole as a function of H2O content of a coexisting H2O-CO2 fluid. With decreasing H2O content of the fluid the melting temperature rises as the concentration of H2O in the silicate liquid phase decreases. Estimates for the H2O content of the silicate liquid are shown in values of weight percent along the melting curve (Holloway 1973).

Pressure (GPa)

7.8 Pargasite

8.8

1070

9.8

1060 Holloway (1973) 1050 0.2 CO2

NHD

HPY-6%

TQ MPY

7.0

3.8 1080

Di+Fo+Sp ±Ne+Lq

5.8

HPY-0.2%

0.4

0.6 XH2O Fluid

0.8

1.0 H2O

Figure 5. Pargasite amphibole stability curves in peridotite bulk compositions at water saturated and under-saturated conditions. HPY-6% denotes the limit of pargasite stability in a Hawaiian pyrolite composition at H2O saturated conditions with 6 wt% H2O (Green 1973). HPY-0.2% is the same peridotite composition at under-saturated conditions with 0.2 wt% H2O. NHD is North Hessian Depression peridotite at H2O undersaturated conditions with 0.15% H2O (Mengel and Green 1989). TQ is H2O undersaturated Tinaquillo lherzolite with 0.2% H2O (Wallace and Green 1991). MPY is H2O undersaturated MORB pyrolite with 0.6% H2O (Niida and Green 1999). For the MPY (MORB pyrolite) composition the vapor absent solidus is also indicated a few degrees below the amphibole breakdown curve by a dashed curve. The dry solidi for these peridotite compositions occur in the grey region.

o

Temperature ( C)

conditions (HPY-6%). Under both H2O-saturated and -undersaturated conditions the studies are consistent with amphibole disappearing from the assemblage a few degrees above the solidus. The fluid-absent solidus, indicated as the dashed line for MORB pyrolite (MPY), is in fact controlled by the amphibole stability field and backbends sharply as amphibole becomes unstable. It should be noted, however, that Mysen and Boettcher (1975) studied 4 peridotite compositions at water-saturated conditions and found significantly lower solidus temperatures and lower amphibole thermal stability limits than those reported by Green (1973). Kushiro (1970) and Millhollen et al. (1974), however, reported similar phase relations to those of Green (1973). Possible causes for these differences are discussed by Gilbert et al. (1982; page 299). The variable extent of the amphibole stability field under fluid absent conditions (variations by about 0.5 GPa and more than 100 °C) has been attributed to variations in alkali and Ti contents of the bulk compositions (Niida and Green 1999). The bulk composition with the lowest alkali content shows the smallest amphibole stability field (Tinaquillo lherzolite;

Stability of Hydrous Mantle Phases

255

Wallace and Green 1991) whereas the highest alkali content results in the largest stability field (Hawaiian pyrolite; Green 1973). This is somewhat counter intuitive as it means the less fertile, alkali poor samples melt at a lower temperature. Wallace and Green (1991), however, rationalized this in terms of the alkali/H2O ratio. If the alkali/H2O ratio of the bulk composition is low then the proportion of amphibole formed at high temperatures will not be sufficient to account for the entire H2O content. The excess H2O will induce melting, which will strongly partition alkalis to the melt and lead to amphibole breakdown. Figure 6 shows how pargasite alkali contents increase in both A and B sites with increasing pressure. The data cover temperatures between 925-1150 °C and there is also an increase in amphibole alkali contents over this temperature range. In addition to pargasite, the Na content of coexisting clinopyroxene also increases with pressure. As there is no other alkali-bearing phase a decrease in the modal abundance of pargasite must also occur with increasing P and T. It is most likely this relationship that ultimately leads to an excess of H2O over alkalis and melting.

Apatite Relatively little experimental work exists on the stability of apatite Ca5(PO4)3(OH,F,Cl) particularly in natural rock systems at high pressures. The melting curves of pure hydroxyl Ca5(PO4)3(OH) and fluor Ca5(PO4)3(F) apatites were approximately determined by Murayama et al. (1986) with the melting temperature of hydroxyl apatite rising from 1613 °C at atmospheric pressure to over 2000 °C at 7.7 GPa. Melting temperatures of fluor apatite are slightly higher. Above 10 GPa both end-members breakdown to gamma-Ca3(PO4)2 plus either CaF or CaO + H2O. Several studies have examined the solubility of apatite in a variety of melt compositions at high pressures. Watson (1980) observed that the high solubility of apatite in basic silicate melt compositions meant that it was unlikely to remain as a residual phase in the mantle during significant partial melting and may not, therefore, influence the trace element 1

Na + K (per formula unit)

0.8

A site

0.6

0.4 TQ-A TQ-B NHD-A NHD-B MPY-A MPY-B HPY-A HPY-B

0.2

B site

0 0

0.5

1

1.5 2 2.5 Pressure (GPa)

3

3.5

Figure 6. Pressure variation of the total alkali content (Na+K) in the A site and Na content in the B site of pargasitic amphibole (general formula A0−1B2C5T8O22(OH)2) from the water-under saturated experimental studies of Green (1973), Mengel and Green (1989), Wallace and Green (1991) and Niida and Green (1999). Peridoitie composition abbreviations are the same as in Figure 5.

256

Frost

composition of melts. Watson (1980) reasoned, however, that it might remain as a residual phase during melting at lower temperatures in the presence of H2O. Baker and Wyllie (1992) measured apatite solubility in equilibrium with low degree H2O bearing carbonatite melts in equilibrium with mantle peridotite and found apatite to melt out close to the solidus. For fluid undersaturated conditions compatible with rocks containing only apatite, the thermal stability of apatite may be quite high, but the onset of melting due to the presence of other hydrous phases such a pargasite may strongly reduce apatite thermal stability.

Phlogopite Studies by Sato et al. (1997) and Trønnes (2002) on the stability field of phlogopite at high pressure report very similar decomposition curves (Fig. 7). Trønnes (2002) found that synthetic end member phlogopite in the KMASH system undergoes dehydration melting to an assemblage of pyrope, forsterite and liquid up to ca. 8 GPa and an assemblage of pyrope, Phase X, forsterite and liquid above 9 GPa. Sato et al. (1997) used a natural phlogopite composition that was Al-rich (ca. 10% eastonite component) and found an incipient breakdown at high pressure to produce phlogopite, closer to the end member composition, and garnet and fluid.

12

Breakdown curves for: A Natural phlogopite C K2Mg6Al2Si6O20(OH)4 phlogopite D K2Mg6Al2Si6O20(OH)4 phlogopite E Natural phlogopite + enstatite

Sato et al. (1997) Trønnes (2002) Yoder & Kushiro (1969) Sato et al. (1997)

10 8

s

A

lidu Ph+en so

Pressure (GPa)

C

6

Gar+Fo +L

Ph+ gar+fl

Gar+Sp +Fo+L

B

4

Ph Fo+L

2

F E

D

0 900 1000 1100 1200 1300 1400 1500 Temperature (oC) Figure 7. The stability of phlogopite and phlogopite plus enstatite from various experimental studies. (A) The stability field of a natural phlogopite K2.1Na0.1Mg5.4Fe0.2Al2.5Si5.7O20(OH,F)4 (Sato et al. 1997) with the breakdown products at various pressures indicated on the diagram. Above the curve (B) excess Al in the natural phlogopite is expelled and produces garnet and fluid. The curves labeled (C) and (D) show the breakdown of synthetic K2Mg6[Si6Al2]O20(OH)4 phlogopite from Trønnes (2002) and Yoder and Kushiro (1969), respectively. The breakdown products are similar to those found by Sato et al, (1997) except that curve (B) was not observed and above 9 GPa the products were pyrope, Phase X, forsterite and liquid. (E) shows the breakdown curve of a natural phlogopite plus enstatite assemblage determined by Sato et al. (1997) which occurs above the solidus indicated by curve (F). Data of Modreski and Boettcher (1972) constrain both (E) and (F) at pressures below 4 GPa.

Stability of Hydrous Mantle Phases

257

The onset of this breakdown is indicated by curve (F) in Figure 7. These results imply that the AlAlMg−1Si−1 eastonite substitution becomes unstable at high pressures. The similar thermal stability range of the ideal phlogopite end member (Yoder and Kushiro 1969; Trønnes 2002) and the natural Al-rich phlogopite (Sato 1997) may be coincidental. The natural phlogopite also contained significant amounts of F and Fe, which could have stabilizing and destabilizing effects, respectively. Sato et al. (1997) also studied the stability of the same natural phlogopite composition coexisting with enstatite. They observed the same shift in phlogopite composition with pressure and a similar stability field for phlogopite above 4 GPa. The solidus, however, is displaced to lower temperatures by approximately 50 with phlogopite coexisting with melt over this temperature interval. Below 4 GPa the stability field of phlogopite and enstatite becomes significantly smaller than that of pure phlogopite as shown by the data of Modreski and Boettcher (1972). Two studies have examined the stability of phlogopite plus diopside assemblages in the KCMASH system (Sudo and Tatsumi 1990; Luth 1997) as shown in Figure 8. In a Ca bearing system K-richterite amphibole is stable and becomes a high-pressure product of phlogopite decomposition. Sudo and Tatsumi (1990) proposed that in idealized terms this reaction is 2K2Mg6Al2Si6O20(OH)4 + 2CaMgSi2O6= phlogopite

diopside

K2CaMg5Si8O22(OH)2 + CaMg5Al4Si6O24+2Mg2SiO4+ [K2O+3H2O] KK-richterite

garnet

forsterite

(5)

fluid

In the Na-free system, K occupies both A and B(M4) sites in amphibole and is referred to a KK-richterite, as opposed to ordinary mantle K-richterites which are closer to the general formula KNaCaMg5Si8O22(OH)2. Sudo and Tatsumi (1990) showed that reaction (5) was divarient and occurred over a pressure interval of approximately 5 GPa. The results of Luth (1997) combined with those of Sudo and Tastumi (1990) indicate that phlogopite starts to

12 B

Pressure (GPa)

10 8

ph+cpx+gar +ol+kr+fl L+cpx +gar+ol

Ph+cpx+gar+ol+fl A

6 4

cpx+gar +ol+kr+fl

Ph+cpx

L+ol+cpx C

2 0 900 1000 1100 1200 1300 1400 1500 1600

Temperature (oC)

Figure 8. The stability of phlogopite coexisting with diopside in the KCMASH system from Luth (1997). All labeled fields and solid thick curves refer to those reported by Luth (1997). Curves (A) and (B) are those reported by Sudo and Tatsumi (1990). Curve (C) shows the stability of pure phlogopite from Trønnes (2002) and Yoder and Kushiro (1969).

258

Frost

breakdown to produce garnet at pressures lower than where K-richterite is observed in the experimental charges. This results in a field of phlogopite+ diopside + garnet, which must also coexist with fluid. Luth (1997) bracketed the high temperature breakdown of phlogopite, by melting of the diopside + phlogopite + garnet assemblage, at between 1400 and 1450 °C at 7.5 GPa. This is one of the highest temperature occurrences reported for phlogopite in ultramafic systems and is difficult to reconcile with the lower thermal stability of pure phlogopite. Konzett and Ulmer (1999) studied the stability of K-bearing phases at high pressure and temperature in an analogue KNCMASH lherzolite-30% olivine composition and in a natural lherzolite composition. As shown in Figure 9 in the KNCMASH system, the maximum pressure and temperature stability of phlogopite is much smaller than that reported by Luth (1997). The lower pressure stability of phlogopite in the bulk composition studied is explained by the presence of enstatite, which instigates phlogopite breakdown via the reaction, 2K2Mg6Al2Si6O20(OH)4 + CaMgSi2O6 + Mg2Si2O6 = phlogopite

diopside

enstatite

K2CaMg5Si8O22(OH)2 + 2Mg3Al2Si3O12 + 2Mg2SiO4 + [K2O + 3H2O] KK-richterite

garnet

forsterite

(6)

fluid

as initially proposed by Sudo and Tatsumi (1990) for a Na-free system. In the bulk composition studied by Konzett and Ulmer (1999) phlogopite persisted to higher pressure as a result of enstatite being exhausted by reaction (6), whereas in natural compositions, where the modal abundance of phlogopite is likely less than enstatite, this would not occur. In the natural lherzolite composition studied by Konzett and Ulmer (1999) reaction (6) is displaced to lower pressure by approximately 1 GPa, most likely as a result of Fe which partitions more favorably into garnet on the right side of Equation (6). At lower pressure the stability of phlogopite in a natural lherzolite composition minus 60% olivine and with 0.4% K2O and H2O was studied by Mengel and Green (1989). Phlogopite was observed to persist above the solidus, which below 2.5 GPa appears to be coincident with the melting out of pargasitic amphibole.

Phlogopite bearing Phlogopite free

12

Pressure (GPa)

10

Ph out 8

A

Kr in

C

D

6 4 2

B

Ph out

0 900 1000 1100 1200 1300 1400 1500 1600

Temperature (oC)

Figure 9. The stability of phlogopite in a synthetic KNCMASH lherzolite30% olivine composition Konzett and Ulmer (1999). Closed and open symbols bracketing curve (A) indicate assemblages where phlogopite was present and absent, respectively. Olivine, garnet and clinopyroxene are present at all conditions. Orthopyroxene is present at pressures below curve (C), which is Equation (7) given in the text, above which K-richterite becomes stable. The temperature stability of phlogopite determined by Mengel and Green (1989) in a North Hessian Depression peridotite composition is shown by curve (B) whereas the phase relations of Luth (1997) for phlogopite plus diopside are shown by the dashed curve labeled (D).

Stability of Hydrous Mantle Phases 3

Al (per formula unit)

As shown in Figure 10 experimental studies show that phlogopite alumina contents in peridotite bulk compositions decrease with pressure (Konzett and Ulmer 1999; Mengel and Green 1989). This is in good agreement with natural phlogopite occurrences from lower pressure spinel and higher pressure garnet peridotites (Fig. 2a). In addition, the data at 6.5 GPa, covering an experimental temperature range from 850 to 1100 °C, show a small increase of Al content with increasing temperature.

2.5

2

1.5 Konzett and Ulmer (1999) Mengel and Green (1989) 1

K-richterite

259

0

2

4

6

8

10

Hübner and Papike (1970) first Pressure (GPa) synthesized K-richterite (KNaCaMg5 Figure 10. Phlogopite Al content in atoms per formula Si8O22(OH)2) and recognized its potential unit versus pressure for experimental samples produced high pressure stability as it has a smaller within peridotite bulk compositions by Konzett and molar volume than the corresponding Ulmer (1999) and Mengel and Green (1989) between 850 and 1195 °C. product assemblage phlogopite and diopside. The high pressure stability field of the pure KNaCaMg5Si8O22(OH)2 phase, as bracketed by Trønnes (2002), extends to pressures slightly greater than 14 GPa (Fig. 11). The maximum thermal stability is approximately 1450 °C and occurs at 10 GPa. The Na-free KK-richterite (K2CaMg5Si8O22(OH)2) has a stability field that extends to higher pressure by approximately 1 GPa, as determined by Inoue et al. (1998). K-richterite could form in the mantle in two quite different bulk compositions. The first, which accounts for virtually all natural xenolith samples, is in garnet-free peralkaline ultrabasic rocks such as the MARID suite and the strongly metasomatized peridotites termed PKP where (Na2O+K2O)/Al2O3 > 1. The second is not naturally observed but would be normal subalkaline lherzolite rocks at pressures above 6-7 GPa, as discussed later. Konzett et al. (1997) and Konzett and Fei (2000) have shown that the K/Na ratio of K-richterite increases with pressure in both peralkaline and subalkaline bulk compositions. The K/Na ratio of experimentally produced K-richterite reflects the K/Na ratio of the bulk composition, which, as Konzett et al. (1997) point out, is quite different to natural MARID rocks. K-richterites in MARIDs have a generally narrow range of K/Na ratio even though the bulk rocks have a far more variable range. As Konzett et al. (1997) reasoned, this means that MARID rocks themselves cannot represent the entire liquid from which the rocks crystallized, i.e., MARID rocks are cumulates. The results of Konzett et al. (1997) and Konzett and Fei (2000) demonstrate that in peralkaline bulk compositions, where diopside is stable but garnet is only present at pressures >8 GPa, K-richterite stability is very close to that of the pure phase. Only the thermal maxima is reduced by approximately 100 °C compared to the pure phase stability determined by Trønnes (2002). The vast majority of mantle peridotite rocks, on the other hand, are subalkaline (i.e., Na2O+K2O)/Al2O3 > 1). K-richterite is only stable in subalkaline lherzolitic bulk compositions above 6-7 GPa as a result of the reaction: 0.5K2Mg6Al2Si6O20(OH)4 + CaMgSi2O6 + NaAlSi2O6 + Mg2Si2O6 = phlogopite

in cpx

opx

KNaCaMg5Si8O22(OH)2 + Mg3Al2Si3O12 K-richterite

garnet

(7)

260

Frost

Pressure (GPa)

16

Di+St +Cen+fl

5

Di+Cen +Wad+X+fl

12

8

1

K-richterite

Di+Cen +Wad+fl 4

2

4

Di+En+fl

K-richterite stable or 3

unstable in peridotite

0 900 1000 1100 1200 1300 1400 1500 1600 o

Temperature ( C) Figure 11. The stability field of pure KNaCaMg5Si8O22(OH)2 K-richterite as bracketed by Trønnes (2002) is shown by curve (A). All of the indicated named products are with respect to curve (A) with Wad being wadite-structured K2Si4O9, X is Phase X, en and Cen are enstatite and clinoenstatite, Di is diopside and St is stishovite. Curves (B) and (C) are stability fields of K-richterite determined by Foley (1991) and Gilbert and Briggs (1974), respectively. The closed and open symbols indicate the presence and absence of K-richterite in a synthetic KNCMASH subalkaline peridotite assemblage, as determined by Konzett and Ulmer (1999) and Konzett and Fei (2000). In these experiments the coexisting assemblage always contained olivine, garnet, clinopyroxene and enstatite or clinoenstatite. The breakdown products at high pressure also contain Phase X and below curve (D) K-richterite breaks down to a phlogopite-bearing assemblage. Curve (E) shows the high pressure stability of KK-richterite K2CaMg5Si8O22(OH)2 as determined by Inoue et al. (1998).

Equation (7), the Na present equivalent of Equation (6) proposed by Sudo and Tatsumi (1990), demonstrates that the breakdown of minor amounts of phlogopite in the presence of pyroxenes containing some jadeite component can produce the observed natural mantle Krichterite without any fluid release. The identical K/OH-ratio of phlogopite and K-richterite (but not KK-richterite) is the fundamental requirement for such a fluid-free reaction (Konzett and Ulmer 1999). Equation (7) and the general lack of K-richterite in garnet-bearing mantle xenoliths (e.g., Erlank et al. 1987), indicate that the majority of such xenoliths equilibrated at depths shallower than 200 km.

EXPERIMENTAL STUDIES ON THE STABILITY OF POTENTIAL HIGH PRESSURE HYDROUS MANTLE MINERALS Whereas most mantle xenoliths originate from the upper 200 km of the mantle or up to approximately 7 GPa, experimental studies at higher pressures have identified a number of other hydrous minerals that are potentially stable in the deeper parts of the mantle, although mainly in subduction zones. These phases generally don’t have mineral names and are simply referred to by letters e.g., A, B, superhydrous B, D, E and X. Their stability in subduction zones is covered in this volume by Kawamoto (2006). The criterion for evaluating their presence in the ambient mantle is their compatibility with typical mantle minerals and their high temperature stability,

Stability of Hydrous Mantle Phases

261

which can only be assessed through experimental studies. Here I only consider phases with upper thermal stability limits that are close to an average mantle adiabat.

Phase X The enigmatically named Phase X has been observed in several studies as a high-pressure product of the decomposition of K-richterite. Phase X has a variable composition with reported K2O contents of between 10 and 19 wt%. In the KCMSH system Inoue et al. (1998) reported Phase X with the approximate composition K4Mg8Si8O25(OH)2 whereas Trønnes (2002) reported the composition K3.7Mg7.4Al0.6Si8O25(OH)2 in the KMASH system. In addition to Phase X with the formula K1.54Mg1.93Si1.89O7H1.04, Yang et al. (2001) synthesized and solved the structures of sodic Phase X, Na1.16K0.01Mg1.93Al0.14Si1.89O7H1.04, and the anhydrous end members K1.85Mg2.06Si2.01O7 and Na1.78Mg1.93Al0.13Si2.02O7. Phase X is composed of layers of brucite-like MgO6 octahedra linked by Si2O7 tetrahedral dimers and K cations (Yang et al. 2001; Mancini et al. 2002). Yang et al. (2001) proposed the general formula A2−xM2Si2O7Hx where A can be K and/or Na, M can be Mg or Al and x = 0-1. An increase in the K content of Phase X is therefore coupled to a decrease in the H content. The only measurement of the H2O content of Phase X, performed using SIMS, yielded a value of 1.7±0.1 wt% H2O (Inoue et al. 1998), which is significantly below the theoretical maximum of 3.51 wt%. No studies have been performed on the stability of any pure Phase X composition; however, Konzett and Fei (2000) have examined the stability of Phase X in a subalkaline KNCMASH analogue peridotite composition. Phase X coexists with a typical mantle assemblage of olivine/wadsleyite, clinopyroxene and garnet between 14 and 20 GPa and at temperatures up to 1600 °C. Phase X, therefore, has the highest thermal stability of any yet investigated nominally hydrous silicate. As shown in Figure 12 the high temperature stability of Phase X is for the main part undetermined. Konzett and Fei (2000) showed that the reactions that produce Phase X from K-richterite in a mantle peridotite composition release fluid because the K/H ratio of Phase X is higher than that of K-richterite. These results also show that the K content and K/Na ratio of Phase X both increase with pressure, which implies a decrease in the H2O content of Phase X with pressure. The change in K/Na ratio occurs as Na is partitioned into coexisting garnet with increasing pressure. Between 20 and 22 GPa Phase X breaks down to an assemblage containing K-hollandite (KAlSi3O8).

K-Hollandite bearing

Phase X out

Pressure (GPa)

K

nd lla -ho

Phase X bearing

K-richterite bearing

Temperature (oC)

in ite

Figure 12. The closed and open symbols indicate the presence and absence of phase-X in a synthetic KNCMASH subalkaline peridotite assemblage, as determined Konzett and Fei (2000). In these experiments the coexisting assemblage was that expected for a peridotite composition at the indicated conditions, i.e., olivine or high-pressure polymorphs, garnet and Ca-perovskite at and above 20 GPa. Filled rectangles show conditions where Luth (1997) observed Phase X in a KCMASH bulk composition. At high pressures Phase X breaksdown to an assemblage containing K-hollandite (KAlSi3O8).

262

Frost

Humite and dense hydrous magnesium silicate phases A number of high pressure experimental studies have shown that the humite minerals chondrodite and clinohumite and the dense hydrous magnesium silicate phases A, superhydrous B, D and E can coexist with ultramafic assemblages at various conditions above 6 GPa and below 1200 °C (Kanzaki 1991; Kawamoto et al. 1995; Ohtani et al. 1995; Frost and Fei 1998; Irifune et al. 1998). The stability fields of these phases are significantly below reasonable average mantle adiabats and they are therefore only expected to be stable in the cooler regions of subduction zones, provided that significant H2O is in fact present within such regions at pressures above 6 GPa. Although the stability fields have been examined in natural systems (Luth 1995; Kawamoto et al. 1995; Frost 1999; Kawamoto 2004) there remains some question as to whether the strong partitioning of some element by a particular hydrous phase may cause some increase in thermal stability. In addition the large amounts of H2O added in some bulk compositions may result in the breakdown of hydrous phases at a lower temperature than we might expect in the mantle as a result of excessive melting. Experiments in relatively low-H2O bulk compositions show, however, that the presence of Al and Fe in phases A and E, superhydrous phases B and D has a limited effect on stability relations in comparison to the MSH system (Luth 1995; Frost 1999). Humite minerals have a preference for Ti and F. Titanian clinohumite is a common accessory mineral in metamorphosed ultrabasic rocks and occurs in serpentinites and kimberlites (López Sánchez-Vizcaíno et al. 2005). The stability of titanian clinohumite is below 1000 °C at 8 GPa although the pure fluorine clinohumite end-member is stable to over 1400 °C at 3 GPa (Weiss 1997; Ulmer and Trommsdorf 1999). The experiments of Kawamoto (2004) contained Ti and showed clinohumite and chondrodite stability to be limited to below 1100 °C at 11 GPa. Phase D is the highest-pressure dense hydrous magnesium silicate and its stability in the lower mantle is ultimately controlled by the reaction, MgSi2O4(OH)2 Phase D

+

MgO periclase

=

2MgSiO3 MgSi-perovskite

+

H2O

(8)

Liquid

The slope of this reaction is not clear however. From a Schreinemakers analysis of the existing experimental data Komabayashi et al. (2004) reported a negative Clapyron slope at approximately 25 GPa with a maximum thermal stability for phase D of 1100°C. Laser heated diamond cell experiments of Shieh et al. (1998) indicate that this reaction leads to the breakdown of phase D at 44 GPa at temperatures between 1000 and 1400 °C, which would be consistent with a convex shape of the reaction boundary of Equation (8), like many other dehydration reactions. Phase D may therefore be stable at temperatures higher than 1100 C at pressures between 25 and 44 GPa but it is probably unlikely that these temperatures approach that of the mantle adiabat. In natural systems phase D contains significant amounts of Al and ferric and ferrous Fe but not in quantities higher than coexisting silicate perovskite, so they have little effect on the thermal stability of phase D (Frost 1999; Frost unpublished data).

THE STABILITY OF HYDROUS PHASES IN ULTRAMAFIC LITHOSPHERE AND THE CONVECTING MANTLE In considering the significance of hydrous minerals in the mantle it is not only of interest to define stability fields, but it is also important to assess the proportion of hydrous minerals that may exist at particular conditions, identify how much of the mantle’s water budget they may account for and examine further factors, such as H2O activity, that may affect their stability. Changing redox conditions as a function of depth in the upper mantle and transition zone may also control fluid speciation and H2O activity, which, in turn, may affect the stability of the hydrous phases.

Stability of Hydrous Mantle Phases

263

Figure 13 shows the stability fields of the major mantle hydrous phases derived from the previously described experimental studies. The experimental data employed are from studies where hydrous phases formed in equilibrium with typical ultramafic mantle assemblages at H2O undersaturated conditions. A mantle adiabat with a potential temperature of 1600 K (i.e., the temperature at the surface when extrapolated through the melting region) is shown with branching geotherms for Achean cratonic and oceanic lithosphere. A water saturated peridotite solidus interpolated from the data of Mysen and Boettcher (1975) to 4 GPa and Kamamoto (2004) >4 GPa is shown. The solidus is not followed into the region of dense hydrous magnesium silicate stability because huge amounts of H2O are required to produce a melt at these conditions and the solvus between fluid and melt may anyway disappear. Figure 13 indicates that the only hydrous mineral to be stable along an average mantle adiabat (AMA) is Phase X, which could be present in the mantle between depths of 400 and 600 km. The Archean lithospheric geotherm (ACL), which branches off the average mantle adiabat at temperatures approaching 1400 °C, misses the stability field of K-richterite but enters the phlogopite stability field at pressures of approximately 6.8 GPa at 1280 °C. In Figure 13 the data on phlogopite and K-richterite are taken from experiments in Fe-free systems (Konzett and Ulmer 1999; Konzett and Fei 2000). Preliminary experiments seem to indicate that Fe destabilizes these hydrous phases further (Konzett and Ulmer 1999) and the extent of the stability fields in Figure 13 may, therefore, be slightly overestimated. It is important to reiterate that in nature K-richterite occurs in mantle xenoliths of peralkline rocks where the K-richterite stability field extends to much lower pressures (Konzett et al. 1997) than in normal subalkaline

700

D

SB

Phase X out

DHMS

600

Phase X

E 400

A Phlo

300

K-richterite AMA

gopit e ou t

K-richterite in

ACL Phlogopite

Depth (km)

Pressure (GPa)

500

200

OL

100 Na-amphibole

Temperature (oC) Figure 13. Stability fields of hydrous minerals in mantle of peridotite composition at H2O-undersaturated conditions. Data are combined from Figures 5,9,11 and 12. The grey shaded region shows where the dense hydrous magnesium silicate phases A, E, super hydrous phase B (sB) and D are stable from Kawamoto (2005). Thick grey curves show the peridotite solidus under H2O saturated and dry conditions. An average mantle adiabat (AMA) and geotherms for Archean cratonic lithosphere (ACL) and 100 million year old oceanic lithosphere (OL) are shown by thin black lines.

264

Frost

peridotite compositions that are depicted in Figure 13. It seems that the only environment where K-richterite could exist in subalkaline mantle rocks is in a subduction zone. The oceanic lithosphere geotherm (OL) passes into the stability field of phlogopite and pargasitic amphibole below 3 GPa. The proportion of hydrous minerals that form along typical geotherms will depend on the Na and K content of the mantle, which in turn will depend on the degree of depletion and metasomatism. Using a primitive mantle composition as a benchmark, it is possible to appreciate the degree of metasomatic enrichment necessary for significant hydrous minerals to form in the lithosphere. If we consider a typical primitive mantle Na2O content of 0.3 wt% then from the experiments of Niida and Green (1999) we can calculate that along an oceanic geotherm at approximately 70 km the lherzolitic assemblage could contain 9 wt% pargasitic amphibole which would accommodate approximately 1500 ppm H2O, assuming stoichiometric amphibole OH contents. At only 50 km this rises to approximately 25 wt% pargasite which would host 4000 ppm H2O in the bulk. Significant amounts of amphibole can, therefore, form in lithospheric mantle with typical Na contents, mostly at the expense of clinopyroxene, by adding relatively small amounts of H2O alone. Primitive mantle K contents, on the other hand, are generally 10 times lower than corresponding Na contents. Therefore, along an Archean craton lithospheric geotherm at approximately 150 km depth 0.03 wt% K2O in the bulk rock will allow a maximum of just 0.2 wt% phlogopite to form, which will host 90 ppm H2O, using data from Konzett and Ulmer (1999). In comparison, metasomatized garnet phlogopite peridotite rocks (GPP) reported by Erlank et al. (1987) have average bulk K2O contents of 0.16% which would result in 1.4 wt% phlogopite forming at 150 km with a bulk H2O content of 600 ppm. Erlank et al. (1987) classified GPP rocks as the least metasomatized, whereas PKP rocks, which are considered to be the most metasomatized, have average K2O contents of approximately 1%. The presence of significant phlogopite in some mantle xenoliths means, therefore, that there are processes that occur in the mantle that strongly concentrate K while having much smaller effects on other major elements and in particular Na. One possibility is that such high K-bearing liquids are produced by the breakdown of the white mica phengite in subducting lithosphere (Schmidt et al. 2004). If this is the only explanation then all K-rich metasomatism of the lithosphere must be related to subduction. Another possibility is that high-pressure metasomatic fluids or melts are K-rich because Na becomes compatible in clinopyroxene during melting at high pressures and low temperatures (Blundy et al. 1995). Clinopyroxene/melt partition coefficients for K, on the other hand, are normally 2 orders of magnitude below those of Na. The compositions of low-fraction hydrous melts or fluids at pressures above 3 GPa are poorly constrained but further study may provide important insights into metasomatic agents in the lithosphere. Phase X is the only hydrous mineral that could be stable along an average mantle adiabat in the convecting mantle, at least given the available experimental data that extend to lower mantle conditions (>660 km). Assuming a primitive mantle bulk composition we can calculate how much Phase X could form at the top of the transition zone (410 km) and how much of the convecting mantle’s water budget it could account for at these depths. Using the data of Konzett and Fei (2000) who used a K2O enriched KLB-1 peridotite composition the K2O content of Phase X expected in the transition zone is 14.5 wt% and the stoichiometric H2O content is approximately 3 wt%. If the bulk rock contains 0.03 wt% K2O then 0.1 wt% Phase X can form with the proportion of H2O hosted by Phase X being just 30 ppm. In addition to the low K content of the primitive mantle, at these conditions high Ca-clinopyroxene also contains as much as 0.1% K2O and given the uncertainty on some of the values it is quite possible that all K2O and H2O may be accommodated by the nominally anhydrous assemblage. It is of course also possible that K is inhomogeneously distributed in the convecting mantle in a similar way to that found in the lithosphere, resulting in regions with higher proportions of Phase X. If the bulk of the transition zone has a primitive mantle composition, however, then the formation of these regions must leave the remaining transition zone depleted in K and the amount of H2O stored

Stability of Hydrous Mantle Phases

265

by Phase X over the bulk of the transition zone cannot be much greater than 30 ppm. It seems clear, therefore, that nominally anhydrous minerals and melts, or possibly fluids at reducing conditions, must host the majority of hydrogen stored in the ambient convecting mantle. As shown in Figure 4 and explained previously for pargasitic amphibole, hydrous phases display the highest thermal stability at fluid-absent conditions. Figure 13 should, therefore, depict the maximum thermal stability within ultramafic bulk compositions with respect to water activity. Lower H2O activities or H2O-saturated conditions should lead to lower hydrous mineral stability fields. One of the problems of relating experimental studies of hydrous mineral stability to natural mantle mineral assemblages is that we generally have only circumstantial evidence for the nature of the metasomatic or igneous melt/fluid phase from which the minerals formed. The activity of H2O at the conditions of formation are, therefore, poorly constrained. For this reason the previously described methodology of Popp et al. (1995) to determine H2O activity through the use of Equation (2) is particularly attractive. Above subduction zones for example where high concentrations of H2O may enter the mantle wedge the maximum thermal stability of hydrous minerals such as pargasite or phlogopite may be closer to that of H2O-saturated conditions, shown for pargasite in Figures 4 and 5, which may be a few hundred degrees below those in Figure 13. Another poorly constrained factor is that the oxygen fugacity of the mantle may decrease with depth causing C-O-H fluids to become richer in CH4 and lowering the activity of H2O (Woermann and Rosenhauer 1985; Wood et al. 1990). Several studies have argued for a lowering of mantle ƒo2 with depth as a result of the pressure effect on the ferric-ferrous equilibria that likely define mantle ƒo2 and due to changes in the solubility of ferric iron in major mantle minerals (Wood et al. 1990; Gudmundsson and Wood 1995; O’Neill et al. 1993; Ballhaus and Frost 1994; Frost et al. 2004). Several oxygen thermobarometry studies on garnet peridotite xenoliths have observed a decrease in ƒo2 with depth from values of around FMQ-1 (one log unit below the fayalite-magnetite-quartz oxygen buffer) close to the spinel peridotite field at 80 km depth down to FMQ-4 at approximately 200 km (McCammon et al. 2001; Woodland and Koch 2003; McCammon and Kopylova 2004). O’Neill et al. (1993) argued that oxygen fugacities in the transition zone may be close to the iron-wüstite buffer (IW i.e., ~FMQ-5). At these conditions a C-O-H fluid may contain over 50% CH4 and up to 5% H2 although values are uncertain as equations of states for reduced gas phases are poorly constrained at these conditions (Holloway 1987; Belonoshko and Saxena 1992). H2 contents may also increase depending on the activity of carbon and components such as H2S could also be relevant. Although there is very little experimental data on their behavior, reduced fluid phases may be more mobile in the mantle as their components likely have lower solubilities in minerals and melts and the solubilities of silicate components in these fluids may be low. As previously discussed, Taylor and Green (1988) observed an increase in the fluid-saturated peridotite solidus between 1.0 and 3.5 GPa at low ƒo2 (~FMQ-4) where CH4 became a major fluid component. This occurred because CH4 lowered the H2O activity in the fluid, which lowered the H2O solubility in the coexisting silicate melt. The presence of a reduced fluid phase with a low H2O activity in the mantle may affect hydrous phase stability and may also lower the solubility of hydroxyl in nominally anhydrous phases. The mobility and low density of a reduced fluid phase in the deeper convecting mantle may help to redistribute hydrogen and might even tend to focus H2O in the upper more oxidized regions of the upper mantle.

ACKNOWLEDGMENTS I am tremendously grateful to Jurgen Konzett, Reidar Trønnes and Alan Woodland for lengthy discussions and for making numerous comments on an earlier version of the manuscript. I also appreciate the comments and corrections of Hans Keppler and John Winter.

266

Frost REFERENCES

Adam J, Green TH, Sie SH (1993) Proton microprobe determined partitioning of Rb, Sr, Ba, Y, Nb and Ta between experimentally produced amphiboles and silicate melts with variable F content. Chem Geol 109: 29-49 Adam J, Green TH, Sie SH, Ryan CG (1997) Trace element partitioning between aqueous fluids, silicate melts and minerals. Eur J Mineral 9:569-584 Arai S, Matsukage K, Isobe E, Vysotskiy S (1997) Concentration of incompatible elements in oceanic mantle: effect of melt/wall interaction in stagnant or failed melt conduits within peridotite. Geochim Cosmochim Acta 61:671-675 Agrinier P, Mevel C, Bosch D, Javoy M (1993) Metasomatic hydrous fluids in amphibole peridotites from Zabargad island (Red Sea). Earth Planet Sci Lett 120:187-205 Baker MB, Wyllie PJ (1992) High-pressure apatite solubility in carbonate-rich liquids: implications for mantle metasomatism. Geochim Cosmochim Acta 56:3409-3422 Bailey DK (1982) Mantle metasomatism-continuing chemical change within the Earth. Nature 296:525-530 Ballhaus C, Frost BR (1994) The generation of oxidized CO2-bearing basaltic melts from reduced CH4-bearing upper mantle sources. Geochim Cosmochim Acta 58:4931-4940 Belonoshko AB, Saxena SK (1992) A unified equation of state for fluids of C-H-O-N-S-Ar composition and their mixtures up to very high-temperatures and pressures. Geochim Cosmochim Acta 56:3611-3626 Bureau H, Keppler H (1999) Complete miscibility between silicate melts and hydrous fluids in the upper mantle: experimental evidence and geochemical implications. Earth Planet Sci Lett 165:187-196 Blundy JD, Falloon TJ, Wood BJ, Dalton JA (1995) Sodium partitioning between clinopyroxene and silicate melts. J Geophys Res 100:15501-15515 Bodinier J-L, Dupuy C, Dostal J (1988) Geochemistry and petrogenesis of Eastern Pyrenean peridotites. Geochim Cosmochim Acta 52:2893-2907 Bose K, Ganguly J (1995) Experimental and theoretical studies of the stabilities of talc, antigorite and phase A at high pressures with applications to subduction processes. Earth Planet Sci Lett 136:109-121 Boyd FR (1959) Hydrothermal investigations of amphiboles. In: Researches in Geochemistry Vol 1. Abelson PH (ed) Wiley, p 377-396 Boyd FR, Mertzman SA (1987) Composition and structure of the Kaapvaal lithosphere, South Africa. In: Magmatic Processes: Physiochemical Principles. Mysen B (ed) Geochemical Society Special Publication 1, p 13-24 Boyd FR (1990) Mantle metasomatism: evidence from a MARID-harzburgite compound xenolith. Carnegie Institution Washington Yearbook 90:18-23 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 Bryndzia LT, Wood BJ (1990) Oxygen thermobarometry of abyssal spinel peridotites: the redox state and the C-O-H volatile composition of the Earth’s sub-oceanic upper mantle. Am J Sci 290:1093-1116 Bureau H, Keppler H (1999) Complete miscibility between silicate melts and hydrous fluids in the upper mantle: experimental evidence and geochemical implications. Earth Planet Sci Lett 165:187-196 Chazot G, Menzies MA, Harte B (1996) Determination of partition coefficients between apatite, clinopyroxene, amphibole, and melt in natural spinel lherzolites from Yemen: implications for wet melting of the lithospheric mantle. Geochim Cosmochim Acta 60:423-437 Cohen RE, O’Nions RK, Dawson JB (1984) Isotope geochemistry of xenoliths from East Africa: implications for development of mantle reservoirs and their interaction. Earth Planet Sci Lett 68:209-220 Conceicao RV, Green DH (2004) Derivation of potassic (shoshonitic) magmas by decompression melting of phlogopite+pargasite lherzolite. Lithos 72:209-229 Dawson JB, Smith JV (1977) The MARID (mica-amphibole-rutile-ilmenite-diopside) suite of xenoliths in kimberlite. Geochim Cosmochim Acta 41:309-323 Dawson JB, Smith JV (1982) Upper-mantle amphiboles: a review. Mineral Mag 45:35-46 Dawson JB (1984) Contrasting types of upper-mantle metasomatism? In: Kimberlites II: The mantle and CrustMantle Relationships. Kornprobst J (ed) Elsevier, pp 289-294 Dyar MD, Mackwell SJ, McGuire AV, Cross, LR, Robertson JD (1993) Crystal chemistry of Fe3+ and H+ in mantle kaersutite: implications for mantle metasomatism. Am Mineral 78:968-979 Edgar AD, Vukadinovic D (1992) Implications of experimental petrology to the evolution of ultrapotassic rocks. Lithos 28:205-220 Eggler DH (1978) The effect of CO2 upon partial melting of peridotite in the system Na2O-CaO-Al2O3-MgOSiO2-CO2 to 35 kb, with an analysis of melting in a peridotite-H2O-CO2 system. Am J Sci 278:305-343 Elkins-Tanton LT, Grove TL (2003) Evidence for deep melting of hydrous metasomatized mantle: Pliocene high-potassium magmas from the Sierra Nevadas. J Geophys Res 108:2350, doi:10.1029/2002JB002168

Stability of Hydrous Mantle Phases

267

Erlank AJ, Waters FG, Hawkesworth CJ, Haggerty SE, Allsopp HL, Rickard RS, Menzies M (1987) Evidence for mantle metasomatism in peridotite nodules from the Kimberley pipes, South Africa. In: Mantle Metasomatism. Menzies MA, Hawkesworth CJ (eds). Academic Press, p 313-361 Fabries J, Lorand J-P, Guiraud M (2001) Petrogenesis of the amphibole-rich veins from the Lherz orogenic lherzolite massif (Eastern Pyrenees, France): a case study for the origin of the orthopyroxene-bearing amphibole pyroxenites in the lithospheric mantle. Contrib Mineral Petrol 140:383-403 Field SW, Haggerty, SE, Erlank AJ (1989) Subcontinental metasomatism in the region of Jagersfontein, South Africa. In: Kimberlites and Related Rocks. Geol Soc Aust Spec Publ 14. Ross J (ed). Geological Society of Australia, p 771-784 Fleet ME (2003) Micas. The Geological Society Foley S (1991) High-pressure stability of the fluor- and hydroxyl-endmembers of pargasite and K-richterite. Geochim Cosmochim Acta 55:2689-2694 Foley S (1992) Vein-plus-wall-rock melting mechanisms in the lithosphere and the origin of potassic alkaline magmas. Lithos 28:435-453 Frey FA, Prinz M (1978) Ultramafic inclusions from San Carlos, Arizona: petrological and geochemical data bearing on their petrogenesis. Earth Planet Sci Lett 38:129-176 Frost DJ, Fei Y (1998). Stability of phase D at high pressure and high temperature. J Geophys Res 103B:74637474 Frost DJ (1999) The stability of dense hydrous magnesium silicates in Earth’s transition zone and lower mantle. In Mantle Petrology: Field Observations and High-Pressure Experimentation. Fei Y, Bertka CM, Mysen BO(ed) Geochemical Society, Special Publication No. 6, p 283-295 Frost DJ, Liebske C, Langenhorst F, McCammon CA, Tronnes RG, Rubie DC (2004) Experimental evidence for the existence of iron-rich metal in the Earth’s lower mantle. Nature 428:409-412 Gilbert MC, Briggs DF (1974) Comparison of the stabilities of OH- and F-potassic richterites- a preliminary report. Trans Am Geophys Union 55:480-481 Gilbert MC, Helz RT, Popp RK, Spear FS (1982) Experimental studies of amphibole stability. Rev Mineral 9B: 229-353 Green DH (1973) Experimental melting studies on a model upper mantle composition at high pressure under water-saturated and water-undersaturated conditions. Earth Planet Sci Lett 19:37-53 Gregoire M, Lorand JP, O’Reilly SY, Cottin JY (2000) Armalcolite-bearing, Ti-rich metasomatic assemblages in harzburgitic xenoliths from the Kerguelen Islands: Implications for the oceanic mantle budget of high-field strength elements. Geochim Cosmochim Acta 64:673-694 Gregoire M, Bell DR, Le Roex AP (2003) Garnet lherzolites from the Kaapvaal craton (South Africa): trace element evidence for a metasomatic history. J Petrol 44:629-657 Gudmundsson G, Wood BJ (1995) Experimental tests of garnet peridotite oxygen barometry. Contrib Mineral Petrol 119:56-67 Halliday AN, Lee D-C, Tommasini S, Davies GR, Paslick CR, Fitton JG, James DE (1995) Incompatible trace elements in OIB and MORB and source enrichment in the sub-oceanic mantle. Earth Planet Sci Lett 133: 379-395 Harte B (1993) Mantle peridotites and processes- the kimberlite sample. In: Continental Basalts and Mantle Xenoliths. Hawkesworth CJ, Norry MJ (eds) Shiva, p 46-91 Hauri EH, Shimizu N, Dieu JJ, Hart SR (1993) Evidence for hotspot-related carbonatite metasomatism in the oceanic upper-mantle. Nature 365:221-227 Hawkesworth CJ, Erlank AJ, Marsh JS, Menzies MA, Van Calsteren P (1983) Evolution of the continental lithosphere: evidence from volcanics and xenoliths in southern Africa. In: Continental Basalts and Mantle Xenoliths. Hawkesworth MJ, Norry MJ (eds). Shiva, p 111-138 Hoal KEO, Hoal BG, Erlank AJ, Shimizu N (1994) Metasomatism of the mantle lithosphere recorded by rare earth elements in garnets. Earth Planet Sci Lett 126:303-313 Holloway JR (1973) The system pargasite H2O-CO2:a model for melting of a hydrous mineral with a mixedvolatile fluid. I. Experimental results to 8 kbar. Geochim Cosmochim Acta 37:651-666 Holloway JR (1987) Igneous fluids. Rev Mineral 17:211-233 Hübner SJ, Papike JJ (1970) Synthesis and crystal chemistry of sodium-potassium richterite (Na,K)NaCaMg5 Si8O22(OH,F)2): a model for amphiboles. Am Mineral 55:1973-1992 Huckenholz HG, Gilbert MC, Kunzmann T (1992) Stability and phase relations of calcic amphiboles crystallized from magnesio-hastingsite compositions in the 1 to 45 kbar pressure range. N Jahr Miner Abh 164:229-268 Inoue T, Irifune T, Yurimoto H, Miyagi I (1998) Decomposition of K-amphibole at high pressures and implications for subduction zone volcanism. Phys Earth Planet Int 107:221-231 Ionov DA, Hofmann AW (1995) Nb-Ta-rich mantle amphiboles and micas: implications for subduction-related metasomatic trace element fractionations. Earth Planet Sci Lett 131:341-356 Ionov D (1998) Trace element composition of mantle-derived carbonates and coexisting phases in peridotite xenoliths from alkaline basalts. J Petrol 39:1931-1941

268

Frost

Irifune T, Kubo T, Isshiki M, Yamasaki Y (1998) Phase transformations in serpentine and transportation of water into the lower mantle. Geophys Res Lett 22:117-120 Jenkins DM (1983) Stability and composition relations of calcic amphiboles in ultramafic rocks. Contrib Mineral Petrol 83:375-384 Jones AP, Smith JV, Dawson JB (1982) Mantle metasomatism in 14 veined peridotites from Bultfontein Mine, South Africa. J Geol 90:435-453 Jones AP (1989) Upper mantle enrichment by kimberlitic or carbonatitic magmatism. In: Carbonates: Genesis and Evolutions. Bell K (ed) Unwin Hyman, p 448-463 Kanzaki M (1991) Stability of hydrous magnesium silicates in the mantle transition zone. Phys Earth Planet Inter 66:307–312 Kawamoto T, Leinenweber K, Hervig RL, Holloway JR (1995) Stability of hydrous minerals in H2O-saturated KLB-1 peridotite up to 15 GPa. In: Volatiles in the Earth and Solar System. Farley KA (ed) American Institute of Physics, p 229-239 Kawamoto T (2004) Hydrous phase stability and partial melt chemistry in H2O-saturated KLB-1 peridotite up to the uppermost lower mantle. Phys Earth Planet Int 143-144:387-395 Kawamoto T (2006) Hydrous phases and water transport in the subducting slab. Rev Mineral Geochem 62: 273-289 King PL, Hervig RL, Holloway JR, Vennemann TW, Righter K (1999) Oxy-substitution and dehydrogenation in mantle-derived amphibole megacrysts. Geochim Cosmochim Acta 63:3635-3651 Kinny PD, Compston JW, Bristow JW, Williams IS (1989) Archaean mantle xenocrysts in a Permian kimberlite: two generations of kimberlitic zircon in Jwaneng DK2, southern Botswana. In: Kimberlites and related rocks. Geol Soc Aust Spec Publ 14. Ross J (ed). Geological Society of Australia, p 833-842 Kinny PD, Meyer HOA (1994) Zircon from the mantle: a new way to date old diamonds. J Geol 102:475-481 Komabayashi T, Omori S, Maruyama S (2004) Petrogenetic grid in the system MgO-SiO2-H2O up to 30 GPa, 1600 degrees C: Applications to hydrous peridotite subducting into the Earth’s deep interior. J Geophys Res 109:Art. No. B03206 Konzett J, Sweeney RJ, Thompson AB, Ulmer P (1997) Potassium amphibole stability in the upper mantle: an experimental study in a peralkaline KNCMASH system to 8.5 GPa. J Petrol 38:537-568 Konzett J, Ulmer P (1999) The stability of hydrous potassic phases in lherzolitic mantle- an experimental study to 9.5 GPa in simplified and natural bulk compositions. J Petrol 40:629-652 Konzett J, Fei Y (2000) Transport and storage of potassium in the Earth’s upper mantle and transition zone: an experimental study to 23 GPa in simplified and natural bulk compositions. J Petrol 41:583-603 Kushiro I (1970) Stability of amphibole and phlogopite in the upper mantle. Carnegie Inst Wash Year Book 68: 245-247 Kushiro I (1987) A petrological model of the mantle wedge and lower crust in the Japanese island arcs. In: Magmatic Processes: Physiochemical Principles. Geochem Soc Spec Publ 1. Mysen BO (ed) The Geochemical Society, p 165-181 Lambert IB, Wyllie PJ (1968) Stability of hornblende and a model for the low velocity zone. Nature 215:12401241 LaTourrette T, Hervig RL, Holloway JR (1995) Trace element partitioning between amphibole, phlogopite and basanite melt. Earth Planet Sci Lett 135:13-50 Leake BE, Woolley AR, Arps CES, Birch WD, Gilbert MC, Grice JD, Hawthorne FC, Kato A, Kisch HJ, Krivovichev VG, Linthout K, Laird J, Mandarino JA, Maresch WV, Nickel EH, Rock NMS, Schumacher JC, Smith DC, Stephenson NCN, Ungaretti L, Whittaker EJW, Guo YZ (1997) Nomenclature of amphiboles: Report of the subcommittee on amphiboles of the International Mineralogical Association, commission on new minerals and mineral names. Am Mineral 82:1019-1037 López Sánchez-Vizcaíno V, Trommsdorff V, Gómez-Pugnaire MT, Garrido CJ, Müntener O, Connolly JAD (2005) Petrology of titanian clinohumite and olivine at the high-pressure breakdown of antigorite serpentinite to chlorite harzburgite (Almirez Massif, S. Spain). Contrib Mineral Petrol 149:627-646 Luth RW (1995) Is phase A relevant to the Earth’s mantle? Geochim Cosmochim Acta 59:679-682 Luth RW (1997) Experimental study of the system phlogopite-diopside from 3.5 to 17 GPa. Am Mineral 82: 1198-1209 Lykins RW, Jenkins DM (1992) Experimental determination of pargasite stability relations in the presence of orthopyroxene. Contrib Mineral Petrol 112:405-413 McCammon CA, Griffin WL, Shee SR, O’Neill HSC (2001) Oxidation during metasomatism in ultramafic xenoliths from Wesselton kimberlite, South Africa. Implications for the survival of diamond. Contrib Mineral Petrol 141:287-296. McCammon C, Kopylova MG (2004) A redox profile of the Slave mantle and oxygen fugacity control in the cratonic mantle. Contrib Mineral Petrol 148:55-68 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

Stability of Hydrous Mantle Phases

269

McNeil AM, Edgar AD (1987) Sodium-rich metasomatism in the upper mantle: implications of experiments on the pyrolite-Na2O-rich fluid system at 950 °C, 20 kbar. Geochim Cosmochim Acta 51:2285-2294 Mancini F, Harlow GE, Cahill C (2002) The crystal structure and cation ordering of phase-X-(K1−x−n)2(Mg1−n [Al,Cr]n)2Si2O7H2x: A potential K- and H-bearing phase in the mantle. Am Mineral 87:302-306 Matson DW, Muenow DW, Garcia MO (1986) Volatile contents of phlogopite micas from South African kimberlite. Contrib Mineral Petrol 93:399-408 Mengel K, Green DH (1989) Stability of amphibole and phlogopite in metasomatized peridotite under watersaturated and water-undersaturated conditions. In: Kimberlites and related rocks. Geol Soc Aust Spec Publ 14. Ross J (ed) Geological Society of Australia, p 571-581 Menzies MA, Murthy VR (1980) Nd and Sr isotope geochemistry isotope geochemistry of hydrous mantle nodules and their host alkali basalts: implications for local heterogeneities in metasomatically veined mantle. Earth Planet Sci Lett 46:323-334 Menzies MA, Rogers N, Tindle A, Hawkesworth CJ (1987) Metasomatic enrichment processes in lithospheric peridotites, an effect of asthenospheric-lithospheric interaction. In: Mantle Metasomatism. Menzies MA, Hawkesworth CJ (eds) Academic Press, p 313-361 Merrill RB, Wyllie PJ (1975) Kaersutite and kaersutite eclogite from Kakanui, New Zealand- water-excess and water-deficient melting to 30 kilobars. Geol Soc Am Bull 86:555-570 Michael PJ (1988) The concentration, behavior and storage of H2O in the sub-oceanic upper mantle: implications for mantle metasomatism. Geochim Cosmochim Acta 52:555-566 Millhollen GL, Irving AJ, Wyllie PJ (1974) Melting interval of peridotite with 5.7 percent water to 30 kilobars. J Geol 82:575-587 Miyagi I, Matsubaya O, Nakashima S (1998) Change in D/H ratio, water content and color during dehydration of hornblende. Geochem J 32:33-48 Modreski PJ, Boettcher AL (1972) Stability of phlogopite + enstatite at high pressure: A model for micas in the interior of the Earth. Am J Sci 272:852-869 Murayama JK, Nakai S, Kato M, Kumazawa M. (1986) A dense polymorph of Ca3(PO4)2:a high pressure phase of apatite decomposition and its geochemical significance. Phys Earth Planet Int 44:293-303 Mysen BO, Boettcher AL (1975) Melting of a hydrous mantle: I Phase relations of natural peridotite at high pressures and temperatures with controlled activities of water, carbon dioxide, and hydrogen. J Petrol 16: 520-548 Mysen BO, Virgo D, Popp RK, Bertka CM (1998) The role of H2O in Martian magmatic systems. Am Mineral 83:942-946 Niida K, Green DH (1999) Stability and chemical composition of pargasitic amphibole in MORB pyrolite under upper mantle conditions. Contrib Mineral Petrol 135:18-40 Obata M (1980) The Ronda peridotite: garnet-, spinel- and plagioclase-lherzolite facies and the P-T trajectories of a high-temperature mantle intrusion. J Petrol 21:533-572 Obata M, Morten L (1987) Transformation of spinel lherzolite to garnet lherzolite in ultramafic lenses of the Austridic crystalline complex, northern Italy. J Petrol 28:599-623 O’Hara MJ, Mercy ELP (1963) Petrology and petrogenesis of some garnetiferous peridotites. Trans Royal Soc Ed 65:251-314 O’Neill HStC, Rubie DC, Canil, D, Geiger CA, Ross CR, Seifert F, Woodland AB (1993) Ferric iron in the upper mantle and in transition zone assemblages: Implications for relative oxygen fugacities in the mantle. Geophys Monograph 74 IUGG 14:73-89 O’Reilly SY, Griffin WL (2000) Apatite in the mantle: implications for metasomatic processes and high heat production in Phanerozoic mantle. Lithos 53:217-232 O’Reilly SY, Griffin WL (1988) Mantle metasomatism beneath western Victoria, Australia: I. Metasomatic processes in Cr-diopside lherzolites. Geochim Cosmochim Acta 52:433-447 Ohtani, E, Shibata T, Kubo T, Kato T (1995) Stability of hydrous phases in the transition zone and the upper most part of the lower mantle. Geophys Res Lett 22:2553-2556 Popp RK, Virgo D, Yoder HS, Hoering TC, Phillips MW (1995) An experimental study of phase equilibria and Fe oxy-component in kaersutitic amphibole: implications for the ƒH2 and aH2O in the upper mantle. Am Mineral 80:534-548 Righter K, Dyar MD, Delaney JS, Vennemann TW, Hervig RL, King PL (2002) Correlations of octahedral cations with OH−, O2−, Cl− and F− in biotite from volcanic rocks and xenoliths. Am Mineral 87:142-153 Robinson P, Spear FS, Schumacher JC (1982) Phase relations of metamorphic amphiboles: natural occurrence and theory. Rev Mineral 9B:1-227 Roden MF, Murthy R (1985) Mantle metasomatism. Ann Rev Earth Planet Sci 13:269-296 Ryabchikov ID, Boettcher AL (1980) Experimental evidence at high pressure for potassic metasomatism in the mantle of the Earth. Am Mineral 65:915-919 Sato K, Katsura T, Ito E (1997) Phase relations of natural phlogopite with and without enstatite up to 8 GPa: implication for mantle metasomatism. Earth Planet Sci Lett 146:511-526

270

Frost

Scambelluri M, Hermann J, Morten L, Rampone E (2006) Melt- versus fluid-induced metasomatism in spinel to garnet wedge peridotites (Ulten Zone, Eastern Italian Alps): clues from trace element and Li abundances. Contrib Mineral Petrol 151:372-394 Schmidt MW, Poli S (1998) Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation. Earth Planet Sci Lett 163:361-379 Schmidt MW, Vielzeuf D, Auzanneau E (2004) Melting and dissolution of subducting crust at high pressures: the key role of white mica. Earth Planet Sci Lett 228:65-84 Schneider ME, Eggler DH (1986) Fluids in equilibrium with peridotite minerals: implications for mantle metasomatism. Geochim Cosmochim Acta 50:711-724 Schrauder M, Navon O (1994) Hydrous and carbonatitic mantle fluids in fibrous diamonds from Jwaneng Botswana. Geochim Cosmochim Acta 58:761-771 Seyler M, Mattson PH (1989) Petrology and thermal evolution of the Tinaquillo peridotite (Venezuela). Journal Geophys Res 94:7629-7660 Shaw CSJ, Eyzaguirre J (2000) Origin of megacrysts in the mafic alkaline lavas of the West Eifel volcanic field, Germany. Lithos 50:75-95 Shen AH, Keppler H (1997) Direct observation of complete miscibility in the albite-H2O system. Nature 385: 710-712 Shieh SR, Mao HK, Hemley RJ, Ming LC (1998) Decomposition of phase D in the lower mantle and the fate of dense hydrous silicates in subducting slabs. Earth Planet Sci Lett 159:13-23 Shimizu N (1975) Rare earth elements in garnets and clinopyroxenes from garnet lherzolite nodules in kimberlites. Earth Planet Sci Lett 25:26-32 Sweeney RJ, Thompson AB, Ulmer P (1993) Phase relations of a natural MARID composition and implications for MARID genesis, lithospheric melting and mantle metasomatism. Contrib Mineral Petrol 115:225-241 Sudo A, Tatsumi Y (1990) Phlogopite and K-amphibole in the upper mantle: implication for magma genesis in subduction zones. Geophys Res Lett 17:29-32 Tatsumi Y, Hamilton DL, Nesbitt RW (1986) Chemical characteristics of fluid phase released from a subducted lithosphere and origin of arc magmas: evidence from high pressure experiments and natural rocks. J Volcanol Geotherm Res 29:293-310 Taylor WR, Green DH (1988) Measurement of reduced peridotite-C-O-H solidus and implications for redox melting of the mantle. Nature 332:349-352 Thibault Y, Edgar AD, Lloyd FE (1992) Experimental investigation of melts from a carbonated phlogopite lherzolite: implications for metasomatism in the continental lithospheric mantle. Am Mineral 77:784-794 Tiepolo M, Vannucci R, Oberti R, Foley S, Bottazzi P, Zanetti A (2000) Nb and Ta incorportation and fractionation in titanian pargasite and kaersutite: crystal-chemical constraints and implications for natural systems. Earth Planet Sci Lett 176:185-201 Thompson AB (1992) Water in the Earth’s upper mantle. Nature 358:295-302 Thompson JB, Laird J, Thompson AB (1981) Reactions in amphibolite, greenschist and Blueschist. J Petrol 23: 1-27 Trønnes RG, Edgar AD, Arima M (1985) A high pressure-high temperature study of TiO2 solubility in Mg-rich phlogopite: implications to phlogopite chemistry. Geochim Cosmochim Acta 49:2323-2329 Trønnes RG (2002) Stability range and decomposition of potassic richterite and phlogopite end members at 5-15 GPa. Mineral Petrol 74:129-148 Ulmer P, Trommsdorff V (1999) Phase relations of hydrous mantle subducted to 300 km. In Mantle Petrology: Field Observations and High-pressure Experimentation. Fei Y, Bertka CM, Mysen BO (ed) Geochemical Society, Special Publication No. 6, p 259-281 Ulmer P (2001) Partial melting in the mantle wedge - the role of H2O in the genesis of mantle-derived ‘arcrelated’ magmas. Phys Earth Planet Int 127:215-232 van Achterbergh E, Griffin WL, Stiefenhofer J (2001) Metasomatism in mantle xenoliths from the Letlhakane kimberlites: estimation of element fluxes. Contrib Mineral Petrol 141:397-414 Wagner C, Deloule E, Mokhtari A (1996) Richterite-bearing peridotites and MARID-type inclusions in lavas from North Eastern Morocco: mineralogy and D/H isotopic studies. Contrib Mineral Petrol 124:406-421 Wallace ME, Green DH (1988) An experimental determination of primary carbonatite magma composition. Nature 335:343-346 Wallace ME, Green DH (1991) The effect of bulk rock composition on the stability of amphibole in the upper mantle: implications for solidus positions and mantle metasomatism. Mineral Petrol 44:1-19 Wallace PJ (1998) Water and partial melting in mantle plumes: Inferences from the dissolved H2O concentrations of Hawaiian basaltic magmas. Geophys Res Lett 25:3639-3642 Waters FG (1987) A suggested origin of MARID xenoliths in kimberlites by high pressure crystallization of an ultrapotassic rock such as lamproite. Contrib Mineral Petrol 95:523-533 Watson BE (1980) Apatite and phosphorus in mantle source regions: an experimental study of apatite/melt equilibria at pressures to 25 kbar. Earth Planet Sci Lett 51:322-335

Stability of Hydrous Mantle Phases

271

Weiss M (1997) Clinohumites: a field and experimental study. ETH Dissertation No. 12202, 168 pp. Wilkinson JFG, Le Maitre RW (1987) Upper mantle amphiboles and micas and TiO2, K2O and P2O5 abundances and 100Mg/(Mg+Fe2+) ratios of common basalts and andesites: implications for modal mantle metasomatism and undepleted mantle compositions. J Petrol 28:37-73 Wilshire HG, Shervais JW (1975) Al-augite and Cr-diopside ultramafic xenoliths in basaltic rocks from Western United States. Phys Chem Earth 9:257-276 Witt-Eickschen G, Seck HA, Reys CH (1993) Multiple enrichment processes and their relationships in the subcrustal lithosphere beneath the Eifel (Germany). J Petrol 34:1-22 Witt-Eickschen G, Kramm U (1998) Evidence for the multiple stage evolution of the subcontinental lithospheric mantle beneath the Eifel (Germany) from pyroxenite and composite pyroxenite/peridotite xenoliths. Contrib Mineral Petrol 131:258-272 Woermann E, Rosenhauer M (1985) Fluid phases and the redox state of the earths mantle. Fortschr Mineral 63: 263-349 Wood BJ, Bryndzia LT, Johnson KE (1990) Mantle oxidation state and its relationship to tectonic environment and fluid speciation. Science 248:337-345 Woodland AB, Kornprobst J, McPherson E, Bodinier J-L, Menzies MA (1996) Metasomatic interactions in the lithospheric mantle: petrologic evidence from the Lherz massif, French Pyrenees. Chem Geol 134:83-112 Woodland AB, Koch M (2003) Variation in oxygen fugacity with depth in the upper mantle beneath the Kaapvaal craton, South Africa. Earth Planet Sci Lett 214:295-310 Wyllie PJ (1978) Mantle fluid compositions buffered in peridotite-CO2-H2O by carbonates, amphibole, and phlogopite. J Geol 86:687-713 Yang HX, Konzett J, Prewitt CT (2001) Crystal structure of Phase X, a high pressure alkali-rich hydrous silicate and its anhydrous equivalent. Am Mineral 86:1483-1488 Yang H-J, Frey FA, Clague DA (2003) Constraints on the source components of lavas forming the Hawaiian North Arch and the Honolulu volcanics. J Petrol 44:603-627 Yaxley GM, Green DH, Kamenetsky V (1998) Carbonatite metasomatism in the southeastern Australian lithosphere. J Petrol 39:1917-1930 Yoder HS, Kushiro I (1969) Melting of a hydrous phase: phlogopite. Am J Sci 267A: 558-582 Young ED, Virgo D, Popp RK (1997) Eliminating closure in mineral formulae with specific application to amphiboles. Am Mineral 82:790-806 Zanetti A, Mazzucchelli M, Rivalenti G, Vannucci R (1999) The Finero phlogopite-peridotite massif: an example of subduction-related metasomatism. Contrib Mineral Petrol 134:107-122

12

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 273-289, 2006 Copyright © Mineralogical Society of America

Hydrous Phases and Water Transport in the Subducting Slab Tatsuhiko Kawamoto Institute for Geothermal Sciences Graduate School of Science Kyoto University Beppu 874-0903, Japan e-mail: [email protected]

INTRODUCTION Arc volcanoes are typically located 90-180 km above the surface of downgoing slabs, as shown by Wadati-Benioff deep seismic foci (Gill 1981; Tatsumi 1989). The intimate relationship between the dip angles of the subducting slab and the locations of volcanic arcs indicates that subduction zone magmatism is triggered by material input from the subducting slab (Tatsumi and Eggins 1995). The slab-derived components are thought to be aqueous fluids or H2O-rich partial melts of subducted oceanic crust. Therefore, knowledge of the stability of hydrous phases and the chemical and physical properties of aqueous fluids in downgoing slabs is essential to understand the material transport in subduction zones. In this section, I will review the stability of hydrous phases in downgoing peridotite, basalt and sediment systems, and the chemical and the wetting properties of aqueous fluids. Recent experimental studies indicate that 3-4 GPa, equivalent to 90-120 km depth, is a key pressure, where (1) the chemical compositions of silicate components dissolved in aqueous fluids equilibrated with mantle minerals approach the composition of mantle peridotite itself (Stalder et al. 2001; Mibe et al. 2002; Kawamoto et al. 2004), (2) the dihedral angle between olivine and aqueous fluids starts becoming smaller than 60° (Watson et al. 1990; Mibe et al. 1998, 1999), and (3) the immisciblity gap between peridotitic melts and aqueous fluids disappears and consequently hydrous minerals liberate supercritical aqueous fluids (Mibe et al. 2004a, 2006). The similarity between these pressures and the depths of downgoing slab underneath volcanic fronts, where the maximum numbers of volcanoes are formed, 124 ± 38 km (Gill 1981) or 112 ± 19 km (Tatsumi 1986), suggests that subduction zone magmatism can be triggered by the input of supercritical fluids from the downgoing peridotite and basalt.

LOW-PRESSURE HYDROUS MINERALS AND HIGH-PRESSURE HYDROUS PHASES Many hydrous crystalline phases are stable in peridotite, basalt and sediment systems over a wide range of pressure. Their chemical formulae and H2O contents are summarized together with those of nominally anhydrous minerals in Table 1. Some hydrous phases have been found only in high-pressure and high-temperature experimental products and have not yet been found in nature: dense hydrous magnesium silicates (DHMS) or alphabet phases (Ringwood and Major 1967), phase Egg (Eggleton et al. 1978), phase Pi (Wunder et al. 1993a), topaz-OH (Wunder et al 1993b), and δ-AlOOH (Suzuki et al. 2000). Although phase D, F, and G were originally suggested as different phases, these phases seem to be identical (Frost 1999; Ohtani et al. 2001). The chemical compositions of DHMS are plotted in Figure 1 with the estimated 1529-6466/06/0062-0012$05.00

DOI: 10.2138/rmg.2006.62.12

274

Kawamoto

H 2O

1 GPa

3 GPa 5 GPa Br

A Nor sB B

MgO

10 GPa

8 GPa 10 Å

E

AhyB Ol

Atg Chn Hywd Chm En

D Talc Ant

SiO2

Figure 1. Compositions of hydrous minerals and dense magnesium hydrous silicates stable in peridotite system plotted with compositions of silicates dissolved into aqueous fluids coexisting with forsterite and enstatite at 1100 °C at 1-10 GPa estimated by Zhang and Frantz (2000) and Mibe et al. (2002) in the MgOSiO2-H2O system. Phase D, E, antigorite, and 10 Å phase are non-stoichiometric phases. Humite is located between chondrodite and clinohumite. Abbreviations are in Table 1.

chemistry of aqueous fluids equilibrated with forsterite + enstatite in the MgO-SiO2-H2O system (Mibe et al. 2002). The hydrous crystalline phases can be divided into three major groups with respect to their stability range (Fig. 2): (1) low-pressure hydrous minerals such as chlorite (clinochlore), talc, and amphibole (the relevant end members are listed in Table 1), which are commonly observed in metamorphic rocks, (2) high-pressure hydrous phases such as DHMS (Fig. 1), K-richterite, topaz-OH, and phase Egg, and (3) middle-pressure hydrous minerals such as phlogopite, antigorite, Mg-sursassite and 10 Å phase in peridotite, lawsonite in basalt, and phengite in sediment. The last group is stable between 5 and 7 GPa, and may be important for delivering H2O from low-pressure hydrous minerals to high-pressure hydrous phases (Fig. 2). Liu (1987) recognized that phase A, a DHMS, can accommodate much more water than amphibole or phlogopite. Therefore he emphasized the important reaction forsterite + H2O = phase A + enstatite, and he described this reaction boundary as a “water-line,” implying that a region deeper than the water-line can be a H2O reservoir in the mantle. In Figure 2, the water-line is shown by the low-pressure stability of DHMS. Kawamoto et al. (1996) identified the presence of a “choke point” in a down going slab. A choke point represents a pressure and temperature condition along a PT path where low-pressure and middle-pressure hydrous minerals get dehydrated at certain pressure conditions and cannot deliver H2O to high-pressure hydrous phases (Fig. 2). The choke point curve, the curve connecting the array of choke points, represents the high-pressure and high-temperature stability limit of the lowpressure and middle-pressure hydrous minerals. In the MgO-SiO2-H2O system, the invariant point composed of antigorite, phase A, enstatite, forsterite and H2O represents the lowest temperature and highest pressure of the choke point. In recent literature, this point is at around 6.2 GPa and 620 °C (Iwamori 2004), and at 5.1 GPa and 550 °C (Komabayashi et al. 2005). In Figure 2, based on the KLB-1 peridotite data, TiO2 stabilizes chondrodite and clinohumite. Therefore, in the peridotite systems, the PT conditions where antigorite meets chondrodite and clinohumite represent the lowest temperature and highest pressure choke point. In the MgO-

Hydrous Phases & Water Transport in the Subducting Slab

Temperature (°C)

moto 400 2

600

HT

2 Talc

800 1000 1200 1400 1600

Chl Par +Opx

LT

10 12 14

ol-fl Present Mantle Adiabat

Phl + Cpx

Phe

100

200

K-ric

gt-fl

E + Chm E + Chm Top +A Eg E

300

Ol Ol + Wd

400

Depth (km)

Pressure (GPa)

8 Chm + Chn

Lws

Wet solidus

+ Ol 4 Atg 6

275

Wd

16 18

Wd + Rg 500

D + sB

20

Rg Hy-wd

22

Hy-rg

24 26

D + sB

Rg Pv + Mw sB

600

700

Figure 2. Pressure and temperature diagram showing stability of hydrous minerals/phases in peridotite (Kawamoto 2004a) with some hydrous phases in basalt/sediment systems. The wet solidus is from Kawamoto and Holloway (1997). Since a second critical endpoint between peridotite melt and aqueous fluids is located at around 3.8 GPa (Mibe et al. 2004a; 2006), the wet solidus is drawn by dashed line at pressures higher than 4 GPa. Stability of lawsonite in basalt is indicated by solid dots and stability boundaries among phengite, topaz-OH, and phase Egg is drawn by open dots, respectively. The stabilities of Par, Chl, Talc, Atg, Phl, K-rich, Lws, Top, Eg are after Schmidt and Poli (1998), Pawley (2003), Ulmer and Trommsdorff (1995), Sudo and Tatsumi (1990), and Ono (1998); phase boundaries among Ol, Ol + Wd, Wd, and Wd + Rg - (Mg0.9Fe0.1)2SiO4 and Rg - (Mg0.9Fe0.1)2SiO4 and Mg-perovskite (Mg-Pv) + magnesium wüstite (Mw) in dry conditions are after Katsura and Ito (1989), and Ito and Takahashi (1989), respectively. The phase boundary of Hy- wd and Hy-rg (dashed line) is at higher pressure than under dry conditions. The 60° isopleths of the dihedral angle in garnet-garnet-fluid (gt-fl) and olivine-olivine-fluid (ol-fl) are also shown (thick gray line). The data of the dihedral angle are compiled in Figure 5. HT and LT represent PT paths of high-temperature and low-temperature subducting slab surface, respectively (Peacock and Wang 1999). Abbreviations are in Table 1.

276

Kawamoto

Table 1. Formula of hydrous minerals/phases and nominally anhydrous minerals in metamorphic basalts, metamorphic sediments, and peridotite (after Wunder and Schreyer 1992, Pawley and Wood 1995, Mysen et al. 1998, Ono 1999, Forneris and Holloway 2003). Name

Symbols

Formula

(Amphibole groups) Tremolite Pargasite Barroisite Glaucophane K-richterite

Trm Par Bar Gln K-ric

Ca2Mg5Si8O22(OH)2 Na2Ca3Mg8FeAl3Si13O44(OH)4 NaCaMg3Al2Si7AlO22(OH)2 Na2Mg3Al2Si8O22(OH)2 K1.9Ca1.1Mg5Si7.9Al0.1O22(OH)2

(Peridotite system) Chlorite Talc Serpentine Antigorite Clinohumite Humite Chondrodite Norbergite Phase A Brucite Phase B Superhydrous B Anhydrous B Phase E Phase D/F/G Anthophyllite Talc 10 Å phase Mg-sursassite Hydrous wadsleyite Hydrous ringwoodite

Chl Tlc Serp Atg Chm Hm Chn Nor A Br B sB AhyB E D Ant Talc 10 Å MgS Hy-wd Hy-rg

(Mg5Al)(AlSi3)O10(OH)8 Mg6Si8O20(OH)4 Mg3Si2O5(OH)4 Mg48Si34O85(OH)62 Mg9Si4O16(OH)2, Ti0.5Mg8.5Si4O17(OH) Mg7Si3O12(OH)2 Mg5Si2O8(OH)2 , Ti0.5Mg4.5Si2O9(OH) Mg3SiO4(OH)2 Mg7Si2O8(OH)6 Mg(OH)2 Mg24Si8O38(OH)4 Mg10Si3O14(OH)4 Mg14Si5O24 Mg2.27Si1.26H2.4O6 MgSi2H2O6 Mg7Si8O22(OH)2 Mg3Si4O10(OH)2 Mg3Si4O10(OH)2 xH2O Mg5Al5Si6O21(OH)7 Mg1.75SiO4(OH)0.5 Mg1.75SiO4(OH)0.5

(Basalt and sediment systems) Zoisite/clinozoisite Staurolite Apatite Sphene Phlogopite Phase Egg Topaz-OH Phase Pi Lawsonite Chloritoid Phengite δ-AlOOH

Zo / Czo Sta Ap Spn Phl Eg Top Pi Lws Cld Phe δ-Al

Ca2Al3Si3O12(OH) (Mg,Fe)2(Al,Fe)9Si4O22(O,OH)2 Ca5(PO4)3(OH,F,Cl) CaTiSiO4(O,OH,F) KMg2Si3AlO10(OH)2 AlSiO3(OH) Al2SiO4(OH)2 Al3Si2O7(OH)3 CaAl2Si2O7(OH)2 H2O (Mg, Fe)2(Al,Fe)4Si2O10(OH)4 K(Al2-xMgx)(Si3+xAl1-x)O10(OH,F)2 AlOOH

(Nominally anhydrous minerals) Olivine/Wadsleyite/Ringwoodite Clinopyroxene Ca-perovskite Orthopyroxene/ Majorite/ Akimotoite/ Perovskite Quartz/ Coesite/ Stishovite Spinel Garnet

Ol / Wd / Rg Cpx Ca-pv Opx/ Mj / Ak / Pv Qz / Coe / St Sp Gt

Mg2SiO4 (Na,Ca)(Mg,Al)Si2O6 CaSiO3 MgSiO3 SiO2 MgAl2O4 (Fe,Mg,Ca)3Al2Si3O12

wt% H2O 2.2 2.2 2.3 2.3 2.1 13 4.8 13 12.3 2.9 - 1.4 3.75 5.3 - 2.6 9.0 11.8 30.9 2.4 1.6 11.4 10.1 2.3 4.75 7.6 - 13 7.2 3.3 3.3 2 2 1.8 1.5 4.8 7.5 10.0 9.0 11.5 8 4.6 15

Hydrous Phases & Water Transport in the Subducting Slab

277

Al2O3-SiO2-H2O system, Mg-sursassite (Gottschalk et al. 2000), which was previously called MgMgAl-pumpellyite (for example, Domanik and Holloway 1996), is stabilized at higher temperature than this invariant point (Fig. 3; Bromiley and Pawley 2003), and its presence therefore increases the temperature of the choke point. The transition zone (410-660 km depth) is also characterized by the high H2O storage capacity of hydrous wadsleyite and hydrous ringwoodite (Fig. 3; Smyth 1987; Inoue et al. 1995; Kawamoto et al. 1996; Kohlstedt et al. 1996; Kudoh et al. 1996; Smyth and Kawamoto 1997; Smyth et al. 1997; Demouchy et al. 2005). Therefore, the transition zone could play a significant role as a large H2O -reservoir formed by crystallization of hydrous wadsleyite and ringwoodite from a hydrous magma ocean. Kawamoto and Holloway (1997) measured the partition coefficient of H2O between hydrous wadsleyite/ringwoodite and hydrous partial melts of peridotite, and suggested the possible existence of a hydrous transition zone in the early history of the Earth. Upwelling from such a hydrous reservoir could generate partial melting at 410 km and produce komatiitic magmas. Through partial melting of a hydrous transition zone, in this hypothesis, the transition zone has been getting drier during the geological time, because the choke point prevents H2O from subducting into the transition zone. Therefore the present transition zone has much less ability to produce komatiite magmas. This hypothesis thus explains why komatiites were produced mainly in the Archean period.

STABILITY OF HYDROUS PHASES IN DOWNGOING PERIDOTITE There are two potentially-hydrated peridotite layers in subduction zones. One is the harzburgite/lherzolite of the subducting lithospheric mantle, which is overlain by oceanic basaltic crust and sediments. The other is downdragged mantle at the base of the mantle wedge. To what extent the peridotite layers are hydrated remains uncertain. Along transform faults, serpentine minerals (antigorite, lizardite, chrysotile) can be formed by seawater alteration. However, the rest of the subducting lithospheric mantle may not be hydrated. The downdragged mantle peridotite at the base of the mantle wedge should be hydrated by aqueous fluids liberated by dehydration reactions of hydrous minerals in downgoing sediment and basalt layers. Nicholls and Ringwood (1973) suggested that subducting basalt will be almost dry beneath the fore-arc region. Sakuyama and Nesbitt (1986), therefore, suggested that downdragged peridotite in the mantle wedge will be hydrated through H2O released by dehydration of the hydrous minerals in the basaltic layer and may carry H2O beneath the volcanic arc. Iwamori (2004) compiled the stability of hydrous phases in the MgO-SiO2-H2O, the MgO-Al2O3-SiO2-H2O, and KLB-1 peridotite systems, and presented the distribution of maximum H2O contents bound in mantle peridotite (Fig. 3). Komabayashi et al. (2004) also presented a similar stability diagram of hydrous phases based on Schreinemakers’ net analysis. They noticed two main differences of hydrous phase stability between the peridotite system and simple systems: (1) the addition of Al2O3 expands the stability field of phase E to the lower pressures and (2) the addition of TiO2 enhances the stability field of clinohumite and chondrodite (Fig. 2). The addition of fluorine is also found to expand the stability of clinohumite into a lower pressure range (Stalder and Ulmer 2001). According to Fumagalli et al. (2001), the10 Å phase (Table 1) is reported to be stable in the peridotite system at 5.2 GPa and 680 °C. Fumagalli and Poli (2005) found that the 10 Å phase has high Al2O3 contents (about 10 wt%) and suggested that this phase is a mixed layer of chlorite and pure 10 Å phase formed in the MgO-SiO2-H2O system. The stability field of this Al-rich 10 Å phase is close to the stability of Mg-sursassite (Bromiley and Pawley 2002). These phases cover some regions of the choke point (Fig. 3), though the H2O content contributed by Mg-sursassite and Al-rich 10 Å phase to peridotite is limited to 0.7 (Iwamori 2004; Fig. 3) and 1 wt% (Fumagalli and Poli 2005), respectively.

Figure 3. Phase diagram showing maximum H2O contents bound in hydrous minerals/phases in the peridotite system. The phase assemblages of fields numbered are shown on the right hand side. Abbreviations are in Table 1. This gray figure was made by the courtesy of Hikaru Iwamori. The original diagram is in full color with better resolution (Iwamori 2004).

278 Kawamoto

Hydrous Phases & Water Transport in the Subducting Slab

279

Amphibole was historically thought to be the most important phase to deliver H2O beneath the volcanic arc (fields 2, 3, 4, 6, and 7 in Fig. 3; Tatsumi 1986; Schmidt and Poli 1998; Niida and Green 1999). According to the compilation by Schmidt and Poli (1998), pargasite can be stable up to between 2.2 GPa and 3.0 GPa depending on the bulk rock chemistry of the system. Although Schmidt and Poli (1998) adopted the lowest pressure (2.2 GPa) for pargasite in harzburgite, it is important to realize that pargasite can be stable up to 3 GPa in more enriched peridotite such as enriched pyrolite (Niida and Green 1999). The 3.5 GPa for the high-pressure stability limit of pargasite adopted by Tatsumi (1986) seems overestimated as Schmidt and Poli (1998) suggested. In Figure 2, 2.8 GPa was adopted as a pressure limit for pargasite according to the recent experimental study by Fumagalli and Poli (2005). The stability of antigorite (line between fields 5 and 8 in Fig. 3) also depends on bulk composition and the effect of Al was evaluated by Bromiley and Pawley (2003). The stability of antigorite in Figure 2 is drawn with the data reported by Ulmer and Trommsdorff (1995).

STABILITY OF HYDROUS PHASES IN DOWNGOING BASALT AND SEDIMENT There are many hydrous minerals observed in metamorphic basalt and sediments. Several experimental studies have explored their high PT stabilities. Concerning the stability of amphibole in the basalt system (Fig. 4), there is a discrepancy between Schmidt and Poli (1998) and Forneris and Holloway (2003). According to Schmidt and Poli (1998), in subducting basalt, amphibole and zoisite dehydrate first, then along a colder path, zoisite and chloritoid dehydrate, and finally lawsonite with or without chloritoid can retain H2O to the deep mantle (Fig. 4B). Along a warmer path, instead of lawsonite, zoisite becomes the only hydrous phase to possess H2O after amphibole dehydration, and then zoisite dehydrates liberating H2O (Fig. 4B). In contrast, according to Forneris and Holloway (2003), amphibole and zoisite at higher temperatures and amphibole with lawsonite at lower temperatures are stable up to 2.5-3.2 GPa (Fig. 4A). Then amphibole and zoisite dehydrate and lawsonite

500 550 2.0

Pressure (GPa)

2.5 3.0

Temperature (ºC)

600

Lws + Amp

Amp + Zo Zo Zo

3.5

4.0

Temperature (ºC)

650 700 750 800 500 550 600 650 700 750 800

Lws

Amp

Cld + Zo

HT

Amp + Zo Zo

HT

Lws + Cld

Cpx + Grt + Fluid

Lws

Cpx + Grt + Fluid

4.5 5.0

A, Forneris & Holloway (2003)

B,

Schmidt & Poli (1998)

Figure 4. Phase diagrams showing stability of hydrous minerals in the MORB system. (A) Forneris and Holloway (2003), Amp represents barroisite at high temperatures and glaucophane at low temperatures, (B) Schmidt and Poli (1998). This figure is after Forneris and Holloway (2003). According to Forneris and Holloway (2003), the chloritoid in B is likely to be formed by metastable crystallization, see text. HT represent a PT path of high-temperature subducting slab surface, and a PT path of low-temperature one (LT in Figure 2) is outside of this PT diagram (Peacock and Wang 1999).

280

Kawamoto

becomes the only hydrous phase (Fig. 4A). Forneris and Holloway (2003) suggested that a possible explanation for the discrepancy was the crystallization of metastable chloritoid (Fig. 4B) during short experimental run durations in the former experiments, perhaps due to the chemical difference between their starting materials: bulk compositions studied by Forneris and Holloway (2003) contained more MgO and Al2O3 than the starting materials of Schmidt and Poli (1998). The appearance of metastable chloritoid depresses the stability of amphibole. According to Schmidt and Poli (1998) and Forneris and Holloway (2003), basalt can possess 0.5-0.8 and 0.3 wt% H2O at 650 °C and 3 GPa. Lawsonite is the most important hydrous phase in subducting basalt because it is stable at relatively high temperature (Pawley and Holloway 1993; Pawley 1994; Poli and Schmidt 1995; Schmidt and Poli 1998; Ono 1998; Forneris and Holloway 2003; Schmidt et al. 2004). In particular, its stability covers the choke points in the stability of hydrous minerals of the peridotite system from 3 to 9 GPa (Fig. 2), and therefore lawsonite could re-hydrate the downdragged peridotite layer under those pressures when it dehydrates. At a temperature region higher than the lawsonite stability field, Schmidt and Poli (1998) observed phengite in basalt. The modal proportion of phengite is, however, limited in basalt because MORB has a low concentration of K and also if K is available in the system, K is partitioned preferentially into fluid. In the system relevant to sediments, Domanik and Holloway (1996) and Ono (1998) reported the stability of phengite, Mg-sursassite, topaz-OH, and phase Egg. These hydrous phases are characterized by their higher temperature stability than hydrous phases in peridotite and basalt systems as seen in Figure 2. Phengite has dehydration conditions similar to that of phlogopite (Fig. 2). The reaction boundary between topaz-OH and phase Egg is identical to the olivine - wadsleyite boundary (Fig. 2). Ono (1998) demonstrated that subducting sediment can bring 2 wt% H2O in phengite to 7 GPa, 0.7 wt% H2O in topaz-OH to 9 GPa, and 0.4 wt% H2O in phase Egg up to 15 GPa, and that subducting basalt can bring about 1 wt% H2O in lawsonite to 6 GPa and 800 °C. Phase Egg could be stable at least up to the transition zone, while lawsonite could dehydrate at around 10 GPa. This means that phase Egg could be formed in the sediment layer by H2O coming from dehydration of lawsonite in the basaltic layer. In addition to these phases, the δ-AlOOH phase, a high-pressure polymorph of diaspore, was proposed to be an important H2O host in sediment or basalt systems (Suzuki et al. 2000). However, it is still uncertain whether this phase is stable in sedimentary or basaltic systems (Litasov and Ohtani 2005).

PRESSURE - TEMPERATURE CONDITIONS AND DEHYDRATION REACTIONS IN THE SUBDUCTING SLAB Obviously the PT conditions of the downgoing slab are critical to determine the dehydration processes of hydrous phases in the slab. Furukawa (1993), Peacock (1993) and Peacock and Wang (1999) suggested several PT paths for subducting slabs (Fig. 2). These calculations have large uncertainties of 100-200 °C in the temperature at 90 km (3 GPa), because steep temperature gradients exist near to slab surfaces. Iwamori (2004) suggested that a kinematic critical parameter comprising the product of subduction angle, potential temperature, slab velocity and slab age, must be exceeded for PT paths to pass below the choke point at 6.2 GPa and 620 °C. When the downgoing hydrous peridotite follows relatively warm PT paths, antigorite breaks down, followed by talc and chlorite (HT path in Fig. 2). Beyond the chlorite-out reaction the subducting peridotite will be almost free of H2O bound in crystals except for a small amount in phlogopite at around 2.5 – 6.5 GPa. This means that when downgoing hydrous peridotite goes on paths like this, the hydrous minerals should encounter a “choke point” at 2.5 GPa (Fig. 2). If there is enough K2O to stabilize phlogopite in the mantle, the downgoing hydrous peridotite will carry a small amount of H into the deeper mantle. At 6.5-11 GPa, the

Hydrous Phases & Water Transport in the Subducting Slab

281

phlogopite breaks down into K-richterite, which has an equal H/K atomic ratio and is stable at least up to 13 GPa (Sudo and Tatsumi 1990) and then dehydrates again into another hydrous phase containing lower H/K (Trønnes et al. 1988; Inoue et al. 1998). The crystal structure of this phase remains to be investigated. When the downgoing hydrous peridotite follows relatively cold PT paths, antigorite breaks down at 6 GPa (LT path in Fig. 2) and beyond which small amounts of H2O may retain in phlogopite. Lawsonite in subducting basaltic crust contains ~11 wt% H2O and is stable beyond the choke point (Figs. 2, 4). Therefore, since the lower pressure stability of DHMS overlaps with the high-pressure stability of lawsonite, DHMS such as chondrodite, clinohumite, phase A and phase E in the downdragged base of the mantle wedge could absorb H2O from decomposing lawsonite in the basaltic layer and become H2O carriers in a cold subduction zone (LT path in Fig. 2) to the deeper mantle beyond the choke point. Phengite, topaz-OH and phase Egg in downgoing sediment could also pass H2O into DHMS because of their high-temperature stability (Fig. 2).

COMPOSITION AND DIHEDRAL ANGLES OF AQUEOUS FLUIDS IN MANTLE PERIDOTITE Since the pioneering work by Nakamura and Kushiro (1974), the chemical compositions of silicates dissolved in aqueous fluids have been assumed to be characterized by an SiO2rich component at relatively shallow depths corresponding to pressures between 1 and 3 GPa (Ryabchikov et al. 1982; Zhang and Frantz 2000). In contrast, recent experimental data above 3 GPa suggest that aqueous fluids coexisting with enstatite (MgSiO3) and forsterite (Mg2SiO4) exhibit higher Mg/Si ratios as the pressure increases from 3 GPa up to 10 GPa (Fig. 1; Stalder et al. 2001; Mibe et al. 2002). When the dihedral angles between crystals and fluids are smaller than 60°, permeable flow is allowed even if the porosity is small. The dihedral angles at triple junctions between forsterite crystals and aqueous fluid change from >60° to <60° from 1 to 3 GPa at 1000 °C (Fig. 5A; Watson and Brenan 1987; Watson et al. 1990; Mibe et al. 1998, 1999). This finding is coincident with (1) the chemical change in the aqueous fluids from an SiO2-rich regime to an enstatite-rich regime and (2) the increase in silicate solubility at around 3 GPa (Fig. 1). The dihedral angle of H2O fluid in forsterite continues to decrease in the pressure range 3-5 GPa (Mibe et al. 1998; 1999). This may be due to an increasing amount of silicate in the aqueous fluids (Takei and Shimizu 2002) and/or a chemical change from an enstatite-rich to a more MgO-rich constitution (Fig. 1). Ono et al. (2002) reported that the dihedral angles of aqueous fluids in pyrope garnet (Mg/Si = 1) increase with increasing pressure from 4 to 9 GPa. Mibe et al. (2003) measured the dihedral angles of aqueous fluids in pyrope garnet and clinopyroxene, which are major constituents of eclogite (Fig. 5B). The PT conditions at which the dihedral angles of olivine - olivine - fluid (ol - fl), and garnet - garnet - fluid (gt - fl) are equal to 60°, are plotted in Figure 5C. Comparison of the PT conditions of the 60° isopleths and the stability of hydrous phases (Fig. 2) suggests the following: (1) The PT conditions of the 60° isopleths between garnet and fluids lie along the stability limit of lawsonite. The fluid liberated from lawsonite can readily migrate upward in the eclogite (basalt) system. (2) The PT conditions of the 60° isopleths between olivine and fluid looks similar to the stability limits of pargasite and chlorite. This means that the fluids liberated by their dehydration reactions in peridotite may readily migrate upward. (3) The liberated fluids in peridotite system re-enter into the stability fields of chlorite or pargasite when they migrate upward, and again they are trapped in these hydrous minerals. The aqueous fluid will be released again when chlorite/pargasite re-dehydrate in the downgoing flow beyond their stability fields. Davies and Stevenson (1992) called this process the lateral

282

Kawamoto

70

Dihedral angle (degree)

70

(a)

Mibe et al Watson et al

65 60

800

55 50 45 1200

40 35

1000

1

2

3

4

5

6

Pressure (GPa) 0

(c)

2

Pressure (GPa)

65 60 55

> 60 °

< 60 ° 6

> 60 °

10 700

gt - fl

800

900

1000

1100

800

Gt-fluid

45 1000

900

40

2

4

Mibe et al Ono et al

6

8

10

12

14

Pressure (GPa)

Figure 5. (A, B) Dihedral angle in (A) olivineolivine-fluid (Watson et al. 1990; Mibe et al. 1998, 1999) and (B) garnet-garnet-fluid (Ono et al. 2002; Mibe et al. 2003) versus pressure. Schematic contours at constant temperatures (numerals) are drawn. The contours at 700 and 800 °C of gt-fluid are assumed to be parallel to the contour at 900 °C. (C) The 60° isopleth of the dihedral angle in olivine-olivine-fluid (olfl) and garnet-garnet-fluid (gt-fl). The plotted data in C are from A and B.

ol - fl

4

8

700

50

35 0

(b)

Ol- fluid

Dihedral angle (degree)

75

1200

1300

Temperature (°C)

H2O transport and suggested that this mechanism delivers H2O laterally to the partial melting zone. Such a process was also considered in numerical calculations by Iwamori (1998). A hydrous mineral is carried deeper by induced mantle flow until the stability limit of the hydrous mineral is reached. H2O is then liberated and fluids are trapped as structural H2O in the hydrous minerals and as immobile fluids due to their dihedral angle >60°. These would get dragged down in the induced flow, and this process would be repeated until the fluid reaches the zone of partial melting.

SECOND CRITICAL ENDPOINT BETWEEN MAGMAS AND AQUEOUS FLUID: IMPLICATIONS FOR SLAB-DERIVED COMPONENT Simple silicate melts and aqueous fluids can mix completely under certain PT conditions (Fig. 6A,B). At pressure conditions equivalent to the Earth’s upper mantle, silicate melts and aqueous fluids cannot be distinguished from each other at the temperature-pressure conditions beyond a second critical endpoint, where a critical temperature meets its wet solidus (Kennedy et al. 1962; Paillat et al. 1992; Shen and Keppler 1997; Bureau and Keppler 1999). Following the visual demonstration of the complete mixing between albite melt and H2O (Shen and Keppler 1997), Bureau and Keppler (1999) reported complete miscibility between aqueous fluids and K2O-bearing nepheline melt, pure jadeite melt, haplogranitic melt, Ca-bearing haplogranitic melt and dacite in the SiO2-Al2O3-Na2O-K2O-CaO-MgO system. Sowerby and Keppler (2002) demonstrated complete miscibility between B2O3 - F enriched albite melt or pegmatite and H2O

Hydrous Phases & Water Transport in the Subducting Slab

Temperature

(A)

dry solidus

Tc fluid

fluid +melt

fluid +H H2O

H A supercritical fluid +A

(B)

increasing pressure

wet solidus

melt +A fluidabsent solidus melt +H A + H

melt

supercritical fluid Tc melt fluid fluid+melt fluid +H H2O (C)

melt +H

A + H

H A supercritical fluid +A

supercritical fluid No wet solidus but practical solidus

A + H

fluid +H H2O

H A

Figure 6. Schematic phase diagrams in the system of mineral A and H2O (Kawamoto et al. 2004). H is a hydrous mineral. (A, B) As pressure increases, a critical temperature (Tc) between H2O-bearing silicate melt and silicate-bearing H2O fluid decreases. (C) The Tc meets the H2O-saturated solidus temperature in the system at a second critical endpoint. At pressures beyond that of the second critical endpoint, there is no difference between melts and fluids. In this case there is no H2O-saturated solidus temperature. The practical solidus represents a temperature above which a detectable amount (more than a few percent) of silicate melt is formed (Iwamori 1998).

283

(Sowerby and Keppler 2002). Kawamoto (2004b) reported similar observations of mixing relationships between aqueous fluids and natural andesitic/dacitic melt (Fig. 7). The critical PT conditions observed in the andesite/dacite- H2O system are similar to those observed in the other simple silicateH2O systems (Fig. 8; Shen and Keppler 1997; Bureau and Keppler 1999). Experiments to determine H2O-saturated solidus temperatures often identify them from abrupt changes in chemical composition of the minerals and/or the appearance of dendritic textures with increasing temperature at a given pressure (Inoue 1994; Kawamoto and Holloway 1997; Irifune et al 1998; Stalder et al. 2001; Mibe et al. 2002). Some workers have distinguished two types of dendritic texture, one quenched from partial melt and the other from aqueous fluids (Irifune et al. 1998; Litasov and Ohtani 2002). However, they mentioned that it is difficult to distinguish between these types of texture at pressures greater than 10-13 GPa. As the critical temperature between aqueous fluids and silicate melts decreases with increasing pressure (Paillat et al. 1992; Shen and Keppler 1997; Bureau and Keppler 1999), it should meet an H2O-saturated solidus temperature with increasing pressure (Fig. 6C). It is difficult to melt basaltic compositions in a Bassett-type diamond anvil cell due to its temperature limitation of 1100 °C (Bureau and Keppler 1999). Therefore, a supercritical behavior between aqueous fluids and mafic magmas equilibrated with mantle peridotite had remained to be investigated for years. Recently Mibe and his coworkers experimentally determined the PT conditions of a second critical endpoint between peridotite/basalt melts and aqueous fluids by the use of a Kawai-type large volume press and synchrotron X-ray radiography (Kanzaki et al. 1987; Mibe et al. 2004a). They reported that a second critical endpoint between peridotite/basalt melts and aqueous fluids may be located at 3.8 and 3 GPa, respectively (Mibe et al. 2004b, 2005, 2006). This pressure range is lower than that estimated by Kessel et al. (2005). Although

284

Kawamoto

50 °C

1000 °C

920 °C

910 °C

890 °C

950 °C

1025 °C

980 °C

970 °C

Figure 7. Microphotographs showing supercritical behavior between Fuji 1707 andesite and H2O using Bassett-type externally heated diamond anvil cell (Kawamoto 2004b). (A) Chips of the andesitic glass and water are in the rhenium gasket (gasket hole is 0.5 mm) with a small bubble (right) at 50 °C. (B) At 1000 °C and about 1 GPa, a homogeneous fluid, with several grains of unidentified crystals. (C) On cooling to 920 °C, a milky appearance due to tiny droplets of andesite melt in aqueous fluid is seen. (D, E) At 910-890 °C, melt globules are growing in the aqueous fluid. (F) Then during re-heating to 950 °C, the boundary disappears the fluid homogenizes. The crystals are also melting. (G) After heating at 1025 °C, there are no crystals left, and (H, I) during the subsequent cooling, the sample turns milky and separates into andesite globules and aqueous fluid. The difference among the critical temperatures on the first cooling (920 °C, in C, D), the heating (950 °C, in F), and the second cooling (980 °C, in H) could be due to a pressure decrease during the experiment.

Kessel et al. (2005) suggested that there is still a melt-fluid solvus at 4 and 5 GPa, they did not show the coexistence of two phases at 4 or 5 GPa. Therefore, the data shown in Figure 5 of Kessel et al. (2005) can be interpreted as evidence that the fluid compositions observed at 4 and 5 GPa vary continuously with temperature as in Figure 6C and these pressures are already beyond the second critical endpoint. In contrast, Mibe et al. (2004b, 2005, 2006) observed melts and fluids up to 3.8 and 3 GPa in peridotite- H2O and basalt- H2O systems, respectively, and found no coexisting two phases at higher pressures. Although X-ray radiography method is not able to detect a small difference between fluids and melts under certain conditions, Mibe et al. (2004a, 2006) tightly constrain the second critical endpoint between peridotite melt and aqueous fluids at 3.8 GPa, 1000 °C and with 55 wt% H2O. The pressure of 3.8 GPa is equivalent to the depth of the Wadati-Benioff zone beneath the volcanic front. If supercritical fluids are common at the base of the mantle wedge beneath volcanic arcs, the traditional H2O-saturated solidus temperature may represent a temperature where the concentration of silicate components dissolved into aqueous fluids increases drastically and should therefore be described as a practical solidus (Fig. 6C; Iwamori 1998). If supercritical

Hydrous Phases & Water Transport in the Subducting Slab

0

Critical temperature (°C) 500 600 700 800 900 1000 1100 Andesite

Dacite

0.5 1 Pressure (GPa)

285

Ab

1.5

Jd

Ne 2

Ab Hgr

Ca-Hgr

2.5 3

Figure 8. Critical temperatures observed between aqueous fluids and albite (Ab, Shen and Keppler 1997), nepheline (Ne), jadeite (Jd), and haplogranitic (Hgr) melts, Ca bearing haplogranitic melts (Ca-Hgr) and dacite (Bureau and Keppler 1999), and natural calc-alkaline andesite/dacite (Kawamoto 2004b). The estimated second critical endpoints between albite (Stalder et al. 2000), basalt (Mibe et al. 2005), and peridotite (Mibe et al. 2004a, 2006) and aqueous fluids are also plotted.

Basalt

3.5 Peridotite 4

fluids commonly exist in the mantle wedge in subduction zones, such a supercritical fluid could separate into a silicate melt and an aqueous fluid when PT conditions become below the second critical endpoint along its migration to the surface (Fig. 8; Bureau and Keppler 1999). In this case, partitioning of elements between aqueous fluids and silicate melts should occur (Bureau et al. 2004). Such elemental fractionation may affect the chemical characteristics of the volcanic rocks. The existence of a second critical endpoint underneath the volcanic arcs suggests that dense supercritical fluids can come from the slab and separate into the aqueous fluid and melt in the mantle wedge. Otherwise, the slab component would be an aqueous fluid in cold subduction zones or a partial melt in warm subduction zones. Detailed studies of the critical curvatures in the peridotite, basalt, sediment systems will shed light on establishing a quantitative model for the magma generation and H2O transport in subduction zones (Manning 2004).

CONCLUDING REMARKS Our knowledge of the stability of hydrous phases in the downgoing slab has increased dramatically in the last decade (Figs. 2, 3, 4). Recently we have also learned much about the chemical features of aqueous fluids under upper mantle conditions. First, the chemical compositions of silicate components dissolved into aqueous fluids coexisting with mantle peridotite change from silica-rich at pressures lower than 3 GPa to magnesium-rich at pressures greater than 3 GPa (Fig. 1; Stalder et al. 2001; Mibe et al. 2002; Kawamoto et al. 2004). This means that the aqueous fluids in the mantle have peridotitic compositions beneath volcanic arcs. Second, dihedral angles formed between olivine and aqueous fluids change from >60° to <60° at 1000 °C and 3 GPa (Fig. 5; Mibe et al. 1998). This change was suggested to control the location of volcanic arcs (Mibe et al. 1999). Both the change of chemical composition and of wetting properties of mantle fluids with pressure may be related to the onset of complete miscibility between silicate melt and aqueous fluids as the pressure and temperature conditions approach the critical endpoint (Fig. 8).

286

Kawamoto ACKNOWLEDGMENT

I thank Ken Koga, Kenji Mibe, Tetsuya Komabayashi, Shigeaki Ono, John Winter and Junji Yamamoto for their careful readings of the manuscript, Hikaru Iwamori for his courtesy to make Figure 3 for this chapter, and Hans Keppler and Joe Smyth for their encouragements and editorial efforts. Careful and constructive reviews by Alison Pawley and an anonymous reviewer improved the manuscript greatly. This paper presents my ideas developed during my recent experimental experiences with Kenji Mibe, Masami Kanzaki, Shigeaki Ono, and Kyoko Matsukage supported by Nissan Science Foundation, Ministry of Education, Culture, Sports, Science and Technology, and SPring-8.

REFERENCES Bromiley GD, Pawley AR (2002) The high-pressure stability of Mg-sursassite in a model hydrous peridotite: a possible mechanism for the deep subduction of significant volumes of H2O. Contrib Mineral Petrol 142: 714-723 Bromiley GD, Pawley AR (2003) The stability of antigorite in the systems MgO-SiO2-H2O (MSH) and MgOAl2O3-SiO2-H2O (MASH): The effects of Al3+ substitution on high-pressure stability. Am Mineral 88: 99-108 Bureau H, Keppler H (1999) Complete miscibility between silicate melts and hydrous fluids in the upper mantle; experimental evidence and geochemical implications. Earth Planet Sci Lett 165:187-196 Bureau H, Ménez B, Malavergne V, Somogyi A, Munoz M, Simionovici A, Massare D, Burchard M, Kubsky S, Shaw C (2004) In situ determination of Pb, Sr, Rb, Zr partitioning between hydrous melts and aqueous fluids at high pressure and temperature. Eos Trans. AGU Fall Meet. Suppl. V11C-05 Davies JH, Stevenson DJ (1992) Physical model of source region of subduction zone volcanics. J Geophys Res 97:2037-2070 Demouchy S, Deloule E, Frost DJ, Keppler H (2005) Pressure and temperature-dependence of water solubility in Fe-free wadsleyite. Am Mineral 90:1084-1091 Domanik KJ, Holloway JR (1996) The stability and compositions of phengitic muscovite and associated phases from 5.5 to 11 GPa: Implications for deeply subducted sediments. Geochim Cosmochim Acta 60:41334150 Eggleton R, Boland J, Ringwood A (1978) High pressure synthesis of a new aluminium silicate Al5Si5O17(OH). Geochem J 12:191-194 Forneris JF, Holloway JR (2003) Phase equilibria in subducting basaltic crust: implications for H2O release from the slab. Earth Planet Sci Lett 214:187-201 Frost DJ (1999) The stability of dense hydrous magnesium silicates in Earth’s transition zone and lower mantle. In: Mantle Petrology; Field Observations and High-pressure Experimentation; a Tribute to Francis R. (Joe) Boyd. Fei Y, Bertka C, Mysen BO (eds) Geochemical Society, p 283-296 Fumagalli P, Stixrude L, Poli S, Snyder D (2001) The 10 Å phase: a high-pressure expandable sheet silicate stable during subduction of hydrated lithosphere. Earth Planet Sci Lett 186:125-141 Fumagalli P, Poli, S (2005) Experimentally determined phase relations in hydrous peridotites to 6.5 GPa and their consequences on the dynamics of subduction zones. J Petrol 46:555-578 Furukawa Y (1993) Magmatic processes under arcs and formation of the volcanic front. J Geophys Res 98: 8309-8319 Gill JB (1981) Orogenic Andesite and Plate Tectonics: Springer-Verlag Gottschalk M, Fockenberg T, Grevel KD, Wunder B, Wirth R, Schreyer W, Maresch WV (2000) Crystal structure of the high-pressure phase Mg4(MgAl)Al4[Si6O21(OH)7]: an analogue of sursassite. Eur J Mineral 12:935-945 Inoue T (1994) Effect of water on melting phase relations and melt composition in the system Mg2SiO4MgSiO3-H2O up to 15 GPa. Phys Earth Planet Inter 85:237-263 Inoue T, Yurimoto H, Kudoh Y (1995) Hydrous modified spinel, Mg1.75SiH0.5O4: a new water reservoir in the mantle transition region. Geophys Res Lett 22:117-120 Inoue T, Irifune T, Yurimoto H, Miyagi I (1998) Decomposition of K-amphibole at high pressures and implications for subduction zone volcanism. Phys Earth Planet Inter 107:221-231 Irifune T, Kubo N, Isshiki M, Yamasaki Y (1998) Phase transformations in serpentine and transportation of water into the lower mantle. Geophys Res Lett 25:203-206 Ito E, Takahashi E (1989) Postspinel transformations in the system Mg2SiO4-Fe2SiO4 and some geophysical implications. J Geophys Res 94:10,637-10,646 Iwamori H (1998) Transportation of H2O and melting in subduction zones. Earth Planet Sci Lett 160:65-80

Hydrous Phases & Water Transport in the Subducting Slab

287

Iwamori H (2004) Phase relations of peridotites under H2O-saturated conditions and ability of subducting plates for transportation of H2O. Earth Planet Sci Lett 227:57-71 Kanzaki M, Kurita K, Fujii T, Kato T, Shimomura O, Akimoto S (1987) A new technique to measure the viscosity and density of silicate melts at high pressure. In: High-Pressure Research in Mineral Physics. Manghnani MH, Syono Y (eds) Am Geophys Union, p 195-200 Kanzaki M (1991) Stability of hydrous magnesium silicates in the mantle transition zone. Phys Earth Planet Inter 66:307-312 Katsura T, Ito E (1989) The system Mg2SiO4-Fe2SiO4 at high pressures and temperatures: precise determination of stabilities of olivine, modified spinel, and spinel. J Geophys Res 94:15,663-15,670 Kawamoto T, Hervig RL, Holloway JR (1996) Experimental evidence for a hydrous transition zone of the early Earth’s mantle. Earth Planet Sci Lett 142:587-592 Kawamoto T, Holloway JR (1997) Melting temperature and partial melt chemistry of H2O-saturated mantle peridotite to 11 gigapascal. Science 276:240-243 Kawamoto T (2004a) Hydrous phase stability and partial melt chemistry of H2O-saturated KLB-1 peridotite up to the uppermost lower mantle conditions. Phys Earth Planet Inter 143-144:387-395 Kawamoto T (2004b) Critical phenomena between magmas and aqueous fluid (in Japanese with English abstract). Rev High Press Sci Tech 14:235-241 Kawamoto T, Matsukage KN, Mibe K, Isshiki M, Nishimura K, Ishimatsu N, Ono S (2004) Mg/Si ratios of aqueous fluids coexisting with forsterite and enstatite based on the phase relations in the Mg2SiO4-SiO2H2O system. Am Mineral 89:1433-1437 Kennedy GC, Wasserburg GJ, Heard HC, Newton RC (1962) The upper three-phase region in the system SiO2H2O. Am J Sci 260:501-521 Kessel R, Ulmer P, Pettke T, Schmidt MW, Thompson AB (2005) The water-basalt system at 4 to 6 GPa: Phase relations and second critical endpoint in a K-free eclogite at 700 to 1400 °C. Earth Planet Sci Lett 237: 873-892 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in α, β and γ phases of (Mg, Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Komabayashi T, Omori S, Maruyama S (2004) Petrogenetic grid in the system MgO-SiO2-H2O up to 30 GPa, 1600 °C: Applications to hydrous peridotite subducting into the Earth’s deep interior. J Geophys Res 109, doi:10.1029/2003JB002651 Komabayashi T, Hirose K, Funakoshi K-i, Takafuji N (2005) Stability of phase A in antigorite (serpentine) composition determined by in-situ X-ray pressure observations. Phys Earth Planet Inter 151:276-289 Kudoh Y, Inoue T, Arashi H (1996) Structure and crystal chemistry of hydrous wadsleyite, Mg1.75SiH0.5O4: Possible hydrous magnesium silicate in the mantle transition zone. Phys Chem Minerals 23:461-469 Litasov K, Ohtani E (2002) Phase relations and melt compositions in CMAS-pyrolite-H2O system up to 25 GPa. Phys Earth Planet Inter 134:105-127 Litasov K, Ohtani E (2005) Phase relations in hydrous MORB at 18-28 GPa: implications for heterogeneity of the lower mantle. Phys Earth Planet Inter 150:239-263 Liu L-G (1987) Effects of H2O on the phase behaviour of the forsterite - enstatite system at high pressures and temperatures and implications for the Earth. Phys Earth Planet Inter 49:142-167 Manning CE (2004) The chemistry of subduction-zone fluids. Earth Planet Sci Lett 223:1-16 Mibe K, Fujii T, Yasuda A (1998) Connectivity of aqueous fluid in the Earth’s upper mantle. Geophy Res Lett 25:1233-1236 Mibe K, Fujii T, Yasuda A (1999) Control of the location of the volcanic front by aqueous fluid connectivity in the mantle wedge. Nature 401:259-262 Mibe K, Fujii T, Yasuda A (2002) Composition of aqueous fluid coexisting with mantle minerals at high pressure and its bearing on the differentiation of the Earth’s mantle. Geochim Cosmochim Acta 66:2273-2285 Mibe K, Yoshino T, Ono S, Yasuda A, Fujii T (2003) Connectivity of aqueous fluid in eclogite and its implications for fluid migration in the Earth’s interior. J Geophys Res 108, doi:10.1029/2002JB001960 Mibe K, Kanzaki M, Kawamoto T, Matsukage KN, Fei Y, Ono S (2004a) Second critical endpoint in peridotiteH2O system and its bearing on the magmatism in subduction zones. AGU Fall meeting (abstract), V11C07 Mibe K, Kanzaki M, Kawamoto T, Matsukage KN, Fei Y, Ono S (2004b) Determination of the second critical end point in silicate-H2O systems using high pressure and high-temperature X-ray radiography. Geochim Cosmochim Acta 68:5189-5195 Mibe K, Kanzaki M, Kawamoto T, Matsukage KN, Fei Y, Ono S (2005) Second critical endpoint in the basaltH2O system. AGU Fall meeting (abstract), V33C-03 Mibe K, Kanzaki M, Kawamoto T, Matsukage KN, Fei Y, Ono S (2006) Second critical endpoint in peridotiteH2O system. (in prep.) Nakamura Y, Kushiro I (1974) Composition of the gas phase in Mg2SiO4-SiO2-H2O at 15 kbar. Carnegie Inst Wash Yearb 73:255-258

288

Kawamoto

Nicholls IA, Ringwood AE (1973) Effect of water on olivine solubility in tholeiites and the production of SiO2saturated magmas in the island arc environment. J Geol 81:285-300 Niida K, Green DH (1999) Stability and chemical compositions of pargasitic amphibole in MORB pyrolite under mantle conditions. Contrib Mineral Petrol 135:18-40 Ohtani E, Toma M, Litasov K, Kubo T, Suzuki A (2001) Stability of dense hydrous magnesium silicate phases and water storage capacity in the transition zone and lower mantle. Phys Earth Planet Inter 124:105-117 Ono S (1998) Stability limits of hydrous minerals in sediment and mid-ocean ridge basalt compositions: implications for water transport in subduction zones. J Geophys Res 103:18253-18267 Ono S (1999) Water in the Earth’s mantle (in Japanese with English abstract). Rev High Press Sci Tech 9:3-10 Ono S, Mibe K, Yoshino K (2002) Wetting of pyrope by H2O fluid: implications for water transport by subducted oceanic crust. Earth Planet Sci Lett 203:895-903 Paillat O, Elphick SC, Brown WL (1992) Solubility of water in NaAlSi3O8 melts: a re-examination of Ab-H2O phase relationships and critical behavior at high pressures. Contrib Mineral Petrol 112:490-500 Pawley A (1994) The pressure and temperature stability limits of lawsonite: implications for H2O recycling in subduction zones. Stability of lawsonite to 100 kbars. Contrib Mineral Petrol 118:99-108 Pawley AR. (2003) Chlorite stability in mantle peridotite: the reaction clinochlore + enstatite = forsterite + pyrope + H2O. Contrib Mineral Petrol 144:449 - 456 Pawley AR, Holloway JR (1993) Water sources for subduction zone volcanism: New experimental constraints. Science 260:664-667 Pawley AR, Wood BJ (1995) The high-pressure stability of talc and 10 Å phase - potential storage sites for H2O in subduction zones. Am Mineral 80:998-1003 Peacock SM (1993) Large-scale hydration of the lithosphere above subducting slabs. Chem Geol 108:49-59 Peacock SM, Wang K (1999) Seismic consequences of warm versus cool subduction metamorphism: examples from Southwest and Northeast Japan. Science 286:937-939 Poli S, Schmidt MW (1995) Water transport and release in subduction zones: experimental constraints on basaltic and andesitic systems. J Geophys Res 100:22299-22314 Ringwood AE, Major A (1967) Some high-pressure transformations of geophysical interest. Earth Planet Sci Lett 2:106-110 Ryabchikov ID, Schreyer W, Abraham K (1982) Compositions of aqueous fluid in equilibrium with pyroxenes and olivines at mantle pressures and temperatures. Contrib Mineral Petrol 79:80-84 Sakuyama M, Nesbitt RW (1986) Geochemistry of the Quaternary volcanic rocks of the northeast Japan arc. J Volcanol Geotherm Res 29:413-450 Schmidt M, Poli S (1998) Experimentally based water budgets for dehydrating slabs and consequences for arc magma generations. Earth Planet Sci Lett 163:361-379 Schmidt MW, Vielzeuf D, Auzanneau E (2004) Melting and dissolution of subducting crust at high pressures: the key role of white mica. Earth Planet Sci Lett 228:65-84 Shen AH, Keppler H (1997) Direct observation of complete miscibility in the albite-H2O system. Nature 385: 710-712 Smyth JR (1987) β-Mg2SiO4: a potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR, Kawamoto T (1997) Hydrous wadsleyite II: A new, dense, high pressure, hydrous magnesium iron silicate phase. Earth Planet Sci Lett 146:E9-E16 Smyth JR, Kawamoto T, Jacobsen SD, Swope RJ, Hervig RL, Holloway JR (1997) Crystal structure of monoclinic hydrous wadsleyite (β-(Mg, Fe)2SiO4). Am Mineral 82:270-275 Sowerby JR, Keppler H (2002) The effect of fluorine, boron and excess sodium on the critical curve in the albiteH2O system. Contrib Mineral Petrol 143:32-37 Stalder R, Ulmer P, Thompson AB, Gunther D (2000) Experimental approach to constrain second critical endpoints in fluid/silicate systems: near solidus fluids and melts in the system albite-H2O. Am Mineral 85:68-77 Stalder R, Ulmer P (2001) Phase relations of a serpentine composition between 5 and 14 GPa: significance of clinohumite and phase E as water carriers into the transition zone. Contrib Mineral Petrol 140:670-679 Stalder R, Ulmer P, Thompson AB, Gunther D (2001) High pressure fluids in the system MgO-SiO2-H2O under upper mantle conditions. Contrib Mineral Petrol 140:607-618 Sudo A, Tatsumi Y (1990) Phlogopite and K-amphibole in the upper mantle: Implication for magma generation in subduction zones. Geophys Res Lett 17:29-32 Suzuki A, Ohtani E, Kamada T (2000) A new hydrous phase delta-AlOOH synthesized at 20.9 GPa and 1000 °C. Phys Chem Mineral 27:689-693 Takei Y, Shimizu I (2003) The effects of liquid composition, temperature, and pressure on the equilibrium dihedral angles of binary solid - liquid systems inferred from a lattice-like model. Phys Earth Planet Inter 139:225-42 Tatsumi Y (1986) Formation of the volcanic front in subduction zones. Geophys Res Lett 13:717-720 Tatsumi Y (1989) Migration of fluid phases and genesis of basalt magmas in subduction zones. J Geophys Res 94:4697-4707

Hydrous Phases & Water Transport in the Subducting Slab

289

Tatsumi Y, Eggins S (1995) Subduction zone magmatism. Blackwell Science Trønnes RG, Takahashi E, Scarfe CM (1988) Stability of K-richterite and phlogopite to 14 GPa. Trans Am Geophys Union 69:1510-1511 Ulmer P, Trommsdorff V (1995) Serpentine stability to mantle depths and subduction-related magmatism. Science 268:858-861 Watson EB, Brenan JM (1987) Fluids in the lithosphere, 1. Experimentally determined wetting characteristics of CO2-H2O fluids and their implications for fluid transport, host-rock physical properties, and fluid inclusion formation. Earth Planet Sci Lett 85:497-515 Watson EB, Brenan JM, Baker DR (1990) Distribution of fluids in the continental mantle. In: Continental Mantle. Menzies MA (ed) Oxford University Press, p 111-125 Yamamoto K, Akimoto S (1977) The system MgO-SiO2-H2O at high pressures and temperatures - stability field for hydroxyl-chondrodite, hydroxyl-clinohumite and 10 Å -phase. Am J Sci 277:288-312 Zhang YG, Frantz JD (2000) Enstatite-forsterite-water equilibria at elevated temperatures and pressures. Am Mineral 85:918-925 Wunder B, Medenbach O, Krause W, Schreyer W (1993a) Synthesis, properties and stability of Al3Si2O7(OH)3 (phase Pi), a hydrous high-pressure phase in the system Al2O3-SiO2-H2O (ASH). Eur J Mineral 5:637649 Wunder B, Rubie DC, Ross IC, Medenbach O, Seifert F, Schreyer W (1993b) Synthesis, stability, and properties of Al2SiO4(OH)2: A fully hydrated analogue of topaz. Am Mineral 78:285-297

2

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 29-52, 2006 Copyright © Mineralogical Society of America

The Structure of Hydrous Species in Nominally Anhydrous Minerals: Information from Polarized IR Spectroscopy Eugen Libowitzky and Anton Beran Institut für Mineralogie und Kristallographie Universität Wien - Geozentrum Althanstraße 14, A-1090 Wien, Austria [email protected]

[email protected]

INTRODUCTION Hydrogen is a major constituent in a wide variety of minerals in the Earth’s crust. Usually H is bonded to oxygen forming H2O molecules or OH− groups. In rare cases, H3O+, H3O2− and H5O2+ units are also formed. The occurrence of NH4+, CHx, etc. in minerals will not be considered here. Hydrogen occurs stoichiometrically in hydrous compounds such as hydrates (e.g., gypsum), (oxy)hydroxides (e.g., goethite), and in many rock-forming silicates (e.g., micas), as well as in nonstoichiometric major amounts in microporous minerals such as zeolites and clay minerals, which are of considerable economic and ecologic importance. Hydrogen also occurs as a minor or trace constituent in minerals that by definition (and by their formulae) do not contain hydrogen at all, i.e., the so-called nominally anhydrous minerals (NAMs). NAMs include common rock-forming minerals in the Earth’s crust (e.g., quartz, feldspars) and upper mantle (e.g., olivine, pyroxene and garnet), but also high-P and high-T phases (e.g., wadsleyite, ringwoodite, and majorite garnet) stable in the mantle transition zone (410-660 km depth). The aim of the present chapter is to review the use of polarized infrared (IR) spectroscopy as it applies to detecting traces of hydrogen in minerals and to characterizing its speciation and structural environment in nominally anhydrous minerals. The basic theoretical background will be supplemented by a number of examples from recent research. It is important to note that many of the concepts described here for NAMs can also be applied to synthetic compounds of importance in the materials sciences.

The importance of hydrous species in NAMs Hydrogen may be incorporated as defects in nominally anhydrous minerals of the Earth’s mantle. Due to the large volume of rock in the Earth’s mantle, even trace concentrations of H in NAMs have the potential to constitute a significant reservoir of H2O comparable in size to all oceans combined (Beran 1999; Ohtani 2005). Such a large potential water reservoir in the mantle would influence many geologic processes and has several major implications for the evolution of the Earth and atmosphere. Moreover, it is likely that the hydrogen content of the mantle has not been constant over geologic time, but constantly changes in a dynamic water cycle between the hydrosphere and the interior via subduction and volcanism (e.g., Thompson 1992; Dixon et al. 2002; Jacobsen and van der Lee 2006). Water strongly influences melting temperatures and rheology (e.g., Hirth and Kohlstedt 1996) and thus plutonism, volcanism, and convection. Water penetrates rocks, leaches 1529-6466/06/0062-0002$05.00

DOI: 10.2138/rmg.2006.62.2

30

Libowitzky & Beran

elements and leads to the formation of hydrothermal minerals and ore deposits. Release of water and reduced mechanical strength of minerals by dehydration and hydrolytic weakening (Griggs 1967) may even trigger earthquakes in the subducting slab (e.g., Dobson et al. 2002; Jung and Green 2004). Even small amounts of hydrogen, in the form of structural defects, can dramatically change the physical properties of minerals (e.g., Mackwell et al. 1985; Karato 1990). Hydration also influences sound wave velocities and the elastic constants of minerals (e.g., Jacobsen 2006, this volume), which may enable seismologists to detect lateral variations in hydrogen concentrations in the mantle if the effects are large enough and H is present in sufficient quantity (e.g., van der Meijde 2003). In general, incorporation of hydrogen affects the thermodynamic properties of minerals and also influences their kinetic behavior during diffusion (Kohlstedt and Mackwell 1998) and phase transformations (Smyth and Frost 2002). It has to be emphasized that hydrogen-related changes of physical properties are not only interesting with regard to natural processes in the Earth interior, but also for critical parameters in technical applications of crystals (e.g., Buse et al. 1997).

Why use IR spectroscopy? If hydrogen occurs as a stoichiometric constituent in crystalline solids it can be investigated by diffraction methods (single-crystal X-ray and preferably neutron scattering techniques) and by spectroscopic methods (IR and Raman spectroscopy, proton-NMR spectroscopy). A combination of both techniques provides mutually valuable information on hydrogen in the structure. However, because diffraction depends upon long range order and probes the average structural sites of the major elements in a crystal, it cannot be applied to investigate structural details of non-crystalline materials such as glasses, melts and amorphous minerals, or of trace constituents, such as hydrogen in defect sites of a host crystal. Spectroscopic methods, on the other hand, are highly site specific and do not rely on crystalline periodicity, so they can also be applied to disordered species in liquids and gases. Among vibrational spectroscopic techniques (IR and Raman spectroscopy), IR spectroscopy is ideally suited to investigate even very low concentrations of hydrogen in geologic samples because it is highly sensitive to polar O-H bonds in the structure. When thick samples greater than ~1 mm are used, IR absorption can detect as little as a few wt. ppm H2O. Moreover, due to the low mass of the H atom, the O-H vibrations and their absorption bands can be easily assigned in the high-energy part of the spectrum between 3000-4000 cm−1 and therefore in most cases do not overlap with other peaks from the host lattice.

History Whereas first simple IR measurements on minerals date back to the beginning of the 20th century, systematic IR investigations of silicate minerals were not made until the 1930s (see Geiger 2004 for a review). However, only in the mid-1950s did IR studies begin to focus on hydrogen in minerals. Early studies were aimed at hydrous minerals such as muscovite (Tsuboi 1950), layered silicates (Serratosa and Bradley 1958), and azurite (Tillmanns and Zemann 1965), which helped to constrain the positions of the hydrogen atoms at a time when it was not possible with X-ray diffraction. Later, IR spectroscopy was also applied to detection of hydrogen in NAMs and confirmed traces of hydrous species in almost every supposedly “dry” mineral (for an earlier review see Rossman 1988). Finally, comparison to well-known structures of stoichiometrically hydrous phases, improvement of quantitative IR data and use of polarized radiation facilitated the development of models for structural incorporation mechanisms. The application and calibration of IR spectroscopy towards analytical water determination was an additional advance in quantitative IR spectroscopy, e.g., Wilkins and Sabine (1973), see also Rossman (2006) in this volume.

Structure of Hydrous Species Using Polarized IR Spectroscopy

31

CONCEPTS OF INFRARED SPECTROSCOPY Introduction to IR spectroscopy Setup of measurements. In a standard IR absorption experiment, a beam of polychromatic IR radiation is emitted from an IR light source, modulated in intensity across the energy (frequency, wavenumber) range by an interferometer and directed through a plane-parallel sample (Fig. 1). Certain energies of light are absorbed by excitation of characteristic vibrations of the material (e.g., by O-H stretching vibrations), and the transmitted light intensity is registered at an IR detector. In modern interferometer-based instruments, the interferogram observed by the detector is mathematically Fourier-transformed to an IR spectrum (thus they are called Fourier transform infrared “FTIR” spectrometers). In an IR absorption experiment, a background spectrum is acquired first (the same setup as above but without sample), recording the incident light intensity at any wavenumber. The sample spectrum is then related to (divided by) this background spectrum giving the transmittance spectrum of the sample. Finally, spectra are usually further converted to an absorbance spectrum (see below). Basic units and relations. The transmittance (T) is defined by the ratio of the transmitted intensity to the incident intensity (Table 1). T is dimensionless, but conveniently multiplied by 100%. However, a more useful quantity is absorbance (A), which is the negative logarithm of T. Absorbance is linearly proportional to the thickness (t) of the sample and the concentration c of the absorbing species in the sample (see e.g., Libowitzky and Rossman 1996). This relation is expressed by the Beer-Lambert’s law: A = ε · t · c, where ε is the molar absorption coefficient. It is obvious (and an advantage of IR absorption spectroscopy) that very low concentrations of an absorbing species can be compensated by using a thicker sample. The energy scale is conventionally given in wavenumbers (unit: cm−1), because this is proportional to energy and frequency (whereas wavelength is inversely proportional). The frequency of an absorption band equals a vibrational frequency in the sample which is pro-

Power supply and instrument control

t S

IR light source

I0 I

A (P) MM BS

FM

S

D

PC

Interferometer Figure 1. Basic experimental setup of an FTIR spectrometer. A = aperture, BS = beam splitter, D = detector, FM = fixed mirror, MM = moving mirror, (P) = polarizer (optional), S = sample. The close-up shows the exponential decrease of light intensity by absorption with increasing sample thickness.

32

Libowitzky & Beran Table 1. Basic relations and units used in infrared spectroscopy.

T = I / I0

T ... transmittance

A ... absorbance

T [%] = I / I0 · 100%

I0 ... incident intensity

I ... transmitted intensity

−A

A = - log T

T = 10

A=ε·t·c

(Beer-Lambert’s law)

ε ... molar absorption coefficient [L cm−1 mol−1]

Ai = εi · t · c

(Beer-Lambert’s law)

εi ... integr. molar absorption coefficient [L cm−2 mol−1]

Ax = A · cos2α

Ay = A · cos2β

Az = A · cos2γ

Atot = Ax + Ay + Az

ν = 12 π f / µ

ν ... frequency [s−1]

f ... force constant

µ ... reduced mass

t ... thickness [cm]

ν=c/λ

c ... speed of light [3.10 cm s ]

E=h·ν

E ... energy

h ... Planck’s constant

ν = 1/ λ

ν [cm −1 ] = 104 / λ [µm]

ν ... wavenumber [cm −1 ]

10

−1

c ... concentration [mol/L]

λ ... wavelength [µm]

portional to the square root of the force constant (strength of the chemical bond) divided by the mass of the vibrating atoms (e.g., Libowitzky and Beran 2004). Therefore, the stretching vibrations of H atoms in water and hydroxyl groups (if not or only weakly H bonded) are the highest-energetic fundamental modes in any vibrational spectrum at around 3200-3700 cm−1. However, their band positions are strongly dependent upon the effective strength of the O-H bond which is influenced by the structural environment such as hydrogen bonding and nearestneighbor cations (see below).

Sample requirements In order to obtain bulk spectra of minerals, samples are commonly prepared in the form of powder pellets pressed from a mixture of mineral powder heavily diluted in KBr (e.g., 1:200), which acts as an IR-transparent sample medium. Alternatively, the mineral powder is pressed and measured in an ATR (attenuated total reflectance) accessory. However, structural investigations of hydrogen traces in NAMs require oriented, cut and polished slabs of single crystals. These slabs must be free of any impurities such as alterations, inclusions, etc. and ideally be of gem-quality. Because these requirements are almost impossible for large natural samples and many synthetic materials, optically clean areas of available samples are selected by narrowing the beam with a mechanical aperture and/or an IR microscope (see below). Nevertheless, critical evaluation of spectra is necessary to distinguish true structural defects from any kind of extrinsic inclusions (Khisina et al. 2001), especially those that are present at a submicroscopic level.

Experimental equipment The basic equipment consists of an (FT)IR spectrometer with an energy range of 400-4000 (or even better to 8000) cm−1, a set of apertures with various sizes and an IR polarizer (see below). The given spectral range can be obtained with a single type of light source, beam splitter and detector. However, various alternatives are available which may give higher sensitivity in a more restricted region of the spectrum. Further technical details are given by e.g., Griffiths and de Haseth (1986). If small samples are investigated (or if small areas of samples have to be selected), a microfocus accessory may be necessary or, even better, an IR microscope. The latter usually facilitates also in-situ investigations under non-ambient temperatures or pressures, using a heating/cooling stage or a diamond anvil cell. Because IR microscopes with conventional black-body light sources are limited to sample areas >20 µm, synchrotron IR radiation was

Structure of Hydrous Species Using Polarized IR Spectroscopy

33

increasingly used as a light source in the last decade. The extreme brightness facilitates measurements with excellent signal-to-noise ratio down to the diffraction limits <5 µm and detailed mapping of larger sample areas. A review of IR microspectrometry using synchrotron radiation is given by Dumas and Tobin (2003).

QUANTITATIVE DATA FROM INFRARED SPECTROSCOPY The distance - frequency correlation of hydrogen bonds If the hydrogen atom of a water or hydroxyl group is attracted by another electronegative atom, usually oxygen or in some cases a halogen, a hydrogen bond is formed. In an O-H···O hydrogen bond, the first oxygen atom is called the donor and the latter the acceptor atom. Though most common in organic molecules and in hydrous compounds, hydrogen bonds are important binding forces in minerals, e.g., between the layers of chlorite-group minerals (Kleppe et al. 2003), and in some cases the only binding force that keeps molecules and coordination polyhedra in a structure together, e.g., in ice (Hall and Wood 1985) and natron (Libowitzky and Giester 2003). According to the d(O···O) hydrogen bond length, hydrogen bonds are classified as very strong (very short) with d(O···O) < 2.5 Å, strong (short) with 2.5 < d(O···O) < 2.7 Å, and weak (long) with d(O···O) > 2.7 Å (Emsley et al. 1981). The lower and upper limits are found at 2.4 Å (without external pressure) and beyond 3 Å (with a continuous transition to non-bonded entities). Due to the attractive force of the acceptor, the hydrogen atom is pulled away from the donor and the O-H bond is attenuated compared to a non-bonded unit. Whereas the O-H distance is approximately 0.98 Å in a free hydrous group, it is successively elongated up to 1.20 Å in the shortest hydrogen bonds (which are therefore symmetric without distinction between donor and acceptor atoms). Due to the variability in hydrogen bond distances (and forces) the strength of a hydrogen bond correlates closely with the frequency of its stretching vibration over a wide range of wavenumbers . The common regions are 3200-3750 cm−1 for weak H bonds (and non-bonded units), 1600-3200 cm−1 for strong H bonds, and 700-1600 cm−1 for very strong H bonds. 4000

O-H stretching wavenumber (cm-1)

The bond length vs. stretching frequency correlation of H bonds has been investigated theoretically by Bellamy and Owen (1969), and empirical correlation diagrams have been published since the 1950s, e.g., Nakamoto et al. (1955), Novak (1974), Mikenda (1986), Libowitzky (1999). The diagram for O···O bond lengths in Figure 2 shows the typical positive and curved trend line of the correlation. Scatter of data is caused by deviation of bonds from a straight O-H···O geometry and by influence of cations (see below). In general, it is observed that (very) strong H bonds tend to be more linear, whereas weak ones (such as those mostly observed in hydrous defects) are frequently bent. Further correlation diagrams that may be useful under certain circumstances

3500

3000

2500

2000

1500

1000 2.4

2.6

2.8

3.0

3.2

3.4

d(O···O) (Å)

Figure 2. Correlation between hydrogen bond length d(O···O) and O-H stretching frequency (wavenumbers) after Libowitzky (1999).

34

Libowitzky & Beran

are stretching frequency vs. d(H···O) by e.g., Libowitzky (1999), stretching vs. d(O-H) by Novak (1974), stretching vs. bending for OH groups (Novak 1974), hydrogen bond strength vs. ν1-ν3 splitting of the H2O molecule (Schiffer et al. 1976). All these empirical correlation diagrams have been obtained by comparison of spectroscopic data with structural data from X-ray and neutron diffraction of hydrous compounds. Thus, in the case of NAMs where only spectroscopic data are available on hydrous defects, they may give important structural information. However, as discussed below, distances obtained from correlation diagrams must be used with caution, because stoichiometric compounds with well-defined hydrogen sites may not be directly comparable with extremely low concentrations of hydrogen atoms located at locally distorted defect sites. Another correlation with hydrogen bond strength is observed in the band widths of O-H stretching bands. Weak H bonds at high wavenumbers in general reveal sharp bands with small full width at half maximum (FWHM), e.g., a few cm−1. With increasing H bond strength and decreasing wavenumber the band width increases up to several hundred cm−1 in case of very strong H bonds (Novak 1974). The reason for this behavior is the increasing anharmonicity of the vibration that correlates also with the increasing H bond strength (Szaly et al. 2002). These extremely broad bands centered at very low wavenumbers that resemble uneven background lines may be recognized in polarized spectra of a few stoichiometric hydrates with high water contents and very strong H bonds (e.g., Hammer et al. 1998), however they have never been observed in NAMs. It may be speculated that they do not exist in the form of defects or that they are simply invisible due to the peculiar background-like shape and the low concentration. In contrast, broad bands in the common O-H stretching region (~2500-3800 cm−1) have been observed in a number of NAMs, e.g., enstatite (Mierdel and Keppler 2004), ringwoodite (Smyth et al. 2003), wadsleyite (Jacobsen et al. 2005), and originate from strong H bonding or other phenomena such as structural disorder. In general, these broad bands and uneven background lines, aggravated by insufficient S/N ratio and small sample size, may affect accurate analysis of water contents in NAMs by IR spectroscopy (see Rossman 2006, this volume).

The spatial orientation of hydrous species Symmetry considerations. IR radiation traveling through a crystal never affects one individual O-H bond in a single unit cell, but rather many of them at the same time and phase. Thus, the vibrations are not independent and couple in-phase and out-of-phase in various combinations for all symmetry-equivalent entities. The rules for coupling according to symmetry are given by group theory in the form of normal mode analysis for molecules and by factor group analysis for crystals (e.g., Fadini and Schnepel 1989). As an example, the three atoms of a single H2O molecule in the gas phase (besides 3 translations and rotations in the three coordinates of space) possess three fundamental vibrations: a bending mode (ν2) above 1600 cm−1 and the symmetric and antisymmetric stretching modes (ν1 and ν3) with slightly different frequencies above 3600 cm−1. Thus, because of symmetry the vibrations do not occur independently along each O-H vector direction but in a coupled way along the vector sum and vector difference, i.e., parallel and perpendicular to the molecular axis. If more than one H2O molecule were contained in the primitive unit cell of a stoichiometric hydrate, further combinations of vibrations were possible. It is an advantage of low concentrations of hydrogen defects in NAMs that the vibrating species are diluted, and coupling of vibrations across many unit cells does not affect the vibrational energies. However, if symmetry-equivalent O-H bonds are grouped together in close vicinity within a unit cell, e.g., in the form of H2O molecules or clustered OH defects, splitting of bands by symmetry must be considered. Polarized radiation. In an optically anisotropic (non-cubic) crystal (IR) light is split into two rays with perpendicularly oriented polarization directions vibrating parallel to the main axes of the indicatrix section. In an absorption experiment, light that is already polarized parallel to

Structure of Hydrous Species Using Polarized IR Spectroscopy one of the indicatrix directions (X, Y, Z) is affected only by the component (Ax , Ay , Az) of an absorber which is parallel to this polarization direction, i.e., when the electric vector E of the light wave is parallel to (a component of) the oscillating dipole (Fig. 3). This component has a simple cosine squared relationship (Table 1) to the magnitude of total absorbance (Libowitzky and Rossman 1996). Thus, by measuring a crystal section in the two principal polarization directions the orientation of the absorber in this section is obtained. By measuring all three principal polarization directions of the indicatrix ellipsoid, the spatial orientation of the absorber is obtained. Moreover, only the sum of all three polarized component spectra yields the full magnitude of the absorber, i.e., the total absorbance (Libowitzky and Rossman 1996).

35

z

Az

A g

x

a b Ay

y Ax

Figure 3. Spatial orientation of an absorber A in an orthogonal optical axis system X, Y,

Polarizers for IR radiation are available acZ. Only components of absorption Ax, Ay, Az cording to two construction principles: (a) wire can be accessed during an IR absorption exgrid polarizers on an IR transparent material (or periment with polarized radiation, and facilitate calculation of the spatial orientation of even without a support), absorbing radiation parthe absorber and the total absorbance (after allel to the extremely fine, parallel (gold) wires, Libowitzky and Rossman 1996). polarize radiation over a wide angular range but their efficiency is usually limited to ~1:100. (b) Crystal polarizers, constructed similar to the well-known Nicol’s prisms, are made from IRtransparent but strongly birefringent material (e.g., LiIO3). They operate only in a narrow angular range, but their efficiencies may be as high as 1:105.

Total absorbance: a first step towards quantitative water analysis Due to the logarithmic relation between transmittance and absorbance (see above) only the total absorbance is proportional to the concentration of an absorber. Therefore unpolarized measurements and powder samples of optically anisotropic crystals are not recommended for quantitative measurements. Even the use of low-quality polarizers may bias results (Libowitzky and Rossman 1996). In cases where oriented single-crystals cannot be prepared, statistical analysis of polarized measurements on randomly oriented mineral grains in a rock section can be treated by comparing the measured spectra with polarized reference spectra of the same material (Asimov et al. 2006). In general, it must be emphasized that only integrated measurement of absorbance (Ai), i.e., the area of a band with properly treated background results in reasonable quantitative data. In that way the various band widths (FWHMs) and even overlapping peaks are reliably evaluated. Correct subtraction of the background line is of high importance. Though a linear background line can be chosen in many cases, problems may be encountered in the case of curved background shape, broad bands (see above) and very low band heights. Once the total absorbance and thickness of the sample have been measured, the concentration can be calculated according to Beer-Lambert’s law. Unfortunately, the molar absorption coefficient is not a unique constant for hydrogen in minerals. In contrast, it varies by orders of magnitude depending upon hydrogen bond strength and stretching wavenumber. Though the linear relation between ε and the wavenumber of the O-H stretching vibration can be used for a general water calibration trend (Libowitzky and Rossman 1997), mineral specific calibrations (in reference to other analytical methods) are preferred. A detailed review of this topic is given by Rossman (2006) in this volume.

36

Libowitzky & Beran CONCEPTS OF STRUCTURAL MODELS FROM INFRARED DATA

At the beginning of this paragraph it should be stressed that all concepts of structural incorporation models for traces of water in NAMs (whether in the form of H2O or OH− defects) have been developed from structural and crystal chemical observations of more or less hydrous minerals, where the information has been extracted from both diffraction and spectroscopy experiments in many cases. Therefore the examples at the end of this chapter contain also hydrous phases with stoichiometric hydrogen.

Charge balance and substitution Among the principles of crystal chemistry Pauling’s five rules (Pauling 1960) represent the most basic ideas on stable ionic compounds. Whereas the first rule comments on bond distances and coordination numbers resulting from the sums and ratios of effective ionic radii, respectively, the second rule comments on charge neutrality. The sum of charges arriving from the ligands at the center of a stable coordination polyhedron equals the (negative) charge of the central atom itself, referred to as the bond strength sum (e.g., Gibbs et al. 2003). In a more general way charge is compensated in the immediate surrounding (coordination sphere) of a charged particle. This principle is also employed in modern structural analysis to check the consistency of a crystal structure and to find hidden hydrogen atoms (missing charges!) in X-ray structural refinements (Brown 1981). This so-called bond valence analysis may even be applied to find preferred oxygen sites for trace hydroxyl substitution in a crystal structure. Thus, the most underbonded O atom in a structure may be considered an ideal docking site for H, e.g., O1 in wadsleyite (Smyth 1987). Equivalent ideas can be applied to the incorporation of a hydrogen defect in a host crystal structure. If a hydrogen atom (actually a H+ ion or “proton”) enters a crystal structure, its positive charge must be compensated. Or, in other words, if an O2− atom in a crystal structure is replaced by an OH− group, the missing negative charge must be compensated. An easy way to do so is to change the charge of a neighboring element with different valence states, e.g., Fe3+ + O2− + ½ H2 ↔ Fe2+ + OH− (e.g., Skogby and Rossman 1989; Koch-Müller et al. 2005). Another common coupled substitution in silicates may involve tetrahedral Si - Al exchange: Si4+ + O2− + ½ H2 ↔ Al3+ + OH− (e.g., Andrut et al. 2003). Whereas the former process involves only electronic charge transfer, the latter requires exchange of framework atoms and appears more likely to occur during crystal growth than by later diffusion processes. Another substitution mechanism which may easily occur during growth and which does not even require charge compensation is the incorporation of OH− groups for halogen atoms (F−, Cl−) such as in apatite (Baumer et al. 1985) and topaz (see example below). In general, unlike crystal growth the later gain and loss of hydrogen and charge compensating neighbor atoms require diffusion processes which are described in more detail by Ingrin (2006) in this volume. Probably the simplest way to compensate for the positive charge of an additional H atom is the simultaneous creation of a cation vacancy. This type of defect is known in olivine (Libowitzky and Beran 1995), perovskite (Beran et al. 1996) and others. Even in synthetic high-P phases, e.g., wadsleyite (Jacobsen et al. 2005), ringwoodite (Smyth et al. 2003), a clear correlation between H2O content (up to 1 wt %) and cation vacancies was established. In the hydrogarnet substitution, a cluster of four OH− groups is facilitated by a Si4+ vacancy at a tetrahedral site (see below). The latter has not only been observed as a trace defect but also as a major constituent of natural grossular garnets containing more than 1 wt% of H2O (Rossman and Aines 1991). Even if investigation of the correlation mechanisms of hydrogen defects with other substituents by chemical analysis may be an easy task for minerals with considerable H contents and characteristic trace element concentrations, it is almost impossible in cases of hydrogen trace defects (of the order of tens of wt. ppm) because the concentrations of accompanying minor and trace elements in the investigated minerals are commonly higher by orders of magnitude.

Structure of Hydrous Species Using Polarized IR Spectroscopy

37

Electrostatic considerations on defect geometry Whereas the electrostatic considerations above provide charge neutrality around a hydrous defect site, the charges of the surrounding atoms also constrain the orientation of an OH− or H2O group. Both hydrous species are polar with the negative end at the oxygen and the positive end at the hydrogen atom(s). Therefore, their orientation in a structure is strongly influenced by the attractive and repelling forces of surrounding ions. Because the oxygen atom of an OH− group is usually part of the crystal structure, it is connected to cations in its first coordination sphere. The same holds true for the H2O molecule in a number of stoichiometric hydrates. A characteristic coordination environment around the oxygen atom of a hydroxyl group is a flat trigonal pyramid with the cations at the corners of the base triangle and the OH group on top with the H atom pointing upwards, i.e., perpendicular to the base triangle, away from the other positive charges. This coordination type is an important part of the brucite sheet structure (e.g., Nagai et al. 2000). The orientation of the O-H vector perpendicular to the basal cation plane is caused only by three ions of equal charge, e.g., Mg2+, Mg2+, Fe2+. Substitution of cations by atoms with different valence, e.g., Al3+, Li+, or even by a vacancy (thus changing also the coordination number) results in considerable deviation of the O-H vector from the normal. A nice example for this type of coordination and the influence of (missing) cations and changed coordination is observed in the mica minerals (Fig. 4). In trioctahedral micas the octahedral layer builds up the regular brucite-type environment of the OH group and thus the O-H vector is aligned exactly perpendicular to the octahedal layer. In dioctahedral micas one third of the octahedral layer cations are missing. Because of this asymmetric distribution of only two positive charges around each OH group, the OH vector is strongly inclined towards the octahedral layer (Beran 2002). At the proton end, attractive forces of H bond acceptors may influence and distort the orientation of the hydrous defect and lead to elongated O-H bonds resulting in a decrease of O-H stretching frequencies (see above). A contrasting effect (i.e., increase of the stretching wavenumber beyond 3700 cm−1) is observed if cations opposite to the proton cause a compressed O-H bond, such as in amphiboles with occupied A site (Rowbotham and Farmer 1973).

a)

H1 O4 Mg Fe Mg

b) H1 O6 Al

Al

Figure 4. Octahedral layer with OH groups in two mica minerals. (a) Trioctahedral biotite (structural data from Brigatti and Davoli 1990): Three cations (Mg, Fe2+) around the hydroxyl group cause the O-H vector to be approximately perpendicular to the layer (brucite-type coordination). (b) Dioctahedral muscovite (structural data from Rothbauer 1971): The missing third cation (dashed square) in the coordination of the hydroxyl group forces the O-H vector into a tilted direction.

38

Libowitzky & Beran

In practice, once the orientation of an O-H vector has been obtained from polarized IR absorption measurements, the probable defect site is considered best by using a three-dimensional ball-and-stick model of the host structure. Thus, electrostatic constraints (see above) can be verified and, moreover, the necessary space (see below) to host a hydrous defect can be investigated. Two-dimensional structure drawings and even animated computer plots may be helpful but they remain always limited in information (as do abstract lists of bond lengths and angles).

Space requirements: ideal and distorted models Though hydrogen is a very small and mobile ion, the possible arrangements of O-H bonds and of O-H···O hydrogen bonds impose certain space requirements, which are not available at any position in a crystal structure and so help to constrain possible sites of hydrogen incorporation. Data for hydrogen bond lengths derived from stretching vibrations (see above) further help to develop a probable model for a suitable site of hydrogen incorporation. However, as mentioned above in the course of charge balance considerations, the incorporation of hydrous species is charge-compensated by other defects such as different cations or even vacancies in the close neighborhood. Similar to major substitution in solid solution series these structural defects cause distortions of the structure (limiting the use of H bond length calculations) which may be considered in two contrasting ways (e.g., Urusov 1992; Andrut et al. 2004). VCA model. In the virtual crystal approximation (VCA) model, no structural relaxation around a site of substitution or a defect is assumed and thus the surrounding bond distances represent an arithmetic average of the substituted and unsubstituted geometries, according to the amount of substitution (Vegard’s rule). In case of trace defects this would imply an almost unchanged defect environment where calculated H bond distances could be easily applied. Because of the “averaging” and “bulk” character of diffraction methods, their results frequently seem to support the VCA model. Hard sphere model. In the hard sphere model, full relaxation of the structure around a “hard” substituent or defect is assumed. Thus the defect environment, e.g., bond distances, develops undisturbed, as if the whole structure would consist to 100% of this substitution or defect. Because of the two different environments (with and without defect/substitution) at a single site, spectroscopic methods are superior to reveal the real situation in the structure. The real situation, expressed by the degree of relaxation (Urusov 1992), is usually found between these two extremes. The example of the hydrogarnet substitution (see below) elucidates the power of IR spectroscopy to identify the true defect environment and points out the limits of theoretical H bond calculations.

Influence on band energies from cation substitution Substitution of cations by elements with different valences and even by vacancies in the vicinity of hydrous defects has been discussed above as an important mechanism to achieve charge neutrality. Another aspect of cationic substitution is that different cations in the neighborhood and coordination sphere of, for example an OH− group may affect the energy of the O-H stretching vibration by more than 50 cm−1. In turn, shifts and splitting of absorption bands may indicate different cationic surroundings of OH defects in crystal structures. Substitution of Mg by different cations (e.g., Mn, Zn, Ni, Fe2+, Fe3+) and formation of solid solution series in common silicates with the brucite-type OH coordination such as amphiboles and talc shows interesting results. The band shift to lower wavenumbers is linearly correlated with increasing electronegativity of the substituting element, i.e., from Mg to Fe2+ and further to Fe3+ (Strens 1974). Moreover, even the number of substituents can be derived from the spectra. An intermediate Mg-Fe actinolite shows four equally spaced OH stretching bands (Fig. 5), which can be correlated according to their intensities and by comparison to pure endmember tremolite to four cationic environments around the OH group (Burns and Strens 1966): MgMgMg (~3670 cm−1), MgMgFe, MgFeFe, FeFeFe (~3625 cm−1).

Structure of Hydrous Species Using Polarized IR Spectroscopy

39

Figure 5. The four O-H stretching bands in an intermediate Mg-Fe-actinolite and their assignment to different cationic environments. Modified after Burns and Strens (1966).

Discrimination among hydrous defects Single OH− group. A single hydroxyl group, i.e., a trace defect without any symmetryequivalence in its close vicinity, is characterized by a single band whose position is mainly dependent upon the strength of hydrogen bonding (see above). Nevertheless, it is commonly observed in the 3000-3750 cm−1 region. However, because this single defect may occur with different cationic environments in different parts of the crystal (e.g., MgMgMg or MgFeMg, see above) every type of environment and even a vacant site may cause a separate band. Further vibrations which are caused by an OH− group are a Me-O-H bending mode around 700-1400 cm−1 (depending also upon H bond strength) and the combination mode of the stretching and bending vibrations around 4500 cm−1. However, if a hydroxyl defect occurs only in trace concentration, the former is usually hidden by the strong vibrations of the host mineral, and the latter may be too weak to be detected. Single H2O molecule. Because of its symmetry an undistorted water molecule possesses two stretching vibrations, one symmetric and one antisymmetric mode. As for all O-H vibrations their band positions depend upon H bonding, but are frequently observed in the 3000-3750 cm−1 region, meaning that they cannot be distinguished from bands of OH− groups. Fortunately, the water molecule also has a bending mode at approximately 1600-1650 cm−1, which is a characteristic feature of this unit. Another characteristic band may be observed at ~5200 cm−1, which is the combination of stretching and bending vibrations. However, because of its weak intensity it may be invisible at trace concentrations. H3O+ group. Hydronium (if present as the only hydrous species) is identified by four vibrations with a characteristic bending mode around 1100 cm−1 (Nakamoto 1977). All these vibrations are similar to water and hydroxyl stretching and bending modes, so that they cannot be distinguished unambiguously from a combination of different H2O or H2O + OH− species. H3O2− group. This unit, which has been considered to possess a symmetric, very short hydrogen bond in its center, has only been observed in stoichiometric phases (e.g., Beran et al. 1997). These investigations confirmed the (very) strong central hydrogen bond, but indicated also the non-symmetric configuration of the bond. Thus, the unit is considered as a linked H2O plus OH− group. The observed vibrations are the stretching modes of the terminal O-H units at high wavenumbers, the stretching mode of the central (very) strong H bond at low wavenumbers, and the bending mode of an H2O unit. Because of the very broad band shape of

40

Libowitzky & Beran

the (very) strong central H bond (see above) it is unlikely that this feature is found at minor or trace concentration levels. H2 molecule. Due to its high symmetry the H2 molecule is only Raman active with a stretching vibration at ca. 4155 cm−1. Though observed by micro-Raman spectroscopy in special fluid inclusions in melts and glasses (if present as a major constituent), it is unlikely to occur commonly as a structural defect in minerals. Nevertheless, it should be stressed that, except under very peculiar circumstances (distortion of the molecular symmetry by attraction of H2 to the host structure), IR spectroscopy is not suited to detect H2 in minerals. Clusters of hydrous defects. Clusters of vibrating O-H units (in principle, an H2O molecule is a simple cluster of two O-H vectors with a common O atom) are identified by more than one vibration according to their symmetry (see above). Unfortunately, this is not an absolute necessity, as is demonstrated in the case of (OH−)4 clusters in hydrogarnet (see below), which are characterized by a single OH stretching band (Rossman and Aines 1991). However, charge balance considerations (in this case a vacancy at the Si4+ position) indicate the necessity of more than one OH− group. Hydrous inclusions. One of the pitfalls of IR spectroscopic identification of hydrous defects is that even microscopically clear, gem-quality samples may include sub-microscopic fluid inclusions that resemble true structural defects. Fluid inclusions are readily identified by their characteristic broad water bands around 3400 cm−1 and by the appearance of sharp ice bands upon freezing. However, inclusions of hydrous minerals may be very difficult to identify. This problem has been discussed by Khisina et al. (2001). In other cases, however, IR spectroscopy may be the perfect tool to identify “invisible” mineral inclusions by their characteristic fingerprints in the OH stretching region, e.g., kaolinite in kyanite (Wieczorek et al. 2004) or corundum (Beran and Rossman 2006). During heating (either by nature or by experiment) these hydrous inclusions may act as a source of hydrogen to incorporate further structural defects.

Deuteration Though the fingerprint of O-H stretching vibrations can usually be distinguished from the fundamentals and overtones of the host mineral, problematic cases need additional treatment. Because diffusion of hydrogen at elevated temperatures is rapid, the isotope deuterium (D) can be incorporated into the material in exchange for hydrogen (Ryskin 1974). Due to the dependence of vibrational frequencies on mass (see above), corresponding O-D bands are observed at lower wavenumbers, shifted by a factor of ~1.35 (depending upon H bonding and anharmonicity, the latter also causing deviation from the ideal value √2). The reduced anharmonicity of O-D stretching vibrations in comparison to O-H modes may also result in sharper peak shapes helping to deconvolute interfering absorption bands. The problem of overlapping bands may also be solved by cooling samples to liquid nitrogen temperature in a cooling stage (see above), which may lead both to reduced FWHM and variable shifts of the peaks.

EXAMPLES Vesuvianite: orientation and hydrogen bonding of hydroxyl groups Vesuvianite, ~Ca19(Mg,Fe)3(Al,Fe)10Si18O70(OH,F)8, is an ideal first example to demonstrate the power of quantitative IR data using the wavenumber vs. hydrogen bond distance correlation and polarized spectra to constrain the O-H vector orientations. Though a chemically complicated sorosilicate (Groat et al. 1992), two different hydroxyl groups can be clearly distinguished in the structure, which have also been determined by neutron diffraction (Lager et al. 1999). The latter is the reason for the selection of a hydrous mineral as example rather than a NAM.

Structure of Hydrous Species Using Polarized IR Spectroscopy

41

Figure 6 shows polarized IR absorption spectra of a (hk0) slab of tetragonal vesuvianite with the E vector of light vibrating parallel and perpendicular to the c axis, respectively. Bands of the E//c spectrum are obviously more intense than those of the E⊥c direction. There is a strong band around 3100-3200 cm−1 and a group of strong bands between 3450 and 3700 cm−1. The former is quite broad and its low wavenumber indicates a hydrogen bond distance of d(O···O) ~ 2.70 Å (Libowitzky 1999). Its intensity perpendicular to c is zero and thus indicates an O-H orientation parallel to the c axis. The latter bands are sharper and their wavenumbers indicate only weak or no hydrogen bonding. Exact evaluation of the band areas in both polarization directions confirms an O-H vector orientation of ~35° tilted from the c axis (Bellatreccia et al. 2005). The identical pleochroic behavior of all high-energy bands indicates one hydroxyl site with different cationic environment resulting in the slightly different positions of these bands. The inset of Figure 6 shows a detail of the vesuvianite structure and confirms the band assignment from above. There is actually a moderately strong hydrogen bond at H(2) with d(O10···O10) ~ 2.72 Å (Lager et al. 1999), the O(10)-H(2) vector pointing exactly parallel to the c axis. Another hydroxyl group is located at O(11)-H(1) confirming the acute angle towards the c axis and various cations in its environment. Its weak, bifurcated H bond is in good agreement with the high wavenumbers of the IR bands. Quantitative measurements of the OH content in vesuvianites by IR spectroscopy and SIMS analyses (Bellatreccia et al. 2005) confirmed the general calibration trend of Libowitzky and Rossman (1997), although it had been demonstrated to be inaccurate for a number of NAMs.

Hydrogarnet substitution - the (OH)44− cluster Garnets contain a wide range of water concentrations, starting from a few wt. ppm in mantle garnets up to several wt% in samples of the grandite (grossular-andradite) series (Beran and Libowitzky 2006). However, their optically isotropic character, their complicated chemistry

Absorbance

1.5

1.0

Vesuvianite Rotkopf, Zillertal, Austria

Y(1)

O(10) ~ 2.7 Å

X(3)

H(2) X(3) O(10)

0.5 E || c E^c 0.0

3800

3600

3400

3200

3000

-1

Wavenumber (cm ) Figure 6. Polarized IR absorption spectra of vesuvianite (data from Kurka 2002). The strong band at ~3100 cm−1 occurs only in the c spectrum and indicates by its low wavenumber a moderately strong hydrogen bond, which is readily assigned to O(10)-H(2)···O(10) in the structure of vesuvianite (modified after Lager et al. 1999).

42

Libowitzky & Beran

due to a number of solid solution series, and a wide variety of observed OH stretching modes at rather high wavenumbers (3500-3700 cm−1) (indicating absence of hydrogen bonding) makes unambiguous identification of distinct OH− defects in silicate garnets difficult. In contrast, the high concentration of water in certain grossular garnets facilitated investigation by diffraction, analytical and spectroscopic methods, which revealed four OH− groups substituting for a SiO44− group, i.e., the (OH)44− cluster in the so-called hydrogarnet (hydrogrossular) substitution. Figure 7 shows the configuration of this cluster in comparison with a common silicate tetrahedron. With regard to the theoretical considerations on defects in crystal structures above, a number of interesting features are observed.

4-

(SiO4)

(O4H4)

4-

Figure 7. A silicate tetrahedron (left) and the hydrogarnet substitution (right) indicating the increased size of the tetrahedron, the empty Si4+ position (square) and the H atoms above the tetrahedral faces (structural data from Lager et al. 1987).

The incorporation of four protons is charge-compensated by a Si4+ vacancy. Thus, charge balance is achieved in the closest vicinity (coordination sphere) of the defect site. However, the four hydrogen atoms are not placed inside the tetrahedron (pointing towards the empty silicon site) as was proposed in an earlier paper (Sacerdoti and Passaglia 1985). Although this configuration might be considered an ideal mechanism for local charge compensation for the missing Si4+, it is not favorable due to electrostatic repulsion of the four protons in close proximity to each other. More recent papers confirm that the positions of the H atoms are, instead, rather slightly above (outside) the faces of the tetrahedron (e.g., Lager et al. 1987).

Because of the missing central charge of Si4+, the size of the hydrogarnet tetrahedron is increased by ~20 % with respect to the silicate tetrahedron (Si-O ~ 1.65 Å, -O ~ 1.95 Å) in pure endmember hydrogrossular. The corresponding IR spectrum shows a single absorption band at 3660 cm−1 (Rossman and Aines 1991). This high wavenumber is in agreement with only weak or no hydrogen bonding along the edge of the tetrahedron (O···O > 3 Å). Intermediate solid solutions that contain both silicate and hydrogrossular tetrahedra are characterized by two bands at 3600 and 3660 cm−1 with different intensities. This classical two-mode behavior was interpreted by a pure hydrogrossular environment (band at 3660 cm−1) and an (OH)44− defect surrounded by silicate tetrahedra (3600 cm−1). The latter wavenumber is still in agreement with only weak hydrogen bonding in a strongly inflated tetrahedron and thus confirms the hard sphere model. If, in contrast, the VCA model (see above) were pertinent, wavenumbers at low hydrogarnet concentrations would be expected at rather low wavenumbers due to short O···O distances in a SiO44− tetrahedron with almost unchanged size. Finally, it should be emphasized that the hydrogarnet substitution is not limited to grossular garnets, but has also been observed at low concentration levels in pyrope (Beran et al. 1993; Geiger et al. 2000) and even other minerals outside the garnet group, e.g., hydrozircon (Caruba et al. 1985). Moreover, the replacement of Si4+ by a cluster of four protons has been proposed as an important hydrogen incorporation mechanism by atomistic simulations (see also Wright 2006, this volume) for e.g., olivine (Braithwaite et al. 2003) and ringwoodite (Blanchard et al. 2005). With Ti4+ substituting for Al3+ in close vicinity to the tetrahedral (vacant) site, even an incomplete cluster of [(OH)3O]5− was suggested in pyrope by Khomenko et al. (1994). The combination of the hydrogarnet cluster with moderately strong hydrogen bonding was observed in the tetragonal garnet henritermierite (Armbruster et al 2001), where the distorted octahedron around Mn3+ provides an oxygen atom acting as H bond acceptor at rather close distance. In a

Structure of Hydrous Species Using Polarized IR Spectroscopy

43

suite of non-cubic garnets of the grossular-uvarovite join, Andrut et al. (2002) observed a number of varieties of the hydrogrossular substitution related to pleochroic IR absorption bands.

Water molecules in structural cavities: beryl and cordierite The framework silicates beryl, Be3Al2Si6O18, and cordierite, Mg2Al4Si5O18, contain structural units of 6-membered rings of tetrahedra which are stacked in such a way that channels parallel to the hexagonal c axis (in beryl) and parallel to the two-fold c axis (in orthorhombic low-cordierite) are formed. Both sets of channels are lined with oxygen atoms from the tetrahedral ligands, and with a maximum width of ~5.1 Å, separated by bottlenecks of ~2.8 Å, they can be occupied by alkalis, H2O and CO2 molecules (Kolesov and Geiger 2000a,b). Vibrational spectra contain water stretching and bending modes with wavenumbers around 3550-3700 cm−1 for stretching vibrations (Aurisicchio et al. 1994) indicative of weak or no hydrogen bonding. This is consistent with water molecules contained in the wide cavities of the channels (Fig. 8). The pleochroism of H2O stretching and bending vibrations in Raman and IR spectroscopic experiments confirms two possible orientations of the H2O molecule: Type I is oriented with the molecular axis perpendicular to the channel axis, whereas type II is oriented parallel to c. It is interesting to note that, although both stretching vibrations of H2O (ν1 symmetric, ν3 antisymmetric stretching) are IR and Raman active, due to strongly different activation cross sections, only ν3 occurs at 3700 cm−1 in IR spectra with E parallel to c for type I H2O, whereas Raman spectra yield only ν1 at 3607 cm−1 (Kolesov and Geiger (2000a). Type I H2O is predominantly observed, if alkalis are absent from the channels of the structure, whereas type II is found together with alkali ions (Aurisicchio et al. 1994). Considering the polar character of the water molecule, the structural incorporation model according to Figure 8 was developed. Moreover, as suspected from IR spectra, another possible type of incorporation in the form of an OH− group attached to a large alkali ion can be derived. Studies at variable temperature showed that type I water is dynamic and “rotates” about the channel axis down to very low temperatures (Winkler 1996). A recent study by Gatta et al. (2006) shows that this “rotation” is better described by a dynamic disorder of the water molecule over 6 equivalent positions. With increasing temperature both types of water approach the gaseous state followed by dehydration without destruction of the mineral structure (Aines and Rossman 1984).

H2O I

H2O II

OH

-

9.2 Å

Rb,Cs Na

Na

2.8 Å Alkali-free

5.1 Å Na-rich

Alkali-rich

Figure 8. Three types of hydrous species in the structural channels of beryl (and similarly of cordierite). H2O I occurs preferably in alkali-free channels, H2O II in alkali (Na)-rich channels, and OH− ions are considered in correlation with large alkalis (Rb, Cs). Modified after Aurisicchio et al. (1994).

44

Libowitzky & Beran

OH substitution in topaz Although topaz has been studied by IR and Raman spectroscopy in a number of investigations (Gebert and Zemann 1965; Aines and Rossman 1985; Beny and Piriou 1987; Wunder et al. 1999; Bradbury and Williams 2003), all these papers lack one or another aspect of spectral evaluation. Therefore, we decided to demonstrate a worked example on a new topaz data set that has not been published previously. Moreover, we chose topaz because it contains a single OH group (with a single IR absorption band) substituting for the fluorine atom in its structure and because its orthorhombic symmetry is ideal for geometric considerations. A clear, colorless, gem-quality topaz crystal from Spitzkoppe, Namibia with a size of 9 × 11 × 17 mm was chosen for this study. The F-rich composition of this specimen, Al2SiO4F1.85(OH)0.15, (OH)/(OH + F) = 0.075, was confirmed by the correlation of Ribbe and Rosenberg (1971) using the lattice parameters from X-ray powder diffraction (space group Pbnm, Z = 4, a = 4.652 Å, b = 8.804 Å, c = 8.390 Å, CuKα, 5% Si standard). The sample was oriented according to the excellent cleavage parallel to (001) and along the optical extinction directions (optical setting: a = X, b = Y, c = Z). Three platelets (100), (010), (001) were cut from the sample, attached with crystal bond epoxy resin to a glass plate sample holder and diamond-polished to a final thickness of ~15 µm (uncertainty due to resin layer). To retain the large size of the extremely thin sections they were not removed from the glass plate, and spectra were corrected for background absorption from glass and epoxy resin. Figure 9 shows the polarized spectra of topaz in the O-H stretching region parallel to the three main axis directions with a strong absorption band at 3649 cm−1 and Figure 10 gives the angular absorption plots of the integrated absorbance of the OH stretching band in all three crystal sections. Both figures confirm that the absorption is almost zero parallel to the b axis (Y) direction. Thus the O-H dipole must be aligned within the (010) plane. Further inspection of the (010) angular absorbance plot (Fig. 10) and comparison of the X and Z directions (Fig. 9) indicate a preferred O-H orientation along the c axis (Z) direction. The ratio of the Z:X integrated band intensities is ~2:1, and application of the cosine squared relation (Table 1) results in an angle of ~35° between the O-H dipole and the c axis direction. This result is in excellent agreement with diffraction data (Zemann et al. 1979; Parise et al. 1980; Belokoneva et al. 1993) that yield ~29°, and also with crystal chemical considerations. Figure 11 shows the environment of the OH group in the structure of topaz. The repelling forces of the two Al atoms coordinating the OH group are such that the H atom is aligned almost exactly within the Al-O-Al plane and bisects the Al-O-Al angle. The weak H bonds around the H atom are in agreement with the high wavenumber position of the OH absorption band and have only a very minor influence on the alignment of the O-H vector. Finally it should be emphasized that the OH stretching band of F-rich topaz is not a single band but rather contains another component at ~3640-3646 cm−1 (depending upon peak fit constraints). This feature has been frequently ignored in older literature on topaz with low H content, but it was definitely described and discussed in recent papers on synthetic OH-rich topaz, e.g., Wunder et al. (1999). These details of peak fitting will be presented and discussed in a separate paper (Libowitzky, in prep.).

OH incorporation in diopside Pyroxenes contain significant amounts of hydrogen with concentrations ranging from a few 10s to more than 1000 wt. ppm H2O. Thus, pyroxene may indeed be a major storage site for hydrogen in the Earth’s upper mantle (Skogby 2006, this volume, and references therein). IR spectra of clinopyroxenes (cpx), i.e., diopside-hedenbergite, augite, and omphacite, are characterized by four regions of pleochroic OH stretching bands centered at 3630-3640, 35303540, 3450-3470 and 3350-3360 cm−1 (the latter only in a number of diopsides). Two different

Structure of Hydrous Species Using Polarized IR Spectroscopy

1.2

45

Topaz

Absorbance

1.0

Figure 9 (left). Polarized IR absorption spectra of colorless topaz from Spitzkoppe, Namibia, in the O-H stretching region parallel to the three principal axis directions. Sample thickness: ~15 µm, Perkin Elmer 1760X FTIR spectrometer (ceramic light source, KBr beam splitter, TGS detector), gold wire grid polarizer (efficiency ~ 1:100), circular sample aperture: 4 mm diameter, spectral resolution: 4 cm−1, 32 scans each averaged.

0.8 0.6 0.4 Z

0.2

X Y

0.0 3750

3700

3650

3600

3550

Wavenumber (cm-1)

20

15

10

Topaz (100)

Topaz (010)

Topaz (001)

Z

Z

X

5

Y

0 0

5

10

15

15

10

5

20

0

2.38

0.98

Al

H1

O1

5

10

15

8

6

4

2

Y

0 0

2

4

6

8

Figure 10 (above). Absorbance figures depicting the pleochroic scheme of the O-H stretching band of topaz (integrated absorbance vs. sample to polarizer angle). Lack of absorbance parallel to the b axis (Y) indicates an O-H vector orientation in the (010) plane. The anisotropic absorbance in (010) indicates an O-H orientation closer to the c axis (Z direction).

2.23

2.40

X

0

2.29

Al

c Si

a b

Figure 11 (left). The environment of the hydroxyl group in the structure of topaz. The view was chosen in such a way that the OH group in the center of the picture and the two coordinating Al atoms (broken circles, connected by bold broken lines) are in the plane of projection. Broken lines indicate weak H···O bonds, the numbers give distances in Å. Structural data from Zemann et al. (1979).

46

Libowitzky & Beran

types of pleochroic behavior can be distinguished. Bands in the 3630-3640 cm−1 region are α- and β-polarized (group I bands), the lower energetic bands are γ-polarized (group II bands). Compare to Figure 12 in this chapter, and Table 1 in Skogby (2006), this volume. The two types of pleochroic bands suggest that at least two types of OH positions exist simultaneously in the diopside structure (Beran 1976; Ingrin et al. 1989; Skogby and Rossman 1989; Skogby et al. 1990). The position and pleochroism of the absorption bands are similar for different cpx samples, but the absolute intensities vary strongly, e.g., spectra of omphacites with jadeite-rich compositions show a strong γ-polarized absorption band at 3460-3470 cm−1, whereas those with diopside-rich compositions reveal a strong α-polarized band at 3620 cm−1 (Smyth et al. 1991). As an example for OH defect characterization by polarized IR spectroscopy, the study of a hydrothermally formed, gem-quality diopside crystal from Rotkopf, Tyrol, Austria, is given below (Andrut et al. 2003). The extremely strong pleochroism of the high-energy group I band in (010) sections at 3647 cm−1 (Fig. 12) suggests a strong preferred orientation of the OH dipole approximately parallel to the α index of refraction, i.e., the direction of the long diagonal of the unit cell projection parallel to [010] (Fig. 13). The moderate pleochroism of this band in (100) with a stronger component parallel to [010] (equivalent to nβ) indicates a strong deviation of the OH vector direction from an alignment within the (010) plane. These results confirm the model proposed by Beran (1976) that OH defects partially replace the O2 “zigzag” oxygen atoms pointing to the O3 oxygen atom of a neighboring silicate chain (Fig. 13). O2 is coordinated by 1 Mg, 1 Ca and 1 Si, thus forming the top of a flat slightly distorted trigonal pyramid, being an ideal candidate for a partial OH replacement. This replacement mode also occurs in a 1100 °C temperaturetreated crystal and evidently represents a very stable OH defect position. Another model of OH defect incorporation on O2 sites (with similar O-H vector orientation) can also be derived under the assumption of a vacant M1 site, resulting in a coordination of the OH defect by Ca and Si. Owing to the pleochroism of the low-energy band doublet at 3464 and 3359 cm−1 in (010) (Fig. 12), the OH dipole direction must be oriented roughly parallel to the γ index of refraction, i.e., the direction of the short diagonal in the (010) section of the unit cell. In addition, a slight deviation from the (010) plane is indicated. An OH dipole direction that is in agreement with the observed pleochroic behavior can be provided under the assumption of M2 vacancies. OH defects coordinated by 1 Mg and 1 Si are generated by a partial replacement of O2 oxygen atoms with an orientation pointing strongly above the Ca vacancy site. The separation of the low-energy bands is explained by a replacement of the coordinating Mg by Fe or Si by Al.

-1

Linear absorption coefficient (cm )

3.0 2.5

group I group II

2.0 1.5 1.0

Figure 12. Polarized OH absorption spectra of lightgreen diopside from Rotkopf, Zillertal, Tyrol, Austria, measured on (100) and (010) plates (modified after Andrut et al. 2003).

a in (010)

b in (100)

0.5 g in (010)

0.0

3800

3700

3600

3500

3400 -1

Wavenumber (cm )

3300

Structure of Hydrous Species Using Polarized IR Spectroscopy OH defects in perovskite

The IR spectrum of OH containing CaTiO3 perovskite consists of two bands with maxima centered at 3394 and 3326 cm−1 (Fig. 14). From the weak pleochroism of the bands in (001) and the more distinct pleochroism in (110), with a stronger component of absorption perpendicular to [001], an OH direction roughly pointing along [110] with the O2 oxygen atoms acting as donor is deduced (inset of Fig. 14). Using the hydrogen bond length vs. stretching frequency correlation of Libowitzky (1999), excellent agreement between the expected (calculated) H bond lengths and the actual O···O distances in the structure (2.75-2.78 Å) is obtained. The O-H vector orientation is facilitated only by the presence of a vacant Ca site (Beran et al. 1996, inset of Fig. 14). The OH defect positions coordinated by two Ti and one Ca atoms correspond to those proposed by Meade et al. (1994) in synthetic high-pressure MgSiO3 perovskite, where OH bands occur at 3483 and 3423 cm−1. Similar to CaTiO3 perovskite, the assumption of a vacant Mg position seems necessary from geometric and electrostatic considerations.

c

g a O3

Ca/Mg

Si

O1

O2 Ca/Mg O3 O3 Si

O1

O1 O3

O3 a

Ca/Mg

O2

O2

Ca/Mg

a/2

Figure 13. Part of the diopside structure projected parallel to [010] with OH defects at O2 pointing to O3 of a neighboring silicate chain (modified after Beran 1999). Dark grey atoms, labels and tetrahedron belong to the silicate chain behind the light grey units. 0.10 b

Ca Ti

a

0.08

Absorbance

(Mg,Fe)SiO3 perovskite is most likely the major mineral phase in the Earth’s lower mantle and its role as a storage site for hydrogen has been recently discussed in review articles by Bolfan-Casanova (2005) and Ohtani (2005). On the other hand, though being a rare mineral, natural CaTiO3 perovskite forms in various geological environments, including kimberlitic rocks, and shows a wide range of compositions (Hu et al. 1992). Based on Paterson’s (1982) calibration, Beran et al. (1996) reported about 70 wt ppm H2O in perovskite of metasomatic origin.

47

[Ca]

O2

0.06

0.04

0.02 3600

3400

3200

3000

2800

-1

Wavenumber (cm ) Figure 14. IR absorption spectra of perovskite. The inset shows the structure of perovskite. The two arrows, deviating from the plane of projection, indicate possible O-H vectors close to a vacant Ca site (modified after Beran et al. 1996).

In contrast, high-P,T experiments performed by Bolfan-Casanova et al. (2000) in the MgO-SiO2-H2O system did not detect any OH in MgSiO3 perovskite. Whereas MgSiO3 akimotoite, coexisting with MgSiO3 perovskite, synthesized at 24 GPa and 1600°C dissolved significant amounts of water (see below), perovskite was essentially dry.

48

Libowitzky & Beran

The conflicting experimental findings of Bolfan-Casanova et al. (2000) and Meade et al. (1994) might be related to the different synthesis conditions of the two studies. In the former study the sample was held at P and T for 3.5 hours, in the latter study for only several minutes. Therefore, any defects in MgSiO3 perovskite may have been annealed out in the Bolfan-Casanova et al. (2000) study whereas they were retained in the latter. Therefore, the presence of vacancies at the Mg site (see above) may be a key factor whether MgSiO3 perovskite incorporates OH groups in its crystal structure or not (Ross et al. 2003). Moreover, under natural conditions, the lower mantle contains elements, such as Fe and Al, which have not been involved in the synthesis and which may facilitate OH incorporation in perovskite (Bolfan-Casanova 2005).

OH traces in corundum The presence of OH in corundum as accessory mineral of mantle rocks is rather speculative. IR spectra of a 320 µm thick corundum from a South African eclogite assemblage showed no indication for the presence of OH (Rossman and Smyth 1990). Natural ruby and sapphire samples from crustal origin showed extremely weak absorption bands at 3310, 3230, and 3185 cm−1 (Beran 1991). Smith et al. (1995) confirmed OH bands also in sapphires from Southern Vietnam. In a comprehensive IR spectroscopic study of about 150 corundum samples from worldwide localities, Beran and Rossman (2006) established the presence of OH defects in corundum of crustal occurrences, however at concentration levels around only 0.5 wt ppm H2O or even lower. On the other hand, knowledge of possible OH defect incorporation mechanisms in this hexagonally close-packed mineral structure has a definite geophysical interest due to its close relation to MgSiO3 akimotoite in the high-PT regions of the Earth’s interior. Therefore, a number of synthetic samples have been studied in the past. OH groups in hydrothermally grown corundum were originally recognized by Belt (1967). A polarized IR spectroscopic study of a suite of Verneuil-grown corundum crystals (Beran 1991) revealed that variously colored samples show a distinct variability in the region of the OH fundamental vibration. Narrow strongly polarized OH bands with varying intensities are centered at 3310, 3230, and 3185 cm−1 (see above). Additional weak bands at 3290 cm−1 occur in (V Cr Fe Ti)-doped “alexandrite” sapphires, weak bands at 3280 and 3160 cm−1 appear in colorless corundum. Due to the strong polarization with maximum absorption perpendicular to the c axis and the deduced orientation of the OH dipoles perpendicular to c, Beran (1991) proposed a model where, under the assumption of vacant Al sites, OH defects are coordinated by two Al atoms, forming groups of face-sharing [Al2(OH)O8] double octahedra. According to Moon and Phillips (1991) the OH defects appear to be correlated to vacant Al sites as well as to the presence of Ti4+. In addition, two types of OH absorption bands were reported for hydrothermally treated synthetic sapphires by Kronenberg et al. (2000). The first type of bands observed at 3308, 3293, 3278, 3231, 3208, 3183, and 3163 cm−1 is characterized by narrow bands and strong pleochroism, the second type consists of a broad isotropic band centered at 3400 cm−1, resembling closely the OH bands of hydrothermally grown quartz crystals. Polarized IR spectra of synthetic high-P MgSiO3 akimotoite (Bolfan-Casanova et al. 2002), consist of five pleochroic OH absorption bands—three sharp strong bands at 3390, 3320, and 3300 cm−1 and two weak bands at 3260 and 3050 cm−1. Based on Paterson’s (1982) calibration the H2O content was calculated to 350 wt ppm. The bands at 3320 and 3300 cm−1 are strongly polarized perpendicular to the c axis. Similar to corundum, under the assumption of Mg vacancies, the pleochroic behavior is consistent with OH groups oriented nearly parallel to the plane of the shared face between two SiO6 octahedra. The two OH bands of corundum (at 3310 and 3230 cm−1) with the same pleochroic behavior occur at lower frequencies compared to those of MgSiO3 akimotoite. The band at 3390 cm−1 has maximum intensity

Structure of Hydrous Species Using Polarized IR Spectroscopy

49

parallel to the c axis and is therefore consistent with OH groups pointing into a tetrahedral void of the close-packed oxygen sublattice. This band, however, has no analogue in the IR spectrum of corundum.

ACKNOWLEDGMENTS The authors wish to thank H. Keppler and J. Smyth for the invitation to contribute to the present MSA volume. S.D. Jacobsen, H. Keppler, and an anonymous referee helped to improve the quality of the manuscript. The topics of this paper were partly sponsored by the European Commission, Human Potential-Research Training Network, HPRN-CT-2000-0056.

REFERENCES Aines RD, Rossman GR (1984) The high temperature behavior of water and carbon dioxide in cordierite and beryl. Am Mineral 69:319-327 Aines RD, Rossman GR (1985) The high temperature behavior of trace hydrous components in silicate minerals. Am Mineral 70:1169-1179 Andrut M, Wildner M, Beran A (2002) The crystal chemistry of birefringent natural uvarovites. Part IV. OH defect incorporation mechanisms in non-cubic garnets derived from polarized IR spectroscopy. Eur J Mineral 14:1019-1026 Andrut M, Brandstätter F, Beran A (2003) Trage hydrogen zoning in diopside. Mineral Petrol 78:231-241 Andrut M, Wildner M, Rudowicz CZ (2004) Optical absorption spectroscopy in geosciences. Part II: Quantitative aspects of crystal fields. In: Spectroscopic Methods in Mineralogy. Beran A, Libowitzky E (eds) EMU Notes in Mineralogy, vol 6. Eötvös University Press, p 145-188 Asimov PD, Stein LC, Mosenfelder JL, Rossman GR (2006) Quantitative polarized infrared analysis of trace OH in populations of randomly oriented mineral grains. Am Mineral 91:278-284 Armbruster T, Kohler T, Libowitzky E, Friedrich A, Miletich R, Kunz M, Medenbach O, Gutzmer J (2001) Structure, compressibility, hydrogen bonding, and dehydration of the tetragonal Mn3+ hydrogarnet, henritermierite. Am Mineral 86:147-158 Aurisicchio C, Grubessi O, Zecchini P (1994) Infrared spectroscopy and crystal chemistry of the beryl group. Can Mineral 32:55-68 Baumer A, Ganteaume M, Klee WE (1985) Determination of OH ions in hydroxyfluorapatites by infrared spectroscopy. Bull Minéral 108:145-152 Bellamy LJ, Owen AJ (1969) A simple relationship between the infrared stretching frequencies and the hydrogen bond distances in crystals. Spectrochim Acta A25:329-333 Bellatreccia F, Della Ventura G, Ottolini L, Libowitzky E, Beran A (2005) The quantitative analysis of OH in vesuvianite: a polarized FTIR and SIMS study. Phys Chem Minerals 32:65-76 Belokoneva EL, Smirnitskaya YuYa, Tsirel’son VG (1993) Electron density distribution in topaz Al2(SiO4)(F,OH)2 as a result of accurate X-ray diffraction study. Zhurnal Neorganicheskoi Khimii 38:1346-1350 Belt FR (1967) Hydrothermal ruby: Infrared spectra and X-ray topography. J Appl Phys 38:2688-2689 Beny JM, Piriou B (1987) Vibrational spectra of single-crystal topaz. Phys Chem Minerals 15:148-154 Beran A (1976) Messung des Ultrarot-Pleochroismus von Mineralen. XIV. Der Pleochroismus der OHStreckfrequenz in Diopsid. Tschermaks Min Petr Mitt 23:79-85 Beran A (1991) Trace hydrogen in Verneuil-grown corundum and its colour varieties - an IR spectroscopic study. Eur J Mineral 3:971-975 Beran A (1999) Contribution of IR spectroscopy to the problem of water in the Earth’s mantle. In: Microscopic Properties and Processes in Minerals. Wright K, Catlow R (eds) NATO Science Series. Kluwer Acad. Publishers, p 523-538 Beran A (2002) Infrared spectroscopy of micas. Rev Mineral Geochem 46:351-369 Beran A, Libowitzky E (2006) Water in natural mantle minerals II: olivine, garnet and accessory minerals. Rev Mineral Geochem 62:169-191 Beran A, Rossman GR (2006) OH in naturally occurring corundum. Eur J Mineral 18, in press Beran A, Langer K, Andrut M (1993) Single crystal infrared spectra in the range of OH fundamentals of paragenetic garnet, omphacite and kyanite in an eklogitic mantle xenolith. Mineral Petrol 48:257-268 Beran A, Libowitzky E, Armbruster T (1996) A single-crystal infrared spectroscopic and X-ray diffraction study of untwinned San Benito perovskite containing OH groups. Can Mineral 34:803-809 Beran A, Giester G, Libowitzky E (1997) The hydrogen bond system in natrochalcite-type compounds - An FTIR spectroscopic study of the H3O2− unit. Mineral Petrol 61:223-235

50

Libowitzky & Beran

Bolfan-Casanova N (2005) Water in the Earth’s mantle. Mineral Mag 69:229-257 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett 182:209-221 Bolfan-Casanova N, Keppler H, Rubie DC (2002) Hydroxyl in MgSiO3 akimotoite: A polarized and highpressure IR study. Am Mineral 87:603-608 Blanchard M, Wright K, Gale JD (2005) A computer simulation study of OH defects in Mg2SiO4 and Mg2GeO4 spinels. Phys Chem Minerals 32:585-593 Bradbury SE, Williams Q (2003) Contrasting bonding behavior of two hydroxyl-bearing metamorphic minerals under pressure: Clinozoisite and topaz. Am Mineral 88:1460-1470 Braithwaite JS, Wright K, Catlow CRA (2003) A theoretical study of the energetics and IR frequencies of hydroxyl defects in forsterite. J Geophys Res 108(B6):2284 Brigatti MF, Davoli P (1990) Crystal structure refinement of 1M plutonic biotites. Am Mineral 75:305-313 Brown ID (1981) The bond-valence method: An empirical approach to chemical structure and bonding. In: Structure and Bonding in Crystals, vol II. O’Keeffe M, Navrotsky A (eds), Academic Press, p 1-30 Burns RG, Strens RGJ (1966) Infrared study of the hydroxyl bands in clinoamphiboles. Science 153:890-892 Buse K, Breer S, Peithmann K, Kapphan S, Gao M, Krätzig E (1997) Origin of thermal fixing in photorefractive lithium niobate crystals. Phys Rev B56:1225-1235 Caruba R, Baumer A, Ganteaume M, Iacconi P (1985) An experimental study of hydroxyl groups and water in synthetic and natural zircons: a model of the metamict state. Am Mineral 70:1224-1231 Dixon JE, Leist L, Langmuir C, Schilling J-G (2002) Recycled dehydrated lithosphere observed in plumeinfluenced mid-ocean-ridge basalt. Nature 420:385-389 Dobson DP, Meredith PG, Boon SA (2002) Simulation of subduction zone seismicity by dehydration of serpentine. Science 298:1407-1410 Dumas P, Tobin MJ (2003) A bright source for infrared microspectroscopy: synchrotron radiation. Spectr Eur 15/6:17-23 Emsley J, Jones D, Lucas J (1981) Detecting and measuring strong hydrogen bonds: recent developments. Rev Inorg Chem 3:105-140 Fadini A, Schnepel F-M (1989) Vibrational spectroscopy. Ellis Horwood Limited Gatta GD, Nestola F, Bromiley GD, Mattauch S (2006) The real topological configuration of the extraframework content in alkali-poor beryl: A multi-methodological study. Am Mineral 91:29-34 Geiger, CA (2004) An introduction to spectroscopic methods in the mineral sciences and geochemistry. In: Spectroscopic Methods in Mineralogy. Beran A, Libowitzky E (eds) EMU Notes in Mineralogy, vol 6. Eötvös University Press, p 1-42 Geiger CA, Stahl A, Rossman GR (2000) Single-crystal IR- and UV/VIS- spectroscopic measurements on transition-metal-bearing pyrope: the incorporation of hydroxide in garnet. Eur J Mineral 12:259-271 Gibbs GV, Rosso KM, Cox DF, Boisen Jr MB (2003) A physical basis for Pauling’s definition of bond strength. Phys Chem Minerals 30:317-320 Griffiths PR, de Haseth JA (1986) Fourier Transform Infrared Spectrometry (Chem Anal 83). John Wiley & Sons Griggs D (1967) Hydrolytic weakening of quartz and other silicates. Geophys J R Astr Soc, 14:19-31 Groat LA, Hawthorne FC, Ercit TS (1992) The chemistry of vesuvianite. Can Mineral 30:19-48 Gebert W, Zemann J (1965) Messung des Ultrarot-Pleochroismus von Mineralen. III. Der Pleochroismus der OH-Streckfrequenz in Topas. N Jb Miner Mh 1965:380-384 Hall D, Wood MK (1985): A molecular-packing analysis of the crystal structures of ice. Acta Cryst B41:169172 Hammer VMF, Libowitzky E, Rossman GR (1998) Single-crystal IR spectroscopy of very strong hydrogen bonds in pectolite, NaCa2[Si3O8(OH)], and serandite, NaMn2[Si3O8(OH)]. Am Mineral 83:569-576 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: Implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Hu Meisheng, Wenk H-R, Sinitsyna D (1992) Microstructures in natural perovskites. Am Mineral 77:359-373 Ingrin J, Blanchard M (2006) Diffusion of hydrogen in minerals. Rev Mineral Geochem 62:291-320 Ingrin J, Latrous K, Doukhan J-C, Doukhan N (1989) Water in diopside: an electron microscopy and infrared spectroscopy study. Eur J Mineral 1:327-431 Jacobsen SD (2006) Effect of water on the equation of state of nominally anhydrous minerals. Rev Mineral Geochem 62:321-342 Jacobsen SD, van der Lee S (Eds) (2006) Earth’s Deep Water Cycle. Am Geophys Union Monogr Ser, in press Jacobsen SD, Demouchy S, Frost DJ, Boffa Ballaran T, Kung J (2005) A systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single-crystal X-ray diffraction: Oxygen sites for hydrogen storage in Earth’s interior. Am Mineral 90:61-70

Structure of Hydrous Species Using Polarized IR Spectroscopy

51

Jung H, Green HW (2004) Experimental faulting of serpentinite during dehydration: Implications for earthquakes, seismic low-velocity zones, and anomalous hypocenter distributions in subduction zones. Int Geol Rev 46:1089-1102 Karato S (1990) The role of hydrogen in the electrical conductivity of the upper mantle. Nature 347:272-273 Khisina NR, Wirth R, Andrut M, Ukhanov AV (2001) Extrinsic and intrinsic mode of hydrogen occurrence in natural olivines: FTIR and TEM investigation. Phys Chem Minerals 28:291-301 Khomenko VM, Langer K, Beran A, Koch-Müller M, Fehr T (1994): Titanium substitution and OH-bearing defects in hydrothermally grown pyrope crystals. Phys Chem Minerals 20:483-488 Kleppe AK, Jephcoat AP, Welch MD (2003) The effect of pressure upon hydrogen bonding in chlorite: A Raman spectroscopic study of clinochlore to 26.5 GPa. Am Mineral 88:567-573 Koch-Müller M, Abs-Wurmbach I, Rhede D, Kahlenberg V, Matsyuk, S (2005) Hydration and dehydration of omphacite. Berichte der Deutschen Mineralogischen Ges, Beih z Eur J Mineral 17, No 1:69 Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kolesov AB, Geiger CA (2000a) A single-crystal Raman study of the orientation and vibrational states of molecular water in beryl. Phys Chem Minerals 27:557-564 Kolesov AB, Geiger CA (2000b) Cordierite II: The role of CO2 and H2O. Am Mineral 85:1265-1274 Kronenberg AK, Castaing J, Mitchell TE, Kirby SH (2000) Hydrogen in α-Al2O3 and water weakening of sapphire and alumina ceramics between 600 and 1000 °C – I. Infrared characterization of defects. Acta Mater 48:1481-1494 Kurka A (2002) IR-spektroskopische Untersuchungen im OH-Streckschwingungsbereich von Vesuvian und Gossular. MSc thesis, University of Vienna, Austria Lager GA, Armbruster T, Faber F (1987) Neutron and X-ray diffraction study of hydrogarnet, Ca3Al2(O4H4)3. Am Mineral 72:756-765 Lager GA, Qianyen X, Ross FK, Rossman GR, Armbruster T, Rotella FJ, Schultz AJ (1999) Hydrogen-atom positions in P4/nnc vesuvianite. Can Mineral 37:763-768 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H···O hydrogen bond lengths in minerals. Mh Chemie 130:1047-1059 Libowitzky E, Beran A (1995) OH-defects in forsterite. Phys Chem Minerals 22:387-392 Libowitzky E, Beran A (2004) IR spectroscopic characterisation of hydrous species in minerals. In: Spectroscopic methods in mineralogy. Beran A, Libowitzky E (eds) EMU Notes in Mineralogy, Vol 6. Eötvös University Press, p 227-279 Libowitzky E, Giester G (2003) Washing soda (natron), Na2CO3 · 10H2O, revised: Crystal structures at low and ambient temperatures. Mineral Petrol 77:177-195 Libowitzky E, Rossman GR (1996) Principles of quantitative absorbance measurements in anisotropic crystals. Phys Chem Minerals 23:319-327 Libowitzky E, Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82:11111115 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90B:11319-11333 Meade C, Reffner JA, Ito E (1994) Synchrotron infrared absorbance measurements of hydrogen in MgSiO3 perovskite. Science 264:1558-1560 van der Meijde M, Marone F, Giardini D, van der Lee S (2003) Seismic evidence for water deep in Earth’s upper mantle. Science 300:1556-1558 Mierdel K, Keppler H (2004) The temperature dependence of water solubility in enstatite. Contr Mineral Petrol 148:305-311 Mikenda W (1986) Stretching frequency versus bond distance correlation of O-D(H)···Y (Y = N, O, S, Se, Cl, Br, I) hydrogen bonds in solid hydrates. J Mol Struct 147:1-15 Moon AR, Phillips MR (1991) Defect clustering in H,Ti:α-Al2O3. J Phys Chem Solids 52:1087-1099 Nagai T, Hattori T, Yamanaka T (2000) Compression mechanism of brucite: An investigation by structural refinement under pressure. Am Mineral 85:760-764 Nakamoto K (1977) Infrared and Raman spectra of inorganic and coordination compounds. Wiley, New York Nakamoto K, Margoshes M, Rundle RE (1955) Stretching frequencies as a function of distances in hydrogen bonds. J Am Chem Soc 77:6480-6486 Novak A (1974) Hydrogen bonding in solids: Correlation of spectroscopic and crystallographic data. Struct Bond 18:177-216 Ohtani E (2005) Water in the mantle. Elements 1:25-30 Parise JB, Cuff C, Moore FH (1980) A neutron diffraction study of topaz: evidence for lower symmetry. Min Mag 43:943-944 Paterson MS (1982) The determination of hydroxyl by infrared absorption in quartz, silicate glasses and similar materials. Bull Minéral 105:20-29

52

Libowitzky & Beran

Pauling L (1960) The Nature of the Chemical Bond. Cornell Univ. Press Ribbe PH, Rosenberg PE (1971) Optical and X-ray determinative methods for fluorine in topaz. Am Mineral 56:1812-1821 Ross NL, Gibbs GV, Rosso KM (2003) Potential docking sites and positions of hydrogen in high-pressure silicates. Am Mineral 88:1452-1459 Rossman GR (1988) Vibrational spectroscopy of hydrous components. Rev Mineral 18:193-206 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Rossman GR, Aines RD (1991) The hydrous components in garnets: Grossular-hydrogrossular. Am Mineral 76: 1153-1164 Rossman GR, Smyth JR (1990) Hydroxyl contents of accessory minerals in mantle eclogites and related rocks. Am Mineral 75: 775-780 Rothbauer R (1971) Untersuchung eines 2M1-Muskovits mit Neutronenstrahlen. N Jb Mineral Mh 1971:143154 Rowbotham G, Farmer VC (1973) The effect of “A” site occupancy upon hydroxyl stretching frequency in clinoamphiboles. Contrib Mineral Petrol 38:147-149 Ryskin YI (1974) The vibrations of protons in minerals: hydroxyl, water and ammonium. In: The Infrared Spectra of Minerals. Farmer VC (ed) Mineralogical Society, p 137 -181 Sacerdoti M, Passaglia E (1985) The crystal structure of katoite and implications within the hydrogrossular group of minerals. Bull Minéral 108:1-8 Schiffer J, Intenzo M, Hayward P, Calabrese C (1976) The FS correlation and a systematic analysis of the stretching absorptions of water in the condensed state. J Chem Phys 64:3014-3020 Serratosa JM, Bradley WF (1958) Determination of the orientation of OH bond axes in layer silicates by infrared absorption. J Phys Chem 62:1164-1167 Skogby H (2006) Water in natural mantle minerals I: pyroxenes. Rev Mineral Geochem 62:155-167 Skogby H, Rossman GR (1989) OH− in pyroxene: An experimental study of incorporation mechanisms and stability. Am Mineral 74:1059-1069 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: Variations in the natural environment. Am Mineral 75:764-774 Smith CP, Kammerling RC, Keller AS, Peretti A, Scarratt KV, Khoa ND, Repetto S (1995) Sapphires from southern Vietnam. Gems Gemology 31:168-186 Smyth JR (1987) β-Mg2SiO4: A potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR, Frost DJ (2002) The effect of water on the 410-km discontinuity: An experimental study. Geophys Res Lett 29:10.1029/2001GL014418 Smyth JR, Bell DR, Rossman GR (1991) Incorporation of hydroxyl in upper-mantle clinopyroxenes. Nature 351:732-735 Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Langenhorst F, McCammon CA (2003) Structural systematics of hydrous ringwoodite and water in Earth’s interior. Am Mineral 88:1402-1407 Strens RGJ (1974) The common chain, ribbon, and ring silicates. In: The Infrared Spectra of Minerals. Farmer VC (ed) Mineralogical Society, p 305-330 Szalay V, Kovács L, Wöhlecke M, Libowitzky E (2002) Stretching potential and equilibrium length of the OH bond in solids. Chem Phys Lett 354:56-61 Thompson AB (1992) Water in the Earth’s upper mantle. Nature 358:295-302 Tillmanns E, Zemann J (1965) Messung des Ultrarot-Pleochroismus von Mineralen. I. Der Pleochroismus der OH-Streckfrequenz in Azurit. N Jb Miner Mh 1965:228-231 Tsuboi M (1950) On the position of the hydrogen atoms in the crystal structure of muscovite, as revealed by the infra-red absorption study. Bull Chem Soc Japan 23:83-88 Urusov VS (1992) A geometric model of deviations from Vegard’s rule. J Solid State Chem 98:223-236 Wieczorek A, Libowitzky E, Beran A (2004) A model for the OH defect incorporation in kyanite based on polarised IR spectroscopic investigations. Schweiz Min Petr Mitt 84:333-343 Wilkins RWT, Sabine W (1973) Water content of some nominally anhydrous silicates. Am Mineral 58:508-516 Winkler B (1996) The dynamics of H2O in minerals. Phys Chem Minerals 23:310-318 Wright K (2006) Atomistic models of OH defects in nominally anhydrous minerals. Rev Mineral Geochem 62:67-83 Wunder B, Andrut M, Wirth R (1999) High-pressure synthesis and properties of OH-rich topaz. Eur J Mineral 11:803-813 Zemann J, Zobetz, E, Heger G, Völlenkle H (1979) Strukturbestimmung eines OH-reichen Topases. Anz Österr Akad Wiss, Math-Naturwiss Kl 116:145-147

13

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 291-320, 2006 Copyright © Mineralogical Society of America

Diffusion of Hydrogen in Minerals Jannick Ingrin and Marc Blanchard Laboratoire des Mécanismes et Transferts en Géologie CNRS, Université Paul-Sabatier, Observatoire Midi-Pyrénées 14 avenue Edouard Belin – 31400, Toulouse, France e-mail: [email protected]

INTRODUCTION Measurement of the rate of diffusion of hydrous species in anhydrous minerals has been a subject of growing interest for the last ten years. Hydrogen diffusion studies are of fundamental importance to estimate the rate of transfer of hydrous species in natural conditions but also to identify the nature of the reactions involved in the speciation of hydrogen in minerals. For low concentration hydrous point defects, spectroscopic techniques are usually not sufficient to fully determine the location of hydrogen in the mineral structure. In these cases, kinetics studies provide additional constraints on the number of different defects stored in the mineral and their specific reaction rates. Numerous experimental data of hydrogen diffusion in minerals have now been published in the Earth Sciences literature. This review provides a digest to these data for a variety of minerals with a special focus on hydrogen in nominally anhydrous minerals. The review will not address the question of molecular water diffusion in minerals, melts or glasses. For a recent review of that question see for instance Doremus (2002). After a brief outline of basic diffusion concepts and methods of measurement, we summarize the main features of atomic diffusion linked to hydrogen isotope exchange and hydrogen reaction in minerals.

BASIC CONCEPTS OF DIFFUSION IN MINERALS Atomic diffusion is defined as the transport of matter in response to a driving force, which can be a chemical potential gradient or a temperature gradient. Atomic diffusion induced by a thermal gradient is very limited in solid earth sciences whereas the occurrence of strong chemical gradients is much more frequent in minerals. Atoms move from one region to the other in order to equilibrate the chemical potentials µ and reduce the Gibbs free energy of the system. The disequilibrium induced by a chemical potential gradient (dµ/dx)i of a component i within a crystal is directly proportional to its gradient in concentration (dC/dx)i if its activity ai is proportional to the concentration Ci (the activity coefficient γi is assumed constant in the crystal; see for instance Brady 1993; Doremus 2002). Most of the applications of diffusion in minerals are developed under this assumption. In this case, for two diffusion components, without any applied external forces, the diffusion coefficient, D > 0 (m2s−1), relates the flux of one component J (mol m−2s−1) to its one-dimensional gradient of the concentration dC/dx (mol·m−4) following: ⎛ ∂C ⎞ J = − D⎜ ⎟ ⎝ ∂x ⎠ t

Under general, non steady state conditions, the flux in each point, varies with time. In order to satisfy mass balance within the crystal the flux must obey the continuity equation: 1529-6466/06/0062-0013$05.00

DOI: 10.2138/rmg.2006.62.13

292

Ingrin & Blanchard ∂J ∂C =− ∂x ∂t

Combining both equations leads to Fick’s second law: ⎛ ∂ ⎡ ⎛ ∂C ⎞ ⎤ ⎞ ⎛ ∂C ⎞ ⎜ ∂t ⎟ = ⎜⎜ ∂x ⎢ D ⎜ ∂x ⎟ ⎥ ⎟⎟ ⎠ t ⎥⎦ ⎠ t ⎝ ⎠ x ⎝ ⎢⎣ ⎝

(1)

In this general law, the diffusion coefficient, D, is a function of concentration, C. If the diffusion coefficient is independent of concentration then Equation (1) can be simplified to ⎛ ∂ 2C ⎞ ⎛ ∂C ⎞ ⎜ ∂t ⎟ = D ⎜⎜ 2 ⎟⎟ ⎝ ⎠x ⎝ ∂x ⎠ t

(2)

This assumption is usually considered valid when the concentration of the diffusion species is very small, for instance for hydrogen in nominally anhydrous minerals. For hydrogen diffusion, authors have argued only in few cases, for a dependence of D on concentration when the quality of the fits by Equation (2) was not satisfactory (see for instance Wang et al. 1996). In this work we consider only volume diffusion through the crystal lattice, which is the dominant diffusion process at high temperature. The isotope diffusion coefficients for hydrogen in nominally anhydrous minerals are regarded as impurity tracer diffusion coefficients, which refer to diffusion of species at infinitely small dilution, while isotope diffusion coefficients for hydrous minerals refer to self-diffusion of intrinsic components of the mineral. Most of these coefficients can be assimilated to the impurity-diffusion or selfdiffusion of hydrogen. On the other hand, the effective diffusion coefficients determined from experiments in which hydrogen is extracted or incorporated correspond to the interdiffusion of species with different diffusivities. The relationship between the effective diffusivity Deff and the individual diffusivities of the species is directly linked to the involved reaction (see for instance, Kohlstedt and Mackwell 1998). If a single mechanism of diffusion is involved, the temperature dependence of D can be described by an Arrhenius equation of the form: ⎛ − ∆H ⎞ D = D0 exp ⎜ ⎟ ⎝ RT ⎠

(3)

where D0 (m2·s−1) is the pre-exponential factor, ∆H (J·mol−1) is the activation enthalpy, R is the gas constant (8.314 J·K−1mol−1) and T is the temperature (K).

EXPERIMENTAL METHODS All diffusion experiments aim to control the environment around the mineral (i.e., temperature; pressure; fugacities in oxygen, hydrogen, water; metal activities) in order to measure the kinetics of the mechanism of interest. Depending on the kind of minerals and reactions studied, the mineral sample can be placed directly into a furnace, which may be open or flushed by a gas (N2, H2, Ar + H2O, Ar + D2O...), or the sample can be sealed in a closed container with water and buffers, for instance, a silica ampoule placed in a furnace or a metal capsule which is then loaded in a piston cylinder or an internally-heated pressure vessel. Whatever the equipment used, two kinds of experiments can be distinguished: the bulk powder-fluid exchange experiments or experiments performed on single-crystals. In bulk powder-fluid (liquid or gas) exchange experiments, only the average extent of the exchange as a function of time is known from the analysis. The diffusion coefficients are then de-

Diﬀusion of Hydrogen in Minerals

293

termined from Fick’s second law (Eqn. 1 and 2) solved for the geometry and the boundary conditions imposed by the experimental settings. Some examples of these solutions are given in the next paragraph. As the material is crushed, an assumption must be made about the size and shape of grains. It is important then to have the most homogeneous grain population possible since depending on the grain geometry chosen, the uncertainties of the diffusivities can be significant. This technique also assumes that diffusion is isotropic which is not true for many minerals. Most of these experiments were originally performed by geochemists in order to measure the isotopic fractionation of hydrogen in hydrous minerals; determination of the diffusion coefficients was not the main objective of these studies. However, even though these results are based on relatively rough assumptions, they provide useful results on diffusion rates of hydrogen. Of the experiments performed on single-crystals, we distinguish the mass-loss experiments from those measuring diffusion profiles directly. In the first case, the sample undergoes successive periods of heating under the same conditions. After each heating step, the average hydrogen (or deuterium) concentration lost or remaining in the sample is measured, for instance, with a mass spectrometer, an infrared spectrometer or a thermobalance. In the second case, in profiling experiments, there is only one heating event, which lasts the time necessary for the formation of a full diffusion profile. This requires an assessment a priori of the diffusivity to choose the experiment duration. The diffusion profile is then measured directly by ion beam depth profiling analysis (nuclear reaction analysis or proton-proton scattering), through surface analysis following successive sectioning (scintillation counting) or the sample is cut in slices in order to measure the diffusion profile in the initial sample thickness by infrared absorption spectroscopy. In both mass-loss and diffusion profile measurements, the sample geometry is of great importance. First it simplifies the analytical treatment of the measurements. For mass-loss experiments, a sample with a plate shape with a small ratio of thickness over lateral sizes allows one to assume that the diffusion is unidirectional (i.e., normal to the plate surface) whereas for profiling experiments, one could assume that the diffusion profiles along the three perpendicular directions are independent by cutting the sample with appropriate length ratios. Second, once the sample is oriented with respect to crystallographic directions, it is possible to determine the anisotropy of diffusion. This point represents a great advantage of experiments using single-crystals over powder experiments. The equations used to fit the experimental measurements are solutions to Fick’s second law (Eqn. 1) for different geometries and boundary conditions (Carslaw and Jaeger 1959; Crank 1975). We consider here only the simplest situation where the diffusion coefficient is not a function of the concentration. All the following solutions correspond to diffusional transport-in experiments for some common sample geometries. For a semi-infinite solid with a homogeneous initial concentration, C0, in contact with an infinite, well-mixed outside reservoir of concentration, C1, the diffusion profile along the direction x perpendicular to the surface can be fitted by the following solution. C ( x, t ) − C 0 C1 − C 0

⎛ x ⎞ = erfc ⎜⎜ ⎟⎟ ⎝ 2 Dt ⎠

( 4)

In experiments where the sample geometry can be compared to a solid with homogeneous initial concentration, C0, bounded by two infinite parallel planes (thickness, 2L) and in contact with an infinite reservoir, the diffusion coefficient can be determined from the diffusion profile along x, by using the following solution C ( x, t ) − C 0 C1 − C 0

=1−

4 π

⎛ − D ( 2 n + 1)2 π 2 ⎜ exp ∑ ⎜ 4 L2 n =0 2n + 1 ⎝ ∞

( −1)

n

⎛ ( 2 n + 1) π x ⎞ t⎞ ⎟ cos ⎜ ⎟⎟ ⎜ ⎟ 2L ⎝ ⎠ ⎠

(5)

294

Ingrin & Blanchard

where the origin is located at the mid-point of the slab. If the measurements provide the average concentration rather than the concentration profile like in mass-loss experiments on single-crystals or powder, then Equation (5) has to be integrated over the sample thickness. The solution is then C ( t ) − C0 C1 − C0

=1−

∞

8 π2

∑

n =0

1

( 2n + 1)

2

⎛ − D ( 2 n + 1)2 π 2 exp ⎜ ⎜ 4 L2 ⎝

t⎞ ⎟ ⎟ ⎠

(66)

For the same kind of geometry, if the sample is not thin enough to assume unidirectional diffusion, Equation (5) becomes C ( x, y, z, t ) − C0 C1 − C0

=1−

64 π3

( −1) ∑ ∑ ∑ ( 2l + 1) ( 2m + 1) ( 2n + 1) l =0 m =0 n =0 ∞

∞

l+m+n

∞

( 7)

⎡ − π 2 t ⎛ D ( 2l + 1)2 D ( 2 m + 1)2 D ( 2 n + 1)2 ⎞ ⎤ y ⎜ x ⎟⎥ exp ⎢ + + z ⎢ 4 ⎜ ⎟⎥ b2 L2 a2 ⎝ ⎠⎦ ⎣ ⎛ ( 2l + 1) π x ⎞ ⎛ ( 2 m + 1) π y ⎞ ⎛ ( 2 n + 1) π z ⎞ cos ⎜⎜ ⎟⎟ cos ⎜⎜ ⎟⎟ ⎟⎟ cos ⎜⎜ 2L 2a 2b ⎝ ⎠ ⎝ ⎝ ⎠ ⎠

This corresponds to the case of a 2a × 2b × 2L parallelepiped where the diffusion coefficients along the three crystallographic directions are different. As before, the expression of the average concentration is obtained by integrating this equation over the volume analyzed. By integrating over the whole volume (2a × 2b × 2L), the average concentration is expressed as follows: C ( t ) − C0 C1 − C0

=1−

512 π6

∞

∞

∞

∑∑∑

l =0 m =0 n =0

⎡ − π2 exp ⎢ ⎢ 4 ⎣

1

( 2l + 1) ( 2m + 1) ( 2n + 1) 2

2

2

(8)

2 2 2 Dy ( 2 m + 1) Dz ( 2 n + 1) ⎞ ⎤ t ⎛⎜ Dx ( 2l + 1) ⎟⎥ + + ⎜ ⎟⎥ b2 L2 a2 ⎝ ⎠⎦

The corresponding solutions for a spherical geometry (radius, R) are, respectively, C ( r, t ) − C0 C1 − C0

=1+

2R πr

C ( t ) − C0 C1 − C0

∞

∑

( −1) n

n =1

=1−

n

6 π2

∞

⎛ − D n 2 π2 t ⎞ ⎛nπr ⎞ exp ⎜⎜ ⎟⎟ sin ⎜ ⎟ 2 R ⎝ R ⎠ ⎝ ⎠ 1

⎛ − D n 2 π2 R2 ⎝

∑ n 2 exp ⎜⎜ n =1

t⎞ ⎟⎟ ⎠

(9)

(10)

MEASUREMENT TECHNIQUES In this section, we describe briefly the techniques of analysis that have been reported in the literature for measuring hydrogen diffusion (infrared spectroscopy, mass spectrometry, nuclear reaction analysis, thermogravimetry, scintillation counting) as well as some other new techniques that have not been used for diffusion measurements but are very promising like proton-proton scattering. More details on analytical methods used for measuring water in minerals can be found in this volume (Rossman 2006). We end this section with a short review of the theoretical simulations that contribute to our understanding of the diffusion mechanisms.

Diﬀusion of Hydrogen in Minerals

295

Infrared spectroscopy Infrared spectroscopy is the most frequently used method to measure hydrogen diffusion. The vibrational modes of the OH dipole within the sample interact with the infrared beam and give rise to absorption bands. The concentration of OH is directly related to the intensity of the bands, and the concentration can be determined if the spectra are measured accurately; IR spectra of anisotropic minerals should be measured in polarized mode (cf. Libowitzky and Rossman 1996a) and the relation between absorption and concentration must be calibrated against an independent hydrogen analysis method. The position in wavenumber of the absorption band depends on the strength of the hydrogen bond, bond geometry and neighbors. Therefore polarized spectra also provide information about the structure of OH (Libowitzky and Beran 2006). Advantages of the IR technique are very high sensitivity (< ppm H2O), its ability to distinguish hydroxyl ions from adsorbed and intrinsic water molecules, and the ability to distinguish OH appearing in inclusions of hydrous phases from OH structurally in the parent phase (e.g., amphibole lamellae in clinopyroxene, Ingrin et al. 1989; Skogby and Rossman 1989). For diffusion experiments only the relative change of hydrogen content is necessary and no independent calibration is required. This technique can be used for measurements of the average concentration of hydrogen and deuterium in mass-loss experiments or measurements of the diffusion profiles. When the spectrometer is equipped with a microscope, measurements with a spatial resolution better than 50 µm can be easily achieved. Sample preparation, for profile measurements is described in Figure 1 and an example of a profile measurement is shown in Figure 2 for a garnet sample that has undergone partial hydrogen extraction in air.

Figure 1. Sample preparation for the measurement of diffusion profiles in a single-crystal by infrared spectroscopy.

296

Ingrin & Blanchard

a.

b.

10h at 1073K in air 100

80

60

40 -13

ms

-13

ms

D = 3.0 x 10

20

D = 4.0 x 10

2 -1 2 -1

-13

ms

0.3

0.4

D = 5.0 x 10

2 -1

0 -0.5 -0.4 -0.3 -0.2 -0.1 0.0

0.1

0.2

0.5

distance (mm)

Figure 2. Infrared spectra recorded through a single-crystal of grossular partially dehydrogenated (a) and corresponding fits of the normalized H concentration profile assuming constant diffusion coefficients (b). (after Kurka 2005).

Mass spectrometry Mass spectrometry techniques offer efficient means of measuring hydrogen isotope exchange (Graham 1981). The main disadvantage is that it is a destructive technique and has relatively poor spatial resolution. The sample is ionized. The ions of differing masses are separated and their relative abundances are recorded by measuring the intensities of ion flux (Graham et al. 1980). Secondary ion mass spectrometry (SIMS) represents an improved method by measuring directly hydrogen-deuterium exchange profiles over a few microns in length (Vennemann et al. 1996). This technique involves bombarding the sample surface with 16O− primary ion beam.

Diﬀusion of Hydrogen in Minerals

297

The secondary ions emitted from the sample are then measured with the mass spectrometer. Because the primary beam erodes the surface, a depth profile can be obtained. Vennemann et al. (1996) and Suman et al. (2000) report measurements of H-D diffusion profiles in hornblende, epidote and pargasite done by this method.

Thermogravimetry A novel thermogravimetric method has been developed to study H2O-D2O exchange in lawsonite (Marion et al. 2001). The mass difference between hydrogen and deuterium is sufficient to monitor the global weight change due to H-D exchange with a thermobalance. This simple experimental setup has proved to be efficient and accurate for hydrous minerals but cannot apply to minerals with low water content such as nominally anhydrous minerals.

Nuclear reaction analysis Nuclear reaction analysis allows measurements of hydrogen or deuterium diffusion profiles (Dersch et al. 1997). To detect the hydrogen, the surface sample is irradiated with 15N with variable energies in order to produce the following nuclear reaction under resonant conditions: 15

N + 1H → 12C + α + γ

The number of γ-rays emitted at any incident energy is proportional to the hydrogen concentration at the respective depth. To contribute to the nuclear reaction, the 15N nuclei have to slow down in the sample to reach the resonance energy (6.385 MeV). Thus each initial kinetic energy corresponds to a depth in the sample where the reaction occurs. The hydrogen diffusion profile is obtained by measuring the yield of the characteristic reaction γ-rays versus the beam energy. To probe the deuterium, a 3He beam with a fixed energy is used. The depth of the deuterium atom in the sample is known from the energy of the proton emitted by the reaction: 3

He + D → 4He + p

Ion-beam analysis displays a depth resolution of few nanometers decreasing slightly with the depth while the size of the beam is of the order of a millimeter. This technique has been used to investigate hydrogen diffusion in quartz (e.g., Dersch et al. 1997; Dersch and Rauch 1999). Figure 3 shows an example of a deuterium concentration profile measured in quartz.

Figure 3. Proton spectrum from the analysis of a deuterated quartz sample with a 700 keV 3He beam. The inset shows the depth profile determined from the spectrum. [Used with kind permission of SpringerVerlag from Dersch and Rauch 1999).

298

Ingrin & Blanchard

Liquid scintillation counting Tritium is sometimes used as a tracer of hydrogen diffusion (e.g., Shaffer et al. 1974). This hydrogen isotope, 3H, is radioactive and its concentration is measured by liquid scintillation counting. Tritium transforms by beta decay into stable helium (half life = 12.3 years). This analysis technique involves the detection of beta decay within the sample via capture of beta emissions in a scintillation cocktail containing organic solvents and solutes, fluor for instance. Beta particles emitted from the sample excite the solvent molecules, which in turn transfer the energy to the solute. The excited solute molecules dissipate the energy by emitting photons, which can be detected via a photomultiplier tube within a scintillation counter. The low-energy beta particles emitted from tritium have a small penetration length (~1 µm). This makes liquid scintillation counting a surface technique and depth resolution is ~1 µm. Tritium concentration profiles can be determined by performing successive steps of grinding and analysis. After each grinding and polishing event, the single crystal is immersed into the cocktail with its back face masked. Analyzing the activity of the removed material can be used to check the results provided by this procedure.

Proton-proton scattering A new analytical procedure has recently been developed to measure hydrogen depth profiles by elastic proton-proton scattering (Wegdén et al. 2005). A beam of 2.8 MeV protons (5-10 µm in diameter) in normal incidence is scanned over the sample. The scattered proton and recoiled target proton are detected coincidentally in the forward direction with an annular surface barrier detector. The summed energy of every detected proton-proton pair and the difference in their energy is used in an indirect approach to determine the depth location for every hydrogen event. The depth resolution is on the order of a micrometer. The major advantages of this method are the high detection cross section and the lowest possible irradiation damage effects compared to other ion-beam techniques (e.g., nuclear reaction analysis). However, sample preparation represents this technique’s main disadvantage. The sample must be thinner than 10-15 µm so that proton pairs from the entrance surface can travel through the whole sample thickness. Thus, diffusion profiles can only be determined to a limited depth. This technique has been tested for several nominally anhydrous minerals with hydrogen concentrations of 10 to 100 ppm H2O (Wegdén et al. 2005) but has not been applied yet to diffusion studies.

Theoretical techniques Many theoretical studies of the migration of protons (H+), neutral hydrogen atoms (H0) or hydrogen molecules (H2) in various materials are reported in the literature. Beyond providing important information on the microscopic mechanisms involved in hydrogen diffusion, these studies have industrial applications such as, for example, the trapping-detrapping of incident hydrogen in fusion devices, hydrogen storage in fuel cells, hydrogen separation processes in molecular sieves and the use of perovskite-type oxides with high proton conductivity as a separator material in electrochemical cells. The modeling techniques used are numerous and follow the technological progress in computational resources. Among them, we distinguish two main methods of determining diffusion laws. Trajectories of the diffusing particle can be obtained at any temperature in molecular dynamics simulations by integrating Newton’s equations of motion. The diffusion coefficients are then computed according to the Einstein relation for random walk: r (t ) − r (0)

2

=6 Dt

where r(t) is the position of the particle at time t and D is the diffusion coefficient. As in experimental studies, the diffusion law (activation energy and pre-exponential factor) is determined by plotting the diffusion coefficients on an Arrhenius graph. This method implies long simula-

Diﬀusion of Hydrogen in Minerals

299

tions (several nanoseconds) recording many hopping events in order to achieve a satisfactory statistics. Therefore only classical molecular dynamics employing an empirical atomistic model can be used to simulate the interatomic interactions. The second method is based on the equation: D=

1 2 ⎛ Q ⎞ ν l exp ⎜ − ⎟ 2 ⎝ kT ⎠

where ν is the attempt frequency, l is the hopping distance between energy minima, Q is the activation energy, k is the Boltzmann constant and T is the temperature. The hopping distance, l, is given by the crystal geometry. The activation energy corresponds to the difference between the energy of the system when hydrogen is located at the most energetically favorable position and when hydrogen is at the saddle point connecting the two energy minima. It is then known from the calculation of the potential energy surface. Any static calculations can provide this information. The simulation techniques can be based either on empirical interatomic potentials or on first-principles solving the Schrödinger equations with some approximations. On the other hand, the attempt frequency is determined by molecular dynamics. The attempt frequency is much faster than the rate of hopping events, and molecular dynamics calculations can be much shorter than in the first method (several picoseconds). Quantum molecular dynamics calculations are much more computationally expensive and can be performed following the classical molecular dynamics calculation. In this case, the ground state energy for each atomic configuration is calculated using first-principles techniques.

DETECTION OF H DIFFUSION THROUGH ISOTOPE EXCHANGE Isotopic exchange is a basic tool to estimate the diffusion coefficients of hydrogen in mineral structures by measurement of deuterium (D) or tritium (T). Such isotopic diffusion experiments also provide a direct measurement of the kinetics of isotope exchange in natural rocks and an understanding of the conditions that preserve the observed H-D isotope disequilibrium in nature.

Anhydrous minerals Rutile. Except for rutile there are few data for true isotopic exchange in single oxide minerals. In rutile, data have been collected for the three isotopes, H, D and T (Johnson et al. 1975; Cathcart et al. 1979). Exchanges have been done under equilibrium conditions; the hydrous content of the sample was in equilibrium with the enriched H2O, D2O or T2O gas. The diffusion laws obtained in the two crystallographic directions a, and c, for H-D and H-T exchange were recalculated using York’s least-squares fit method, assuming an uncertainty of 5 K in T and 0.15 in logD (York 1966; Table 1): ⎛ −121 ± 6 kJmol −1 ⎞ H DTiO = D exp ⎜⎜ ⎟⎟ , with l og D0 = −4.55 ± 0.229 0 a 2 RT ⎝ ⎠ −1 ⎛ −56 ± 2 kJmol ⎞ H DTiO = D0 exp ⎜⎜ ⎟⎟ , with log D0 = −6.78 ± 0.12 2c RT ⎝ ⎠ −1 ⎛ −106 ± 3 kJmol ⎞ T = D0 exp ⎜⎜ DTiO ⎟⎟ , with log D0 = −5.84 ± 0.13 2a RT ⎝ ⎠ −1 ⎛ −70 ± 2 kJmol ⎞ T DTiO = D0 exp ⎜⎜ ⎟⎟ , with log D0 = −6.25 ± 0.10 2c RT ⎝ ⎠

300

Ingrin & Blanchard

Table 1. Isotope diffusion data. Sample orientation

Diffusing T range species (K)

P (MPa)

Rutile TiO2; ~ 17 ppm H2O

Single xtal // a

H-D exch.

887-994

7×10−3

Rutile TiO2; ~ 17-45 ppm H2O

Single xtal // c

H-D exch.

623-973

7×10−3

Rutile TiO2; ~ 3 ppm H2O

Single xtal // a

H-T exch.

773-1183

4×10−4

Rutile TiO2; ~ 3 ppm H2O

Single xtal // c

H-T exch.

527-973

4×10−4

Quartz-α; ~ 30 ppm H2O

Single xtal // c

H-D exch.

673-893

2.5; H2O/D2O vapor

Quartz-β; ~ 30 ppm H2O

Single xtal // c

H-D exch.

893-1273

2.5; H2O/D2O vapor

Quartz-β; ~ 13 ppm H2O

Single xtal // c

H-D exch.

973-1173

1100; D2O fluid

Quartz-β; ~ 13 ppm H2O

Single xtal ⊥ c

H-D exch. 1073-1173

1100; D2O fluid

Diopside Fe/(Fe+Mg) = 0.036; ~ 10-40 ppm H2O

Single xtal // c and a*

H-D exch.

873-1139

0.1; 90%Ar + 10% D2

Diopside Fe/(Fe+Mg) = 0.036; ~ 10-40 ppm H2O

Single xtal // b

H-D exch.

973-1173

0.1; 90%Ar + 10% D2

Forsterite with 0.25 wt% Fe; (Libowitzky and Beran 1995) < 100 ppm H2O

Single xtal // c

H-D exch.

973-1423

0.1; 90%Ar + 10% D2

Mineral / range of proton solubility Nominally anhydrous minerals

γ-Spinel (synthetic, Mg2GeO4);

Single xtal

H-D exch.

873-973

0.1; 90%Ar + 10% D2

Pyrope (Gr3Alm15Py81); ~ 13-36 ppm H2O

Single xtal

H-D exch.

973-1223

0.1; 90%Ar + 10% D2

Grossular (Gr84And14Py2); 220 ppm H2O

Single xtal

H-D exch. 1073-1323

0.1; 90%Ar + 10% D2

Grossular (Gr73And23Py2); 1400 ppm H2O

Single xtal

H-D exch.

973-1223

0.1; 90%Ar + 10% D2

Andradite (Gr1And99); 1500 ppm H2O

Single xtal

H-D exch. 1073-1173

0.1; 90%Ar + 10% D2

Ilvaite; ~ CaFe2Fe3+Si2O8 (OH)

powder

H-D exch.

5-20; H2O/D2O fluid

Zoisite; ~ Ca2Fe0.1Al2.9Si3O12(OH)

powder

H-D exch.† 623-923

200 or 400; H2O/D2O fluid

Epidote; ~ Ca2Fe0.9Al2.1Si3O12(OH)

powder

H-D exch.† 523-923

200 or 400; H2O/D2O fluid

Epidote; ~ Ca2FeAl2Si3O12(OH)

Single xtal // b

H-D exch.

200; 99%D2O

Hydrous minerals 623-923

473-873

Epidote

?

H-D exch.

423-673

Lawsonite; CaAl2Si2O7(OH)2H2O‡

powder

H-D exch.

648-698

0.1; Ar + D2O

Tourmaline; 14.27% FeO, 1.93% MgO, 31.34% Al2O3

powder

H-D exch.

723-1073

15-25; H2O/D2O fluid

Hornblende; ~ (NaK) Ca2(Mg2.4Fe1.8Al0.8)(Si6.5Al1.5)O22(OH)2

powder

H-D exch.† 623-823

200 to 800; H2O/D2O fluid

Kaersutite; (Na0.4K0.3)(Ca1.6Na0.4) (Mg2.6Fe2+1Fe3+0.5Ti0.5Al0.3)(Si5.9Al2.1)O22(OH)2

Single xtal // b

H-D exch.

873-1173

0.1; 90%Ar + 10% D2

Tremolite; ~ Ca2(Mg4.8Fe0.2.)Si8O22(OH)2

powder

H-D exch.

623-1073

200 to 400; H2O/D2O fluid

Actinolite; ~ Ca2(Mg4 Fe1)Si8O22(OH)2

powder

H-D exch.† 673-943

200; H2O/D2O fluid

Chlorite; ~ (Mg0.7Al0.3Fe)12Si5.5Al2.5O20(OH)16

powder

H-D exch.† 773-973

200 or 500; H2O/D2O fluid

Muscovite; ~ K2Al4Si6Al2O20(OH,F)4

powder

H-D exch.† 723-1023

200 or 400; H2O/D2O fluid

Notes: ‡ Data from two different samples. † possible H-D leak through the capsule References: [1] Johnson al. (1975); [2] Cathcart et al. (1979); [3] Kats et al. (1962); [4] Kronenberg et al. (1986); [5] Hercule and Ingrin (1999); [6] Ingrin unpublished data; [7] Hertweck and Ingrin (2005a, 2005b); [8] Blanchard and Ingrin (2004a); [9] Kurka et al. (2005); [10] Kurka (2005); [11] Yaqian and Jibao (1993); [12] Graham (1981);

Diﬀusion of Hydrogen in Minerals

∆H (kJ•mol−1)

301

log D0 (m2s−1)

Comments

3×10−3-7×10−3 121±6

H: −4.55±0.29 D: −4.72±0.30

Sequential IR measurement

[1]

3×10−3-7×10−3 56±2

H: −6.78±0.12 D: −6.92±0.12

Sequential IR measurement

[1]

10−30-10−6

106±3

−5.84±0.13

Liquid-scintillation counting; No effect of pO2

[2]

10−39-10−17

70±2

−6.25±0.10

Liquid-scintillation counting; No effect of pO2

[2]

unbuffered

69±12

−8.95±0.83

Sequential IR measurement

[3]

unbuffered

169±15

−3.66±0.67

Sequential IR measurement

[3]

unbuffered

215±92

−0.45±4.45

Bulk IR after a single annealing, only 3 data

[4]

unbuffered

156

−3.45

Bulk IR after a single annealing, only 2 data

[4]

10−26-10−19

149±16

−3.4±0.8

Sequential IR measurement

[5]

10−23-10−18

143±33

−5.0±1.7

Sequential IR measurement

[5]

10−23-10−13

134±7

−7.5±0.3

Sequential IR measurement

[6]

10−23-10−13

140±34

−5.8±1.9

Sequential IR measurement

[7]

10−23-10−18

140±38

−5.8±1.9

Sequential IR measurement

[8]

10−21-10−16

102±45

−7.6

Sequential IR measurement

[9]

10−23-10−18

185±28

−3.8±1.3

Sequential and profile IR measurement

[10]

10−21-10−19

70

− 8.9

Sequential and profile IR measurement Only 2 data points

[10]

unbuffered

115-119

−7.0 to –7.4

Bulk analysis by mass spectrometer. 1D-2D diffusion model (overestimation of D). Grain size ~ 50 µm

[11]

unbuffered

100-103

−7.8 to –8.4

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain size ~ 68-75 µm

[12]

unbuffered

52 to 58 (T≥723 K) 128 (T≤623 K)

−9 to −9.5 (T≥723 K) −4 to −3 (T≤623 K)

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain size ~ 68-75 µm

[12]

unbuffered

67

−13.6

SIMS, in contradiction with earlier paper from the same authors (Chacko et al. 1999)

[13]

unbuffered

77

−9.7

SIMS, and/or Mass spectrometer?

[14]

80±70

−9±5

Bulk analysis by thermogravimetry. Grain size ~ 77 µm. Anisotropic diffusion evidenced by Libowitzky and Rossman (1996b)

[15]

unbuffered

123-128

−9.6 to –10.0

Bulk analysis by mass spectrometer. 1D-2D diffusion model. Grain size ~ 50 µm

[16]

unbuffered

79 to 84

−10.8 to –11.6

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain lengths (30-50 µm)

[17]

10−26-10−18

104±12

−8.7±0.7

Sequential IR measurement

[18]

unbuffered

71.5

−10.7 to –12.1

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain lengths (30-50 µm)

[17]

unbuffered

99

−9.2

Bulk analysis by mass spectrometer. 2D diffusion model, grain size assumed = 56 µm

[19]

unbuffered

166 to 172

−5.2 to –7.3

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain sizes (30-100 µm)

[20]

unbuffured

121

−8 to −11

Bulk analysis by mass spectrometer. 1D-2D diffusion model, grain size ~ 75-150 µm

[21]

O2 (MPa)

Ref.

[13] Suman et al. (2000); [14] Vennemann et al (1996); [15] Marion et al. (2001); [16] Jibao and Yaqian (1997); [17] Graham et al. (1984); [18] Ingrin and Blanchard (2000); [19] Graham et al. (1984); from Suzuoki and Epstein (1976); [20] Graham et al. (1987); [21] Graham (1981) from Suzuoki and Epstein (1976)

302

Ingrin & Blanchard

The diffusion coefficients for deuterium are lower than those for hydrogen by a factor of 1.3 to 1.5 but have the same activation enthalpy. The results from experiments with tritium are very close too but with a slightly different activation enthalpy (Fig. 4). The differences have been attributed to isotopic behavior linked to the mass difference of isotopes and experimental uncertainties (Cathcart et al. 1979). Hydrogen diffusion in TiO2 has also been studied by Bates et al. (1979) performing static calculations based on semi-empirical interatomic potentials. The calculated activation energies for the diffusion along the c- and a-axes fall within the experimental uncertainties. This agreement supports the idea that hydrogen diffusion in rutile involves the migration of protons from one lattice oxygen to another. Moreover the strong anisotropy observed along with the difference in activation enthalpy with transport direction can be explained by looking at the migration paths. In the rutile structure, large open channels between the TiO6 octahedra are aligned parallel to the c-axis. Thus the diffusion path in the c direction describes a helix as the proton moves from one oxygen ion to another along the channel. In the direction perpendicular to the c direction, on the other hand, the mechanism consists of a combination of rotations of the hydroxyl, OH−, to move the proton from one channel to another and jumps that are required along the channel. Quartz. A large number of studies have been dedicated to water in quartz; however, there are few results for true isotope diffusion experiments. The main results come from studies by Kats et al. (1962) and Kronenberg et al. (1986) (Table 1). Hydrogen diffusion anisotropy was only addressed by Kronenberg et al. (1986). These authors observed no anisotropy of diffusion (Fig. 5), but the small number of data, their uncertainties and the small ratio of samples diameter over thickness (6.3 mm/3 mm) limit the weight of this statement. The fit and the uncertainties of the parameters of the diffusion laws determined from these two studies were recalculated

T (K) -9.5

1173

973

873

773

623

523

Rutile

-10

logD (m2s-1)

-10.5 -11 -11.5

//c

-12 -12.5

//a

-13 -13.5 0.8

1

1.2

1.4

1.6

1.8

2

103/T (K-1) Figure 4. Compilation of isotope diffusion data for rutile. Open symbols: H diffusion (Johnson et al. 1975); solid symbols: D diffusion (Johnson et al. 1975); symbols with cross: T diffusion (Cathcart et al. 1979).

Diﬀusion of Hydrogen in Minerals

303

with York’s method assuming individual uncertainties of diffusion coefficients equal to those plotted in Figure 5 (Table 1). The two studies roughly agree for β quartz within the range of uncertainties. It must also be noticed that the change of diffusion behavior observed by Kats et al. (1962) below 893 K is not necessarily linked to the change of structure from β to α quartz but to the presence of unstable OH bands below this temperature (Kats et al. 1962). Bongiorno et al. (1997) have simulated the diffusion of hydrogen in the different SiO2 structures by classical molecular dynamics. The calculated activation enthalpies of about 125 and 60 kJ·mol−1 for α- and β-quartz respectively, are in close agreement with the experimental values. The theoretical enthalpy for α-quartz has later been confirmed by static ab initio calculations, which indicated an energy of about 135 kJ·mol−1 (Bunson et al. 1999). It is not clear, however, why the calculated activation enthalpy for β-quartz is smaller than for α-quartz while the observed activation enthalpy above 893 K is greater than at lower temperatures. In any case, the main conclusion drawn from the computer simulations is that hydrogen diffusion in quartz should be anisotropic. The calculated hydrogen trajectories in α-quartz differs slightly from that in β-quartz due to the higher crystal symmetry but in both cases, the diffusion path is confined in the small hexagonal channels parallel to the c-axis. No quantitative value of the anisotropy was deduced from the modeling. As discussed above this anisotropy is not clearly observed but a difference up to a factor 5 might be hidden behind the experimental uncertainties. Diopside. In silicates, the most detailed isotopic study has been performed in diopside. The diffusion coefficients have been measured in three crystallographic orientations. The study has not been performed exactly under equilibrium conditions, but the very fast kinetics of H-D exchange, compared to hydrogen uptake in diopside, precludes any significant loss

T (K) 1173 1073

973

873

773

673

-9

Quartz

-10

logD (m2s-1)

-11

-12

-13

-14

-15 0.8

1

1.2 3

1.4

1.6

1

Figure 5. Compilation of isotope diffusion data in quartz. Circles: H-D diffusion data after Kronenberg et al. (1986) (solid symbols // c, open symbols ⊥ c); squares: H-D diffusion above 893 K (Kats et al. 1962); triangles: H-D diffusion below 893K (// c, Kats et al. 1962).

304

Ingrin & Blanchard

of hydrogen species during the H-D exchange experiments (Hercule and Ingrin 1999). The results compiled in Table 1 show that hydrogen diffusion is faster along c and a* directions than along b, but transport in the three directions have comparable activation enthalpies around 140-150 kJ·mol−1. As for rutile, the same type of anisotropy is observed for oxygen diffusion in dry diopside (Ingrin et al. 2001). Olivine and spinel phases. No data exist for mantle olivine; the only data for H-D exchange are for forsterite. The activation enthalpy is close to those of diopside but the diffusion coefficients are two to three orders of magnitude lower than for natural diopside (Fig. 6). Recent data for synthetic Mg2GeO4 spinel give a comparable activation enthalpy for H-D exchange (Fig. 6, Table 1). Garnets. Isotope diffusion studies have been performed only recently for garnets: in Dora Maira pyrope, two grossulars and one andradite single crystals (Table 1). It was not possible to perform the same type of H-D exchange experiments in mantle pyropes because they have comparable kinetics of hydrogen uptake (Blanchard and Ingrin 2004a). The three types of garnets have diffusion coefficients that compare favorably within one order of magnitude. A summary for all garnet data gives the general law (Fig. 7): ⎛ −130 ± 15 kJmol −1 ⎞ Dgarnet = D0 exp ⎜⎜ ⎟⎟ , with l og D0 = −6.28 ± 0.72 RT ⎝ ⎠

Hydrous minerals Numerous experiments of H-D exchange have been reported for hydrous minerals; most of them have been done on powder samples. Most of the studies assume that the crystal grains

T (K) 1423

-10

1173

logD (m2s-1)

-11

1073

973

diopside

873

// c, a*

-12

// b -13

forsterite

Ge-spinel

-14

// c

-15

-16 0.6

0.7

0.8

0.9

1

1.1

1.2

103/T (K-1) Figure 6. Compilation of H-D diffusion data in diopside, forsterite and Mg2GeO4 spinel. Circles: diopside (Hercule and Ingrin 1999); squares: forsterite (Ingrin unpublished data); triangles: spinel (Hertweck and Ingrin 2005a,b).

Diﬀusion of Hydrogen in Minerals

305

T (K) 1373

1273

1173

1073

973

-10

H-D Garnets

logD (m2s-1)

-11

-12

-13

-14

-15 0.7

0.8

0.9

103/T

1

1.1

(K-1)

Figure 7. Compilation of H-D diffusion data in garnets. Solid circles: pyrope (Blanchard and Ingrin 2004a); squares: grossular (Kurka et al. 2005); diamonds: grossular (Kurka 2005); triangles: andradite (Kurka 2005).

have infinite plate shapes, with diffusion being essentially perpendicular to the plate, following the solution of Equation (6) or the grains are modeled as infinite cylinders with a contribution from radial diffusion alone (these two assumptions are labeled 1D and 2D respectively in Table 1). This approach renders relatively inaccurate measures of diffusion coefficients and does not take account for possible diffusion anisotropy. A summary of these data for the minerals of the epidote group is presented in Figure 8. The activation enthalpy for these minerals is around 80 to 110 kJ·mol−1. The data of Graham (1981), which were originally fitted with two different laws (Table 1), can also be fitted by a single law with an activation enthalpy around 90 kJ·mol−1. Furthermore the reported difference in temperature dependence has not been confirmed by later studies (Vennemann et al. 1996; Suman et al. 2000). The only data collected on single crystals are reported by Suman et al. (2000) and diffusion coefficients from this study are more than 3 orders of magnitude lower than those determined using powder methods. These data were only published in an abstract and never confirmed later; they must be analyzed with some caution. Amphiboles and lawsonite have activation enthalpies in the same range as values for epidote, between 80 and 100 kJ·mol−1; only tourmaline and sheet silicates have higher activation enthalpies from 120 to 150 kJ·mol−1 (Fig. 9; Table 1). A study performed on single crystal kaersutite amphiboles confirms that H-D exchange is anisotropic in amphibole with a diffusion coefficient along the c direction that is five times faster than along the b direction (Ingrin and Blanchard 2000). The activation enthalpy for diffusion along b direction agrees with those found for actinolite and hornblende from powder experiments. Almost all the data for hydrous minerals were collected on powders with small grain size. This explains why these data are available only for temperature conditions lower than

306

Ingrin & Blanchard T (K) 923

823

723

623

523

-12

epidote (graham 1981)

-13

-1

2 log D (m s )

-14 -15

ilvaite zoisite

-16 -17

// b

-18

epidote (Suman et al 2000)

-19

epidote (Vennemann et al. 1996)

-20 0.6

0.8

1

1.2

3

1.4

-1

1.6

1.8

2

2.2

10 /T (K ) Figure 8. Compilation of H-D diffusion laws for hydrous minerals of the epidote group: epidote, zoisite, ilvaite (see Table 1 for details). Only laws deduced from the “infinite cylinder” model of grains are presented here.

T (K) 1073

923

823

723

623

-12

// c

-13

// b

-1

2 log D (m s )

-14

kaersutite

-15

lawsonite

-16

actinolite

-17

tourmaline

-18

tremolite hornblende

-19 -20 0.6

0.8

1

1.2 3

1.4 -1

1.6

1.8

2

10 /T (K ) Figure 9. Compilation of H-D diffusion laws for lawsonite, tourmaline, amphiboles and sheet silicates (chlorite: dashed line, muscovite: dotted line; circles: kaersutite single crystal; see Table 1 for details). Only laws deduced from the “infinite cylinder” model of grains are represented.

Diﬀusion of Hydrogen in Minerals

307

data obtained for anhydrous minerals. As a general trend, the diffusivities of hydrous minerals compare well with those of anhydrous minerals but the activation enthalpies seem lower (close to 100 rather than 140 kJ·mol−1; Fig. 10). Of course, this statement is only true if the extrapolation of the data of hydrous minerals is valid above 1173 K (i.e., if there is no change of mechanism with temperature). During H-D exchange in silicates, hydrogen is expected to migrate mainly as a single proton much as in oxides, jumping between successive OH positions (Norby and Larring 1997). It has been suggested that only a minor part of the activation enthalpy is due to proton transfer between the two oxygen atoms; the larger part of the activation is linked to the vibrational energy necessary to decrease the oxygen-oxygen distance of the oxygen sublattice sufficiently to allow the proton jump (Norby and Larring 1997). In this case, the activation enthalpy of hydrogen diffusion in minerals should follow that of oxygen diffusion. The activation enthalpy of oxygen self-diffusion in hydrous minerals like hornblende and tremolite (<200 kJ·mol−1) is indeed lower than oxygen self-diffusion in anhydrous silicates (quartz, diopside, olivine, garnets; Brady 1993).

T (K) 1423

1173

873

723

623

1.4 -1

1.6

523

-10 -11

-1

2 log D (m s )

-12

diopside garnet

-13

spinel forsterite

-14 -15

-16 -17 -18 0.6

0.8

1

1.2 3

1.8

2

10 /T (K ) Figure 10. Comparison of H-D diffusion laws for anhydrous and hydrous minerals (hydrous phases: grey lines; anhydrous phases: symbols and dark lines).

EXTRACTION/INCORPORATION REACTIONS IN ANHYDROUS MINERALS Olivine The first extensive study on the kinetics of hydrogen uptake in olivine has been performed by Mackwell and Kohlstedt (1990). This study was later followed by more complete studies in olivine and forsterite (Kohlstedt and Mackwell 1998; Demouchy and Mackwell 2003; Table 2). The authors interpret their results through two different mechanisms:

308

Ingrin & Blanchard

Table 2. Hydrogen uptake diffusion data. Mineral / range of proton solubility

Sample orientation

Diffusing species

T range (K)

P (MPa)

Forsterite synthetic Mg2SiO4; ~ 1-7 ppm H2O Forsterite synthetic Mg2SiO4; ~1-7 ppm H 2O Forsterite synthetic Mg2SiO4; ~1-7 ppm H2O

Single xtal // c

H uptake

1173-1383

1500 & 200; H2O fluid

Single xtal // b

H uptake

1273-1383

1500 & 200; H2O fluid

Single xtal // a

H uptake

1273-1373

1500 & 200; H2O fluid

Olivine San Carlos Mg0.91Fe0.09 Ni0.003)2SiO4 ; ~2-10 ppm H2O Olivine San Carlos Mg0.91Fe0.09 Ni0.003)2SiO4; ~2-10 ppm H2O

Single xtal // a

H uptake

1073-1273

300; H2O fluid

Single xtal // c

H uptake

1073-1273

300; H2O fluid

Olivine San Carlos (Mg0.91Fe0.09 Ni0.003)2SiO4;~2-10 ppm H2O Olivine San Carlos (Mg0.91Fe0.09 Ni0.003)2SiO4; ~ 2-10 ppm H2O Olivine San Carlos Mg0.91Fe0.09 Ni0.003)2SiO4; ~ 2-10 ppm H2O Olivine San Carlos Mg0.91Fe0.09 Ni0.003)2SiO4; ~ 2-10 ppm H2O Olivine San Carlos (Mg0.91Fe0.09 Ni0.003)2SiO4; ~ 2-10 ppm H2O Olivine San Carlos (Mg0.91Fe0.09 Ni0.003)2SiO4; ~2-10 ppm H2O Diopside Fe/(Fe + Mg) = 0.07; ~8 ppm H2O

Single xtal // c

H uptake

1073-1273

300; H2O fluid

Single xtal // a

H uptake

1073-1273

300; H2O fluid

Single xtal // b

H uptake

1073-1273

300; H2O fluid

Single xtal // c

H uptake†

1173-1273

200; H2O fluid

Single xtal // a

H uptake†

1173-1273

200; H2O fluid

Single xtal // b

H uptake†

1173-1273

200; H2O fluid

Single xtal // a, c*

H extraction

973-1123

CO/CO2

Single xtal // b

H extraction

1023-1123

CO/CO2

Single xtal // a*, b, c

H extraction in 973-1273 air; H uptake at 0.01 and 0.1 MPa of H2 H extraction 873-1173

Air

Single xtal // ~[212]; ~70° from b Single xtal

H extraction

973-1173

Air

H uptake†

1473

Single xtals // a, b

H extraction in air

973-1173

Annealed in air; then at 1273K; pH2 = 0.1 MPa Air; H2

Enstatite, Kilbourne Hole En90Fs70; Fe3+/Fetotal = 3.5%

Single xtals // a, b, c

Air

Pyropes 8 different megacrysts Py67-72Alm14-21Gr10-14; 22 – 112 ppm H2O Pyrope from Dora Maira Gr3Alm15Py81; ~ 13-36 ppm H2O Grossular (Gr84And14Py2); 220 ppm H2O Grossular (Gr73And23Py2); 1400 ppm H2O

H extraction in 973 air; H uptake at 0.1 MPa H extraction 965-1223

Single xtal (2 ≠ OH groups) Single xtal Single xtal

H extraction

1073-1323

Air; 90%Ar + 10% H2

H extraction H extraction

1073-1323 973-1223

Air Air

H extraction

973-1223

Air

Quartz-β; ~5 ppm H2O Quartz-β; ~13 ppm H2O

Single xtal (2 ≠ OH groups) Single xtal // c Single xtal // c

T uptake H uptake

993-1123 973-1173

0.06 T2O vapor 890-1540; H2O fluid

Quartz-β; ~13 ppm H2O

Single xtal ⊥ c

H uptake

1073

890; H2O fluid

Adularia feldspar; Structural H2O; Ab9Or90Cs1; ~90 ppm H2O 0.11 wt% Fe2O3, 0.03 wt% FeO Andesine plagioclase feldspar, Structural OH defects; Ab66An30Or3; 510 ppm H2O 0.11 wt% Fe2O3, 0.03 wt% FeO

Single xtal // [001]*

H extraction and few uptake exp. H extraction

773-1173

0.1 in air

1073-1273

0.1 N2

Diopside Fe/(Fe + Mg) = 0.07; ~8 ppm H2O Diopside Fe/(Fe + Mg) = 0.036; ~10-40 ppm H2O Diopside Fe/(Fe + Mg) = 0.05; ~5-20 ppm H2O Diopside Fe/(Fe + Mg) = 0.126; ~2-10 ppm H2O Diopside Fe/(Fe + Mg) = 0.036; ~10-40 ppm H2O Enstatite, synthetic

Andradite (Gr1And99); 1500 ppm H2O

Single xtal // c

Single xtals

Single xtal // [010]*, ⊥ [010]*,

Air; 90% Ar + 10% H2; H2

Air and N2

Notes: † H uptake involving cation vacancies. References: [1] Demouchy and Mackwell (2003); [2] Mackwell and Kohlstedt (1990); [3] Kohlstedt and Mackwell (1998); [4] Carpenter-Wood et al. (2000); [5] Ingrin et al. (1995); [6] Hercule and Ingrin (1999); [7] Hercule (1996);

Diﬀusion of Hydrogen in Minerals

O2 (MPa)

∆H (kJ•mol−1)

log D0 (m2s−1)

Comments

Ni/NiO buffer 10−13-10−10 Ni/NiO buffer 10−13-10−10 Ni/NiO buffer 10−13-10−10 Fe/FeO, Ni/NiO buffer 10−19-10−11 Fe/FeO, Ni/NiO buffer 10−19-10−11

210±33

−3.3±1.3

205±31

−4.1±1.2

225±40

−3.8±1.6

130±30

−4.22±0.30

130±30

−5.30±0.70

Fe/FeO buffer; 10−19-10−13 Fe/FeO buffer 10−19-10−13 Fe/FeO buffer 10−19-10−13 Ni/NiO buffer 10−13-10−11 Ni/NiO buffer 10−13-10−11 Ni/NiO buffer 10−13-10−11 10−15; olivine buffered

110±50

252±51 and –17±2.1 through York’s fit. (9 pts). IR profiles 188±15 and –4.8±0.6 through York’s fit. (6 pts). IR profiles 217±52 and –4.2±2.0 through York’s fit. (4 pts). IR profiles 5×D//b < D//c; 118±27 and –4.71±1.19 through York’s fit. (7pts). IR profiles IR profiles. The same activation enthalpy than direction c is assumed. 62±27 and –8.40±1.16 through York’s fit. (3-4pts) IR profiles. 109±47 and –6.59±2.06 through York’s fit. (5pts). IR profiles. 145±25 and –3.53±1.13 through York’s fit. (5pts). IR profiles. 188±50 and –3.21±2.21 through York’s fit. (5pts). IR profiles; Fit from 2 pts; ∆H = 315; logD0 = 1.60 IR profiles; Fit from 2 pts; ∆H = 286; logD0 = -0.97 IR profiles; Fit from 2 pts; ∆H = 349; logD0 = 2.01 Sequential and profile IR measurement. No obvious anisotropy, single fit of data give: ∆H = 160±21; logD0 = -3.06±1.05 profile IR measurement.

145±30 180±50 260±20 assumed 260±20 assumed 260±20 assumed a: 181±38 c*: 153±32

10−15; olivine buffered 0.021 in air; 126±24 10−17-10−21 in Ar/10%H2

a: −2.1±1.9 c*: −3.4±1.6 One order of mag. below a, c −6.7±1.1

309

Ref. [1] [1] [1] [2] [2] [3] [3] [3] [3] [3] [3] [4] [4]

Sequential IR measurement. No detectable anisotropy, no dependence with pO2 or pH2

[5], [6]

0.021

107±30

−6.64±1.46

Sequential IR measurement (4pts.)

[7]

0.021

109

−5.7

Sequential IR measurement (2pts.)

[7]

D = 5±2 10−131

IR profile measurement. Isotropic diffusion of related point defects is assumed Sequential IR measurement, isotropic, only 2 pts per direction

[8], [9] [10]

logD = a: −13.6; b: −13.2; c: −14.2

Sequential IR measurement

[10]

0.021 0.021

295±55

0.021 in air

0.14

unbuffered

253±13

0.2 to 0.8

Sequential and profile IR measurement

[11]

0.021 in air 10−21-10−16 0.021 0.021

OH3650: 277±22 OH3600: 329±21 323±46 180±10

OH3650: 0.5±1 OH3600: 1.9±0.9 1.0 −3.7±0.3

[12]

0.021

OH3620: 271±35 OH3560: 209±25 100±9 200±20

OH3620: 0.0±1.6 OH3560: −1.8±0.8 −10.37±0.45 −0.85

Sequential and profile IR measurement. pO2 dependence: pO20.12, pO20.15, resp. Sequential IR measurement (5 points) Sequential and profile IR measurement (8 points). pO2 dependence Sequential IR measurement (3 and 4 points resp.) Liquid-scintillation counting Bulk IR after a single annealing, only 3 data, thick sample Bulk IR after a single annealing, only 1 data, thick sample Sequential IR measurement (only one heating step per data points)

unbuffered unbuffered

D = (2±1.15)×10−11

unbuffered 0.021

172±15

−3.21

unbuffered

224±33

−3.24±0.25

Sequential polarized IR measurement Diffusion is isotropic (8 data points)

[13] [14] [14] [15] [16] [16] [16] [17]

[8] Guilhaumou et al. (1999); [9] Ingrin and Skogby (2000); [10] Stalder and Skogby (2003); [11] Wang et al. (1996); [12] Blanchard and Ingrin (2004b); [13] Kurka et al. (2005); [14] Kurka (2005); [15] Shaffer et al. (1974); [16] Kronenberg et al. (1986); [17] Johnson (2003)

310

Ingrin & Blanchard .

1.

a relatively fast redox-exchange reaction involving polarons h (electron holes), localized on iron atoms, following the relation: Fex + H·i = Fe· + ½H2 or H·i = h· + ½H2 (11) . where superscripts x, and subscript i denote neutral and positive charges and interstitial sites, respectively.

2.

a slower exchange process controlled by the mobility of metal vacancies point defects following the reaction: .

MeO + V''Me + 2H i = Mex + H2O

(12)

where superscripts ' and subscript Me denote negative charge and metal sites, respectively. The first type of exchange implies that the flux of protons are charge compensated by a counter flux of polarons: JH = −Jh. Kohlstedt and Mackwell (1998) suggested that the effective diffusivity Deff measured during hydrogen uptake experiments at low temperature is related to the diffusivities of polarons Dh and hydrogen DH by the relation: Deff =

2 DH Dh DH + Dh

(13)

If Dh >> DH, then Deff ≈ 2DH. However, no data on H-D exchange has yet been performed in olivine in order to prove this conclusion. We report only data of effective diffusivity in Table 2. For the exchange Reaction (12), Kohlstedt and Mackwell assume that the flux of metal vacancies is charge compensated by the counter flux of protons (JH = −2JVMe) and Dh >> DH, and conclude that the effective diffusivity is given by: Deff =

3DH DVMe 2 DVMe + DH

(14)

where DVMe is the diffusivity of metal vacancies. If DH >> DVMe we have Deff ≈ 3DVMe. For the same reasons as for Relation (13), only the effective diffusivities are reported in Table 2. The diffusivity measured from H-D exchange in natural forsterite along c is lower than the effective diffusivity of hydrogen uptake in synthetic forsterite (Fig. 11). More data need to be collected to confirm this for synthetic forsterite and the other crystallographic direction to learn whether the assumption DH >> DVMe is valid. A summary of the effective diffusivities measured in San Carlos olivine is presented in Figure 12. H-D exchange experiments have not been reported in San Carlos olivine, but experience shows that when we try to do it at room pressure, the experiment fails due to the concurrent extraction of hydrogen following Reaction (11). Thus we expect that the redox-exchange reaction is limited by the diffusion of hydrogen DH, with Dh >> DH, at least for diffusion along the slowest directions, b and c (Fig. 12). Despite the conclusion of Mackwell and Kohlstedt (1990) suggesting that diffusion is anisotropic with different kinetics along b and c directions, it is not clear yet from the latest data of Kohlstedt and Mackwell (1998) that the diffusion rates of hydrogen along these two directions are really different (Fig. 12). The assumption that DH >> DVMe is justified for directions a and b but not for c; the diffusivity of hydrogen along c is comparable to the effective diffusivity of hydrogen uptake attributed to Reaction (12) (see Fig. 12). Metal vacancy diffusivities along a, and b directions are not expected to be significantly different considering the data of effective diffusivities plotted in Figure 12.

Diopside Extensive hydrogen extraction/incorporation experiments have been performed on natural diopsides with different iron contents (Ingrin et al. 1995; Hercule 1996; Hercule and Ingrin

Diﬀusion of Hydrogen in Minerals

311

T (K) 1423

1173

1073

973

873

-10

-11

Fo -1

2 log D (m s )

H-VMe

-12

// b // c // a

-13

-14

Fo

H-D

-15

-16 0.6

0.7

0.8

3

0.9

-1

1

1.1

1.2

10 /T (K ) Figure 11. Comparison of hydrogen uptake effective diffusivities in synthetic forsterite along the three crystallographic directions (Demouchy and Mackwell 2003) and H-D diffusivity along c in natural forsterite.

T (K) 1423

1173

1073

973

873

-9

Olivine

-1

2 log D (m s )

-10

H-h

// a

-11

Olivine

// c

H-VMe

-12

// c

// b

-13

// b // a

-14

-15 0.6

0.7

0.8

3

0.9

-1

1

1.1

1.2

10 /T (K ) Figure 12. Effective diffusivities of hydrogen uptake along the three crystallographic directions of San Carlos olivine from Kohlstedt and Mackwell (1998). Lines with data points are for redox-exchange reactions (open circles: // a; full circles: // b; triangles: // c) and lines without data points are for exchange involving metal vacancies.

312

Ingrin & Blanchard

1999; Carpenter-Wood et al. 2000). These results can be compared with the diffusivities determined from H-D exchange experiments (Fig. 13). For low iron content (iron molar fraction, XFe = 0.036) the effective diffusivity of H extraction and uptake is isotropic and much lower than hydrogen diffusivities. Effective diffusivity of hydrogen extraction increases with iron content and then shows the same anisotropy as hydrogen diffusivity (Da = Dc >> Db, Jaipur diopside with XFe = 0.07; Fig. 13). For the richest iron content (XFe = 0.07 and 0.126), the extraction diffusivities reach values comparable to hydrogen diffusivity. Studies of hydrogen incorporation in iron-rich synthetic diopside single crystals by infrared, Mössbauer- and optical spectroscopies have shown that hydrogen uptake is accompanied by iron reduction from Fe3+ to Fe2+ (Skogby 1994; Forneris and Skogby 2004), suggesting that hydrogen uptake is controlled by the redox-exchange Reaction (11) much as shown for olivine. It is thus possible to analyze the effective diffusivities of uptake in diopside from Equation (13) if we assume that Dh increases with iron content. However, the concentration of hydrogen point defects XH in Relation (13) is supposed to be close to the concentration of polarons Xh. The general solution is:

Deff =

( X H + X h ) DH Dh X H DH + X h Dh

Deff

or

⎛ Xh ⎞ ⎜1 + ⎟ DH Dh X H ⎠ =⎝ X DH + h Dh XH

(15)

T (K) 1423

1173

1073

973

873

-10

X = 0.07

Diopside

Fe

-10.5

X = 0.126 Fe

2 -1

log D (m s )

-11

X = 0.05 Fe

-11.5

X = 0.036 Fe

D // c, a* H

-12 -12.5

D // b H

-13 -13.5 -14 0.6

0.7

0.8

3

0.9

-1

1

1.1

1.2

10 /T (K ) Figure 13. Effective diffusivities of hydrogen extraction and uptake for diopside single crystals with different iron contents compared to hydrogen diffusivities determined from H-D exchange experiments. Solid lines: hydrogen diffusivities from Hercule and Ingrin (1999). Dashed lines and data points: extraction/ incorporation experiments (diamonds: Jaipur diopside, extraction // a and c; Carpenter-Wood et al. 2000; full triangles: Jaipur diopside, extraction // b; Carpenter-Wood et al. 2000; empty circles: Russian diopside, extraction/incorporation // a*, b and c; Ingrin et al. 1995; Hercule and Ingrin 1999; long dashed line: Baikal diopside, extraction // c; Hercule 1996; short dashed line: Malacacheta diopside, extraction // ≈ [212]; Hercule 1996).

Diﬀusion of Hydrogen in Minerals

313

The assumption XH = Xh can be discussed for olivine but it is certainly wrong for diopside. Figure 14 shows the results of the values of Dh deduced from Equation (15) assuming that DH is given by the diffusion laws obtained from H-D exchange and Xh is arbitrarily fixed equal to 10% of the iron content (Xh/XH is equal to 6, 17, 36 and 85 respectively for diopsides of increasing iron content). At very low XFe, (XFe = 0.036), DH >> XhDh/XH and Dh ≈ Deff/7 where Dh is isotropic. We assume that polaron diffusivity is isotropic for other iron concentrations too. For XFe = 0.05, along the c-direction, we still have DH >> XhDh/XH and Dh ≈ Deff/18. At XFe = 0.07, XhDh/XH is comparable to DH along c and a directions and is much greater than DH along b. Thus Dh is close to DH along b. For XFe = 0.126, the effective diffusivities were not measured along the main crystallographic axes but the data confirm that XhDh/XH >> DH; thus Deff ≈ DH (Fig. 14). The only data that reports uptake involving another reaction than the redox-exchange shows a diffusivity only slightly smaller than the diffusivity of polarons (XFe = 0.036; the square symbol in Fig. 14; Guillaumou et al. 1999; Ingrin and Skogby 2000).

Enstatite Stalder and Skogby (2003) published a study on hydrogen exchange in orthopyroxene. They show that the extraction of hydrogen in enstatite single crystals from a Kilbourne Hole xenolith is two orders of magnitude faster than the extraction from end-member synthetic enstatite at 973 K. In end-member enstatite the effective diffusivity is isotropic and the activation enthalpy is high (295 ± 55 kJ·mol−1; Table 2). Values of the effective diffusivity of extraction/incorporation in Kilbourne Hole enstatite are of the same order as the diffusivity of the iron-poor diopside. The low ferric iron content of this enstatite (Fe3+/Fetotal = 3.5%) suggests that the exchange is controlled by the mobility of polarons (DH >> XhDh/XH). In end-member enstatite, diffusion is slower probably because the exchange is controlled by Reaction (12) and the diffusivity of metal vacancies is much slower than hydrogen diffusivity

T (K) 1423

1173

1073

973

873

-10

Diopside

-10.5 0.07

D

-11

2 -1

log D (m s )

ha,b,c

-11.5

D // c, a* H

-12 -12.5

0.05

0.036

D

D

ha,b,c

hc

D // b H

-13 -13.5 -14 0.6

0.7

0.8

3

0.9

-1

1

1.1

1.2

10 /T (K ) Figure 14. Evolution of polaron diffusivities determined from equation (15), DhXFe, for various iron contents: grey lines. Full square: effective diffusivity of a reaction involving metal vacancies (Guillaumou et al. 1999; Ingrin and Skogby 2000). Other symbols are the same as in Figure 13.

314

Ingrin & Blanchard

(XhDh/XVMe >> DVMe). Stalder and Skogby (2003) report that a second reaction is involved in the uptake of hydrogen in Kilbourne Hole enstatite with the same effective diffusivity as for end-member enstatite suggesting that the same is true for mantle enstatite.

Garnets Water-related defects in garnet have been an important subject of study for decades but it is only recently that the kinetics of hydrogen extraction have been studied (Wang et al. 1996; Blanchard and Ingrin 2004b; Kurka 2005; Kurka et al. 2005). The activation enthalpy of hydrogen extraction in garnets is ~260 kJ·mol−1 (with a range of 180 – 329 kJ·mol−1), much higher than the enthalpies measured in olivine and diopside for redox-exchange reactions (107 – 181 kJ·mol−1; Table 2) and much higher than the average activation enthalpy of hydrogen diffusion determined from H-D exchange (130 kJ·mol−1; Figs. 7, 15). The extraction kinetics in mantle pyropes can be described by a single set of diffusion laws with an activation enthalpy of 253 kJ·mol−1 (Wang et al. 1996; Fig. 15). In this type of pyrope the extraction diffusivity is only slightly higher than the hydrogen diffusivity (Blanchard and Ingrin 2004a). For Dora Maira pyrope, which has a lower concentration of ferric iron than mantle pyrope, the extraction diffusivity is one to two orders of magnitude slower than for mantle pyropes. In that case, extraction diffusivities are comparable to or slightly lower than the hydrogen diffusivity (Fig. 15). Two general features of hydrogen diffusion in garnet, which were not observed in other silicates like olivine or pyroxenes, are the difference of hydrogen extraction kinetics between OH bands (groups of OH bands at 3650 and 3600 cm−1 in Dora Maira pyrope and 3620 and 3560 cm−1 in andradite; Table 2, Figs. 15, 16) and the diffusivity dependence on oxygen partial pressure (Blanchard and Ingrin 2004b; Kurka 2005; Fig. 15). In Dora Maira pyrope, the diffusivity decreases by two orders of magnitude for a pO2 drop of 15 to 20 orders (Fig. 15).

T (K) 1273

1173

1073

-9

logD (m2s-1)

-10

mantle pyropes

-11

-12

DH-D -13

Dora Maira pyrope

-14

0.75

0.8

0.85

OH3650 OH3600 0.9

0.95

103/T (K-1) Figure 15. Effective diffusivities of hydrogen extraction in pyrope. Solid lines: mantle pyropes (Wang et al. 1996); solid lines with data points: Dora Maira pyrope for two different groups of OH bands (OH3650, OH3600; in air: big symbols; in reducing atmosphere: small symbols; Blanchard and Ingrin 2004b); dashed line: average law of H-D exchange in garnet.

Diﬀusion of Hydrogen in Minerals

315

In calcium-rich garnets, the diffusivities of hydrogen extraction roughly increase with ferric iron content (Table 2, Fig. 16). The kinetics dependence on iron concentration, like for diopside, suggests that a redox-exchange reaction controls the hydrogen extraction. The implication of Reaction (11) is also confirmed by the simultaneous decrease of Fe2+ concentration measured in grossular by Mössbauer during its hydrogen extraction (sample Gr73And23Py2, Table 2; Forneris and Skogby 2004). However, complex kinetics and high activation enthalpy (≈260 kJ·mol−1) of hydrogen loss from garnet, close to the enthalpies found for reactions involving the migration of metal vacancies in olivine and enstatite, are difficult to interpret using Relations (13) and (15). In their experiments, Wang et al. (1996) could not fit the diffusion data of mantle garnets by a solution of Fick’s law with D independent of the concentration. Doremus (2002) explains this behavior, assuming that the mobile species is molecular water, reacting with the oxygen sublattice to produce OH species like in glass. However, this effect was not observed in other garnets; a diffusion profile in grossular is fitted with a constant diffusivity in Figure 2. Moreover, the departure from a constant diffusion coefficient can be explained without assuming concentration dependence. The hydrogen species giving rise to different OH bands of garnet have different kinetics, and they must be considered separately in order to determine a diffusion law (Blanchard and Ingrin 2004b; Kurka 2005). In addition, the spatial resolution of infrared profiles is highly dependent on sample thickness and its measure close to the edge of the crystal is sometimes tricky. A dependence of D with hydrogen concentration for a single OH band closely resembles a relation given by Equation (15) where the effective diffusivity is a function of the ratio Xh and XH, as Xh and XH vary all along the exchange. The analysis of

T (K) 1323

-10

1223

1123

And99 3620

1073

973

Ca-garnets

-11

logD (m2s-1)

DH-D

And23

And14 -12

-13

And99 3560 -14

-15 0.7

0.75

0.8

0.85

103

/T

0.9

0.95

1

1.05

(K-1)

Figure 16. Diffusivity of hydrogen extraction in calcium-rich garnets. Data from Kurka et al. (2005) and Kurka (2005): “Andxx” gives the concentration of the garnet in the andradite component, And99xxxx corresponds to the two types of OH bands present in the endmember andradite sample. Dashed line: average law of H-D exchange in garnet.

316

Ingrin & Blanchard

these results in terms of a specific reaction model cannot be done before the evolution during the exchange of both species Fe3+ and H is determined.

Quartz Very few diffusion experiments of extraction or incorporation of hydrogen have been done in quartz. Only the study of Kronenberg et al. (1986) reports hydrothermal experiments where the mobility of hydrogen, and not water, was involved during uptake. The diffusivity of uptake is close to the diffusivities for H-D exchange and the diffusion seems to be isotropic (Kats 1962; Kronenberg et al. 1986; Fig. 17). An experiment of tritium uptake under low water pressure by Shaffer (1974) gives values lower by a factor of 104 to 105 (Fig. 17). Shaffer noticed that the diffusivity increases with the decrease of water concentration in the sample. This behavior is contrary to that expected for water exchange; however, the values Shaffer reports are very close to data for water and oxygen diffusion in quartz (Cordier et al. 1988) suggesting that tritium uptake in Shaffer’s experiments was due to the mobility of water.

Feldspars To our knowledge, there is no experimental study providing a diffusion law for H-D exchange in feldspar and only two studies determining hydrogen extraction/incorporation diffusivities. Kronenberg et al. (1996) measured hydrogen diffusivity from extraction experiments in air from adularia feldspars that contain H2O structural defects. Despite a debate as to whether the mobile defect is really the hydrogen atom or the water molecule (see Doremus 1998; Kronenberg et al. 1998), the results are interesting because the measured kinetics are very fast and may provide a lower limit for H-D exchange kinetics in feldspar. These results are plotted in Figure 18 with the diffusion law for H-D exchange in quartz from Kats (1962); the two laws are very close. Kronenberg et al. (1996) performed one extraction experiment in a direction parallel to b suggesting that the diffusion is probably isotropic. They also performed a few uptake experiments at 1000 MPa H2O but no value of diffusivity was proposed.

T (K) 1273

-9

-10

1173

1073

973

Uptake Kronenberg et al. (1986)

H-D Kronenberg et al. (1986)

logD (m2s-1)

-11

H-D Kats (1962) -12

-13

-14

-15

-16 0.7

T uptake Shaffer (1974)

Quartz

0.75

0.8

0.85

103

0.9

/T

0.95

1

1.05

1.1

(K-1)

Figure 17. Diffusivity of hydrogen extraction in quartz. Empty circles are for hydrogen uptake experiments done parallel to c, full circle for direction perpendicular to c.

Diﬀusion of Hydrogen in Minerals

317

T (K) 1273

1173

1073

973

873

773

-10

Feldspars

logD (m2s-1)

-11

H-D Quartz Kats (1962) -12

K feldspar Kronenberg et al. (1996) -13

Plagioclase Johnson (2003)

-14

-15 0.7

0.8

0.9

1

103 /T

1.1

1.2

1.3

(K-1)

Figure 18. Diffusivity of hydrogen extraction in feldspars (symbols and solid lines) compared to the H-D exchange kinetics in quartz (dashed line).

Pegmatite minerals like adularia are H2O-bearing feldspars but plagioclases from volcanic origin contain essentially structural OH groups (Johnson and Rossman 2004). Extraction experiments in nitrogen at atmospheric pressure in andesine crystal by Johnson (2003) give much lower diffusivities and a higher activation enthalpy than are given by the data on adularia (Fig. 18). The activation term is close to values measured for garnet and the diffusivities compare favorably to the diffusivities measured in iron-poor grossular (Figs. 16, 18; Table 2). The iron concentration in andesine is three times lower than the hydrogen concentration. Thus a redoxexchange reaction like (11) cannot be the main mechanism of extraction (Johnson 2003).

CONCLUSION AND FUTURE DIRECTIONS We have deliberately not considered migration of water molecules in hydrous or anhydrous minerals; the review was mainly focused on hydrogen diffusion in anhydrous minerals. Like for oxides, there are very few minerals with absolutely no evidence of the presence of hydrogen, so it is possible that the review has omitted some few diffusion data. However, the amount of published diffusion data on anhydrous minerals is now large enough to underline some general behavior:

•

The values of hydrogen diffusivities determined from H-D exchanges in anhydrous minerals spread on 3 log units and the activation enthalpy is generally between 100 to 200 kJ·mol−1. The activation enthalpies for hydrous minerals are generally lower than those for anhydrous minerals (Fig. 10).

•

Fast hydrogen uptake is controlled by a redox-exchange reaction involving the Fe3+/Fe2+ couple (11) in most of minerals containing iron like olivine, pyroxenes and garnets. It is theoretically possible to determine the effective diffusivity of hydrogen exchange from the diffusivity of hydrogen and the diffusivity of polarons (15).

318

Ingrin & Blanchard •

In garnets and iron-free minerals the activation enthalpy of hydrogen uptake is frequently higher than 200 kJ·mol−1 (pure forsterite, pure enstatite, plagioclase). For slow hydrogen uptakes controlled by the diffusion of vacancies (12), the activation enthalpy is also higher than 200 kJ·mol−1.

•

Anisotropy of diffusion is important in rutile, olivine, diopside, epidote and amphibole, but surprisingly very low in quartz and feldspar.

The recent development of the infrared microscopy and techniques like ion and nuclear microprobes will increase the number of studies reporting hydrogen profiles in minerals (Demouchy 2004; Kurka 2005; Peslier et al. 2006). It will facilitate the study of the anisotropy of diffusion in single crystals and the systematic use of anisotropic data in modeling. Such profiles can be used to estimate the kinetics of natural processes like for instance the rate of magma ascent (Demouchy 2004; Peslier et al. 2006). However, there are still many experimental data missing even for well-studied minerals like olivine or diopside. We need these data in order to correctly analyze diffusion results and to build quantitative models of hydrogen uptake. For instance, there are no data for isotopic diffusion in olivine, orthopyroxenes and feldspars. The knowledge of the kinetics of H-D exchange is essential to identify the expression of the effective diffusivity and to validate simplified equations like (12) or (13). Experiments have demonstrated the importance of redox-exchange reactions in silicates but there are few if any studies that follow the Fe3+/Fe2+ change during hydrogen extraction/incorporation. Future works on hydrogen diffusion in anhydrous minerals will systematically incorporate studies of the iron oxidation during dehydrogenation. It is only with these new data that we will be able to understand the mechanisms of hydrogen uptake in major silicate minerals.

ACKNOWLEDGMENTS We thank Andreas Kurka, Juliette Forneris and Henrik Skogby for the communication of their most recent results. We are also grateful for helpful reviews from Andreas Kronenberg and Henrik Skogby. This study was supported by the EU, through the Human Potential Program HPRM-CT-2000-0056.

REFERENCES Bates JB, Wang JC, Perkins RA (1979) Mechanisms for hydrogen diffusion in TiO2. Phys Rev B 19:41304139 Blanchard M, Ingrin J (2004a) Kinetics of deuteration in pyrope. Eur J Mineral 16:567-576 Blanchard M, Ingrin J (2004b) Hydrogen diffusion in Dora Maira pyrope. Phys Chem Mineral 31:593-605 Bongiorno A, Colombo L, Cargnoni F (1997) Hydrogen diffusion in crystalline SiO2. Chem Phys Lett 264: 435-440 Brady JB (1993) Diffusion data for silicate minerals, glasses, and liquids. In: Handbook of Physical Constants. Vol 2. Arhens TH (ed) Am Geophysical Union, p 269-290 Bunson PE, Di Ventra M, Pantelides ST, Schrimpf RD, Galloway KF (1999) Ab initio calculations of H+ energetics in SiO2: Implications for transport. IEEE Trans Nuclear Sci 46:1568-1573 Carpenter-Wood S, Mackwell SJ, Dyar D (2000) Hydrogen in diopside: Diffusion profiles. Am Mineral 85: 480-487 Carslaw HS, Jaeger JC (1959) Conduction of Heat in Solids. Oxford University Press Cathcart JV, Perkins RA, Bates JB, Manley LC (1979) Tritium diffusion in rutile (TiO2). J Appl Phys 50:41104119 Chacko T, Riciputi LR, Cole DR, Horita J (1999) A new technique for determining equilibrium hydrogen isotope fractionation factors using the ion microprobe: Application to the epidote-water system. Geochim Cosmochim Acta 63:1-10 Cordier P, Boulogne B, Doukhan JC (1988) Water precipitation and diffusion in wet quartz and wet berlinite AlPO4. Bull Minéral 111:113-137 Crank J (1975) The Mathematics of Diffusion. Oxford University Press

Diﬀusion of Hydrogen in Minerals

319

Demouchy S (2004) Water in the Earth’s interior: thermodynamics and kinetics of hydrogen incorporation in olivine and wadsleyite. PhD Dissertation, Universität Bayreuth, Germany Demouchy S, Mackwell SJ (2003) Water diffusion in synthetic iron-free forsterite. Phys Chem Mineral 30: 486-494 Dersch O, Rauch F (1999) Water uptake of quartz investigated by means of ion-beam analysis. Fres J Anal Chem 365:114-116 Dersch O, Zouine A, Rauch F, Ericson JE (1997) Investigation of water diffusion into quartz using ion beam analysis techniques. Fres J Anal Chem 358:217-219 Doremus (1998) Comment on “Stationary and mobile hydrogen defects in potassium feldspar” by AK Kronenberg, RA Yund and GR Rossman. Geochim Cosmochim Acta 62:377-378 Doremus RH (2002) Diffusion of Reactive Molecules in Solids and Melts. Wiley Forneris JF, Skogby H (2004) Is hydrogen loss via iron oxidation an important mechanism in nominally anhydrous minerals? Goldschmidt J Conf Abstr 1.1.23, A34 Graham CM (1981) Experimental hydrogen isotope studies III: Diffusion of hydrogen in hydrous minerals, and stable isotope exchange in metamorphic rocks. Contrib Mineral Petrol 76:216-228 Graham CM, Harmon RS, Sheppard SMF (1984) Experimental hydrogen isotope studies: hydrogen isotope exchange between amphibole and water. Am Mineral 69:128-138 Graham CM, Sheppard SMF, Heaton THE (1980) Experimental hydrogen isotope studies-I. Systematics of hydrogen isotope fractionation in the systems epidote-H2O, zoisite-H2O and AlO(OH)-H2O. Geochim Cosmochim Acta 44:353-364 Graham CM, Viglino JA, Harmon RS (1987) Experimental study of hydrogen-isotope exchange between aluminous chlorite and water and hydrogen diffusion in chlorite. Am Mineral 72:566-579 Guillaumou N, Dumas P, Ingrin J, Carr GL, Williams JP (1999) Microanalysis of fluids in minerals in the micron scale range by synchrotron infrared microspectrometry. Internet J Vibrational Spec [www.ijvs.com] 3: 1-11 Hercule (1996) Cinétique et solubilité de l’hydrogène dans le diopside monocrystallin. PhD Dissertation, University paris XI, Orsay, France Hercule S, Ingrin J (1999) Hydrogen in diopside: Diffusion, kinetics of extraction-incorporation, and mobility. Am Mineral 84:1577-1587 Hertweck B, Ingrin J (2005a) Hydrogen incorporation in a ringwoodite analogue: Mg2GeO4 spinel. Mineral Mag 69:335-341 Hertweck B, Ingrin J (2005b) Hydrogen incorporation in ringwoodite analogue: Mg2GeO4 spinel. Geophys Res Abstr, EGU 7:2690 Ingrin J, Blanchard M (2000) Hydrogen mobility in single crystal kaersutite. EMPG VIII, J Confe Abstr 5:52 Ingrin J, Hercule S, Charton T (1995) Diffusion of hydrogen in diopside: Results of dehydrogenation experiments. J Geophys Res 100:15489-15499 Ingrin J, Latrous K, Doukhan JC, Doukhan N (1989) Water in diopside: an electron microscopy and infrared spectroscopy study. Eur J Mineral 1:327-341 Ingrin J, Pacaud L, Jaoul O (2001) Anisotopy of oxygen diffusion in diopside. Earth Planet Sci Lett 92:347361 Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Mineral 12:543-570 Jibao J, Yaqian Q (1997) Hydrogen isotope fractionation and hydrogen diffusion in the tourmaline-water system. Geochim Cosmochim Acta 61:4679-4688 Johnson EA (2003) Hydrogen in nominally anhydrous crustal minerals. PhD Dissertation, California Institute of Technology, Pasadena, USA Johnson EA, Rossman GR (2004) A survey of hydrous species and concentrations in igneous feldspars. Am Mineral 89:586-600 Johnson OW, Paek SH, Deford JW (1975) Diffusion of H and D in TiO2: Suppression of internal fields by isotope exchange. J Appl Phys 46:1026-1033 Kats A, Haven Y, Stevels JM (1962) Hydroxyl groups in α-quartz. Phys Chem Glasses 3:69-76 Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kronenberg AK, Kirby SH, Aines RD, Rossman GR (1986) Solubility and diffusional uptake of hydrogen in quartz at high water pressure: implications for hydrolytic weakening. J Geophys Res 91:12723-12744 Kronenberg AK, Yund RA, Rossman GR (1996) Stationary and mobile hydrogen defects in potassium feldspar. Geochim Cosmochim Acta 60:4075-4094 Kronenberg AK, Yund RA, Rossman GR (1998) Reply to the comment by Robert H. Doremus on “Stationary and mobile hydrogen defects in potassium feldspar.” Geochim Cosmochim Acta 62:379-382 Kurka A (2005) Hydrogen in Ca-rich garnets: diffusion and stability of OH-defects. PhD Dissertation, University Paul Sabatier, Toulouse, France

320

Ingrin & Blanchard

Kurka A, Blanchard M, Ingrin J (2005) Kinetics of hydrogen extraction and deuteration in grossular. Mineral Mag 69:359-371 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52 Libowitzky E, Rossman GR (1996a) Principles of quantitative absorbance measurements in anisotropic crystals. Phys Chem Mineral 23:319-327 Libowitzky E, Rossman GR (1996b) FTIR spectroscopy of lawsonite between 82 and 325K. Am Mineral 81: 1080-1091 Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: Implications for water in the mantel. J Geophys Res 95:5079-5088 Marion S, Meyer H-W, Carpenter M, Norby T (2001) H2O-D2O exchange in lawsonite. Am Mineral 86:11661169 Norby T, Larring Y (1997) Concentration and transport of protons in oxides. Curr Opin Solid State Mater Sci 2:593-599 Peslier AH, Luhr JF (2006) Hydrogen loss from olivines in mantle xenoliths from Simcoe (USA) and Mexico: Mafic alkalic magma ascent rates and water budget of the sub-continental lithosphere. Earth Planet Sci Lett 242:302-319 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Shaffer EW, Sang SL, Cooper AR, Heuer AH (1974) Diffusion of tritiated water in β-quartz. In: Geochemical Kinetics and Transport. Hofmann AW, Giletti BJ, Yoder H, Yund RA (eds) Carnegie Inst Wash Publ 634: 131-138 Skogby H (1994) OH incorporation in synthetic clinopyroxene. Am Mineral 79:240-249 Skogby H, Rossman GR (1989) OH− in pyroxene: An experimental study of incorporation mechanisms and stability. Am Mineral 74:1059-1069 Stalder R, Skogby H (2003) Hydrogen diffusion in natural and synthetic orthopyroxene. Phys Chem Mineral 30:12-19 Suman KD, Cole DR, Riciputi LR, Chacko T, Horita J (2000) Experimental determination of hydrogen diffusion rates in hydrous minerals using the ion microprobe. J Conf Abstr 5(2):340 Suzuoki T, Epstein S (1976) Hydrogen isotope fractionation between OH-bearing minerals and water. Geochim Cosmochim Acta 40:1229-1240 Vennemann TW, O’Neil JR, Deloule E, Chaussidon M (1996) Mechanism of hydrogen exchange between hydrous minerals and molecular hydrogen: Ion microprobe study of D/H exchange and calculations of hydrogen self-diffusion rates. Goldschmidt. J Conf Abstr 1(1):648 Wang L, Zhang Y, Essene E (1996) Diffusion of the hydrous component in pyrope. Am Mineral 81:706-718 Wegdén M, Kristiansson P, Skogby H, Auzelyte V, Elfman M, Malmqvist KG, Nilsson, Pallon J, Shariff A (2005) Hydrogen depth profiling by p-p scattering in nominally anhydrous minerals. Nucl Instr Methods Phys Res B 231:524-529 Yaqian Q, Jibao G (1993) Study of hydrogen isotope equilibrium and kinetic fractionation in the ilvaite-water system. Geochim Cosmochim Acta 57:3073-3082 York D (1966) Least-squares fitting of a straight line. Can J Phys 44:1079-1086

14

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 321-342, 2006 Copyright © Mineralogical Society of America

Effect of Water on the Equation of State of Nominally Anhydrous Minerals Steven D. Jacobsen Department of Geological Sciences Northwestern University Evanston, Illinois, 60208, U.S.A. e-mail: [email protected]

INTRODUCTION It is possible that the majority of Earth’s H2O budget is present as hydroxyl (OH) structurally incorporated into the major nominally anhydrous minerals (NAMs) of the mantle (e.g., Martin and Donnay 1972). Ringwood (1966) thought as much as five times the surface H2O-mass could be present in the mantle, amounting to ~0.2 wt% H2O if distributed throughout the entire mantle (Harris and Middlemost 1969). We know now the (Mg,Fe)2SiO4 polymorphs of the upper mantle and transition zone can incorporate up to several weight percent of water in their structures (e.g., Smyth 1987; Inoue et al. 1995; Kohlstedt et al. 1996; Bolfan-Casanvoa et al. 2000; Mosenfelder et al. 2006). The possibility of a deep-Earth water cycle leads naturally to the question, is “water” in the mantle, whether regional or globally distributed, detectible seismically? In order to address this question, it is necessary to know, quantitatively, the effects of water (or more precisely structurally bound hydroxyl) on the elastic moduli of mantle minerals. Also required are pressure and temperature derivatives of the elastic moduli for more direct comparison with seismological observation. This chapter will review what is known about the elastic properties of “hydroxylated” NAMs from experimental studies. From a crystal chemical perspective, hydrated NAMs are defect structures because hydrogen is usually incorporated through charge balance by cation vacancies. Therefore, small variations in water content can have a dramatic effect on thermoelastic parameters, more so than any other major geochemical substitution such as iron or aluminum. For example, at one atmosphere the addition of ~1 wt% H2O into ringwoodite has a similar effect on the shear modulus as raising the temperature by 8001000 °C (Wang et al. 2003a; Jacobsen et al. 2004). However, due to elevated pressure derivatives, the difference between anhydrous and hydrous velocities diminishes at higher pressures. In this chapter, elasticity-water systematics will be evaluated in order to make some predictions about the elastic properties of major phases for which shear velocities are not yet available. Seismologists require accurate thermoelastic parameters of hydrated NAMs in order to evaluate velocity anomalies in potentially hydrous regions of the mantle. However, many of the thermoelastic properties of hydrated NAMs have not yet been measured. For example, pressure derivatives of the bulk and shear moduli are needed to estimate both compressional and shear wave velocities (vP and vS) at high pressure, but are currently available only for hydrous ringwoodite (Wang et al. 2003b; Jacobsen and Smyth 2006). Furthermore, the effects of OH on temperature derivatives of the elastic moduli are unknown for any of the major NAMs. Thus, in order to obtain a broader picture of the effects of water on elastic properties, some stoichiometrically hydrous phase will also be considered, such as humites and selected dense-hydrous magnesium silicates (DHMS). Finally, using a model set of thermoelastic parameters, monomineralic velocities are calculated for hydrous (Mg,Fe)2SiO4 polymorphs along a mantle adiabat for simple comparison with global seismic velocity models. 1529-6466/06/0062-0014$05.00

DOI: 10.2138/rmg.2006.62.14

322

Jacobsen

The chapter is divided into sections based on mineral groups stable within the upper mantle and transition zone. The DHMS are treated separately. There are no elasticity data for OH-bearing silicate perovskite or magnesiowüstite in the lower mantle. Within each section the reader will find different types of equation of state data. Pressure-volume (P-V) studies with X-ray or neutron diffraction techniques are static in nature, providing variation of density (and sometimes crystal structure) with pressure, ρ(P). P-V data are usually fit to a third-order Birch-Murnaghan equation of state (e.g., Angel et al. 2000), parameterized in terms of initial volume (V0), isothermal bulk modulus, KT = −V(dP/dV)T, and first pressure derivative K′ = dKT/dP. With a third-order truncation, the second derivative of KT is implied, but probably plays an important role in the variation of density of hydrated NAMs with pressure due to the initial high compressibility of cation vacancies. It is possible that hydrated NAMs become as dense, or even denser than their anhydrous end members due to elevated pressure derivatives, which may also be the case for hydrous melts (Agee 2005; Matsukage et al. 2005; Sakamaki et al. 2006). Volume-temperature studies with X-rays or neutrons provide coefficients of thermal expansion, such as the volume thermal expansivity, αV = V−1(dV/dT)P, needed to calculate mineral density at high temperature. One problem facing thermal expansion studies of hydrated mantle NAMs is the relatively low temperature (~500 °C) that these phases dehydrate or decompose when not under confining pressure (Inoue et al. 2004). To date, data for OHbearing NAMs is restricted to P-V or V-T studies, which means the temperature derivatives of hydrated NAMs at high pressure are not known. Dynamic studies of mineral elastic properties are effectively one derivative ahead of static compression. These include Brillouin spectroscopy, resonant ultrasound spectroscopy, and ultrasonic interferometry (and various adaptations of each technique). With these methods, elastic moduli are obtained from measured sound velocities, resonance frequencies, or elastic wave travel times, and at ambient pressure both the adiabatic bulk modulus, KS = −V(dP/dV)S, and shear modulus (G, sometimes written µ) are obtained. The compressional (P) and shear (S) velocities and moduli are related through the equations: vP =

K S + ( 4 / 3)G ρ

(1)

G ρ

(2)

vS =

High-pressure Brillouin or ultrasonic studies determine KS and G at each pressure, so pressure derivatives of the moduli (KS′ and G′) are usually very accurate in comparison with static compression studies where KT′ is a second-order fitting parameter. Similarly, resonance techniques at elevated temperature determine the moduli at each temperature, resulting in very accurate temperature derivatives.

ELASTIC PROPERTIES OF NOMINALLY ANHYDROUS MINERALS IN THE UPPER MANTLE Olivine Olivine, α-(Mg,Fe)2SiO4, is the most abundant mineral in the upper mantle. An understanding of the effects of water on the elastic properties of olivine is therefore of primary importance. Natural olivines from mantle xenoliths exhibit a wide range of water contents from <1 to several hundred ppm wt. H2O (e.g., Miller et al. 1987; Kitamura et al. 1987; Bell and Rossman 1992; Rossman 1996). More recent studies of synthetic olivine samples demonstrate

Eﬀect of Water on Equations of State

323

that water concentrations in olivine can reach 0.6-0.8 wt% H2O at conditions of 12-14 GPa and 1200 °C (Chen et al. 2002; Smyth et al. 2005a; Mosenfelder et al. 2006). It is possible the relatively low water content observed in natural olivine is due to water loss during ascent (Ingrin and Skogby 2000; Demouchy et al. 2006), but given the potential for there to be thousands of ppm wt. H2O in mantle olivine, elastic properties measurements are highly warranted. Relatively little is known about the effects of water on the elastic properties of olivine. The bulk modulus of single-crystal Fo95-olivine containing 0.8 wt% H2O was measured by X-ray diffraction (static compression) to 8 GPa (Smyth et al. 2005a), showing KT0 = 120(2) GPa, with K′ = 7.0(6). For comparison, the bulk modulus of anhydrous olivine is 127-130 GPa with K′ = 44.5 (Table 1). Thus, the addition of ~1 wt% H2O reduces K by about 6% and elevates K′. As will be shown through additional examples in this chapter, reduced K and elevated K′ is a common feature of water-bearing NAMs in minerals. The shear modulus of hydrous olivine has not been measured, so the effect of water on compressional and shear-wave velocities remains uncertain, but it is hopeful that future Brillouin or ultrasonic studies will resolve this outstanding issue.

Humite-group minerals along the forsterite-brucite join In order to gain a broader view on the effect of water on olivine, it is necessary to consider stoichiometrically hydrous minerals occurring along the olivine-brucite join of the MgOSiO2-H2O ternary. The hydroxyl and Mg-end member humite group minerals are expressed as n[Mg2SiO4]∙Mg(OH2), where n = 1, 2, 3, 4 for norbergite, chondrodite, humite, and clinohumite, respectively, and pure Mg(OH)2 is the mineral brucite. Phase A, the only DHMS to occur along the forsterite-brucite join, will be discussed separately in the section on DHMS. Pure hydroxylclinohumite Mg9Si4O16(OH)2 contains 2.9 wt% H2O and has a bulk modulus of about 120(1) GPa with K′ = 4.8(2), measured by static compression (Ross and Crichton 2001). A natural Fe and Fluorine-bearing clinohumite from Mount Somma, Vesuvius, Italy, was studied by Brillouin spectroscopy and has a bulk modulus of 125(2) GPa, and shear modulus of 73(5) GPa (Fritzel and Bass 1997). End-member hydroxylchondrodite, Mg5Si2O8(OH)2, contains 5.3 wt% H2O and has a bulk modulus of 116(1) GPa with K′ = 4.9(2) by static compression (Ross and Crichton 2001). A natural Fe- and Fluorine-bearing hydroxylchondrodite from the Tilley Foster Mine, Brewster, NY, was studied by Brillouin spectroscopy and has a bulk modulus of 118(2) GPa and a shear modulus of 75.6(7) GPa (Sinogeikin and Bass 1999). The MgO-H2O end-member, brucite, has the formula Mg(OH)2. A single-crystal X-ray diffraction study of brucite to 15 GPa shows KT0 = 42(2) GPa with K′ = 5.7(5) (Duffy et al. 1995a). Brucite has also been studied by high-pressure Brillouin spectroscopy to 14 GPa, giving KS0 = 43.8(8) GPa and G0 = 35.2(3), with pressure derivatives of 6.8(2) and 3.4(1), respectively (Jiang et al. 2006). Elastic properties of minerals along the olivine-brucite join are summarized in Table 1. In agreement with the observation for olivine that K is reduced and K′ is elevated upon hydration (Smyth et al. 2005a), consideration of phases along the entire olivine-brucite join verifies that the trend of decreasing K and increasing K′ continues, with brucite having the lowest K and G and highest K′ and G′. Elasticity-water systematics along the olivine-brucite join are plotted in Figures 1 and 2.

Garnet Hydrogen enters the garnet structure through the hydrogarnet substitution of H4O4 = SiO4 (e.g., Lager et al. 2005), effectively replacing the silicate tetrahedron with a somewhat larger H4O4 tetrahedron. Natural Mg-silicate garnets from the upper mantle below South African show typically 5-10 ppm wt. H2O, and occasionally 50-100 ppm (Bell and Rossman 1992). While the hydrogarnet substitution is documented in Mg-rich garnets in greater amounts, such as pyrope from Kaalvallei, South Africa, with ~200 ppm wt. H2O (Rossman 1996), hydration is more favorable in Ca-rich garnets andradite (Ca3Fe2Si3O12) and grossular (Ca3Al2Si3O12).

2.38 2.38

3.07 3.23 3.23

3.14 3.26

3.24

3.23 3.23 3.34 3.34

ρSTP (g/cm3)

30.9 30.9

5.3 ~2.5 ~3.5

2.9 ~2

0.8

0 0 0

H2O (wt%)

42(2) 43.8(8)

115.7(8) 117.0(4) 118.4(16)

119.4(7) 125(2)

120(2)

128.8(5) 127(4) 130.3(4) 129.9(6)

K (GPa)

5.7(5) 6.8(2)

4.9(2) 5.6(1)

4.8(2)

7.0(6)

4.2(2) 4.2(8) 4.61(11) 4.0 fixed

K′ (dK/dP)

35.2(3)

75.6(7)

3.4(1)

1.61(4)

77.4(2)

73(5)

1.4(1)

G′ (dG/dP)

81.6(2)

G (GPa)

6.17

8.24

8.26

8.36

8.58

vP (km/s)

3.85

4.84

4.73

4.81

5.03

vS (km/s)

XRD BS

XRD XRD BS

XRD BS

XRD

BS XRD US XRD

method*

14 14

8 10 0

8 0

8

16 10 8 32

Pmax (GPa)

[10] [11]

[6] [8] [9]

[6] [7]

[5]

[1] [2] [3] [4]

Ref.

*BS: Brillouin spectroscopy, US: Ultrasonic interferometry, XRD: X-ray diffraction References: [1] Zha et al. 1996; [2] Zhang 1998; [3] Liu et al. 2005; [4] Zha et al. 1998 [5] Smyth et al. 2005; [6] Ross and Crichton 2001; [7] Fritzel and Bass 1997; [8] Friedrich et al. 2002 [9] Sinogeikin and Bass 1999; [10] Duffy et al. 1995a; [11] Jiang et al. 2006

Mg(OH)2 Mg(OH)2

brucite

Mg5Si2O8(OH)2 Mg4.7Fe0.3Si2O8(F0.58OH0.42)2 Mg4.7Fe0.3Si2O8(F0.31OH0.67)2

OH-chondrodite

Mg9Si4O16(OH)2 Mg8.2Fe0.6Si4O16(F0.29OH0.71)2

OH-clinohumite

Mg1.84Fe0.09SiH0.127O4

hydrous olivine

Mg2SiO4 Mg2SiO4 (Mg0.9Fe0.1)2SiO4 (Mg0.9Fe0.1)2SiO4

anhydrous olivine

phase

Table 1. Elastic properties of minerals along the olivine-brucite join [nMg2SiO4·Mg(OH)2].

324 Jacobsen

Eﬀect of Water on Equations of State

325

Figure 1. Plot of the bulk modulus against water-content from various studies of nominally anhydrous and hydrous minerals. See tables for references.

Figure 2. Plot of the shear modulus against water content from various studies of nominally anhydrous and hydrous minerals. See tables for references.

326

Jacobsen

Natural grossular containing ~1 wt% H2O and andradite containing ~6 wt% H2O have been reported (Rossman 1996). The petrologic significance of silicate garnet in the upper mantle warrants examination of the affects of water on the compressibility of garnet.

Grossular-hydrogrossular The hydrogarnet substitution in grossular is expressed as Ca3Al2(SiO4)3−x(H4O4)x, where hydrogrosular with 0 < x < 1.5 is called hibschite, and for 1.5 < x ≤ 3.0 the mineralogical name is katoite. The elastic constants of a natural hibschite containing about 11 wt% H2O from Crestmore, California, have been determined by Brillouin spectroscopy (O’Neill et al. 1993). The single-crystal elastic constants of this hibschite are C11 = 186.5(11) GPa, C44 = 63.9(5) GPa, and C12 = 56.5(14) GPa, with KS0 = 99.8(10) GPa and G0 = 64.3(5) GPa. In comparison, pure grossular (Bass 1989) has C11 = 321.7(8) GPa, C44 = 104.6(4) GPa, C12 = 91.4(9) GPa, KS0 = 168.4(7) GPa and G0 = 108.9(4) GPa. Thus, in going from grossular to hibschite, the Cij and aggregate moduli are reduced equally by ~40%. The compressibility of katoite has been studied by powder X-ray (Olijnyk et al. 1991) and powder neutron diffraction (Lager and Von Dreele 1996), giving KT0 = 66(4) GPa and 52(1) GPa, respectively. Single-crystal X-ray diffraction experiments to 8 GPa (Lager et al. 2002) show a phase transition from Ia3d to I43d symmetry at ~5 GPa. Below 5 GPa, the bulk modulus of katoite was determined to be 58(1) GPa with K′ = 4.0(7) (Lager et al. 2002). The increased compressibility of hydrogarnets with increasing water is attributed to the positive volume change upon introduction of H4O4 tetrahedron compared with the smaller and more rigid silicate tetrahedron. Although the larger volume of the H4O4 tetrahedron has been used as an argument against H4O4 = SiO4 as a favorable substitution at high pressure, natural hydrogarnets equilibrated at 180-km depth are known (O’Neill et al. 1993). Spectroscopic studies of hibschite to 25 GPa indicate the absence of a phase transition or glass transition and thus hydrogarnet is stable throughout the pressure range of the upper mantle (Knittle et al. 1992). Tetragonal majorite garnets (MgSiO3) synthesized at 17.5 GPa and 1500 °C show water contents of 600-700 ppm wt. H2O (Bolfan-Casanova et al. 2000), although the compressibility of OH-bearing majorite has not been measured. Elastic properties of garnets along the grossular-hydrogrossular join are provided in Table 2 and plotted in Figures 1 and 2.

Pyroxene After olivine, pyroxene is the next most abundant mineral in the upper mantle, and among NAMs they tend to have the highest OH-concentrations. Natural orthoenstatite (Mg,Fe)SiO3 of mantle origin commonly contains several hundred ppm wt. H2O (Bell and Rossman 1992; Skogby et al. 1990). From synthesis experiments at 1100 °C, the water content of pure-Mg enstatite was found to increase with pressure from about ~150 ppm at 1 GPa to ~700 ppm at 10 GPa (Rauch and Keppler 2002). At 1.5 GPa, the water content increases slightly with temperature from ~50 ppm at 700 °C, to ~150 ppm at 1100 °C (Mierdel and Keppler 2004). At mantle temperatures and above ~7 GPa, both orthoenstatite (Pbca) and low-clinoenstatite (P21/c) adopt the high-clinoenstatite (C2/c) structure (Angel et al. 1992), and there appear to be some differences in the IR-spectra of enstatites quenched from the high-clino stability field (Bromiley and Bromiley 2006). Enstatite synthesized at 15 GPa and 1500 °C contains a maximum of about 500 ppm of water (Bolfan-Casanova et al. 2000). The concentration of water in enstatite also increases with aluminum content (Stalder and Skogby 2002). Despite extensive works on the incorporation of water into enstatite, no previous studies on the elastic properties of OH-bearing enstatite pyroxenes could be found. Diopside (CaMgSi2O6) and jadeite (NaAlSi2O6) form a solid solution referred to as omphacite. Natural omphacites are known to contain up to ~2000 ppm H2O (Smyth et al. 1991). No systematic studies were found on the effects of OH on omphacite compressibility. However, McCormick et al. (1989) measured the compressibility of two omphacites, both having Jd58

2.54 2.54 2.54

3.03

3.60 3.60 3.60 3.60

ρSTP (g/cm3)

29 29 29

11

0 0 0 0

H2O (wt%)

66(4) 52(1) 58(1)

99.8(1)

168(25) 169.3(12)

168.4(7)

K (GPa)

4.1(5) 4.0 fixed 4.0(7)

5.46 6.2(4) 5.92(14)

K′ (dK/dP)

64.3(5)

108.9(4)

G (GPa)

1.1

G′ (dG/dP)

7.83

9.33

vP (km/s)

4.61

5.50

vS (km/s)

XRD ND XRD

BS

BS BS XRD XRD

method*

42 9 5

0

0 10 18 37

Pmax (GPa)

*BS: Brillouin spectroscopy, US: Ultrasonic interferometry, XRD: X-ray diffraction, ND: Neutron diffraction References: [1] Bass 1989; [2] Conrad et al. 1999; [3] Olijnyk et al. 1991; [4] Pavese et al. 2001; [5] O’Neill et al. 2993; [6] Lager and Von Dreele 1996; [7] Lager et al. 2002

Ca3Al2(H4O4)3 Ca3Al2(H4O4)3 Ca3Al2(H4O4)3

hydrogrossular: katoite

Ca3Al2(SiO4)1.72(H4O4)1.28

hydrogrossular: hibschite

Ca3Al2Si3O12 Ca3Al2Si3O12 Ca3Al2Si3O12 Ca3Al2Si3O12

grossular

phase

Table 2. Elastic properties of hydrous garnets.

[3] [6] [7]

[5]

[1] [2] [3] [4]

Ref.

Eﬀect of Water on Equations of State 327

328

Jacobsen

composition (58% jadite, 42% diopside), but one showing about 2% M2-site vacancies and the other having about 7% M2-site vacancies. The high vacancy omphacite is attributed to increasing Ca-Eskola component (CaAl2Si4O12). The bulk modulus of pure diopside, vacancypoor omphacite, and the vacancy-rich omphacite were reported to be 75, 70, and 65 GPa, respectively. Although the samples were not analyzed for water, it is possible the vacancy-rich ompacite from the study of McCormick et al. (1989) contained significant amounts of water added as HAlSi2O6, which is similar to the Ca-Eskola component, Ca0.50.5AlSi2O6. The measurable difference in compressibility between low- and high-vacancy omphacites suggests that a compressibility study of OH-bearing pyroxene should be carried out.

ELASTIC PROPERTIES OF NOMINALLY ANHYDROUS MINERALS IN THE TRANSITION ZONE Wadsleyite The high-pressure polymorphs of olivine, wadsleyite β-(Mg,Fe)2SiO4 and ringwoodite γ(Mg,Fe)2SiO4, together have the potential to store more water as hydroxyl in the transition zone alone than is present in the hydrosphere. These phases can store at least several times the maximum amount in nominally anhydrous upper mantle minerals. Smyth (1987) recognized that wadsleyite, which contains Si2O7 sorosilicate groups, has an oxygen site (O1) that is not bonded to Si. This so-called non-silicate oxygen is highly underbonded, with a Pauling bond strength sum about 1/3 hydroxyl and 2/3 oxygen in character. Smyth (1987) predicted that wadsleyite, therefore, should contain variable amounts of water, with 3.3 wt% being the theoretical maximum if every O1 were hydroxyl. Soon after, IR studies of wadsleyite samples from high-pressure studies showed strong absorbance in the IR at ~3300 cm−1 and the presence of structurally bound hydroxyl in wadsleyite was confirmed (McMillan et al. 1991; Young et al. 1993). Inoue et al. (1995) synthesized pure-Mg wadsleyite containing 3.1 wt% H2O, measured by secondary-ion mass spectrometry (SIMS), and samples containing up to 2.5 wt% according to both SIMS and IR are routinely reported (e.g., Kohlstedt et al. 1996; Smyth et al. 1997; Kudoh and Inoue 1999). Attempts to synthesize nominally anhydrous wadsleyite without any H through careful firing and handling of starting materials and the high-pressure assembly resulted in no less than about 50 ppm of H2O in pure-Mg wadsleyite (Jacobsen et al. 2005), so it is conceivable that all wadsleyite samples in elasticity experiments of nominally anhydrous samples contain some amount of water. Although polarized infrared spectra are now available (Jacobsen et al. 2005), the exact location of hydrogen in wadsleyite remain somewhat uncertain due to positional disorder (Kohn et al. 2002). The bulk modulus of nominally anhydrous pure-Mg and Fo90 wadsleyite is about 170 GPa, with a shear modulus of about 112 GPa for pure-Mg wadsleyite and 107 GPa for Fo90 wadsleyite (Table 1). The bulk modulus of hydrous wadsleyite containing 2.5 wt% H2O (SIMS) was measured by X-ray diffraction, resulting in KT0 = 155(2) GPa with K′ fixed at 4.3 (Yusa and Inoue 1997). Thus, hydration to 2.5 wt% lowers the bulk modulus of wadsleyite by almost 10%. Also, the density is reduced by about 3% at this level of hydration. An intermediate sample containing ~1 wt% was also measured by static compression and shows KT0 = 162(2) GPa when K′ is fixed at 4.3, and KT0 = 152 GPa when K′ is allowed to refine with K′ = 6(1) (Holl et al. 2003). Thus, in wadsleyite the trends are consistent with the reduced K and elevated K′ seen along the olivine-humite join. Wadsleyite elastic properties are given in Table 3 and plotted in Figure 1. The volume thermal expansivity (αV) of pure-Mg hydrous wadsleyite was recently measured at room pressure to its dehydration temperature of ~450 °C, resulting in αV = 30(1) × 10−6 K−1 (Inoue et al. 2004), being slightly lower than αV for anhydrous wadsleyite with αV = 34.0(5) × 10−6 K−1 measured to 700 °C in the same study (Inoue et al. 2004), but about the same as the

3.51 3.49

3.41 3.41 3.36

3.47 3.47 3.57 3.60

ρSTP (g/cm3)

2.1 2.8

~1 ~1 2.5

0 0 0 0

H2O (wt%)

151(6) 145.6(2.8)

152(3) 162(2) 155(2)

170(2) 170 170(3) 172(2)

K (GPa)

6(2.5) 6.1(7)

6(1) 4.3 fixed 4.3 fixed

4.6(1)

4.3(2) 4.24(10)

K′ (dK/dP) 115(2) 108 108(2) 106(1)

G (GPa)

1.5(1)

1.4(2) 1.49(3)

G′ (dG/dP) 9.65 9.51 9.51 9.50

vP (km/s) 5.76 5.58 5.58 5.53

vS (km/s)

XRD XRD

XRD XRD XRD

BS US BS US

method*

5 10

9 9 8.5

14 12.5 0 9.6

Pmax (GPa)

*BS: Brillouin spectroscopy, US: Ultrasonic interferometry, XRD: X-ray diffraction References: [1] Zha et al. 1997; [2] Li et al. 1996; [3] Sinogeikin et al. 1998; [4] Li and Liebermann 2000; [5] Holl et al. 2003; [6] Yusa and Inoue 1997; [7] Smyth et al. 2005

Mg1.71Fe0.18Al0.01Si0.96H0.33O4 Mg1.60Fe0.22Al0.01Si0.97H0.44O4

wadsleyite-II

β-Mg1.9SiH0.2O4 β-Mg1.9SiH0.2O4 β-Mg1.8SiH0.4O4

hydrous wadsleyite

β-Mg2SiO4 β-Mg2SiO4 β-(Mg0.92Fe0.08)2SiO4 β-(Mg0.88Fe0.12)2SiO4

anhydrous wadsleyite

phase

Table 3. Elastic properties of wadsleyite.

[7] [7]

[5] [5] [6]

[1] [2] [3] [4]

Ref.

Eﬀect of Water on Equations of State 329

330

Jacobsen

anhydrous value of αV = 29.4(8) × 10−6 K−1 from the study of Suzuki et al. (1980) to 800 °C. Temperature derivatives of the elastic moduli for hydrous wadsleyite have not been measured.

Wadsleyite-II Wadsleyite-II, discovered by Smyth and Kawamoto (1997), is a variably hydrous modified spinel similar in structure to spinelloid IV in the nickel-aluminosilicate system. Wadsleyite-II has the same a and c-axis as wadsleyite, but twice the b-axis unit cell parameter as wadsleyite, and is slightly denser (Table 3). While the powder X-ray diffraction pattern and IR-spectra of wadsleyite-II are very similar to wadsleyite, there are subtle differences allowing positive identification of wadsleyite-II (Smyth et al. 2005b). The compressibility of two different wadsleyite-II samples of approximately Fo90 composition and containing 2.1 and 2.8 wt% H2O (SIMS) have been measured by single-crystal X-ray diffraction (Smyth et al. 2005b). The bulk modulus of the 2.1 wt% H2O sample is KT0 = 151(6) GPa with K′ = 6(2), and for the 2.8 wt% H2O sample KT0 = 146(3) GPa with K′ = 6.1(7). The available wadsleyite-II data are listed in Table 3 and plotted in Figure 1.

Ringwoodite Ringwoodite (γ-Mg2SiO4) has the spinel structure and is the stable phase of Mg2SiO4 in the lower part of the transition zone between 520 and 660 km depth. The bulk modulus of nominally anhydrous pure-Mg and Fo90 ringwoodite is about 185 GPa, and the shear modulus is about 120 GPa (Table 4). Synthesis experiments show ringwoodite readily contains 1-2.5 wt% of water (Kohlstedt et al. 1996; Bolfan-Casanova et al. 2000; Smyth et al. 2003). A sample of pure-Mg ringwoodite containing 2.2 wt% H2O (SIMS) was studied at ambient pressure by Brillouin spectroscopy, showing KS0 = 155(4) GPa and G0 = 107(3) GPa (Inoue et al. 1998), representing a 16% decrease in K and 10% decrease in G upon hydration. Subsequently, a sample of similar composition (2.3 wt% H2O) was measured again with Brillouin showing a somewhat higher KS0 = 165.8(5) GPa but similar G0 = 107.4(3) GPa (Wang et al. 2003a). Ultrasonic interferometry was used to determine the bulk and shear moduli of Fo90 hydrous ringwoodite containing ~1 wt% of water, giving KS0 = 176(7) GPa and G0 = 103(5) GPa (Jacobsen et al. 2004). The compressibility of hydrous ringwoodite has also been measured in several X-ray diffraction studies. A pure-Mg powder sample with 2.8 wt% H2O was compressed to 6 GPa showing KT0 = 148(1) GPa with K′ fixed to 5 (Yusa et al. 2000). A single-crystal Fo90 sample containing ~1 wt% H2O was compressed to 11 GPa giving KT0 = 169(3) GPa with K′ = 7.9(9) (Smyth et al. 2004), and a powder sample of similar composition (Fo90, ~1 wt% H2O) was compressed to 45 GPa resulting in KT0 = 175(3) GPa and K′ = 6.2(6) (Manghnani et al. 2005). The compressional and shear wave velocities of Fo90 hydrous ringwoodite were recently measured to 9 GPa (Jacobsen and Smyth 2006). Gigahertz ultrasonic interferometry in the diamond anvil cell was used to determine KS0 = 177(4) GPa, K′ = 5.3(4), G0 = 103.1(9) GPa and G′ = 2.0(2) for hydrous Fo90 ringwoodite containing ~1 wt% H2O (Jacobsen and Smyth 2006). For comparison, nominally anhydrous Fo90 ringwoodite (Sinogeikin et al. 2003) shows KS0 = 188(3) GPa, K′ = 4.1(2), G0 = 120(2) GPa, and G′ = 1.3(1). Thus, hydration of Fo90 ringwoodite to ~1 wt% H2O reduced the bulk and shear moduli by about 6% and 14%, respectively. Considering temperature derivatives of the elastic moduli for anhydrous ringwoodite, dKS/dT = −0.021 GPa/K and dG/dT = −0.015 GPa/K (Sinogeikin et al. 2003; Mayama et al. 2005), hydration of ringwoodite to 1 wt% H2O at room pressure has the same effect on the moduli as increasing the temperature by about 600 °C for KS and 1000 °C for G (Jacobsen et al. 2004). Equation of state parameters for ringwoodite are summarized in Table 4 and plotted in Figures 1 and 2. In addition to compression and sound velocity studies, the strength of hydrous ringwoodite has been measured by non-hydrostatic radial diffraction in a diamond anvil cell (Kavner 2003). The differential stress in hydrous Fo90 ringwoodite increases from 2.9 to 4.5 GPa in the pressure range of 6.7 to 13.2 GPa, indicating a significant water weakening when compared with results

3.47 3.43 3.45 3.65 3.65 3.65

3.56 3.56 3.70

ρSTP (g/cm3)

2.2 2.3 2.8 ~1 ~1 ~1

0 0 0

H2O (wt%)

155(4) 165.8(5) 148(1) 176(7) 177(4) 175(3)

185(3) 185(2) 188(3)

K (GPa)

5.3(4) 6.2(6)

5.0 fixed

4.5(2) 4.1(2)

K′ (dK/dP)

103(5) 103.1(9)

107(3) 107.4(3)

120(2) 120(1) 120(2)

G (GPa)

2.0(2)

1.5(1) 1.3(1)

G′ (dG/dP)

9.26 9.28

9.27 9.50

9.84 9.84 9.69

vP (km/s)

5.31 5.31

5.56 5.60

5.81 5.81 5.68

vS (km/s)

BS BS XRD US US XRD

BS US BS

method*

0 0 6 0 9 45

0 12 16

Pmax (GPa)

[4] [5] [6] [7] [8] [9]

[1] [2] [3]

Ref.

*BS: Brillouin spectroscopy, US: Ultrasonic interferometry, XRD: X-ray diffraction References: [1] Jackson et al. 2000; [2] Li 2003; [3] Sinogeikin et al. 2003; [4] Inoue et al. 1998; [5] Wang et al. 2003a; [6] Yusa et al. 2000; [7] Jacobsen et al. 2004; [8] Jacobsen and Smyth 2006; [9] Manghnani et al. 2005

γ-Mg1.89Si0.97H0.33O4 γ-Mg1.85Si0.99H0.35O4 γ-Mg1.84Si0.98H0.42O4 γ-(Mg0.85Fe0.11)2H0.16SiO4 γ-(Mg0.85Fe0.11)2H0.16SiO4 γ-(Mg0.85Fe0.105)2H0.13SiO4

hydrous ringwoodite

γ-Mg2SiO4 γ-Mg2SiO4 γ-(Mg0.92Fe0.08)2SiO4

anhydrous ringwoodite

phase

Table 4. Elastic properties of ringwoodite.

Eﬀect of Water on Equations of State 331

332

Jacobsen

from anhydrous ringwoodite, which supports 6-8 GPa over the same pressure range (Kavner and Duffy 2001). The volume thermal expansivity of pure-Mg hydrous ringwoodite was measured to its dehydration temperature of ~400 °C, resulting in αV = 27.3(9) × 10−6 K−1 (Inoue et al. 2004), being slightly lower than αV for anhydrous ringwoodite with αV = 30.7(6) × 10−6 K−1 measured to 700 °C in the same study (Inoue et al. 2004), but actually higher than the anhydrous value of αV = 24.7(5) × 10−6 K−1 from the study of Suzuki et al. (1980) to 750 °C. Temperature derivatives of the elastic moduli for hydrous ringwoodite have not been measured.

DENSE HYDROUS MAGNESIUM SILICATES The DHMS are discussed separately because though their stability at various mantle conditions is well known (e.g., Shieh et al. 2000a; Angel et al. 2001; Komabayashi et al. 2005), their actual presence in the mantle is more speculative. However, if subduction of antigorite to 5 GPa occurs below ~550 °C, a dehydration reaction would transfer water to phase A, representing an important choke point for deep-water circulation in the mantle (Bose and Ganguly 1995). Furthermore, if slab temperatures are below 1000 °C at 30 GPa, phase D has the potential to carry water into the upper part of the lower mantle (Komabayashi et al. 2005). Thus, the importance of DHMS to Earth’s potential deep-water cycle warrants consideration of their elastic properties.

Phase A Phase A, Mg7Si2O8(OH)6, is very hydrous with 11.8 wt% H2O and is about 8% less dense than olivine (Table 5). Phase A is stable up to 15 GPa and 1000 °C, whereupon it breaks down to form phase B and brucite (Yamamoto and Akimoto 1977). A single-crystal X-ray diffraction study of phase A shows a bulk modulus of 97.5(4) GPa with K′ = 5.97(14) (Crichton and Ross 2002). A recent Brillouin study of an Fe-bearing phase A shows a slightly higher KS = 106(1) GPa, presumably due to the addition of iron, and shear modulus of 61(1) GPa (Sanchez-Valle et al. 2006). A powder X-ray compression study of the same Fe-bearing phase A sample shows KT0 = 107(3) GPa with K′ = 5.7(3) (Holl et al. 2006).

Phase-B group minerals Phase B, Mg12Si4O19(OH)2, and superhydrous phase B (sometimes written SHy-B), Mg10Si3O14(OH)4, are stoichiometrically hydrous, but have a nominally anhydrous end-member called anhydrous phase B (AHy-B). Phase B group minerals are part of the dense hydrous magnesium silicates (DHMS). Phase B is stable at transition zone pressures and temperatures, it contains 2.4 wt% H2O, and has about the same density as wadsleyite (Table 5). SHy-B is also stable at TZ conditions, contains 5.8 wt% H2O, and is less dense, with about the same density as anhydrous olivine. The bulk modulus of anhydrous phase B is 151.5(9) GPa with K′ = 5.5(3) (Crichton et al. 1999) and the bulk modulus of phase B is 143(2) GPa, with K′ = 7.0(5) (Crichton and Ross 2000a). Surprisingly, the bulk modulus of superhydrous phase B is slightly higher than phase B, with KT0 = 144.9(7) GPa and K′ = 5.1(2), despite having about twice as much water and a slightly lower density than phase B (Crichton and Ross 2000a). Equation of state parameters for phase-B group minerals are summarized in Table 5 and plotted in Figure 1.

Phase D Phase D, MgSi2H2O6, also has about the same density as hydrous wadsleyite but has Si exclusively in 6-coordination with hydroxyl. Phase D is the only known mineral with VISi(OH)6 groups other than thaumasite (Jacobsen et al. 2003). The bulk modulus of phase D is highest of the DHMS, with KT0 = 166(3) GPa and K′ = 4.1(3) (Frost and Fei 1999), determined by powder X-ray diffraction to 30 GPa. The single-crystal elastic constants of phase D have been determined by Brillouin Spectroscopy (Liu et al. 2004), with calculated aggregate moduli of KS0 = 175(15) GPa and G0 = 104(14) GPa, comparable with anhydrous wadsleyite.

2.92 2.84

3.34 3.32

3.33 3.33 3.35

3.37

3.44

2.96 2.96 2.98 2.98

ρSTP (g/cm3)

13.6 18.7

16 18

5.8 5.8 5.8

2.4

0

11.8 11.8 11.7 11.7

H2O (wt%)

93(4) 92.9(7)

175(15) 166(3)

144.9(7) 156(2) 142.6(8)

142.6(17)

151.5(9)

97.5(4) 105(4) 107(3) 106(1)

K (GPa)

5(1) 7.3(2)

4.1(3)

5.1(2) 4.0 fixed 5.8(2)

7.0(5)

5.5(3)

5.97(14) 3.9(8) 5.7(3)

K′ (dK/dP)

104(14)

61(1)

G (GPa)

G′ (dG/dP)

9.70

7.93

vP (km/s)

5.59

4.52

vS (km/s)

XRD XRD

BS XRD

XRD XRD XRD

XRD

XRD

XRD XRD XRD BS

method*

15 6.7

0 30

8 20 8

7

7

7.6 11 33 0

Pmax (GPa)

[10] [11]

[8] [9]

[6] [7] [5]

[6]

[5]

[1] [2] [3] [4]

Ref.

*BS: Brillouin spectroscopy, XRD: X-ray diffraction References: [1] Crichton and Ross 2002; [2] Kuribayashi et al. 2003; [3] Holl et al. 2006 [4] Sanchez-Valle et al. 2006; [5] Crichton et al. 1999; [6] Crichton and Ross 2000a; [7] Shieh et al. 2000a; [8] Liu et al. 2004; [9] Frost and Fei 1999; [10] Shieh et al. 2000b; [11] Crichton and Ross 2000b

Mg2.23Si1.18H2.8O6 Mg1.96Fe0.072Si1.04H3.7O6

phase E (Mg2SiH4O6 ideal)

Mg1.02Si1.71H3.12O6 Mg1.11Si1.6H3.4O6

phase D (MgSi2H2O6 ideal)

Mg10Si3O14(OH)4 Mg10Si3O14(OH)4 Mg9.4Fe0.4Si3.1O14(OH)4

superhydrous phase B (SHy-B)

Mg12Si4O19(OH)2

phase B

Mg14Si5O24

anhydrous phase B (AHy-B)

Mg7Si2O8(OH)6 Mg7Si2O8(OH)6 Mg6.85Fe0.14Si2O8(OH)6 Mg6.85Fe0.14Si2O8(OH)6

phase A

phase

Table 5. Elastic properties of the dense hydrous magnesium silicates (DHMS).

Eﬀect of Water on Equations of State 333

334

Jacobsen

Phase E Phase E, Mg2SiH4O6 has the lowest density of the DHMS and the most water, with 14-20 wt% H2O. The bulk modulus of pure-Mg phase E was measured by powder X-ray diffraction to 15 GPa with KT0 = 93(4) GPa and K′ = 5(1) (Shieh et al. 2000b), and a single-crystal Xray compression study of an Fe-bearing phase E shows KT0 = 92.9(7) GPa with K′ = 7.3(2) (Crichton and Ross 2000b). The elastic properties of these DHMS are summarized in Table 5. DHMS equations of state have been reviewed by Crichton and Ross (2005).

WATER-ELASTICITY SYSTEMATICS The available bulk and shear moduli of hydrated NAM’s, their anhydrous end-members, plus some relevant dense hydrous magnesium silicates (Tables 1 through 5) are plotted in Figures 1 and 2. While references for OH-bearing NAM’s and the DHMS are almost inclusive of the available studies to date, it was necessary to select only a few studies for the anhydrous end members (i.e., olivine, wadsleyite, ringwoodite), which have been studied extensively for many years. Several papers were chosen on the basis that they reflect results from different methods (usually Brillouin spectroscopy, ultrasonic interferometry, or static compression), which generally show agreement between methods. Inspection of the data compiled Figures 1 and 2 reveals one obvious trend, which is that both K and G are measurably reduced upon hydration. It is also observed that the compositional derivative (∂M0/∂XH2O, where M0 is the modulus and XH2O is the water content expressed as wt% H2O) tends to decrease in magnitude as the anhydrous end-member’s M0 value decreases. This observation is clarified in Figure 3, where the anhydrous M0 is plotted against |(∂M0/∂XH2O)|. A dependence of the compositional derivative on the anhydrous moduli (AHyM0) may provide a way of predicting the moduli of hydrated NAM’s for which data are not yet available. The fits to AHyM0 in Figure 3 are: ∂K 0 ⎛ AHyK 0 − 114 ⎞ = −⎜ ⎟ GPa/wt% H 2O ∂X H 2 O 9 ⎝ ⎠

(3)

∂G0 ⎛ AHyG0 − 65 ⎞ = −⎜ ⎟ GPa/wt% H 2O ∂X H 2 O 10 ⎝ ⎠

( 4)

where AHyK0 and AHyG0 are the bulk and shear moduli of the anhydrous end member, respectively. The correlation coefficients for these fits are 0.984 and 0.991, respectively. At the time of writing, the shear modulus of hydrous wadsleyite has not been measured. Taking AHyG0-wadsleyite to be ~110 GPa (Table 2), Equation (4) gives ∂G0/∂XH2O = −4.5 GPa/wt%, as shown by the dashed line in Figure 3. Thus, the predicted shear modulus of wadsleyite containing ~1 wt% of water is ~105 GPa, and for the hypothetical hydrous end member Mg1.5SiH0.5O3 with 3.33 wt% of water, the predicted shear modulus is ~95 GPa. Two points of caution; while the dependence of ∂M0/∂XH2O on the anhydrous M0 appears valid (Fig. 3), it should be noted that it is an empirical relationship without any physical basis. In particular, while it appears reasonable for dense silicate phases, Equations (3) and (4) would predict that for anhydrous end-members with K less than 114 GPa and G less than 65 GPa, the addition of water increase the moduli. Secondly, the relationship implies that ∂M0/∂XH2O is linear, which, over the large water content range to ~12 wt% appears broadly true, but it is also possible that ∂M0/∂XH2O is non linear, especially for low-water contents very near the anhydrous end member. Based upon the available data for (Mg,Fe)2SiO4 polymorphs with ~1 wt% of water, the relationships probably underestimate the effects of water. Therefore, the relationship may be limited in the scope of mineral phases and water contents to which it is applied.

Eﬀect of Water on Equations of State

335

Figure 3. Relationship between the anhydrous bulk and shear moduli (M0) and the compositional derivative |(dM/dXH2O)|. Phases with higher anhydrous M0 show a stronger dependence on water content.

CALCULATED HYDROUS VELOCITIES IN THE UPPER MANTLE AND TRANSITION ZONE In order to illustrate the effect of ~1 wt% of water on seismic velocities in the upper mantle and transition zone with the limited availability of elasticity data, it is necessary to make some assumptions about the effects of water on the shear modulus of olivine and wadsleyite, and on the temperature derivatives of the moduli for all three hydrous phases. The main point is to graphically illustrate the effects of lowering K and G while increasing K′ and G′ on velocities. A model set of thermoelastic parameters for the (Mg,Fe)2SiO4 polymorphs and their hydroxylated analogs is presented in Table 6. Monomineralic velocities were calculated along an adiabat with foot temperature of 1673 K using finite strain theory (e.g., Davies and Dziewonski 1975; Duffy and Anderson 1989). For anhydrous phases, bulk moduli of 129, 172, and 188 GPa are used for olivine, wadsleyite and ringwoodite respectively (Duffy et al. 1995b; Li et al. 1996; Zha et al. 1997; Zha et al. 1998; Li et al. 1998; Li 2003; Sinogeikin et al. 2003; Liu et al. 2005). For all three phases, K′ is assumed to be 4.3 since the measured variation of K′ between them is negligible within experimental uncertainty (Sinogeikin et al. 2003). Shear moduli of 82, 106, and 120 GPa were taken for anhydrous olivine, wadsleyite, and ringwoodite (Duffy et al. 1995b; Li et al. 1998; Li 2003; Sinogeikin et al. 2003; Liu et al. 2005), and a pressure derivative G′ = 1.4 was used for all three phases. For anhydrous olivine, dKS/dT of −0.016 GPa/K and dG/dT = −0.013 GPa/K were used (Liu et al. 2005). For anhydrous wadsleyite, dKS/dT = −0.019 GPa/K and dG/dT = −0.017 GPa/K were used (Meng et al. 1993; Jackson and Rigden 1996; Li et al. 1998; Katsura et al. 2001). For ringwoodite, dKS/dT of −0.021 GPa/K and dG/dT = −0.015 GPa/K were used (Sinogeikin et al. 2003; Mayama et al. 2005). A linear coefficient of volume thermal expansion of 27 × 10−6 K−1 was assumed for all phases. For hydrous olivine, hydrous wadsleyite, and hydrous ringwoodite, bulk moduli of 120, 155, and 177 GPa were used (Yusa and Inoue 1997; Holl et al. 2003; Jacobsen et al. 2004;

336

Jacobsen

Table 6. Model thermoelastic parameters used to estimate the effect of ~1 wt% H2O on monomineralic seismic velocities of the (Mg,Fe)2SiO4 polymorphs in the upper mantle and transition zone (see Figs. 4 and 5). ρ0 (kg/m3)

KS0 (GPa)

KS′ (dKS/dP)

G0 (GPa)

anhydrous olivine

3340

129

4.3

82

1.4

−0.016

−0.013

hydrous olivine

3240

120

5.3

72

1.8

−0.016

−0.013

anhydrous wadsleyite

3600

172

4.3

106

1.4

−0.019

−0.017

hydrous wadsleyite

3400

155

5.3

93

1.8

−0.019

−0.017

anhydrous ringwoodite

3700

188

4.3

120

1.4

−0.021

−0.015

hydrous ringwoodite

3600

177

5.3

105

1.8

−0.021

−0.015

phase

G′ dKS/dT (dG/dP) (GPa/K)

dG/dT (GPa/K)

Smyth et al. 2005). Shear moduli are available only for hydrous ringwoodite, with G ~105 GPa (Inoue et al. 1998; Wang et al. 2003a; Jacobsen et al. 2004). Compared with anhydrous ringwoodite, G is reduced by ~12%, so the shear modulus of hydrous ringwoodite and hydrous wadsleyite were shifted down from their anhydrous values by this amount, resulting in 72 GPa and 93 GPa for hydrous olivine and hydrous wadsleyite, respectively. Given the trends in Figure 3, this assumption likely overestimates the effect of water on G for olivine and wadsleyite, but for the purpose of this illustration it can be considered a maximum affect. Pressure derivatives of all three hydrous polymorphs were assumed to be K′ = 5.3 and G′ = 1.8, and since there is no information on temperature derivatives of the moduli for hydrous NAMs, the anhydrous values were assumed. Monomineralic compressional and shear-wave velocities for anhydrous and hydrous (Mg,Fe)2SiO4 polymorphs from the data in Table 6 are plotted in Figures 4 and 5, along with several seismic velocity models; PEMC (Dziewonski et al. 1975), PREM (Dziewonski and Anderson 1981), and IASP91 (Kennett and Engdahl 1991). In addition, velocities for (anhydrous) majorite garnet are plotted using ρ0 = 3.53 kg/m3, KS0 = 165 GPa and G0 = 88 GPa (Sinogeikin et al. 1997; Gwanmesia et al. 1998) with the same pressure derivatives as used for the other anhydrous phases (K′ = 4.3 and G′ = 1.4) and temperature derivatives for pyrope (Suzuki and Anderson 1983). The general effect of decreasing M and increasing M′ on velocities is illustrated in Figures 4 and 5 since temperature derivatives of the moduli have been assumed to be the same for the individual phases (Table 6). For P-waves, the elevated derivatives bring hydrous values up to anhydrous values, within uncertainty, below about 250 km depth for olivine, below about 350 km depth for wadsleyite, and below about 450 km depth for ringwoodite. Therefore, bulk P-wave velocities alone are not good indicators of local hydration in the mantle, and PREM is equally consistent with hydrous and anhydrous vP in the transition zone (Fig. 4). A similar effect is observed for shear-wave velocities (Fig. 5), however since water has a larger relative effect on G than K, hydrous values do not approach anhydrous values within respective stability fields, but remain 1-3% below anhydrous S-velocities. For S-waves, anhydrous olivine is most consistent with PREM, while hydrous olivine dvS/dP is more consistent with PEMC and IASP91 (Fig. 5). In the transition zone, anhydrous wadsleyite and hydrous wadsleyite are equally consistent with the velocity models, although in the lower part of the transition zone hydrous ringwoodite, or an equal mixture of anhydrous ringwoodite and majorite are most consistent with the velocity models.

Eﬀect of Water on Equations of State

337

Figure 4. Monomineralic P-wave velocities for anhydrous (black lines) and hydrous (~1 wt% H2O, grey lines) phases calculated along a 1673 K (foot) adiabat using finite strain equations of state and model parameters presented in Table 6. Also shown are estimated S-velocities for anhydrous majorite (dash-dotted line). See text for details.

Figure 5. Monomineralic S-wave velocities for anhydrous (black lines) and hydrous (~1 wt% H2O, grey lines) phases calculated along a 1673 K (foot) adiabat using finite strain equations of state and model parameters presented in Table 6. Also shown are estimated S-velocities for anhydrous majorite (dash-dotted line) See text for details.

338

Jacobsen CONCLUSIONS AND FUTURE RESEARCH OPPORTUNITIES

Hydroxylated NAMs are defect structures, and therefore exhibit modified elastic properties compared with OH-free NAMs. At ambient conditions, 0.2 wt% H2O (2000 ppm water by weight) causes a measurable reduction of elastic moduli, on the order of ~2% for 0.2 wt% of H2O added (see Figs. 1-3). Still, ambient shear moduli for OH-bearing olivine and wadsleyite are needed, as well as their pressure derivatives. Temperature derivates of the moduli, ideally at simultaneous high pressure, are needed for all three Mg2SiO4 polymorphs and other NAMs such as pyroxenes. While this chapter has focused on equations of state and elastic wave velocities, reflecting the availability of current experimental data, it is also expected that hydroxyl defects will also influence anelasticity (Dodd and Fraser 1965) and therefore seismic wave attenuation. Due to its importance in seismology and given the measurable affects of “water” on seismic velocities, and especially S-wave velocities, future studies in the direction of understanding the effects of water on seismic-wave attenuation are of high priority. Identifying hydration, or lateral variation of hydration in the TZ, if present at all, would require several types of seismic data together with a comprehensive view of the regional mantle structure. Since water is expected to affect S-wave velocities more than P-wave velocities, the best possible seismic parameter to consider is the vP/vS ratio. Calculated velocities in Figures 4 and 5 indicate that lateral variations in water would manifest in the transition zone as reduced Swave velocities and elevated vP/vS ratios at the level of several percent. Temperature derivatives of the elastic moduli, for hydrated olivine, wadsleyite, and ringwoodite are needed to make an improved estimate of velocities in the upper mantle and transition zone. From seismology, high-resolution local velocity structure is needed in candidate locations, preferably away from subduction zones where thermal anomalies are likely minimal. In the same regions, detailed discontinuity structure (i.e., discontinuity depths, intervals, and total TZ thickness) will be required. Water alone is not the only possible cause of local S-wave anomalies; especially temperature variations, but also variation in the volume fraction of other mineralogical phases such as majorite may influence the local velocity structure in similar ways. Still, the available data indicate that hydration is likely to have a larger influence on S-wave velocities than temperature, at least within the possible variations of each within the TZ. The presence of a broad and elevated 410-km discontinuity, together with a depressed 660-km discontinuity and intervening low S-wave anomalies and high vP/vS ratios would be compatible with hydration considering the available information from mineral physics.

ACKNOWLEDGMENTS I wish to thank Joseph Smyth, Christopher Holl, Carmen Sanchez-Valle, Stanislav Sinogeikin, Fuming Jiang, Tom Duffy, Suzan van der Lee, Sylvie Demouchy, Dan Frost, Ross Angel, Nancy Ross, Hans Keppler, Bjorn Mysen, and Russell Hemley for discussions and making studies in progress available. This study has been supported in part by NSF EAR0440112, a fellowship from the Carnegie Institution of Washington, and the Carnegie/DOE Alliance Center (CDAC).

REFERENCES Agee CB (2005) Static compression of hydrous komatiite liquid. EOS Trans, Am Geophys Union 86:MR11A03 Angel RJ (2000) Equations of state. Rev Min Geochem 41:35-59 Angel RJ, Chopelas A, Ross NL (1992) Stability of high-density clinoenstatite at upper-mantle pressures. Nature 358:322-324 Angel RJ, Frost DJ, Ross NL, Hemley R (2001) Stabilities and equations of state of dense hydrous magnesium silicates. Phys Earth Planet Int 127:181-196

Eﬀect of Water on Equations of State

339

Bass JD (1989) Elasticity of grossular and spessartite garnets by Brillouin spectroscopy. J Geophys Res 94: 7621-7628 Bell DR, Rossman GR (1992) Water in Earth’s mantle: the role of nominally anhydrous minerals. Science 255: 1391-1397 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett 182:209-221 Bose K, Ganguly J (1995) Experimental and theoretical studies of the stabilities of talc, antigorite and phase A at high pressures with applications to subduction processes. Earth Planet Sci Lett 136:109-121 Bromiley GD, Bromiley FA (2006) High-pressure phase transitions and hydrogen incorporation in MgSiO3 enstatite. Am Mineral 91:1094-1101 Chen J, Inoue T, Yurimoto H, Weidner DJ (2002) Effect of water on olivine-wadsleyite phase boundary in the (Mg,Fe)2SiO4 system. Geophys Res Lett 29:1875 Conrad PG, Zha CS, Mao HK, Hemley RJ (1999) The high-pressure, single-crystal elasticity of pyrope, grossular, and andradite. Am Mineral 84:374-383 Crichton WA, Ross NL (2000a) Single crystal equation of state measurements on Mg-end-members of the B-group minerals. In: Manghnani MH, Nellis WJ, Nicol MF (eds) Science and Technology of High Pressure, Proc AIRAPT-17:587-590 Crichton WA, Ross NL (2000b) Equation of state of phase E. Mineral Mag 64(3):561-567 Crichton WA, Ross NL (2002) Equation of state of dense hydrous magnesium silicate phase A, Mg7Si2O8(OH)6. Am Mineral 87:333-338 Crichton WA, Ross NL (2005) Equations of state of dense hydrous magnesium silicates: results from singlecrystal X-ray diffraction. Mineral Mag 69:273-287 Crichton WA, Ross NL, Gasparik T (1999) Equations of state of magnesium silicates anhydrous B and superhydrous B. Phys Chem Minerals 26:570-575 Davies GF, Dziewonski AM (1975) Homogeneity and constitution of the Earth’s lower mantle and outer core. Phys Earth Planet Int 10:336-343 Demouchy S, Jacobsen SD, Gaillard F, Stern CR (2006) Rapid magma ascent recorded by water diffusion profiles in mantle olivine. Geology 34:429-432 Dodd DM, Fraser DB (1965) The 3000-3900 cm−1 absorption bands and anelasticity in crystalline α-quartz. J Phys Chem Solids 26:673-686 Duffy TS, Anderson DL (1989) Seismic velocities in mantle minerals and the mineralogy of the upper mantle. J Geophys Res 94:1895-1912 Duffy TS, Shu J, Mao HK, Hemley RJ (1995a) Single-crystal X-ray diffraction of brucite to 14 GPa. Phys Chem Minerals 22:277-281 Duffy TS, Zha CS, Downs RT, Mao HK, Hemley RJ (1995b) Elasticity of forsterite to 16 GPa and the composition of the upper mantle. Nature 378:170-173 Dziewonski AM, Anderson DL (1981) Preliminary reference earth model, Phys Earth Planet Int 25:297-356 Dziewonski AM, Hales AL, Lapwood ER (1975) Parametrically simple earth models consistent with geophysical data. Phys Earth Planet Int 10:12-48 Friedrich A, Lager GA, Ulmer P, Kunz M, Marshall WG (2002) High-pressure single-crystal X-ray and powder neutron study of F,OH/OD-chondrodite: compressibility, structure and hydrogen bonding. Am Mineral 87:931-939 Fritzel TLB, Bass JD (1997) Sound velocities of clinohumite, and implications for water in Earth’s upper mantle. Geophys Res Lett 24:1023-1026 Frost DJ, Fei Y (1999) Static compression of the hydrous magnesium silicate phase D to 30 GPa at room temperature. Phys Chem Minerals 26:415-418 Gwanmesia GD, Chen G, Liebermann RC (1998) Sound velocities in MgSiO3-garnet to 8 GPa. Geophys Res Lett 25:4553-4556 Harris PG, Middlemost AK (1969) The evolution of kimberlites. Lithos 3:77-88 Holl CM, Smyth JR, Jacobsen SD (2003) The effect of water on the compressibility of wadsleyite. EOS Trans, Am Geophys Union 84:T11C-0414 Holl CM, Smyth JR, Manghnani MH, Amulele GM, Sekar M, Frost DJ, Prakapenka VB, Shen G (2006) Crystal structure and compression of an iron-bearing phase A to 33 GPa. Phys Chem Minerals 33:in press Ingrin J, Skogby H (2000) Hydrogen in nominally anhydrous upper-mantle minerals: concentration levels and implications. Eur J Min 12:543-570 Inoue T, Tanimoto Y, Irifune T, Suzuki T, Fukui H, Ohtaka O (2004) Thermal expansion of wadsleyite, ringwoodite, hydrous wadsleyite and hydrous ringwoodite. Phys Earth Planet Int 143-144:279-290 Inoue T, Weidner DJ, Northrup PA, Parise JB (1998) Elastic properties of hydrous ringwoodite (γ-phase) in Mg2SiO4. Earth Planet Sci Lett 160:107-113 Inoue T, Yurimoto H, Kudoh Y (1995) Hydrous modified spinel, Mg1.75SiH0.5O4: a new water reservoir in the mantle transition region. Geophys Res Lett 22:117-120

340

Jacobsen

Jacobsen SD, Demouchy S, Frost DJ, Ballaran TB, Kung J (2005) Systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single-crystal X-ray diffraction: oxygen sites for hydrogen storage in Earth’s interior. Am Mineral 90:61-70 Jacobsen SD, Smyth JR (2006) Effect of water on the sound velocities of ringwoodite in the transition zone. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (eds) Am Geophys Union Monogr Ser (in press) Jacobsen SD, Smyth JR, Spetzler H, Holl CM, Frost DJ (2004) Sound velocities and elastic constants of ironbearing hydrous ringwoodite. Phys Earth Planet Int 143-144:47-56 Jacobsen SD, Smyth JR, Swope RJ (2003) Thermal expansion of hydrated six-coordinate silicon in thaumasite, Ca3Si(OH)6(SO4)12H2O. Phys Chem Minerals 30:321-329 Jackson I, Rigden SM (1996) Analysis of P-V-T data: constraints on the thermoelastic properties of highpressure minerals. Phys Earth Planet Int 96:85-112 Jiang F, Speziale S, Duffy TS (2006) Single-crystal elasticity of brucite, Mg(OH)2, to 15 GPa by Brillouin scattering. Am Mineral 91, in press Katsura T, Mayama N, Shouno K, Sakai M, Yoneda A, Suzuki I (2001) Temperature derivatives of elastic moduli of (Mg0.91Fe0.09)2SiO4 modified spinel. Phys Earth Planet Int 124:163-166 Kavner A (2003) Elasticity and strength of hydrous ringwoodite at high pressure. Earth Planet Sci Lett 214: 645-654 Kavner A, Duffy TS (2001) Strength and elasticity of ringwoodite at upper mantle pressures. Geophys Res Lett 28:2691-2694 Kennett BLN, Engdahl ER (1991) Travel times for global earthquake location and phase identification. Geophys J Int 105:429-465 Kitamura M, Kondoh S, Morimoto N, Miller GH, Rossman GR, Putnis A (1987) Planar OH-bearing defects in mantle olivine. Nature 328:143-145 Knittle E, Hathorne A, Davis M, Williams Q (1992) A spectroscopic study of the high-pressure behavior of the O4H4 substitution in garnet. In: High-pressure Research: Application to Earth and Planetary Sciences. Syono Y, Manghnani MH (eds) Am Geophys Union, p 297-304 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ (2002) Ordering of hydroxyl defects in hydrous wadsleyite (β-Mg2SiO4). Am Mineral 87:293-301 Komabayashi TS, Omori S, Maruyama S (2005) Experimental and theoretical study of stability of dense hydrous magnesium silicates in the deep upper mantle. Phys Earth Planet Int 153:191-209 Kudoh Y, Inoue T (1999) Mg-vacant structural modules and dilution of the symmetry of hydrous wadsleyite, β-Mg2−xSiH2xO4 with 0.00 ≤ x ≤ 0.25. Phys Chem Minerals 26:382-388 Kuribayashi T, Kudoh Y, Tanaka M (2003) Compressibility of phase A, Mg7Si2H6O14, up to 11.2 GPa. J Mineral Pet Sci 98:215-234 Lager GA, Downs RT, Origlieri M, Garoutte R (2002) High-pressure single-crystal X-ray diffraction study of katoite hydrogarnet: evidence for a phase transition from Ia3d to I43d symmetry at 5 GPa. Am Mineral 87:642-647 Lager GA, Marshall WG, Liu Z, Downs RT (2005) Re-examination of the hydrogarnet structure at high pressure using neutron powder diffraction and infrared spectroscopy. Am Mineral 90:639-644 Lager GA, Von Dreele RB (1996) Neutron powder diffraction of hydrogarnet to 9.0 GPa. Am Mineral 81: 1097-1104 Li B (2003) Compressional and shear wave velocities of ringwoodite γ-Mg2SiO4 to 12 GPa. Am Mineral 88: 1312-1317 Li B, Gwanmesia GD, Liebermann RC (1996) Sound velocities of olivine and beta polymorphs of Mg2SiO4 at Earth’s transition zone pressures. Geophys Res Lett 23:2259-2262 Li B, Liebermann RC (2000) Sound velocities of wadsleyite β-(Mg0.88Fe0.12)2SiO4 to 10 GPa. Am Mineral 85: 292-295 Li B, Liebermann RC, Weidner DJ (1998) Elastic moduli of wadsleyite (β-Mg2SiO4) to 7 Gigapascals and 873 Kelvin. Science 281:675-677 Liu W, Kung J, Li B (2005) Elasticity of San Carlos olivine to 8 GPa and 1073 K. Geophys Res Lett 32: L16301 Liu LG, Okamoto K, Yang YJ, Chen, CC, Lin CC (2004) Elasticity of single-crystal phase D (a dense hydrous magnesium silicate) by Brillouin spectroscopy. Solid State Comm 132:517-520 Manghnani MH, Amulele G, Smyth JR, Holl CM, Chen G, Prakapenka V, Frost DJ (2005) Equation of state of hydrous Fo90 ringwoodite to 45 GPa by synchrotron powder diffraction, Mineral Mag 69:317-323 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554-570 Matsukage K, Jing Z, Karato SI (2005) Density of hydrous silicate melt at the conditions of Earth’s deep upper mantle. Nature 438:488-491 Mayama N, Suzuki I, Saito T, Ohno I, Katsura T, Yoneda A (2005) Temperature dependence of the elastic moduli of ringwoodite. Phys Earth Planet Int 148:353-359

Eﬀect of Water on Equations of State

341

McCormick TC, Hazen RM, Angel RJ (1989) Compressibility of omphacite to 60 kbar: role of vacancies. Am Mineral 74:1287-1292 McMillan PF, Akaogi M, Sato RK, Poe B, Foley J (1991) Hydroxyl groups in β-Mg2SiO4. Am Mineral 76: 354-360 Meng Y, Weidner DJ, Gwanmesia GD, Liebermann RC, Vaughan MT, Wang Y, Leinenweber K, Pacalo RE, Yeganeh-Haeri A, Zhao Y (1993) In situ high P-T X-ray diffraction studies on three polymorphs (α, β, γ) of Mg2SiO4. J Geophys Res 98:22,199-22,207 Mierdel K, Keppler H (2004) The temperature dependence of water solubility in enstatite. Contrib Mineral Petrol 148:305-311 Miller GH, Rossman GR, Harlow GE (1987) The natural occurrence of hydroxide in olivine. Phys Chem Minerals 14:461-472 Mosenfelder JL, Deligne NI, Asimow PD, Rossman GR (2006) Hydrogen incorporation in olivine from 2-12 GPa. Am Mineral 91:285-294 Olijnyk H, Paris E, Geiger CA, Lager GA (1991) Compressional study of katoite [Ca3Al2(O4H4)3] and grossular garnet. J Geophys Res 96:14313-14318 O’Neill B, Bass JD, Rossman GR (1993) Elastic properties of hydrogrossular garnet and implications for water in the upper mantle. J Geophys Res 98:20031-20037 Pavese A, Levy D, Pischedda V (2001) Elastic properties of andradite and grossular by synchrotron X-ray diffraction at high pressure conditions. Eur J Mineral 13:929-937 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Ringwood AE (1966) The chemical composition and origin of the Earth. In: Advances in Earth Sciences. Hurley PM (ed) MIT Press, p 287-356 Ross NL, Crichton WA (2001) Compression of synthetic hydroxylclinohumite [Mg9Si4O16(OH)2)] and hydroxylchondrodite [Mg5Si2O8(OH)2]. Am Mineral 86:990-996 Rossman GR (1996) Studies of OH in nominally anhydrous minerals. Phys Chem Minerals 23:299-304 Sakamaki T, Suzuki A, Ohtani E (2006) Stability of hydrous melt at the base of the Earth’s upper mantle. Nature 439:192-194 Sanchez-Valle C, Sinogeikin SV, Smyth JR, Bass JD (2006) Single-crystal elastic properties of dense hydrous magnesium silicate phase A. Am Mineral 91:961-964 Shieh SR, Mao HK, Hemley RJ, Ming LC (2000a) In situ X-ray diffraction studies of dense hydrous magnesium silicates at mantle conditions. Earth Planet Sci Lett 177:69-80 Shieh SR, Mao HK, Konzett J, Hemley RJ (2000b) In-situ high pressure X-ray diffraction of phase E to 15 GPa. Am Mineral 85:765-769 Sinogeikin SV, Bass JD (1999) Single-crystal elastic properties of chondrodite: implications for water in the upper mantle. Phys Chem Minerals 26:297-303 Sinogeikin SV, Bass JD, O’Neill B, Gasparik T (1997) Elasticity of tetragonal end-member majorite and solid solutions in the system Mg4Si4O12-Mg3Al2Si3O12. Phys Chem Minerals 24:115-121 Sinogeikin SV, Bass JD, Katsura T (2003) Single-crystal elasticity of ringwoodite to high pressures and high temperatures: implications for 520 km seismic discontinuity, Phys Earth Planet Int 136:41-66 Sinogeikin SV, Katsura T, Bass JD (1998) Sound velocities and elastic properties of Fe-bearing wadsleyite and ringwoodite. J Geophys Res 103:20819-20825 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxenes: variations in the natural environment. Am Mineral 75:764-774 Smyth JR (1987) β-Mg2SiO4: a potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR, Frost DJ, Nestola F (2005a) Hydration of olivine and the Earth’s deep water cycle. Geochim Cosmochim Acta 69:A746 Smyth JR, Holl CM, Frost DJ, Jacobsen SD (2004) High pressure crystal chemistry of hydrous ringwoodite and water in the Earth’s interior. Phys Earth Planet Int 143-144:271-278 Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Langenhorst F, McCammon CA (2003) Structural systematics of hydrous ringwoodite and water in Earth’s interior. Am Mineral 88:1402-1407 Smyth JR, Holl CM, Langenhorst F, Lausten HMS, Rossman GR, Kleppe A, McCammon CA, Kawamoto T, van Aken PA (2005b) Crystal chemistry of wadsleyite II and water in the Earth’s interior. Phys Chem Minerals 31:691-705 Smyth JR, Kawamoto T (1997) Wadsleyite II: a new high pressure hydrous phase in the peridotite-H2O system. Earth Planet Sci Lett 146:E9-E16 Smyth JR, Kawamoto T, Jacobsen SD, Swope JR, Hervig RL, Holloway JR (1997) Crystal structure of monoclinic hydrous wadsleyite [β-(Mg,Fe)2SiO4]. Am Mineral 82:270-275 Smyth JR, Rossman GR, Bell DR (1991) Incorporation of hydroxyl in upper mantle clinopyroxenes. Nature 351:732-735 Stalder R, Skogby H (2002) Hydrogen incorporation in enstatite. Eur J Mineral 14:1139-1144 Suzuki I, Anderson OL (1983) Elasticity and thermal expansion of a natural garnet up to 1000 K. J Phys Earth 31:125-138

342

Jacobsen

Suzuki I, Ohtani E, Kumazawa M (1980) Thermal expansion of modified spinel, β-Mg2SiO4. J Phys Earth 28: 273-280 Wang J, Sinogeikin SV, Inoue T, Bass JD (2003a) Elastic properties of hydrous ringwoodite. Am Mineral 88: 1608-1611 Wang J, Sinogeikin SV, Inoue T, Bass JD (2003b) Elastic properties of hydrous ringwoodite at high pressures. EOS Trans, Am Geophys Union 84:V31D-0971 Yamamoto K, Akimoto S (1977) The system MgO-SiO2-H2O at high pressures and temperatures - stability field for hydroxyl-chondrodite, hydroxyl-clinohumite, and 10 Å phase. Am J Sci 277:288-312 Young TE, Green HW II, Hofmeister AM, Walker D (1993) Infrared spectroscopic investigation of hydroxyl in β-(Mg,Fe)2SiO4 and coexisting olivine: implications for mantle evolution and dynamics. Phys Chem Minerals 19:409–422 Yusa H, Inoue T (1997) Compressibility of hydrous wadsleyite (β-phase) in Mg2SiO4 by high pressure X-ray diffraction. Geophys Res Lett 24:1831-1834 Yusa H, Inoue T, Ohishi Y (2000) Isothermal compressibility of hydrous ringwoodite and its relation to the mantle discontinuities. Geophys Res Lett 27:413-416 Zhang L (1998) Single crystal hydrostatic compression of (Mg,Mn,Fe,Co)2SiO4 olivines. Phys Chem Minerals 25:308-312 Zha CS, Duffy TS, Downs RT, Mao HK, Hemley RJ (1996) Sound velocity and elasticity of single-crystal forsterite to 16 GPa. J Geophys Res 101:17535-17545 Zha CS, Duffy TS, Downs RT, Mao HK, Hemley RJ (1998) Brillouin scattering and X-ray diffraction of San Carlos olivine: direct pressure determination to 32 GPa. Earth Planet Sci Lett 159:25-33 Zha CS, Duffy TS, Mao HK, Downs RT, Hemley RJ, Weidner DJ (1997) Single-crystal elasticity of β-Mg2SiO4 to the pressure of the 410 km seismic discontinuity in the Earth’s mantle. Earth Planet Sci Lett 147:9-15

15

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 343-375, 2006 Copyright © Mineralogical Society of America

Remote Sensing of Hydrogen in Earth’s Mantle Shun-ichiro Karato Department of Geology and Geophysics Yale University New Haven, Connecticut, 06520-8109, U.S.A. e-mail: [email protected]

INTRODUCTION 1

Hydrogen in Earth’s interior is known to play a key role in a number of processes. Consequently, inferring the distribution of hydrogen is a critical step in our study of dynamics and evolution of Earth. Usually, hydrogen distribution is inferred from two types of samples at the surface. First, a magma may contain hydrogen (water), and under some conditions the water content of the magma can be quenched upon cooling. In these cases, measurements of the water content of the magma provide us with some constraints on the hydrogen (water) content of the source region (if we know the partitioning of hydrogen between magmas and the source rocks, and the degree of melting). Second, hydrogen content of some xenoliths transported by magma can be measured. They provide a direct clue as to the hydrogen content in a region from which a xenolith has been carried. However, these direct, petrologic approaches have two major problems. Firstly, the sampling is limited by the distribution of volcanoes, and even if there are volcanoes that carry rocks from Earth’s interior, the depth extent that volcanoes sample rocks is limited (usually <200 km). Secondly, there is no guarantee that the hydrogen content that one measures on these samples actually represents the hydrogen content in a region where these samples came from. For example, hydrogen is known to diffuse very easily so hydrogen dissolved in minerals could diffuse out during the transport of a rock, or conversely, a piece of rock may acquire extra hydrogen during its ascent. Also hydrogen atoms dissolved in minerals may precipitate in the mineral to form fluid inclusions or micro-scale hydrous minerals. In summary, the direct method to infer the distribution of hydrogen from rock samples has major limitations, and an alternative approach, i.e., remote sensing hydrogen content from geophysical observations appears to be worth serious consideration. In this chapter, I will review some of the recent progress in inferring hydrogen content based on geophysical observations. The major advantage of this approach is the fact that hydrogen in much broader regions can be inferred because geophysical observations span much broader regions, particularly the depth. Because hydrogen has drastic effects on physical properties, some of the geophysically measurable properties are sensitive to hydrogen content. However, such an approach is new and a number of challenges exist. The key, from mineral physics point of view, in this attempt is to establish the relationship between hydrogen content and geophysical properties including the analysis of a range of trade-offs. Establishing these relationships including the developments of methodology to differentiate the influence of hydrogen from other factors requires in-depth analysis of microscopic physics of various 1

I will use a term “hydrogen” in most part of this paper to imply that it is hydrogen and not water molecule that affects the physical and chemical properties of silicates. In the geological literature, however, the concentration of chemical species is usually measured as wt% of oxide. Therefore when the concentration of hydrogen is discussed I will use water content in wt% following geological literatures. 1 wt% of water in olivine corresponds to 1.5×105 ppm H/Si (atomic ratio) and not molecular water.

1529-6466/06/0062-0015$05.00

DOI: 10.2138/rmg.2006.62.15

344

Karato

properties in addition to a sound understanding of geophysical methods of measuring relevant properties. In this chapter, I will first provide a brief tutorial of geophysical methods of inferring some of the key parameters that may be sensitive to hydrogen. Then I will give a detailed discussion on the mineral physics bases for inferring hydrogen content from geophysical observations. This will be followed by the discussion on some examples showing how these methods can be applied to infer hydrogen contents in various regions of Earth’s mantle.

GEOPHYSICAL OBSERVATIONS When an attempt is made to interpret any geophysical observations, it is critical to understand the nature of uncertainties in each geophysical technique. For example, in seismological studies, the nature of structures that one can resolve depends on the wavelength of waves used and hence the nature of uncertainties is different between body wave and surface wave observations. Also there is a complication in inferring anisotropic structures from seismology. Understanding these technical issues is critical to interpret geophysical observations. Therefore I will provide a brief summary of some of the geophysical techniques.

Electrical conductivity The physical basis of the methods for inferring electrical conductivity in the Earth is electromagnetic induction. When there is a variation in the magnetic field outside of the Earth (due to solar storms for example), then electric current is induced inside the Earth whose magnitude depends on the electrical conductivity. The measurements of electromagnetic field are in most cases made on Earth’s surface. The observed electromagnetic field comes both from the “source” (i.e., field from the outside of Earth) and the “induced” field inside the Earth. Therefore the first step is to separate them using the spherical harmonic analysis. The field coming from inside is interpreted in terms of distribution of electrical conductivity. So the basic equations to govern this phenomenon is an induction equation, viz., ∇2F = κ

∂F ∂t

(1)

where F is either the electric or magnetic field, and κ ≡ 4πσcond·µmag where σcond is the electrical conductivity and µmag is the magnetic susceptibility. For a one-dimensional case, the solution of this equation has a form z F ∝ exp ⎛⎜ − ⎞⎟ exp ( i ωt ) z ⎝ 0⎠

(2)

with z0 =

1 2 πσ cond µ mag ω

(3)

where z is the depth, ω is the (angular) frequency, t is time, σcond is the electrical conductivity and µmag is the magnetic susceptibility. Consequently, the magnetic disturbance with a long period will reflect the conductivity at greater depth. The magnetic susceptibility is nearly constant in Earth, so one infers the distribution of electrical conductivity from this type of observation. A lower frequency disturbance penetrates deeper into Earth’s interior, and can sense the electrical conductivity in the deeper portions. In other words, by investigating the time-variation of the electromagnetic field with various frequencies one can determine the depth variation of conductivity. In some cases, when the anisotropy in the electromagnetic field is determined, one will obtain some constraint on the anisotropy in conductivity. One

Remote Sensing of Hydrogen in the Earth’s Mantle

345

limitation with this approach is that because the amplitude of electromagnetic disturbance decays with depth exponentially, the depth resolution of this technique is not high. Similar to surface wave studies, sharp discontinuities are difficult to detect by electromagnetic induction. Since the depths of discontinuities are not well constrained from the geomagnetic studies, these depths are usually assumed and the magnitude of conductivity jumps and a smooth variation in conductivity between these discontinuities are determined by the inversion. Therefore it is critical to know the depth at which a discontinuity in conductivity may occur. An important progress in this connection was the experimental study by Xu et al. (1998b) who showed that the electrical conductivity of wadsleyite and ringwoodite are much higher than that of olivine and comparable to that of silicate perovskite2. These studies motivated inversion of electromagnetic induction data assuming jumps in conductivity at 410 km and 660 km. If one does not assume any discontinuities in conductivity, the best-fit models tend to be biased towards those with a smooth depth variation. So in short, among a range of models of electrical conductivity, those with assumed discontinuities at 410-km (and 660-km) best resolve any jumps in conductivity caused by a discontinuity in water content. Technical issues of inverting for the electromagnetic sounding are discussed by Parker (1980), Constable et al. (1987) and de Groot-Hedlin and Constable (2004). Figure 1 illustrates one example of such a study showing a marked increase in electrical conductivity at 410-km discontinuity in the mantle beneath the Pacific ocean (from Utada et al. 2003). Note also that the electrical conductivity also has a large lateral variation. Utada et al. (2005) reported a higher conductivity in the transition zone beneath Hawaii, as well as some regions of the mantle beneath Philippine Sea. A large jump in electrical conductivity at ~410-km is frequently observed in both oceanic and continental upper mantle (e.g., Olsen 1999; Tarits et al. 2004).

Figure 1. Electrical conductivity-depth profiles inferred from electromagnetic sounding in the Pacific ocean. Three profiles correspond to the results of inversion based on different assumptions about the discontinuity (after Utada et al. 2003). The results show a jump in electrical conductivity of a factor of ~10. Electrical conductivity shows a large regional variation, particularly in the upper mantle. However, a jump of a factor of ~10 is often observed at 410-km discontinuity (e.g., Olsen 1999; Tarits et al. 2004). 2

As I will explain later, the interpretation of the data by Xu et al. (1998) turned out to be incorrect. But ironically this incorrect interpretation promoted detailed studies on the depth variation in conductivity across the transition zone.

346

Karato

Seismic wave velocities Variation of seismic wave velocities in Earth can be determined by various techniques. There are excellent textbooks on this topic, but I need to summarize some of the crucial points for our purpose. The key question that I address in this chapter is the distribution of hydrogen as inferred from the geophysical observations on the upper mantle and the transition zone. In this context, it is important to appreciate the difference in the nature of seismological observations inferred from short-period body-waves and relatively long-period surface waves. The use of surface waves is critical to resolve the depth-dependent structure of the upper mantle where body-waves have little sensitivity. However, by its very nature, surface wave data are not sensitive to sharp but small discontinuities. Therefore the surface wave data are good data source to determine a gross depth-dependent structure, but are not a good data set to identify small but sharp discontinuities. The latter type of structures is better constrained by body-wave observations, particularly those using data of converted waves at discontinuities. An important case to illustrate this point is the inference of upper mantle structure of the oceanic upper mantle. A majority of studies on upper mantle structure is based on surface wave data, and these studies on the oceanic upper mantle show a gradual increase in the thickness of oceanic lithosphere associated with a gradual increase in the wave velocity in the asthenosphere (at the same depth) (e.g., Yoshii 1973; Forsyth 1975). In contrast, some recent studies using short wavelength body waves show a rather constant depth of the discontinuity at around ~60-70 km depth (e.g., Gaherty et al. 1996; Evans et al. 2005). Karato and Jung (1998) interpreted this sharp, age-independent discontinuity as a result of stiffening due to dewatering as originally proposed by Karato (1986). (Hirth and Kohlstedt (1996) adopted the same model and added some petrological details.) One of the common difficulties in inverting geophysical observations in terms of physical and chemical state of Earth is non-uniqueness. The same low velocity anomaly can be attributed to high temperature, high iron content, partial melting or a high water content. To help reduce the non-uniqueness, simultaneous inversion of multiple data set is a useful approach. Not only P-wave but also S-wave velocity anomalies should be used. However, one must make sure that the resolutions of each technique are similar in order to combine the results of different data sets. Some technical issues on simultaneous inversion of P and S waves are discussed by Masters et al. (2000).

Seismic wave attenuation Velocity anomalies are the primary data from seismology that we can get with reasonably high resolution. Velocities of seismic waves reflect the “stiffness” (i.e., elastic constants) (and density) of Earth materials. However, in some cases, we can also obtain some information on “softness” of materials from seismological data. This is seismic wave attenuation that reflects non-elastic deformation of Earth materials which is very sensitive to temperature and a small amount of hydrogen. Reviews on seismic wave attenuation from seismological view-points include Bhattacharya et al. (1996) and Romanowicz and Durek (2000). Seismic wave attenuation, or the degree to which energy is lost during the wave propagation, is characterized by a Q factor defined by Q −1 ≡

∆E 2πE

( 4)

where ∆E is the energy dissipated during one cycle of wave propagation and E is energy stored in the system. The Q factor can be determined by the amplitude decay of seismic waves that can be determined by the frequency dependence of amplitude, viz., ⎛ ω ⎞ A ( x, ω) = A0 exp ⎜ − x⎟ ⎝ 2 vQ ⎠

(5)

Remote Sensing of Hydrogen in the Earth’s Mantle

347

where v is the velocity of seismic waves, x is distance and ω is the (angular) frequency of seismic waves. Usually, determining the attenuation is more difficult than seismic wave velocities because a number of other factors can contribute to the amplitude including geometrical focusing or defocusing, and scattering. In particular, the effects of geometrical focusing or defocusing are large for surface wave studies and consequently careful analysis of seismic records is required including waveform inversion (e.g., Gung and Romanowicz 2004). Nevertheless, when highdensity seismic sources and receivers are available, one can calculate a decent three-dimensional mapping of seismic wave attenuation. An example of results of attenuation tomography is shown together with velocity tomography (Shito et al. 2006). In this special case, attenuation factor, Q, was determined from the ratio of amplitude of P- and S-waves from the same data set assuming that QP/QS is constant throughout the study region. This procedure minimizes the influence of source and geometrical effects and if one has a dense array of seismic stations and sources (earthquakes), then one can obtain high-resolution tomographic maps (Fig. 2).

Seismic anisotropy Seismic anisotropy can provide additional constraints on the distribution of hydrogen. The possible mechanism for hydrogen to alter seismic anisotropy in the upper mantle was proposed by Karato (1995) based on experimental observations showing the high anisotropy in hydrogen weakening effects in olivine single crystals (Mackwell et al. 1985). Seismic anisotropy can be determined by a range of methods, but two methods are most widely used. These are shear-wave splitting measurements, and the measurements of vSH /vSV polarization anomalies. Both of these methods use polarization anisotropy as opposed to

Figure 2. Tomographic images of a cross section of the upper mantle beneath the Philippine Sea. The left hand side figures show anomalies in attenuation, and the right hand side figures show anomalies in P-wave velocities (from Shito et al. 2006). See Plate 1 for color figure.

348

Karato

azimuthal anisotropy. The polarization anisotropy is a type of anisotropy that describes the difference in the velocity of two types of shear waves with different polarizations. The major advantage of using this method is that one can determine seismic anisotropy from a single seismic record, and therefore the results are free from the influence of lateral heterogeneity. Azimuthal anisotropy can also be determined if a dense coverage of seismic rays is available for a given region, but in general, the errors in azimuthal anisotropy are larger than those for polarization anisotropy. An example of high-resolution shear wave splitting measurements is shown in Figure 3.

Topography of discontinuity The topography of some of the seismic discontinuities may also depend on hydrogen content. This is the case when a seismic discontinuity is caused by a phase transformation and when there is a large partitioning of hydrogen between the two minerals. An example is the olivine to wadsleyite transformation that occurs at ~410-km. The topography of a discontinuity can be measured by using seismic waves converted or reflected at a boundary (see e.g., Shearer 2000). Topography on a given discontinuity could be caused by several reasons including temperature anomalies and water content anomalies. Therefore, similar to velocity anomalies, if only topographic anomalies are known, it is difficult to obtain unique conclusion about the cause. Therefore it is important to obtain not only the topography on the discontinuity but also some other parameters from the same region. For example, as will be explained in the next section, topography on the discontinuity could be due to a temperature anomaly or to an anomaly in hydrogen content. Both factors affect the topography and velocity in different ways, so the simultaneous inversion of these two parameters will provide a tight constraint on hydrogen content. Since inferring the depth of a discontinuity relies on the knowledge of seismic wave velocities, there is a trade-off between the determination of topography and velocity anomalies. Gu et al. (2003) discussed the methods to reduce the uncertainties caused by this trade-off. By combining the results of topography on the “410-km” discontinuity and velocity anomalies

(a)

(b)

Figure 3. An example of spatially varying pattern of shear wave splitting using regional seismic sources and dense stations (from Nakajima and Hasegawa 2004). (a) A map of study area (Tohoku Japan) where plate subduction occurs from the east. (b) The results of shear wave splitting measurements. An abrupt change in the pattern occurs near the volcanic front. Near the trench the direction of polarization of fast Swave is near trench parallel, whereas it becomes nearly normal to the trench at a far distance. Such a trend is commonly observed in many subduction zones (e.g., Smith et al. 2001; Long and van der Hilst 2005).

Remote Sensing of Hydrogen in the Earth’s Mantle

349

in the transition zone, Blum and Sheng (2004) inferred the hydrogen content in the transition zone. Suetsugu et al. (2006) conducted a similar study for the western Pacific. However, in these studies, topography and velocity anomalies were not determined in a consistent way, but rather determined separately. Therefore the results of these earlier studies have large uncertainties.

Sharpness of discontinuities The sharpness of discontinuities may depend on the hydrogen content if the sharpness is related to a phase transformation loop involving two minerals with a large contrast in hydrogen solubility (Wood 1995). The sharpness can be inferred from a range of seismic observations particularly those relying on the amplitude of converted phases at a discontinuity (e.g., Benz and Vidale 1993; van der Meijde et al. 2003). In essence the inference of sharpness is based on the amplitude of converted waves or reflected waves as a function of wavelength. However, because these methods use the information on the amplitude of waves, the uncertainties are large as compared to the measurements of the depth of seismic discontinuities. In addition, the mineral physics background for interpreting the sharpness is more complicated than the interpretation of the topography of a discontinuity as I will discuss later.

PHYSICAL BASIS FOR INFERRING HYDROGEN CONTENT FROM GEOPHYSICAL OBSERVATIONS Electrical conductivity The first quantitative work on the influence of hydrogen on electrical conductivity is Karato (1990) who pointed out that a high diffusivity and solubility of hydrogen in olivine implies that hydrogen contributes to electrical conductivity. In this earlier suggestion I used the Nernst-Einstein relation for electrical conductivity that allows us to calculate electrical conductivity from the concentration and mobility of charged species, viz., σ cond = f

Dcq 2 RT

(6)

where σcond is electrical conductivity, f is a constant about unity, D is the diffusion coefficient of a charged species, c is the concentration of the charged species, q is the electric charge of the species, R is the gas constant and T is temperature. Given this relation, one can calculate electrical conductivity from the known concentration of the charged species and its diffusion coefficient. The simplest assumption that I made in that paper was (i) all the hydrogen atoms dissolved in mineral (olivine) contribute to electrical conductivity equally (i.e., c is the total concentration of hydrogen), and (ii) the (chemical) diffusion coefficient reported by Mackwell and Kohlstedt (1990) can be used to calculate the electrical conductivity. This hypothesis was used to interpret geophysical observations by many scientists without experimental tests (e.g., Lizarrale et al. 1995; Hirth et al. 2000; Simpson 2002; Tarits et al. 2004; Evans et al. 2005; Simpson and Tommasi 2005). The first experimental test of this hypothesis was reported by Huang et al. (2005) for wadsleyite and ringwoodite (Fig. 4) and a similar study has been performed by Wang et al. (2006) for olivine. These experimental studies have shown the basic validity of Karato (1990)’s hypothesis that hydrogen enhances electrical conductivity, but in detail some of the assumptions of Karato (1990) are not supported by the experimental observations. Two points must be noted. First the hydrogen content exponent, i.e., the value of r in σ cond ∝ CWr is found to be ~0.6-0.7 for olivine, wadsleyite and ringwoodite, whereas the Karato (1990) model predicts r = 1. Second, for olivine and wadsleyite, where we have data on electrical conductivity and diffusion, the activation enthalpy of diffusion is considerably higher than that of electrical conductivity (Wang et al. 2006). These observations show that the charge carrier

350

Karato

Figure 4. The influence of water content on electrical conductivity in wadsleyite (after Huang et al. 2005). The numbers next to each line are the temperatures (K).

is not the majority of hydrogen atoms (protons) dissolved in minerals, i.e., protons trapped at M-site vacancies, (2H)×M , but rather some other species with higher mobility (lower activation enthalpy). An immediate conclusion is that one cannot calculate the electrical conductivity of a mineral from the solubility and diffusion coefficient of “hydrogen” as originally proposed by Karato (1990). Huang et al. (2005) proposed that electrical conductivity is due to the motion of free protons that can be produced by the ionization reaction (2H)×M ↔ H′M + H i

(7)

where H′M is a M-site vacancy that contains one proton, and H• is a free proton. If free proton is easier to move, as expected, then the activation enthalpy for conduction will be smaller than that of diffusion of (2H)×M . Therefore I consider that a better hypothesis for the hydrogenenhanced electrical conductivity is that the charge carrier is free proton, viz., σ cond ∝ [H i ] ⋅ µ H i ∝ CWr ∝ fHr 2 O

(8)

where CW is the water content, [H•] is the concentration of free protons and µH• is their mobility and r is a constant that depends on the types of the dominant charged defects. Here I made a relation that CW ≈ [(2H)×M ] ∝ fH 2 O . When the charge neutrality condition is [H′M ] = [ Fe iM ] as suggested by Karato (1989a) (see also Mei and Kohlstedt 2000a), then this model predicts r = ¾, whereas if the charge neutrality condition is the same as in the water-poor conditions, i.e., ′′ ] = [ Fe iM ], r = ½. These predictions are consistent with the experimental observations of 2[ VM r = 0.6-0.7. An immediate implication of this model is that anisotropy in conductivity can be different from the anisotropy of chemical diffusion of hydrogen because anisotropy of diffusion of a free proton may be different from that of (2H)×M. In many previous studies, anisotropy in conductivity predicted from the Karato (1990) model was used to interpret geophysical observations (e.g., Lizarrale et al. 1995; Hirth et al. 2000; Simpson 2002; Evans et al. 2005; Simpson and Tommasi 2005). I conclude that there is no strong mineral physics basis to support this assumption. Direct experimental studies on electrical conductivity under hydrous conditions are needed to determine the possible anisotropy in electrical conductivity.

Remote Sensing of Hydrogen in the Earth’s Mantle

351

I should also note that there are several issues that also need more attention before electrical conductivity can be used as a sensor for hydrogen content with high confidence. First, so far we have made a simple assumption that the electrical conductivity is dominated by volumetrically dominant minerals such as olivine (in the upper mantle) and wadsleyite or ringwoodite (in the transition zone). The validity of this assumption needs further examination. In the shallow upper mantle, the hydrogen solubility in aluminous orthopyroxene is significantly higher than that of olivine (e.g., Rauch and Keppler 2002). In this case, the electrical conductivity of the upper mantle could be controlled by orthopyroxene. Similarly, the influence of garnet must be examined for the deep upper mantle and the transition zone. Second, so far little attention was paid to the influence of grain-size on electrical conductivity. ten Grotenhus et al. (2004) showed that for hydrogen-poor iron-free olivine (forsterite), the electrical conductivity increases with decreasing grain-size. If the grain-boundary effect is important, then the anisotropy of conductivity will be weaker than expected from single crystal data. Our preliminary data on olivine indicate that grain-boundary effect is weak for iron-bearing olivine for the grain-size of ~ 10 µm to ~ 1 mm, but the details are not known yet. Electrical conductivity may also be enhanced by hydrogen through the enhancement of diffusion of some ionic species (e.g., Karato 1990). This is an obvious possibility since Karato et al. (1986) inferred the enhancement of diffusion by hydrogen (see also Mei and Kohlstedt 2000a). Hier-Majumder et al. (2005) investigated this through an experimental study, and concluded that this effect is minor compared to the direct influence by proton conduction. I should also mention an important technical issue in the experimental study of hydrogenrelated properties. In a paper by Xu et al. (1998b), they reported that wadsleyite and ringwoodite have much higher electrical conductivity than olivine compared at similar pressure and temperature. Similar studies were made on cation diffusion (e.g., Farber et al. 1994; Chakraborty et al. 1999). It was considered (incorrectly) that the electrical conductivity and diffusion in wadsleyite (and ringwoodite) are intrinsically higher (faster) than those processes in olivine. However, later studies in my lab clearly showed that the contrast in electrical conductivity between these minerals is almost entirely due to the (then unrecognized) difference in hydrogen content in these minerals: compared at the same pressure, temperature and hydrogen content, electrical conductivity in olivine is similar to that of wadsleyite (or ringwoodite). Since the affinity of hydrogen to wadsleyite (or ringwoodite) is much higher than that of olivine, wadsleyite and ringwoodite tend to dissolve more hydrogen from the surrounding medium during high-pressure experiments than olivine, which enhances electrical conductivity and other defect-related properties enormously. In any experimental study on hydrogen-related properties, one must determine the hydrogen content in the sample both before and after each experiment.

Seismic properties Seismic wave velocities and attenuation. The most obvious effect of hydrogen on seismic wave velocities is that the addition of hydrogen reduces seismic wave velocities by reducing the bond strength. The magnitude of such an effect was calculated by Karato (1995) assuming that the incorporation of hydrogen in a crystal will create a region with zero elastic modulus, and the average elastic modulus of a mineral containing these weak regions is calculated by a homogeneous strain model (i.e., Voigt model). This model predicts ~1% reduction for 1 wt% addition of water, which roughly agrees with the later experimental observations (e.g., Inoue et al. 1998; Jacobsen et al. 2004), see also Jacobsen, this volume for more detail of this effect). This is an effect on bond strength at an elastic limit. In other words, this is an effect on seismic wave velocities at infinite frequency, viz., v ∞ = v0∞ (1 − A ⋅ CW )

( 9)

where CW is water content (in wt%), v0∞ is the unrelaxed velocity for hydrogen-free mineral,

352

Karato

and A is a constant (~1-2). An immediate conclusion from this analysis is that this direct effect is important only for very large values of water content, ~0.1-1 wt%. If the water content is less than 0.1 wt%, the influence of water on seismic wave velocity is less than ~0.5%, so is not important. The water content in the upper mantle is less than ~0.1 wt%. Consequently, Karato (1995) concluded that this effect is not very important in the upper mantle, although this effect is important in the transition zone if water content is as high as ~1 wt%. In addition to the effects on unrelaxed velocity, hydrogen may affect seismic wave velocities through its effect on anelasticity. The importance of anelasticity in seismic wave propagation was pointed out by Gueguen and Mercier (1973), Karato (1993), and Minster and Anderson (1980). Karato (1995, 2003) suggested that the effect of hydrogen on seismic wave velocity is mostly due to the enhancement of anelasticity (and through changes in anisotropy, see later part of this paper). Experimental evidence for this is provided by Jackson et al. (1992) who showed that anelasticity of a dunite sample (Åheim dunite) is reduced significantly after the majority of water is removed from the sample by heat treatment. Since the heat treatment used to remove hydrogen resulted in other complications, Jackson et al. (1992) did not conclude that the change in anelasticity is due to the difference in hydrogen content. However the observations by Jackson et al. (1992) are similar to those by Chopra and Paterson (1984) on exactly the same dunite who found that the creep strength was significantly increased after the heat treatment. The results of Chopra and Paterson (1984) were supported by later works on synthetic olivine polycrystals (e.g., Karato et al. 1986; Mei and Kohlstedt 2000a,b), and are interpreted as a result of change in hydrogen content. Therefore the simplest explanation for the observation by Jackson et al. (1992) is to attribute it to the change in hydrogen content. As such the experimental evidence on the influence of hydrogen on anelasticity is preliminary, and the exact functional form by which seismic wave attenuation depends on hydrogen content is not known. However, based on the frequency dependence of seismic wave attenuation, and the experimental results on the relationship between hydrogen content and various kinetic processes, one can propose a plausible model. Let us recall that in most of solid-state elastic wave attenuation at low frequencies and high-temperatures, attenuation will follow the following form (e.g., Karato and Spetzler 1990), Q −1 ∝ ω− α

(10)

with α = 0.3±0.1 (e.g., Jackson 2000). Now since attenuation is a non-dimensional quantity, such frequency dependence must imply that the seismic wave attenuation depends on relaxation time as Q −1 ∝ ( ωτ )

−α

∝ τ−α

(11)

The relaxation time depends on the rate of some kinetic processes such as the motion of dislocations or grain-boundary sliding. Hydrogen is known to enhance all of these processes, and in most cases the characteristic time of these processes depend on hydrogen content as τ ∝ CW−r

(12)

where r is a constant that depends on the process (Table 1). Consequently, I assume that Q −1 ∝ CWαr

(13)

So in summary, the functional form for the dependence of seismic wave velocity on hydrogen content is πα −1 ⎡ 1 ⎤ v = v0∞ (1 − A ⋅ CW ) ⎢1 − cot ⋅ Q ( CW ) ⎥ 2 ⎣ 2 ⎦

(14)

Remote Sensing of Hydrogen in the Earth’s Mantle

353

Table 1. The dependence of various kinetics processes on hydrogen content. The values of r for various properties are shown where a power-law relation R ∝ CWr is assumed (R: the property e.g., strain-rate, diffusion coefficient). property

r

diffusion coefficient

~1.0 (oxygen, quartz)1

diffusion creep

~0.7-1.0 (olivine)2

dislocation creep

~1.0-1.2 (olivine, quartz)3,4,5

grain-boundary mobility

~2.0-2.3 (for wadsleyite)6 (r > 1 for olivine)7

electrical conductivity

~0.7-0.9 (for olivine, wadsleyite, ringwoodite)8

1

2

Farver and Yund (1991); Mei and Kohlstedt (2000a); 3 Kohlstedt et al. (1995); 4 Mei and Kohlstedt (2000b); 5 Karato and Jung (2003); 6 from grain-growth (Nishihara et al. 2006); 7 from dynamically recrystallized grain-size (Jung and Karato 2001a); 8 Huang et al. (2005), Wang et al. (2006)

The relations (13) and (14) provide a basis for inferring hydrogen content from seismic wave velocity and attenuation. Figure 5 illustrates how hydrogen content affects seismic attenuation for a range of parameters. I should also discuss the issue of grain-size sensitivity of seismic wave attenuation. In contrast to rather preliminary observations on the influence of hydrogen, there have been solid results on the grain-size sensitivity of seismic wave attenuation in olivine aggregates (e.g, Tan et al. 1997, 2001; Gribb and Cooper 1998; Cooper 2002; Jackson et al. 2002). These studies showed Q −1 ∝ L− αs

(15)

with α~0.25 and s~1 where L is the grain-size. I note that the functional form of grain-size dependence is the same as that of water content dependence, both cases are represented by a power-law formula, Q−1 ∝ Xβ where X is water content or inverse grain-size with β = 0.25-0.50. Therefore the influence of grain-size cannot be distinguished from that of hydrogen from the observations on attenuation. Therefore the interpretation of the attenuation tomography needs to be made on the basis of some other considerations such as geodynamic plausibility (e.g., Shito et al. 2006). I should emphasize that although more experimental observations are available

Figure 5. The relation between seismic wave attenuation and water content for a power-law relation (Eqn. 13) for α = 0.25 and r = 1-2. Jackson et al. (1992)’s data indicate that dried and undried (“wet”) dunite have Q values that are different by a factor of ~2. For exactly the same sample, this drying procedure changes the strainrate by a factor of ~10-20 (Chopra and Paterson 1984). Using the water content dependence of strain-rate, this can be translated into the difference in water content of a factor of ~10-20. Therefore I conclude that the results by Jackson et al. (1992) are consistent with the relation (13) with r = 1-2.

354

Karato

for the grain-size sensitivity than for hydrogen sensitivity, the degree to which grain-size influences attenuation in Earth’s interior is likely much smaller than that of a likely effect of water. This is simply due to the fact that grain-size in a typical upper mantle has a narrow range, ~3-10 mm (e.g., Mercier 1980; Karato 1984) whereas water content has a broad range, ~10−4-10−1 wt%, depending on the geological processes such as melting (Karato 1986; Hirth and Kohlstedt 1996). The expected change in Q corresponding to these changes will be a factor of ~1.3 for grain-size and a factor of ~6-30 for hydrogen (water) content. Therefore, as a first approximation, it is safe to ignore the influence of grain-size compared to the influence of hydrogen. A key issue in applying these relations to infer the hydrogen content in Earth’s mantle is how to distinguish the effects of hydrogen from other effects. The issue of partial melting is separately discussed in a later section, and the main conclusion is that in most of Earth’s upper mantle, there is no clear evidence so far to suggest any significant effects of partial melting on seismic wave velocities or attenuation. Here I will briefly review the methodology to distinguish the influence of hydrogen from that of temperatures and major element chemistry (for detail, see Shito et al. 2006). Key points are (i) the major element chemistry has relatively small influence on seismic wave attenuation, but has important effects on seismic wave velocities, whereas (ii) hydrogen has a large effect on seismic wave attenuation, and (iii) temperature has effects on both seismic wave velocity and attenuation. Consequently, when one plots the velocity and attenuation anomalies in a certain region, anomalies due to major element chemistry will show large variation in velocities but not in attenuation. If anomalies are due mainly to the variation in hydrogen content, then there will be large lateral variation in attenuation with relatively small velocity anomalies. Finally, if anomalies are due to temperature variation, both attenuation and velocities will show some variations (Fig. 6). To determine the hydrogen content as well as other variables from tomographic data, we need to perform a formal inversion of the observed data in terms of unknown parameters. Because there are a large number of unknowns, it is important to use as many independent observations as possible to constrain unknowns. Generally, if anomalies in seismic wave velocities, attenuation and density are obtained from one region, then one can write a general equation, m

m

j =1

j =1

δ log X i = ∑ Aij ⋅ δY j = ∑

Figure 6. A schematic diagram showing the difference in the influence of temperature, water (hydrogen) content and major element chemistry on seismic wave velocities and attenuation (after Shito et al. 2006).

∂ log X i δY j ; ∂Y j

i = 1… n

(16)

Remote Sensing of Hydrogen in the Earth’s Mantle

355

which can be inverted for five unknowns. Here δlogXi (I = 1…n) is a set of observed data such as (δ log VP δ log VS δ log QP−1 δ log QS−1 δ log ρ) and δYj (j = 1…m) is a set of unknowns such as (δT δ log X1 … δ log X n δξ), and Aij ≡ ∂logXi/∂Yj is a matrix made of partial derivatives of seismological observations with respect to physical/chemical variables (e.g., ∂logVP/∂T). This matrix is made of elements that need to be evaluated based on mineral physics. Note that in general, one has a large number of unknowns particularly because there are a large number of elements to specify the chemical composition, and the solution is non-unique. Matsukage et al. (2005) analyzed the compositional data and elasticity of constituent minerals in mantle peridotites and concluded that in most cases the chemical variation of mantle peridotite can be specified in terms of a small number of parameters. In the simplest case of peridotites in the oceanic environment, a single parameter, i.e., Mg# (mole fraction of Mg relative to Fe) is enough to specify the compositional dependence of seismic wave velocities of peridotite. With this simplification (and the assumed null effects of partial melting), one can invert seismic anomalies in terms of anomalies in water content, temperature and major element chemistry if at least three data are obtained for each point. The details of the inversion scheme are described in Shito et al. (2006). The quality of such an inversion depends strongly on the quality of seismological data and of the mineral physics-derived partial derivatives. In general, the inversion is non-linear because the values of matrix elements depend on the values of unknowns such as hydrogen content and temperature. Seismic anisotropy is in most cases caused by the lattice-preferred orientation (LPO) of elastically anisotropic minerals such as olivine (e.g., Nicolas and Christensen 1987; Chapter 21 of Karato 2006a). The possible influence of hydrogen on LPO of olivine was suggested by Karato (1995). This hypothesis was proposed based on the experimental study by Mackwell et al. (1985) who showed that the effect of hydrogen to enhance plastic deformation of olivine is anisotropic: deformation by slip systems with b = [001] is more enhanced by hydrogen than deformation by b = [100] slip systems. Consequently, Karato (1995) postulated that at high water fugacity conditions, slip systems with b = [001] (e.g., [001](010), [001](100)) might become the dominant (easiest) slip system, and consequently the LPO will be different from that usually observed at low water fugacity conditions. This hypothesis has been tested by experimental studies in my lab (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006). Based on these results, a variety of olivine LPOs have been identified (see Fig. 7), and each LPO has its own anisotropic signature (Table 2). A fabric transition is a commonly observed phenomenon in a material where multiple slip systems operate that have relatively small contrast in strength (e.g., quartz, Lister 1979). A fabric transition will occur when the relative easiness of two slip systems change, i.e., σ σ ⎛ T ⎞ ⎛ T ⎞ , ,C ⎟ = ε2 ⎜ , ,C ⎟ ε1 ⎜ W W ( ) ( , ) ( ) ( , ) T P T P T P T P µ µ ⎝ m ⎠ ⎝ m ⎠

(17)

where ε1,2 is strain-rate with the slip system 1, 2, T is temperature, Tm(P) is melting temperature, σ is stress, and µ(T,P) is shear modulus. Consequently, the conditions for a fabric transition will be given by σ ⎛ T ⎞ , , CW ⎟ = 0 F⎜ ( ) µ ( , ) T P T P ⎝ m ⎠

(18)

A boundary between different fabric types is characterized by a hyper-surface in the space defined by three variables, [T/(Tm(P)), σ/(µ(T,P)), CW]. Several points may be noted on the nature of fabric transitions. First, because the fabric boundary is defined by the relative easiness of two slip systems, strain-rate does not explicitly enter the equation for a fabric boundary. In other words, the fabric diagram determined for

356

Karato Table 2. Seismological signature of various olivine LPOs corresponding to the horizontal shear. fabric

fast S-wave polarization

vSH/vSV

A-type

parallel to flow

>1

B-type

normal to flow*

>1

C-type

parallel to flow

<1

E-type

parallel to flow

>1 (weak)

* This relation holds also for the vertical shear.

Figure 7. Various olivine deformation fabrics found in the experimental studies by (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006); pole figures on the equal area projection on the lower hemisphere. See Plate 2 for color figure.

a certain range of strain-rates can be applied to slower strain-rates without any (explicit) problems. Second, I note that a conventional power-law formula as applied to olivine does not predict stress-induced fabric transformation. The most detailed study on dislocation creep in olivine single crystal is the work by Bai et al. (1991) who showed a highly complicated creep laws for olivine single crystals, but a common feature they reported is that the stress exponent is common to all slip systems, n~3.5. In this case, Equation (18) will not contain stress as a variable and one should not expect stress-induced fabric transformations.

Remote Sensing of Hydrogen in the Earth’s Mantle

357

To see these points, let us consider the following simple power-law creep constitutive relationship, ⎛ H* ⎞ n ε1,2 = A1,2 exp ⎜ − 1,2 ⎟ ⋅ σ 1,2 ⎜ RT ⎟ ⎝ ⎠

(19)

where A1,2 are the pre-exponential factors, H1*,2 are the activation enthalpy, and n1,2 are the stress exponent for the 1,2 slip systems respectively. Equating ε1 = ε 2, one gets the conditions for the fabric boundary, viz.,

( n1 − n2 ) log σ =

H1* − H 2* A − log 1 A2 RT

(20)

This equation does not contain strain-rate, so the fabric boundary does not explicitly depend on strain-rate. Also if n1 = n2 as Bai et al. (1991) showed, then the boundary will be given by [( H1* − H 2* ) / RT ] − log( A1 / A2 ) = 0 and would not depend on stress. The latter point is inconsistent with some of the experimental results including Carter and Avé Lallemant (1970) and the B- to C-type, B- to E-type or B- to A-type transition observed in our study (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006). Therefore there is a need to go beyond a simple power-law constitutive relation to interpret the observed fabric transitions. One way is to incorporate a subtle deviation from the power-law formula observed in some of the experimental studies at high stress levels. This power-law breakdown occurs beyond a certain stress (>100-200 MPa), and can be explained by the stress dependence of activation enthalpy, H*(σ) ⎛ H * ( σ ) ⎞ n1,2 ε1,2 = A1,2 exp ⎜ − 1,2 ⎟⋅σ ⎜ RT ⎟⎠ ⎝

(21)

A simple case is a linear stress dependence, viz., H1*,2 = H1*,02 − B1,2 ⋅ σ

(22)

where B1,2 is a constant related to the resistance for dislocation motion (e.g., the Peierls stress). With this formula, Equation (20) becomes

( n1 − n2 ) log σ =

H1*0 − H 2*0 B1 − B2 A − σ − log 1 A2 RT RT

(23)

In most cases, n1 = n2 = 2, so that one has

( H1*0 − H 2*0 ) − ( B1 − B2 ) σ − RT log A1 = 0 A

(24)

2

A formula similar to (24) has been shown to be consistent with the observations on B- to C-type fabric transition observed in the lab as well as in naturally deformed peridotites (Katayama and Karato 2006). This type of transition may be classified as a stress- (and temperature-) induced fabric transition. How about the hydrogen effect? To include the effect of hydrogen on deformation in addition to the stress-dependence of activation enthalpy, the relation (19) can be extended to

(

)

⎛ H* ( ) ⎞ n ε1,2 = A1,2 + C1,2 ⋅ f Hr1,2O ⋅ exp ⎜ − 1,2 σ ⎟ ⋅ σ 1,2 2 RT ⎠ ⎝

(25)

where fH2O is the fugacity of water, C1,2 is a constant, and r1,2 is a non-dimensional constant that

358

Karato

depends on the mechanisms of hydrogen weakening. From this formula, one obtains

( n1 − n2 ) log σ =

A1 + C1 f Hr12O H1* ( σ ) − H 2* ( σ ) − log RT A2 + C2 f Hr2 O

(26)

2

This is a more general form for a fabric boundary that depends on temperature, stress and hydrogen content. If n1 = n2, and the stress dependence of activation enthalpy is weak, then we will have a hydrogen-induced fabric transition for which the boundary will follow 0=

A1 + C1 f Hr12O H1*0 − H 2*0 − log RT r A2 + C 2 f H22O

(27)

Both stress-induced transitions and hydrogen-induced transitions have been identified in olivine. The B- to C-type (and the B- to E-type, B- to A-type) transition is the stress- (and temperature-) induced transition, and the A- to E-type (and the E- to C-type) is the hydrogeninduced transition. The [T/(Tm(P)), σ/(µ(T,P)), CW] space under which various fabrics dominate is shown in Figure 8. Five types of olivine fabrics have so far been identified (see Fig. 7 that shows four of them). Among them A-, B-, C- and E-type fabrics are particularly relevant for Earth. Given the data on the distribution of crystallographic orientation, one can calculate the macroscopic elastic constants of aggregates. Seismic anisotropy resulting from these fabrics can readily be calculated from these elastic constants. Among many aspects of seismic anisotropy, those frequently used in seismology are summarized in Table 2. Notable points are:

Figure 8. A three-dimensional fabric diagram of olivine in the [T/(Tm(P)), σ/(µ(T,P)), CW] space. Dots represent the data from our lab (P = 0.5-2.0 GPa). There are a large number of data for A-type fabric that are not shown. The boundaries between A- and E-type, E-and C-type fabrics are water (hydrogen) content sensitive but not sensitive to stress nor temperature. In contrast, the boundary between B- and C-type (also B- and Eor A-type) is sensitive to stress and temperature. At low water content conditions (i.e., depleted lithosphere), A-type fabric dominates. As water (hydrogen) content increases, E-type and then C-type fabric dominates at relatively high temperatures. When temperature decreases, the domain of B-type becomes important. At low temperature (T/Tm < 0.5), B-type dominates in most cases. The results by Couvy et al. (2004) obtained at P = 11 GPa are also plotted after normalization by T/(Tm(P)), σ/(µ(T,P) (a rectangular normal to the T/Tm axis in high water (hydrogen) content region). It is seen that their results lie in the region for the C-type fabric.

Remote Sensing of Hydrogen in the Earth’s Mantle

359

(i) The fast olivine [100] axis, is normal to the flow direction for the B-type fabric, and consequently, the direction of polarization of the faster S-wave (which is reported as a shear-wave splitting observations) is normal to the flow direction. When deformation is horizontal shear, then the direction of polarization of the faster S-wave will be orthogonal to the macroscopic shear direction. Similarly, when shear occurs in a vertical plane, then the direction of polarization of faster S-wave will be in the plane but orthogonal to the flow direction. (ii) The C-type fabric will show a similar anisotropy signature to the A-type fabric in terms of shear wave splitting but vSH /vSV anisotropy will be different because the fast olivine axis is normal to the shear plane in this fabric as opposed to the case of the A-type fabric. (iii) The E-type fabric is qualitatively similar to the A-type fabric. However, because the fastest axis ([100] axis) and the slowest axis ([010] axis) are on the shear plane, the amplitude of shear wave splitting corresponding to the horizontal flow will be large for this fabric, and the amplitude of vSH /vSV anisotropy will be small. These features can be compared with seismological observations to obtain some insights into the distribution of physical and chemical conditions in Earth particularly the distribution of hydrogen. Some of the results will be discussed in the next section. I should also comment on two recent papers in which similar fabric transitions were reported but different causes were suggested. First, Holtzman et al. (2003) reported that olivine [001] axis is subparallel to the maximum elongation and olivine [010] axis is normal to the shear plane (B-type fabric) when olivine is deformed with a small amount of melt that contains a large amount of chromite or FeS. They noted that olivine fabric in their sample has strong [010] along the direction normal to the shear plane, but olivine [100] and [001] axes assume a girdle when strong shear bands are not formed (their results are different from those by Zimmerman et al. (1999) who observed a typical A-type fabric in which olivine [100] direction has a peak at a direction subparallel to the shear direction). Clear shear bands were formed in these cases (due presumably to a smaller compaction length due to the presence of chromite or FeS). After clear shear bands are formed (at larger strains), olivine [001] peak starts to strengthen along the direction normal to the shear direction. Holtzman et al. (2003) interpreted this evolution of olivine fabric in terms of deformation geometry. In their experimental setup, a significant compression component exists and therefore extrusion of sample occurs. The “B-type fabric” they observed is likely due to the anisotropic extrusion: more extrusion normal to shear direction Holtzman et al. 2003). However, the reason for this deformation geometry is due to an artifact caused by the sample geometry. First of all, extrusion is the result of compression that would not occur in truly simple shear deformation. Furthermore, the anisotropic extrusion is likely a result of oblate shape of their sample. Otherwise there is no obvious reason for the selective extrusion normal to the shear direction. Consequently, I conclude that the results reported by Holtzman et al. (2003) on LPO of olivine in partially molten olivine are due to experimental artifacts and the relevance of their observation to seismic anisotropy is highly questionable. In fact, if such an olivine fabric develops beneath a mid-ocean ridge, one would expect an anisotropic structure in the oceanic lithosphere that is totally inconsistent with observations. I conclude that the results of Holtzman et al. (2003) are unlikely to be relevant to Earth science. Second, Mainprice et al. (2005) argued that the fabric reported by Couvy et al. (2004) is due to a pressure-induced fabric transition. However, their samples contained a large amount of hydrogen (on the order of ~2000 ppm H/Si, see Couvy et al. 2004), and a comparison of the deformation conditions of their experiments with those by our group (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006) shows that the samples that show C-type fabric in Couvy et al. (2004)’s experiments were deformed precisely in the

360

Karato

[T/(Tm(P)), σ/(µ(T,P)), CW] conditions where the C-type fabric is developed (see Fig. 8). There have been similar results using D-DIA apparatus showing C- or B-type fabrics under highpressures but with a low water content (Ratteron, Li and Weidner, private communication, 2005). However, in these cases, the magnitude of deviatoric stress is high (~500-900 MPa). The dominance of slip systems involving b = [001] in olivine under high-stress conditions has been well known (e.g., Carter and Avé Lallemant 1970). Therefore I conclude that the observation by Couvy et al. (2004) can be naturally attributed to the high hydrogen content, and those by Ratteron, Li and Weidner are likely due to the high stresses. Obviously, there is a possibility that pressure might change the rate of deformation by different slip systems differently than the homologous temperature and normalized stress scaling would imply. However, in order to demonstrate intrinsic pressure effects on deformation fabrics, one needs to show different fabrics for two samples deformed at different pressures but otherwise nearly identical conditions (including stress levels). Such a study has not been reported to my knowledge. I therefore consider that the hypothesis of pressure-induced fabric transition in olivine proposed by Mainprice et al. (2005) and Ratteron, Li and Weidner has little experimental support and remains highly speculative at this stage. Topography and sharpness of discontinuities. Dissolution of hydrogen reduces the free energy of a material. Experimental studies show that a significantly larger amount of hydrogen can be dissolved in wadsleyite than in olivine (e.g., Young et al. 1993; Kohlstedt et al. 1996). Therefore the dissolution of hydrogen will expand the stability field of wadsleyite relative to that of olivine (the depth of “410-km” boundary will be shallower if a large amount of hydrogen is present). A more subtle effect is the change in the width of the “410-km” discontinuity with hydrogen content. When the system is considered to be a binary system, i.e., Mg2SiO4-Fe2SiO4, there is a range of pressure (at a fixed temperature) during which the phase transformation is completed. When the upper and the lower boundaries are affected by hydrogen differently, then the width of the boundary will be modified by hydrogen. The degree to which the dissolution of hydrogen affects the free energy depends on the atomistic mechanisms of hydrogen dissolution. In his paper on this topic, Wood (1995) used a model of hydrogen dissolution in wadsleyite by Smyth (1987, 1994) and in olivine by Bai and Kohlstedt (1993). These models are not consistent with the recent experimental observations, and here I use a model that is consistent with the current experimental observations. In my model, the dissolution mechanism is identical for both olivine and wadsleyite, hydrogen is dissolved mainly as (2H )×M, and only the magnitude of solubility is different between the two phases. In calculating the phase diagram, I use a simplifying assumption that the system under consideration can be treated as an ideal mixture of three components, Mg, Fe and H for the M-sites. In this approximation, the chemical potential of each component is a function of concentration of each component as µ ij = µ ij0 + RT log xij

(28)

where µij is the chemical potential, µ ij0 is the chemical potential for a pure material and xij is the mole fraction of M-sites for a component i (Mg, Fe and H) in phase j (olivine and wadsleyite). Chemical equilibrium demands 0 µ io0 + RT log xio = µ iw + RT log xiw

for i = Mg, Fe, and H

(29)

where subscript o(w) refers to olivine (wadsleyite). The M-site is shared by three elements so that

∑ xio =∑ xiw = 1 i

(30)

i

There are six unknowns (three components in a two phase system) with five equations (Eqns. 29 and 30). Another necessary parameter is the water fugacity that determines the hydrogen

Remote Sensing of Hydrogen in the Earth’s Mantle

361

content (in olivine, xHo). In the first calculation, I fixed the water fugacity, i.e., I assumed an open system behavior. I use the various values of hydrogen content in olivine (corresponding to the various values of water fugacity) and the ratio of water content in wadsleyite to olivine is fixed to be xHw/xHo = 10, which is assumed to be independent of pressure (this is justifiable because in my model, the mechanism of hydrogen dissolution is identical between olivine and wadsleyite, but not valid for the model assumed by Wood (1995)). Thermodynamic data for Mg and Fe components of olivine and wadsleyite are from Akaogi et al. (1989). The results are shown in Figure 9a, showing that the depth of transition decreases with the increase in hydrogen content. Both the upper and the lower boundaries move nearly the same amount, and the width of the transition does not change with hydrogen in this case (open system behavior). The shift of the transformation pressure by water is given by ⎛ ∂z ∆z = ⎜ ⎝ ∂CW

⎞ ⎟ ⋅ CW ⎠

(31)

with (∂z/∂CW ~ 30 km/wt% where water content is water in wadsleyite. In general, where both water content and temperature vary with lateral position, the depth to the “410-km” discontinuity will change as ⎛ ∂z ⎞ ⎛ ∂z ⎞ ∆z = ⎜ ⎟ ⋅ CW + ⎜ ⎟ ⋅ ∆T ⎝ ∂T ⎠CW ⎝ ∂CW ⎠T

(32)

where (∂z/∂T)CW ≈ 0.13 km/K is the temperature dependence of transformation depth from olivine to wadsleyite (Akaogi et al. 1989). When a phase transformation occurs in a closed system with hydrogen-under-saturated condition, then a progressive phase transformation will change the concentrations of hydrogen in each phase according to the degree of transformation (water fugacity will change with the progress of a phase transformation). Consequently, when the pressure just reaches the

Figure 9. Influence of hydrogen on the olivine-wadsleyite phase boundary. (a) Solid lines: results for an open system. Broken curves: results for a closed system with the total water content of 50% saturation. Addition of water expands the stability field of wadsleyite relative to that of olivine, leading to a shift of the phase boundary, ∆z = (∂z/∂CW)·CW with (∂z/∂CW) ~30 km/wt%. When the transformation occurs at a fixed total hydrogen content, then the broadening of the transition occurs due to the presence of hydrogen. (b) Variation of the width of the olivine-wadsleyite boundary with hydrogen saturation (saturation in olivine corresponds to ~0.1 wt% water or 1.5 × 104 ppm H/Si).

362

Karato

minimum pressure at which wadsleyite phase appears, most of the hydrogen is in olivine, whereas the hydrogen concentration in olivine decreases as the volume fraction of wadsleyite increases. Therefore the pressure at which olivine starts to transform to wadsleyite is controlled by the initial water content in olivine, and is significantly lower than the pressure at which this transformation would start in a hydrogen-free system. In contrast, the pressure at which this transformation is completed is controlled by the hydrogen content in olivine at this stage that is significantly lower than the hydrogen content at the beginning of the olivine to wadsleyite transformation (exactly the same can be applied to a case when a phase transformation starts from wadsleyite). A phase diagram for this more realistic case can be calculated from the results for a fixed hydrogen content by incorporating the mass balance requirement (see broken curves in Fig. 9a). As a result of this gradual change in hydrogen content in each phase, there is a broadening of the binary loop as first pointed out by Wood (1995). The width of the olivine to wadsleyite binary loop calculated in this way is shown in Figure 9b. Note that although the width of the boundary increases with hydrogen content similar to the model by Wood (1995), the degree to which the width increases with hydrogen content is somewhat different between the present model and the model by Wood (1995). This difference is caused mainly by the difference in the model of hydrogen dissolution and the choice of thermodynamic parameters. A few comments are in order here. (1) The model results for an open, saturated system agree with the experimental observations by Chen et al. (2002) and Smyth and Frost (2002) or an open, water-saturated system. But the closed system behavior predicted by the model has not been tested experimentally. Chen et al. (2002) and Smyth and Frost (2002) compared their results for an open system directly with those by Wood (1995) for a closed system and discussed that their results did not agree with those by Wood (1995). This is misleading. The variation of the width of transformation by hydrogen content predicted by Wood (1995) model occurs only in a closed system but not in an open system. (2) The validity of the assumption of a closed system behavior in real Earth is not necessarily obvious. If the amount of water (hydrogen) in Earth’s transition zone exceeds a critical value (~0.05 wt%, see the next section), the phase transformation from wadsleyite to olivine in a upwelling current could cause partial melting (Bercovici and Karato 2003). In this case, hydrogen can be removed from the system during the phase transformation and the assumption of a closed system behavior will be violated. (3) A comparison of the model results on the width of the “410-km” boundary with seismological observation is not straightforward due to the fact that the actual depth variation of acoustic properties in a phase loop may not a simple function of the volume fraction of each phase (Stixrude 1997). In summary, I conclude that the use of the width of the “410-km” discontinuity to infer the hydrogen content as proposed by Wood (1995) (see also van der Meijde et al. 2003) is subject to large uncertainties. Other techniques such as the use of electrical conductivity or seismic wave velocities (or attenuation) provide more robust estimate of hydrogen contents (e.g., Karato 2003; Huang et al. 2005).

Partial melting? A frequently asked question when water content is to be inferred is what about partial melting? Partial melting may also explain the majority of geophysical anomalies (high electrical conductivity, low seismic wave velocities, high attenuation). In fact, throughout the geophysical literature, these anomalies (low seismic wave velocities, high attenuation and high electrical conductivity) have often been attributed to partial melting (e.g., (Gutenberg 1954; Shankland et al. 1981)). This classical view has been questioned on various grounds. First, based on mineral physics considerations, Gueguen and Mercier (1973) proposed a solid-state mechanism of anelasticity could explain a high attenuation and low velocity zone. The follow-up studies include Minster and Anderson (1980), Karato (1993), Karato and Jung (1998) and Faul and Jackson (2005) who quantified this notion. Although low velocity and high attenuation could be attrib-

Remote Sensing of Hydrogen in the Earth’s Mantle

363

uted to sub-solidus processes, high electrical conductivity was considered to be difficult to be attributed to sub-solidus processes. Consequently, Shankland et al. (1981) argued that it is the high electrical conductivity in the asthenosphere that provides the strongest constraint on the presence of partial melt. Karato (1990) challenged this view by showing that high conductivity can also be attributed to sub-solidus process if high diffusivity and solubility of hydrogen in minerals such as olivine is taken into account. This hypothesis has now been supported by laboratory studies (e.g., Wang et al. 2006). So in short, the current status of our understanding of the role of partial melting is that all of the observed geophysical anomalies on the asthenosphere can be attributed to subsolidus processes if the role of hydrogen is included. In other words, much of the anomalous properties of the asthenosphere can be explained by a high hydrogen content as well as high temperature without invoking the influence of partial melting. One natural question, then, is if both high water content and high temperature are needed to explain geophysical anomalies of the asthenosphere, why doesn’t partial melting occur and contribute to some of these geophysical anomalies in the asthenosphere? To answer this question, one needs to understand the fact that the partial melting in the upper mantle likely occurs as a two-stage process, first (in the deeper part) by hydrogen-assisted melting and then (in the shallow part) as dry (hydrogen-free) melting. A close look at the phase diagram of the upper mantle system shows that the conditions of the asthenosphere corresponds to hydrogen-assisted melting regime, and the degree of melting there is controlled by hydrogen content and is estimated to be ~0.1-0.2% (Plank and Langmuir 1992). A larger degree of melting occurs only in the shallow region near mid-ocean ridges where temperature exceeds the dry solidus, and even if the degree of melting is well constrained to be ~10% (from the thickness of the oceanic crust), the fraction of melt that determines the degree of change in physical properties by partial melting, can be much smaller. Indeed there are strong constraints on the fraction of melt near mid-ocean ridges to be ~0.1% or less from geochemical observations (see e.g., (Spiegelman and Kenyon 1992; Spiegelman and Elliott 1993) and the MELT experiment at the east Pacific Rise failed to detect evidence of melt from seismic anisotropy (Wolfe and Solomon 1998). Also Shito et al. (2004) found that the frequency dependence of seismic wave attenuation in the upper mantle beneath the Philippine Sea is not consistent with the presence of a significant amount of melt. I conclude that the water (hydrogen) content in the asthenosphere (~0.01 wt%) is low in a petrologic sense: with this amount of water, a significant amount of melt does not exist in the majority of the asthenosphere to cause detectable change in seismic wave propagation or electrical conductivity (melt fraction is less than ~0.2%). If the degree of melting is at this level, a large fraction of total hydrogen will stay in solid minerals. This amount of water (~0.01 wt%) is, however, large compared to the concentration of defects in solid minerals at hydrogen-free conditions. Consequently, many of the solid-state processes including electrical conductivity, seismic wave attenuation are markedly affected by this much of hydrogen in the asthenosphere. In other words, the asthenosphere has anomalous physical properties because of the absence of a large fraction of melt as opposed to the conventional model as first proposed by Karato and Jung (1998). However, it is still possible that a small amount of melt exists in the asthenosphere that causes a velocity reduction without affecting attenuation (e.g., Karato 1977). These effects appear to occur only in limited regions according to the geophysical, petrological and geochemical observations (see also Shito et al. 2006).

SOME EXAMPLES Water content in the transition zone The transition zone minerals such as wadsleyite and ringwoodite are known to have large solubility of hydrogen (to ~3 wt% as water, Kohlstedt et al. 1996). However, the actual

364

Karato

hydrogen content in the transition zone was not well constrained. In fact, based on geodynamic modeling assuming whole mantle convection, Richard et al. (2002) showed that it is hard to keep a high hydrogen content in the transition zone, and the transition zone might be an empty hydrogen reservoir. In contrast, Bercovici and Karato (2003) proposed that melting at ~410km is likely and material circulation may occur in such a way that incompatible elements including hydrogen may be sequestered in the deep mantle by melting. This latter model implies that the hydrogen content in the transition should be higher than that of the upper mantle, whereas a conventional model would imply a similar hydrogen content between the upper mantle and the transition zone. Consequently, the determination of hydrogen content in the transition zone provides a good test for the models of mantle circulation. The hydrogen content in the transition zone has been inferred using seismological observations or electric conductivity. Wood (1995) inferred that hydrogen content of the transition zone is ~0.02 wt% (global average) based on the observed width of the 410-km boundary. Van der Meijde et al. (2003) applied the same method to obtain somewhat different results for the transition zone beneath Europe. Blum and Sheng (2004) used both the velocity anomalies in the transition and the topography of the 410km to infer hydrogen content and temperature anomalies. They concluded that the hydrogen content in the transition zone beneath South Africa is ~0.1 wt%. These two studies suffer major limitations. Wood (1995)’s method is based on a model of hydrogen dissolution that is not consistent with our latest knowledge. Karato (2006b) showed that a model consistent with the latest knowledge of mechanisms of hydrogen dissolution in these minerals gives different results. In addition the inferring the thickness of the boundary is not trivial as discussed in the previous section. The inference of hydrogen content from velocity anomalies and topography of the 410-km is more straightforward although estimating these two parameters from seismology involves some uncertainties (Gu et al. 2003). One major problem with the earlier work by Blum and Sheng (2004) is the ignorance of anelasticity. Huang et al. (2005) determined the relation between electrical conductivity and hydrogen content (plus temperature) for wadsleyite and ringwoodite, and by comparing these results with geophysically inferred electrical conductivity, they inferred the water content in the transition zone to be ~0.1-0.2 wt% beneath the Pacific Ocean (Fig. 10b). Electrical conductivity of the transition zone varies from one region to another. The conductivity of the transition zone in the Philippine Sea region is significantly higher than average Pacific, and also the transition zone beneath Hawaii has a higher conductivity than surrounding regions (Utada et al. 2005). It is likely that a major cause of this regional variation in conductivity is the regional variation is hydrogen content (regional variation in temperature also causes regional variation in conductivity, but the influence of temperature is less important than hydrogen, see Fig. 10a). A more robust analysis of distribution of hydrogen was made by Huang et al. (2006) who used a jump in conductivity at 410-km to infer the jump in hydrogen content. Utada et al. (2003) presented a model of electrical conductivity of the Pacific region that is characterized by a factor of ~10 jump in conductivity across the 410-km discontinuity. A similar jump at ~410km is observed in other regions (e.g., Olsen 1999; Tarits et al. 2004). This jump in conductivity can be translated into jumps in physical and chemical conditions between the transition zone and the upper mantle as +

+

σ410 − σ410

=

3 −1 8 ⎛ 410 + ⎞ 4 ⎜ fH O ⎟ 2 ⎜ ⎟ = ⎜ 410 − ⎟ ⎜ fH O ⎟ ⎠ ⎝ 2 ⎠

⎛ ⎞ ⎟ σ wad ⎜ f O410 2 ⎜ ⎟ − ⎟ ⎜ σ oli ⎜ f 410 ⎟ ⎝ O2

+

−1 8 ⎛ 410 + ⎜ Cw ⎜ − ⎜ C 410 ⎝ w ⎠

⎛ ⎞ ⎟ σ was ⎜ f O410 2 ⎜ ⎟ − ⎟ ⎜ σ oli ⎜ f 410 ⎟ ⎝ O2

3 ⎞4 ⎟ ⎟ ⎟ ⎠

(33)

where σ410 /σ410 is the contrast in electrical conductivity across the 410-km discontinuity (~10), and any quantities with 410+ (410−) means quantities just below (above) the 410-km, i.e., the uppermost transition zone (the lowest upper mantle). In writing this I note that a possible +

−

Remote Sensing of Hydrogen in the Earth’s Mantle

365

(a) Figure 10. (a) A trade-off between temperature and hydrogen (water) effect on the electrical conductivity in wadsleyite (numbers next to each curve are electrical conductivity in Sm−1), (b) a comparison of electrical conductivity profile in the transition zone for various water contents showing ~0.1-0.2 wt% of water content is consistent with the geophysical observations (after Huang et al. 2005).

(b)

temperature jump at 410-km (~50-100 K) yields less than ~30% change in conductivity, so this effect is ignored. From the relation (33), one can calculate the combinations of jumps in water (hydrogen) content and oxygen fugacity across the 410-km that are consistent with the observed jump in electrical conductivity. O’Neill et al. (1993) suggested that if the oxygen to metal ion ratio is constant throughout the transition zone and the upper mantle, then the oxygen fugacity of the transition zone should be significantly (a factor of ~103-104) higher than that of the upper mantle. Figure 11 shows the trade-off between these two factors. In order to explain the observed jump in electrical conductivity, the variation in water (hydrogen) content and oxygen fugacity must satisfy the relations shown by thick lines. Such an estimate contains some uncertainties related to the calibration of hydrogen content based on FT-IR, so I included a range of values corresponding to this uncertainty. This analysis shows that the influence of oxygen fugacity is small (because of a weak dependence of conductivity on oxygen fugacity), and if the water content were the same between the transition zone and the upper mantle, the

2

log10

2

fO410

Karato

fO410

366

log10

410 CW 410 CW

Figure 11. A diagram showing the range of combination of a jump in water content and oxygen fugacity between the upper mantle and the transition zone that is consistent with the observed jump by a factor of ~10 jump in electrical conductivity at ~400 km (after Huang et al. 2006). Two lines correspond to the two different choices of hydrogen content calibration curves.

jump in oxygen fugacity must be on the order of 108-1012, which is unacceptable. Therefore I conclude that the current geophysical observations combined with the available experimental data on electrical conductivity in olivine and wadsleyite strongly suggest that there is a jump in water content across the 410-km discontinuity. Major remaining uncertainties in this approach include (i) the role of secondary phases such as garnets, and (ii) the role of grain-size. Our current data on grain-size indicate only a small effect (a comparison of the conductivity data for olivine with ~10 µm grain-size and ~1 mm grain-size shows less than a factor of ~3 difference). However, the dependence of electrical conductivity of garnet on hydrogen content is not known. Garnet is the second most abundant mineral in both the deep upper mantle and the transition zone, and the determination of electrical conductivity of this mineral is urgent.

Distribution of hydrogen in the upper mantle Some comments on petrologic approach. A large number of petrological or geochemical data are available to infer the distribution of hydrogen in the upper mantle. This approach uses either the hydrogen contents of minerals in mantle rocks (mostly xenoliths) or the water contents in the magmas. There have been numerous publications on this topic (see e.g., Martin and Donnay 1972; Michael 1988; Jambon and Zimmermann 1990; Thompson 1992; Bell and Rossman 1992; Stolper and Newman 1994; Hirth and Kohlstedt 1996; Kurosawa et al. 1997; Wallace 1998; Jamtveit et al. 2001; Katayama et al. 2005), so I will give only a brief review for completeness. These studies have shown the following general trend: (i) MORB source regions have generally low water content (~0.01 wt%), (ii) the source regions of arc magmas have high water content (~1 wt%), and (iii) the source regions of OIB have intermediate water content (~0.02-0.05 wt%). A major advantage of this approach is that this provides a direct measurement of water (hydrogen) content from real rocks, so there is little ambiguity as to what one obtains. However, there are three major sources of uncertainties or limitations in

Remote Sensing of Hydrogen in the Earth’s Mantle

367

this approach. (i) The region from which one can infer hydrogen content is limited. (ii) In case where magmas are used to infer the water content of the mantle, in order to estimate the hydrogen content of the source region from the hydrogen content in magmas, one needs to know the partition coefficient and the degree of melting (and the mode of melting, either batch melting or fractional melting). (iii) In all cases, there is a potential addition of loss or addition of hydrogen during the transport of these rocks from Earth’s interior. The last point is particularly relevant for xenoliths, and I will address this issue in some detail. It is well known that (chemical) diffusion of hydrogen in many minerals is fast (e.g., Kohlstedt and Mackwell 1998; Mackwell and Kohlstedt 1990). Based on this observation, it is often argued that hydrogen may have escaped (or been added) very easily from the minerals during their ascent to the surface. Although this is true, there are potential complications. (i) In cases where hydrogen dissolution is coupled with the dissolution of another species whose diffusion coefficients are low, the majority of hydrogen atoms in the mineral will be preserved. This is the case of hydrogen in orthopyroxene where dissolution of hydrogen is often coupled with dissolution of Al2O3 (e.g., Rauch and Keppler 2002). (ii) Even though hydrogen diffusion is fast, not all hydrogen will escape from a mineral, but some of it will precipitate. This was observed in a laboratory experiment on olivine (unpublished data by Karato 1984). After heating olivine single crystals that contained a large amount of hydrogen (at room pressure with a controlled oxygen fugacity), I found that hydrogen precipitated as water-filled bubbles in addition to some hydrogen loss. In such a case, some fraction of original water is preserved in a different form as the original hydrogen in the crystalline lattice as defects. The precipitated water will react with the host mineral to form hydrous minerals at low temperatures. The time scale of diffusion of hydrogen-related species during the change in P-T conditions can be analyzed based on the experimental data on diffusion coefficients. The characteristic time for the motion of hydrogen-related species with a distance d is given by τ ≈ d 2/π2D where D is the relevant diffusion coefficient. If the time-scale of a given process is much less than this time scale, then a hydrogen-related species will be kept in its original form. In contrast, if the time scale is much larger than the characteristic time, then hydrogen-related species will either escape (or to be added) to a crystal or precipitate in a crystal. Figure 12

Figure 12. A diagram showing the conditions where hydrogen-related features (hydrogen content, lattice-site where hydrogen sits) can be preserved in a process with a given time and space-scale. Thick lines show the characteristic time for hydrogen diffusion, τ, for a characteristic distance, d, using τ ≈ d 2/π2D, where D is the diffusion coefficient of hydrogen (a range corresponds to temperature of 1300 to 1800 K). Shown together are some timescales and length-scales corresponding to laboratory experiments and some geological processes. If (τ,d) for a given process falls below the thick lines, diffusion is efficient and diffusion-loss or change in the speciation will occur, whereas if (τ,d) falls above the lines, these hydrogen-related features will be preserved (after Karato 2006a).

368

Karato

summarizes this analysis. I conclude that for most of xenoliths, the characteristic time (time for xenolith transport) is similar to the characteristic time for (chemical) diffusion. Therefore some hydrogen is preserved but some hydrogen is likely lost or gained during the ascent. It must also be noted, from this diagram, that the characteristic time for diffusion among various lattice-sites is very short (less than a micro second). Therefore it is impossible to preserve the lattice sites where hydrogen occupied at high P-T during the quenching in any experimental studies. This means that the room P-T measurements such as FTIR measurements provide us with the data that reflect crystallographic sites of hydrogen at room P-T or some other P-T during quenching, but these observations do not necessarily reflect the lattice site that hydrogen occupies at high P-T. A great care needs to be exercised in interpreting the results of FTIR measurements at room P-T on a sample prepared at high P and T. Hydrogen in the upper mantle. Several geophysical observations can be used to infer the distribution of hydrogen in the upper mantle. (i) From the electrical conductivity of the asthenosphere determined by Lizarrale et al. (1995) or Evans et al. (2005) (~0.1 S/m), the hydrogen content in the asthenosphere is estimated to be ~0.01 wt% (I assumed the temperature of 1600 K and used our latest experimental results on the relation between hydrogen content and electrical conductivity in olivine (Wang et al. 2006)). This value agrees well with an estimate from the petrologic approach. Obviously, the electrical conductivity in the upper mantle (asthenosphere) varies from one region to another suggesting a regional variation in hydrogen content (as well as temperature). (ii) The onset of seismic low velocity zone as determined by high-frequency body waves using reflected (or converted) waves showed nearly age-independent depth of the onset of a low velocity zone (e.g., Gaherty et al. 1996). This is in contrast to the well-known feature of age-dependent change in lithosphere thickness and velocities as inferred from surface wave studies (e.g., Forsyth 1975; Yoshii 1973). The age-independent sharp change in velocity has been attributed to a sharp contrast in hydrogen content caused by partial melting near midocean ridges (Karato and Jung 1998). I conclude that in order to explain an age-independent velocity jump at ~60-70 km depth detected by short wavelength body wave studies, hydrogeninduced anelasticity provides a good explanation. The age-dependent smooth variation in velocity detected by surface wave studies can be attributed to the temperature effects as has been known long time (e.g., Gueguen and Mercier 1973; Karato 1977; Minster and Anderson 1980 and recent similar works with new parameters Faul and Jackson 2005; Stixrude and Lithgow-Bertelloni 2005). In order to explain both surface wave and body wave observations, one needs to invoke both hydrogen and temperature effects on anelasticity. (iii) The spatial distribution of shear wave splitting in the subduction zone can be interpreted in terms of spatial variation in hydrogen content, stress and temperature. This observation does not provide strong constraint on water content, but does require some water (>20 ppm wt) in the wedge mantle (Kneller et al. 2005). (iv) A joint inversion of velocity and attenuation tomography provides a constraint on the distribution of hydrogen in the upper mantle (Fig. 13, Shito et al. 2006). Variation in hydrogen content by a factor of ~10-100 is found. A hydrogen-rich region is identified in the deep (~300400km) upper mantle beneath the Philippine Sea. This deep hydrogen-rich region is likely caused by the deep transportation of water by hydrous minerals by fast and cold subducting slabs in this region (e.g., Rüpke et al. 2004). (v) The fabric type in the asthenosphere is likely not the A-type fabric as in the lithosphere. The fabric type in the asthenosphere is either E- or C-type according to the results summarized in Figure 8. The fabric type of the asthenosphere can be identified by a close examination of seismic anisotropy in that region, which will provide a useful constraint on the hydrogen content in that layer.

Remote Sensing of Hydrogen in the Earth’s Mantle

369

Figure 13. Results of a joint inversion of velocity and attenuation tomography using mineral physicsbased inversion scheme (after Shito et al. 2006). The tomographic maps shown in Figure 2 are inverted for anomalies in the temperature (δT), major element composition [δ(Mg/Mg+Fe)] and water content anomalies (δlnCW).

Hydrogen in the lower mantle Currently very little is known about the distribution of hydrogen in the lower mantle. Generally, the solubility of hydrogen in lower mantle minerals is lower than those in the upper mantle or transition zone minerals (e.g., Bolfan-Casanova et al. 2000, 2003). One major limitation in inferring distribution of hydrogen in the lower mantle is our lack of any mineral physics data on the relationship between hydrogen content and physical properties of lower mantle minerals. However, we can make some inferences based on our knowledge on these relationships for upper mantle and transition zone minerals. (i) The direct effects of hydrogen on seismic wave velocities, namely the effects of hydrogen on unrelaxed seismic wave velocities, will be negligibly small in the lower mantle. The direct effect of hydrogen can be roughly calculated from a simple model (Karato 1995). Applying this model to lower mantle minerals where the maximum water content is ~0.1 wt% or less, then one will conclude that the maximum degree of hydrogen to change the seismic wave velocities will be less than ~0.1% for the lower mantle. (ii) Hydrogen may enhance seismic wave attenuation. Recently, Lawrence and Wysession (2006) reported a broad high attenuation region in the lower mantle beneath Asia. This region shows only modest low velocity anomalies. In this sense, the nature of velocity and attenuation anomalies in this region is similar to those in the deep upper mantle found by Shito and Shibutani (2003) and Shito et al. (2006). This suggests that hydrogen may have an important effect of enhancing seismic wave attenuation in lower mantle minerals, but very

370

Karato

little is known about the possible mechanisms by which hydrogen enhances anelasticity in lower mantle minerals. (iii) The influence of hydrogen on electrical conductivity in lower mantle minerals is difficult to assess. However, a comparison of laboratory electrical conductivity data (e.g., Katsura et al. 1998; Xu et al. 1998a) with geophysical observations suggests that hydrogen may not have a large effect to enhance conductivity in lower mantle minerals. Electrical conductivity of silicate perovskite is high due to a high concentration of point defects (Xu and McCammon 2002) and hydrogen solubility in perovskite is low (Bolfan-Casanova et al. 2000) and consequently, hydrogen may not play an important role in electrical conductivity in the lower mantle.

SUMMARY AND OUTLOOK The remote sensing of hydrogen in Earth’s mantle is an exciting new opportunity. The key concept behind this is a notion that hydrogen has strong effects on various physical properties of minerals some of which can be detected by geophysical methods. In earlier papers (e.g., Karato 1990; Karato 1995), possible relationships between hydrogen content and geophysically measurable properties such as electrical conductivity and seismic wave propagation were proposed based on then available sketchy experimental observations and theoretical models on defect-related properties. Many of these hypotheses have now been transformed to more solid models based on detailed experimental studies. However, inferring the distribution of hydrogen from geophysical observations remains challenging, and a number of issues need to be explored in more detail. Here I will list some of the mineral physics issues that are critical to make further progress in this approach: (i) The quantitative relationships between hydrogen content and seismic wave attenuation must be determined for major mantle minerals. Preliminary data exist (e.g., Jackson et al. 1992), but quantitative studies under well-controlled chemical environment are needed. (ii) The quantitative relationships between electrical conductivity and hydrogen content must be determined for all major constituent minerals in the mantle. These studies are needed particularly for orthopyroxene and garnet for the upper mantle and transition zone, and for silicate perovskite (and post-perovskite phase) and (Mg,Fe)O for the lower mantle. (iii) The role of hydrogen on non-elastic deformation and other deformation-related processes (e.g., diffusion, grain-growth) in lower mantle minerals must be clarified. Not only silicate perovskite but also the role of hydrogen in (Mg,Fe)O needs to be investigated. As I emphasized in several places, further developments in geophysical studies are also needed. Because of the trade-off among competing factors (i.e., non-uniqueness), reliable inference of distribution of hydrogen can only be made when high-resolution data are available for several geophysical parameters from the same place. This requires not only the development of dense stations, but also the development of some theoretical approach such as the simultaneous inversion of attenuation and velocity. A combination of developments in two areas and a close conversation among scientists in these two areas are a key to make further progress in this interdisciplinary area of Earth science.

ACKNOWLEDGMENTS This article is based on an extensive set of studies that I have performed with a number of colleagues during the last ten years or so. Financial support for these studies was obtained from National Science Foundation of USA and Japan Society for Promotion of Sciences.

Remote Sensing of Hydrogen in the Earth’s Mantle

371

REFERENCES Akaogi M, Ito E, Navrotsky A (1989) Olivine-modified spinel-spinel transitions in the system Mg2SiO4Fe2SiO4: calorimetric measurements, thermochemical calculation, and geophysical application. J Geophys Res 94:15671-15685 Bai Q, Kohlstedt DL (1993) Effects of chemical environment on the solubility and incorporation mechanism for hydrogen in olivine. Phys Chem Mineral 19:460-471 Bell DR, Rossman GR (1992) Water in Earth’s mantle: The role of nominally anhydrous minerals. Science 255:1391-1397 Benz H, Vidale JE (1993) Sharpness of upper-mantle discontinuities determined from high-frequency reflections. Nature 365:147-150 Bercovici D, Karato S (2003) Whole mantle convection and transition-zone water filter. Nature 425:39-44 Bhattacharya BB, Masters G, Shearer PM (1996) Global lateral variations of shear wave attenuation in the upper mantle. J Geophys Res 101:22273-22289 Blum J, Shen Y (2004) Thermal, hydrous, and mechanical states of the mantle transition zone beneath southern Africa. Earth Planet Sci Lett 217: 367-378 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett182:209-221 Bolfan-Casanova N, Keppler H, Rubie DC (2003) Water partitioning at 660 km depth and evidence for very low water solubility in magnesium silicate perovskite. Geophys Res Lett 30, doi:10.1029/2003GL017182 Carter NL, Avé Lallemant HG (1970) High temperature deformation of dunite and peridotite. Geol Soc Am Bull 81:2181-2202 Chakraborty S, Knoche R, Schulze H, Rubie DC, Dobson D, Ross NL, Angel RJ (1999) Enhancement of cation diffusion rates across the 410-kilometer discontinuity in Earth’s mantle. Science 283:362-365 Chen J, Inoue T, Yurimoto H, Weidner DJ (2002) Effect of water on olivine-wadsleyite phase boundary in the (Mg,Fe)2SiO4 system. Geophys Res Lett 29, 10.1029/2001GRL014429 Chopra PN, Paterson MS (1984) The role of water in the deformation of dunite. J Geophys Res 89:78617876 Constable SC, Parker RL, Constable CG (1987) Occam’s inversion: A practical algorithm for generating smooth models from electromagnetic sounding data. Geophysics 52:289-300 Cooper RF (2002) Seismic wave attenuation: energy dissipation in viscoelastic crystalline solids. Rev Mineral Geochem 51:253-290 Couvy H, Frost DJ, Heidelbach F, Nyilas K, Ungar T, Mackwell SJ, Cordier P (2004) Shear deformation experiments of forsterite at 11 GPa-1400°C in the multianvil apparatus. Eur J Mineral 16:877-889 de Groot-Hedlin C, Constable SC (2004) Inversion of magnetotelluric data for 2D structure with sharp resistivity contrasts. Geophysics 69:78-86 Evans RL, Hirth G, Baba K, Forsyth DW, Chave A, Makie R (2005) Geophysical evidence from the MELT area for compositional control on oceanic plates. Nature 437:249-252 Farber DL, Williams Q, Ryerson FJ (1994) Diffusion in Mg2SiO4 polymorphs and chemical heterogeneity in the mantle transition zone. Nature 371:693-695 Farver JR, Yund RA (1991) Oxygen diffusion in quartz: dependence on temperature and water fugacity. Chem Geol 90:55-70 Faul UH, Jackson I (2005) The seismological signature of temperature and grain size variations in the upper mantle. Earth Planet Sci Lett234:119-134 Forsyth WD (1975) The early structural evolution and anisotropy of the oceanic upper mantle. Geophys J Royal Astro Soc 43:103-162 Gaherty JB, Jordan TH, Gee LS (1996) Seismic structure of the upper mantle in a central Pacific corridor. J Geophys Res 101:22291-22309 Gribb TT, Cooper RF (1998) Low-frequency shear attenuation in polycrystalline olivine: grain boundary diffusion and the physical significance of the Andrade model for viscoelastic rheology. J Geophys Res 103:27267-27279 Gu YJ, Dziewonski AM, Ekström G (2003) Simultaneous inversion for mantle velocity and topography of transition zone discontinuities. Geophys J Inter 154:559-583 Gueguen Y, Mercier JM (1973) High attenuation and low velocity zone. Phys Earth Planet Interior 7:39-46 Gung Y, Romanowicz B (2004) Q tomography of the upper mantle using three-component long-period waveforms. Geophys J Int 157:813-830 Gutenberg B (1954) Low-velocity layers in the earth’s mantle. Bull Seismo Soc Am 65:337-348 Hier-Majumder S, Anderson IM, Kohlstedt DL (2005) Influence of protons on Fe-Mg interdiffusion in olivine. J Geophys Res 110, doi:10.1029/2004JB003292 Hirth G, Evans RL, Chave AD (2000) Comparison of continental and oceanic mantle electrical conductivity: Is Archean lithosphere dry? Geochem Geophys Geosyst 1, doi:10.1029/2000GC000048

372

Karato

Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle - implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Huang X, Xu Y, Karato S (2005) Water content of the mantle transition zone from the electrical conductivity of wadsleyite and ringwoodite. Nature 434:746-749 Huang X, Xu Y, Karato S (2006) A wet mantle conductor? (Reply). Nature 439:E3-E4 Inoue T, Weidner DJ, Northrup PA, Parise JB (1998) Elastic properties of hydrous ringwoodite (γ-phase) of Mg2SiO4. Earth Planet Sci Lett160:107-113 Jackson I (2000) Laboratory measurements of seismic wave dispersion and attenuation: recent progress. In: Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, p 265-289 Jackson I, Fitz Gerald JD, Faul UH, Tan BH (2002) Grain-size sensitive seismic-wave attenuation in polycrystalline olivine. J Geophys Res 107(B12), doi: 10.1029/2001JB001225 Jackson I, Paterson MS, Fitz Gerald JD (1992) Seismic wave dispersion and attenuation in Åheim dunite. Geophys J Int 108:517-534 Jacobsen SD, Smyth JR, Spetzler HA, Holl CM, Frost DJ (2004) Sound velocities and elastic constants of ironbearing hydrous ringwoodite. Phys Earth Planet Interior 143/144:47-56 Jambon A, Zimmermann JL (1990) Water in oceanic basalts: evidence for dehydration of recycled crust. Earth Planet Sci Lett101:323-331 Jamtveit B, Brooker R, Brooks K, Larsen LM, Pedersen T (2001) The water content of olivines from the North Atlantic volcanic province. Earth Planet Sci Lett186:401-425 Jung H, Karato S (2001a) Effect of water on the size of dynamically recrystallized grains in olivine. J Struct Geol 23:1337-1344 Jung H, Karato S (2001b) Water-induced fabric transitions in olivine. Science 293:1460-1463 Jung H, Katayama I, Jiang Z, Hiraga T, Karato S (2006) Effects of water and stress on the lattice preferred orientation in olivine. Tectonophysics 421:1-22 Karato S (1977) Rheological Properties of Materials Composing the Earth’s Mantle. Ph. D, University of Tokyo Karato S (1984) Grain-size distribution and rheology of the upper mantle. Tectonophysics 104:155-176 Karato S (1986) Does partial melting reduce the creep strength of the upper mantle? Nature 319:309-310 Karato S (1989a) Defects and plastic deformation in olivine. In: Rheology of Solids and of the Earth. Karato S, Toriumi M (eds) Oxford University Press, p 176-208 Karato S (1989b) Plasticity-crystal structure systematics in dense oxides and its implications for creep strength of the Earth’s deep interior: a preliminary result. Phys Earth Planet Interior 55:234-240 Karato S (1990) The role of hydrogen in the electrical conductivity of the upper mantle. Nature 347:272-273 Karato S (1993) Importance of anelasticity in the interpretation of seismic tomography. Geophys Res Lett 20: 1623-1626 Karato S (1995) Effects of water on seismic wave velocities in the upper mantle. Proc Japan Academy 71: 61-66 Karato S (2003) Mapping water content in Earth’s upper mantle. In: Inside the Subduction Factory. Eiler JE (ed) American Geophysical Union, p 135-152 Karato S (2006a) Hydrogen-related defects and their influence on the electrical conductivity and plastic deformation of mantle minerals: A critical review. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (eds), American Geophysical Union, in press Karato S (2006b) Microscopic models for the influence of hydrogen on physical and chemical properties of minerals. In: Superplume: Beyond Plate Tectonics. Yuen DA, Maruyama S, Karato S, Windley BF (ed), Springer, in press Karato S, Jung H (1998) Water, partial melting and the origin of seismic low velocity and high attenuation zone in the upper mantle. Earth Planet Sci Lett157:193-207 Karato S, Jung H (2003) Effects of pressure on high-temperature dislocation creep in olivine polycrystals. Philos Mag A 83:401-414 Karato S, Paterson MS, Fitz Gerald JD (1986) Rheology of synthetic olivine aggregates: influence of grain-size and water. J Geophys Res 91:8151-8176 Karato S, Spetzler HA (1990) Defect microdynamics in minerals and solid state mechanisms of seismic wave attenuation and velocity dispersion in the mantle. Rev Geophys 28:399-421 Katayama I, Jung H, Karato S (2004) New type of olivine fabric at modest water content and low stress. Geology 32:1045-1048 Katayama I, Karato S (2006) Effects of temperature on the B- to C-type fabric transition in olivine. Phys Earth Planetary Interior 157:33-45 Katayama I, Karato S, Brandon M (2005) Evidence of high water content in the deep upper mantle inferred from deformation microstructures. Geology 33:623-616

Remote Sensing of Hydrogen in the Earth’s Mantle

373

Katsura T, Sato K, Ito E (1998) Electrical conductivity of silicate perovskite at lower-mantle conditions. Nature 395:493-495 Kneller EA, van Keken PE, Karato S, Park J (2005) B-type olivine fabric in the mantle wedge: Insights from high-resolution non-Newtonian subduction zone models. Earth Planet Sci Lett 237:781-797 Kohlstedt DL, Evans B, Mackwell SJ (1995) Strength of the lithosphere: constraints imposed by laboratory measurements. J Geophys Res 100:17587-17602 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kurosawa M, Yurimoto H, Sueno S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Minerals 24:385-395 Lawrence JF, Wysession ME (2006) Seismic evidence for subduction-transported water in the lower mantle. In: Earth’s Deep Water Cycle. Jacobsen SD, Lee Svd (ed) American Geophysical Union, in press Lister GS (1979) Fabric transitions in plastically deformed quartzites: competition between basal, prism and rhomb systems. Bull Mineral 102:232-241 Lizarrale D, Chave A, Hirth G, Schultz A (1995) Northeastern Pacific mantle conductivity profile from longperiod magnetotelluric sounding using Hawaii-to-California cable data. J Geophys Res 100:1783717854 Long MD, van der Hilst RD (2005) Upper mantle anisotropy beneath Japan from shear wave splitting. Phys Earth Planet Interior 151:206-222 Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: implications for water in the mantle. J Geophys Res 95:5079-5088 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90:11319-11333 Mainprice D, Tommasi A, Couvy H, Cordier P, Frost DJ (2005) Pressure sensitivity of olivine slip systems and seismic anisotropy of Earth’s upper mantle. Nature 433:731-733 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554-570 Masters G, Laske G, Bolton H, Dziewonski AM (2000) The relative behavior of shear velocity, bulk sound speed, and compressional velocity in the mantle: implications for chemical and thermal structure. In: Earth’s Deep Interior. Karato AMFS, Liebermann RC, Masters G, Stixrude L (eds), American Geophysical Union, p 63-87 Matsukage KN, Nishihara Y, Karato S (2005) Seismological signature of chemical evolution of Earth’s upper mantle. J Geophys Res 110, doi:10.1029/2004JB003504 Mei S, Kohlstedt DL (2000a) Influence of water on plastic deformation of olivine aggregates, 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates, 2. Dislocation creep regime. J Geophys Res 105:21471-21481 Mercier J-CC (1980) Magnitude of the continental lithospheric stresses inferred from rheomorphic petrology. J Geophys Res 85:6293-6303 Michael PJ (1988) The concentration, behavior and storage of H2O in the suboceanic upper mantle: implications for mantle metasomatism. Geochim Cosmochim Acta 52:555-566 Minster JB, Anderson DL (1980) Dislocations and nonelastic processes in the mantle. J Geophys Res 85: 6347-6352 Nakajima J, Hasegawa A (2004) Shear-wave polarization anisotropy and subduction-induced flow in the mantle wedge of northern Japan. Earth Planet Sci Lett 225:365-377 Nicolas A, Christensen NI (1987) Formation of anisotropy in upper mantle peridotite: a review. In: Composition, Structure and Dynamics of the Lithosphere-Asthenosphere System. Fuchs K, Foridevaux C (eds) American Geophysical Union, p 111-123 Nishihara Y, Shinmei T, Karato S (2006) Grain-growth kinetics in wadsleyite: effects of chemical environment. Phys Earth Planet Interior 154:30-43 O’Neill HStC, McCammon CA, Canil D, Rubie DC, Ross CRI, Seifert F (1993) Mössbauer spectroscopy of mantle transition zone phases and determination of minimum Fe3+ content. Am Mineral 78:456-460 Olsen N (1999) Long-period (30 days - 1 year) electromagnetic sounding and the electrical conductivity of the lower mantle beneath Europe. Geophys J Int 138:179-187 Parker RL (1980) The inverse problem of electromagnetic induction: Existence and construction of solutions based upon incomplete data. J Geophys Res 85:4421-4428 Plank T, Langmuir AH (1992) Effects of melting regime on the composition of the oceanic crust. J Geophys Res 97:19749-19770 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536

374

Karato

Richard G, Monnereau M, Ingrin J (2002) Is the transition zone an empty water reservoir? Inference from numerical model of mantle dynamics. Earth Planet Sci Lett 205:37-51 Romanowicz B, Durek JJ (2000) Seismological constraints on attenuation in the Earth: a review. In: Earth’s Deep Interior. Karato AMFS, Liebermann RC, Masters G, Stixrude L (eds), American Geophysical Union, p 161-179 Rüpke LH, Phipps Morgan J, Hort M, Connolly JAD (2004) Serpentine and the subduction zone water cycle. Earth Planet Sci Lett223:17-34 Shankland TJ, O’Connell RJ, Waff HS (1981) Geophysical constraints on partial melt in the upper mantle. Rev Geophys Space Phys 19:394-406 Shearer PM (2000) Upper mantle discontinuities. In: Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, p 115-131 Shito A, Karato S, Matsukage KN, Nishihara Y (2006) Toward mapping water content, temperature and major element chemistry in Earth’s upper mantle from seismic tomography. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (eds), American Geophysical Union, in press Shito A, Karato S, Park J (2004) Frequency dependence of Q in Earth’s upper mantle inferred from continuous spectra of body wave. Geophys Res Lett 31, doi:10.1029/2004GL019582 Shito A, Shibutani T (2003a) Anelastic structure of the upper mantle beneath the northern Philippine Sea. Phys Earth Planet Inter 140:319-329 Shito A, Shibutani T (2003b) Nature of heterogeneity of the upper mantle beneath the northern Philippine Sea as inferred from attenuation and velocity tomography. Phys Earth Planet Interior 140:331-341 Simpson F (2001) Resistance to mantle flow inferred from the electromagnetic strike of the Australian upper mantle. Nature 412:632-635 Simpson F (2002) Intensity and direction of lattice-preferred orientation of olivine: are electrical and seismic anisotropies of the Australian mantle reconcilable? Earth Planet Sci Lett 203:535-547 Simpson F, Tommasi A (2005) Hydrogen diffusivity and electrical anisotropy of a peridotite mantle. Geophys J Int 160:1092-1102 Smith GP, Wiens DA, Fischer KM, Dorman LM, Hildebrand JA (2001) A complex pattern of mantle flow in the Lau back-arc. Science 292:713-716 Smyth JR (1994) A crystallographic model for hydrous wadsleyite (β-Mg2SiO4): An ocean in the Earth’s interior. Am Mineral 79:1021-1024 Smyth JR (1987) β−Mg2SiO4: a potential host for water in the mantle? Am Mineral 75: 1051-1055 Smyth JR, Frost DJ (2002) The effect of water on the 410-km discontinuity: An experimental study. Geophys Res Lett 29: 10.129/2001GL014418 Spiegelman M, Elliott T (1993) Consequences of melt transport for uranium series disequilibrium in young lavas. Earth Planet Sci Lett 118:1-20 Spiegelman M, Kenyon PM (1992) The requirement of chemical disequilibrium during magma migration. Earth Planet Sci Lett 109:611-620 Stixrude L (1997) Structure and sharpness of phase transitions and mantle discontinuities. J Geophys Res 102: 14835-14852 Stixrude L, Lithgow-Bertelloni C (2005) Mineralogy and elasticity of the oceanic upper mantle: Origin of the low-velocity zone. J Geophys Res 110, doi:10.1029/2004JB002965 Stolper EM, Newman S (1994) The role of water in the petrogenesis of Mariana trough magmas. Earth Planet Sci Lett121:293-325 Suetsugu D, Inoue T, Yamada A, Zhao D, Obayashi M (2006) Towards mapping three-dimensional distribution of water in the transition zone from P-wave velocity tomography and 660-km discontinuity depths. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (ed) American Geophysical Union, in press. Tan B, Jackson I, Fitz Gerald JD (1997) Shear wave dispersion and attenuation in fine-grained synthetic olivine aggregates: preliminary results. Geophys Res Lett 24:1055-1058 Tan B, Jackson I, Fitz Gerald JD (2001) High-temperature viscoelasticity of fine-grained polycrystalline olivine. Phys Chem Minerals 28:641-664 Tarits P, Hautot S, Perrier F (2004) Water in the mantle: Results from electrical conductivity beneath the French Alps. Geophys Res Lett 31, doi:10.1029/2003GL019277 ten Grotenhuis SM, Drury MR, Peach CJ, Spiers CJ (2004) Electrical properties of fine-grained olivine: evidence for grain-boundary transport. J Geophys Res 109, doi:10.1029/2003JB002799 Thompson AB (1992) Water in the Earth’s upper mantle. Nature 358:295-302 Utada H, Goto T, Koyama T, Shimizu H, Obayashi M, Fukao Y (2005) Electrical conductivity in the transition zone beneath the North Pacific and its implications for the presence of water. EOS, Trans Am Geophys Union 86:DI41A-1260 Utada H, Koyama T, Shimizu H, Chave AD (2003) A semi-global reference model for electrical conductivity in the mid-mantle beneath the north Pacific region. Geophys Res Lett 30, doi:10.1029/2002GL016092

Remote Sensing of Hydrogen in the Earth’s Mantle

375

van der Meijde M, Marone F, Giardini D, van der Lee S (2003) Seismic evidence for water deep in Earth’s upper mantle. Science 300:1556-1558 Wallace P (1998) Water and partial melting in mantle plumes: inferences from the dissolved H2O concentrations of Hawaii basaltic magmas. Geophys Res Lett 25:3639-3642 Wang D, Mookherjee M, Xu Y, Karato S (2006) The effect of hydrogen on the electrical conductivity in olivine. Nature, submitted. Wolfe CJ, Solomon SC (1998) Shear-wave splitting and implications for mantle flow beneath the MELT region of the East Pacific. Science 280:1230-1232 Wood BJ (1995) The effect of H2O on the 410-kilometer seismic discontinuity. Science 268:74-76 Xu Y, McCammon C (2002) Evidence for ionic conductivity in lower mantle (Mg,Fe)(Si,Al)O3 perovskite. J Geophys Res 107, doi:10.1029/2001JB000677 Xu Y, McCammon C, Poe BT (1998a) Effect of alumina on the electrical conductivity of silicate perovskite. Science 282:922-924 Xu Y, Poe BT, Shankland TJ, Rubie DC (1998b) Electrical conductivity of olivine, wadsleyite, and ringwoodite under upper-mantle conditions. Science 280:1415-1418 Yoshii T (1973) Upper mantle structure beneath the north Pacific and marginal seas. J Phys Earth 21:313-328 Young TE, Green HW II, Hofmeister AM, Walker D (1993) Infrared spectroscopic investigation of hydroxyl in β-(Mg,Fe)2SiO4 and coexisting olivine: implications for mantle evolution and dynamics. Phys Chem Mineral 19:409-422 Zimmerman MR, Zhang S, Kohlstedt DL, Karato S (1999) Melt distribution in mantle rocks deformed in shear. Geophys Res Lett 26:1505-1508

16

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 377-396, 2006 Copyright © Mineralogical Society of America

The Role of Water in High-Temperature Rock Deformation David L. Kohlstedt Department of Geology and Geophysics University of Minnesota Minneapolis, Minnesota, 55455, U.S.A. e-mail: [email protected]

INTRODUCTION The first report of water/hydrolytic weakening of silicate minerals was based on the observation that the strength of quartz decreases significantly in the presence of the water released from talc, the confining medium used in some of high-pressure experiments (Blacic and Griggs 1965). Subsequently, a number of researchers have studied this important phenomenon in several different minerals. Publications examining the influence of water or protons on the creep behavior of nominally anhydrous silicate minerals are listed in Table 1. Initial studies treated the water-weakening phenomenon as an on-off process; that is, minerals and rocks are weak under hydrous conditions but strong under anhydrous conditions. Further investigations, however, demonstrated that the strengths of nominally anhydrous minerals (NAMs) and rocks decrease systematically with increasing hydrogen concentration (Kronenberg and Tullis 1984; Kohlstedt et al. 1995; Post et al. 1996; Mei and Kohlstedt 2000a,b; Karato and Jung 2003). Two quite different approaches have been used in analyzing the effect of water or protons on the strength of nominally anhydrous silicate minerals. In the first model, a mechanism is envisioned in which water hydrolyzes strong Si-O bonds via the reaction Si-O-Si + H2O → Si-OH·OH-Si (Griggs 1967), thus the term water or hydrolytic weakening. As a result, the glide of dislocations becomes easier in wet quartz than in dry quartz since Si-O bonds do not need to be broken if water is present. In this case of water/hydrolytic weakening, the rate limiting step is the propagation of kinks along dislocations, which is facilitated by diffusion of HOH along dislocation cores. The resulting dislocation velocity is assumed to be proportional to the HOH concentration (Griggs 1974). In effect, this analysis implies that the Peierls stress/barrier becomes small as Si-O-Si bridges become hydrolyzed. In the second model, the role of point defects, diffusion, and dislocation climb are emphasized (Hobbs 1981, 1983, 1984; Poumellec and Jaoul 1984; Mackwell et al. 1985). This approach builds in part on the observation that protons diffuse quickly in nominally anhydrous minerals such as olivine (Mackwell and Kohlstedt 1990; Kohlstedt and Mackwell 1998, 1999) so that it is not necessary to move hydroxyl (OH) ions or water (H2O) molecules in order to induce water/hydrolytic weakening, now more appropriated referred to as protonic weakening. As discussed in the following sections, the introduction of a charged species, namely protons, into a nominally anhydrous mineral must be accompanied by an increase in the concentration of negatively charged point defects, for example, cation vacancies. Hence, enhanced rates of ionic diffusion and thus of dislocation climb will necessarily occur. While hydrolysis of Si-O bonds and the associated reduction of the barrier to kink migration and thus to dislocation glide may be important at low temperatures, the increase in ionic diffusivities due to the addition of positively charged protons and the associated increase in jog mobility and thus dislocation climb rate will undoubtedly prevail at high temperatures. 1529-6466/06/0062-0016$05.00

DOI: 10.2138/rmg.2006.62.16

378

Kohlstedt Table 1. Investigations of the effect of water/protons on the strength of nominally anhydrous silicate minerals. Mineral/Rock

Reference

quartz

Griggs and Blacic (1965) Hobbs et al. (1972) Blacic (1975) Kekulawala et al. (1978) Paterson and Kekulawala (1979) Kekulawala et al. (1981) Blacic and Christie (1984) Jaoul et al. (1984) Mainprice and Paterson (1984) Doukhan and, Trépied (1985) Mackwell and Paterson (1985) Ord and Hobbs (1986) Koch et al. (1989) Paterson (1989) Tullis and Yund (1989) Luan and Paterson (1992) Kronenberg (1994) Gleason and Tullis (1995) Post et al. (1996) Post and Tullis (1998) Avé Lallemant and Carter (1970) Blacic (1972) Poumellec and Jaoul (1984) Mackwell et al. (1985) Karato et al. (1986) Borch and Green (1989) Karato (1989) Hirth and Kohlstedt (1996) Chen et al. (1998) Mei and Kohlstedt (2000a) Mei and Kohlstedt (2000b) Hirth and Kohlstedt (2003) Karato and Jung (2003) Avé Lallemant (1978) Boland and Tullis (1986) Hier-Majumder et al. (2005a) Chen et al. (2006) Tullis et al. (1979) Tullis and Yund (1980) Tullis and Yund (1991) Dimanov et al. (1999) Rybacki and Dresen (2000) Stünitz et al. (2003)

olivine

pyroxene

feldspar

Water in High-T Rock Deformation

379

BACKGROUND This chapter specifically examines the role of water—that is, protons—on the hightemperature (T > 2/3 Tm) creep of olivine, a nominally anhydrous mineral, from the point of view of dislocation climb as the rate-limiting step in the deformation process. The usual justification for this approach is the similarity of the value of the activation energy for creep and that for self-diffusion of slowest ionic species (e.g., Weertman 1999), since climb of edge dislocations requires a diffusive flux of ions (e.g., Hirth and Lothe 1968, pp. 506-519; Poirier 1985, pp. 5862). Indeed, a marked one-to-one correlation exists between the activation energies for creep and self-diffusion for a large number of metallic and ceramic materials (e.g., Dorn 1956; Sherby and Burke 1967; Mukerjee et al. 1969; Takeuchi and Argon 1976; Evans and Knowles 1978). Until recently, it appeared that, under dry conditions, the activation energy for creep of olivine-rich rocks, Qcdry ≈ 530 kJ/mol (e.g., Hirth and Kohlstedt 2003), was significantly larger than the activation energy for diffusion of silicon, the slowest ionic species in olivine, QSidry ≈ 290 kJ/mol (Houlier et al. 1990). Consequently, models for climb-controlled creep in olivine attributed the difference in activation energy either to the energy required to form jogs along dislocations (e.g., Gueguen 1979; Karato and Ogawa 1982; Hobbs 1981, 1983, 1984; Ricoult and Kohlstedt 1985; Karato 1989; Mei and Kohlstedt 2000a; Karato and Jung 2003) or to the energy resulting from coupled diffusion of the slowest species and the fastest species in olivine, the tetrahedral and octahedral cations, respectively (Jaoul 1990). Alternatively, this discrepancy between the two activation energies might be considered as evidence against climb-controlled creep of olivine. Newer results, however, yield a larger value of QSidry = 530 ± 40 kJ/mol (Dohmen et al. 2002) for silicon self-diffusion under anhydrous (dry) conditions. In their study, Dohmen et al. (2002) attributed the difference between their value for QSidry and that of Houlier et al. (1990) to artifacts related to very short diffusion distances combined with limits on the spatial resolution of the analytical instruments used to measure diffusion profiles. Experiments designed to study self-diffusion of Si under hydrous conditions yield QSiwet = 450 ± 50 kJ/mol (Chakraborty and Costa 2004). This value for the activation energy for Si selfdiffusion agrees within experimental error with that obtained in high-temperature creep experiments carried out under hydrous conditions, Qcwet = 520 ± 40 kJ/mol (e.g., Hirth and Kohlstedt 2003). This good agreement between the activation energies for Si diffusion and dislocation creep under both anhydrous and hydrous conditions lends support to a deformation mechanism for olivine in which creep rate is limited by dislocation climb. In the following pages, some of the oft-cited models for climb-controlled creep are first introduced, and the case for and against climb-limited creep in olivine is explored. Next, the relationship among climb velocity, jog velocity, and jog concentration is briefly reviewed. Then, the ionic flux induced by a gradient in stress is examined to isolate the appropriate diffusion coefficient governing the climb rate. Subsequently, this diffusivity is written in terms of the concentration and diffusivity of the appropriate point defects (e.g., vacancies) to explicitly introduce the dependence of creep rate on water fugacity and, in turn, on water concentration. Finally, the experimentally determined flow laws for olivine deformed under anhydrous and hydrous conditions are compared to the models for high-temperature creep to further test the applicability of a climb-controlled deformation mechanism to flow of this important mantle mineral.

MODELS OF CLIMB-CONTROLLED CREEP The idea that climb of dislocations is the rate-controlling step during high-temperature, steady-state creep of crystalline materials was suggested over fifty years ago (e.g., Mott 1951, 1956; Weertman 1955, 1957a). Three models for creep in which the creep rate is controlled

380

Kohlstedt

by dislocation climb are briefly discussed below in order to introduce the general form of the flow law relating strain rate, ε, differential stress, σ, and self-diffusivity, D. For more extended discussions of these and related creep equations, the reader is referred to Weertman (1978), Poirier (1985, pp. 94-144), Cannon and Langdon (1988), Evans and Kohlstedt (1995), and Weertman (1999). Based on an analysis of steady-state creep by Weertman (1955, 1957a), strain is produced primarily by dislocation glide while strain rate is controlled by the rate of dislocation climb. The basic premise of this model is that dislocations glide relatively rapidly until they encounter obstacles at which point they slowly climb past the barriers to proceed again by gliding. The average dislocation velocity, ν, is then ν≈

g

νc

(1)

c

where g is the glide distance, c is the climb distance, and νc is the climb velocity (Poirier 1985, p. 110; Weertman 1999). For dislocations with Burgers vector b, the strain rate, ε, is then calculated using the Orowan equation (e.g., Poirier 1985, pp. 62-63) ε = ρbν

(2)

with the dislocation density, ρ, given in terms of the ratio of differential stress, σ, to the shear modulus, G (Poirier 1985, pp. 190-192; Weertman 1999) ⎛ σ ⎞ ρ≈⎜ ⎟ ⎝ Gb ⎠

2

(3)

and the climb velocity as (Hirth and Lothe 1968, pp. 506-519) ν c = 2π

σVm D 1 RT b ln( Ro / rc )

( 4)

In Equation (4), Vm is the molar volume, D is the self-diffusivity, R is the gas constant, and T is the absolute temperature. The average spacing between dislocations, Ro, can be written in terms of the dislocation density as Ro ≈ 1/(ρ)½, while the inner cutoff radius, rc, is generally taken to be as rc ≈ b. Equations (1) to (4) can be combined to yield (Weertman 1999) 3

ε = 2π

GVm ⎛ σ ⎞ D 1 RT ⎜⎝ G ⎟⎠ b 2 ln(G / σ)

g

(5)

c

In the model for dislocation creep developed by Nabarro (1967), strain is produced entirely by climb of dislocations. Dislocations produced by Bardeen-Herring sources (1952) form a network and continuously climb. Operation of the dislocation sources increases their density, while climb of dislocations of opposite sign toward one another results in annihilation of dislocations thus decreasing their density. In balance, a steady state is reached yielding the flow law 3

ε=2

GVm ⎛ σ ⎞ D 1 RT ⎜⎝ G ⎟⎠ b 2 ln(4G / πσ)

(6)

Note that this expression differs from that originally presented by Nabarro (1967) by a factor 2π as corrected by Nix et al. (1971). In an analysis of high-temperature creep by Evans and Knowles (1977), a threedimensional network of dislocations is envisioned with strain produced solely by glide of edge dislocations with strain rate limited by climb. Dislocations, released from pinning points by

Water in High-T Rock Deformation

381

rupture of network nodes or activation of Frank-Read sources, glide from one pinning point to another in the network. The slip distance is taken to be the network spacing, and the rate of release is governed by climb. If diffusion occurs through the interiors of grains (rather than, for example, along dislocation cores), then the strain rate is given by 3

ε=

4.2 3π GVm ⎛ σ ⎞ D 1 ⎜ ⎟ 2 2 RT ⎝ G ⎠ b ln(αG / 2σ) α

( 7)

where α = 1.6 (Evans and Knowles 1978). By introducing the glide step, the strain rate predicted by Equation (7) is a factor of 4-5 times greater than that obtained from the climbonly model in Equation (6). This flow law fits the experimentally determined creep data for UO2, Al2O3, and MgO within a factor of ~2 (Evans and Knowles 1978).

THE CASE FOR OLIVINE As discussed below, the flow laws given by Equations (5), (6), and (7) can be applied directly to olivine if the self-diffusivity, D, is replaced by the self-diffusivity of the slowest ionic species. For olivine, under both anhydrous and hydrous conditions, DMe >> DO >> DSi. Hence, in the above equations, D = DSi. For anhydrous conditions, the diffusion results for (i) Me are from Hermeling and Schmalzried (1984) and Chakraborty (1997), (ii) O are from Ryerson et al. (1989), Gerard and Jaoul (1989), and Dohmen et al. (2002), and (iii) Si are from Houlier et al. (1990) and Dohmen et al. (2002). For hydrous conditions, the diffusion results for (i) Me are from Wang et al. (2004) and Hier-Majumder et al. (2005b), (ii) O are from Charkraborty and Costa (2004), and (iii) Si are from Chakraborty and Costa (2004). Possibly the most compelling argument in favor of climb-controlled deformation of olivine under both anhydrous and hydrous conditions is the near equality of the activation energies for creep, Qcdry = 530 ± 40 kJ/mol and Qcwet= 520 ± 40 kJ/mol (e.g., Hirth and Kohlstedt 2003), and those for self-diffusion of Si, QSidry = 530 ± 40 kJ/mol (Dohmen et al. 2002) and QSiwet = 450 ± 50 kJ/mol (Chakraborty and Costa 2004). Other observations that favor climb-controlled creep over other possible rate-limiting mechanisms involve the temperature and stress dependencies of creep rate. First, models for cross slip-controlled (Poirier 1979; Poirier 1985, pp. 97-101), glide-controlled (Weertman 1957b; Poirier 1985, pp. 101-103), and barrier- or lattice resistance-controlled creep (Kocks et al. 1975, pp. 171-229, 237-255; Frost and Ashby 1982, pp. 8-9) generally include a marked dependence of the thermally activated term on differential stress rather than the simple dependence of strain rate on temperature predicted by Equations (5) to (7) (Evans and Kohlstedt 1995). This more complex behavior is not observed for high-temperature creep of either polycrystalline samples (e.g., Hirth and Kohlstedt 2003) or single crystal samples of olivine (Bai et al. 1991). Second, of the various models for creep, only those involving dislocation climb yield a simple power-law dependence of strain rate on differential stress (Poirier 1985, pp. 94-117; Evans and Kohlstedt 1995). For olivine, creep behavior exhibits a simple power-law relationship between strain rate and differential stress with a stress exponent of n = 3.5 ± 0.3 over a very wide range of thermomechanical conditions (for the experimentalist’s point of view on this topic, see Hirth and Kohlstedt 2003). These observations on the temperature and stress dependencies of strain rate support basic climb-controlled models for creep of olivine. However, the reader is urged to review the debate on the general topic of “is power law creep diffusion controlled?” in the papers of Poirier (1978, 1979) and Sherby and Weertman (1979). Probably the most demanding test of climb-controlled creep models or, for that matter, any creep model, comes from a comparison of the experimentally determined creep rate for a series of samples with the rate of creep predicted by each of the models, for example, by

382

Kohlstedt

Equations (5), (6), and (7). Such a comparison is presented in Figure 1 for olivine-rich rocks deformed under anhydrous conditions. The models for climb-limited creep described by Equation (5) (Weertman 1999), Equation (6) (Nabarro 1967), and Equation (7) (Evans and Knowles 1977) are compared with the experimentally determined flow law of Chopra and Paterson (1984) as modified by Hirth and Kohlstedt (1996). The model of Nabarro (1967) in which all of the strain is accomplished by climb yields a strain rate at a given stress that is a factor of ~1000 smaller than the measured strain rate. Likewise, the model of Evans and Knowles (1977) in which all of the strain is generated by glide but the creep rate is governed by climb yields a strain rate that is a bit less than a factor of 100 smaller than the experimentally determine strain rate. Finally, the model of Weertman (1999) in which most of the strain is produced by glide with a small fraction produced by climb while the strain rate is controlled by climb can be brought into good agreement with the laboratory-derived flow law if the ratio of the glide distance to the climb distance is set to g/ c ≈ 200. A similar comparison is made for dunite deformed under hydrous conditions in Figure 2. In this case, the flow law from the review paper of Hirth and Kohlstedt (2003) is used. This flow law was obtained from a detailed reanalysis of the data published in the studies of Chopra and Paterson (1984), Mei and Kohlstedt (2000a), and Karato and Jung (2003). The self-diffusivity for Si in olivine was taken from Chakraborty and Costa (2004). As in the dry case, the strain rates predicted by the Nabarro (1967) and Evans and Knowles (1977) models are about 3 and 2 orders of magnitude smaller, respectively, than that determined experimentally. A ratio of glide distance to climb distance of g/ c ≈ 50 is required to bring the Weertman (1999) model into agreement with the experimentally derived flow law. This ratio of glide-to-climb distance is similar to that reported in Jaoul (1990) based on his analysis of high-temperature creep of forsterite single crystals. In addition, the pronounced crystallographic fabrics (lattice preferred

Figure 1. Log-log plot of strain rate versus differential stress comparing the experimentally determined flow law for dunite deformed under anhydrous (dry) conditions by Chopra and Paterson (1984) as modified by Hirth and Kohlstedt (1996) with the models for climb-limited creep from Nabarro (1967), Evans and Knowles (1977), and Weertman (1999). The comparison is made at T = 1400 °C. The flow law is given by the power law relation ε = Aσ n exp(−Q / RT ) with A = 4.85×104 s−1 MPa−n, n = 3.5, Q = 535 kJ mol−1. In the three models, the diffusion coefficient for Si from Dohmen et al. (2002) was used with values of b = 0.485 nm, G = 52 GPa, and Vm = 43.8×10−6 m3 with a Poisson’s ratio of 0.245.

Water in High-T Rock Deformation

383

Figure 2. Log-log plot of strain rate versus differential stress comparing the experimentally determined flow law for dunite deformed under hydrous (wet) conditions modified from the review of Hirth and Kohlstedt (2003) with the models for climb-limited creep from Nabarro (1967), Evans and Knowles (1977), and Weertman (1999). The comparison is made at T = 1200 °C and P = 2 GPa with a corresponding fH2O = 11.8 GPa. The flow law is given by the power law relation ε = A fH1.20O σn exp[ −( E + PV ) / RT ] with A = 9.84×103 s−1 MPa−(n+1.0), n = 3.5, E = 520 kJ mol−1, V = 22×10−6 m3 mol−1. In this equation, E and V are activation energy and activation volume for dislocation creep, while P is pressure. Note that Hirth and Kohlstedt (2003) used a water fugacity exponent of 1.2 rather than 1.0. In the three models, the diffusion coefficient for Si obtained at P = 2 GPa by Chakraborty and Costa (2004) was used with values of b = 0.485 nm, G = 52 GPa, and Vm = 43.8×10−6 m3 with a Poisson’s ratio of 0.245.

orientations) observed in sheared olivine aggregates indicate a substantial contribution of glide to the high-temperature creep of olivine (Zhang and Karato 1995; Bystricky et al. 2000; Jung and Karato 2001). Thus, not only do the values of the activation energy and stress exponent support climb-limited creep for olivine, but also the absolute value of the deformation rate is consistent with diffusion-limited dislocation creep in which a major portion of the strain is accomplished by glide.

DISLOCATION CLIMB The dislocation climb process is well described in the books by Hirth and Lothe (1968, pp. 506-529) and Poirier (1985, pp. 58-62). The climb velocity is written in terms of the jog concentration, cj, and the jog migration velocity, vj, as ν c = c jν j

(8)

The usual assumption is that dislocations are fully saturated with jogs, that is, cj = 1 such that the concentration of vacancies is maintained at its local equilibrium value along each dislocation. This approximation is expected to apply unless the stacking fault energy is low (i.e., unless unit dislocations dissociate into widely spaced partial dislocations). The point to be emphasized here for olivine is that it has often been assumed that cj << 1 (e.g., Gueguen 1979; Karato and Ogawa 1982; Hobbs 1983, 1984; Ricoult and Kohlstedt 1985; Karato 1989; Karato and Jung 2003). As a result, the activation energy for climb and consequently for climb-limited creep is composed of two terms, the activation energy for diffusion plus the activation energy for jog formation.

384

Kohlstedt

This approach has been used to justify the apparent difference between the activation energy for high-temperature creep of olivine (~530 kJ/mol, Hirth and Kohlstedt 2003) and early values for the activation energy for Si self-diffusion (~290 kJ/mol, Houlier et al. 1990). Recent selfdiffusion results for silicon in olivine suggest that this discrepancy between the two activation energies does not exist (Dohmen et al. 2002). Consequently, cj ≈ 1, consistent with the high stacking fault energy inferred from transmission electron microscopy observations of partial dislocations in olivine (Vander Sande and Kohlstedt 1976; Fujino et al. 1993a).

DIFFUSION The diffusion coefficient in the above creep equations enters through the climb velocity, as expressed in Equation 4. The climb velocity is calculated by considering the flux of atoms (ions) to or from a dislocation. In a simple metal, the proper diffusivity to use in this equation is clear. In a compound, however, the climb velocity and thus the creep rate are determined by the slowest diffusing constituent since complete lattice molecules must be transported to or from dislocations. Consequently, the fluxes of all of the constituent ions must be coupled (Ruoff 1965; Readey 1966; Gordon 1973; Dimos et al. 1988; Jaoul 1990; Schmalzried 1995, pp. 345-346). The rate of climb-controlled creep of a ternary system such as olivine, Me2SiO4, is expected to be governed by the rate of diffusion of the slowest of the three ionic species. (In the case of ferromagnesian olivine, the Me site can be occupied by either Mg or Fe; therefore, Fe-Mg olivine is treated as a quasi-ternary system.) To examine explicitly the role of diffusion in hightemperature creep of olivine, we start by considering the fluxes of the three ionic components— Me2+, Si4+, and O2−—and the condition that couples these fluxes together, noting that, in his analysis, Jaoul (1990) concluded that creep rate of olivine is controlled by the diffusivity of the slowest species (i.e., Si) times the concentration of the fastest diffusing species (i.e., Me). If cross terms are neglected, the flux equations for the ionic species in Me2SiO4 can be expressed as (e.g., Schmalzried 1981, p.63; Schmalzried 1995, pp. 78-82) j Me2+ = −

(

DMeC Me D C ∇η Me2+ = − Me Me ∇µ Me2+ + 2 F∇Φ RT RT

)

(9a )

(

)

(9 b )

(

)

( 9c )

jSi 4+ = −

DSiCSi D C ∇ηSi 4+ = − Si Si ∇µ Si 4+ + 4 F∇Φ RT RT

jO 2 − = −

DOCO D C ∇η O 2 − = − O O ∇µ O 2 − − 2 F ∇ Φ RT RT

and

where the bold indicates a vector quantity; ji is the flux, Di is the self-diffusivity, and ηi the electrochemical potential of the ith ionic species. In Equation (9), the electrochemical potential is expressed in terms of the chemical potential, µi, and the electrical potential, Φ, in the following manner (e.g., Schmalzried 1981, p. 63): η i = µi + z i FΦ

(10)

where zi is the charge on the ith species and F is the Faraday constant. To maintain stoichiometry during diffusion-controlled creep, if ions on different sublattices travel by more than one path (e.g., through the lattice and along dislocations and/or grain boundaries), the ionic fluxes are coupled by the relationship ∇i j Me2+ C Me

=

∇i jSi 4+ CSi

=

∇ i jO 2 − CO

(11a )

Water in High-T Rock Deformation

385

in order to meet the requirement that entire lattice molecules be transferred between sites of repeatable growth such as dislocations and grain boundaries. In the one-dimensional limit in which the various types of ions diffuse along a single path or parallel paths, the flux coupling conditions reduces to the usual form j Me2+ C Me

=

jSi 4+ CSi

=

jO 2 −

(11b)

CO

To obtain an expression for the rate-limiting flux in terms of the gradient in chemical potential and thus in terms of the gradient in stress, the gradient in electrical potential must be evaluated. In a ternary or quasi-ternary silicate, an expression can be obtained for ∇Φ (i) if stress-directed diffusion is rate limited by one of the ionic species or (ii) if Me2SiO4 is a semiconductor as is the case for (Mg,Fe)2SiO4. We consider the first case here and the second case in the Appendix. From Equation (11b), the flux coupling condition for Me2SiO4 is 2jMe2+ = 4jSi4+ = jO2−. Then, noting that DMe >> DO >> DSi, the relation ∇η Me2+ = ∇µ Me2+ + 2 F∇Φ ≈ 0

(12)

must hold in order to ensure that the fluxes of the ions on all three sublattices are of similar finite magnitude. By exploiting the equilibrium condition for Me2+ + O2− = MeO and Si4+ + 2O2− = SiO2, that is, µ Me2+ + µ O2− = µ MeO

(13a )

µ Si 4+ + 2µ O2− = µ SiO2

(13b)

and

expressions for the chemical potentials of the component oxides can then be calculated from the flux coupling conditions. The result is DSi∇µ SiO2 = ( DO + 2 DSi )∇µ MeO

(14)

The flux of the slowest species, Si4+, can now be rewritten in terms of the gradient in the chemical potential of the component oxides as jSi 4+ = −

(

CSi DSi ∇µ SiO2 − 2∇µ MeO RT

)

(15)

and, hence, the chemical potential gradient of olivine, µ Me2 SiO 4 = 2µ MeO + µ SiO2

(16)

as jSi 4+ = −

CSi DSi DO ∇µ Me2 SiO 4 RT DO + 4 DSi

(17)

The chemical potential of olivine, as well as of each oxide component, can now be expressed as the sum of an activity, a, and a stress, σ, contribution in the form ∇µ Me2 SiO 4 = RT ∇ ln a Me2 SiO 4 − VMe2 SiO 4 ∇σ

(18a )

where VMe2SiO4 is the molar volume of olivine (e.g., Ready 1966; Gordon 1973; Dimos et al. 1988; Schmalzried 1995, p. 334). Jaoul (1990) arrived at a similar expression for the gradient in chemical potential for olivine but omitted the second term. Consequently, his expression for the gradient in chemical potential of olivine in terms of the gradient in the concentration // ]. Here, of vacancies on the Me sublattice became ∇µMe2SiO4 = RT∇lnaMe2SiO4 ≈ −RT ∇[ VMe

386

Kohlstedt

the Kröger-Vink (1956) notation is used to indicate the species, site occupancy, and charge for point defects. In this specific case, a vacancy (V) on the metal (Me) sublattice is doubly negatively charged (//) relative to a defect-free sublattice site; the square brackets [ ] denote concentration. The gradient in the concentration in metal vacancies arises due to the gradient in normal stress between sources for vacancies (regions of minimum compressive stress) and sinks for vacancies (regions of maximum compressive stress). Jaoul’s expression for the // )°]VMe2SiO4∇σ, chemical potential gradient of olivine, therefore, becomes ∇µMe2SiO4 ≈ [(VMe // where [(VMe )°] is the concentration of Me vacancies under hydrostatic stress conditions. // As [(VMe )°] is typically <10−3, the second term in Equation (18a) dominates such that ∇µ Me2 SiO 4 ≈ −VMe2 SiO 4 ∇σ

(18b)

CSi DSi VMe2 SiO 4 ∇σ RT

(19)

Therefore, for DO >> DSi, jSi 4+ =

and the climb velocity is of the form given by Equation (4) with D = DSi but without the // [( VMe )°] term that appears in Jaoul’s (1990) analysis.

DEPENDENCE OF CREEP RATE ON WATER FUGACITY If the rate of creep is controlled by dislocation climb and hence by diffusion, then the role of water on creep rate can be examined within the framework of point defect thermodynamics. From this perspective, water influences kinetic properties of silicate minerals because it supplies hydrogen, which enters the crystal structure as charged point defects, namely, as protons. The protons, in turn, affect the concentrations of vacancies and interstitials of the constituent ions and, therefore, the diffusivities of these ions. For diffusion of an ion by a vacancy mechanism, the relation between the self-diffusivity of the ion, Dion, and that of a vacancy on the sublattice of that ion, DV, is given by the relationship Dion = XV DV

(20)

where XV is the mole fraction of vacancies on the corresponding sublattice. To explore the dependence of the concentrations of point defects on the Me2+, Si4+, and O sublattices on water fugacity, first consider the charge neutrality conditions appropriate under anhydrous and hydrous conditions. Based on a series of thermogravimetry, diffusion, and electrical conductivity measurements, the charge neutrality condition for olivine under anhydrous conditions is generally taken to be (e.g., Kohlstedt and Mackwell 1998) 2−

i // [ MeMe ] ≡ [ h i ] = 2[ VMe ]

(21a )

Again, the Kröger-Vink (1956) notation is used with the superscripted dot (•) indicating that the Me cation is singly positively charged relative to the normal charge of the site; that is, a Me3+ occupies a site normally occupied by a Me2+. This defect is often referred to as an electron hole, h•, or a small polaron. With the addition of protons, p•, new point defects enter the charge neutrality condition. One might anticipate that with increasing water fugacity (i) protons will replace electron holes as the dominant positively charged point defect and (ii) defect complexes involving protons and Me vacancies will replace Me vacancies as the dominant negatively charged point defect. Unfortunately, direct experimental measurements confirming this suggestion are not yet available; however, indirect evidence is compelling (Kohlstedt et al. 1996; Kohlstedt and Mackwell 1998, 1999; Zhao et al. 2004; Hier-Majumder et al. 2005b). A general form of the

Water in High-T Rock Deformation

387

charge neutrality condition incorporating the majority point defects for anhydrous conditions with those for hydrous conditions can then be written as // / [ h i ] + [ p i ] = 2[ VMe ] + [H Me ]

{(OH)iO

// / − VMe }

i

(21b)

where is shorthand notation for a defect associate formed ≡ {p between a hydroxyl ion or proton and a metal cation vacancy. With increasing water fugacity, // ] to [p•] = the charge neutrality condition might then be expected to change from [h•] = 2[ VMe // 2[ VMe ] to [p•] = 2[ H /Me]. The dependencies on water fugacity of the concentrations of h•, p•, // VMe , and H /Me for the charge neutrality condition given by Equation (21b) are derived in the Appendix and illustrated in Figure 3. H /Me

// / − VMe },

For clarity, the reader should note that the symbols p• for protons and (OH)iO for hydroxyl ions are used interchangeably in this paper, much as the symbols h• for electron holes and i MeMe for polarons are used interchangeably. The terminologies p• and h• reflect the high i mobility of these defects, while the nomenclatures (OH)iO and MeMe indicate their equilibrium locations within crystalline grains. Although infrared spectroscopy in the 3 µm region detects O-H bonds, it is the movement of a hydrogen ion from one oxygen site to the next that gives rise to protonic charge transport. As the charge neutrality condition changes from one representative of anhydrous conditions to one appropriate for hydrous conditions, so do the concentrations of vacancies and interstitials on the three sublattices of olivine. The dependencies on fH2O of the concentrations // of VMe , VSi////, O i// , and related point defect associates formed between vacancies and protons are summarized in Table 2 for the four limiting charge neutrality conditions represented in // // ], [p•]=2[ VMe ], [h•]=2[ H /Me], and [p•]=2[ H /Me]. The concentration of Equation (21b): [h•]=2[VMe each point defect is proportional to water fugacity to a power r: [ ] ∝ ( fH2O)r.

Figure 3. Log-log plot of point defect concentration versus water fugacity for the general charge neutrality // ] + [H /Me ] based on the defect reactions and equilibrium conditions presented in condition [ h i ] + [ pi ] = 2[ VMe i the Appendix. Under anhydrous conditions (left region), charge neutrality is dominated by h i ≡ MeMe and // . As water fugacity increases, pi ≡ (OH)iO replaces h• as the positively charged majority point defect. As VMe i // / // water fugacity increases further, the defect associate H /Me ≡ {(OH)OH as the negatively − VMe } replaces VMe charged majority point defect.

388

Kohlstedt

Table 2. Dependencies of point defect concentrations on water fugacity for four charge neutrality conditions, expressed as the exponent r in the relationship [ ] ∝ fHr 2 O . [ Fe iMe ] [(OH)iO ] x // /// // / [ VMe ] [H /Me ] [(2H)xMe ] [O i// ] [ VSi//// ] [HSi ] [(2H)Si ] [(3H)Si ] [(4H)Si ] [hi ] [ pi ] // [ h i ] = 2[ VMe ] i

[p ] = [h

i

// ] 2[ VMe

] = [H /Me ]

[ pi ] = [H /Me ]

0

0

1/2

1/2

1

0

0

1/2

1

3/2

2

−1/6

1/3

1/3

2/3

1

1/3

2/3

1

4/3

5/3

2

1/4

−1/2

3/4

1/4

1

−1/2

−1

−1/4

1/2

5/4

2

0

0

1/2

1/2

1

0

0

1/2

1

3/2

2

For climb-controlled deformation, the strain rate is proportional to the flux of olivine, that is, the flux of Si ions, as expressed in Equation (19). Therefore, the strain rate of a deforming olivine sample is directly proportional to the self-diffusivity of Si and thus proportional to the concentration of vacancies on the silicon sublattice, as given by Equation (20). In turn, the concentration of vacancies on the Si sublattice is directly dependent on water fugacity, as indicated in Table 2. Furthermore, defect associates formed between Si vacancies and protons provide an additional reservoir of vacancies that will increase the silicon self-diffusivity. The total concentration of Si vacancies, [ VSitot ], can be expressed in the following manner: /// // / [ VSitot ] = [ VSi//// ] + [H Si ] + [(2H)Si ] + [(3H)Si ] + [(4H)×Si ]

(22)

where /// H Si ≡ {(OH) iO − VSi//// }///

(23a )

// (2H)Si ≡ {2(OH) iO − VSi//// }//

(23b)

/ (3H)Si

≡ {3(OH) iO

− VSi//// }/

(23c)

(4H)×Si

≡ {4(OH) iO

− VSi//// }×

(23d )

Since hydrogen ions diffuse much more rapidly than silicon vacancies and since the binding energy between silicon vacancies and protons is relatively small, the additional silicon vacancies introduced by the presence of protons greatly enhance the diffusivity of silicon, /// // / which is directly proportional to [ VSitot ]. The concentrations of VSi////, H Si , (2H)Si , (3H)Si , and × tot (4H)Si as well as [ VSi ] are plotted as functions of water fugacity in Figure 4. With ε ∝ DSi ∝ [ VSitot ]

(24) fH1.22O± 0.4,

the observed approximately linear dependence of strain rate on water fugacity, ε ∝ (Mei and Kohlstedt 2000a; Hirth and Kohlstedt 2003; Karato and Jung 2003) can be explained if the largest population of silicon vacancies is that formed as point defect associates with // ≡ {(2OH) iO − VSi//// }// ≡ {2p i − VSi//// }//, with charge neutrality given by [p•] two protons, (2H)Si / = [H Me ], as summarized in Table 2. Finally, the strain rate can be written in terms of the concentration of protons by introducing the experimentally determined relationship between water fugacity and hydroxyl concentration (Kohlstedt et al. 1996; Zhao et al. 2004), as has been done by Hirth and Kohlstedt (2003) and Karato and Jung (2003).

CONCLUDING REMARKS // Two points deserve comment. (1) Why are defect associates formed between VMe and

Water in High-T Rock Deformation

389

Figure 4. Log-log plot of the concentration of silicon vacancies and related defect associates, which form between silicon vacancies and one, two, three, or four hydroxyl ions. The total silicon vacancy concentration, [ VSitot ], as given by Equation (22) is indicated by the black circles. The slopes of these curves reflect the slopes summarized in Table 2; the relative locations of the curves have been selected to correspond to the observed dependence of strain rate on water fugacity. As in Figure 3, the general // charge neutrality condition [ h i ] + [ pi ] = 2[ VMe ] + [H /Me ] was used. If strain rate is controlled by the rate of dislocation climb, then the dependence of strain rate should follow that of [ VSitot ].

(OH) iO used in the charge neutrality condition rather than those formed between VSi//// and one or more (OH) iO ions? (2) Are there other constraints on the rate of dislocation climb in olivine?

First, theoretical analyses of mechanisms of incorporation of protons in olivine indicate // the importance of point defect associates involving one or more (OH) iO ions with a VMe //// and/or a VSi (Brodholt and Refson 2000; Braithwaite et al. 2003). These studies suggest that the most energetically favorable mechanism for introducing protons into olivine is as (4H)×Si ≡ {4(OH) iO − VSi//// }×. Since the O-H stretching band that should be produced by this point defect associate is not recorded in infrared spectra from olivine, it was concluded that / (3H)Si ≡ {3(OH) iO − VSi//// } defects will dominate (Braithwaite et al. 2003). However, it should be kept in mind that these calculations are carried out at zero pressure and temperature, so that their application to high-pressure, high-temperature conditions is unclear. Also, the uncertainty in the calculated band energies is at least 10%. A marked increase in the concentration of VSitot in going from an anhydrous to a hydrous environment is clearly indicated by the greater than three order of magnitude increase in DSi produced by increasing the water fugacity from a dry (one-atmosphere) to a wet (2 GPa) environment (Dohmen et al. 2002; Chakraborty and Costa 2004). However, at a confining pressure of 2 GPa and a temperature of 1300 °C (corresponding to a water fugacity of >10 GPa), DMe is almost five orders of magnitude larger than DSi. Since the diffusivities of metal vacancies and silicon vacancies are of the same order of magnitude (Mackwell et al. 1988; tot Wanamaker 1994), the relation [ VMe ] >> [ VSitot ] must hold (see Eqn. 20). Therefore, the concentration of defect associates formed between Me vacancies and protons far exceeds that formed between Si vacancies and protons, such that H /Me is the appropriate defect associate to use in the charge neutrality condition.

390

Kohlstedt

Second, from Equation (4), at 1400 °C in 1 h, dislocations in olivine can climb a distance of only ~1 nm under anhydrous conditions and a few hundred nanometers under hydrous conditions (fH2O ≈ 10 GPa) based on published data for DSi. (Dohmen et al. 2002; Chakraborty and Costa 2004). If this calculation is correct, then deformation of olivine is unlikely to involve a significant amount of climb, at least under anhydrous conditions. However, three observations indicate that climb contributes significantly to high-temperature creep of olivine, even under anhydrous conditions. (i) Prismatic (edge) dislocation loops with diameters of ~0.1 µm in olivine collapse in ~1 h at 1300 °C (Goetze and Kohlstedt 1973). Their collapse requires diffusion of all of the ionic species (i.e., entire unit cells or olivine molecules) at a rate approximately three orders of magnitude faster than values reported for Si self-diffusion (Dohmen et al. 2002). (ii) Analyses of the shape change of olivine single crystals deformed at one-atmosphere and high temperatures also require a significant amount of climb, with climb contributing 20 to 30% of the measured strain (Durham and Goetze 1977). (iii) Low-angle tilt boundaries are prominent features in olivine grains deformed under anhydrous, as well as under hydrous, conditions (e.g., Mackwell et al. 1985; Bai and Kohlstedt 1992). Formation of such boundaries requires diffusion over distances of at least a few nanometers to a few tens of nanometers. All three of these observations indicate an inconsistency between the measured rates of Si self-diffusion and the kinetics of dislocation climb that remains to be resolved.

ACKNOWLEDGMENTS This material is based upon work supported by the National Science Foundation under Grant EAR-0439747. The careful mathematical scrutiny of Shushu Chen and the insightful comments of Georg Dresen, Hans Keppler, and Steve Mackwell are greatly appreciated. The thoughtful teaching of Hermann Schmalzried and critical discussions with Quan Bai provided the essential framework for writing this paper.

REFERENCES Avé Lallemant HG (1978) Experimental deformation of diopside and websterite. Tectonophys 48:1-27 Avé Lallemant HG, Carter NL (1970) Syntectonic recrystallization of olivine and modes of flow in the upper mantle. Geol Soc Am Bull 81:2203-2220 Bai Q, Mackwell SJ, Kohlstedt DL (1991) High-temperature creep of olivine single crystals: 1. Mechanical results for buffered samples. J Geophys Res 96:2441-2463 Bai Q, Kohlstedt DL (1992) High-temperature creep of olivine single crystals, 2. Dislocation structures. Tectonophys 206:1-29 Bardeen J, Herring C (1952) Diffusion in alloys and the Kirkendall effect. In: Imperfections in Nearly Perfect Crystals. John Wiley, p 261-288 Blacic JD (1972) Effects of water in the experimental deformation of olivine. In: Flow and Fracture of Rocks. Heard HC, Borg IY, Carter NL, Raleigh CB (eds) Am Geophys Union, p 109-115 Blacic JD (1975) Plastic deformation mechanisms in quartz: The effect of water. Tectonophys 27:271-294 Blacic JD, Christie JM (1984) Plasticity and hydrolytic weakening of quartz single crystals. J Geophys Res 89:4223-4239 Boland JN, Tullis TE (1986) Deformation behavior of wet and dry clinopyroxenite in the brittle to ductile transition region, In: Mineral and Rock Deformation: Laboratory Studies. Geophys Monogr Ser, BE Hobbs, HC Heard (eds) Am Geophys Union, 36:35-49 Borch RS, Green HW II (1989) Deformation of peridotite at high pressure in a new molten salt cell: Comparison of traditional and homologous temperature treatments. Phys Earth Planet Inter 55:269-276 Braithwaite JS, Wright K, Catlow CRA (2003) A theoretical study of the energetics and IR frequencies of hydroxyl defects in forsterite. J Geophys Res 108(B6), 2284, doi:10.1029/2002JB002126 Brodholt JP, Refson K (2000) An ab initio study of hydrogen in forsterite and a possible mechanism for hydrolytic weakening. J Geophys Res 105:18977-18992 Bystricky M, Kunze K, Burlini L, Burg J-P (2000) High shear strain of olivine aggregates: Rheological and seismic consequences. Science 290:1564-1567 Cannon WR, Langdon TG (1988) Review: Creep of ceramics. Part 2, An examination of flow mechanism. J Mater Sci 23:1-20

Water in High-T Rock Deformation

391

Chakraborty S (1997) Rates and mechanisms of Fe-Mg interdiffusion in olivine at 980° to 1300 °C. J Geophys Res 102:12317-12331 Chakraborty S, Costa F (2004) Fast diffusion of Si and O in San Carlos olivine under hydrous conditions. Geochim Cosmochim Acta 68:A275 Chen J, Inoue T, Weidner DJ, Wu Y, Vaughan MT (1998) Strength and water weakening of mantle minerals, olivine, wadsleyite and ringwoodite, Geophys Res Lett 25:575-578 Chen S, Hiraga T, Kohlstedt DL (2006) Water weakening of clinopyroxene in the dislocation creep regime. J Geophys Res (in press) Chopra PN, Paterson MS (1984) The role of water in the deformation of dunite. J Geophys Res 89:78617876 Dimanov A, Dresen G, Xiao X, Wirth R (1999) Grain boundary diffusion creep of synthetic anorthite aggregates: The effect of water. J Geophys Res 104:10483-10497 Dimos D, Wolfenstine J, Kohlstedt DL (1988) Kinetic demixing and decomposition of multicomponent oxides due to a nonhydrostatic stress. Acta Metall 36:1543-1552 Dohmen R, Chakraborty S, Becker H-W (2002) Si and O diffusion in olivine and implications for characterizing plastic flow in the mantle. J Geophys Res Vol 29, No 21, 2030, doi:10.1029/2002GL015480 Dorn JE (1956) Some fundamental experiments on high temperature creep. In: Creep and Fracture of Metals at High Temperatures. Her Majesty’s Stationary Office, London, pp 89-132 Doukhan JC, Trépied L (1985) Plastic deformation of quartz single crystals. Bull Mineral 108:97-123 Durham WB, Goetze C (1977) Plastic flow of oriented single crystals of olivine. J Geophys Res 5737-5753 Evans B, Kohlstedt DL (1995) Rheology of rocks. In: Rock Physics and Phase Relations: A Handbook of Physical Constants. TJ Ahrens (ed), American Geophysical Union, p 148-165 Evans HE, Knowles G (1977) A model of creep in pure materials. Acta Metall 25:963-975 Evans HE, Knowles G (1978) Dislocation creep in non-metallic materials. Acta Metall 26:141-145 Frost HJ, Ashby MF (1982) Deformation-Mechanism Maps: The Plasticity and Creep of Metals and Ceramics. Pergamon Press Fujino K, Nakazaki H, Hitosshi M, Karato S-i, Kohlstedt DL (1993) TEM observation of dissociated dislocations with b = [01] in naturally deformed olivine. Phys Earth Planet Inter 78:131-137 Gerard O, Jaoul O (1989) Oxygen diffusion in San Carlos olivine. J Geophys Res 94:4119-4128 Gleason GC, Tullis J (1995) A flow law for dislocation creep of quartz aggregates determined with the molten salt cell. Tectonophys 247:1-23 Goetze C, Kohlstedt DL (1973) Laboratory study of dislocation climb and diffusion in olivine. J Geophys Res 78:5961-5971 Gordon RS (1973) Mass transport in the diffusional creep of ionic solids. J Am Ceram Soc 56:147-152 Griggs DT, Blacic JD (1965) Quartz: Anomalous weakness of synthetic crystals. Science 147:292-295 Griggs DT (1967) Hydrolytic weakening of quartz and other silicates. Geophys J R Astr Soc 14:19-31 Griggs DT (1974) A model of hydrolytic weakening in quartz. J Geophys Res 79:1653-1661 Griggs DT, Blacic JD (1965) Quartz: Anomalous weakness of synthetic crystals. Science 147:292-295 Gueguen Y (1979) High-temperature creep of olivine: Evidence for control by edge dislocations. Geophys Res Lett 6:357-360 Hermeling J, Schmalzried H (1984) Tracer diffusion of the Fe cations in olivine (FexMg1-x)2SiO4 (III). Phys Chem Minerals 11:161-166 Hier-Majumder S, Mei S, Kohlstedt DL (2005a) Water weakening of clinopyroxenite in diffusion creep, J Geophys Res 110:1-12, B07406, doi:10.1029/2004JB003414 Hier-Majumder S, Anderson IM, Kohlstedt DL (2005b) Influence of protons on Fe-Mg interdiffusion in olivine. J Geophys Res 110:B02202, doi:10.1029/2004JB003292 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: Implications for rheology, melt extraction and the evolution of the lithosphere, Earth Planet Sci Lett 144:93-108 Hirth G, Kohlstedt DL (2003) Rheology of the upper mantle and the mantle wedge: A view from the experimentalists. In: Inside the Subduction Factory. J Eiler (ed) Geophysical Monograph, American Geophysical Union, 138:83-105 Hirth JP, Lothe J (1968) Theory of Dislocations, McGraw-Hill Book Company Hobbs BE (1981) The influence of metamorphic environment upon the deformation of materials. Tectonophys 78:335-383 Hobbs BE (1983) Constraints on the mechanism of deformation of olivine imposed by defect chemistry. Tectonophys 78:335-383 Hobbs BE (1984) Point defect chemistry of minerals under hydrothermal environment. J Geophys Res 89: 4026-4038 Hobbs BE, McLaren AC, Paterson MS (1972) Plasticity of single crystals of synthetic quartz. In: Flow and Fracture of Rocks (The Griggs Volume). HC Heard, IY Borg, NL Carter, CB Raleigh (eds) Am Geophys Union, 16:29-53

392

Kohlstedt

Houlier B, Cheraghmakani M, Jaoul O (1990) Silicon diffusion in San Carlos olivine. Phys Earth Planet Inter 62:329-340 Jaoul O, (1990) Multicomponent diffusion and creep in olivine. J Geophys Res 95:17631-17642 Jaoul O, Tullis J, Kronenberg AK (1984) The effect of varying water content on the creep behavior of Heavitree quartzite. J Geophys Res 89:4298-4312 Jung H, Karato S-i (2001) Water-induced fabric transitions in olivine. Science 293:1460–1463 Karato S-i (1989) Defects and plastic deformation in olivine. In: Rheology of Solids and of the Earth. Karato S-i, Toriumi M (eds) Oxford Science Publications, p 176-208 Karato S-i, Ogawa M (1982) High pressure recovery in olivine: Implications for creep mechanisms and creep activation volume. Phys Earth Planet Inter 28:102-117 Karato S-i, Jung H (2003) Effects of pressure on high-temperature dislocation creep in olivine. Phil Mag 83: 401–414 Karato S-i, Paterson MS, Fitz Gerald JD (1986) Rheology of synthetic olivine aggregates: Influence of grain size and water. J Geophys Res 91:8151-8176 Kekulawala KRSS, Paterson MS, Boland JN (1978) Hydrolytic weakening in quartz. Tectonophys 46:T1-T6 Kekulawala KRSS, Paterson MS, Boland JN (1981) An experimental study of the role of water in quartz deformation. In: Mechanical Behavior of Crustal Rocks. NL Carter, M Friedman, JM Logan, DW Stearns (eds) Am Geophys Union 24:49-60 Koch PS, Christie JM, Ord A, George RP Jr (1989) Effect of water on the rheology of experimentally deformed quartzites. J Geophys Res 94:13975-13997 Kocks UF, Argon AS, Ashby MF (1975) Thermodynamics and kinetics of slip. In: Progress in Materials Sciences. Chalmers B, Christian JW, Massalski TB (eds) Pergamon Press Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kohlstedt DL, Mackwell SJ (1999) Solubility and diffusion of ‘water’ in silicate minerals. In: Microscopic Properties and Processes in Minerals. Wright K, Catlow R (eds) Kluwer Academic Publishers, p 539559 Kohlstedt DL, Evans B, Mackwell SJ (1995) Strength of the lithosphere: Constraints imposed by laboratory experiments. J Geophys Res 100:17587-17602 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kröger FA, Vink HJ (1956) Relation between the concentrations of imperfections in crystalline solids. In: Solid State Physics 3. Seitz F, Turnball D (eds) Academic Press, p 307-435 Kronenberg AK (1994) Hydrogen speciation and chemical weakening of quartz. Rev Mineral 29:123-176 Kronenberg AK, Tullis J (1984) Flow strengths of quartz aggregates: Grain size and pressure effects due to hydrolytic weakening. J Geophys Res 89:4281-4297 Luan FC, Paterson MS (1992) Preparation and deformation of synthetic aggregates of quartz. J Geophys Res 97:301-320 Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: Implications for water in the mantle. J Geophys Res 95:5079-5088 Mackwell SJ, Paterson MS (1985) Water-related diffusion and deformation effects in quartz at pressure of 1500 and 300 MPa. In: Point Defects in Minerals. Schock RN (ed) American Geophysical Union, p 141-150 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90:11319-11333 Mackwell SJ, Dimos D, Kohlstedt DL (1988) Transient creep of olivine: Point defect relaxation times. Phil Mag A 57:779-789 Mainprice DH, Paterson MS (1984) Experimental studies of the role of water in the plasticity of quartzites. J Geophys Res 89:4257-4270 Mei S, Kohlstedt DL (2000a) Influence of water on plastic deformation of olivine aggregates: 2. Dislocation creep regime. J Geophys Res 105:21471-21481 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates 1. Diffusion creep regime, J Geophys Res 105:21457-21469 Mott NF (1951) The mechanical properties of metals. Proc Phys Soc 64:729-742 Mott NF (1956) A discussion of some models of the rate-determining process in creep. In: Creep and Fracture of Metals at High Temperatures. Her Majesty’s Stationary Office, London, pp 21-24 Mukerjee AK, Dird JE, Dorn JE (1969) Experimental correlations for high-temperature creep. Trans ASM 62: 155-179 Nabarro FRN (1967) Steady-state diffusional creep. Phil Mag 16:231-237 Nix WD, Gasca-Neri R, Hirth JP (1971) A contribution to the theory of dislocation climb. Phil Mag 23:13391349

Water in High-T Rock Deformation

393

Ord A, Hobbs BE (1986) Experimental control of the hydrolytic weakening effect in quartz. In: Mineral and Rock Deformation: Laboratory Studies. Hobbs BE, Heard HC (eds) Am Geophys Union, 36:51-72 Paterson MS (1989) The interaction of water with quartz and its influence in dislocation flow - an overview. In: Rheology of Solids and of the Earth. Karato S-i, Toriumi M (eds) Oxford Univ Press, p 107-142 Paterson MS, Kekulawala KRSS (1979) The role of water in quartz deformation. Bull Mineral 102:92-98 Poirier JP (1978) Is power-law creep diffusion-controlled? Acta Metall 26:629-637 Poirier JP (1979) Reply to “Diffusion-controlled dislocation creep: A defense”. Acta Metall 27:401-403 Poirier JP (1985) Creep of Crystals: High-temperature Deformation Processes in Metals, Ceramics and Minerals. Cambridge University Press Post AD, Tullis J (1998) The rate of water penetration in experimentally deformed quartzite: implications for hydrolytic weakening. Tectonophys 295:117-137 Post AD, Tullis J, Yund RA (1996) Effects of chemical environment on dislocation creep of quartzite. J Geophys Res 101:22143-22155 Poumellec B, Jaoul O (1984) Influence of pO2 and pH2O on the high temperature plasticity of olivine. In: Deformation of Ceramic Materials II. Tressler RE, Bradt RC (eds) Materials Science Research, Plenum Press, 18:281-305 Ready DW (1966) Chemical potentials and initial sintering in pure metals and ionic compounds. J Appl Phys 37:2309-2312 Ricoult DL, Kohlstedt DL (1985) Experimental evidence for the effect of chemical environment on the creep rate of olivine. In: Point Defects in Minerals. Schock RN (ed) American Geophysical Union, p 171-184 Ruoff AL (1965) Mass transfer problems in ionic crystals with charge neutrality. J Appl Phys 36:2903-2907 Rybacki E, Dresen G (2000) Dislocation and diffusion creep of synthetic anorthite aggregates. J Geophys Res 105:26017-26036 Ryerson FJ, Durham WB, Cherniak DJ, Lanford WA (1989) Oxygen diffusion in olivine: Effect of oxygen fugacity and implications for creep. J Geophys Res 94:4105-4118 Schmalzried H (1981) Solid State Reactions. 2nd ed. Verlag Chemie Schmalzried H (1995) Chemical Kinetics of Solids. VCH Publishers Sherby OD, Weertman J (1979) Diffusion-controlled dislocation creep: A defense. Acta Metall 27:387-400 Sherby OD Burke PM (1967) Mechanical behavior of crystalline solids at elevated temperature. In: Progress in Materials Science. Chalmers B, Hume-Rothery W (eds) Pergamon Press, 13:325-390 Stünitz H, Fitz Gerald JD, Tullis J (2003) Dislocation generation, slip systems, and dynamic recrystallization in experimentally deformed plagioclase single crystals. Tectonophys 372:215-233 Tullis J, Yund RA (1980) Hydrolytic weakening of experimentally deformed Westerly granite and Hale albite rock. J Struct Geol 2:439-451 Tullis J, Yund RA (1989) Hydrolytic weakening of quartz aggregates: The effects of water and pressure on recovery. Geophys Res Lett 16:1343-1346 Tullis J, Yund RA (1991) Diffusion creep in feldspar aggregates: Experimental evidence. J Struct Geol 13: 987-1000 Tullis J, Shelton GL, Yund RA (1979) Pressure dependence of rock strength: implications for hydrolytic weakening, Bull Mineral 102:110-114 Vander Sande JB, Kohlstedt DL (1976) Observation of dissociated dislocations in deformed olivine. Phil Mag 34:653-658 Wanamaker BJ (1994) Point defect diffusivities in San Carlos olivine derived from reequilibration of electrical conductivity following changes in oxygen fugacity Geophys Res Lett 21:21-24 Wang Z, Hiraga T, Kohlstedt DL (2004) Effect of H+ on Fe-Mg interdiffusion in olivine, (Mg,Fe)2SiO4. Appl Phys Lett 85:209-211 Weertman J (1955) Theory of steady-state creep based on dislocation climb. J Appl Phys 26:1213-1217 Weertman J (1957a) Steady-state creep through dislocation climb. J Appl Phys 28:362-364 Weertman J (1957b) Steady-state creep of crystals. J Appl Phys 28:1185-1189 Weertman J (1978) Creep laws for the mantle of the Earth. Phil Trans R Soc London A 288:9-26 Weertman J (1999) Microstructural mechanisms of creep. In: Mechanics and Materials: Fundamentals and Linkages. MA Meyers, RW Armstrong, H Kirschner (eds) John Wiley & Sons, p 451-488 Zhang SQ, Karato S-i (1995) Lattice preferred orientation of olivine aggregates deformed in simple shear. Nature 375:774-777 Zhao YH, Ginsberg SG, Kohlstedt DL (2004) Solubility of hydrogen in olivine: Effects of temperature and Fe content, Contrib Mineral Petrol 147:155-161, doi:10.1007/s00410-003-0524-4

394

Kohlstedt APPENDIX

Charge neutrality To construct Figure 3, the dependencies on water fugacity of the concentrations // , and H /Me —with the charge neutrality condition given by Equation (21b), of h•, p•, VMe i i // / [ h ] + [ p ] = 2[ VMe ] + [H Me ] —were determined using a subset of the mass action equations for the reactions listed below. In each case, the dependencies of the defect concentrations on the activity of the oxide phase, aMeO, and on oxygen fugacity, fO2, have been incorporated into the reaction constant, Ki, as illustrated in the first reaction. It is also generally assumed that [ Me×Me ] ≈ 1, since [ Me×Me ] deviates from unity only by the concentration of cation vacancies. Reaction 1: 1 // i O 2 + 3Me×Me → 2 MeMe + VMe + MeO(g) 2

( A1a )

i // K1' [ Me×Me ]3 f O122 = [ MeMe ]2 [VMe ]a MeO

( A1b)

// K1 = [h i ]2 [ VMe ]

⇒ [h i ] =

1 // 1/2 K11/2 [ VMe ]

( A1c)

Reaction 2: 1 1 i H 2O + O 2 + 2 Me×Me → MeMe + H /Me + MeO(g) 2 4

(A2a )

K 2 f H1/22 O = [h i ][H /Me ]

(A2b)

// H 2O + 2O×O + Me×Me → 2(OH) iO + VMe + MeO(g)

( A3b)

Reaction 3:

// K 3 f H 2 O = [p i ]2 [VMe ]

⇒

[p i ] =

K 31/2 f H1/22 O // 1/2 [VMe ]

(A A3b)

Reaction 4: H 2O + O×O + Me×Me → (OH) Oi + H /Me + MeO(g)

(A4a )

K 4 f H 2 O = [p i ][H /Me ]

(A4b)

// H 2O + Me×Me + VMe → 2H /Me + MeO(g)

( A5a )

Reaction 5:

// ] f H 2 O = [H /Me ]2 K 5[ VMe

⇒

// 1/2 1/2 [H /Me ] = K 51/2 [VMe ] f H 2O

(A5b)

Only three of the above five equilibrium equations that were obtained using the law of mass action are independent. In particular, if Equations (A1c), (A3b), and (A5b) are substituted into the general form of the charge neutrality condition expressed in Equation (21b), then an // ] as a function of fH2O: equation is obtained for [ VMe // / [ h i ] + [ p i ] = 2[ VMe ] + [H Me ] ⇒

Rearranging yields

K 31/2 f H1/22 O 1 // // 1/2 1/2 ] f H 2O + = 2[ VMe ] + K 51/2 [VMe // 1/2 // 1/2 K11/2 [ VMe ] [ VMe ]

(A6a )

Water in High-T Rock Deformation

f H1/22 O

⎛⎛ 1 ⎞⎞ // 3 / 2 ⎜ ⎜ 2[ VMe ] − 1/2 ⎟ ⎟ K1 ⎠ ⎟ =⎜⎝ ⎜ K 1/2 − K 1/2 [V // ] ⎟ 3 5 Me ⎝ ⎠

(

395

2

(A6b)

)

the expression used in generating Figure 3.

Flux equations for a semi-conducting silicate Fe-bearing olivine, (Mg,Fe)2SiO4, is a semi-conducting silicate; that is, electron holes are the positively charged majority point defect, as expressed in Equation (21b). In a manner analogous to that used for ionic species, the flux of holes is then given by jh i = −

(

DhC h DC ∇η h i = − h h ∇µ h i + F ∇Φ RT RT

)

( A7)

consistent with Equations (9) and (10). To maintain local electrical neutrality in an open circuit system, the flux of holes must be of similar magnitude to the flux of ions. Since the mobility of holes is much greater than the mobility of the ionic species (i.e., DhCh >> DMeCMe, DOCO, DSiCSi),

(

)

∇η h i = ∇µ h i + F ∇Φ ≈ 0

( A8a )

− F ∇Φ ≈ ∇µ h i

( A8b)

that is,

Recalling that the flux of silicon ions in Equation (9b) is given by jSi 4+ = −

(

DSiCSi D C ∇ηSi 4+ = − Si Si ∇µ Si 4+ + 4 F∇Φ RT RT

)

(A9a )

then using the relation Si4+ = Si + 4h• yields

(

)

DSiCSi D C ∇µ Si 4+ − 4∇µ h i = − Si Si ∇µ Si RT RT

jSi 4+ = −

( A9b)

Assuming that local thermodynamic equilibrium is established (consistent with the premises of irreversible thermodynamics), then from the reaction Si + O2 = SiO2, (local) equilibrium requires that µSi + µO = µSiO2. Thus, jSi 4+ = −

(

DSiCSi ∇µ SiO2 − ∇µ O 2 RT

)

( A9c)

Likewise j Me2+ = −

DMeC Me D C ⎛ 1 ⎞ ∇µ Me = − Me Me ⎜ ∇µ MeO − ∇µ O 2 ⎟ RT RT ⎝ 2 ⎠

(A10)

With DMeCMe >> DOCO, DSiCSi, 1 ∇µ MeO ≈ ∇µ O 2 2

( A11)

Therefore, from the above equations, jSi 4+ = −

(

DSiCSi ∇µ SiO2 − 2∇µ MeO RT

)

( A9d )

396

Kohlstedt

which is identical to Equation (15) and leads directly to Equation (17), jSi 4+ = −

CSi DSi DO ∇µ Me2 SiO 4 RT DO + 4 DSi

( A9e)

17

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 397-420, 2006 Copyright © Mineralogical Society of America

The Effect of Water on Mantle Phase Transitions Eiji Ohtani and K. D. Litasov Institute of Mineralogy, Petrology and Economic Geology Tohoku University Sendai, Miyagi-ken 980-8578, Japan e-mail: [email protected]

INTRODUCTION Water has played an important role in the Earth’s evolution. The incorporation of water as hydroxyl into solid mineral phases or as coexisting hydrous fluids and melts affects the chemical and physical properties of crust and mantle constituents, i.e., it weakens rocks and minerals, reduces viscosity and strength of the materials, and depresses dramatically the melting temperature of silicate minerals (e.g., Karato 1990; Inoue 1994; Hirth and Kohlstedt 1996; Chen et al. 1998; Kubo et al. 1998; Mei and Kohlstedt 2000). Many recent studies have suggested the possible existence of water in the Earth’s mantle especially in the transition zone (e.g., Smyth and Frost 2002; Ohtani et al. 2004; Litasov et al. 2005a; Hae et al. 2006), where wadsleyite and ringwoodite can accommodate up to 3 wt% of H2O in their structures (e.g., Kohlstedt et al. 1996). Low-velocity zones observed seismologically at the top of the 410 km discontinuity may indicate the existence of trapped high-density melt (e.g., Revenaugh and Sipkin 1994; Song et al. 2004; Matsukage et al. 2005; Sakamaki et al. 2006), which is likely to be hydrous as it is not possible to melt the base of the upper mantle without water at these conditions. Electrical conductivity anomalies in the upper mantle and transition zone that are related to subduction zones have also been interpreted as an effect of water in the mantle (e.g., Fukao et al. 2004; Tarits at el. 2004; Hae et al. 2006; Koyama et al. 2006). Studies of the kinetics of the hydrous olivine-wadsleyite transformation (Ohtani et al. 2004; Hosoya et al. 2005) seem to be consistent with seismological observations (e.g., Koper et al. 1998) that indicate the absence of a metastable olivine wedge in subducting slabs. Such a metastable wedge would be expected as a result of the sluggish olivine-wadsleyite transformation under anhydrous conditions (Rubie and Ross 1994). Studies of the elasticity of hydrous wadsleyite and ringwoodite indicate that P- and S- wave velocities of the transition zone are consistent with the existence of hydrated wadsleyite and ringwoodite (Inoue et al. 2004; Jacobsen et al. 2004). These data indicate that a significant amount of water may be stored in the mantle especially in the transition zone. Seismic discontinuities at 410 and 660 km depths are well established on a global scale to the point where they occur in reference velocity models such as PREM (Dziewonski and Anderson 1981). These discontinuities are usually attributed to the phase transformations of olivine in mantle peridotite. Olivine α-(Mg,Fe)2SiO4 transforms to wadsleyite β-(Mg,Fe)2SiO4 at a depth of approximately 410 km, and ringwoodite γ-(Mg,Fe)2SiO4 decomposes to perovskite (Mg,Fe)SiO3 and magnesiowustite (Mg,Fe)O at approximately 660 km depth. The latter transformation is frequently termed the post-spinel transformation. The topography and sharpness of these discontinuities depend on mantle temperatures, chemical compositions and mineral proportions (e.g., Agee 1998; Weidner and Wang 2000; Frost 2003). According to most seismological studies the 410 and 660 km discontinuities are sharp, and the change of density and velocity occurs over a small depth interval, 4-35 km for 1529-6466/06/0062-0017$05.00

DOI: 10.2138/rmg.2006.62.17

398

Ohtani & Litasov

the 410 km discontinuity and <5 km for the 660 km discontinuity (e.g., Benz and Vidale 1993; Shearer 2000). Topography of the discontinuities has also been reported in seismic studies; i.e., elevations of 410 km and depression of 660 km in subduction zones and the opposite effects in hot mantle plume regions (e.g., Flanagan and Shearer 1998). Until recently these variations were believed to be consistent with the experimentally determined Clapeyron slopes of the olivine-wadsleyite and post-spinel transformations. For example, Ito and Takahashi (1989) reported a Clapeyron slope of −3.0 MPa/K, and Bina and Helffrich (1994) calculated a Clapeyron slope of −2.0 MPa/K for the post-spinel transformation. However, recent in situ Xray diffraction studies of the post-spinel transformation in Mg2SiO4 and pyrolite compositions have implied a much gentler negative slope to this boundary (−0.4 to −1.3 MPa/K) (Katsura et al. 2003; Fei et al. 2004a; Litasov et al. 2005b) based on the newly established MgO and Au pressure scales (e.g., Speziale et al. 2001; Shim et al. 2002; Tsuchiya 2003). These results make it difficult to explain the topography of the 660 km discontinuity of about ±20 km (e.g., Gu et al. 1998) through thermal perturbations and indicate that the post-spinel phase transformation may account for less than a half of variations in depth of the 660 km discontinuity. Additional explanations for the observed topographic variations of the 660 km discontinuity are required and may reflect the influence of minor elements or volatiles. In addition, two other phase transformations, which could be important in mantle dynamics and velocity structure of the Earth, are considered here: (1) the wadsleyite-ringwoodite transformation in peridotite, which is believed to be responsible for the seismic discontinuity at approximately 520 km (e.g., Shearer 1996), and (2) the post-garnet (garnet to perovskite) transformation in basalt (eclogite) composition (e.g., Ringwood 1994; Hirose et al. 1999; Litasov et al. 2004). The basaltic oceanic crust component of a subducting slab transforms to eclogite as the slab descends into the deep mantle. This basalt component may accumulate above 660 km due to a density crossover with peridotite mantle and it may create a complex velocity structure near the 660 km discontinuity (Ringwood 1994). In this paper, we summarize recent studies on the effect of water on the location of phase boundaries and phase relations of mantle minerals. In addition, we examine the effect of water on the kinetics of these phase transformations. We also discuss the implications of these recent results with respect to seismic discontinuities and mantle dynamics.

RECENT PROGRESS ON PRESSURE SCALES FOR THE DETERMINATION OF PHASE BOUNDARIES IN MANTLE MINERALS The boundaries of the phase transformations of mantle minerals were previously determined by the quenching technique (e.g., Ito and Takahashi 1989). The pressure scale in this technique is based on the calibration curve, i.e., the relation between the generated pressure and press load determined separately using pressure standards, which were determined by in situ X-ray diffraction technique. However, this procedure contains a large uncertainty due to the reproducibility of the experiments and unexpected temperature dependency of the generated pressure due to a different furnace assembly and starting materials. In order to overcome this difficulty, in situ X-ray diffraction techniques have been introduced. We can observe the reactions occurring in the samples at high pressure and temperature, and we can also measure the pressure values generated at high pressure and temperature by determination of the cell parameters of the standard materials such as Au, NaCl, and MgO. This technique has been used extensively by using intense X-ray from synchrotron radiation (e.g., Katsura et al. 2003). An accurate pressure scale is essential for comparing the pressure of a phase boundary with the depth of a discontinuity in the mantle. A possible ambiguity in the use of Au as an in situ powder X-ray diffraction pressure scale was initially identified through the study of Irifune

Eﬀect of Water on Mantle Phase Transitions

399

et al. (1998) on the post-spinel phase boundary. Irifune et al. (1998) employed the equation of state (EOS) for Au of Anderson et al. (1989) to determine the pressure in their experiments from the unit cell parameters of Au measured in situ. Based on these data Irifune et al. (1998) proposed that the post spinel phase boundary occurred at a pressure that was approximately 2 GPa lower than that expected at the 660 km discontinuity. They suggested the possibility that a compositional change at the base of the transition zone may, therefore, cause the 660 km discontinuity, a possibility that was proposed by Ringwood (1994). Since this report of the inconsistency of the phase boundary with the 660 km discontinuity, attempts have been made to evaluate a number of pressure scales and to improve the scales at high pressure and temperature (e.g., Fei et al. 2004b). Au, Pt, and MgO have been often used as pressure markers for in situ X-ray diffraction studies at high-pressure and temperature in the multianvil apparatus and in the diamond anvil cell (e.g., Fei et al. 2004b). It has been reported that the pressure scale of Au proposed by Anderson et al. (1989) underestimates the pressure at high temperatures due to incorrect evaluation of the thermal pressure (e.g., Matsui et al. 2000), although the Au pressure scale in general has a strong stress effect at low temperature. A pressure scale is essential for accurate determination of phase transformation boundaries and their Clapeyron slopes, however there is no consistency among the different pressure scales especially at pressures above 20 GPa at high temperature. For instance, the pressure differences calculated from the EOS of Au, Pt, and MgO (e.g., Jamieson et al. 1982; Anderson et al. 1989; Shim et al. 2002; Tsuchiya 2003) may be as large as 2.5 GPa at 25 GPa and 2000 K. Fei et al. (2004b) made a comprehensive review of different pressure scales and calibrated the Au and Pt scales based on the MgO EOS (Speziale et al. 2001). Based on the above studies on the pressure scale, the equation of state of MgO is now considered to be one of the most reliable pressure scales at high pressure and temperature, since there are many reliable shock compression, elasticity, and static compression data at high pressure and temperature. The Au pressure scale was also improved by Shim et al. (2002) by using the new reliable compression data of Au. Fei et al. (2004b) made a correction of the Au pressure scale by Shim to fit the MgO pressure scale by Speziale et al. (2001). Tsuchiya (2003) presented a Au pressure scale based on molecular dynamic calculations, and it is generally consistent with that of the Au pressure scale by Fei et al. (2004b) and the MgO pressure scale by Speziale et al. (2001). Thus, we adopt the pressure scale at high temperature based on the unit cell volume of Au using EOS by Tsuchiya (2003) and MgO using EOS by Speziale et al. (2001) in the present review. We found that, the pressures calculated by the MgO scale by Speziale et al. (2001) are generally consistent with those by Au scale by Tsuchiya (2003). We estimated that P (MgO; Speziale et al. 2001) − P (Au; Tsuchiya 2003) = 0.08 ± 0.36 GPa in the pressure range of 20-25 GPa and temperatures between 1500 and 2200 K. However, at lower temperatures of 1200-1500 K the pressure difference may exceed 0.4 GPa (Litasov et al. 2006). The MgO pressure scale proposed by Matsui et al. (2000) is also used in some studies (e.g., Katsura et al. 2004). This pressure scale gives the pressures generally 0.1 GPa lower than those of the MgO scale by Speziale et al. (2001). It is noteworthy that there are several other uncertainties in determination of the phase boundaries, such as the effect of pressure on Electromotive force (EMF), stress differences in sample and standard, and kinetics of the phase transformation. We can minimize the stress differences in the sample and standard by annealing at high pressure and temperature and by placing them in different parts in the furnace assembly. We can also minimize the effect of the reaction kinetics by conducting reversal runs in the in situ X-ray diffraction experiments. However, the most serious problem is the effect of pressure on EMF. The current temperature values without pressure correction for EMF is likely to be underestimated by several tens degrees based on previous estimates (e.g., Ohtani 1979).

400

Ohtani & Litasov EFFECT OF WATER ON PHASE TRANSFORMATION

Dry and wet phase boundaries in the olivine-wadsleyite transformation Water may modify the location of the olivine-wadsleyite transformation boundary as a result of the difference in water solubility between olivine and wadsleyite. At 1300-1500 K wadsleyite contains 5-40 times more water than olivine (e.g., Kohlstedt et al. 1996; Chen et al. 2002), therefore the phase transition boundary can be shifted to lower pressures by the addition of water. The water content in olivine is usually determined by Fourier Transform Infrared Spectroscopy (FTIR) (Paterson et al. 1982) or Secondary Ion Mass Spectrometry (SIMS) (e.g., Kurosawa et al. 1997; Demouchy et al. 2005). Water content is estimated by the integrated intensity of the OH absorption band of FTIR measurement based on the method by Paterson et al. (1982). Recently, the FTIR calibration of Paterson et al. (1982) for the water content in olivine has been reexamined and shown to be about three times as large as those of the previous values (Bell et al. 2003). If we adopt the new calibration, the partition coefficients of water between wadsleyite and olivine at 1300-1500 K range become one third of the previous value, i.e., 1.6-13. However, we need more detailed studies of the water contents in wadsleyite using SIMS and improving the Paterson’s calibration for wadsleyite. Demouchy et al. (2005) provided extensive SIMS data for water content in wadsleyite. The olivine-wadsleyite transformation has been intensively studied both under anhydrous and hydrous conditions and the results are summarized in Figure 1. Katsura and Ito (1989) determined that the Clapeyron slope of the olivine-wadsleyite phase boundary is 2.5 MPa/K and the pressure interval of the olivine-wadsleyite transformation loop for (Mg,Fe)2SiO4 of Fo90 composition is 0.5 GPa at 1900 K. Akaogi et al. (1989) calculated the olivine-wadsleyite phase boundary using calorimetric data, and they suggested a shallower Clapeyron slope of 1.5±0.5 MPa/K, and Bina and Helffrich (1994) recalculated the data by Akaogi et al. (1989) and obtained a Clapeyron slope of 3.0 MPa/K. Recent in situ X-ray diffraction studies have shown that the Clapeyron slope is 3.6 MPa/K (Morishima et al. 1994) to 4.0 MPa/K (Katsura et al. 2004) in Mg2SiO4 and Fo90 systems. Katsura et al. (2004) estimated that the width of the binary loops of the phase transformation is about 0.4 GPa at 1900 K and about 0.6 GPa at 1600 K for Fo90. Litasov et al. (2006) performed preliminary experiments to determine the olivine-wadsleyite transformation in anhydrous and hydrous pyrolite compositions using in situ X-ray diffraction (Fig. 1). Although the Clapeyron slope of the boundary was not determined precisely, it was

(2) (1) (3)

2200

Temperature, K

2000

1800

Ol

1600

(5)

Wd

(4)

in Ol+Wd Wd

1400

1200

1000

9

10

11

12

13

14 15 16 17

Pressure, GPa

Figure 1. Comparison of the olivinewadsleyite transformation boundaries obtained by different studies (Litasov et al. 2006). (1), olivine-wadsleyite boundary in Mg2SiO4 by Morishima et al. (1994); (2), Fo90 by Katsura et al. (2004) under the dry conditions. (3), the phase boundary of pyrolite under the dry conditions (Litasov et al. 2006). (4), the phase boundary of pyrolite under the hydrous conditions (3.0 wt% H2O) (Litasov et al. 2006). (5), Chen et al. (2002). Ol, olivine; Wd, wadsleyite. Pressure was calculated based on the pressure scale of Tsuchiya (2003). Preliminary experiments by Litasov et al. (2006) indicate the field of olivine + wadsleyite broadens under the wet conditions.

Eﬀect of Water on Mantle Phase Transitions

401

found to be consistent with the slope in (Mg,Fe)2SiO4 determined by Katsura et al. (2004) under dry conditions. Using the Clapeyron slope of Katsura et al. (2004) and the Au pressure scale by Tsuchiya (2003), we can obtain a linear equation for the olivine-wadsleyite boundary in pyrolite under dry conditions, which is P (GPa) = 0.0039 T (K) + 7.47. The choice of pressure scale does not significantly affect the slope of the phase transformation. The pressure interval of olivine and wadsleyite coexistence is about 0.2-0.3 GPa in these experiments. Several experimental and theoretical studies on the effect of water on the olivinewadsleyite phase transformation and water partitioning between olivine and wadsleyite have been carried out to date. Wood (1995) argued that hydrogen, being more soluble in wadsleyite than olivine, expands the stability field of wadsleyite to lower pressures due to the effect of hydrogen on the configuration entropy of disorder in wadsleyite. Wood (1995) calculated that the presence of 500 ppm H2O in olivine would expand the olivine-wadsleyite loop interval from 7 km for the anhydrous system to about 22 km for the hydrous system. Smyth and Frost (2002), on the other hand, experimentally observed a shift of the olivine-wadsleyite phase boundary by about 1 GPa to lower pressures using Fo92-98 and the peridotite composition with Fo89 olivine both containing 3 wt% H2O. This result is consistent with that of Litasov et al. (2006) determined by in situ X-ray diffraction experiments. Chen et al. (2002) also determined the olivine-wadsleyite phase transformation boundary in Fo100 and Fo90 under dry and hydrous conditions by heating both dry and hydrous (11 wt% H2O) charges simultaneously at 1473 K and 12.6-14.7 GPa (Fig. 1). They also observed a shift of the olivine-wadsleyite phase boundary to lower pressure, but found a decrease of the pressure interval of the olivinewadsleyite loop to 0.3 GPa.

Wadsleyite-ringwoodite transformation The data for the determination of the wadsleyite-ringwoodite phase boundary are limited compared to those for the olivine-wadsleyite and the post-spinel phase boundaries in the olivine and peridotite systems. The wadsleyite-ringwoodite phase boundary in (Mg,Fe)2SiO4 determined using quenching experiments (Katsura and Ito 1989) has a significant positive Clapeyron slope of +5 MPa/K. Suzuki et al. (2000) reported a linear equation expressed as P (GPa) = 0.0069 T (K) + 8.43 for the phase boundary in Mg2SiO4 using the NaCl pressure scale (Brown 1999). Litasov and Ohtani (2003) observed a minor shift of the boundary to higher pressure in the CMAS-pyrolite system with 2 wt% H2O, whereas Kawamoto (2004) observed wadsleyite at 20 GPa and 1573 K in pyrolite with 13 wt% H2O, which suggests a shift of the phase boundary by 2.5 GPa to higher pressures compared to that observed in the anhydrous system. Inoue et al. (2001) also showed a shift of the phase boundary in Fo80-100 with 1 wt% H2O, which is consistent with the results described above, although the exact shift in pressure of the boundary has not yet been reported. There are no data on the effect of water on the width of the wadsleyite-ringwoodite transformation interval. Katsura and Ito (1989) determined the pressure interval of the wadsleyite-ringwoodite loop to be ~0.9 GPa (24 km) for anhydrous Fo90 at 1473-1873 K, whereas Frost (2003) calculated that it may be reduced to ~0.7 GPa (20 km) in a garnet peridotite. The cause of the shift of the phase boundary under water saturated conditions is not clear. It might be caused by the lower solubility of water in ringwoodite relative to wadsleyite, which is suggested by the data on the maximum water solubilities of these phases; i.e., the maximum water solubility in ringwoodite (2.6 wt%) is lower than that in wadsleyite (3.4 wt%) (e.g., Kohlstedt et al. 1996; Inoue et al. 1998).

Post-spinel transformation In Figures 2(A) and (B) we summarize recent studies on the determination of the postspinel phase boundary in Mg2SiO4 and peridotite under dry conditions using both the multianvil

402

Ohtani & Litasov Depth, km

Depth, km 600

Temperature, K

(2)

A (1)

2200

700

600 2400

(3)

Average mantle geotherm

1600

Subducting slab geotherm

1800

1600

Subducting slab geotherm

(2) (3) (4) (5)

1000

20

21

22

23

MgO Scale

Average mantle geotherm

1200

(6)

(4) (5) 1000

(1)

700

2000

1400

1400

1200

650

2200

2000

1800

B

Au Scale

Temperature, K

2400

650

24

Pressure, GPa

25

26

20

21

22

23

24

25

(6) 26

Pressure, GPa

Figure 2. Comparison between post-spinel transformation boundaries determined by different studies using (A) the Au pressure scale of Tsuchiya (2003) and (B) the MgO pressure scale of Speziale et al. (2001). The average mantle geotherm is after Akaogi et al. (1989). The cold subduction geotherm is after Kirby et al. (1996). (1) the phase boundary in (Mg,Fe)2SiO4 determined by the quenching method by Ito and Takahashi (1989). This boundary was determined by the pressure scale using some phase boundaries determined based on the Au scale by Jamieson et al.(1982) as pressure fixed points. We corrected this Au pressure scale to new scales of Au (Tsuchiya 2003) and MgO (Speziale et al. 2001). (2) the boundary in pyrolite (Litasov et al. 2005b) under dry conditions. (3), (4) the boundary in Mg2SiO4 under dry conditions determined by Katsura et al. (2003) and Fei et al. (2004a), respectively. (5) the post-spinel phase boundary in hydrous pyrolite (Litasov et al. 2005a). (6) the phase boundary in Mg2SiO4 determined by diamond anvil cell (Shim et al. 2001).

press and diamond anvil cell. Most studies made by in situ X-ray diffraction study using the multianvil press indicate that the post-spinel phase boundary is lower than that expected for the depth of the 660 km discontinuity (Figs. 2A,B). Experiments made using the diamond-anvil cell (DAC) (Chudinovskih and Boehler 2001; Shim et al. 2001) and those made with the multi-anvil press with an improved technique by taking into account of kinetics based on following reliable pressure scales (Katsura et al. 2003; Fei et al. 2004a; Litasov et al. 2005b) have shown the pressure of the phase boundary under dry conditions to be in closer agreement with conditions at the 660 km discontinuity (Figs. 2A,B). The post-spinel phase boundary of Mg2SiO4 determined by Fei et al. (2004a) using the MgO pressure scale of Speziale et al. (2001) is consistent with the depth of the 660 km discontinuity, i.e., ∆P660 = (P660 km – Pmeasured at 1850 K) = 0.6 GPa. Litasov et al. (2005b) reported ∆P660 = 1.0 GPa using the Au scale of Tsuchiya (2003) for anhydrous pyrolite. Ignorance of the pressure effect on thermocouple Electromotive force (EMF) in multianvil experiments underestimates temperatures (e.g., Ohtani 1979), and it further reduces discrepancy. The results obtained by Chudinovskikh and Boehler (2001) for Mg2SiO4 using the DAC are consistent with the data of Fei et al. (2004a). The data determined by Shim et al. (2001) using the DAC indicate that the post-spinel transformation boundary (Figs. 2A,B) is close to the depth of the 660 km seismic discontinuity. The results obtained by DAC, however, have a large temperature (100-200 K) uncertainty. DAC results are also relatively scattered, causing a high uncertainty in the estimation of the Clapeyron slope of the phase boundary. It should be noted that the most remarkable results of the recent in situ X-ray diffraction experiments are the gentle slope of the phase boundary, i.e., previous results of the Clapeyron slope, dP/dT , were relatively large around 2 MPa/K (Ito and Takahashi 1989; Irifune et al.

Eﬀect of Water on Mantle Phase Transitions

403

1998), whereas Katsura et al. (2003), Fei et al. (2004a) and Litasov et al. (2005b) indicate that the slope is small and approximately 0.4~1 MPa/K. The improvement for determination of the equilibrium boundary especially at temperatures below 1500 K has been made in the recent in situ study, i.e., there is a possibility that the old data overestimated the pressure of the phase transformation due to a slow transformation kinetics at low temperature (e.g., Kubo et al. 2002a). Effect of slow transformation kinetics was overcome by observation of forward and reversal reactions by changing pressure at a constant high temperature (e.g., Katsura et al. 2003; Litasov et al. 2005b). A small Clapeyron slope of the post-spinel transformation has very important implications for the 660 km discontinuity as discussed below in detail; the dry Clapeyron slope cannot explain the topography of the 660 km discontinuity of about ±20 km (e.g., Gu et al. 1998). Litasov et al. (2006) determined the phase boundary by in situ X-ray diffraction study in the pressure range of 20-26 GPa and temperatures up to 2300 K. The post-spinel phase boundary defined by a boundary of the appearance of Mg-perovskite can be expressed as P (GPa) = −0.0005 T (K) + 23.54 using the Au pressure scale by Tsuchiya (2003) and P (GPa) = −0.0008 T (K) + 24.42 using the MgO pressure scale by Speziale et al. (2001) (see Figs. 2 and 3). Ito and Takahashi (1989) reported that the pressure interval of coexisting ringwoodite and Mg-perovskite + magnesiowustite to be less than 0.1 GPa for (Mg0.9Fe0.1)2SiO4 based on quench experiments. On the other hand, Hirose (2002) and Nishiyama et al. (2004) reported a relatively wide pressure interval (0.5-0.7 GPa) for the post-spinel transformation in pyrolite. Litasov et al. (2005b) reported a narrow pressure interval of coexistence of 0.1-0.5 GPa based on their in situ X-ray diffraction study on anhydrous pyrolite. In a water bearing system, a shift of the post-spinel phase boundary to higher pressure may be expected due to the difference in the water solubility between ringwoodite and Mg-perovskite + magnesiowustite assemblages. Ringwoodite can contain H2O up to 2.6 wt% at 1100 °C and 20 GPa (Kohlstedt et al. 1996; Inoue et al. 1998), whereas Mgperovskite and magnesiowustite contain very limited amounts of H2O (<30 and <100 ppm, respectively)(Bolfan-Casanova et al. 2000; 2002; 2003; Litasov et al. 2003). Higo et al. (2001) studied the influence of water on the phase boundary of the Mg2SiO4 post-spinel phase transformation and reported a shift of 0.2 GPa to higher pressure at 1873 K. However, the effect of water has not been studied in detail. Recently, Litasov et al. 2400

Gt+St+Cpv

We

t)

2000

in (

1600

Dry )

Pv

in (

1800

Pv

Temperature, K

2200

1400

Hirose and Fei, 2002

Mpv+Cpv+St +NAL(CF)

1200

20

22

24 26 28 Pressure, GPa

30

32

Figure 3. Comparison of garnet-perovskite phase boundaries in basalt (MORB) under dry and wet conditions. The dashed line (Litasov et al. 2004) and solid line (Sano et al. 2006) show the first appearance of perovskite determined by in situ X-ray diffraction experiments using pressures calculated based on the Au pressure scale of Tsuchiya (2003). The phase boundary determined using quench experiments by Hirose and Fei (2002) is shown as a dotted line. Gt, garnet; St, stishovite; Cpv, Ca-perovskite; Mpv, Mg-perovskite; NAL, Na-Al hexagonal phase; CF, aluminous orthorhombic phase with Ca-ferrite structure.

404

Ohtani & Litasov

(2005a) made a detailed determination of the post-spinel phase boundary with a hydrous pyrolite composition with 2.0 wt% H2O using in situ X-ray diffraction at high pressure and temperature. The pressure of the post-spinel transformation boundary in a hydrous pyrolite composition was found to be higher by about 0.6 GPa than that of anhydrous pyrolite at 1473 K, whereas there is no obvious difference in pressure of this boundary at higher temperatures (1773-1873 K). The phase boundary defined by the appearance of Mg-perovskite can be expressed as P (GPa) = −0.002 T (K) + 26.3 in the temperature range from 1000 to 1800 K for pyrolite containing 2.0 wt% H2O. Superhydrous phase B coexists both with ringwoodite and Mg-perovskite in hydrous runs made below 1500 K. It should be noted that the Clapeyron slope of the phase boundary is −2 MPa/K in pyrolite containing 2.0 wt% H2O which is greater than that observed under dry conditions. A large Clapeyron slope is consistent with the temperature dependency of the H2O solubility in ringwoodite, i.e., its H2O content is about 2.0 wt% at around 1200 K, whereas it decreases to about 0.5 wt% at 1700 K (Ohtani et al. 2001). Since the shift of the phase boundary can be related to the water content of ringwoodite, a large amount of the pressure shift can be expected at lower temperature and a smaller shift at relatively higher temperature, resulting in a large Clapeyron slope of the phase boundary. Based on the results of in situ X-ray diffraction experiments of the post-spinel transformation in anhydrous and hydrous pyrolite we may conclude that both pyrolite and Mg2SiO4 show smaller negative Clapeyron slopes for the post-spinel phase boundary between −0.4 and −1.0 MPa/K using both Au and MgO pressure scales (Speziale et al. 2001; Tsuchiya 2003), whereas the addition of 2.0 wt% H2O may shift the boundary to higher pressures by ~0.6 GPa (15 km) at 1473 K and produce a larger Clapeyron slope of about –2 MPa/K. The change of the Clapeyron slope has important implications for slab dynamics since it affects the magnitude of the buoyancy force operating in slabs. The topography of the 660 km discontinuity may be at least partly due to the water content in the slabs.

Post-garnet transformation in basalt (MORB) The post-garnet transformation has a positive Clapeyron slope, and it occurs at higher pressure than the decomposition of ringwoodite in peridotite compositions under dry condition [e.g., Kubo and Akaogi 2000]. Based on quench experiments, Hirose and Fei (2002) demonstrated that the post-garnet phase transformation in basalt occurs at 27 GPa and 1900 K with a slightly positive Clapeyron slope (dP/dT = +0.8 MPa/K) as shown in Figure 3. Using in situ X-ray diffraction, Oguri et al. (2000) determined the post-garnet transformation boundary in natural pyrope at 25 GPa and 1900 K with a significant positive dP/dT (+6.4 MPa). On the other hand, Litasov et al. (2004) made an in situ X-ray diffraction study of the phase transformation in basalt under dry conditions, and showed that their data are inconsistent with the previous quench experiments by Hirose and Fei (2002), possibly due to corrections for the thermal pressure resulting in a large positive dP/dT of the post-garnet transformation in basalt (+4.1 MPa/K). The experiments by Litasov et al. (2004) revealed that the post-garnet phase boundary is expressed as P (GPa) = 0.0046 T (K) + 18.40 using the Au pressure scale by Tsuchiya (2003) as shown in Figure 4. The pressure interval of coexistence of garnet and Mg-perovskite is very narrow and less than 0.5 GPa. Using quench experiments, Litasov and Ohtani (2005) observed a shift in the post-garnet transformation boundary in basalt by ~1 GPa to lower pressure with the addition of 2 wt% H2O. Sano et al. (2006) confirmed an approximate 2 GPa shift of the phase boundary under wet conditions (~10 wt% H2O) using in situ X-ray diffraction. They observed a CaMg-perovskite bearing assemblage at temperatures below 1400 K. At 1400-1500 K they observed the transformation of Mg-perovskite to garnet and the reversal transformation at about 22.9-23.1 GPa (Fig. 5). The resulting equation for the post garnet transformation in hydrous basalt can be expressed as P (GPa) = 0.0049 T (K) + 15.94 using the Au pressure scale by Tsuchiya (2003).

Eﬀect of Water on Mantle Phase Transitions

Widths of the wadsleyite rim, µm

1.0

13.5 GPa and 1303 K

C

NaCl+Mg(OH)3 mixture is placed around the olivine crystal NaCl : Mg(OH) 2 = 10:1 (wt.)

400

300

B

NaCl:Mg(OH)2= 500:1

0.8

200

0.6 100

Pure NaCl

0.4

A

Volume fraction of wadsleyite

500

405

0.2 0 0

200

400

600

800

1000

1200

Time, min. Figure 4. The time dependence of the width and volume fraction of a wadsleyite rim growing on a single crystal of San Carlos olivine under dry and wet conditions at 13.5 GPa and 1303 K. Details are given in Kubo et al. (1998). Water was added as brucite Mg(OH)2 and its content was controlled by the ratio of NaCl and brucite surrounding the single crystal sample. (A) dry condition; (B) 500:1 mixture of NaCl and brucite; (C) 10:1 mixture (weight) of NaCl and brucite. The transformation is clearly enhanced by water.

1300

Temperature, K

1500

1200

1400

1100

Metastable ilmenite

1300

1000

Ringwoodite 1200

900

1100 Metastable ilmenite and stishovite

1000

Temperature,

Periclase + perovskite

oC

1600

800 700

22

24

26

28

30

Pressure, GPa Figure 5. Pressure-temperature diagram showing the conditions where metastable assemblages were observed in experiments on the post-spinel transformation (see Kubo et al. 2002a). The dotted line is the equilibrium boundary between Mg2SiO4 ringwoodite and MgSiO3 perovskite + MgO periclase. The conditions where metastable assemblages appear are also shown. Solid circles, ilmenite and stishovite do not appear but stable assemblage of perovskite and periclase appears; open square, ilmenite + periclase assemblage appears; solid squares, both ilmenite + periclase and stishovite + periclase assemblages appear.

406

Ohtani & Litasov

The shift of the post garnet phase boundary to lower pressure cannot be explained by the effect of water solubility in the coexisting phases. Majorite garnet contains water up to 1130-1250 ppm H2O, (Katayama et al. 2003) which is larger than Mg-perovskite (<30 ppm; Litasov et al. 2003; Bolfan-Casanova et al. 2003). Thus a shift of the phase boundary to higher pressure might be expected. Stishovite may contain a significant amount of water (up to 1500 ppm as reported by Chung and Kagi 2002), but it exists in both the garnet- and perovskite-bearing assemblage. The effects of the other phases such as Al-rich NAL phase and Ca-perovskite are not clear. Litasov et al. (2005a) suggested a change in the oxidation state in the hydrous fluid phase under wet conditions causes the shift of the boundary due to a change in ferric iron content in Mg-perovskite. Since aluminous Mg-perovskite can contain a significantly higher amount of Fe3+ (McCammon 1997), the difference in the oxidation state causes a different ferric iron content in Mg-perovskite. We observed a difference in the iron content in Mg-perovskite under the dry and wet conditions (Litasov et al. 2005a; Sano et al. 2006) perhaps due to such a change of ferric iron content in Mg-perovskite. Another explanation is that the fluid can dissolve the major elements such as Mg, Si and Fe (e.g., Kawamoto et al. 2004), and thus it can change the composition of minerals, resulting in the shift of the location of the phase boundary (Sano et al. 2006).

EFFECT OF WATER ON PHASE TRANSFORMATION KINETICS Olivine-wadsleyite phase transformation kinetics Water plays an important role in controlling rheological properties of the mantle, since a trace amount of water lowers the strength of olivine crystals, i.e., hydrolytic weakening (e.g., Karato 1989). Olivine-wadsleyite transformation kinetics has been studied by various authors under dry conditions (e.g., Mosenfelder et al. 2001; Kerschhofer et al. 2000). However, the effect of water on the transformation kinetics has not been studied so much, although water may affect the phase transformation kinetics of mantle minerals. Recently, Kubo et al. (1998) and Hosoya et al. (2005) have conducted experiments to study the effect of water on the kinetics of the olivine-wadsleyite transformation based on both quench experiments using olivine single crystals and in situ X-ray diffraction experiments of polycrystalline olivine Kubo et al. (1998) showed an enhancement of the transformation rate by water based on quench experiments using single crystals of San Carlos olivine that transformed to wadsleyite. The growth rate is roughly estimated from the thickness of the wadsleyite rim on the olivine single crystals. Figure 4 shows the time dependence of the width and volume fraction of wadsleyite rim under the dry and wet conditions. In dry runs, wadsleyite growth was retarded with time and eventually ceased after several hundred minutes as shown in Figure 4. Whereas, in wet runs, the wadsleyite grew more rapidly compared to the dry runs, resulting in a difference in growth rate between dry and wet conditions. These experiments clearly indicate that the growth rate of the wadsleyite rim was enhanced by the presence of water. The volume change associated with the transformation is often large in solid-state first order transformations, resulting in the development of localized stress (e.g., Rubie and Thompson 1985). In dry runs, a localized pressure drop in the relict olivine as a result of the volume change during the transformation possibly causes a decrease in the free energy change of the reaction, resulting in a decrease in the growth rate (Liu et al. 1998). Plastic flow of the outer rim can relax the localized pressure drop and thus controls the growth rate of the wadsleyite rim, whereas a localized pressure drop in olivine may be relaxed perhaps due to water weakening in wet runs. Thus, wadsleyite grows more rapidly compared to the dry runs, resulting in the difference in growth rate between dry and wet conditions. Figure 4 clearly implies that a small amount of water, about 0.12–0.5 wt%, enhances the olivine-wadsleyite phase transformation kinetics corresponding roughly to a temperature increase of about 150 degrees.

Eﬀect of Water on Mantle Phase Transitions

407

Kubo et al. (2004) and Hosoya et al, (2005) studied the effect of water on the phase transformation kinetics by in situ X-ray diffraction using the multianvil press, SPEED1500 at beam line BL04B1 of Spring 8 of Japan Synchrotron Research Institute. They obtained growth rate data at 13.4-15.8 GPa, 730-1100 °C, for 660-5000 ppm weight H2O to determine the pressure, temperature and water content dependences on the growth kinetics of the α-β transformation of Mg2SiO4. The growth in the polymorphic transformation is controlled by interface kinetics (Turnbull 1956). The growth rate as a function of pressure, temperature and water content can be described as follows, where A is a pre-exponential factor, T is absolute temperature, COH is water content in ppm by weight, n is the water content exponent, ∆Ha is the activation enthalpy for growth, P is pressure, V* is the activation volume for growth, R is the gas constant, and ∆Gr is the free energy change of the transformation. This rate equation was fitted to the growth rate data by a weighted least-squares procedure, which yields ln A = −18.0 ± 3.8 ms−1 wt. ppm H2O−3.2, n = 3.2 ± 0.6, ∆Ha = 274 ± 87 kJ/mol, and V* = 3.3 ±3.8 cm3/mol. Results of the fitting of the pressure, temperature, and water content against the growth rate are shown in Figure 6. This study demonstrates that water greatly enhances the growth rate at the olivine transformation. Because the depth of the olivine transformation in cold slabs is controlled by the growth kinetics (Rubie and Ross 1994), the effect of water on the growth kinetics must be considered when estimating fields of olivine metastability.

-22

A

ln{growth rate/T[1-exp(-∆Gr /RT)]}

ln{growth rate/T[1-exp(-∆Gr /RT)]}

-22 -24 P=15 GPa COH= 1000 wt.ppm H2O

-26 -28

H=274 KJ/mol -30 -32 -34

B -24

-28 -30

0.8

0.9

1.0

1000/T(K)

V *= 3.3 cm3/mol

-32 -34

0.7

T=900oC COH=1000 wt.% ppm H2O

-26

13

14 15 Pressure, GPa

16

ln{growth rate/T[1-exp(-∆Gr /RT)]}

-22

C -24

P= = 15 GPa T= = 900oC

Figure 6. The effect of water on the transformation of olivine to wadsleyite studied by in situ X-ray diffraction. Details are given in Hosoya et al. (2005). (A), (B) show the temperature and pressure dependences of the growth rate respectively; (C) the effect of the olivine water content on olivine to wadsleyite transformation. Water enhances the transformation significantly.

-26 -28

n=3.2

-30 -32 -34

102

103 Water content (wt.% ppm H2O)

104

408

Ohtani & Litasov

Post-spinel and post garnet phase transformation kinetics Kinetics of the post-spinel and post-garnet transformations have been studied by Kubo et al. (2002a,b) under dry conditions. Kubo et al. (2002a) showed that a metastable phase assemblage is observed at low temperatures in the initial stage of the post-spinel transformation. They observed decomposition of ringwoodite to periclase + stishovite or periclase + akimotoite (MgSiO3). These assemblages formed in the initial stage of the phase transformation but disappeared after a few minutes, which indicates that these assemblages are metastable. The pressure and temperature conditions of appearance of the metastable assemblage are shown in Figure 5. These observations suggest that a metastable assemblage such as periclase + stishovite or periclase + akimotoite (MgSiO3) might exist in dry and cold subducting slabs. Kubo et al. (2002b) clarified that post-garnet transformation kinetics are very sluggish, and we can expect a metastable garnetite layer in the lower mantle due to low temperatures of the penetrating slabs. Figure 7 shows clearly that the post-garnet transformation is remarkably sluggish compared to the olivine-wadsleyite and post-spinel transformations. The appearance of a metastable assemblage in the post-spinel transformation might cause a relatively fast reaction rate of the transformation. The effect of water on the transformation kinetics in the post-spinel and post-garnet transformations has not yet been studied. However, it might be evaluated qualitatively based on the in situ X-ray diffraction study at high pressure and high temperature (Litasov et al. 2004, 2005a,b; Sano et al. 2006). Litasov et al. (2005b) observed the phase transformation from ringwoodite to magnesiowustite and Mg-perovskite proceeds at 1623 K within 30 minutes. On the other hand, Litasov et al. (2005a) observed that a similar reaction rate in peridotite plus 2.0 wt% water at 1423 K. These results suggest that water (about 2.0 wt%) can enhance reaction rates to a degree that may be equivalent to a temperature increase of about 200 K. In situ X-ray diffraction studies of the post garnet transformation of basalt under dry and wet conditions indicate that the transformation at 1473 K proceeds within 30 minutes under

Transformed volume fraction

1 Post-spinel transformation 28.2 GPa,, 1283 K,, P= 5.4GPa . (Mg ( 2SiO4) d = 10 m 31.4 GPa, 1473 K,, P= 6.0 GPa d = 3.4 m (Pure pyrope)

0.8 0.6

30.8 GPa, 1273 K, P= 5.5 GPa

0.4

28.1 GPa 1600 K P = 3.5 GPa d = 12 m (Natural pyrope)

0.2

Post-garnet transformation

0

0

1000

2000 3000 Time, seconds

4000

Figure 7. The rate of the post-spinel and post-garnet transformations after Kubo et al. (2002a,b). The solid curve indicates the rate of the post-spinel transformation of Mg2SiO4 at 28.2 GPa and 1283 K. The thin dotted curves are the rate of the post-garnet transformation in Mg3Al2Si3O12 at 31.4 GPa and 1473 K and 30.8 GPa and 1273 K. A thick dotted curve is the rate of the post-garnet transformation in natural pyropic garnet (Mg0.724Fe0.184Ca0.111)3(Al0.872Cr0.0044Ti0.010)2Si3.064O12. The rate of the post-spinel transformation is significantly faster than that of the post-garnet transformation. ∆P, overpressure (a pressure interval from the equilibrium boundary): d, grain size of the sample.

Eﬀect of Water on Mantle Phase Transitions

409

the wet conditions (Sano et al. 2006), whereas the temperature has to be raised to 1973 K for the same reaction time scale under dry conditions (Litasov et al. 2004). Water enhances the post-garnet transformation kinetics similar to the post-spinel transformation kinetics, although the reaction rate in the post-garnet transformation is sluggish compared to the post-spinel phase transformation. Although our preliminary studies suggest that both post-spinel and postgarnet transformation kinetics are enhanced by water, we need more quantitative studies by in situ X-ray diffraction on the reaction under wet conditions to quantify the effect of water.

IMPLICATION FOR SEISMIC DISCONTINUITIES AND PHASE TRANSFORMATION BOUNDARIES UNDER DRY AND WET CONDITIONS 410 km seismic discontinuity and olivine-wadsleyite phase boundary The 410 km seismic discontinuity is considered to be caused by the transformation from olivine to wadsleyite in peridotite mantle. The average depth of the discontinuity is 411 km (Gu et al. 1998) and 418 km (Flanagan and Shearer 1998, 1999) using stacking SS and PP precursors. Figure 8 summarizes the shifts of the phase boundaries in the mantle as a result of the addition of water. Katsura et al. (2004) estimated the mean temperature at the 410 km seismic discontinuity to be 1760 ± 45 K for pyrolite mantle using the average depth of

Depth, km 200

300

400

500

600

700

800

2200 Wet

2000

ol Wd

tle man e g ra Ave herm t o ge

Temperature, K

1800

1600

1400

Dry

Rw Mpv +Pe

Gt Mpv

Gt Mpv Rw+ Gt+ Cpx

Wd+Gt

Mpv+Pe +Cpv

B

Ol+Gt+Cpx

slab cting u d b Su erm h geot

Wet

1200

1000

Mpv+Pe+Cpv +Sb

Rw+Gt+ Cpv+Sb

Dry

A Wet Dry

800

6

10

14

18

22

26

30

Pressure, GPa Figure 8. Phase boundaries in peridotite and basalt under dry and wet conditions. The stability fields of wasleyite and ringwoodite expand under wet conditions at low temperatures (Litasov et al. 2005a,b), whereas the post garnet transformation shifts to lower pressures under wet conditions (Litasov et al. 2004; Sano et al. 2006). The grey dotted curve is the average temperature of the mantle (Akaogi et al. 1989). The shaded area represents the temperature range of the slabs (Kirby et al. 1996). The phase boundaries of the post-spinel and post-garnet transformations cross at around 1000 K (A) under dry conditions and 1400 K (B) under wet conditions. Rw, ringwoodite; Sb, superhydrous phase B; Pe, periclase. The other abbreviations are the same as those given in Figures 1 and 3.

410

Ohtani & Litasov

the discontinuity. As discussed in the previous section, the existence of water in the mantle (~1760 K) may elevate the depth of the 410 km discontinuity by a few kilometers. Global topography of the 410 km seismic discontinuity indicates depth variations of approximately ±20 km (Flanagan and Shearer 1998). Local seismological studies indicate greater depth variations in some areas. For example, the maximum elevation of the discontinuity in subduction zones is around 60-70 km, which is consistent with a temperature anomaly of about 1000 K if we apply the Clapeyron slope of the olivine-wadsleyite phase boundary of +3~4 MPa/K (e.g., Collier et al. 2001). Such a temperature anomaly of 1000 K at the 410 km discontinuity is too large given that the normal temperature of the discontinuity is around 1760 K based on the geotherm estimated by some authors (e.g., Akaogi et al. 1989; Katsura et al. 2004). Since water and temperature have a similar effect on the phase boundary, the topography of the 410 km seismic discontinuity in some regions may be also explained by the presence of water in subduction zones, i.e., some fraction of the topography of the discontinuity may be explained by the effect of water in combination with the temperature effect as shown in Figure 8. The width of the 410 km discontinuity varies from 4 to 35 km (Shearer 2000). The phase relations in the olivine composition in peridotite indicate a width of the binary loop of coexistence of olivine and wasleyite is 25 km at 1473 K and 14 km at 1873 K (Katsura and Ito 1989). Frost (2003) reviewed recent experimental and thermodynamic data on the olivine-wadsleyite transformation and discussed the possibility of a 4-6 km width, i.e., the minimum of the range of the width of discontinuity. Frost (2003) found this width to be consistent with the olivinewadsleyite phase transformation in a peridotite composition containing garnet and pyroxenes, which makes the phase boundary sharper due to partitioning of elements between minerals. There are several factors that make the phase boundary broader. First, we can expect a broader transformation in colder, garnet-poor or FeO-rich regions in the mantle based on the equilibrium phase relations. Second, the reaction kinetics at low temperature also tends to broaden the width of the discontinuity since olivine and wadsleyite can coexist metastably under the conditions of a cold subducting slab (e.g., Rubie and Ross 1994; Kubo et al. 1998). Third, the sharpness of the discontinuity may be affected by the presence of water. Wood (1995) used a thermodynamic calculation to estimate that the presence of a small amount of water, i.e., about 100 ppm, can broaden the stability field of coexistence of olivine and wadsleyite by about 3 km. He argued that the sharpness of the 410 km discontinuity indicates that the transition zone is essentially dry. On the other hand, Smyth and Frost (2002) suggested that the hydrogen diffusion and gravitational stratification narrow the phase transformation interval between olivine and wadsleyite, which is an opposite effect from the argument by Wood (1995). The preliminary experiments on determination of the olivine-wadsleyite transformation in peridotite-3.0 wt% water system made by Litasov et al. (2006) suggested that coexistence loop of olivine and wadsleyite expands under the wet conditions, which is consistent with the estimation by Wood (1995). Nolet and Zielhuis (1994), Revenaguh and Sipkin, (1994), and Song et al. (2004) suggested water as a possible cause of anomalies in the deep upper mantle and the transition zone. Recent observations that the width of the 410 km seismic discontinuity beneath the Mediterranean is between 20 and 35 km (Van der Meijde et al. 2003) may be explained by H2O contents in this region. The cold and wet nature of this area may be supported by the high electrical conductivity of the upper mantle in this region, which is consistent with 1000-1500 ppm H2O in olivine (Tarits et al. 2004).

The 660 km seismic discontinuity and the post-spinel transformation: average depth and topography of the 660 km seismic discontinuity The 660 km seismic discontinuity has been considered to be caused by the post-spinel transformation in peridotite (e.g., Ito and Takahashi 1989). Seismological studies show that the 660 km discontinuity is observed globally and the average depth of the discontinuity

Eﬀect of Water on Mantle Phase Transitions

411

is 654 km (Gu and Dziewonski 2002). If we adopt this depth as a global average for the discontinuity, the discrepancy between the pressure of the post-spinel transformation (i.e., decomposition of ringwoodite) determined by experiments and that in seismic studies becomes small. The phase boundary determined experimentally is lower by 0.2-0.3 GPa (i.e., 6-10 km shallower) compared with the pressure of the 660 km discontinuity assuming the temperature of the discontinuity around 1800-2000 K, when we apply the pressure scales of Au and MgO (Tsuchiya 2003; Speziale et al. 2001), which is believed to be the most reliable at present (Figs. 2A,B and Fig. 8). The above small discrepancy may be accounted for by the effect of pressure on the thermocouple EMF, i.e., the temperatures of the high pressure in situ X-ray diffraction study may be underestimated, resulting in a shift of the phase boundary to higher pressure (e.g., Ohtani 1979). The alternative explanation for the discrepancy may be the effect of water on the phase boundary. We observed a change of the Clapeyron slope of the post-spinel phase boundary under hydrous conditions, but found no significant shift of the boundary at higher temperatures around 1773-1873 K as shown in the previous section. On the other hand the boundary appears to shift to higher pressure at lower temperatures; i.e., Higo et al. (2001) and Inoue et al. (2001) reported 0.2 GPa shift of the phase boundary at 1873 K, and Katsura et al. (2003) also suggested 0.6 GPa shift of the post-spinel transition boundary to the higher pressure in hydrous Mg2SiO4 based on their preliminary results at 1663 K. Our data indicate that the post-spinel transformation boundary in hydrous pyrolite shifts to higher pressure by about 0.6 GPa relative to anhydrous pyrolite at 1473 K, whereas there is no obvious shift of this boundary at higher temperatures (1773-1873 K). The Clapeyron slope of the post-spinel transformation determined recently by in situ X-ray diffraction experiments is very gentle (about −0.5 MPa/K) under dry conditions (e.g., Katsura et al. 2003; Fei et al. 2004a; Litasov et al. 2005b). The gentle slope of the postspinel transformation boundary may require a large temperature difference to account for the topography of the 660 km seismic discontinuity (Figs. 2A,B, Fig. 8). The depth of the 660 km seismic discontinuity varies by about ±20 km; about +20 km beneath Northern Pacific Ocean, Atlantic Ocean, and South Africa, whereas it is −20 km beneath subduction zones such as the western Pacific and South America (Flanagan and Shearer 1998; Gu and Dziewonski 2002). Lebedev et al. (2003) proposed smaller variations in the range of ±15 km. The depressions caused by the slabs have been studied by many authors; depression of 20~30 km beneath Tonga (Niu and Kawakatsu 1995) and up to about 50 km beneath Izu-Bonin (e.g., Collier et al. 2001). Elevation of the 660 km discontinuity by 10-20 km is also reported in areas related to hot plumes such as Hawaii (Li et al. 2000) and the South Pacific (Niu et al. 2000). If the topography of the discontinuity is caused only by the effect of temperature on the phase boundary, an elevation of 20 km corresponds to a temperature elevation of about 1300 K and a depression of 50 km corresponds to a temperature decrease of about 3000 K, which implies unusually if not impossibly large temperature variations. This indicates that the variation in depth of the 660 km discontinuity cannot be explained only by the temperature effect, but we need to introduce the other effects to explain the topography. A delay of the phase transformation due to kinetics, or the influence of minor components or volatiles may be additional explanations for the large variations in 660 topography. Kinetics of the post spinel transformation in Mg2SiO4 (Kubo et al. 2002a) indicates that the transformation is very fast compared to the speed of the slab subduction and it completes in 104 years with 1 GPa of overpressure (i.e., a pressure interval from the equilibrium phase boundary) even at the temperatures of 1000 K (Fig. 7). Only for very cold slabs with temperatures below 1000 K, the post-spinel transformation reaction delays by 106-108 years, a meaningful duration to explain a large depression of the 660 km discontinuity. We can expect a delay of transformation and shift of the phase boundary by about 1 GPa (corresponding to 20-25 km) due to kinetics only in unusually cold slabs.

412

Ohtani & Litasov

The most plausible explanation for such a large elevation and depression of the discontinuity may be the presence of water, although the elevation may be smaller due to lower water solubility in ringwoodite in hot environments. The observed shift of the postspinel transformation boundary is relevant to the topography of the 660 discontinuity detected by the seismological studies (Fig. 8). The displacement of the post-spinel phase boundary by 0.5 GPa under the hydrous conditions corresponds to about 15 km in depth, which is a half of the observed depressions of the 660 km discontinuity, 30-40 km, in the hot subduction zones at a temperature about 1473 K. Thus, a large displacement of the 660 km discontinuity may be considered as evidence for existence of water in the transition zone (Litasov et al. 2005a,b).

The density relation of basalt and peridotite near the 660 km discontinuity Ringwood (1994) drew attention to the importance of the density contrast between the basaltic and peridotite layers of a subducting slab. He argued that there is a density crossover between peridotite and basalt due to the pressure difference between the post-spinel transition in the peridotite layer and the post-garnet transition in the basaltic layer of the slabs. Figure 8 summarizes the phase relations of peridotite and basalt under dry and wet conditions and shows the range of temperatures expected for subducting slabs and an average mantle geotherm. Figure 9 shows the density of the peridotite layer and the basalt layer of the slab. Irifune and Ringwood (1993) suggested that the density crossover occurs in the pressure interval of about 2 GPa along the slab geotherm. The density crossover may lead to a separation of the basaltic crust from the peridotite body at a depth of about 660 km resulting in the formation of a garnetite-bearing layer at the base of the transition zone (Ringwood 1994). On the other hand, Hirose et al. (1999) showed that the transformation of the basaltic crust from garnetite to a perovskite lithology occurs near 720 km and they argued that the density crossover between

Depth, km 600

640

680

720

760

800

840

Density, g/cm3

5.0

4.6

Wet MORB Wet peridotite Dry peridotite

4.2

Dry MORB (equilibrium)

Dry MORB (metastable; d== 1mm) d

Density crossover 3.8

20

22

24 26 28 Pressure, GPa

30

32

Figure 9. Density profiles of dry and wet basalt (MORB) and pyrolite demonstrating the density crossover near 660 km. Density calculations were carried out along a normal mantle geotherm for anhydrous systems (Akaogi et al. 1989) and a cold subduction geotherm for hydrous systems (Kirby et al. 1996) using a third order Birch-Murnaghan equation of state and the set of thermoelastic parameters given by Litasov and Ohtani (2005). A density crossover exists between the peridotite (thin dotted curve) and basalt (thin solid curve) in the range from 23 GPa to 27 GPa under dry conditions. The sluggish transformation rate of the postgarnet transformation expands the region of the density crossover under dry conditions. There is no density crossover between the peridotite (thick dotted curve) and basalt (thick solid curve) under wet conditions.

Eﬀect of Water on Mantle Phase Transitions

413

basalt and peridotite layers may be too narrow for separation of the basaltic crust to accumulate at the base of the transition zone. Therefore, the basaltic crust may gravitationally sink into the lower mantle. However, the equilibrium phase relations may not be applicable to the phase transformation in real subducting slabs, because reaction kinetics may be an important factor in controlling the mineralogy and density of subducting slabs. Kubo et al. (2002b) studied the kinetics of the post-garnet transformation in basalt, and they showed a very slow reaction rate allowing metastable garnet to survive for a long time (the order of 10 Ma) after crossing the 660 km seismic discontinuity. Therefore, a wider pressure interval of the density crossover can be expected by taking into account the kinetics of the post-garnet transformation under dry conditions, suggesting a separation of the basaltic layer of the slabs to form a garnetite layer at the base of the transition zone (Fig. 9). Litasov et al. (2004) and Sano et al. (2006) demonstrated that at the low temperature of slabs the density crossover between the peridotite and basalt layers might be absent (Figs. 8 and 9) especially in water-rich subduction zones. The crossover of the post-spinel and postgarnet transformation boundaries locates below the temperature path of a cold subducting slab in anhydrous subduction (point A in Fig. 8), whereas it lies above the temperature path of a hot slab at around 24 GPa and 1500 K (point B in Fig. 8) in hydrous subduction. Therefore, there may be no density crossover between the basalt and peridotite layers of hydrated slabs following cold subduction geotherms. If slabs pass through the 660 km seismic discontinuity, penetration of the basaltic crust component into the lower mantle can occur without gravitational separation from the peridotite body of the slab, at least for hydrous subduction environments.

Seismic reflectors: the possible existence of fluid in the lower mantle Kaneshima and Helfrich (1998) and Niu et al. (2003) reported seismic reflectors in the lower mantle. Niu et al. (2003) indicate that the physical properties of the reflector observed in the upper part of the lower mantle beneath the Mariana subduction zone show a decrease in shear wave velocity by 2-6% and an increase in density by 2-9% within the reflectors, whereas, no difference exists in P-wave velocity (<1%) between the reflector and the surrounding mantle. The thickness of the reflector is estimated to be around 12 km. The origin of the seismic reflectors is one of the most interesting issues in understanding heterogeneity in the lower mantle. The seismic reflectors may correspond to subducted oceanic crust as was suggested by some authors (e.g., Kaneshima and Helffrich 1999). In order to test whether subducted oceanic crust can cause these reflections, we need to test whether oceanic crust in the lower mantle can explain the observed seismological properties of the reflectors. There are three possible scenarios that may influence the seismic properties of subducted crust in the lower mantle. First, the subducted crust may possess an equilibrium lower mantle lithology composed of Mg-perovskite, Ca-perovskite, stishovite, and calcium ferrite type aluminous phase (CF) (or Na-aluminous phase, NAL). Second, it may possess a metastable majorite garnet lithology due to the sluggish transformation kinetics of the garnet-perovskite transformation (Kubo et al. 2002b). The third possibility is that the physical properties of the reflectors cannot be accounted for by the above two lithologies, and an additional factor is needed to explain the properties of the reflectors. The factor may be the presence of fluid or magma at these depths. In order to find a plausible explanation for the properties of the reflectors, we need to estimate physical properties of the oceanic crust under lower mantle conditions. Ohtani (2006) estimated the physical properties of high-pressure minerals in the oceanic crust and mantle. Figure 10 shows the differences between density, vP, and vS of the basaltic crust in the slab and the surrounding mantle peridotite. The density and velocity (vP and vS) differences between the reflector and surrounding mantle observed seismologically (Niu et al. 2003) are also shown in this figure. The properties of MORB with a metastable garnet-bearing lithology are also shown in this figure. A decrease in vS and an increase in density are the characteristic

414

Ohtani & Litasov

A

15

St

10 5

Density jump, % -8

-6

Cpv Mpv Vp

CF

-4

0 -5 -10

Mj -15

2

4

6

B

Vs jump vs density jump Velocity jump,%

Velocity jump,%

Vp jump vs density jump

Density jump, % 8 -8

-6

St 15 10 5

0 -5

Mj

Mpv

CF

-4

Seismic reflector in lower mantle

Cpv

-10

2

4

6

8

Vs Seismic reflector in lower mantle

-15

Figure 10. Differences (%) in density and vP between basalt (MORB) in the slab and surrounding mantle peridotite (A) and those in density and vS between basalt and peridotite (B). The density and velocity (vP and vS) differences between the reflector and surrounding mantle observed seismologically (Niu et al. 2003) are also shown in these figures. Density, vP, and vS of the mantle minerals are calculated at conditions of 30 GPa and 1873 K by using the procedure and parameters given in Ohtani (2006). Open squares and circles represent the vP-density jump and vS-density jump, respectively. The shaded areas indicate that the density and velocity differences of the majorite bearing assemblage and perovskite bearing assemblage. The properties of the reflectors cannot be explained either by a stable perovskite bearing assemblage nor a metastable majorite bearing assemblage. Abbreviations in the Figure are the same as those given in Figures 3 and 8.

properties of the reflector which cannot be accounted for by either the perovskite lithology or metastable majorite lithology of basalt, i.e., by a complete or incomplete phase transformation from majorite to perovskite lithologies, although there is a large uncertainty in seismic velocity, especially vS in basalt. The detailed calculations and the error estimations of the vP and vS are given in Ohtani (2006). We need additional factors to reduce vS, although the density increase can be explained by the lower mantle lithology; i.e., the complete phase transformation from majorite lithology to perovskite lithology shows positive jumps in density, vP and vS relative to the surrounding mantle, whereas the metastable majorite lithology shows physical properties (ρ, vP, and vS) that are smaller than the surrounding mantle. It may be possible to explain the drastic decrease in vS as an effect of fluid or melt films in the subducted oceanic lithosphere in the lower mantle (e.g., Williams and Garnero 1996). Hydrous phase D dehydrates at pressures around 40-50 GPa (Shieh et al. 1998). Therefore, the physical properties of the reflectors might be explained by fluid in the oceanic crust supplied from dehydration of this phase in the underlying hydrous peridotite layer of the slab penetrating into the lower mantle. Although the existence of fluid films in the oceanic crust of the slabs is a plausible mechanism for the change of the physical properties of reflectors, the existence of minor metallic iron formed by the garnet-perovskite transformation in the oceanic crust (Miyajima et al. 1999; Frost et al. 2004) could also cause the reduction of shear wave velocity and increase the density of the reflectors. In addition, several other explanations might also account for the unusual properties of the reflectors. Theoretical calculations, for example, have revealed the existence of an elastic anomaly associated with the phase transformation from stishovite to the post-stishovite CaCl2 phase (e.g., Stixrude 1998). A similar anomaly could result from the transformation in Al2O3 bearing CaSiO3 perovskite from the orthorhombic to cubic phase (Kurashina et al. 2004). The elastic anomalies associated with theses phase transformations, however, are not yet confirmed experimentally.

Eﬀect of Water on Mantle Phase Transitions

415

CONCLUDING REMARKS In this chapter we have summarized recent advances in the study of the effects of water on major phase transformations in the Earth’s mantle with implications for the topography of seismic discontinuities and mantle dynamics based on the results of quenched multianvil and in situ X-ray diffraction experiments at high pressure and temperature. Differences in water solubility between (a) wadsleyite and olivine and (b) ringwoodite and Mg-perovskite and ferropericlase may cause displacement of the phase transition boundaries, which are believed to be responsible for the 410 and 660 km discontinuities, respectively. Experimental results show that water may increase the pressure interval of the binary olivine-wadsleyite transformation loop, i.e., water expands the stability field of wadsleyite to lower pressures. This interval is ~0.3 GPa (or ~6 km) wide at 1700 K in the anhydrous pyrolite, whereas it may be ~1.2 GPa at 1500 K (or 33 km) in pyrolite with 2-3 wt% of H2O. The broadening of the 410 km discontinuity observed in some regions of the mantle is consistent with enrichment by water in the mantle. A significant shift of the boundary of the wadsleyite to ringwoodite phase transformation to the higher pressure through the addition of water to the peridotite system may also be responsible for variations in the depth of the 520 km discontinuity. However, the cause of a significant displacement of the wadsleyite/ringwoodite phase boundary is not clear. In situ X-ray diffraction study of the post-spinel transformation in hydrous pyrolite indicates that the phase boundary is shifted to higher pressures by 0.6 GPa relative to anhydrous pyrolite at 1473 K, whereas it shows no obvious shift at higher temperature of around 1873 K. The displacement of the post-spinel phase transformation boundary in hydrous pyrolite would correspond to a depth variation of about 15 km and may account for approximately half of 30-40 km depression observed for the 660 km discontinuity in subduction zones at temperatures of around 1473 K. This effect should be much stronger at lower temperatures. Since the Clapeyron slope of the post-spinel transformation boundary in anhydrous pyrolite is very gentle (about −0.5 MPa/K), we cannot explain depressions of the 660 km discontinuity as a result of temperature variations alone. Thus, a large depression of the 660 km discontinuity might be considered as direct evidence for the existence of water in the transition zone. In situ X-ray diffraction studies of the post-garnet transformation in anhydrous and hydrous basalt show that the phase boundary shifts to lower pressures by ~2 GPa upon the addition of water. This observation demonstrates that at temperatures of subducting slabs the density crossover between peridotite and basalt near 660 km might be absent under hydrous conditions. The basaltic component of the slab may therefore penetrate into the lower mantle under hydrous conditions and not separate from the peridotite body at the 660 km discontinuity.

ACKNOWLEDGMENTS We thank A. Suzuki, T. Kubo, H. Terasaki, K. Funakoshi, T. Kondo for collaboration during the experiments at SPring-8. We appreciate S.D. Jacobsen and an anonymous reviewer for improving the manuscript. This work was supported by the grants in Aid for Scientific Researches from the Ministry of Education, Culture, Sports, Science and Technology, Japan (No. 14102009 and 16075202) to E. Ohtani. This work was conducted as a part of the 21th Century Center-of-Excellence program Advanced Science and Technology Center for the Dynamic Earth at Tohoku University.

REFERENCES Agee CB (1998) Phase transformations and seismic structure in the upper mantle and transition zone. Rev Mineral 37:165-203

416

Ohtani & Litasov

Akaogi M, Ito E, Navrotsky A (1989) Olivine-modified spinel-spinel transitions in the system Mg2SiO4Fe2SiO4: calorimetric measurements, thermochemical calculation, and geophysical application. J Geophys Res 94:15671-15685 Anderson OL, Issak DG, Yamamoto S (1989) Anharmonicity and the equation state for gold. J Appl Phys 6: 1534-1543 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res 108, doi: 10.1029/2001JB000679 Benz HM, Vidale JE (1993) Sharpness of upper-mantle discontinuities determined from high-frequency reflections. Nature 365:147-150 Bina CR, Helffrich GR (1994) Phase transition Clapeyron slopes and transition zone seismic discontinuity topography. J Geophys Res 99:15853-15860 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett 182:209-221 Bolfan-Casanova N, Mackwell S, Keppler H, McCammon C, Rubie DC (2002) Pressure dependence of H solubility in magnesiowüstite up to 25 GPa: Implications for the storage of water in the Earth’s lower mantle. Geophys Res Lett 29, doi: 10.1029/2001GL014457 Bolfan-Casanova N, Keppler H, Rubie DC (2003) Water partitioning at 660 km depth and evidence for very low water solubility in magnesium silicate perovskite. Geophys Res Lett 30, doi: 10.1029/2003GL017182 Brown JM (1999) High pressure equation of state for NaCl, KCL, and CsCl. J Appl Phys 86:5801-5808 Chen J, Inoue T, Weidner DJ, Wu Y (1998) Strength and water weakening of mantle minerals, α, β, and γ Mg2SiO4. Geophys Res Lett 25:575-578 Chen J, Inoue T, Yurimoto H, Weidner DJ (2002) Effect of water on olivine-wadsleyite phase boundary in the (Mg,Fe) SiO system. Geophys Res Lett 29:1875, doi:10.1029/2001 GL014429 2 Chudinovskikh L, 4Boehler R (2001) High-pressure polymorphs of olivine and the 660 km seismic discontinuity. Nature 411:574-577 Chung JI, Kagi H (2002) High concentration of water in stishovite in the MORB system. Geophys Res Lett 29, doi: 10.1029/2002GL015579 Collier JD, Helffrich GR, Wood BJ (2001) Seismic discontinuities and subduction zones. Phys Earth Plane Inter 127:35-49 Demouchy S, Deloule E, Frost DJ, Keppler H (2005) Pressure and temperature dependence of water solubility in Fe-free wadsleyite. Am Mineral 90:1084-1091 Dziewonski AM, Anderson DL (1981) Preliminary Reference Earth Model. Phys Earth Planet Int 25:297-356 Fei Y, Van Orman J, Li J, van Westrenen W, Sanloup C, Minarik W, Hirose K, Komabayashi K, Walter M, Funakoshi K (2004a) Experimentally determined postspinel transformation boundary in Mg2SiO4 using MgO as an internal pressure standard and its geophysical implications. J Geophys Res 109, doi: 10.1029/ 2003JB002562 Fei Y, Li J, Hirose K, Minarik W, Van Orman J, Sanloup C, van Westrenen W, Komabayashi T, Funakoshi K (2004b) A critical evaluation of pressure scales at high temperatures by in situ X-ray diffraction measurements. Phys Earth Planet Inter 143-144:515-526 Flanagan MP, Shearer PM (1998) Global mapping of topography on transition zone velocity discontinuities by stacking SS precursors. J Geophys Res 103:2673-2692 Flanagan MP, Shearer PM (1999) A map of topography on the 410 km discontinuity from PP precursors. Geophys Res Lett 26:549-552 Frost DJ (2003) The structure and sharpness of (Mg,Fe)2SiO4 phase transformations in the transition zone. Earth Planet Sci Lett 216:313-328 Frost DJ, Liebske C, Langenhorst F, McCammon C, Tronnes RG, Rubie DC (2004) Experimental evidence of iron-rich metal in the Earth’s lower mantle. Nature 428:409-411 Fukao Y, Koyama T, Obayashi M, Utada H (2004) Trans-Pacific temperature field in the mantle transition zone derived from seismic and electromagnetic tomography. Earth Planet Sci Lett 217:425-434 Gu YJ, Dziewonski AM (2002) Global variability of transition zone thickness. J Geophys Res 107, doi: 10/1029/ 2001JB000489 Gu YJ, Dziewonski AM, Agee CB (1998) Global de-correlation of the topography of the transition zone discontinuity. Earth Planet Sci Lett 157:57-67 Hae R, Ohtani E, Kubo T, Koyama T, Utada H (2006) Hydrogen diffusivity in wadsleyite and water distribution in the mantle transition zone. Earth Planet Sci Lett 243:141–148 Higo Y, Inoue T, Irifune T, Yurimoto H (2001) Effect of water on the spinel post-spinel transformation in Mg2SiO4. Geophys Res Lett 28:3505-3508 Hirose K (2002) Phase transitions in pyrolitic mantle around 670 km depth: implications for upwelling of plumes from the lower mantle. J Geophys Res 107, doi: 10.1029/2001JB000597

Eﬀect of Water on Mantle Phase Transitions

417

Hirose K, Fei Y (2002) Subsolidus and melting phase relations of basaltic composition in the uppermost lower mantle. Geochim Cosmochim Acta 66:2099-2108 Hirose K, Fei Y, Ma Y, Mao HK (1999) The fate of subducted basaltic crust in the Earth’s lower mantle. Nature 397:53-56 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: Implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Hosoya T, Kubo T, Ohtani, E, Sano A, Funakoshi K (2005) Water controls the fields of metastable olivine in cold subducting slabs. Geophys Res Lett 32:L17305 doi:10.1029/2005GL023398 Inoue T (1994) Effect of water on melting phase relations and melt compositions in the system Mg2SiO4MgSiO3-H2O up to 15 GPa. Phys Earth Planet Inter 85:237-263 Inoue T, Weidner DJ, Northrup PA, Parise JB (1998) Elastic properties of hydrous ringwoodite (γ-phase) in Mg2SiO4. Earth Planet Sci Lett 160:107–113 Inoue T, Higo Y, Ueda T, Tanimoto Y, Irifune T (2001) The effect of water on the high-pressure phase boundaries in the system Mg2SiO4-Fe2SiO4. Transport of Materials in the Dynamic Earth, Kurayoshi, p 128-129 Inoue T, Tanimoto Y, Irifune T, Suzuki T, Fukui H, Ohtaka O (2004) Thermal expansion of wadsleyite, ringwoodite, hydrous wadsleyite and hydrous ringwoodite. Phys Earth Planet Inter 143-144:279-290 Irifune T, Ringwood AE (1993) Phase transformations in sbducted oceanic crust and buoyancy relationships at depths of 600-800 km in the mantle. Earth Planet Sci Lett 117:101-110 Irifune T, Nishiyama N, Kuroda K, Inoue T, Isshiki M, Utsumi W, Funakoshi K, Urakawa S, Uchida T, Katsura T, Ohtaka O (1998) The postspinel phase boundary in Mg2SiO4 determined by in situ X-ray diffraction. Science 279:1698-1700 Ito E, Takahashi E (1989) Post-spinel transformation in the system Mg2SiO4-Fe2SiO4 and some geophysical implications. J Geophys Res 94:10637-10646 Jacobsen SD, Smyth JR, Spetzler H, Holl CM, Frost DJ (2004) Sound velocities and elastic constants of ironbearing hydrous ringwoodite. Phys Earth Planet Inter 143-144:47-56 Jamieson JC, Fritz JN, Manghnani MH (1982) Pressure measurements at high temperature in X-ray diffraction studies: gold as a primary standard. In: High pressure research in geophysics. Akimoto S, Manghnani MH (ed) Center for Academic Publications Japan, Tokyo, p 27-48 Kaneshima S, Helffrich G (1998) Detection of lower mantle scatterers northeast of the Mariana subduction zone using short-period array data. J Geophys Res 103:4825-4838 Kaneshima S, Helffrich G (1999) Dipping lower-velocity layer in the mid-lower mantle: evidence for geochemical heterogeneity. Science 283:1888-1891 Karato S (1989) Grain growth kinetics in olivine aggregates. Tectonophysics 168:255- 273 Karato S (1990) The role of hydrogen in the electrical conductivity of the upper mantle. Nature 347: 72–273 Katayama I, Hirose K, Yurimoto H, Nakashima S (2003) Water solubility in majoritic garnet in subduction oceanic crust. Geophys Res Lett 30, doi: 10.1029/2003GL018127 Katsura T, Ito E (1989) The system Mg2SiO4-Fe2SiO4 at high pressures and temperatures: Precise determination of stabilities of olivine, modified spinel, and spinel. J Geophys Res 94:15663-15670 Katsura T, Yamada H, Shinmei T, Kubo A, Ono S, Kanzaki M, Yoneda A, Walter MJ, Ito E, Urakawa S, Funakosh Ki, Utsumi W (2003) Post-spinel transition in Mg2SiO4 determined by high P-T in situ X-ray diffraction. Phys Earth Planet Inter 136:11-24 Katsura T, Yamada H, Nishikawa O, Song M, Kubo A, Shinmei T, Yokoshi S, Aizawa Y, Yoshino T, Walter MJ, Ito E, Funakoshi K (2004) Olivine-wadsleyite transition in the system (Mg,Fe)2SiO4. J Geophys Res 109: B02209, doi:10.1029/2003JB002438 Kawamoto T (2004) Hydrous phase stability and partial melt chemistry in H2O-saturated KLB-1 peridotite up to the uppermost lower mantle conditions. Phys Earth Planet Inter 143-144:387-395 Kawamoto T, Matsukage K, Mibe K, Isshik M, Nishimura K, Ishimatsu N, Ono S (2004) Mg/Si ratios of aqueous fluids coexisting with forsterite and enstatite based on the phase relations in the Mg2SiO4-SiO2H2O system. Am Mineral 89:1433-1437 Kerschhofer L, Rubie DC, Sharp TG, McConnell JDC, Dupas-Bruzek C (2000) Kinetics of intercrystalline olivine-ringwood transformation. Phys Earth Planet Inter 121:59-76 Kirby SH, Stein S, Okal EA, Rubie DC (1996) Metastable mantle phase transformations and deep earthquakes in subducting oceanic lithosphere. Rev Geophys 34:261-306 Koyama T, Shimizu H, Utada H, Ohtani E, Hae R (2006) Water contents in the mantle transition zone beneath the north Pacific derived from the electrical conductivity anomaly. In: Earth’s Deep Water Cycle. Jacobsen S, van der Lee S (eds) Am Geophys Union Monogr (in press) Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Koper KD, Wiens DA, Dorman LM, Hildebrand JA, Webb SC (1998) Modeling the Tonga slab: can travel time data resolve a metastable olivine wedge? J Geophys Res 103:30079-30100 Kubo A, Akaogi M (2000) Post-garnet transitions in the system Mg4Si4O12-Mg3Al2Si3O12 up to 28 GPa: phase relations of garnet, ilmenite and perovskite. Phys Earth Planet Inter 121:85-102

418

Ohtani & Litasov

Kubo T, Ohtani E, Kato T, Shinme T, Fujino K (1998) Effect of water on the α-β transformation kinetics in San Carlos Olivine. Science 281:85–87 Kubo T, Ohtani E, Kato T, Urakawa S, Suzuki A, Kanbe Y, Funakoshi K, Utsumi W, Kikegawa T, Fujino K (2002a) Mechanism and kinetics of the post-spinel transformation in Mg2SiO4. Phys Earth Planet Inter 129:153-171 Kubo T, Ohtani E, Kondo T, Kato T, Toma M, Hosoya T, Sano A, Kikegawa T, Nagase T (2002b) Metastable garnet in oceanic crust at the top of the lower mantle. Nature 420:803-806 Kubo T, Ohtani, E Funakoshi K (2004) Nucleation and growth kinetics of the α-β transformation in Mg2SiO4 determined by in situ synchrotron powder X-ray diffraction. Am Mineral 89:285-293 Kurashina T, Hirose K, Ono S, Sata N, Ohishi Y (2004) Phase transition in Al-bearing CaSiO3 perovskite: implications for seismic discontinuities in the lower mantle. Phys Earth and Planet Inter 145:67-74 Kurosawa M, Yurimoto H, Sueno S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Mineral 24:385-395 Lebedev S, Chevrot S, van der Hilst RD (2003) Correlation between the shear-speed structure and thickness of the mantle transition zone. Phys Earth Planet Inter 136:25-40 Li X, Kind R, Priestley K, Sobolev SV, Tilmann F, Yuan X, Weber M (2000) Mapping the Hawaiian plume conduit with converted seismic waves. Nature 405:938-941 Litasov KD, Ohtani E (2003) Stability of hydrous phases in CMAS-pyrolite-H2O system up to 25 GPa. Phys Chem Mineral 30:147-156 Litasov KD, Ohtani E, Langenhorst F, Yurimoto H, Kubo T, Kondo T (2003) Water solubility in Mg-perovskites and water storage capacity in the lower mantle. Earth Planet Sci Lett 211:189-203 Litasov KD, Ohtani E, Suzuki A, Kawazoe T, Funakoshi K (2004) Absence of density crossover between basalt and peridotite in the cold slabs passing through 660 km discontinuity. Geophys Res Lett 31, doi:10.1029/ 2004GL021306 Litasov KD, Ohtani E (2005) Phase relations in hydrous MORB at 18-28 GPa: implications for heterogeneity of the lower mantle. Phys Earth Planet Inter 150:239-263 Litasov KD, Ohtani E, Sano A, Suzuki A, Funakoshi K (2005a) Wet subduction versus cold subduction. Geophys Res Lett 32:L13312, doi:10.1029/2005GL022921 Litasov KD, Ohtani E, Sano A, Suzuki A, Funakoshi K (2005b) In situ X-ray diffraction study of post-spinel transformation in a peridotite mantle: Implication to the 660 km discontinuity. Earth Planet Sci Lett 238: 311-328 Litasov KD, Ohtani E, Sano A (2006) Influence of water on major phase transitions in the earth's mantle. In: Earth’s Deep Water Cycle. Jacobsen S, van der Lee S (eds) Am Geophys Union Monogr (in press) Liu M, Kerschhofer L, Mosenfelder JL, Rubie DC (1998) The effect of strain energy on growth rates during the olivine–spinel transformation. J Geophys Res 103:23897-23909 Matsui M, Parker SC, Leslie M (2000) The MD simulation of the equation of state of MgO: application as a pressure calibration standard at high temperature and high pressure. Am Mineral 85:312-316 Matsukage KN, Jing Z, Karato S (2005) Density of hydrous silicate melt at the conditions of Earth’s deep upper mantle. Nature 438:488-491, doi: 10.1038/nature04241 McCammon C (1997) Perovskite as a possible sink for ferric iron in the lower mantle. Nature 387:694-696 Mei S, Kohlstedt DL (2000) Influence of water on plastic deformation of olivine aggregates. 1. Diffusion creep regime. J Geophys Res 105:21457-21470 Miyajima N, Fujino K, Funamori N, Kodo T, Yagi T (1999) Garnet-perovskite transformation under conditions of the Earth’s lower mantle: an analytical transmission electron microscopy study. Phys Earth Planet Inter 116:117-131 Morishima H, Kato T, Suto M, Ohtani E, Urakawa S, Utsumi W, Shimomura O, Kikegawa T (1994) The phase boundary between α- and β-Mg2SiO4 determined by in situ X-ray observation. Science 265:1202-1203 Mosenfelder JL, Marton FC, Ross CR II, Kershhofer R, Rubie DC (2001) Phys Earth Planet Inter 127:165-180 Nishiyama N, Irifune T, Inoue T, Ando J, Funakoshi K (2004) Precise determination of phase relations in pyrolite across the 660 km seismic discontinuity by in situ X-ray diffraction and quench experiments. Phys Earth Planet Inter 143-144:185-199 Niu F, Kawakatsu H (1995) Direct evidence for the undulation of the 660 km discontinuity beneath Tonga: Comparison of Japan and California array data. Geophys Res Lett 22:531-534 Niu F, Kawakatsu H, Fukao Y (2003) A slightly dipping and strong seismic reflector at mid-mantle depth beneath the Mariana subduction zone. J Geophys Res 108, doi:10.1029/2002JB002384 Niu F, Inoue H, Suegetsu D, Kanjo K (2000) Seismic evidence for a thinner mantle transition zone beneath the South Pacific Superswell. Geophys Res Lett 27:1981-1984 Nolet G, Zielhuis, A (1994) Low S velocities under the Tornquist-Teisseyre zone: Evidence for water injection into the transition zone by subduction. J Geophys Res 99:15813-15821, doi: 10.1029/94JB00083 Oguri K, Funamor N, Uchida T, Miyajima N, Yagi T, Fujino K (2000) Post-garnet transition in a natural pyrope: a multi-anvil study based on in situ X-ray diffraction and transmission electron microscopy. Phys Earth Planet Inter 122:175-186

Eﬀect of Water on Mantle Phase Transitions

419

Ohtani E (1979) Melting relation of Fe2SiO4 up to about 200 kbar. J Phys Earth 27:189-203 Ohtani E (2006) Recent progress in experimental mineral physics: phase relations of hydrous systems and the role of water in slab dynamics. In: Earth’s deep mantle: Structure, Composition, and evolution. Geophys Monograph Series 160. van der Hilst RD, Bass J, Matas J, Trampert J (eds), Am Geophys Union, p 321334 Ohtani E, Touma M, Litasov K, Kubo T, Suzuki A (2001) Stability of hydrous phases and water storage capacity in the transitional zone and lower mantle. Phys Earth Planet Inter 124:105-117 Ohtani E, Litasov K, Hosoya T, Kubo T, Kondo T (2004) Water transport into the deep mantle and formation of a hydrous transition zone. Phys Earth Planet Inter 143-144:255-269 Paterson MS (1982) The determination of hydroxyl by infrared absorption in quartz. silicate glasses and similar materials. Bull Mineral 105:20-29 Revenaugh, J, Sipkin SA (1994) Seismic evidence for silicate melt atop the 410 km mantle discontinuity. Nature 369:474-476 Ringwood AE (1994) Role of the transition zone and 660 km discontinuity in mantle dynamics. Phys Earth Planet Inter 86:5-24 Rubie DC, Thompson AB (1985) Metamorphic Reactions: Kinetics, Textures, and Deformation. In: Advances in Physical Geochemistry. Vol 4. Thompson AB, Rubie DC (eds) Springer-Verlag, p 27-79 Rubie DC, Ross II CR (1994) Kinetics of the olivine-spinel transformation in subducting lithosphere: experimental constraints and implications for deep slab processes. Phys Earth Planet Inter 86:223-241 Sano A, Ohtani E, Litasov KD, Kubo T, Hosoya T, Funakoshi K, Kikegawa T (2006) Effect of water on garnetperovskite transformation in MORB and implications for penetrating slab into the lower mantle. Phys Earth Planet Inter (in press) Sakamaki T, Suzuki A, Ohtani E (2006) Stability of hydrous melt at the base of the Earth’s upper mantle. Nature 439:192-194 Shearer PM (1996) Transition zone velocity gradients and the 520 km discontinuity. J Geophys Res 101:30533066 Shearer PM (2000) Upper mantle seismic discontinuities. In: Earth's Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Geophysical Monograph 117. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) Am Geophys Union, p 115-131 Shieh SR, Mao HK, Ming JC (1998) Decomposition of phase D in the lower mantle and the fate of dense hydrous silicates in subducting slabs. Earth Planet Sci Lett 159:13-23 Shim SH, Duffy TS, Shen G (2001) The post-spinel transformation in Mg2SiO4 and its relation to the 660 km seismic discontinuity. Nature 411:571-574 Shim SH, Duffy TS, Takemura K (2002) Equation of state of gold and its application to the phase boundaries near 660 km depth in Earth’s mantle. Earth Planet Sci Lett 203:729-739 Smyth J, Frost DJ (2002) The effect of water on the 410 km discontinuity: An experimental study. Geophys Res Lett 29, doi:10.1029/2001GL014418 Song TR, Helmberger DV, Grand SP (2004) Low-velocity zone atop the 410 km seismic discontinuity in the northwestern United States. Nature 427:530-533 Speziale S, Zha CS, Duffy TS, Hemley RJ, Mao HK (2001) Quasi-hydrostatic compression of magnesium oxide to 52 GPa: implication for the pressure-volume-temperature equation of state. J Geophys Res 106: 515-528 Stixrude L (1998) Elastic constants and anisotropy of MgSiO3 perovskite, periclase, and SiO2 at high pressure. In: Core-Mantle Boundary Region. Geodynamics Series 28. Gurnis M, Wysession ME, Knittle E, Biffet BA (eds) Am Geophys Union, p 83-96 Suzuki A, Ohtani E, Morishima H, Kubo T, Kanbe Y, Kondo T, Okada T, Terasaki H, Kato T, Kikegawa T (2000) In situ determination of the phase boundary between wadsleyite and ringwoodite in Mg2SiO4. Geophys Res Lett 27:803–806 Tarits P, Hautot S, Perrier F (2004) Water in the mantle: Results from electrical conductivity beneath the French Alps. Geophys Res Lett 31, doi:10.1029/2003GL019277 Tsuchiya T (2003) First-principles prediction of the P-V-T equation of state of gold and the 660 km discontinuity in Earth’s mantle. J Geophys Res 108, doi:10.1029/2003JB002446 Turnbull D (1956) Phase changes, in Solid State Physics. Vol 3. Seitz F, Turnbull D (eds) Elsevier, p 225-306 Van der Meijde M, Marone F, Giardini D, van der Lee S (2003) Seismic evidence for water deep in the Earth’s upper mantle. Science 300:1556-1558 Weidner DJ, Wang Y (2000) Phase transformations: Implications for mantle structure Earth’s Deep Interior. In: Earth's Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Geophysical Monograph Vol. 117. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) Am Geophys Union, p 215-235 Williams Q, Garnero EJ (1996) Seismic evidence for partial melting at the base of the lower mantle. Science 273:1528-1530 Wood BJ (1995) The effect of H2O on the 410-kilometer seismic discontinuity. Science 268:74-76

18

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 421-450, 2006 Copyright © Mineralogical Society of America

Water in the Early Earth Bernard Marty Centre de Recherches Pétrographiques et Géochimiques Centre National de la Recherche Scientifique 15 Rue Notre Dame des Pauvres 54220 Vandoeuvre lès Nancy Cedex France e-mail: [email protected]

Reika Yokochi University of Illinois at Chicago Department of Earth and Environmental Sciences Chicago, Illinois, 60607, U.S.A. e-mail: [email protected]

INTRODUCTION The origin of water is a long standing problem that has fascinated generations of philosophers and scientists since the dawn of humanity. It has to do with processes that took place in the nascent solar system, but, unfortunately, we lack record of what really happened during this dark age. On Earth, the tectonic activity has erased completely any record dating back from this period, and the oldest rocks preserved on Earth crystallized about 600 Ma after start of solar system condensation (ASSC). What we observe today does not represent necessarily what was present when the solar system formed, and it may well be possible that the true water ancestors could have had chemical and isotopic characteristics that are not observed in any reservoir of the present-day solar system. However, the extraterrestrial objects that escaped planetary differentiation (chondrites, comets, interplanetary dust) are probably the best available precursor candidates for the source of volatile elements in Earth. In this context, the nature of such potential contributions can be estimated in view of chemical and isotopic mass balance, taking into account astrophysical and/or thermodynamic constraints on planetary system formation. The Earth is not a water-rich body. Water at the Earth’s surface (1.5 × 1024 g), mostly in the oceans, makes about 0.025% H2O over the whole Earth (5.97 × 1027 g). Most estimates for the mantle water content advocate a few ocean masses, so that the bulk water content of Earth may be <0.2% H2O (see the “Potential Water Contributors” section in this chapter). For comparison, carbonaceous chondrites contain up to 10% equivalent water (Boato 1954; Robert and Epstein 1982; Kerridge 1985) and comets have up to 50% H2O (Delsemme 1988). Thus the contribution of only a small fraction of such water-rich bodies could well account for the terrestrial inventory. It is possible that the amount of water trapped in the Earth was higher early in the Earth’s history, and that water was lost into space subsequently when the atmosphere was not a closed system. However, we do not see the effects of such loss in the isotopic composition of terrestrial water, which suggests that the water content of the Earth has been always low. In contrast, Venus might have had a large amount of water as its elevated atmospheric D/H ratio points to extensive water loss from a steam atmosphere (Hunten and Donahue 1976). Mars contains less water than Earth (Carr and Wänke 1992) and Mercury, the Moon and Vesta are dry. The reason for such differences may be due to heterogeneities in the supply of few water-rich bodies to dominantly dry accreting material, and/or to contrasted evolutions of terrestrial planets. 1529-6466/06/0062-0018$05.00

DOI: 10.2138/rmg.2006.62.18

422

Marty & Yokochi

The distribution of water in the solar system is presumably governed by the radial distance from the Sun, with conditions being dryer with decreasing distance (Boss 1998). A water heliocentric gradient is consistent with the distribution of different classes of meteorites in the inner solar system, as volatile-rich carbonaceous chondrites are located in the asteroid belt (2 < a < 4 AU; 1 AU is the Sun-Earth distance) and might have formed further beyond, whereas ordinary chondrites and enstatite chondrites that contain much less water (<1%, e.g., Robert 2003; however, the hydrated matrix of ordinary chondrites may contain up to 4% water; Deloule and Robert 1995) might have originated from inner regions, possibly the terrestrial formation region for the latter. In the solar nebula, water was only able to condense at distances from the Sun greater than 4.5 AU. In order to account for the occurrence of water in the Earth forming region, either hydrogen from the solar nebula was captured by the growing Earth and subsequently oxidized, or water was delivered by bodies formed in outer regions. Such supply could have occurred by late addition of volatile-rich bodies or, as recently proposed, because the Earth and sister planets formed from a swarm of planetary embryos, a few of them being volatile-rich (Morbidelli et al. 2000). Both models appear plausible in the light of solar system formation dynamics, so that independent constraints are needed to refine these possibilities. In this chapter we review current data and models, and expose our view on the origin of terrestrial water and on processes of delivery. Further recent discussions on the origin of water in Earth can be found in Abe et al. (2000); Dauphas et al.(2000); Morbidelli et al. (2000); Robert (2003), Drake (2005).

ISOTOPIC CONSTRAINTS ON THE ORIGIN OF TERRESTRIAL WATER Hydrogen isotopic ratios The chemical composition of the solar nebula is well represented by that of the Sun, which has been determined by spectroscopy, and is dominated by H2 (90%). Thus the isotopic composition of H (D/H) gives strong clues on the origin of water on Earth. Hydrogen and deuterium have been involved in nuclear reactions within the Sun so that their present-day abundances are not representative of the solar nebula. The D/H ratio of the latter has to be estimated by other means such as models for the evolution of the Sun. Jupiter’s atmosphere, a fragment of the gravitationally captured solar nebula, is also a source of important information. These approaches converge towards an estimate of the solar nebula D/H ratio of 25±5 × 10−6 (Geiss and Gloecker 1998).The D/H ratio of solar system reservoirs (Figure 1) show one order of magnitude variation between the solar nebula D/H ratio on one hand and cometary values around 310±20 × 10−6 observed in 3 comets on another hand (Bockelée-Morvan and al. 1988; Jessberger et al. 1988; Meier et al. 1998). Primitive meteorites present intermediate bulk D/H values within 130-180 × 10−6 (e.g., Kerridge 1985), although more extreme values up to 720 × 10−6 were found in phyllosilicates of the Semarkona LL3 (ordinary) chondrite (Deloule and Robert 1995). Such high values may represent the best preserved D/H fossil of a D-rich reservoir in the nascent solar system (Deloule and Robert 1995). Micrometeorites present D/H ratios which distribution mimics that of carbonaceous chondrites (Engrand et al. 1999a) (Figure 1), thus consistent with a genetic link between these objects and carbonaceous chondrites, otherwise attested by their mineralogy (Kurat et al. 1994) and their oxygen isotope compositions (Engrand et al. 1999b). Ion probe measurements of IDPs have shown their large range of D/H values on a microscale, which, together with other isotopic and elemental measurements, has been taken as an evidence for a mixed origin (cometary, asteroidal, interstellar; Aléon et al. 2001). These D/H isotopic variations among solar system reservoirs are by far the largest isotopic variations recorded so far for any element. For comparison, moderately volatile elements such as potassium present isotopic variation at the per mil level among objects of the solar system (Humayun and Clayton 1995), and refractory elements such as Mo vary by a few epsilon (10−4) units in planetary bodies (Dauphas et al. 2002b). Although it is not the goal of this work to present a discussion on this subject, we shall briefly mention some of the currently

Water in the Early Earth

Solar

423

Interstellar ice

Earth Comets

20

Carbonaceous chondrites

Number of cases

10

30

Ordinary chondrites

Chondrules

20 10

Hydrated minerals

20 16

Micrometeorites Stratospheric IDPs

12 8 4 100

200

300

400

500 1000 5000

D/H (x 106) Figure 1. Histogram of D/H values among different solar system reservoirs, following Engrand et al. (1999a), Robert (2003), and references therein. The y-axis represents the number of analyses.

discussed interpretations. Models generally assume that the solar nebula D/H ratio represents the starting value from which D/H ratios observed in the solar system evolved through either extreme isotopic fractionation, or due to addition of D-rich compounds originating from outside the solar system (Geiss and Reeves 1972). Interstellar water could have been enriched in D through isotopic exchange between the most common hydrogen bearing molecules, H2 and H2O, favoring enrichment of the latter in D (Geiss and Reeves 1972; Millar et al. 1989). Laboratory experiments suggest that the D/H value of water vapor in the solar nebula can hardly exceed 200 × 10−6 in solar system formation conditions (Lécluse and Robert 1994), so that other D-enrichment processes, such as isotopic fractionation during condensation together with a Rayleigh distillation must be advocated to account for the Semarkona and the cometary values. Extending this approach to a turbulent nebula does not solve either the problem: in such a case, the D/H ratio of the vapor cannot exceed 50 × 10−5. Thus it has been postulated that the large D enrichment observed in comets and some of the meteorites were due to isotope fractionation at low temperature during e.g., ion-molecule isotope exchange in molecular clouds (Millar et al. 1989). D/H values as high as 10−3 could be reached at 20 K, and the meteoritic values around 200 × 10−6 require temperatures around 60-80 K (Drouart et al. 1999), which are also consistent with temperatures required for accounting N isotope values of meteorites (Aléon and Robert 2004). This temperature range is higher than the one expected for cold dense clouds and could rather correspond to warmer stages during solar nebula early processes. The terrestrial D/H ratios—153 × 10−6 for the Earth’s surface including oceans (Lecuyer et al. 1998) and 136 × 10−6 for the lowest mantle value measured so far (Deloule et al. 1991)—are within the chondritic range. The Venusian (Hunten and Donahue 1976) and the Martian (Owen

424

Marty & Yokochi

et al. 1988) atmospheres are rich in deuterium compared to terrestrial oceans, a feature interpreted as resulting from isotopic fractionation during atmospheric escape. The Martian mantle may be characterized by a 10-20% depletion in D as the terrestrial mantle is relative to seawater (the H isotope standard), suggesting a common source origin for both planets (Deloule 2002). The similar range of terrestrial and chondritic D/H ratios (Figure 1) has been taken as an evidence for an asteroid, rather than a cometary origin for terrestrial water (Dauphas et al. 2000; Morbidelli et al. 2000). Alternatively, some models ascribe a solar nebula origin for part, or all, terrestrial water (e.g., Hayashi et al. 1979; Drake 2005). In such a case, the isotopic contrast between nebular and terrestrial water requires isotopic fractionation, but we have argued above that this possibility is not consistent with available experimental data. In principle, the terrestrial water D/H value (136-153 × 10−6) could result from mixing between a solar nebula component (25±5 × 10−6) and a cometary component (310±20 × 10−6), and the fact that chondritic values encompass to the terrestrial range could be coincidental. We believe however, based on other stable isotope ratios, that this possibility is unlikely.

Nitrogen and carbon isotopic ratios The nitrogen and carbon isotopic compositions of the proto-solar nebula have been recently estimated from ion probe surface isotope analysis of lunar soil grains (Hashizume et al. 2000, 2004). These studies have proposed that solar N and C isotopic ratios are enriched in the light isotopes compared to terrestrial values (15N/14N = 3.68 × 10−3 for the atmosphere, most of 15 N/14N terrestrial values including the mantle and the crust are within 2% of the atmospheric value, e.g., Marty and Dauphas 2003). A light N isotope composition for the protosolar nebula (15N/14N ≤ 2.8 × 10−3) is consistent with the Galileo data for the Jupiter atmosphere (15N/14N = 2.3±0.3 × 10−3; Owen et al. 2001). As for D/H values, meteorites (3.57 × 10−3 for enstatite chondrites on average, 3.80 × 10−3 for CI and CM carbonaceous chondrites, up to 4.2 × 10−3 for CR chondrites; e.g., Kerridge 1985) and micrometeorites (3.3-4.2 × 10−3; Marty et al. 2005) present heavier 15N/14N ratios which encompass the range of terrestrial values. Remote sensing measurements on comets Hale-Bopp and LINEAR have shown a high degree of isotopic disequilibrium for N between HCN (15N/14N = 3.1±0.4 × 10−3; Jewitt et al. 1997) and CN (15N/14N = 7.1±1.8 × 10−3; Arpigny et al. 2003), highlighting the large isotopic variation of N in the solar system, second in range after H. As for H, the cause of these variations is not fully understood and could represent at least two sources of nitrogen in the solar system. However, the HCN value is consistent with the proto solar nebula end-member value and may be more representative of bulk cometary N than CN which origin may be linked with organics processed in the interstellar medium. When considered together, D/H and 15N/14N isotope variations rule out mixing between extreme end-members as a cause of the similarity between Earth and asteroid bodies (Figure 2). It would be already coincidental to have the Earth and chondrites with the same D/H ratio by mixing of extreme H isotope end-members. The likelihood to have the same mixing ratio for N isotopes, despite drastically different cosmochemical behaviors of H and N, is close to zero. For instance, water in chondrites is mostly hosted by hydrated minerals and was probably gained during aqueous alteration on parent bodies, whereas chondritic N is mainly hosted by organics. A similar picture emerges from carbon isotope variations. According to Hashizume et al. (2004), solar carbon is depleted by ≥ 12% relative to Earth in 13C, the Earth and most of primitive meteorites (excluding CR chondrites which are also anomalous in H and N isotopes relative to other chondrites). As for H and N isotopes, the Earth is within the range of chondritic values (Figure 3), despite drastic differences in the cosmochemical behaviors of H, N and C.

Noble gas isotopic ratios Noble gases in the terrestrial atmosphere present an abundance pattern which is similar to that of chondrites, and has been labeled for this reason “planetary” (e.g., Ozima and

Water in the Early Earth

425

a) 8.E-03

Cometary CN

6.E-03

15

14

N/ N

7.E-03

Micromet. 5.E-03

Earth

CI, CM Cometary HCN

3.E-03

2.E-03

Solar

1.E-03 0.E+00

1.E-04

2.E-04

3.E-04

4.E-04

D/H

b)

4.6E-03

4.2E-03

3.8E-03

15

14

N/ N

CI, CM

3.4E-03

3.0E-03 1.2E-04

Earth

Micromet.

1.6E-04

2.0E-04

D/H Figure 2. a) Coupled H and N isotope variations in different objects and reservoirs of the solar system (see text for data sources). Mixing curves between a solar nebula end-member and cometary and meteoritic end-members are represented for two hydrogen species in the solar nebula. Dotted curves: H2 in the solar end-member; thick curves: nebular H2O (H2/H2O = 104). Two cometary values, corresponding to CN (Arpigny et al. 2003) and HCN (Jewitt et al. 1997) D/H measurements, are represented. b) Enlarged section of Figure 2a. The terrestrial values can be accounted reasonably well for by contribution of carbonaceous chondrites and/or micrometeorites.

Podosek 2002 and refs. therein). Nevertheless, isotopic ratios of terrestrial noble gases differ significantly from meteoritic and solar ones. For instance, atmospheric neon has a 20Ne/22Ne ratio lower by 30% compared to the solar 20Ne/22Ne ratio, and atmospheric xenon is enriched by ~3.6% per atomic mass unit (amu) relative to solar Xe. Atmopheric Ne is also depleted by 9% relative to Ne-Q, a neon component found in meteorites believed to be ubiquitous in

426

Marty & Yokochi 1.2E-02 Earth

1.1E-02 1.0E-02

13

12

C/ C

9.0E-03

Solar

CI, CM

8.0E-03 7.0E-03

Cometary CN

6.0E-03 5.0E-03 4.0E-03 3.0E-03 0.0E+00

1.0E-04

2.0E-04

3.0E-04

4.0E-04

D/H Figure 3. Coupled H and C isotope variations in different objects and reservoirs of the solar system. See text for data sources.

primitive chondrites, and atmospheric Xe is also fractionated by 3.3% per amu relative to Xe-Q. Because noble gases are not involved in chemical reactions, no equilibrium isotopic fractionation is expected to occur. Thus the peculiar isotopic composition of atmospheric neon and xenon has been attributed to kinetic processes specific to their host reservoir, and models of isotopic fractionation during atmospheric escape have been developed (Hunten et al. 1987; Pepin 1991; Sasaki and Nakasawa 1988). However, there are some problems with this possibility. Xenon is elementally depleted relative to krypton, as expected if heavy elements/isotopes were lost preferentially, but its isotopic composition reveals loss of light isotopes. This is the so-called missing xenon problem. If these isotopic variations are indeed due to atmospheric escape fractionation, then it is necessary to consider a suite of fractionation episodes that affected specifically each noble gas, during e.g., mantle-atmosphere exchanges (Pepin 1991; Tolstikhin and Marty 1998). Mixing of different noble gas components could also result in the observed patterns (Owen and Bar-Nun 1996; Dauphas 2003). Atmospheric noble gases might have been first depleted in light elements/isotopes during atmospheric escape and then mixed with cometary noble gases. Indeed comets may be depleted in neon because comet formation temperature did not allow trapping of Ne, and depleted in xenon because trapping of Xe in ice is less efficient then of Ar and Kr (Notesco et al. 2003) Mass balance considerations suggest that in this two-stage process, major volatile elements like water, C or N were not added significantly, because the noble gas/(C+N) ratio is much higher in comets than in the proto-Earth or proto-atmosphere to which cometary material is added. Most models for the early evolution of the atmosphere request atmospheric noble gases to originate from the solar nebula (Pepin 1991; Tolstikhin and Marty 1998; Dauphas 2003). In this case, the amounts of H(2O), C and N contributed by the solar nebula to Earth were very small, because the solar nebula has low H, N, C/noble gas ratios compared to asteroids, planets and comets. This implies that H, C and N were supplied by other source(s) such as comets or asteroids, possibly after the major episodes of noble gas isotopic fractionation. In this case, these contributions should not have erased the Ne and Xe fractionated signature of the atmosphere. This condition, together with D/H systematics of comets, chondrites and Earth, allows one to set strong constraints on the nature of water carriers as well as on the mass of impacting bodies requested to supply water and major volatile elements to Earth (Dauphas

Water in the Early Earth

427

et al. 2000). Results of the mass balance calculation (Figure 4) show that, for reasonable volatile contents and isotopic compositions of carbonaceous chondrites and comets, mass balance solutions request the cometary contribution to be small (<10%, probably <1% of impacting material Dauphas and Marty 2002) relative to the chondritic one. Within these solutions, atmospheric neon, thought to predate this contribution, is not affected. There also exists a subset of solutions that allows preservation of the fractionated isotopic composition of atmospheric xenon. Furthermore, the mass of impacting bodies in this scenario (1 × 1023 2 × 1025 g) is compatible with the mass of impacting bodies on Earth estimated from the lunar cratering record (0.01-7 × 1026 g, Chyba 1990) and with the highly siderophile element (HSE) budget of the mantle (~3 × 1025 g, Chyba 1991) that will be discussed later. In the mixing scenario of Dauphas et al. (2000), up to half of terrestrial water was supplied by this asteroid material addition the rest being already in “early Earth.”

Other isotopic constraints It is well known since three decades that the Earth, the Moon, and enstatite chondrites lie on a common mass fractionation line for oxygen isotopes (e.g., Clayton 2003 and refs. therein), which has been regarded for long as an evidence for Earth accretion from enstatite chondrite-like planetesimals (e.g., Javoy et al. 1986). An enstatite chondrite-like origin for the Earth is consistent with molybdenum (Dauphas et al. 2002a) and chromium (Shukolyukov and Lugmair 2000) isotope variations among inner solar system bodies. However, enstatite

Mass of comets+c-chondrites, Kg

1E+24 Lunar record 1

Atmospheric neon

1E+23

Atmospheric xenon 1E+22 Lunar record 2

1E+21

Lunar record 3

1E+20 0.001

0.01

0.1

1

α = Mcomets /(Mcomets +Mc-chondrites ) Figure 4. The total mass of impacting bodies on Earth as a function of the cometary mass fraction in the total mass of impacting (comets plus asteroids) bodies (after Dauphas et al. 2000). The possible range of impacting bodies that might have fallen onto Earth (“Lunar record” dotted lines parallel to the x-axis) has been estimated by Chyba (1991) from the lunar impact (pre-Nectarian) record. The grey-shaded area represents the mass of chondritic material necessary to account for the amount of highly siderophile elements in the mantle (Chyba 1991), with the presumption that these elements were delivered by late impacts after core formation. The curves represent the range of solutions for such mixing that are allowed by D/H systematics of Earth, comets, and asteroids. The mass of chondritic matter that would be necessary to supply the amount of atmospheric neon and xenon are also indicated by lines parallel to the x-axis. Because the isotopic compositions of atmospheric Ne and Xe differ from the chondritic ones, the set of D/H solutions must also contribute less Ne and Xe than these limits, which is the case for Ne and more difficult, but still possible, for Xe.

428

Marty & Yokochi

chondrites do not contain an appreciable amount of water, and their very reducing character would have prevented formation of water during early evolution of the Earth. Thus it has been proposed a mixed model for Earth formation in which an initially reducing Earth was later contributed by volatile-rich, oxidizing material (Wänke et al. 1984; Javoy 1997). The enstatite chondrite model for the Earth is not without problems. First, the terrestrial mantle has Mg/Si and Al/Si ratios similar to those of the carbonaceous chondrites, particularly of the CV type, and significantly higher than those of the enstatite chondrites (Allègre et al. 2001; Palme 2001). Thus, if the Earth was formed by enstatite chondrite-like material, then there existed a process able to remove a fraction of Si from the mantle. Trapping of silicon in the core may be a possibility, which needs further experimental checking (see discussion in Hillgren et al. 2002). Second, the formation of the Earth from enstatite chondrite-like planetary embryos is difficult to reconcile with dynamic models of terrestrial planet formation (e.g., Becker et al. 2003), which rather call for a mix of different bodies originating from variable heliocentric distances (e.g., Morbidelli 2002). In this case, one may argue that a mix of oxygen isotope signatures of various classes of different classes of meteorites could have lead to the observed match between the Earth and the enstatite chondrites. We do not wish to enter here into this important debate but we point out that, if the Earth was built up in majority by enstatite chondrite material, then contributions from material having different volitile composition is required. To conclude this subsection, it seems unlikely that the solar nebula alone could have supplied the water inventory of Earth. Likewise, a major cometary contribution appears also unlikely. Contribution from asteroid material is compatible with the isotopic signatures of H, C and N of Earth, meteorites and micrometeorites, and therefore imposes strong clues on the origin of terrestrial water. This does not eliminate a possible minor contribution of the two first candidates, and it is possible that isotopic variations of H and N seen among the surface reservoirs (oceans, atmosphere, crust) and the mantle could partly originate from primordial heterogeneities.

POTENTIAL WATER CONTRIBUTORS Examinations of isotopic structures in the Earth and potential water contributors are consistent with a water-rich, chondritic-like material as a major source of terrestrial water. In standard models of solar system formation, such material is thought to have originated beyond 2.5 AU. Water could have been supplied by contribution of asteroid bodies (Chyba 1990), or asteroid dust (Pavlov et al. 1999; Maurette et al. 2000), or could be the result of collisions between planetary embryos with a few of them originating from the “wet” zone of the inner solar system (Morbidelli et al. 2000). First, we review estimates for the Earth’s water content. Water at the Earth’s surface (1.5 × 1024 g) amounts for 250 ppm relative to the bulk Earth. Most estimates are at maximum 10 ocean masses or less in the mantle. A global H2O content (mantle+surface) of ~350±50 ppm is derived from a mantle 145±85 ppm H2O, based on H2O/Ce ratios of MORB (Saal et al. 2002). This estimates holds for the MORB mantle composition and may be a lower limit, given the depleted character of this reservoir. From K2O/H2O relationships in oceanic basalts, Jambon and Zimmermann (1990) proposed a preferred bulk Earth water content of 1,300 ppm, with a possible range of 550-1900 ppm. This estimate is based on the assumption that there exists a volatile-rich lower mantle, based on noble gas systematics of mantle plumes. The existence of such mantle stratification and/or the extent of a large deep mantle, volatile-rich, reservoir is increasingly questioned by recent geochemical and geophysical models. Experimental data on water solubility and partitioning among minerals suggest that most of the water in the mantle is concentrated in the transition zone (e.g. Bolfan-Casanova et al. 2000). Electrical conductivity measurements, on the other hand, suggest that the transition zone does not contain more

Water in the Early Earth

429

than 1000-2000 ppm of water (Huang et al. 2005). This leads to a bulk water content in the Earth’s mantle of no more than half an ocean mass. Based on measurements in nominally anhydrous minerals, Bell and Rossman (1992) estimated that at most ~1.2 × 1024 g H2O could be stored in the whole mantle, which makes, together with surface water, 450 ppm for the bulk Earth. A similar result is obtained by assuming an upper mantle water content around 250 ppm (consistent with data from xenoliths and the water content measured in MORBs) and equilibrium partitioning of water throughout the mantle (Keppler and Bolfan-Casanova 2006) All these estimates converge towards a Earth water content within 350-500 ppm. The amount of water in the Early Earth could have been much higher, up to 50 ocean masses (Abe et al. 2000), but we do not see such extreme water content in the present Earth, either because this estimate is too high, or because most of water was lost during early terrestrial evolution (e.g., the lunar forming event), or both. Indeed, given the extremely dry character of the Moon, it is doubtful that a wet Earth, if it existed before, could have survived this event. The mass of carbonaceous chondrite-type (~10% water) material necessary to supply terrestrial water is at least 3 × 10−3 times the Earth’s mass, which is roughly ten times more than the current mass of the entire asteroid belt (Morbidelli et al. 2000). However, the asteroid belt was probably more massive in the past (Weidenschilling 1977). Timing is also of fundamental importance. Models of solar system formation predict that the Earth and terrestrial planets formed by collisions between planetary embryos (Wetherill 1992). The duration of embryo formation was short (few 105 yr), but collisions were enhanced by orbital perturbations resulting from the formation of Jupiter and Saturn. Jupiter’s formation might have lasted ≤10 m.y., which gives a rough time constraint of the formation of inner planets of the solar system. The 182Hf-182W extinct radioactivity chronometer (T1/2 = 9 m.y.) has recorded a major episode of Earth’s silicate-metal differentiation at ~11-30 Ma ASSC, which could indicate that the size of our planet was already significant at that time, maybe close to its present size. Such a short timescale is, however, under discussion, and longer durations of the order of 100 m.y. have been advocated from geochemical arguments (Allègre et al. 1995) as well as from dynamical ones (Wetherill 1992).

Contribution of water-rich planetary embryos According to Wetherill (1992) and Chambers and Wetherill (1998), the runaway process that led to the formation of planetary embryos was not restricted to the terrestrial planet region (a <2 AU) but also occurred in what is now the region of the main asteroid belt (2 < a < 4 AU). Results of such simulations indicate that in the a <2 AU region a few planets with masses comparable to that of the Earth are formed. Using a similar code, Morbidelli et al. (2000) showed that a few (typically 1 to 3) planetary embryos originating from a >2.5 AU and for this reason presumably rich in volatile elements are incorporated in the growing Earth, with their total mass fraction amounting 9-36%. Furthermore, in 7 of 11 simulations, collisions of these wet embryos occur towards the end of accretion (t > 36 m.y. up to 121 m.y.). This process would have been able to supply much more water than presently estimated in the present-day Earth, with the adequate D/H ratio (Morbidelli et al. 2000). It is important to note that this model assumes that planetary embryos did not undergo differentiation, preventing water degassing during magma evolution. Further modeling that will include planetary differentiation is needed to check the validity of this assumption. Nevertheless, this approach is very attractive because the contribution of water-rich embryos is a direct consequence of dynamical simulations developed on an independent ground.

Asteroid contribution Despite the observation that the present-day flux of meteorites is dominated by ordinary chondrites (Sears and Dodd 1988), there is evidence that volatile (carbon) -rich asteroids (Gaffey et al. 1993) dominate the asteroid belt population (Gradie et al. 1989) and that

430

Marty & Yokochi

most xenoliths in meteoritic regolith breccias resemble carbonaceous chondrites (Anders 1978). This predominance is in agreement with the fact that the lunar regolith contains 12% carbonaceous debris (Keays et al. 1970), that a significant fraction of nitrogen in the lunar regolith may be derived from a volatile-rich chondritic source having a N content characteristic of carbonaceous chondrites (Hashizume et al. 2002) and that micrometeorites, which constitute presently the largest flux of extraterrestrial matter on Earth, resemble most carbonaceous chondrites (Kurat et al. 1994). These observation support the possibility that volatile-rich matter was ubiquitous in the solar system. Morbidelli et al. (2000) investigated the efficiency of water delivery by primitive asteroids which, in their model, are bodies of negligible mass with respect to planetary embryos, formed at a >2.5 AU, and are presumably in majority of the carbonaceous-type and therefore rich in water. Before the formation of Jupiter, the probability of encounter between asteroidal bodies and the proto-Earth was close to 0. After Jupiter accreted its final envelope, eccentricities of asteroids increased drastically and orbital excitation resulted in depletion of the asteroid belt. Asteroid contribution to inner planets peaked at ca. 10 Ma ASSC (the assumed time of the end of Jupiter formation) and became low after 35 Ma ASSC, when the proto-Earth was about 60% of its present mass according to Morbidelli et al’s model. Assuming a 10% water content and 100% retention efficiency during accretion, these authors estimated that 5 × 10−3 terrestrial mass of hydrated asteroidal material is necessary to supply water to Earth, which is about 10 times the current mass of the entire asteroid belt. Thus the delivery of water from this region could have been significant only at a time when the asteroid belt was more massive than at present. This is consistent with a 102-103 more massive asteroid belt in the past (Weidenschilling 1977). Morbidelli et al. (2000) concluded that the delivery of volatile elements to terrestrial planets by dust due to collisions between asteroids might not have been efficient after 35 Ma due to the disappearance of asteroids in their model. According to these authors, such a short timescale compared to a typical duration of 100 m.y. for growing completely the Earth in N-body simulations poses a serious problem concerning the origin of water from asteroids. This problem of timing may however be lifted if the terrestrial growth was faster. In line with this possibility, Martian meteorites have recorded in their extinct radioactivity products major differentiation events that occurred on Mars within 15 Ma ASSC (Halliday et al. 2001), and the last major metal-silicate differentiation event of the Earth occurred at 11-30 Ma ASSC (Kleine et al. 2002; Yin et al. 2002). There is certainly a strong possibility that the Earth and other terrestrial planets were already formed close to their present size in 30 Ma ASSC or less, allowing time for contribution of hydrated asteroids before their dissipation.

Constraints on water delivery by asteroidal material from the terrestrial highly siderophile element budget Highly siderophile elements (HSE) in the mantle are apparently out of equilibrium with the core: they are more abundant in mantle-derived rocks than metal-silicate equilibrium would allow one to predict, and their relative abundance pattern is chondritic. For this reason, it has been proposed that HSE were contributed to Earth by chondritic material after core formation (Chou 1978; Morgan 1986). It is not clear if this contribution represents a late addition of material after core formation, or is also the result of incomplete metal-silicate equilibrium. It has also been argued that the metal-silicate partition coefficients determined at low pressure and temperature are not necessarily relevant to partitioning at mantle P-T condition (see Righter 2003, and refs. therein). Nevertheless the near-perfect chondritic relative abundance of PGE in the terrestrial mantle seems difficult to obtain by an ad hoc combination of partitioning under specific thermodynamic conditions (e.g., Brandon et al. 2006), and strongly suggests that the Earth contains a tiny fraction of chondritic material that never equilibrated with the core. 0.5% to 1% late accretion (with respect to core formation) of chondritic material could account for the mantle HSE concentrations (Morgan 1986).

Water in the Early Earth

431

The nature of the asteroidal source for HSE has been discussed from osmium isotopes. Rhenium and osmium are two HSE that are linked by the β-decay of 187Re to 187Os. The isotopic composition of Os (e.g., 187Os/188Os) has been found to vary slightly among different classes of undifferentiated meteorites, the reason of which being presumably in relation with thermodynamic conditions of the respective formation regions and/or processes leading to variable fractionation of the Re/Os ratio. The Os isotopic composition of the primitive upper mantle (PUM), estimated after correction for age effects and parent/daughter fractionation due to mantle depletion events, has been found to match the Os signature of ordinary chondrites, and to a lesser extent that of enstatite chondrites, rather than the carbonaceous chondrite one (Meisel et al. 2001). Dauphas et al. (2002c) have argued that a difference in Os isotopic composition between the primitive upper mantle and carbonaceous chondrites does not necessarily implies that carbonaceous chondrites did not contribute highly siderophile and volatile elements to Early Earth. These authors proposed an hybrid model in which the Os isotopic composition of the PUM is the result of mixing between a HSE component left after core segregation and late addition of carbonaceous material. This model places strict constraints on Pt, Re, and Os relative partition coefficients between metal and silicate that will be testable in the future. A comparison (Figure 5) between (carbonaceous) chondritic abundances of volatile elements (H2O, C, N, noble gases) and terrestrial (mantle and mantle + surface) suggests that

1E-02

Mantle mantle+surface

Terrestrial, mol/g

H 2O C

1E-06 Terrestrial = chondritic

N 1E-10 Terrestrial = 0.003 x chondritic

36

1E-14 22

Ar

Ne

84 Kr 130

1E-18 1E-18

Xe

1E-14

1E-10

1E-06

1E-02

Chondritic, mol/g Figure 5. Comparison between chondritic (carbonaceous) abundances and terrestrial abundances of volatile elements. Open circles: mantle, squares: mantle + surface (oceans, atmosphere). The terrestrial water content range is the one discussed in the text (350-1,000 ppm H2O), the upper limit being represented by the black triangle. For other volatile elements, we have taken a silicate Earth N content of 1.7 ± 0.7 ppm which is well constrained from N-Ar-K systematics (Marty and Dauphas 2003), and have computed noble gas contents from a N2/40Ar silicate Earth ratio of 80±20, a 40Ar/36Ar ratio of 30,000, and the noble gas isotopic composition of the popping rock computed for 20Ne/22Ne = 13.0 (Moreira et al. 1998). Note that the estimate for the mantle noble gas content will increase significantly if one considers a 40Ar/36Ar ratio of ~5,000 thought to represent the mantle source of plumes (Marty et al. 1998). For carbon, we took a C/N ratio of 535 ± 224 from analysis of oceanic basalts (Marty and Zimmermann 1999).

432

Marty & Yokochi

a depletion factor of ~ 0.003 relative to Chondritic, can account fairly well for the abundances of water, C and N and, partly, for those of noble gases in the Earth’s mantle. This depletion is within the range of the one envisioned for HSE in the terrestrial mantle, suggesting a common delivery origin for HSE and volatile elements. Assuming that the present-day bulk Earth water content (350-1,000 ppm) is somewhat representative of the amount of water supplied during accretion, that is, 100% efficiency for water retention during accretion and early atmospheric events, and further assuming that the late accretion contributed 0.5-1% terrestrial mass based on the HSE budget, then the water content of these contributors should have been between 3.5% and 20%. The upper limit is clearly outside the range observed in meteorites and, if appropriate, would suggest either: (i) contribution of water-rich asteroidal matter not yet recorded in meteoritic collections; (ii) contribution of comets, which is not supported by D/H values; (ii) only part of water has been delivered by this process. Therefore, this possibility gives credit to models which call for water trapped during Earth’s building stages (e.g., Dauphas et al. 2000; Morbidelli et al. 2000). The lower limit is, in contrast, fully compatible with an asteroidal contribution. Indeed, carbonaceous chondrites contain 5-10% H2O (Boato 1954; Robert and Epstein 1982; Kerridge 1985). Furthermore, the hydrated matrix of ordinary chondrites may contain several % H2O, as demonstrated in the case of the Semarkona ordinary chondrite (LL3) matrix that contains 1-4% H2O (Deloule and Robert 1995). An ordinary chondrite (OC) origin for water would be consistent with the Os isotope constraints. However, available bulk H2O data for ordinary chondrites tend to favour low water contents for these meteorites (Robert 1978), so that the OC hypothesis needs to be evaluated by further analyses of this type of meteorites. We still know very little about the origin of water in meteorites. Both ordinary and carbonaceous chondrites show evidence of extensive aqueous alteration on their parent bodies, highlighting the general occurrence of water in the early solar system. Chondrites are themselves bodies of the solar system that gained their H, C and N isotope signatures from sources having very contrasted isotope compositions. The observation that most of bulk isotopic ratios of chondrites are within a limited range supports the existence of homogenization processes between nebular and heavy isotope-rich reservoirs, that could have taken place during previous generations of small bodies. The different chondrite parent bodies that we see today may in fact constitute remnants of these that have survived the collisional period of the nascent solar system. Some of these bodies could have had the right osmium isotopic signature while bearing several percents of water gained during hydration processes. It is probable that considerable dust was generated during this period, which might have carried some water inherited from this lost planetesimals. We explore in the next subsection this possibility.

The case of interplanetary dust as a source of terrestrial water There exist water-rich objects which, at present, dominate the flux of extraterrestrial matter on Earth. These are dusts and grains known as interplanetary dust particles (IDP) for sizes <50 µm as recovered by stratospheric planes, and micrometeorites (MM) for size in the range 50400 µm recovered in polar ices. It is probable that the size of these particles form a continuum, so that we refer in the following to IDPs for both stratospheric IDPs and micrometeorites. The total flux of dust onto Earth is of the order of 103-104 tons/yr (Love and Brownlee 1993), whereas the flux of meteorites is only ~10 tons/yr (Bland et al. 1996). IDPs might have variable origins, cometary and asteroidal. Notably, Antarctic micrometeorites share resemblance with carbonaceous chondrites (Engrand et al. 1999b), especially of the CM type (Kurat et al. 1994) which contain ≥4% water, and have D/H ratios comparable to the terrestrial ones (Engrand et al. 1999a). Hence interplanetary dust might have contributed significantly volatile elements during formation and early evolution of terrestrial planets (Pavlov et al. 1999; Maurette et al. 2000), provided that they also dominated the extraterrestrial flux onto terrestrial surfaces during these periods. Maurette et al. (2000) estimated that, assuming a IDP accretion rate 106 times higher in the first 100 Ma ASSC, all atmospheric neon and carbon present at the Earth’s

Water in the Early Earth

433

surface could have been delivered by this way, but that the IDP flux with this rate was too short to supply the oceans, thus requiring other source(s) of water for the Earth. Pavlov et al. (1999) proposed that IDPs of presumably cometary origin would have been loaded with solar wind hydrogen during transit to Earth. These authors noted that (i) implantation of solar wind hydrogen in interplanetary space is a fast process (102 yr) compared to the transit time of IDPs (103-105 yr), and (ii) the intermediate value of terrestrial D/H ratio between solar nebula and cometary ones would have resulted form mixing of cometary H with solar wind H. Ultimately, hydrogen from IDPs entering the terrestrial atmosphere would have been oxidized to form water by reaction with oxygen and/or OH. Whether or not hydrogen implanted in IDPs could have survived atmospheric entry, and/or oxidation would have taken place at this epoch needs certainly further investigation. There is strong evidence, from the study of extrasolar planetary systems, that dust was abundant during the first tens to hundreds of million years in planetary systems. Dust originates either from comet evaporation, or from collision between asteroids and planetary embryos. Small dust grains (<1.5 µm) are blown off from planetary disks by the star radiation pressure over typical timescale of several tens of years. In contrast, larger grains absorb the stellar photons and re-radiate them in a way that cause them to spiral in towards the central star over timescales of 103-105 yr. Hence the occurrence of dust for long periods of time reflects collisions between larger objects. Collisions themselves are extremely fast events so that dust is swept up rapidly, however collisions induce debris and orbital destabilizations which themselves result in further collisions on much longer time scales up to several tens of million years. Hence the production of dust can continue for long periods of time even if major collisions have ended. Okamoto et al. (2004) reported the observation of amorphous silica dust rings around β-Pictoris, a ~12 m.y.-old planetary system, that were interpreted as reflecting planetary formation in the disc. Within the most inner rings, a band with no dust was interpreted as representing a planet at 12 AU that has collected dust along its orbit. From infrared observations of dust around stars of various ages, Habing et al. (1999) reported that planetary systems younger than 400 m.y. contain abundant dust whereas 90% of those older than this time interval are cleaned up. We believe that interplanetary dust supplied a significant fraction of terrestrial water, for the following reasons. First, there is strong evidence that nascent planetary systems were dust-rich and that dust might have been present during several tens to hundreds of million years. Second, the process of delivery is more gentle than in the case of planetary embryos, so that volatile loss from the impactor-target system is unlikely. Third, the D/H ratios of micrometeorites present a distribution that centers around the ocean value. The nitrogen isotopic ratios of a few micrometeorites define a range that also encompasses the terrestrial values (Marty et al. 2005). Finally, the timing may be also adequate, as we suggested in the preceding subsection. The amount of volatile-rich dust of the carbonaceous chondrite type necessary to supply terrestrial water is in the range of ~ 10−3 terrestrial mass (see Figure 4) which, considering the possible delivery efficiency from the asteroid belt, is high compared with the amount that the nascent asteroid belt could have supplied (Weidenschilling 1977), but not impossible. An estimate of the water delivery flux by interplanetary dust can be assessed by assuming a H2O content of ~ 5-10% characterizing carbonaceous chondrites and thought to be the best analogues of micrometeorites (MMs). Interplanetary dust particles (IDPs) sampled in the stratosphere may contain more water (Aléon et al. 2003; Messenger 2000). However the IDP mass flux is ~0.01 times that of micrometeorites, (Love and Brownlee 1993), so that their water contribution appears limited, perhaps at best similar to that of MMs. Comets contain about 50% H2O and might have contributed significantly to cosmic dust. However, we have seen that mass balance considerations involving volatile and siderophile elements together with D/H ratios of the terrestrial oceans suggest that the bulk cometary contribution to the terrestrial inventory of volatile elements was very limited, of the order of 10−3 relative to the

434

Marty & Yokochi

total mass of contributors (Dauphas and Marty 2002). We make the conservative hypothesis that the initial water content of cosmic dust before atmospheric entry is on the order of ~10%, probably within a factor of 2. As discussed previously, the asteroid contribution necessary to account for the HSE budget, 0.5-1% of terrestrial mass, could also account for the terrestrial water content, provided that the water content of the contributing material was within 3.5-20%. With this respect, IDPs fit well this requirement, but there subsists the problem of the nature of this contributor raised by osmium isotopes that requires ordinary chondrite (or, marginally, enstatite chondrite) material, whereas oxygen isotope measurements of Antarctic micrometeorites confirmed the link between carbonaceous chondrites and AMM (nevertheless, a few O isotope data plot in the field of ordinary chondrites; Engrand et al. 1999b). Although Os isotopes have not been yet analyzed in IDPs, the Os isotope constraint must be considered seriously. A possible way to turn around this problem is to envision a mixed origin for HSE and water, in which the HSE and a fraction of water could have been supplied by ordinary chondrite type material whereas a major fraction of water would have been supplied by water-rich IDPs. For example, a mix of 20% CC-like IDP containing 10% H2O with 80% OC-like hydrated material containing 2% H2O in the hydrated matrix could fit both the Os isotope composition of the primitive upper mantle, the D/H composition of the Earth, and amount of terrestrial water. The cosmic dust flux might not have varied dramatically since 3.8 Ga ago, except for a significant increase in the last 0.5 b.y. (Grieve and Shoemaker 1994; Culler et al. 2000; Hashizume et al. 2002). The cratering record of the Moon is also compatible with a nearconstant planetary contribution within a factor of 2 since 3.8 Ga ago (Hartmann et al. 2000). Thus we assume, as done previously (Chyba and Sagan 1992), that the cosmic dust flux remained constant since 3.8 Ga at a rate similar to the present-day one. Another important question is the flux ratio between dust and larger objects. The present-day mass ratio between dust and meteorites is of the order of 102-103. Although highly imprecise, the mass contribution due to large objects, which is dominated by rare events of km or more sized objects, might have been comparable to the cosmic dust flux, averaged over the last 3 b.y. (Kyte and Wasson 1986; Anders 1989; Trull 1994). Here we assume the flux of cosmic dust varied in proportion to that of larger objects. A constant cosmic dust flux similar to the present-day one (~30,000 tons/yr) integrated over 3.8 b.y. could have supplied ~ 1 × 1019 g H2O to the atmosphere, which is negligible compared with the bulk water content of the Earth (2.1-6.0 × 1024 g). There is an active debate about past fluxes before 3.8 Ga. Although some argue a steep decline of the cratering rate between 4.5 Ga and 3.8 Ga, there is compelling evidence that a spike of bombardment took place 4.0-3.8 Ga ago and that in the time interval 4.4-4.0 Ga ago the impacting flux was not dramatically high. The following estimates are based on the recent review and arguments given by Hartmann et al. (2000). The total mass of impactors during the last spike of bombardment around 3.9 Ga was ~6 × 1021 g. The content of siderophile elements in the ancient highlands suggests that the amount of interplanetary mass accumulated by the Moon in the 4.4-4.0 Ga period is about the same as that required to form the 3.9 Ga basins, so that the post 4.4 Ga contribution to the lunar surface might have been ~1.2 × 1022 g. This is ~20 times less than the contribution suggested by models with a declining bombardment since 4.4 Ga ago (Hartmann et al. 2000). This contribution includes both bolides (evidenced by remnants of the lunar cataclysm) and cosmic dust, all of them being integrated in the siderophile element record. Extrapolation of the lunar record to the case of Earth is rendered difficult by the possibility that bolides larger than those impacted on the lunar surface could have struck the Earth, given the larger dimension and therefore higher encounter probability of our planet. For cosmic dust, differences between Earth and Moon in collecting cosmic dust are mainly due to contrasted surface area and gravitational focusing (Kortenkamp et al. 2001). A lower limit for the Earth’s efficiency over Moon to collect cosmic dust is given by the ratio between the

Water in the Early Earth

435

surfaces of the two bodies (13 for Earth and Moon). Gravitational focusing possibly increases the efficiency of collection by a factor of ~3 (Hashizume et al. 2002). Therefore, the range of extraterrestrial material collected by Earth between 4.4 Ga and 3.8 Ga might have been ~2.47.2 × 1023 g. This contribution represents ~2-7 × 1022 mol H2O, that is ≤5% of the minimum water budget of our planet and can be neglected as an important water contributor to Earth. Thus terrestrial water was probably delivered before 4.4 Ga ago, during the first 100 m.y.

PROCESSES OF WATER INCORPORATION IN EARTH Solar nebula Some models of planetary accretion, known as the “Kyoto School” advocate the presence of nebular gas during accretion of terrestrial planets (Hayashi et al. 1979). This view is consistent with the relatively short time interval for the accretion of Mars-sized bodies of about 106 yr (e.g., Wetherill 1990; Weidenschilling et al. 1997; Morbidelli 2002) on one hand, and on the possible existence of the solar nebula up to 107 yr (Podosek and Cassen 1994; Feigelson and Montmerle 1999) on another hand. The recent observation of hot Jupiters much closer to their central star than 1 AU (Bodenheimer and Lin 2002) give direct evidence for the transit of volatile-rich bodies in central regions of nascent extrasolar systems. Gravitational capture of the gas nebula supplies a huge quantity of hydrogen on Earth. Thus in such a case a significant concentration of nebular water could have been trapped in the Earth’s proto-mantle and some vestiges could still be found in the mantle, as solar-like Ne is. Indeed, the mantle contains a neon component with 20Ne/22Ne up to 13.0±0.2 (Yokochi and Marty 2004), close to the solar value of 13.7±0.5 (Benkert et al. 1993; Wiens et al. 2004) that suggests incorporation of solar nebula gas into the growing Earth (Marty 1989; Honda et al. 1991; Hiyagon et al. 1992; Yokochi and Marty 2004). If such process would have governed the water content of the Earth, then the D/H ratio of mantle water should have kept a record of this contribution. Interestingly, the δD value (where δD is the deviation, in parts per mil, of the sample D/H ratio with respect to terrestrial ocean D/H ratio) of the deep mantle has been constrained to be ≤ −120‰ (Deloule et al. 1991), to be compared with the protosolar nebula value of −830‰. This comparison suggests that ca. 10% of mantle water could have originated from dissolution of a primordial, solar-like atmosphere in the proto-mantle. It should be noted that alternative models for trapping of noble gases in Earth exist. For instance, Trieloff et al. (2000) have argued that the mantle has neon with 20Ne/22Ne = 12.5 that characterizes matter irradiated by the Sun rather than the solar nebula. In such a case, the argument based on D/H for a nebularlike origin of some of terrestrial water may still hold since material irradiated by the Sun in space may also be rich in solar hydrogen. In order for a planet to retain a solar nebula atmosphere, it must reach a size large enough to gravitationnally retain a gas phase while the solar nebula is present. The most serious uncertainty in estimating the contribution of a solar nebula component on Earth is therefore the relative timing of the dissipation of the nebula relative to that of the accretion of the Earth. Gaseous molecules and atoms near an accreting planet are gravitationally captured if the gravitational energy of the gas constituents exceeds their kinetic energy as well as the gravitational energy of the nearby star. Neglecting the latter, the lower size limit for a planet to retain an atmosphere for hydrogen is given by: G M 3k B T > r 2µ m H

(1)

where G, kB, mH are the gravitational constant, the Boltzman constant and the mass of the hydrogen atom, respectively, M and r are the mass and radius of the accreting planet, respectively, and µ and T are the molecular weight of the gas and the temperature of the

436

Marty & Yokochi

planetary surface, respectively. The minimum planet size required for atmosphere retention is a function of the surface temperature, as shown in Figure 6 for an hydrogen atmosphere and a water vapor atmosphere between 300 and 3000 K, respectively (the planet density is assumed to be 5 g/cm3). The size of the planet needs to be at least 200 km in radius to retain a waterdominated atmosphere, and will be larger if the temperature is higher and/or the molecular weight of the atmosphere is lighter (i.e., H2). The temperature of the accreting planet is controlled by (i) solar radiation, (ii) gravitational energy release and (iii) radial heat loss to the space. Sasaki (1990) presented a model temperature of the planetary surface. He assumed that the atmosphere is in radiative equilibrium so that the surface temperature is given by: T = G Ms

µ mH 4 kB r

(2)

where Ms is the mass of the accreting Earth and other parameters have been defined above. The model temperature at a given radius is shown in Figure 6 with the planet size required to retain an atmosphere as a function of temperature. Note that this model provides a minimum value for temperature since no blanketing effect is taken into account. The real curve is therefore likely to be shifted to the right hand side as discussed later, thus perturbing the stability of the accreted atmosphere. According to curve (a), melting of the proto-Earth will occur when its radius reaches 3800 km (0.2 ME). An alternative scenario for nebular H2O as a source of terrestrial water has been recently proposed in which water was trapped from the solar nebula due to adsorption on dust from which larger bodies formed (Drake 2005). The problem with this issue is mainly isotopic, that is, the drastic difference in D/H values between the solar nebula and the Earth, and the apparent impossibility to raise the nebular D/H value to the terrestrial one under adequate thermodynamic conditions and during a reasonable interval of time (Lécluse and Robert 1994; Drouart et al. 1999).

7000

(a) 6000

H2

Model Temperature

R (km)

5000

Blanketing Effect

4000 3000

H2O

2000 1000 0 0

1000

2000

3000

4000

5000

6000

7000

T (K)

Figure 6. Evolution of surface temperature as a function of the proto-planet size (radius, in km), according to the model of Sasaki (1990). The solid line indicates the temperature (K) in function of planetary size, and defines a lower limit for temperature because no blanketing effect the atmosphere is taken into account. The dotted lines represent the planetary size required to retain an atmosphere of water or of hydrogen as a function of temperature.

Water in the Early Earth

437

Impact degassing As we have seen from isotope consideration, a significant fraction of volatile elements now in the atmosphere and oceans were originally trapped in solid bodies. They could have been extracted by melting and by impact degassing. The effect of impacts on water incorporated in the accreting body has been investigated theoretically and experimentally. Theoretical studies were first inspired by the discovery of the impact craters on the Moon through the Apollo missions (Ahrens and O’Keefe 1972). Lange and Arhens (1982) applied the Entropy Method to investigate the critical pressure at which a complete dehydration occurs from non-porous serpentine. They estimated a critical shock pressure of 60 GPa and less for porous materials. Subsequent shock-loading experiments were performed on water-bearing minerals (serpentine and brucite; Boslough et al. 1980; Lange and Ahrens 1982, 1984; Lange et al. 1985; Tyburczy et al. 1990) and a bulk carbonaceous chondrite (Murchison; Tyburczy et al. 1986, 2001). As shown in Figure 7, their experimental results indicate that impact degassing of water from porous hydrous minerals begins for a peak impact pressure of 20 GPa. The impact pressure of the projectile depends on its velocity. In the case of planetary accretion, a minimum value of this velocity is given by the escape velocity from the colliding planet, which depends on the size of the growing planet: ve = (2GM/r)1/2. The shock pressure of the colliding body is given by: Ps = ρ 0 ve ( C 0 + S ve )

(3)

where ρ0 is the initial density, and C0 and S are shock wave parameters. According to experimental results and theoretical work on dehydration, accreting bodies start to release water when the radius of the proto-Earth reached 1500 km (porous) and 3200 km (non-porous, Figure 7), corresponding to 0.6% - 16% of its present mass. Going back to the gravitational energy required for the retention of the atmosphere, water released during a shock event of porous objects can be kept on the planetary surface even if the surface temperature is high (2000 K). Hence we conclude that impact degassing, atmosphere retention, and planetary melting all are likely to have occurred during growth of the proto-Earth.

Figure 7. Fraction of water loss as a function of shock pressure. Open symbols are serpentine (Boslough et al. 1980; Lange and Ahrens 1982, 1984; Lange et al. 1985; Tyburzcy et al. 1990) and solid squares are bulk carbonaceous chondrite (Murchison: Tyburzcy et al. 1986, 2001). Dotted line is the best fit curve for antigonite. Serpentine data are from Lange et al. (1985).

438

Marty & Yokochi

Impact erosion As the planet grows, the impact energy of the impactors increases accordingly. Consequently, the material at the surface of the accreting Earth can be evaporated and/or possibly blown away from the gravitational field. When a giant impact occurs, the atmosphere near the impact site is expelled by the expansion of vapor plumes generated at the impact site (Ahrens 1993). A giant impact also creates a strong shock wave that travels through the planetary interior, inducing a global ground motion. If the ground velocity is higher than the escape velocity of the planet, a fraction of the atmosphere can be ejected into space (Chen and Ahrens 1997). The Lunar cataclysm, indeed, is likely to have resulted in this impact erosion, which is unfortunately difficult to quantify given the numerous sources of uncertainties, like the state of atmosphere at the time of impact and the size and chemistry of the impactor. Interestingly, such atmospheric loss could have fractionated isotopically noble gases in order to match the unique noble gas isotopic composition of the terrestrial atmosphere (Pepin 1997).

Post-accretional role of a proto-atmosphere in the Early Earth’s evolution Water is likely to have exerted some control on the early evolution of our planet. Water vapor has strong adsorption bands for infrared radiation, thus the existence of an H2O atmosphere plays an important role in preventing the impact energy released at the surface of a growing planet from escaping into interplanetary space (blanketing effect). Abe and Matsui (1986) suggested that the temperature of the Earth started to increase due to addition of water vapor from impact degassing when the Earth reached a radius of 1200 km. In their model, the surface temperature reached the liquidus temperature when r = 2400 km, assuming a mass accretion rate of 50 m.y. for 1 terrestrial mass. The blanketing effect of water induced the formation of the magma ocean, which allowed interaction between gas and accreting solid phases. Moreover, the chemistry of the proto-atmosphere could have buffered the oxygen fugacity of the magma ocean, leading to core formation. As the surface temperature reaches the melting temperature of silicates in a primary H2-He atmosphere (see Figure 6), silicates should undergo reduction by H2 and supply H2O to the atmosphere (Sasaki 1990). As a result, the H2O abundance at the bottom of the atmosphere is determined by chemical equilibrium at the planetary surface, and the H2O/H2 ratio of is a function of the oxygen fugacity (fO2). Assuming that metallic iron existed at the planetary surface to form the planetary core, i.e., fO2 = IW or slightly below, the H2O/H2 ratio would have been ~0.1 (Sasaki 1990), much higher than that of solar nebula (6 × 10−4). The timing of both gravitational capture of solar nebula, magma ocean, start of impact degassing and core formation is critical. Without the blanketing effect of water, melting of Earth would have occurred when the radius reached 3800 km. Thus it is unlikely that metal reduction due to silicate melting in the presence of nebular H2 occurred before accumulation of impact-degassed water (r > 1200 km). Sasaki (1990) also examined the chemical effect of addition of impact degassed water to a solar (H2-rich) atmosphere. As the proto-atmosphere becomes increasingly enriched in oxygen due to the supply of impact-degassed water, the temperature increases due to blanketing effect and the H2O/H2 ratio approaches the value imposed by IW or IM buffers, without any reaction with silicates. According to this model, it is unlikely that the solar hydrogen atmosphere contributed significantly to the core formation. An interesting possibility to overcome this perspective is dissolution of hydrogen molecules into a magma ocean, but experimental data to test this hypothesis is presently lacking. Once a magma ocean is formed, volatile elements will be partitioned between the atmosphere and molten silicates. The quantity of volatile elements dissolved in the silicates at the surface of the Earth can be estimated using the Henry’s law, provided that the atmospheric pressure is known. According to Fricker and Reynolds (1968) (see also Dixon et al. 1995, for recent water solubility data), the amount of water dissolved in silicate melt is given by the

Water in the Early Earth

439

following equation as a function of pressure: Xsol(%) = 2.08 × 10−4 P0.54 (Pa)

(4)

When the surface begins to melt, the pressure at the Earth surface is about 105 hPa (100 atm), and about 1% H2O dissolves in silicate melts. In the model of Abe and Matsui (1986), this water dissolution mechanism buffers the surface temperature of the accreting Earth: the decrease in water vapor pressure decreases the temperature of the Earth so that the Earth solidifies and no more water can be dissolved. Further impact degassing then increases the partial pressure of water in the atmosphere, and the proto-Earth melts again, allowing further water dissolution, etc. Due to this feedback effect, the model calculation of Matsui and Abe (1986) indicated that an impact induced proto-atmosphere should contain the same quantity of water vapor as the one present in the oceans, suggesting that such a mechanism was indeed operational. The concentration of neon and helium in the deep, presumably less-degassed, mantle can be estimated from He and Ne isotopes produced in the mantle by U and Th (Yokochi and Marty 2004). Assuming that Ne was introduced in the mantle following vapor-molten silicate equilibrium (Henry’s coefficient = 11 × 10−12 mol·g−1·hPa−1 for Ne) and was not outgassed during crystallization (closed system conditions), the partial pressure of the proto-atmosphere is estimated to be 0.019 × 10−12 mol/g 22Ne, requiring 0.0017 hPa of Ne partial pressure. As the fraction of 22Ne in the solar nebula is 13 ppmv, the partial pressure of solar gas is estimated to be 7.5 × 103 hPa (~7.5 bar), much lower (7.5%) than the total pressure estimated from the impact degassing model above. Interestingly, this fraction is in agreement with the ≈10% solar nebula in mantle hydrogen derived from the isotopic constraints advocated in the preceding section.

A summary of volatile delivery processes and of their inherent uncertainties Volatile elements were supplied to the Earth’s surface by a succession of processes that can be summarized as follows. First, the accretion of planetesimals occurred probably in the presence of the solar nebula. The existence, or not, of a solar-like proto-atmosphere depended on the timing of Earth’s growth relative to that of the dissipation of the solar nebula, which constitutes the first source of uncertainty. The contribution of this proto-atmosphere appears anyway minor, according to the H-C-N isotope constraints presented above. During planetary collisions, dehydration of the impactors took place when the Earth reached a critical size. By this time, the size of the planet had grown large enough to retain the degassed volatiles, provided that the temperature of the Earth’s surface was not too high, which is a function of the accretion rate. This constitutes the second main source of uncertainty. Depending on the chemistry (third source of uncertainty), the proto-atmosphere could have had limited the radiative heat loss to the space. In case of a low accretion rate, this atmosphere could have been retained so that a blanketing effect would have caused melting of the Earth’s surface. If the accretion rate was high, the gravitational energy might have been high enough to heat up and melt the planet. Impacts of large planetesimals could have blown off this protoatmosphere (to an unknown extent, fourth source of uncertainty). The final amount of volatile elements accreted by the Earth depended on the cumulative mass of late accreting material, its degassing state prior to terrestrial accretion (fifth source of uncertainty) and atmospheric escape processes that were likely to have taken place during the first tens to hundreds of years (sixth source of uncertainty). In the absence of direct insights into this dark age, indirect studies such as geochemical constraints mentioned above, a better understanding of the solid Earth evolution, and comparison with young extrasolar system are essential.

Cooling of the primordial Earth As the accretion ended, radiative heat loss exceeded the supply of gravitational energy through accretion, and the proto-Earth started to cool. For the condensation of liquid water, i.e., formation of oceans, two conditions are required. First, the temperature need to be higher than that of the triple point (273.15 K at 103 hPa) and lower than that of critical point (647.1 K)

440

Marty & Yokochi

so that liquid water can exist. Second, the abundance of water molecules needs to exceed the saturation vapor pressure at the given temperatures. Provided that an impact-derived water atmosphere existed, the second condition is likely to be met so that the ocean started to form as the temperature of the Earth surface became lower than that of the critical point. Once water condensation started, cooling of Earth might have proceeded rapidly because the thick blanket of water vapor no longer existed. As the critical temperature is much lower than that of the solidus of silicates, the solidification of the Earth’s surface occurred before the formation of the oceans. The distribution of water in the solid Earth would depend on whether the crystallization proceeded from the bottom to the surface or inversely. In the former case, water was partitioned between three (crystal, melt and gas) phases whereas dissolved water would have been partitioned into mantle-forming minerals in the latter case. Both cases have been modeled by Porcelli and Pepin (2000).

THE WATER CYCLE IN THE HADEAN The origin of the solar system is dated by U-Pb at 4.566±0.002 Ga ago (Allègre et al. 1995) or 4.5672±0.0006 Ga ago (Amelin et al. 2002). N-body simulations of planetary growth predict that bodies of the size of Moon and Mars could have formed within a few million years after start of solar system condensation (ASSC) The Earth experienced a major metal-silicate fractionation at 11-30 Ma ASSC, that could have been linked to the Moon forming event. There is geochemical evidence for crystallization of the lunar magma ocean 4.527±0.010 Ga ago (Kleine et al. 2005), that is, ~40 Ma ASSC. The age of crystallization of the terrestrial magma ocean is less constrained. but could have been 70 Ma ASSC according to coupled extinct radioactivities of 129I (T1/2 = 16 m.y.) and 244Pu (T1/2 = 82 m.y.) producing 129Xe and 131−136 Xe, respectively (Kunz et al. 1998). 142Nd excesses in Isua supracrustal belt rocks (age: 3.7-3.8 Ga ago) have been attributed to a major fractionation of the 142Sm-142Nd couple the former decaying to the latter with T1/2 = 103 m.y. (Boyet et al. 2003; Caro et al. 2003). This episode might have corresponded to global mantle differentiation 4.460±0.115 GA ago, that is, 100±115 Ma ASSC. Although imprecise, the last time constraint indicates that large scale differentiation took place within the first 1-2 hundreds of million years. Recent high precision analysis of 142Nd in primitive meteorites have led Boyet and Carlson (2005) to suggest global differentiation of the Earth 4.53 Ga ago, that is, 30-40 Ma ASSC. Hence there is evidence that the Earth differentiated very early within a few tens of Ma ASSC. What was the fate of water at this epoch is highly speculative. It is not clear how much water could have survived the Moon forming event, although some models predict that oceans predated this event and could have survived (Genda and Abe 2005). Later on, water was further contributed to this initial inventory by asteroids and interplanetary dust. Part of volatile elements contributed during these late episodes could have been introduced in the proto-mantle through “cold” subduction (Tolstikhin and Hofmann 2005), that is, burial of late accreting matter into the mantle under relatively mild thermal condition that would have prevented volatile degassing. Nowadays, recycling of surface material at subduction zones entrains extensive degassing of subducted volatile elements, and Tolstikhin and Hofmann (2005) have argued that this was not the case in the early Hadean. This assumption is highly speculative because there is independent evidence that the Hadean mantle was hotter than today’s (i.e., higher radioactive isotope content), so that it is also possible that HSE, but not water and other volatile elements, were subducted. Mantle or crustal rocks dating back from this epoch are lacking, and we must rely on indirect records for mantle evolution in the Hadean (4.5-4.0 Ga ago). The earliest terrestrial samples we have in hands are detrital zircons found in Archean sedimentary rock from Western Australia. Ion probe analysis have revealed U-Pb ages in excess of 4.0 Ga (Froude et al. 1983)

Water in the Early Earth

441

up to 4.3-4.4 Ga (Mojzsis et al. 2001; Wilde et al. 2001). Zircons form in granitic magmas, and granites originate mainly from melting of pre-existing crustal material under hydrous condition. Oxygen isotopic ratios (Mojzsis et al. 2001; Wilde et al. 2001) are consistent with zircon protoliths having interacted with liquid water, thus implying mild temperatures at the Earth’s surface, although other processes than water-rock interaction might have determined the oxygen isotopic ratios. Independent studies on trace elements in Hadean zircons also point to the occurrence of wet melting conditions (Watson and Harrison 2005) and of widespread occurrence of continental crust at that time (Harrison et al. 2005). These results demonstrate without ambiguity that water was present early in Earth’s history, but they do not allow to settle constraints on the water content of the Hadean Earth. As a guide for the behavior of water, we can tentatively rely on noble gases for which there exists several isotopes produced by extinct radioactivities. Water and noble gases are both incompatible during partial melting and so should behave similarly during mantle differentiation. However, the behavior of water during silicate-vapor interaction is likely to be different from that of noble gases, because water degasses only at low pressure but noble gases are extremely insoluble in molten silicates. Thus this approach has its limitations and can be used only as a guideline for possible events having affected mantle water. 129 Xe has been produced by 129I with a half life of 16 million years. The amount of terrestrial iodine, represented by its only stable isotope 127I, has been estimated at ~ 10 ppb for the whole silicate (mantle+surface) Earth (Deruelle et al. 1992). In contrast, the amount of radiogenic 129 Xe present in the atmosphere, crust and mantle represents only about 1% of the amount that has been produced by 129I. The corresponding time interval to decrease this amount is of the order of 100 m.y. (Allègre et al. 1995). Three possibilities could account for this depletion: (i) delayed accretion of the Earth, in which case Earth-forming material did not condense (retain) xenon until 110 Ma ASSC, this is highly unlikely as terrestrial embryos were formed within a few million years at maximum; (ii) there exists a hidden reservoir of xenon in the Earth in which 129 Xe is sequestrated with other Xe isotopes; (iii) xenon was degassed early in the atmosphere and subsequently lost into space from an open atmosphere. With respect to (ii), the possibility of a hidden reservoir of Xe is often advocated as it could also account for the missing xenon problem (Ozima and Podosek 1999). So far, this reservoir has not been identified but this remains an attractive working hypothesis. Possibility (iii) that the terrestrial mantle has been extensively degassed during the first 110 m.y. is fully consistent with early mantle differentiation recorded in other extinct radioactivity systems and also with the occurrence of a magma ocean in the first ~100 m.y. as the cause of mantle early differentiation (Caro et al. 2005). Thus 129Xe, and possibly iodine, could have been transferred to the Earth’s surface and atmosphere during these magma ocean episodes. Models for evolution of early terrestrial atmosphere all advocate loss of atmospheric elements into space in order to account for the isotopic fractionation of noble gases ( Hunten et al. 1987; Sasaki and Nakasawa 1988; Pepin 1991; Tolstikhin and Marty 1998; Dauphas 2003). The 129Xe budget allows one to predict that such escape lasted for at least ~100 m.y., possibly more depending on the escape rate. This escape might have been linked with the occurrence of enhanced UV solar flux relative to present, that are otherwise observed to last ~108 yr for other solar systems (Feigelson and Montmerle 1999). During this early differentiation event, water might have also been transferred to the Earth’s surface and subsequently lost into space, thus decreasing drastically its original content in the mantle. As we have argued in the Isotope Section, there is no evidence in the D/H ratio of the mantle or of the oceans that water was lost during such escape so that a large fraction water might have been trapped in the mantle due to its high solubility.

Xenon has also recorded in its isotopic composition independent evidence for drastic fractionation in the Hadean. In addition to the decay of 129I producing 129Xe with T1/2 = 16 m.y., the fission of the now extinct 244Pu isotope (T1/2 = 82 m.y.) has produced 131−136Xe

442

Marty & Yokochi

isotopes with a specific yield, and the fission of still extant 238U (T1/2 = 4.45 b.y.) also produces 131−136 Xe isotopes, but with a different yield. Hence three decay schemes are recorded in the isotopic composition of Xe, allowing one to address first, the building era of the Earth (100 Ma ASSC), second, part of the Hadean (in practice, 600 Ma ASSC) and third, the geological time from the Archean to present (Yokochi and Marty 2005). Although the isotopic composition of atmospheric Xe has been drastically altered by atmospheric processes, that of mantle Xe stills retains a record of mantle differentiation events. Identifying the respective contributions of 244Pu and 238U in Xe-depleted mantle-derived samples is a hard task that will not be detailed here and only the main results will be presented. Because noble gases are extremely sensitive to alteration and in-situ production of radiogenic isotopes, a precise picture of the Xe composition of the mantle is mainly derived from the analysis of present-day mantlederived samples. Both mid-ocean ridge (Kunz et al. 1998) and mantle plume-derived (Poreda and Farley 1992; Trieloff et al. 2000; Trieloff and Kunz 2005; Yokochi and Marty 2005) samples have recorded a similar fractionation in the 129I-129Xe and 244Pu-131−136Xe systems. In these systems, the main possibility for fractionation between parents and daughter isotopes is degassing, which for the mantle can only occur either during large-scale magma episodes such as magma oceans, or later on when the mantle solidified, through convection allowing decompression melting. The major episodes of mantle and atmosphere evolution recorded in Xe isotopes and that could be applied to the case of water are (Yokochi and Marty 2005): (i) both mantle (MOR and plume) sources behaved similarly in term of convection during the Hadean, thus requiring isolation of these reservoirs to have occurred at a later time; (ii) extensive loss of volatile elements from the proto-mantle occurred during the first ~100 m.y. due to terrestrial accretion and magma ocean episodes; and (iii) extensive mantle convection and degassing, as recorded in the 244Pu-131−136Xe system, continued in the Hadean for 400700 Ma ASSC. As a result, the mantle lost at least one order of magnitude fission Xe from 244 Pu, that is, isotopes that have been produced with a half life of 82 m.y., after the first tens of million years. This indicates that the Hadean mantle was very active and could have also lost a significant amount of water. This approach also allows one to predict that the atmosphere was open for 100-200 Ma ASSC, in agreement with previous atmospheric evolution models (Pepin 1991; Tolstikhin and Marty 1998), and therefore allowing loss of water into space. The fate of water could have been however different from that of xenon. Water is readily recycled into the mantle whereas noble gases are not (xenon may constitute an exception since part of mantle Xe could be atmospheric in origin, but such recycling was probably limited as indicated by the detection of extinct radioactivity isotope anomalies in the mantle that were not erased by atmospheric Xe recycling). Hence, after atmospheric closure, water might have cycled between the surface and the mantle at a rate greater than at present. These models are quite firmly constrained because the abundances of the parent elements (I and Pu) can be well estimated from cosmochemical consideration and from their predicted geochemical behavior in magmatic system. The Hadean terminated possibly with a peak of bombardment that was recorded on the Lunar surface by an increase in the cratering rate 3.8-4.0 Ga ago and is known as the late heavy bombardment (LHB). Recently, Gomes et al. (2005) have proposed that the LHB was due to a rapid migration of giant planets after a long period of stability that triggered a massive delivery of planetesimals from outside the orbits of the planets towards the inner solar system. The integrated mass that impacted the Moon during the LHB is 1.6 × 1021 g (Hartmann et al. 2000). Assuming a maximum scaling factor of 30 between Earth and Moon, the amount of LHB material that could have fallen onto Earth, 4.8 × 1022 g is unlikely to have had a profound impact on the terrestrial water budget. Indeed, assuming a maximum water content of 50% as some, or all, of these objects could have been cometary, the amount of water delivered during the LHB could have contributed at most 1.6% to oceans and 1% of less to the bulk Earth inventory of water.

Water in the Early Earth

443

WATER CONTENT OF THE ARCHEAN MANTLE FROM THE COMPOSITION OF KOMATIITES Komatiites are ultramafic magmas that erupted mainly in the Archean and Early Proterozoic and which are very rich in MgO (up to 28%). If dry, this type of magma would erupt at ≥1500 °C and must have come from extremely hot mantle, 200-300 °C hotter than temperatures inferred for the Archean mantle (Herzberg 1992; Nisbet et al. 1993). Thus it has been proposed that komatiitic magmas originated from Archean plume heads. However, there is the possibility that komatiites originated from an hydrated mantle source (Allègre 1982; Viljoen and Viljoen 1969). Because the presence of water decreases the temperature liquidus of ultramafic magmas (e.g., Gasparik 1993; Inoue 1994), a water-rich mantle source would have allowed the production of high MgO lavas with more “normal” mantle temperature (e.g., Parman et al. 1997). Intuitively, it may seem logical to expect a greater radioactive element heat production and greater gravitational energy release in the Hadean and Archean and therefore a hotter mantle than today. However, there are a number of poorly constrained feedback effects that could profoundly affect the Archean convection regime. For example, because the Rayleigh number is proportional to the temperature gradient, a Archean mantle hotter than the present-day one would also dissipate heat much more rapidly due to enhanced convection. Thus, estimating the water content of the mantle in the past has far reaching implications for understanding the thermal state of Earth through time Komatiites erupted early in the Archean tend to be depleted in Al and present fractionated trace element patterns that are consistent with melting in the presence of garnet, whereas komatiites from the late Archean and Proterozoic do not show evidence that the parent magmas coexisted with this phase. Together with determination of phase equilibria of dry peridotite mantle assemblage, this observation is consistent, in the case of dry melting, with a varying melting regime due to secular cooling of the mantle (Nisbet et al. 1993). However, determination of phase equilibria for hydrous peridotite show that both compositions can be achieved by melting in a hydrous mantle at different pressures (Asahara et al. 1998; Inoue 1994), with liquidus temperatures being 100-200 °C lower than in the case of dry peridotite for 1-2% H2O. There is a number of field and mineralogical observations that have been regarded as supporting hydrous generation of komatiitic magmas (e.g., Grove et al. 1997), but other authors challenged these interpretations (e.g., Arndt et al. 1998). The least disputed evidence for a wet scenario is the occurrence of igneous hydroxyl-amphibole in ultramafic flows and sills from Boston Creek, Ontario (Abitibi formation), showing that at least some komatiitic magmas contained water (Stone et al. 1997). These authors estimated that the initial melts (50% partial melting) contained as much as 2% H2O. There have been several reports of water content in glass inclusions encapsulated in olivine or chromite from Belingwe komatiite, Zimbabwe, with contrasted conclusions. McDonough and Danyushevsky (1995) proposed that the host komatiitic melt contained 0.2% H2O, not drastically different from modern basalts, whereas Shimizu et al. (1997) estimated a higher water amount of 1.8% for 11.6% MgO. Arndt et al. (1998) noted that observations of high water contents in some of the komatiite mineral phases did not necessarily imply a mantle origin for water, since assimilation of hydrated basalts, a process that might have been favored by the inferred high temperature of komatiites, could have led to the same result, thus explaining the range of water contents observed for a single unit. The debate has been somewhat focused on the interpretation of the spinifex structure of komatiites. Spinifex, one of the most spectacular magmatic textures, refers to large olivine plates that form decimeter-sized dentritic crystals which are oriented perpendicular to the roof of the komatiite unit. In laboratory conditions, such texture can only be reproduced under high cooling rates (typically 50 °C/h) whereas cooling rates of spinifex-textured komatiite flows are typically ≤1 °C/h (Donaldson 1982). In order to resolve this discrepancy, Grove et al. (1997) proposed from field observation that komatiites are hydrous intrusive rocks. Under these P, T

444

Marty & Yokochi

conditions, quench crystallization in liquids that became supercooled due to intensive volatile loss, or rapid growth of skeletal crystals in hydrous magmas due to the depolymerizing effect of water both could have resulted in the formation of the spinifex texture. The debate is not ended because the intrusive nature of komatiites is not proven according to other authors (e.g., Arndt et al. 1998) and recent experiments tend to show that spinifex texture can be reproduced in the laboratory by slow cooling under a thermal gradient (7-35 °C/cm) that may well apply to cooling conditions of komatiitic flows (Faure et al. 2006).

CONCLUSIONS It is probable that several cosmochemical reservoirs contributed water to the nascent Earth. Both planetary modeling and D/H isotope systematics suggest that a fraction of water was trapped early in the Earth, possibly from iron-silicate reactions in the presence of a solartype atmosphere that favored oxidation of solar nebula H2. The present day fraction of such water component remaining in the mantle may be ~10%. The major fraction of water present in the Earth is probably from sources sharing similar D/H, 13C/12C and 15N/14N ratios with asteroidal matter. Given current D/H measurements of comets, the contribution by these objects appears limited to a few percents at best. This view is possibly biased by the very limited data base concerning the isotopic compositions of extraterrestrial reservoirs including the solar nebula, the parent bodies of meteorites, and the comets. Water might have been ubiquitously present on planetesimals and planetary embryos, as evidenced by aqueous alteration that affected equally parent bodies of ordinary chondrites and of carbonaceous chondrites. According to recent simulations of solar system dynamics, water could have been contributed by wet embryos towards the end of the Earth’s growth (Morbidelli et al. 2000). These simulations predict the contribution of bodies originating from the solar system snowline or beyond, that would have the adequate D/H ratio. There is also a growing number of evidence that interplanetary dust, otherwise observed in abundance in young extrasolar planetary systems, could have contributed a significant fraction of terrestrial water. Interplanetary dust particles are likely to be rich in volatile elements, as suggested by the chemical link between micrometeorites and carbonaceous chondrites on one hand, and the high volatile element content observed by ion probe analysis on stratospheric IDPs on another hand. The delivery of volatile elements to terrestrial planets is a natural consequence of IDPs being generated beyond 1 AU and spiraling towards the Sun through the orbits of the inner planets. Yet, there is no relevant dynamical simulation of IDP generation and motion in the early solar system that would allow one to test of this possibility. We also lack at present information that would permit a better characterization of the chemistry and isotopic composition of Kuiper belt objects, but the situation is likely to evolve rapidly with space missions aimed to better characterize bodies beyond the inferred snowline of the solar system. Noble gas models for the Hadean are extended to the case of water, H2O was extensively exchanged between the surface and the mantle by whole mantle convection during at least 400-700 Ma ASSC. Because the atmosphere became closed at maximum 200 Ma ASSC, no water loss into space was expected after this date and the total amount of water in the terrestrial system should be approximated by the sum of present-day inventories of water in the oceans and in the mantle.

ACKNOWLEDGMENTS Reika Yokochi acknowledges support at the University of Illinois at Chicago from the Camille and Henry Dreyfus Postdoctoral Program in Environmental Chemistry. Bernard Marty wishes to thank Neil Sturchio and Nicolas Dauphas for welcoming him at the University of

Water in the Early Earth

445

Illinois at Chicago and the University of Chicago, respectively, during the preparation of this work, and for discussions. François Faure and François Robert provided advices on komatiites, and the water content and isotopic composition of chondrites, respectively. We thank Steve Mojzsis, an anonymous reviewer and Hans Keppler for very constructive comments. This is CRPG contribution 1810.

REFERENCES Abe Y, Matsui T (1986) Early evolution of the Earth: accretion, atmosphere formation, and thermal history. J Geophys Res, Proc 17th Lunar Planet Sci 91:E291-E302 Abe Y, Ohtani E, Okuchi T, Righter K, Drake MJ (2000) Water in the early Earth. In: Origin of the Earth and Moon. Righter K, Canup RM (eds), Univ. Arizona Press, p 413-433 Ahrens TJ (1993) Impact Erosion of Terrestrial Planetary-Atmospheres. Ann Rev Earth Planet Sci 21:525-555 Ahrens TJ, O’Keefe JD (1972) Shock melting and vaporization of lunar rocks and minerals. Moon 4:214-249 Aléon J, Engrand C, Robert F, Chaussidon M (2001) Clues on the origin of interplanetary dust particles from the isotopic study of their hydrogen-bearing phases. Geochim Cosmochim Acta 65:4399-4412 Aléon J, Robert F (2004) Interstellar chemistry recorded by nitrogen isotopes in Solar System organic matter. Icarus 167:424-430 Aléon J, Robert F, Chaussidon M, Marty B (2003) Nitrogen isotopic composition of macromolecular organic matter in interplanetary dust particles. Geochim Cosmochim Acta 67:3773-3783 Allègre CJ (1982) Genesis of Archaean komatiites in a wet ultramafic subducted plate. In: Komatiites. Arndt NT, Nisbet EG (eds), Allen & Unwin, p 495-500 Allègre CJ, Manhès G, Göpel C (1995) The age of the Earth. Geochim Cosmochim Acta 59:1445-1457 Allègre CJ, Manhes G, Lewin E (2001) Chemical composition of the Earth and the volatility control on planetary genetics. Earth Planet. Sci Lett 185:49-69 Amelin Y, Krot AN, Hutcheon ID, Ulyanov AA (2002) Lead isotopic ages of chondrules and calcium-aluminumrich inclusions. Science 297:1678-1683 Anders E (1978) Most stony meteorites come from the asteroid belt. In: Asteroids: An Exploration Assessment. Morrisson D, Wells DC (eds), NASA, p 57-75 Anders E (1989) Abiotic organic matter from comets and asteroids. Nature 342:255-257 Arndt N, Ginibre C, Chauvel C, Albarède F, Cheadle M, Herzberg C, Jenner G, Lahaye Y (1998) Were komatiites wet? Geology 26:739-742 Arpigny C, Jehin E, Manfroid J, Hutseméker D, Schulz R, Stüwe JA, Zucconi JM, Ilyin I (2003) Anomalous nitrogen isotope ratio in comets. Science 301:1522-1524 Asahara Y, Ohtani E, Suzuki A (1998) Melting relations of hydrous and dry mantle compositions and the genesis of komatiites. Geophys Res Lett 25:2201-2204 Becker RH, Clayton RN, Galimov EM, Lammer H, Marty B, Pepin RO, Wieler R (2003) Isotopic signatures of volatiles in terrestrial planets - Working group report. Space Sci Rev 106:377-410 Bell DR, Rossman GR (1992) Water in the Earth’s mantle: the role of anhydrous minerals. Science 255:13911397 Benkert JP, Baur H, Signer P, Wieler R (1993) Helium, neon, and argon form the solar wind and solar energetic particles in lunar ilmenites and pyroxenes. J Geophys Res 98:13,147-13,162 Bland PA, Smith TB, Jull AJT, Berry FJ, Bewan AWR, Cloudt S, Pillinger CT (1996) The flux of meteorites to the Earth over the last 50,000 years. Mon Not R Astron Soc 283:551-565 Boato G (1954) The Isotopic composition of hydrogen and carbon in the carbonaceous chondrites. Geochim Cosmochim Acta 6:209-220 Bockelée-Morvan D, Gautier D, Lis DC, Young K, Keene J, Phillips T, Owen T, Crovisier J, Goldsmith PF, Bergin EA, Despois D, Wootten A (1998) Deuterated water in comet C/1996 B2 (Hyakutake) and its implications for the origin of comets. Icarus 193:147-162 Bodenheimer P, Lin DNC (2002) Implications of extra solar planets for understanding planet formation. Ann Rev Earth Planet Sci 30:113-148 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: Implications for the distribution of water in the Earth´s mantle. Earth Planet Sci Lett 182:209 - 221 Boslough MB, Weldon RJ, Ahrens TJ (1980) Impact-induced water loss from serpentine, nontronite and kernite. Proc Lunar Planet Sci Conf 11:2145-2158 Boss AP (1998) Temperatures in protoplanetary disks. Ann Rev Earth Planet Sci 26:53-80 Boyet M, Blichert-Toft J, Rosing M, Storey M, Telouk P, Albarède F (2003) Nd-142 evidence for early Earth differentiation. Earth Planet Sci Lett 214:427-442

446

Marty & Yokochi

Boyet M, Carlson RW (2005) Nd-142 evidence for early (>4.53 Ga) global differentiation of the silicate Earth. Science 309:576-581 Brandon AD, Walker RJ, Puchtel IS (2006) Platinum–osmium isotope evolution of the Earth’s mantle: Constraints from chondrites and Os-rich alloys. Geochim Cosmochim Acta 70:2093-2103 Caro G, Bourdon B, Birck JL Moorbath S (2003) Sm-146-Nd-142 evidence from Isua metamorphosed sediments for early differentiation of the Earth’s mantle. Nature 423:428-432 Caro G, Bourdon B, Wood BJ, Corgne A (2005) Trace-element fractionation in Hadean mantle generated by melt segregation from a magma ocean. Nature 436:246-249. Carr MH, Wänke H (1992) Earth and Mars - Water Inventories as Clues to Accretional Histories. Icarus 98: 61-71 Chambers JE, Wetherill CW (1998) Making the terrestrial planets: N-body integrations of planetary embryos in three dimensions. Icarus 136:304-327 Chen GQ, Ahrens TJ (1997) Erosion of terrestrial planet atmosphere by surface motion after a large impact. Phys Earth Planet Int 100:21-26 Chou CL (1978) Fractionation of siderophile elements in the Earth’s upper mantle. Proc Lunar Planet Sci Conf 9th:219-230 Chyba CF (1990) Impact delivery and erosion of planetary oceans in the inner solar system. Nature 343:129133 Chyba CF, Sagan C (1992) Endogenous production, exogenous delivery and impact-shock synthesis of organic molecules: an inventory for the origins of life. Nature 355:125-132 Chyba CF (1991) Terrestrial mantle siderophiles and the Lunar impact record. Icarus 92:217-233 Clayton RN (2003) Oxygen isotopes in the solar system. Space Sci Rev 106:19-32 Culler TS, Becker TA, Muller RA, Renne PR (2000) Lunar impact history from Ar-40/Ar-39 dating of glass spherules. Science 287:1785-1788 Dauphas N (2003) The dual origin of the terrestrial atmosphere. Icarus 165: 326-339. Dauphas N, Marty B (2002) Inference on the nature and the mass of Earth’s late veneer from noble metals and gases. J Geophys Res-Planets 107:12-1 - 12-7 Dauphas N, Marty B, Reisberg L (2002a) Inference on terrestrial genesis from molybdenum isotope systematics. Geophys Res Lett, 29:8-1 - 8-3 Dauphas N, Marty B, Reisberg L (2002b) Molybdenum evidence for inherited planetary scale isotope heterogeneity of the protosolar nebula. Astrophys J 565:640-644 Dauphas N, Reisberg L, Marty B (2002c) An alternative explanation for the distribution of highly siderophile elements in the Earth. Geochem J 36:409-419 Dauphas N, Robert F, Marty (2000) The late asteroidal and cometary bombardment of Earth as recorded in water deuterium to protium ratio. Icarus 148:508-512 Deloule E (2002) D/H ratio ion probe measurements on magmatic minerals from Martian meteorites: Implications for degassing of the Martian mantle. Lunar Planet Sci XXXIII:1607 Deloule E, Albarède F, Sheppard SMF (1991) Hydrogen isotope heterogeneities in the mantle from ion probe analysis of amphiboles from ultramafic rocks. Earth Planet Sci Lett 105:543-553 Deloule E, Robert F (1995) Interstellar water in meteorites. Geochim Cosmochim Acta 59:4695-4706 Delsemme AH (1988) The chemistry of comets. Phil Trans R Soc London A325:509-523 Deruelle B, Dreibus G, Jambon A (1992) Iodine Abundances in oceanic basalts - Implications for Earth dynamics. Earth Planet Sci Lett 108:217-227 Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of water and carbon dioxide solubilities in mid ocean ridge basaltic liquids .1. Calibration and solubility models. J Petrol 36:1607-1631 Donaldson CH (1982) Spinifex-textured komatiites: a review of textures, compositions and layering. In: Komatiites. Arndt NT, Nisbet EG (eds) Allen & Unwin, p 213-244 Drake MJ (2005) Origin of water in the terrestrial planets. Meteorit Planet Sci 40:519-527 Drouart A, Dubrulle B, Gautier D, Robert F (1999) Structure and transport in the solar nebula from constraints on deuterium enrichment and giant planets formation. Icarus 140:129-155 Engrand C, Deloule E, Robert F, Maurette M, Kurat G (1999a) Extraterrestrial water in micrometeorites and cosmic spherules from Antarctica: an ion microprobe study. Meteorit Planet Sci 34:773-786 Engrand C, McKeegan KD, Leshin LA (1999b) Oxygen isotopic compositions of individual minerals in Antarctic micrometeorites: Further links to carbonaceous chondrites. Geochim Cosmochim Acta 63: 2623-2636 Faure F, Arndt N, Libourel G (2006) Formation of spinifex texture in komatiites: an experimental study. J Petrol 47:1591-1610 Feigelson ED, Montmerle T (1999) High-energy processes in young stellar objects. Ann Rev Astron Astrophys 37:363-408 Fricker PE, Reynolds RT (1968) Development of atmosphere of Venus. Icarus 9:221-234 Froude DO, Ireland TR, Kinny PD, Williams IS, Compston W, Williams IR, Myers JS (1983. Ion microprobe identification of 4,100-4,200 Myr-old terrestrial zircons. Nature 304:616-618

Water in the Early Earth

447

Gaffey MJ, Burbine TH, Binzel RP (1993) Asteroid spectroscopy: Progress and perspectives. Meteoritics 28: 161-187 Gasparik T (1993) The role of volatiles in the transition zone. J Geophys Res-Solid Earth 98:4287-4299 Geiss J, Gloecker G (1998) Abundances of deuterium and helium in the protosolar cloud. Space Sci Rev 84: 239-250 Geiss J, Reeves H (1972) Cosmic and solar system abundances of deuterium and helium-3. Astron Astrophys 18:126-132 Genda H, Abe Y (2005) Enhanced atmospheric loss on protoplanets at the giant impact phase in the presence of oceans. Nature 433:842-844 Gomes R, Levison HF, Tsiganis K, Morbidelli A (2005) Origin of the cataclysmic Late Heavy Bombardment period of the terrestrial planets. Nature 435:466-469 Gradie JC, Chapman CR, Tedesco EF (1989) Distribution of taxonomic classes and the compositional structure of the asteroid belt. In: Asteroids II. Binzel RP (ed), Univ. Arizona Press, p 316-335 Grieve RAF, Shoemaker EM (1994) The record of past impact on Earth. In: Hazards due to comets and asteroids. Gerhels T (ed) Univ. Arizona Pressn, p 417-462 Grove TL, de Wit MJ, Ashwal LD (1997) Komatiites from the Komati Type Section, Barbeton, South Africa. In: Greenstone belts. de Wit MJ, Aswhal LD (eds) Oxford Sci Publ, p 422-437 Habing HJ et al. (1999) Disappearance of stellar debris disks around main-sequence stars after 400 million years. Nature 401:456-458 Halliday AN, Wänke H, Birck JL, Clayton RN (2001) The accretion, composition and early differentiation of Mars. Space Sci Rev 96:1-34 Harrison TM, Blichert-Toft J, Muller W, Albarède F, Holden P, Mojzsis SJ (2005. Heterogeneous Hadean hafnium: Evidence of continental crust at 4.4 to 4.5 Ga. Science 310:1947-1950 Hartmann WK, Ryder G, Dones L Grinspoon D (2000) The time-dependent intense bombardment of the primordial Earth/Moon system. In: Origin of the Earth and Moon. Canup RM, Righter K (eds) Univ. Arizona Press, p 493-512 Hashizume K, Chaussidon M, Marty B, Robert F (2000) Solar wind record on the moon: Deciphering presolar from planetary nitrogen. Science 290:1142-1145 Hashizume K, Chaussidon M, Marty B, Terada K (2004) Protosolar carbon isotopic composition: Implications for the origin of meteoritic organics. Astrophys J 600:480-484 Hashizume K, Marty B, Wieler R (2002) Analyses of nitrogen and argon in single lunar grains: towards a quantification of the asteroidal contribution to planetary surfaces. Earth Planet Sci Lett 202:201-216 Hayashi C, Nakazawa K, Mizuno H (1979) Earth’s melting due to the blanketing effect of the primordial dense atmosphere. Earth Planet Sci Lett 43:22-28 Herzberg C (1992) Depth and degree of melting of komatiites. J Geophys Res-Solid Earth 97:4521-4540 Hillgren V, Gessmann CK, Li J (2002) An experimental perspective on the light element in Earth’s core. In: Origin of the Earth and Moon. Righter K, Canup RM (eds) Univ. Arizona Press, p 245-264 Hiyagon H, Ozima M, Marty B, Zashu S, Sakai H (1992) Noble gases in submarine glasses from mid-oceanic ridges and Loihi Seamount - constraints on the early history of the Earth. Geochim Cosmochim Acta 56: 1301-1316 Honda M, McDougall I, Patterson DB, Doulgeris A, Clague DA (1991) Possible solar noble-gas component in Hawaiian basalts. Nature 349: 149-151 Huang XG, Xu YS, Karato SI (2005) Water content in the transition zone from electrical conductivity of wadsleyite and ringwoodite. Nature 434:746 - 749 Humayun M, Clayton RN (1995) Potassium isotope cosmochemistry - Genetic implications of volatile element depletion. Geochim Cosmochim Acta 59:2131-2148 Hunten DM, Donahue TM (1976) Hydrogen loss from terrestrial planets. Ann Rev Earth Planet Sci 4:265-292 Hunten DM, Pepin RO, Walker JCB (1987) Mass fractionation in hydrodynamic escape. Icarus 69:532-549 Inoue T (1994) Effect of water on melting phase-relations and melt composition in the system Mg2SiO4MgSiO3-H2O up to 15 Gpa. Phys Earth Planet Int 85:237-263 Jambon A, Zimmermann JL (1990) Water in oceanic basalts - evidence for dehydration of recycled crust. Earth Planet Sci Lett 101:323-331 Javoy M (1997) The major volatile elements of the Earth: Their origin, behavior, and fate. Geophys Res Lett 24:177-180 Javoy M, Pineau F, Delorme H (1986) Carbon and nitrogen isotopes in the mantle. Chem Geol 57:41-62 Jessberger EK, Christoforidis A, Kissel J (1988) Aspects of the major element composition of Halley’s dust. Nature 332:691-695 Jewitt DC, Matthews HE, Owen T, Meier R (1997) Measurements of C-12/C-13, N-14/N-15, and S-32/S-34 ratios in comet Hale-Bopp (C/1995 O1). Science 278:90-93 Keays RR, Ganapathy R, Laul JC, Anders E, Herzog GF, Jeffery PM (1970) Trace element and radioactivity in lunar rocks: Implications for meteorite infall, solar-wind flux, and formation conditions of the Moon. Science 167:490-493

448

Marty & Yokochi

Keppler H, Bolfan-Casanova N (2006) Thermodynamics of water solubility and partitioning. Rev Mineral Geochem 62:193-230 Kerridge F (1985) Carbon, hydrogen and nitrogen in carbonaceous chondrites: abundances and isotopic compositions in bulk samples. Geochim Cosmochim Acta 49:1707-1714 Kleine T, Münker C, Metzger K, Palme H (2002) Rapid accretion and early core formation in asteroids and the terrestrial planets from Hf-W chronometry. Nature 418:952-955 Kleine T, Palme H, Mezger M, Halliday AN (2005) Hf-W chronometry of lunar metals and the age and early differentiation of the Moon. Science 310:1671-1674 Kortenkamp SJ, Dermott SF, Fogle D, Grogan K (2001) Sources and orbital evolution of interplanetary dust accreted by Earth. In: Accretion of Extraterrestrial Matter Throughout Earth’s History. Peuker-Ehrenbrink B, Schmitz BP (ed) Kluwer Academic, p 13-30 Kunz J, Staudacher T, Allègre CJ (1998) Plutonium-fission xenon found in the Earth’s mantle. Science 280: 877-880 Kurat G, Koeberl C, Presper T, Branstätter F, Maurette M (1994) Petrology and geochemistry of Antarctic micrometeorites. Geochim Cosmochim Acta 58:3879-3904 Kyte FT, Wasson JT (1986) Accretion rate of extraterrestrial matter: iridium deposited 33 to 67 million years ago. Science 232:1225-1229 Lange MA, Ahrens TJ (1984) FeO and H2O and the homogeneous accretion of the Earth. Earth Planet Sci Lett 71:111-119 Lange MA, Arhens TJ (1982) The evolution of impact-generated atmosphere. Icarus 51:96-120 Lange MA, Lambert P, Ahrens TJ (1985) Shock effects on hydrous minerals and implications for carbonaceous meteorites. Geochim Cosmochim Acta 49:1715-1726 Lécluse C, Robert F (1994) Hydrogen isotope-exchange reaction-rates - origin of water in the inner solarsystem. Geochim Cosmochim Acta 58:2927-2939 Lecuyer C, Gillet P, Robert F (1998) The hydrogen isotope composition of seawater and the global water cycle. Chem Geol 145:249-261 Love S, Brownlee DE (1993) A direct measurement of the terrestrial mass accretion rate of cosmic dust. Science 262:550-553 Marty B (1989) Neon and xenon isotopes in MORB: implications for the earth-atmosphere evolution. Earth Planet Sci Lett 94:45-56 Marty B, Dauphas N (2003) The nitrogen record of crust-mantle interaction from Archean to Present. Earth Planet Sci Lett 206:397-410 Marty B, Robert P, Zimmermann JL (2005) Nitrogen and noble gases in micrometeorites. Meteorit Planet Sci 40n:881-894 Marty B, Tolstikhin IN, Kamensky IL, Nivin V, Balaganskaya E, Zimmermann JL (1998) Plume-derived rare gases in 380 Ma carbonatites from the Kola region (Russia) and the argon isotopic composition in the deep mantle. Earth Planet Sci Lett 164:179-192 Marty B, Zimmermann JL (1999) Volatiles (He, C, N, Ar) in mid-ocean ridge basalts: Assesment of shallowlevel fractionation and characterization of source composition. Geochim Cosmochim Acta 63:3619-3633 Maurette M, Duprat J, Engrand C, Gounelle M, Kurat G, Matrajt G, Toppani A (2000) Accretion of neon, organics, CO2, nitrogen and water from large interplanetary dust particles on the early Earth. Planet Space Sci 48:1117-1137 McDonough WF, Danyushevsky WF (1995) Water and sulfur contents of melt inclusions from Archean komatiites. Eos, Trans Amer Geophys Union 76:266 Meier R, Owen TC, Matthews HE, Jewitt DC, Bockelée-Morvan D, Biver N, Crovisier J, Gautier D (1998) A determination of the DHO/H2O ratio in comet C/1995 O1 (Hale-Bopp). Science 279:842-844 Meisel T, Walker RJ, Irving AJ, Lorand JP (2001) Osmium isotopic compositions of mantle xenoliths: A global perspective. Geochim Cosmochim Acta 65:1311-1323 Messenger S (2000) Identification of molecular-cloud material in interplanetary dust particles. Nature 404: 968-971 Millar TJ, Bennett A, Herbst E (1989) Deuterium fractionation in dense interstellar clouds. Astrophys J Lett 340:906-920 Mojzsis SJ, Harrison TM, Pidgeon RT (2001) Oxygen-isotope evidence from ancient zircons for liquid water at the Earth’s surface 4,300 Myr ago. Nature 409:178-181 Morbidelli A (2002) Modern integrations of solar system dynamics. Ann Rev Earth Planet Sci 30:89-112 Morbidelli A, Chambers J, Lunine JI, Petit JM, Robert F, Valsecchi GB, Cyr K (2000) Source regions and timescales for the delivery of water to the Earth. Meteorit Planet Sci 35:1309-1320 Moreira M, Kunz J, Allègre CJ (1998) Rare gas systematics in Popping Rock: Isotopic and elemental compositions in the upper mantle. Science 279:1178-1181 Morgan JW (1986) Ultramafic xenoliths - Clues to Earth’s late accretionary history. J Geophys Res-Solid Earth 91:2375-2387

Water in the Early Earth

449

Nisbet EG, Cheadle MJ, Arndt NT, Bickle MJ (1993) Constraining the potential temperature of the Archean mantle - a review of the evidence from komatiites. Lithos 30:291-307 Notesco G, Bar-Nun A, Owen T (2003) Gas trapping in water ice at very low deposition rates and implications for comets. Icarus 162:183-189 Ohtani E (2005) Water in the mantle. Elements 1:25-30 Okamoto YK et al (2004) An early extrasolar planetary system revealed by planetesimal belts in beta Pictoris. Nature 431:660-663 Owen T, Bar-nun A (1996) Comets, Meteorites and Atmospheres. Earth Moon Planet 72:425-432 Owen T, Mahaffy PR, Niemann HB, Atreya S, Wong M (2001) Protosolar nitrogen. Astrophys J Lett 553:L77L79 Owen T, Maillard JP, Debergh C, Lutz BL (1988) Deuterium on Mars - the abundance of HDO and the value of D/H. Science 240:1767-1770 Ozima M, Podosek FA (1999) Formation age of Earth from I-129/I-127 and Pu-244/U-238 systematics and the missing Xe. J Geophys Res-Solid Earth 104:25493-25499 Ozima M, Podosek FA (2002) Noble Gas Geochemistry. Cambridge University Press Palme H (2001) Chemical and isotopic heterogeneity in protosolar matter. Philos Trans R Soc London Ser A 359:2061-2075 Parman SW, Dann JC, Grove TL, de Witt MJ (1997) Emplacement conditions of komatiite magmas from the 3.49 Ga Komati Formation, Barberton Greenstone Belt, South Africa. Earth Planet Sci Lett 150:303-323 Pavlov AA, Pavlov AK, Kasting JF (1999) Irradiated interplanetary dust particles as a possible solution for the deuterium/hydrogen paradox of Earth’s oceans. J Geophys Res-Planets 104:30725-30728 Pepin RO (1991) On the origin and early evolution of terrestrial planetary atmospheres and meteoritic volatiles. Icarus 92:1-79 Pepin RO (1997) Evolution of Earth’s noble gases: Consequences of assuming hydrodynamic loss driven by giant impact. Icarus 126:148-156 Podosek FA, Cassen P (1994) Theoretical, observational, and isotopic estimates of the lifetime of the solar nebula. Meteorit Planet Sci 29:6-25 Porcelli D, Pepin RO (2000) Rare gas constraints on Early Earth history. In: Origin of the Earth and Moon. Canup RM, Righter K (eds) University of Arizona Press, p 435-458 Poreda RJ, Farley KA (1992) Rare-gases in Samoan xenoliths. Earth Planet Sci Lett 113:129-144 Righter K (2003) Metal-silicate partitioning of siderophile elements and core formation in the early Earth. Ann Rev Earth Planet Sci 31:135-174 Robert F (1978). Teneur en eau et en deuterium des chondrites. Unpubl. Thesis, University of Paris 7 Robert F (2003) The D/H ratio in chondrites. Space Sci Rev 106:87-101 Robert F, Epstein S (1982) The concentration and isotopic composition of hydrogen, carbon and nitrogen in carbonaceous chondrites. Geochim Cosmochim Acta 46:81-95 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 Sasaki S (1990) The primary solar-type atmosphere surrounding the accreting Earth: H2O-induced high surface temperature. In: Origin of the Earth. Newson HE, Jones JH (eds) Oxford University Press, p 195-209 Sasaki S, Nakasawa K (1988) Origin of isotopic fractionation of terrestrial Xe, hydrodynamic fractionation during escape of the primordial H2-He atmosphere. Earth Planet Sci Lett 89:323-334 Sears DWG, Dodd TR (1988) Overview and classification of meteorites. In: Meteorites and the Solar System. Kerridge JF, Matthews MS (eds) University of Arizona Press, Tucson, p 3-3. Shimizu N, Komiya T, Maruyama S. Hirose K (1997) Water content of melt inclusion in Cr-spinel of 2.7 Ga komatiite from Belingwe Greenstone belt, Zimbabwe. Eos, Trans Amer Geophys Union 78:750 Shukolyukov A, Lugmair GW (2000) On the Mn-53 heterogeneity in the early solar system. Space Sci Rev 92: 225-236 Stone WE, Deloule E, Larson MS, Lesher CM (1997 Evidence for hydrous high-MgO melts in the Precambrian. Geology 25:143-146 Tolstikhin IN, Hofmann AW (2005) Early crust on top of the Earth’s core. Phys Earth Planet Int 148:109-130 Tolstikhin IN, Marty B (1998) The evolution of terrestrial volatiles: a view from helium, neon, argon and nitrogen isotope modelling. Chem Geol 147:27-52 Trieloff M, Kunz J (2005) Isotope systematics of noble gases in the Earth’s mantle: possible sources of primordial isotopes and implications for mantle structure. Phys Earth Planet Int 148:13-38 Trieloff M, Kunz J, Clague DA, Harrison CM, Allègre CJ (2000) The nature of pristine noble gases in mantle plumes. Science 288:1036-1038 Trull T (1994) Influx and age constraints on the recycled cosmic dust explanation for high 3He/4He ratios at hotspot volcano’s. In: Noble gas geochemistry and cosmochemistry. Matsuda JI (ed) Terra Scientific Publishing Co, p 77-88 Tyburczy JA, Frisch B, Ahrens TJ (1986) Shock-induced volatile loss from a carbonaceous chondrite Implications for planetary accretion. Earth Planet Sci Lett 80:201-207

450

Marty & Yokochi

Tyburczy JA, Krishnamurthy RV, Epstein S, Ahrens TJ (1990) Impact-induced devolatilization and hydrogen isotopic fractionation of serpentine - Implications for planetary accretion. Earth Planet Sci Lett 98:245260 Tyburczy JA, Xu XM, Ahrens TJ, Epstein S (2001) Shock-induced devolatilization and isotopic fractionation of H and C from Murchison meteorite: some implications for planetary accretion. Earth Planet Sci Lett 192:23-30 Viljoen MJ, Viljoen RP (1969) The geology and geochemistry of the lower ultramafic unti of the Onverwacht Group and a proposed new class of igneous rocks. Spec Pub Geol Soc South Africa 2:55-85 Wänke H, Dreibus G, Jagoutz E (1984) Mantle chemistry and accretion history of the Earth. In: Archean Geochemistry. Kroner A (ed) Springer, p 1-24 Watson EB, Harrison TM (2005) Zircon thermometer reveals minimum melting conditions on earliest Earth. Science 308:841-844 Weidenschilling SJ (1977) Distribution of Mass Planetary System and Solar Nebula. Astrophys Space Sci 51: 153-158 Weidenschilling SJ, Spaute D, Davies DR, Marzari F, Ohtsuki K (1997) Accretional evolution of a planetary swarm. Icarus 128:429-438 Wetherill GW (1990) Formation of the Earth. Ann Rev Earth Planet Sci 18:205-256 Wetherill GW (1992) An alternative model for the formation of the asteroids. Icarus 100:307-325 Wiens RC, Bochsler P, Burnett DS, Wimmer-Schweingruber RF (2004) Solar and solar wind isotopic compositions. Earth Planet Sci Lett 222:697-712 Wilde SA, Valley JW, Peck WH, Graham CM (2001)Evidence from detrital zircons for the existence of continental crust and oceans on the Earth 4.4 Gyr ago. Nature 409: 175-178 Yin Q, Jacobsen SB, Yamashita K, Telouk P, Blichert-Toft J, Albarède, F (2002) A short timescale for terrestrial planets from Hf-W chronometry. Nature 418:949-952 Yokochi R, Marty B (2004) A determination of the neon isotopic composition of the deep mantle. Earth Planet Sci Lett 225:77-88 Yokochi R, Marty B (2005) Geochemical constraints on mantle dynamics in the Hadean. Earth Planet Sci Lett 238:17-30

19

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 451-473, 2006 Copyright © Mineralogical Society of America

Water and Geodynamics Klaus Regenauer-Lieb Geophysics & Geodynamics Johannes Gutenberg-Universität Mainz D-55099 Mainz, Germany e-mail: [email protected]

INTRODUCTION Hydrogen is the most abundant element (Fig. 1) in the galaxy and our solar system (Lodders 2003). Therefore it is not astonishing that hydrogen is a key player in the geodynamic evolution of planets. Its fate in the early Earth, after condensation of the solar nebula, the accretion of our planet and hydrogen reprocessing through early asteroidal and cometary bombardment (Dauphas et al. 2000) and segregation of a proto-Earth into iron core and silicate mantle is described elsewhere in this volume (Marty and Yokochi 2006). This chapter concerns itself with the geodynamics of the modern Earth, where nowadays hydrogen occupies only a very small mass fraction (less than 0.1 wt%). Although its abundance is thus drastically diminished, it is still believed to have a profound influence on the geodynamics of the present planet. The reason why hydrogen can have a profound influence on geodynamics, even when it is diluted to ppm-levels, relies on its fundamental influence on both rock mechanics (Kohlstedt 2006) and the melting relationships (Kohn 2006). However, both the micromechanical mechanism and the large scale implications are not yet fully understood. In this review geodynamic processes which may be triggered by a rock mechanical hydrogen threshold value are reviewed in addition to resulting styles of planetary convection controlled by melt and rock mechanics. This review will conclude with a more speculative discussion of Solar System Abundances 100000 10000 1000

H He

100

O

C

10

Ne

N

1

Si Mg

0.1

Na

Al

0.01 0.001 0.0001

F Li

B

0.00001 0.000001

Be

0.0000001 H He Li Be B

C

N

O

F Ne Na Mg Al Si

Figure 1. Abundance of elements in the solar system in numbers of atoms per 106 atoms of silicon (Lodders 2003). 1529-6466/06/0062-0019$05.00

DOI: 10.2138/rmg.2006.62.19

452

Regenauer-Lieb

the possible implications for planetary evolution from an early Earth to the present day and highlight areas of future work. Following current practice in the literature the term “water” will be used interchangeably for hydrogen. Recent breakthroughs in computational geodynamics are throwing a new light on the possible evolution of planetary tectonics. Simulations have shown that the outer boundary layer of the Earth, the lithosphere, has a major impact on solid-state convection in the Earth’s interior. The key factor, the lithospheric strength, is in turn significantly influenced by the Earth’s water content. Thermo-mechanical calculations based on laboratory flow laws suggest that for the onset of lithosphere failure the critical water content must be above 200 ppm H/ 106Si, thus defining a threshold strength of the lithosphere. Accordingly, a terrestrial planetary system develops into three different styles of mantle convection: (1) below a critical strength mobile lid convection occurs, in which the lithosphere actively participates in planetary convection; this is the present plate tectonic mode on the Earth; (2) for a stiffer lithosphere heat escapes through episodic lid convection by which the whole lithosphere, resurfaces after long periods of inactivity through catastrophic flushing events as inferred on Venus 300-600 Ma ago; (3) at yet lower water content or for a more rapidly cooling planet a fully stagnant lid develops in which the surface of the planet does no longer participate in the convection.

WATER IN THE LITHOSPHERE The success of plate tectonics relies on the assumption of essentially rigid plates. On the other hand, plates must apparently also be broken and slide efficiently past or above each other. This happens at plate boundaries of which about 85% are only narrow zones ~1-60 km wide (Gordon and Stein 1992). Most of the remaining diffuse plate boundaries lie within continents and it is convenient to start with a discussion of the mechanics of the oceanic lithosphere, which is simpler and also much better constrained in terms of chemistry, water content and thermal evolution. The complex behavior of continents follows the same principles but there is an additional complexity owing to fundamentally different material behavior of quartz in deformation and melt segregation. Obviously, since plate tectonics is known to have evolved with death of old plate boundaries and nucleation of new plate boundaries, three important questions arise with a logical order: What mechanism sustains the rigidity of (oceanic) plates? What mechanism causes nucleation of new plate boundaries? What controls the evolution of these plate boundaries? A definite answer to these questions does not exist. Water is, however, known to play a fundamental role in controlling rigidity and weakness of plates. In the sections to come a summary of these dependencies is presented.

Water and the rigidity of (oceanic) plates Hydrogen is an incompatible element. The mineral melt partition coefficient of water is similar to the rare earth element cerium (Michael 1995). The process of lithosphere formation is commonly viewed as one of melt extraction which extracts REE’s in particular. However the process of lithospheric formation can also be viewed as a dehydration process, which also strengthens the lithosphere and prevents it from further recycling. The water content of the upper mantle can be estimated from analyses of primitive MORB glasses, which show a range between 0.1-0.2 wt% H2O (Danyushevsky et al. 2000; Michael 1995). Of particular interest for the mechanics of the lithosphere is the water content in the most abundant mineral in the mantle. Olivine governs the mechanical strength of the lithosphere (Tullis et al. 1991). For

Water & Geodynamics

453

the mineral olivine an estimate of 810 ± 490 ppm H/106Si (Hirth and Kohlstedt 1996) can be inferred from the water content of MORB glasses. This estimate of water content does not apply to the oceanic lithosphere but is strictly only valid for the asthenosphere. The oceanic lithosphere (the MORB source material) is likely to be left in a dry state (80-100 ppm H/106Si) down to a depth of 65 km (Hirth and Kohlstedt 1996; Karato and Jung 1998) since water is highly soluble in MORB melts. This water depleted layer is formed right at the mid-ocean ridges and is likely to be retained for several hundred millions of years because water diffusion from the asthenosphere into the lithosphere is inefficient at low lithospheric temperatures (Regenauer-Lieb and Kohl 2003). Figure 2 summarizes the conceptual view of the chemically dehydrated oceanic lithosphere and highlights its implication for lithospheric strength. The quantitative estimate of water content of the dehydrated lithosphere and the water content in Figure 2 (Hirth and Kohlstedt 1996) is for a dry oceanic lithosphere only. For the continental lithosphere there is a much longer history, and therefore there is much greater variability. Since a peak lithospheric stress of 8 kbar is difficult to achieve with normal geodynamic conditions (Kemp and Stevenson 1996) oceanic plates are commonly rigid plates. The rigidity of plates, which is a fundamental prerequisite of plate tectonics, is directly related to the dehydration of the oceanic lithosphere.

Ocean

Shear strength (MPa) 0

200

400

600

800

Brittle crust 10

Lithospheric Mantle (dry) < 100 ppm H+/Si

20

200

Weakening by brittle-ductile instabilities

400

30

Strong elastic core 40 50

600

Weakening by creep / ductile instability

800

60

Transfer of H 2O ? Convecting Upper Mantle (wet)

Olivine (main mechanical constituent)

800-900 ppm H+/Si

Figure 2. Classical yield strength envelope of the dry oceanic lithosphere (Kohlstedt et al. 1995). The normal oceanic lithosphere is dry (80 ppm H/ 106Si) because it has been formed through melt extraction. A dry oceanic lithosphere cannot break under normal geodynamic forcing. This is due to the fact that there will always be some other lithosphere which has a higher water content than dry oceanic lithosphere. The mechanism for hydrating a lithosphere is unknown but from xenolith studies we know very well that the continental mantle can be far from dry. Such a wet lithosphere has a lower strength (dotted line) and will fail through a thermo-mechanically enhanced weakening mechanism before any dry oceanic lithosphere fails. Consequently, the oceanic lithosphere has a strong elastic core and oceanic plates are essentially rigid. Continents may be deformable.

454

Regenauer-Lieb

Water and the nucleation of (new) plate boundaries In plate tectonics new plate boundaries are formed sporadically. Initiation of a new subduction system or continental breakup and formation of new oceans are typical examples. Figure 2 highlights the problem of high strength of oceanic plates. In particular subduction initiation poses a problem because as laid out in the previous section the negatively buoyant oceanic lithosphere is likely to be strong. The obvious idea is that some “damage” mechanism may be necessary to cause formation of new plate boundaries. Perhaps damage is also the key for keeping a memory of pre-existing plates. However, if the concept of lithosphere formation is one of dehydration of the lithosphere, then obviously new plate boundaries cannot form because the new plate will be always stronger (through melt extraction) than any pre-existing plate boundary. How can a new plate boundary be formed?

Co ntinuity

The classical approach to plate failure relies on phenomenological (continuum mechanical) approaches. In these approaches plate boundaries are simply parameterized by an assigned friction (Bird 2003) or by more complex Mohr-Coulomb rheology which is extrapolated down to depth (Lavier and Buck 2002; Hall et al. 2003). Plate failure is assumed if critical stresses are reached. Plate failure is then solved using classical continuum mechanics (Mohr-Coulomb plasticity). For time dependence, arbitrary weakening laws are postulated, often by referring to high fluid pressures on the frictional shear planes or by assigning a strain weakening law (Huismans and Beaumont 2002). It is obvious that such analyses are very useful descriptions of the observed deformation, but they are not designed to predict whether a new plate boundary can form or whether it cannot form. This is because the approaches do not model the physics of the underlying processes explicitly. More precisely, Figure 3 shows the shortcut that is made in classical continuum mechanics. Fluid Normally, what is solved is an equilibrium Dynamics of momentum and continuity equations. En e r The energy equation is visited without gy full coupling to the mechanics. This is often useful, since the thermal conduction process is much slower than the deformaContinuum tion, and the coupling of thermomechanics to flow laws is not considered. However, on Mechanics the contrary in fluid dynamics the energy equation is the master equation for convection. While it is always possible to postuum nt e late some “damage” law for the presence m Mo of water and hence explain plate tectonics through continuum mechanics, one only Figure 3. Classical continuum/solid mechanical deals with a small part of the whole story. approaches to the formation of new plate boundarThe heat loss of the Earth is the driving ies consider just the continuity (mass conservation) mechanism of plate tectonics and a merged and the momentum balance. The energy equation is normally visited only for solving the temperature fluid dynamics and continuum mechanical and sometimes also the buoyancy. However, in fluid approach, i.e., a fully coupled thermaldynamics the energy equation is considered as the mechanical approach including the energy key element for convection. It is normally coupled to equation, is necessary. Such approaches are momentum and continuity balance. Shear heating is now being developed in both the geodyoften neglected since fluids have a lower tendency to form shear zones than solids. Since plate tectonics innamics and the engineering communities. volves both quasi-fluid deformation (convection) and The appendix gives a short introduction failure of a solid (lithosphere) a new fully coupled for the very simple case of isothermal thermo-mechanical approach comprising the fluid deformation, which is a useful approach dynamic setup with elasticity is the currently being in engineering applications. It appears that developed.

Water & Geodynamics

455

thermodynamics has a lot to offer for geomechanics (Collins 2005). New developments for non-isothermal geodynamic modeling are briefly discussed in the next paragraphs.

Water and the evolution of plate boundaries For modeling time dependence it is necessary to consider the energy equation because this is the main time dependent equation that is solved in computational geodynamics. Energy approaches to geodynamic problems are currently the field of more theoretical approaches in computational geodynamics. Classically, these theoretical analyses solve the continuity, momentum and energy equations of a mechanical system in a fully coupled way. For incompressible approaches only the energy equation contains time dependence explicitly, and for compressible approaches the continuity equation has time dependence, which is usually incorporated into an equation of state approach. Continuum mechanics on the other hand, is classically time independent since it only solves for coupled momentum and continuity balance. Continuum mechanical approaches (like the Mohr Coulomb approach) are therefore often called “quasistatic,” since time does not appear explicitly in the equations but it is equated to a process time or a characteristic velocity. Time dependent energy fluxes are encapsulated in the theory of thermodynamics. Hence, in fully coupled geodynamics two basic thermodynamic potential functions are solved, which are the Helmholtz free energy (or its Legendre transform, the Gibbs free energy) and the dissipation function (see Appendix). Since both functions play a fundamental role in the geodynamic evolution, both potentials are currently investigated as possible roots for the evolution of plate boundaries. One mechanism relies on modification of the stored energy potential (Helmholtz free energy) of the minerals, and the other on the dissipation potential. Two different reviews give summaries of the approaches. One review is based on a fluid mechanical approach ignoring the role of elasticity. It acknowledges and summarizes known effects in the dissipation potential in terms of shear heating (Bercovici and Karato 2003). The other review is based on the continuum mechanics approach that includes elasticity (Regenauer-Lieb and Yuen 2003). Both reviews emphasize the role of the energy equation. Water has been argued to have an enhancing effect in both reviews. Accordingly, the catalytic role of water for plate tectonics on the Earth (Tozer 1985) is explained by a critical process within the Earth lithosphere. Before going into a quantitative description, the qualitative arguments for a “damage” related mechanism are presented and discussed on the basis of deformation of rocks within the lithosphere. A quick glance at the thermodynamic approach laid out in the appendix is beneficial before continuing to read.

Water and the stored energy potential Ψ Plate boundaries are without doubt engraved through some form of modification of the stored energy of the sheared material. The stored energy could appear as microstructural (micro-strain αk, see Appendix) modifications through dynamic recrystallization (Montesi and Zuber 2002) or other ductile microstructural “damage” (surface energy) giving rise to favorable alignment of a weak phase through feedback between fluid and/or rigid inclusion (e.g., void-volatile feedback (Regenauer-Lieb 1999) or some brittle form of dilatancy (Bercovici and Ricard 2003). For the “damage” mechanism water is commonly not considered, explicitly, because it is only present at ppm level in the oceanic lithosphere (Hirth and Kohlstedt 1996). Therefore it is unlikely to build a mechanically important water phase in an undisturbed mantle. Water can, however, find access and appear implicitly in the form of introducing a melt phase and wetting the grain boundaries, in which case it could boost its volumetric effect and become truly important. From outcrop evidence of exhumed mylonitic shear zones there is indeed

456

Regenauer-Lieb

ample evidence for access of melts (Dijkstra et al. 2004) at plate boundaries. This effect in turn can cause a grain size reduction via a reaction-enhanced softening mechanism (Fernandez et al. 1997). It is difficult to estimate its role in plate tectonics since a quantitative description of the mechanism does not yet exist. Finally, not all mylonitic shear zones in the mantle (Drury et al. 1991) show this mechanism. There are certainly a lot of plate boundaries that do not show any magmatic activity (Boillot and Froitzenheim 2001). So from observational evidence we would conclude that melts are an extremely potent mechanism for shear zone formation in rocks dominated by olivine rheology but certainly they are not the universal driver for plate boundary forming shear zones. A different form of grain size reduction is more widely observed. It can lead to a grain size sensitive creep mechanism which can also be enhanced by access of water (Mei and Kohlstedt 2000a) if diffusion creep is the dominant creep mechanism. Within the strong part of the oceanic lithosphere the dominant creep mechanism is thought to be dislocation creep (Kohlstedt et al. 1995). The diffusion creep regime can be reached through reducing the grain size by dislocation creep. Like the reaction enhanced softening mechanism grain size sensitive creep will act as a microstructural enhancement of the localization mechanism described in the next chapter. However, it is important to note that while grain size reduction is characteristic for shear zones in the mantle, it is not necessarily evidence for the formation mechanism of the shear zone. Grain size enhanced diffusional flow processes probably do not play a significant role at the early stages of plate failure. At significant strain through deformation enhanced access of water this may change and the (modified) stored energy terms becomes significant. At this stage plate boundaries are well and truly established and the stored energy term causes the memory of shear zones. Another microstructural mechanism causing weakening at modest water content (2001000 ppm H/Si) is probably more generally applicable. It operates at low stress and high temperatures (Katayama et al. 2004). This mechanism is due to water assisted mobility of dislocations with ensuing anisotropy through enhanced slip on the water assisted [100](001) slip system. Such hydrogen induced fabric transition are often found in nature (Michibayashi and Mainprice 2004). The mechanism can be treated, again, as an additional booster of the mechanism described in the next section. The next section simply deals with the isotropic form of hydrolytic enhancement of dislocation motion without the additional complexity of the change in microstructure. Before going into the description of dissipative processes it should be emphasized that any dissipative process, like the process of moving a dislocation through the crystalline lattice, is converting mechanical work into heat, thus it has always a positive feedback on flow localization. This is because flow stress depends on temperature and a positive increment in temperature will facilitate further deformation. However, microstructural modifications can have a negative feedback on localization at the early stage of shear zone formation. An example would be an increase in the density of dislocations which is equivalent to an increase in the internal stress (see Appendix) in the crystal, hence an increase in stored energy. This has a negative influence on the mobility of dislocation which has to move through the tangle of other dislocations. Hence, this implies an increase in the flow stress (strain hardening). Microstructural modification can thus have a negative feedback on localization. These microstructural modifications equilibrate, however, after a transient phase of deformation, where the stored energy reduces from the beginning of deformation to a situation of quasisteady state. Initially there is more than 50% of the total mechanical work converted into stored energy and very rapidly after only 8% strain the stored energy terms reduce to only 10-20% of the total work done; see Figure 9 in Chrysochoos et al. (1989), the remainder appears as heat. In continuum mechanics (e.g., metal deformation) modeling of the transient phase is considered through empirical strain hardening laws. However, for geodynamic modeling

Water & Geodynamics

457

comprehensive empirical databases do not exist, because, amongst other reasons, the time scale for suitable experiments is prohibitive. Instead geodynamicists try to come up with simple fundamental solution on the grounds of basic physics. Presently, we mostly consider experimental results in the quasi-steady state limit, where the stored energy term is small and deformational work is efficiently converted to heat. This quasi-steady state limit is described in the next chapter.

Water and the dissipated energy potential Φ For a fundamental analysis which can be used globally for all plate boundaries, without requiring a special mechanism such as demanding the existence of melts or any other microstructural transitions, we are left with a discussion of a very basic mechanism. Water has a dramatic effect in rock deformation experiments causing an enhanced mobility of dislocations (Griggs 1967; Paterson and Kekulawala 1979; Kronenberg and Tullis 1984; Hobbs 1985). This is the first potential weakening mechanism that will operate before any large strain micro-structure has been created. It is hence the most basic solution to the formation of plate boundaries. The dissipation potential Φ is defined as the double dot product of the dissipative strain rates times the stress tensor (see Appendix). An important mechanism of microstrain in minerals that leads to dissipation (shear heating) is dislocation glide (and climb) and an average strain rate measure can be described by Orowans equation (Poirier 1985).

ε = ρ b v (T ,σ, P, COH )

(1)

where ρ is the density of mobile dislocations (length of dislocation line per unit volume) and b is their Burgers vector and v their average velocity which depends on temperature, stress, pressure and the hydroxyl solubility COH. Thus, the mobility of dislocations is enhanced by water. The mechanism has first been discovered for quartz (Griggs 1967; Paterson and Kekulawala 1979; Kronenberg and Tullis 1984; Hobbs 1985) and dubbed “hydrolytic weakening.” It since has been confirmed for many minerals, in both diffusion creep (Mei and Kohlstedt 2000a) and dislocation creep regimes (Mei and Kohlstedt 2000b). Water is noted to affect the activation energy and volumes. More importantly, by inserting Orowans equation into the particular flow law, the strain rates are boosted up for the same applied stress by roughly linear scaling through the hydroxyl solubility COH (r close to unity): ⎛ − P∆VOH ⎞ ε ∝ C OH r f (σ ) = A(T ) fH 2 0 exp ⎜ ⎟ f (σ ) ⎝ RT ⎠

(2)

where A(1375 K) is 1.1 ppm [H/106Si (MPa)−k] and fH2O is the water fugacity in MPa for 10.6 cm3mol−1 volume change of olivine upon incorporation of water ∆VOH, respectively (Kohlstedt et al. 1996). The function f(σ) depends upon the flow law. Although Orowans equation and the resulting creep equations describe the effect of water empirically, it is fair to say that the exact underlying mechanism of “hydrolytic weakening” is not fully understood (Brodholt and Refson 2000). Since the dissipation potential is defined by the product of the stress × strain rates (see Appendix) water leads to an enhanced dissipation potential Φ and consequently (through maximum entropy production, see appendix) also a complimentary enhancement of the Helmholtz free energy potential Ψ. For the case of lithosphere deformation the dissipation potential and the Helmholtz free energy potential are both almost totally saturated by the brittleductile transition. This is because T/Tm < 0.3 and the highest stresses within the lithosphere are reached at the brittle-ductile transition. At such low temperatures the glide of dislocations in the direction of their Burgers vector governs crystalline plasticity (Poirier 1985).

458

Regenauer-Lieb

Coming back to the important problem of lithosphere failure, one can expect from this that the mechanically intact mantle has its strength maximum in the dislocation creep (dislocation glide and climb) regime where the laboratory results are often fitted empirically by three different empirical fits, an exponential fit, a power law and if the grain size is small enough by a linear creep law. Such flow laws can be directly applied to geodynamics but a few steps are implicit which need to be explained. Experimental fits are reported for quasi-steady state creep, i.e., the sample is put under constant stress and strain is recorded as a function of time. After a transient stage the strainrate becomes approximately constant (quasi-steady state). For this behavior a good empirical description is sought which depends on the ambient temperature. The tensorial properties like isotropy or anisotropy of the creep law are often not investigated. In geodynamics it is therefore common practice to interpret laboratory experiments in terms of an isotropic flow law where stress and strain rate tensors are assumed to have a common coaxial orientation. This is expressed in an associated flow rule (tensor quantities are bold) ε diss =

f (J 2 ) σd J2

(3)

Thereby the viscous dissipative strain rate tensor is assumed to be associated with the deviatoric stress σ d defined by σ d = σ − P, i.e., the Cauchy stress tensor σ minus the pressure P. The flow law also incorporates an effective stress measure J2 which is known as the second invariant of the deviatoric stress tensor. For a triaxial laboratory experiment this is defined as (Regenauer-Lieb and Yuen 2003) J2 ≡

3 σd : σd 2

( 4)

Three flow laws are used for empirical fit (see Table 1 for parameters) ε Peierls = A

⎡ H ⎛ J ⎞2 ⎤ σd exp ⎢ − L ⎜ 1 − 2 ⎟ ⎥ J2 ⎢⎣ RT ⎝ τo ⎠ ⎥⎦

⎛ Q + PVP ⎞ ε Power = B fHr 2 O J 2n −1 σ d exp ⎜ − P ⎟ RT ⎝ ⎠ ε Diffusion = C fHr 2 O J 2m −1

σd ⎛ Q + PVD ⎞ exp ⎜ − D ⎟ l RT g ⎝ ⎠

(5)

(6) (7)

Different fits are used for different temperatures. For the low temperature domain Equation (5) lists a fit from Goetze and Evans (1979), Equation (6) and Equation (7) are from a tandem set of papers from Mei and Kohlstedt (2000a,b). It is not necessary that only one particular microstrain mechanism is operating at a particular temperature. However, for the high temperature extreme the flow laws can clearly be attributed to a diffusion dominant mechanism (Eqn. 6). Note, that this equation is in fact another power law suggesting a slight non-linearity (power law exponent 1.1) thus a departure from the classical linear law. For low temperatures the exponential fit is interpreted as lattice controlled dislocation glide mechanism, here labeled as the Peierls mechanism. The power law fit is the most widely applicable flow law and it can be tuned with varying power law exponents to fit both the low and high stress extremes. Some modelers in geodynamics use this approach (Kameyama et al. 1999). Another approach is presented here (Ashby and Verall 1977). A composite flow law is obtained by assuming that the various creep mechanism are operating in a concurrent framework (see Appendix). The most important micro-mechanism thereby is the one that produces the fastest

Water & Geodynamics

459

Table 1. Hydrolytic weakening for Peierls mechanism, power law and diffusion in olivine.1 Variable R Adry Awet HLdry HLwet T0dry T0wet B QP VP r n C QD VD D L m

Value [units] −1

Explanation −1

8.3144 [J mol K ] 5.7 × 10+11 [s−1] 1.2 × 10+12 [s−1] 536 [kJ/mol] 498 [kJ/mol] 8.5 10+9 [Pa] 9.1 × 10+9 [Pa] 4.6× 103 [µm3MPa−3.7s−1] 470 [kJ/mol] 2.0 × 10−5 [m3/mol] 0.98 3 4.8 × 104 [μm3MPa−2.1s−1] 295 [kJ/mol] 2.0 × 10−5 [m3/mol] [µm] 3 1.1

Universal gas constant Prefactor Peierls stress Prefactor Peierls stress Activation enthalpy (Peierls dry) Activation enthalpy (Peierls wet) Peierls stress (dry) Peierls stress (wet) Prefactor Power Law Activation energy Power Activation Volume Power fH2O exponent Power Power Law exponent Prefactor Diffusion Activation Energy Diffusion Activation Volume Diffusion. Grain size Grain size exponent Diffusion Law exponent

1 Data for Peierls mechanism from Evans and Goetze (1979), Goetze and Evans (1979). Data for the Power Law and Diffusion Creep from Mei and Kohlstedt (2000a, 2000b)

strain rates, however, the other mechanism can also contribute to the deformation. This approach is easily extended to allow brittle processes in terms of a plastic yield mechanism. As shown in the appendix it can be expanded to also include elastic strain rates. Brittle (plastic) deformation is activated only if the stress is high enough. This setup allows a simple deformation mechanism map for geodynamic modeling. ε diss = ε Peierls + ε Power + ε Diffusion + ε Plastic

(8)

ε total = ε el + ε diss

(9)

The dominant role for reducing lithosphere strength through addition of water is clearly expressed by the fH2Or factors in the diffusion and the power laws. For the Peierls mechanism the same hydrolytic weakening is expected as for power and diffusion laws but it has never been quantified in scaled experiments. Based on theoretical analyses and experiments for spinel and quartz a strong sensitivity of the Peierls mechanism on the water content was observed (Evans 1984; Donlon et al. 1998). The enhancement of creep by hydrogen captured in the Orowan Equation (1) may also be related to pipe diffusion along dislocations in dislocation glide. In a numerical simulation, pipe diffusion was found to be assisted by active pumping of water through the dislocations (Heggie 1992). Using the Peierls law cited in Table 1 as an upper and lower bound one obtains a conservative estimate of the potential effect of water at low temperatures of the brittle-ductile transition. In Table 1 the stronger of the two laws is labeled dry and the weaker is labeled wet. The resulting strain-rate map of Equation (8) is shown in Figure 4. It illustrates the same dry conditions as discussed by Kohlstedt et al. (1995) and shown in Figure 2 but explores the full strain-rate space from 10−20-10−6 s−1. There are three differences: One difference is that Figure 2 shows the shear strength only for one particular strain rate of 10−15 s−1, while Figure 4 shows the entire geologically relevant strain rate space and beyond. The second difference is that the brittle law is simplified in Figure 4 into a single brittle failure line. The third difference

460

Regenauer-Lieb Strain Rate (s-1) -5

log10 İ diss

-8

-5

-11 -8 -14 -11

-17

-14

1000 St r e

750

Vd ss

-17 500

a)

( MP

250 0

1100 1000 800 900

-20

700 500 600 T (K) 300 400 rature e p m e T

Figure 4. Deformation mechanism map of temperature (depth) versus shear stress for the same case as in Figure 2. However, in this plot the full additive strain-rate Equation (8) is used. The strain-rates cover the full geologically relevant space and beyond. The black area in the plot defines the elastic core of the plate which is only broken if the stress can reach the depth dependent yield limit. Since the water content in the flow laws scales the flow stress through the fH20 factor a plate with a dry rheology will be considerably stronger (have a much larger elastic core) than a wet plate.

is that the peak stress in the lithosphere is limited by the Peierls mechanism which is not considered in Figure 2. Figure 4 suggests that the effect of water on the Peierls mechanism deserves further attention in order to improve our understanding for lithosphere rheology. However, the following discussion would also hold if more data on the Peierls mechanism or more information on the brittle-ductile transition became available. The results would simply be corrected to a higher or lower stress for failure of the lithosphere. Alternatively, one could also incorporate the presence of some other (semi-brittle) mechanism not considered here, clipping the highest strength at the brittle-ductile transition. This point will be discussed further in the conclusion to the chapter on subduction initiation. Figure 4 shows basically two different domains. One in which, under any geologically relevant strain-rate, there exists an area shown in black, where the material has not reached its flow/yield stress. Consequently this area is deforming elastically before reaching the yield stress. It is called the elastic core of the lithosphere. This domain is classically modeled by continuum mechanics neglecting the energy equation (Fig. 3). Elastic deformation can be significant, like in the case of the flexure of the lithosphere under the Hawaiian Emperor seamount load (Walcott 1970). Such significant deformation processes need to be considered in any geodynamic modeling attempt for the lithosphere. There is another domain at higher temperatures, where the elastic core appears to be missing. The material can flow without significant applied shear stress. This domain is classically modeled by fluid dynamics where the energy equation is the master equation (Fig. 3). The fH2O factor in the flow laws has the fundamental effect in shifting solid versus fluid dynamic domains in the deformation mechanism map, by reducing the peak stress in the elastic core, thus it brings the lithosphere closer to critical conditions for solid mechanical failure.

Water & Geodynamics

461

Solid versus fluid dynamic modeling setups With this generic, rheological model given, it is possible to investigate the problem of the critical water content inside the lithosphere necessary to promote the existence of plate boundaries. The inverse problem, what is the water condition to prevent lithosphere failure can also be investigated. Subduction initiation and the continental breakup problems can be used as fundamental strength-meters for the lithosphere. Since continental breakup involves a quartz dominated crust the behavior is more complicated and we discuss only the subduction initiation problem. A hybrid solid-fluid mechanical setup is used in which the lithosphere is modeled by an extended Maxwell body (Fig. 5). For short term loading the lithosphere behaves elastically but for long term tectonics it also has the potential to creep by a combination of dominant creep mechanisms as described in the previous chapter. The mode of failure is selected by the dominant strain rate in the model and not prescribed by a working hypothesis. The rheology below the bottom of the lithosphere is assumed to be fully viscous with a quasi-elastic response from the differential buoyancy of lithosphere and mantle. The coupled solid-fluid model setup shown in Figure 5 has been extended by an active mantle flow solver in order to investigate the mantle back flow induced by the subduction problem (Morra and Regenauer-Lieb 2006). Solid mechanical failure is concerned with the stress loading, deformation and failure of solid materials and structures, while fluid mechanics classically deals with material flow without localized failure. Rice (1993) notes that all material (i.e., fluids and solids) can support normal forces and suggests following definition of solids: “We call a material solid rather than fluid if it can support a substantial shearing force over the time scale of some natural process or technological application of interest.” Obviously, the importance of elastic strain (stored energy potential) is crucial for shearing strength and the appearance of solid-like behavior, but

Extended Maxwell body

Rheology inside the lithosphere

Lithosphere

H

H Elastic H Plastic H Peierls H Power H Diffusion

elastic

plastic

non-linear viscous dash-pots

Viscous dashpot foundation (new)

classical Winkler foundation

+ Lithosphere

Rheology at the bottom of the lithosphere

Mantle

F

§ dz dx · K¨ ¸ © dt dt ¹

+

F

' U gz

Figure 5. The rheology inside the lithosphere considers the full strain rate space in a deformation mechanism map (Fig. 4) derived from the additive strain-rate decomposition. The rheology at the bottom of the lithosphere is simplified by only considering viscous half space and intrinsic buoyancy loads.

462

Regenauer-Lieb

an additional constraint may be that all kind of localized failure (dissipative energy potential) must be dealt within the framework of solid-mechanics. Conversely, wherever distributed creep mechanisms are dominant we deal with the classical framework of fluid-dynamics. Therefore in an extended Maxwell rheology as shown in Figure 5, the elastic spring and, if present the plastic body, is equivalent to a solid property. The creep “dash pots” can define a weak solid or fluid according to their strain rates. A weak form of the definition of a solid (without elasticity) sometimes is used in fluid dynamics. This is done in order to simplify the basic equations. Note that this approach neglects significant elastic deformation before flow such as the flexure of the oceanic lithosphere and also neglects significant redistribution of stress inside the solid which are necessary before actual failure conditions are reached. In the fluid dynamic approach the plastic body without elasticity is used to mimic solid-like behavior of the lithosphere within fluid dynamic codes (e.g., Trompert and Hansen 1998; Solomatov and Moresi 2000; Tackley 2000). The dash-pot is the classical rheological body describing the mechanical properties of a fluid body. Depending on the degree of non-linearity of viscosity (e.g., power law with a high power law exponent) the dash-pot is transforming to a pseudo-plastic rheology. The nonlinear composite rheology defines a transitional property, which can be neither assigned fully solid or fluid, if both deformation by the dash-pot and the elastic or plastic body are of equal magnitude.

Application to subduction initiation Subduction initiation and subsequent subduction evolution have traditionally been approached by two different working hypotheses. One hypothesis is based on assuming continuum-mechanical failure (Fig. 6A), supposing a critical stress-state to exist that ruptures the

A) Solid Mechanical Failure

B) Fluid Mechanical Failure

of the Lithosphere on fluid foundation (e.g. Cloetingh et al. 82, Branlund et al. 2000)

forced Rayleigh-Taylor Instability (e.g. Faccenna et al. 99, Becker et al. 99)

Sediment Load

weak zone formation

C) Convection + Weak Fault Subduction on existing weak fault (e.g. Casey and Dewey 84, Toth and Gurnis 98)

no weak zone

D) Solid + Fluid Mechanical Failure Coupling of A) and B) generates weak fault (R.-L. et al. 2001,R.-L. and Kohl 2003)

weak zone assumed

weak zone formation

Figure 6. Four different models of planetary failure. Only the right panel solves the energy equation and is thermodynamically self-consistent. The coupled solid-fluid failure of the lithosphere shown in D) is controlled by a critical yield stress of the lithosphere (Solomatov 2004) which is given by the water content (Eqns. 2, 6 and 7).

Water & Geodynamics

463

entire lithosphere strength envelope (Wortel and Vlaar 1988; Cloetingh et al. 1989; Shemenda 1992; Giunchi et al. 1996; Hassani et al. 1997; Branlund et al. 2000). This stress state is thought to be reached independent of time through quasistatic tectonic loading. The other hypothesis approximates the long-term behavior of the lithosphere by a high viscosity, slowly creeping fluid, acknowledging the effect of deformation time for conditioning of forced Rayleigh-Taylor style (Fig. 6B) fluid dynamic instabilities (Gurnis 1992; Fullsack 1995; Larsen et al. 1996; Becker et al. 1999; Ellis et al. 1999; Faccenna et al. 1999; van Hunen et al. 2000; Doin and Henry 2001; Hall and Kincaid 2001; Winder and Peacock 2001). While the latter approach is easily reaching the condition for instabilitiy for a hypothetical material, provided that sufficient deformation time has elapsed, its inherent flaw lies in ignoring the physics of solids tentatively described in the former approach. As shown in Figure 4 it will never reach conditions for instability of the elastic core of the lithosphere if laboratory derived creep laws are considered. Thus in order to postulate a fluid dynamic instability it is necessary to assume that the laboratory derived flow laws are not applicable for geological time scales. The yield strength of the lithosphere gives indeed very tight controls on modes of solid-mechanical rupture allowing failure only for a very young thermal age of the lithosphere (Kemp and Stevenson 1996). It has also been shown that fluid mechanical models fail when applying Earth-like parameters (Toth and Gurnis 1998). Only when assuming convection with a pre-existing weak fault is subduction possible (Fig. 6C). Numerical experiments have been performed for estimating the potential of coupled solid/ fluid-dynamical failure modes of the lithosphere under a sedimentary load (Figs. 6D and 7) (Regenauer-Lieb et al. 2001). These experiments predict (Fig. 7) that elasticity in the strongest part of the lithosphere is crucial for modulating two fundamentally different failure modes. In the dry case scenario the strongest core of the lithosphere always remains elastic, i.e., it never yields. This elastic core couples surface deformation following the sedimentary load into fluid mechanical instabilities at depth. However there is a significant time lag of about 50 Ma between surface instabilities and instabilities at depth. This time lag ensures that there is no positive feedback through coupling of sediment loads with the negative buoyancy load of the Rayleigh-Taylor instabilities at the bottom of the lithosphere. The lithosphere cannot be broken. In the wet case there also remains a strong elastic part as a communicator of surface to deep deformation, however this strong part has the opposite effect than in the dry case. Instead of creating a subhorizontal layer, thus protecting the entire lithosphere from failure, a shear zone grows from the bottom of the elastic layer upwards (Fig. 7 wet rheology at 16 Ma). This shear zone is induced by a secondary tensile load which slowly builds from below the elastic core during continuing sediment loading because the bottom of the lithosphere yields and flows away. This phenomenon is not observed in the case of the dry lithosphere, where insufficient flow in the lower lithosphere does not shift the core upwards. This shift preconditions failure of the lithosphere and allows the generation of a solid mechanical so called “plastic hinge line” (Johnson et al. 1974) at about 16 Ma after initial loading. Such a hinge line ruptures the lithosphere and leads to downward rotation the future subducted lithosphere much like on a real lubricated hinge. It is interesting to note that the hinge line does not initially propagate from the surface of the lithosphere but from the deep ductile part. This shows the important role of elastic transmission of surface loads to deep ductile deformation. The process has already been identified in the early 90’s (Kusznir 1982) and the term “visco-elastic stress amplification” has been coined to describe this phenomenon which does not occur in fluid dynamics without elasticity. Another important observation is that the shear zone nucleates in the regime where power law creep dominates. Coming back to the poor knowledge of the maximum strength of the lithosphere at the brittle-ductile transition: The observation that the shear zone grows from the bottom upwards implies that it is likely that visco-elastic stress amplification has the potential to rupture the strong part, once the shear zone is initiated in the deeper power law regime.

464

Regenauer-Lieb

Figure 7. Model for the initiation of subduction: Right is the wet > 200 ppm H/106 Si and left the dry < 200 ppm H/106Si case. Both cases show the importance of Rayleigh-Taylor like instabilities that develop at the bottom of the lithosphere after about 50 Ma sediment loading have elapsed. The significant difference is, however, that the Rayleigh-Taylor instability is coupled to a solid-mechanical instability in the wet case, while it is decoupled in the dry case. The crucial role for synchronizing solids and fluid instabilities appears to be attributed to the presence of water in the power law creep regime of the lithosphere. In addition elasticity plays a crucial role in acting as a communicator between solid and fluid parts of the lithosphere. See Plates 3 and 4 for the colorized version of this figure.

Water & Geodynamics

465

While ultimately, the strength of the plate and the load to break the plate is thus controlled by hydrolytic weaking at conditions of the brittle-ductile transition, the mode switching from failure (wet) to non-failure (dry) appears to be chiefly controlled by hydrolytic weakening in the power law regime around a value of approximately 200 ppm H/106 Si. While the incorporation of better brittle-ductile rheology is necessary for future studies, the fundamental role of water as a toggle switch between failure or no-failure modes will remain unchanged.

WATER IN THE CONVECTING MANTLE Recent breakthroughs in large scale computing of planetary solid-state convection (Schubert 2001) can now be applied to assess basic modes of convection in a cooling and degassing planet. The key driver for developing highly sophisticated numerical models of convection of the roughly 3000 km thick outer spherical shell of the Earth was the question: “Why, so far, have we not yet discovered other planets that have plate tectonics?” In the early days of plate tectonics it was accepted that for planetary convection, comprising thousands of kilometers, it is an unnecessary complexity to describe exactly the behavior of something so small (often less than 100 km) as the cool outer shell of the planet, the so-called lithosphere. We now know that it is exactly this outer boundary layer that has decisive control on styles of planetary convection. Highly non-linear fluid dynamics models (Bercovici 1993; Lenardic et al. 1995; Solomatov 1995; Moresi and Solomatov 1998; Tackley 1998; Trompert and Hansen 1998) have identified that the yield stress of the outer layer of planets has a first order impact on triggering different modes of heat exchange of a planet with its surroundings. A cooling and degassing/dewatering planet is expected to hence go through an increase in yield stress of the lithosphere. Three fundamentally different modes of heat transfer are caused by this change in yield stress of the lithosphere. These modes are here illustrated in terms of the time evolution of the so-called “Nusselt number”. This number defines the efficiency of heat transfer to the surface of a planet, i.e., the ratio of heat transfer by conduction over the actual heat transfer. A spike in the Nusselt number implies that the outer rigid surface of the planet does not transmit heat by conduction but founders episodically and dives into its interior. This is much like the stiff surface of a lava lake that dives episodically back into the lava lake (Turcotte 1995). Accordingly, in a basic planetary convection models three distinct modes with decaying, periodic or absent Nusselt spikes are inferred. The amplitude of the Nusselt spikes and the period of episodicity is entirely modulated by the water controlled yield stress (Moresi and Solomatov 1998). Firstly, for a low lithosphere yield stress case, episodic convective flushing events are expected. Eventually these decay into a plate tectonic-like stable mode of heat transfer shown by the solid line in Figure 8a. This is the case for the present Earth. Note that the Nusselt number is larger than unity implying that through plate tectonics the planets looses more efficiently its heat than by conduction only. With increasing yield stress (labeled intermediate yield stress in Fig. 8a) the convection mode switches again into a fully episodic behavior, which is the case for the present Venus where global planetary resurfacing has occurred 300-600 Ma ago (Nimmo and McKenzie 1998). Upon even higher strength of the lithosphere a switch to a third mode with a frozen lithosphere occurs (high yield stress Fig. 8a). This is presumably the case for Mars, where the lithosphere is punctured only through the giant volcano Olympus Mons (Stein et al. 2004).

DISCUSSION AND CONCLUSIONS The insight that the yield stress of the lithosphere controls basic modes of convection possibly provides the key to understanding the uniqueness of plate tectonics by linking it to

466

Regenauer-Lieb

(a)

70

Intermediate yield stress

60

Nusselt number

50

40

Low yield stress

30

20

10

High yield stress 0

(b)

0

0.1 Time

0.2

(c)

Figure 8. Basic convection styles of a planet as first pointed out by Moresi and Solomatov (1998). The three different modes are triggered by different lithosphere strength and have since been confirmed as robust features in basic mantle convection models with many different formulations. Panel a) shows the Nusselt number versus non-dimensional time. Panels b) and c) show contours of nondimensional temperature (0 = cold surface temperature, 1 = bottom temperature) and flow vectors. The system has a prominent tendency for resurfacing, i.e., oscillation between the two solutions shown in panels b) and c) with long periods of phase c) and short burst of b).

the role of water as a crucial element in planetary evolution (Kaula 1994). For a long time it has been a puzzle why the effective yield stress of plate boundaries of the present Earth appears to defy any reasonable material science based extrapolation of laboratory conditions. Plate boundaries appear to be at least an order of magnitude weaker than the plates themselves (McKenzie 1977; Bird 1998). As laid out in this review the influence of very small amounts of water can resolve this paradox (Regenauer-Lieb et al. 2001). The observed weakening mechanism relies on hydrogen controlled thermal-mechanical feedback in temperature sensitive elasto-visco-plastic creep. These thermal mechanical calculations suggest that this mechanism only operates above a threshold value of around 200 ppm H/106Si and can lead to vast local strength contrast. Water is incorporated in the dehydrated solid rock matrix of the oceanic lithosphere presumably only in very small proportions of 80-100 ppm H/106Si (Hirth and Kohlstedt 1996), however, our results show that a doubled water content can substantially alter the mechanical behavior of the lithosphere. It thus appears crucial to explicitly study the volatile flux within the Earth’s mantle and assess threshold values of volatile content/strength for the emergence and death of

Water & Geodynamics

467

plate tectonics. In particular the question how the lithosphere is hydrated (Fig. 2) needs further attention. This finding should change our view on how planetary tectonics operates. It should also fundamentally affect geological interpretations thereof causing a paradigm shift in geosciences. It now becomes more important to look at what happens exactly inside the lithosphere. In the past five years we have made fast progress in developing the tools to do just this resulting in detailed analyses of the visco-elasto-plastic lithosphere, while retaining the aspect of mantle convection (Moresi et al. 2002; Muhlhaus and Regenauer-Lieb 2005). The convection models also illustrate that the Earth’s plate tectonic mode is something special and that plate tectonics only occurs for a very narrow parameter range of yield stress/ water content of the lithosphere. These models also imply that plate tectonics as we know it today did not operate throughout the entire evolution history of the Earth. Plate tectonics most likely did not emerge as a sudden mode switching but it was preceded by pulses of convective overturn. Planets such as Venus and Mars may have had plate tectonics during their early stages but the evidence is now lost through whole scale planetary resurfacing. The global temperature of a planet reduces with age and so does the water content, in particular if water is not coming in through cometary impact and the atmosphere does not prohibit hydrogen escape into space (Bullock and Grinspoon 1996). Quantitative models of Earth atmosphere-biosphere-geosphere co-evolution (McGovern and Schubert 1989; Franck and Bounama 1995; Bounama et al. 2001; Franck et al. 2002; von Bloh et al. 2003) have been pioneered using parametric convection models. These models have significantly improved our understanding of climate-geosphere feedback. However, they are blind to the major transitions in the convective patterns of the Earth reported in the chapter on mantle convection. Such changes were previously implemented as model assumptions in an ad hoc manner. A good test for a planetary scale atmosphere co-evolution model is that the mode switching is not hardwired, but that the convection models should produce different scenario by reaching the threshold values in volatile fluxes. Ultimately, vastly different atmospheres of the neighboring planets Earth (N = 77%, CO2 = 0.03%, O2 = 21%) Venus (N < 2%, CO2 = 95%, O2 = 0%) and Mars (N < 3%, CO2 = 95%, O2 = 0%) should emerge from a mild variation of initial conditions and a consideration of the different sizes and the biosphere-geosphere feedback processes of the individual planets.

REFERENCES Ashby MF, Verall RA (1977) Micromechanisms of flow and fracture, and their relevance to the rheology of the upper mantle. Phil Trans Roy Soc London 288:59-95 Becker TW, Faccenna C, O’Connell RJ, Giardini D (1999) The development of slabs in the upper mantle: Insights from numerical and laboratory experiments. J Geophys Res 104(B7):15207-15226 Bercovici D (1993) A simple model of plate generation from mantle flow. Geophy J Int 114:635-650 Bercovici D, Karato S (2003) Theoretical analysis of shear localization in the lithosphere. Rev Mineral Geochem 51:387-420 Bercovici D, Ricard Y (2003) Energetics of a two-phase model of lithospheric damage, shear localization and plate-boundary formation. Geophys J Int 152:581-596 Bird P (1998) Testing hypotheses on plate-driving mechanisms with global lithosphere models including topography, thermal structure, and faults. J Geophys Res 103(B5):10115-10129 Bird P (2003) An updated digital model of plate boundaries. Geochem Geophy Geosystems 4 Boillot G, Froitzenheim N (2001) Non-volcanic rifted margins, continental break-up and the onset of sea-floor spreading: some outstanding questions. In: Non-Volcanic Rifiting of Continental Margins. Wilson RC, Whitmarsh RB, Taylor B, Froitzenheim N (eds), The Geological Society London, p 9-30 Bounama C, Franck S, von Bloh W (2001)The fate of the Earth’s ocean. Hydrology and Earth System Sci 5(4): 569-575 Branlund J, Regenauer-Lieb K, Yuen D (2000) Fast ductile failure of passive margins from sediment loading. Geophys Res Let 27(13):1989-1993

468

Regenauer-Lieb

Brodholt JP, Refson K (2000) An ab initio study of hydrogen in forsterite and a possible mechanism for hydrolytic weakening. J Geophys Res 105(B8):18977-18982 Bullock MA, Grinspoon DH (1996) The stability of climate on Venus. J Geophys Res101:7521-7530 Chrysochoos A, Maisonneuve O, Martin G, Caumon H, Chezeaux JC (1989) Plastic and dissipated work and stored energy. Nucl Eng Des 114:323-333 Cloetingh S, Wortel R, Vlaar NJ (1989) On the initiation of subduction zones. Pure Appl Geophy 129:7-25 Collins IF (2005) The concept of stored plastic work or frozen elastic energy in soil mechanics. Geotechnique 55(5):373-382 Danyushevsky LV, Eggins SM, Falloon TJ, Christie DM (2000) H2O abundance in depleted to moderately enriched mid-ocean ridge magmas; Part I: Incompatible behavior, implications for mantle storage, and origin of regional variations. J Petrology 41(8):1329-1364 Dauphas N, Marty B, Robert F (2000) The late asteroidal and cometary bombardment of earth as recorded in water deuterium to protium ratio. Icarus 148(2):508 Dijkstra AH, Drury MR, Vissers RLM, Van Roermund HLM, Newman J (2004) Shear zones in the upper mantle: Evidence from alpine- and ophiolite-type peridotite massifs. Geol Soc Spec Pub (224):11 Doin MP, Henry P (2001) Subduction initiation and continental crust recycling: the roles of rheology and eclogitization. Tectonophy 342(1-2):163-191 Donlon WT, Heuer AH, Mitchell TE (1998) Compositional softening in Mg-Al spinel. Philos Mag A 78(3): 615-641 Drury MR, Vissers RLM, Van Der Wal D, Hoogerduijn Strating EH (1991) Shear localisation in upper mantle peridotites. Pure Appl Geophy 137(4):439 Ellis S, Beaumont C, Pfiffner OA (1999) Geodynamic models of crustal-scale episodic tectonic accretion and underplating in subduction zones. J Geophys Res 104(B7):15169-15190 Evans B (1984) The effect of temperature and impurity content on indentation hardness of quartz. J Geophys Res 89(B6):4213-4222 Evans B, Goetze C (1979) The temperature variation of the hardness of olivine and its implications for the polycrystalline yield stress. J Geophys Res 84:5505-5524 Faccenna C, Giardini D, Davy P, Argentieri A (1999) Initiation of subduction at Atlantic-type margins: Insights from laboratory experiments. J Geophys Res 104(B2):2749-2766 Fernandez C, Perello V, Casillas R, Ahijado A, Hernandez-Pacheco A (1997) Shear zones as a result of intraplate tectonics in oceanic crust: The example of the Basal Complex of Fuerteventura (Canary Islands). J Struct Geol 19(1):41 Franck S, Bounama C (1995) Rheology and volatile exchance in the framework of planetary evolution. Adv Space Res 10:1079-1086 Franck S, Kossaki K, von Bloh W, Bounama C (2002) Long-term evolution of the global carbon cycle: historic minimum of global surface temperature at present. Tellus 54B:325-243 Fullsack P (1995) An arbitrary lagrangian-eulerian formulation for creeping flows and its application in tectonic models. Geophys J Int 120(1):1-23 Giunchi C, Sabadini R, Boschi E, Gasperini P (1996) Dynamic models of subduction: Geophysical and geological evidence in the Tyrrhenian Sea. Geophys J Int 126(2):555-578 Goetze C, Evans B (1979) Stress and temperature in the bending lithosphere as constrained by experimental rock mechanics. Geophys J Roy Astr Soc, 59:463-478 Gordon RG, Stein S (1992) Global tectonics and space geodesy. Science 256:333-342 Green AE, Naghdi PM (1965) A dynamical theory of interacting continua. Arch Rational Mech Anal 24:243263 Griggs DT (1967) Hydrolytic weakening of quartz and other silicates. Geophys J Roy Astron Soc 14:19-32 Gurnis M (1992) Rapid continental subsidence following the initiation and evolution of subduction. Science 255:1556-1558 Hall CE, Gurnis M, Sdorlias M, Lavier LL, Mueller RD (2003) Catastrophic initiation of subduction following forced convergence across fracture zones. Earth Planet Sci Lett 212:15-30 Hall PS, Kincaid C (2001) Diapiric flow at subduction zones: A recipe for rapid transport. Science 292(5526): 2472-2475 Hassani R, Jongmans D, Chery J (1997) Study of plate deformation and stress in subduction processes using two-dimensional numerical models. J Geophys Res 102(B8):17951-17965 Heggie M (1992) A molecular water pump in quartz dislocations. Nature 355:337-339 Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle: implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Hobbs BE (1985) The hydrolytic weakening effect in quartz. Geophys Monograph 31(Mineral Physics 1): 151-170 Huismans RS, Beaumont C (2002) Asymmetric lithospheric extension: The role of frictional plastic strain softening inferred from numerical experiments. Geology 30(3):211-214

Water & Geodynamics

469

Johnson W, Chitkara NR, Ranshi AS (1974) Plane-stress yielding of cantilevers in bending due to combined shear and axial load. J Strain Anal 9(2):67-77 Kameyama C, Yuen DA, Karato S (1999) Thermal-mechanical effects of low temperature plasticity (the Peierls mechanism) on the deformation of a viscoelastic shear zone. Earth Planet Sci Lett 168:159-162 Karato S, Jung H (1998) Water, partial melting and the origin of the seismic low velocity and high attenuation zone in the upper mantle. Earth Planet Sci Lett 157(3-4):193-207 Katayama I, Karato SI, Jung H (2004) New type of olivine fabric from deformation experiments at modest water content and low stress. Geology 32(12):1045 Kaula WM (1994) The tectonics of Venus. Philos Trans Royal Soc London Ser A 349(1690):345-355 Kemp DV, Stevenson DJ (1996) A tensile, flexural model for the initiation of subduction. Geophys J Int 125: 73-94 Kocks UF (1987) Constitutive behavior based on crystal plasticity. In: Unified Equations for Creep and Plasticity. Miller AK (ed) Elsevier Applied Science, p 1-88 Kocks UF, Argon AS, Ashby MF (1975) Thermodynamics and Kinetics of Slip. Pergamon Press Kohlstedt DL (2006) The role of water in high-temperature rock deformation. Rev Mineral Geochem 62: 377-396 Kohlstedt DL, Evans B, Mackwell SJ (1995) Strength of the lithosphere: Constraints imposed by laboratory measurements. J Geophys Res 100(B9):17587-17602 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohn SC (2006) Structural studies of OH in nominally anhydrous minerals using NMR. Rev Mineral Geochem 62:53-66 Kronenberg AK, Tullis J (1984) Flow strength of quartz aggregates: grain size and pressure effects due to hydrolytic weakening. J GeophysRes 89:42981-42997 Kusznir NJ (1982) Lithosphere response to externally and internally derived stresses - a viscoelastic stress guide with amplification. Geophys J Roy Astron Soc 70(2):399-414 Larsen TB, Yuen DA, Malevsky AV, Smedsmo JJ (1996) Dynamics of strongly time-dependent convection with non- Newtonian temperature-dependent viscosity. Phys Earth Planet Interiors 94(1-2):75-103 Lavier LL, Buck WR (2002) Half graben versus large-offset low-angle normal fault: Importance of keeping cool during normal faulting J Geophys Res 107(B6), Art. No. 2122 JUN 2002 Lenardic A, Kaula WM, Bindschadler DL (1995) Some effects of a dry crustal flow law on numerical simulations of coupled crustal deformation and mantle convection on Venus. J Geophys Res-Planets 100(E8):16949-16957 Lodders K (2003) Solar system abundances and condensation temperatures of the elements. Astrophys J 591(2 I):1220 Marty B, Yokochi R (2006) Water in the early Earth. Rev Mineral Geochem 62:421-450 McGovern P, Schubert G (1989) Thermal evolution of the Earth: effects of volatile exchange between atmosphere and interior. Earth Planet Sci Lett 96:27-37 McKenzie DP (1977) The initiation of trenches: A finite amplitude instability. In: Island Arcs Deep Sea Trenches and Back-Arc Basins. Maurice Ewing Ser., Vol. 1. Talwani M, Pitman WC (eds) American Geophysical Union, p 57-61 Mei S, Kohlstedt DL (2000a) Influence of water on plastic deformation of olivine aggregates: 1. Diffusion creep regime. J Geophys Res 105(B9):21457-21469 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates: 2. Dislocation creep regime. J Geophys Res 105(B9):21471-21481 Michael P (1995) Regionally distinctive sources of depleted MORB; evidence from trace elements and H2O. Earth Planet Sci Lett 131(3-4):301-320 Michibayashi K, Mainprice D (2004) The role of pre-existing mechanical anisotropy on shear zone development within oceanic mantle lithosphere: An example from the Oman ophiolite. J Petrology 45(2):405 Montesi LGJ, Zuber MT (2002) A unified description of localization for application to large-scale tectonics. J Geophys Res 107(B3):art no.-2045 Moresi L, Dufour F, Muhlhaus H (2002) Mantle convection models with viscoelastic/brittle lithosphere: Numerical methodology and plate tectonic modeling. Pageoph 159(10):2335-2356 Moresi L, Solomatov V (1998) Mantle convection with a brittle lithosphere: thoughts on the global tectonic styles of the Earth and Venus. Geophys J Int 133(3):669-682 Morra G, Regenauer-Lieb K (2006) A coupled solid-fluid method for modeling subduction. Philos Mag 86(2122):3307-3323 Muhlhaus H, Regenauer-Lieb K (2005) A self-consistent plate-mantle model that includes elasticity: Computation aspects and application to basic modes of convection. Geophys J Int163(2):788-800 Nemat-Nasser S (1979) Decomposition of strain measures and their rates in finite deformation elasto-plasticity. Int J Solids Struct 15:155-166

470

Regenauer-Lieb

Nimmo F, McKenzie D (1998) Volcanism and tectonics on Venus. Annu Rev Earth Planet Sci 26:23-51 Paterson MS, Kekulawala KRSS (1979) The role of water in quartz deformation. Bull Mineral 102:92-98 Poirier J-P (1985) Creep of Crystals: High Temperature Deformation Processes in Metals, Ceramics and Minerals. Cambridge University Press Regenauer-Lieb K (1999) Dilatant plasticity applied to alpine collision: ductile void growth in the intraplate area beneath the Eifel volcanic field. J Geodyn 27:1-21 Regenauer-Lieb K, Kohl T (2003) Water solubility and diffusivity in olivine: Its role for planetary tectonics. Mineral Mag 67:697-715 Regenauer-Lieb K, Yuen D, Branlund J (2001) The initation of subduction: criticality by addition of water? Science 294:578-580 Regenauer-Lieb K, Yuen DA (2003) Modeling shear zones in geological and planetary sciences: solid- and fluid- thermal- mechanical approaches. Earth Sci Rev 63:295-349 Rice JR (1993) Mechanics of Solids, Encyclopedia Britannica, p 734 - 747 Schubert G (2001) Breakthroughs in our knowledge and understanding of the earth and planets. Ann Rev Earth Planet Sci 29:1-15 Shemenda AI (1992) Horizontal lithosphere compression and subduction: constraints provided by physical modeling. J Geophys Res 97(B7):11097-11116 Solomatov V (1995) Scaling of temperature-dependent and stress-dependent viscosity convection. Phys Fluids 7(2):266-274 Solomatov V, Moresi L (2000) Scaling of time-dependent stagnant lid convection: Application to small-scale convection on Earth and other terrestrial planets. J Geophys Res 105(B9):21795-21817 Solomatov VS (2004) Initiation of subduction by small-scale convection. J Geophys Res B: Solid Earth 109(1) B01412 Stein C, Schmalzl J, Hansen U (2004) The effect of rheological parameters on plate behavior in a selfconsistent model of mantle convection. Phys Earth Planet Interiors 142(3-4):225-255 Tackley P (1998) Self-consistent generation of tectonic plates in three-dimensional mantle convection. Earth Planet Sci Lett 157:9-22 Tackley P (2000) Self-consistent generation of tectonic plates in time- dependent, three-dimensional mantle convection simulations, 1. Pseudoplastic yielding. Geochem Geophys Geosystems 3 01(23):1525 Toth G, Gurnis M (1998) Dynamics of subduction initiation at preexisting fault zones. J Geophys Res 103: 18053-18067 Tozer DC (1985) Heat transfer and planetary evolution. Surveys in Geophysics 7(3):213-247 Trompert R, Hansen U (1998) Mantle convection simulations with rheologies that generate plate-like behavior. Nature 395:686-689 Tullis TE, Horowitz FG, Tullis J (1991) Flow laws for polyphase aggregates from end member flow laws. J Geophys Res 96(B1):8081-8096 Turcotte DL (1995) How does Venus lose heat? J Geophys Res 100(E8):16931-16940 van Hunen J, van den Berg AP, Vlaar NJ (2000) A thermo-mechanical model of horizontal subduction below an overriding plate. Earth Planet Sci Lett 182(2):157-169 von Bloh W, Bounama C, Franck S (2003) Cambrian explosion triggered by geosphere-biosphere feedbacks. Geophys Res Lett 30(1):(6)1-5 Walcott RI (1970) Flexural Rigidity, Thickness, and Viscosity of the Lithosphere. J Geophys Res 75(20): 3941-3951 Winder RO, Peacock SM (2001) Viscous forces acting on subducting lithosphere. J Geophys Res 106(B10): 21937-21951 Wortel MJR, Vlaar NJ (1988) Subduction zone seismicity and the thermomechanical evolution of downgoing lithosphere. Pure Appl Geophys 128(3-4):625-659

Water & Geodynamics

471

APPENDIX: THERMOMECHANICAL APPROACH The basis for a thermomechanical approach to rock deformation (Green and Naghdi 1965) lies in the local strain energy rate density which is nothing else but the local power (work rate) of a reference volume. Thermomechanical approaches have recently been rediscovered as a new paradigm in continuum mechanics (Fig. 3). Here, following Collins (2005) a simplified isothermal deformational work rate is discussed. A more complete non-isothermal formulation can be found elsewhere (Regenauer-Lieb and Yuen 2003). Consider an isothermal reference volume +Φ σ : ε ≡ W = Ψ

(10)

where Ψ is the Helmholtz free energy function and W is the rate of working equivalent to the double dot product (scalar product) of the stress and the strain rate tensor dε/dt of the applied stress σ tensor (Cauchy stress); Φ is the rate of dissipation. The Clausius Duhem Inequality (second law of thermodynamics) states that: Φ≥0

(11)

Expanding the isothermal deformational work rate as: σ : ε =

∂Ψ ∂Ψ ∂Φ : ε + k : α k + k : α k ∂ε ∂α ∂α

(12)

a split of dissipated and non-dissipated microstructural processes attributed to microstrain α k is obtained assuming a summation over k of the various microstructural processes. Now the Cauchy stress tensor is defined as σ≡

∂Ψ ∂ε

(13)

ε≡

∂G ∂σ

(14)

and the small strain tensor as:

for which a maximum entropy production is given if no zero work rate term is allowed in the energy balance. From the partial derivatives of the Helmholtz free energy and the dissipation potential over their microstrains and microstrain rates, respectively we obtain two familiar quantities. One is the recoverable elastic small strain measure and the other is a dissipative small strain measure, i.e., we obtain the additive elasto-dissipative strain rate decomposition (Nemat-Nasser 1979) as ε =

∂ 2G ∂ 2G : σ + : α k = ε elastic + ε diss ∂σ 2 ∂σ∂α k

(15)

This is the standard thermodynamic formulation in terms of time derivatives where the power is considered instead of the energy. It may be more convenient to discuss the concepts on the basis of a standard thermomechanical formulation where instead of the rates virtual variations δ are considered (Kocks et al. 1975). The small strain addition then reads δε = δε elastic + δε diss

(16)

and the small macroscopic strain in the unit volume V is the average of the microscopic, local strains δε =

1 δε Loc dV V∫

(17)

472

Regenauer-Lieb

similarly the macroscopic stress then is the average of microscopic stresses σ=

1 σ Loc dV V∫

(18)

As an example thermodynamics can be used to derive micromechanical equilibrium/ flow conditions of the three basic rheological laws applied here, i.e., elastic deformation, dislocation, and diffusion creep. Using the same isothermal formulation, considering all nonmechanical work terms neglected, the second law of thermodynamics (Eqn. 11) is now understood in terms of energy dissipated during deformation by virtual displacements. Vσ σ δε − δψ ≥ 0

(19)

Thermodynamics is capable of describing equilibrium conditions for microscopic, internal, hidden processes. An estimate of the microscopic equilibrium can therefore be derived in terms of macroscopic strain. Defining the microstructural stress as an internal stress variable τ=

1 δΨ V δε

(20)

then the change of the dissipation during virtual displacements is 1 δΦ dt = σ − τ δε

∫V

(21)

This can be used to recast the microscopic equilibrium condition in Equation (19) in terms of macroscopic strain. The internal stress now defines the internal deformation resistance describing the changes of the free energy of a body. For elastic deformation the free energy changes due to the stretching of the atomic bonds. This is proportional to the macroscopic strain times the elastic modulus. When the applied stress is exactly equal to the internal stress, static mechanical equilibrium is achieved and Hooke’s Law applies. There is no change in the dissipation and the deformation is reversible, i.e., Equation (21) is zero. On the other extreme scale of thermodynamic conditions lies diffusion creep where there is no static equilibrium and there is only dissipation. Diffusion of vacancies does not change the free energy. So the internal stress τ =0 and there can be no equilibrium condition for any applied stress σ , however, if the response to the driving stress is linear then Newtonian viscosity follows. The third case is that of dislocation glide which is a mixed case of the two end members. In this case the condition for macroscopic dislocation motion or slip requires that the applied stress must be larger than the (microstructural) internal stress σ≥τ

(22)

Kocks et al. (1975) show that the concept of a mechanical threshold follows naturally out of this thermodynamic approach. The thermodynamic conditions for macroscopic dislocation motion, or slip, is that the work done per unit volume, by the applied stress must at least equal the stored free energy of deformation per unit volume. This prescribes the lower limit to the applied stress required for flow to occur; below this stress no steady flow can occur even at high temperatures; see page 17 Kocks et al. (1975). Another interesting case is the one in which several microstrain processes happen simultaneously. In this case a good approach is simply an extension of the above (Ashby and Verall 1977; Kocks 1987). A useful approach is to grade the microstrain processes according to increasing structural scale. Whenever the structural dimensions for the different threshold mechanism differs by an order of magnitude the resistance of the finest structure can be

Water & Geodynamics

473

considered smooth and added to the one of the next larger scale. There are several possibilities of superposition. Here only the two principle cases are discussed. If microstrain is achieved through a set of independent states in parallel so that e.g., the dislocation merely selects the easiest available path, macrostrain is achieved through an addition of strain rates. This is the basic assumption underlying Equation (8). If the microstrain had to progress through a set of states in series then the times spent for each displacement were to be added. For this case, in an approximate way, the inverse strain rates are additive. This is for instance the case for obstacle controlled and lattice controlled dislocation glide. Inverse strain rates are added to give the total inverse of the strain rate (Ashby and Verall 1977).

15

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 343-375, 2006 Copyright © Mineralogical Society of America

Remote Sensing of Hydrogen in Earth’s Mantle Shun-ichiro Karato Department of Geology and Geophysics Yale University New Haven, Connecticut, 06520-8109, U.S.A. e-mail: [email protected]

INTRODUCTION 1

Hydrogen in Earth’s interior is known to play a key role in a number of processes. Consequently, inferring the distribution of hydrogen is a critical step in our study of dynamics and evolution of Earth. Usually, hydrogen distribution is inferred from two types of samples at the surface. First, a magma may contain hydrogen (water), and under some conditions the water content of the magma can be quenched upon cooling. In these cases, measurements of the water content of the magma provide us with some constraints on the hydrogen (water) content of the source region (if we know the partitioning of hydrogen between magmas and the source rocks, and the degree of melting). Second, hydrogen content of some xenoliths transported by magma can be measured. They provide a direct clue as to the hydrogen content in a region from which a xenolith has been carried. However, these direct, petrologic approaches have two major problems. Firstly, the sampling is limited by the distribution of volcanoes, and even if there are volcanoes that carry rocks from Earth’s interior, the depth extent that volcanoes sample rocks is limited (usually <200 km). Secondly, there is no guarantee that the hydrogen content that one measures on these samples actually represents the hydrogen content in a region where these samples came from. For example, hydrogen is known to diffuse very easily so hydrogen dissolved in minerals could diffuse out during the transport of a rock, or conversely, a piece of rock may acquire extra hydrogen during its ascent. Also hydrogen atoms dissolved in minerals may precipitate in the mineral to form fluid inclusions or micro-scale hydrous minerals. In summary, the direct method to infer the distribution of hydrogen from rock samples has major limitations, and an alternative approach, i.e., remote sensing hydrogen content from geophysical observations appears to be worth serious consideration. In this chapter, I will review some of the recent progress in inferring hydrogen content based on geophysical observations. The major advantage of this approach is the fact that hydrogen in much broader regions can be inferred because geophysical observations span much broader regions, particularly the depth. Because hydrogen has drastic effects on physical properties, some of the geophysically measurable properties are sensitive to hydrogen content. However, such an approach is new and a number of challenges exist. The key, from mineral physics point of view, in this attempt is to establish the relationship between hydrogen content and geophysical properties including the analysis of a range of trade-offs. Establishing these relationships including the developments of methodology to differentiate the influence of hydrogen from other factors requires in-depth analysis of microscopic physics of various 1

I will use a term “hydrogen” in most part of this paper to imply that it is hydrogen and not water molecule that affects the physical and chemical properties of silicates. In the geological literature, however, the concentration of chemical species is usually measured as wt% of oxide. Therefore when the concentration of hydrogen is discussed I will use water content in wt% following geological literatures. 1 wt% of water in olivine corresponds to 1.5×105 ppm H/Si (atomic ratio) and not molecular water.

1529-6466/06/0062-0015$05.00

DOI: 10.2138/rmg.2006.62.15

344

Karato

properties in addition to a sound understanding of geophysical methods of measuring relevant properties. In this chapter, I will first provide a brief tutorial of geophysical methods of inferring some of the key parameters that may be sensitive to hydrogen. Then I will give a detailed discussion on the mineral physics bases for inferring hydrogen content from geophysical observations. This will be followed by the discussion on some examples showing how these methods can be applied to infer hydrogen contents in various regions of Earth’s mantle.

GEOPHYSICAL OBSERVATIONS When an attempt is made to interpret any geophysical observations, it is critical to understand the nature of uncertainties in each geophysical technique. For example, in seismological studies, the nature of structures that one can resolve depends on the wavelength of waves used and hence the nature of uncertainties is different between body wave and surface wave observations. Also there is a complication in inferring anisotropic structures from seismology. Understanding these technical issues is critical to interpret geophysical observations. Therefore I will provide a brief summary of some of the geophysical techniques.

Electrical conductivity The physical basis of the methods for inferring electrical conductivity in the Earth is electromagnetic induction. When there is a variation in the magnetic field outside of the Earth (due to solar storms for example), then electric current is induced inside the Earth whose magnitude depends on the electrical conductivity. The measurements of electromagnetic field are in most cases made on Earth’s surface. The observed electromagnetic field comes both from the “source” (i.e., field from the outside of Earth) and the “induced” field inside the Earth. Therefore the first step is to separate them using the spherical harmonic analysis. The field coming from inside is interpreted in terms of distribution of electrical conductivity. So the basic equations to govern this phenomenon is an induction equation, viz., ∇2F = κ

∂F ∂t

(1)

where F is either the electric or magnetic field, and κ ≡ 4πσcond·µmag where σcond is the electrical conductivity and µmag is the magnetic susceptibility. For a one-dimensional case, the solution of this equation has a form z F ∝ exp ⎛⎜ − ⎞⎟ exp ( i ωt ) z ⎝ 0⎠

(2)

with z0 =

1 2 πσ cond µ mag ω

(3)

where z is the depth, ω is the (angular) frequency, t is time, σcond is the electrical conductivity and µmag is the magnetic susceptibility. Consequently, the magnetic disturbance with a long period will reflect the conductivity at greater depth. The magnetic susceptibility is nearly constant in Earth, so one infers the distribution of electrical conductivity from this type of observation. A lower frequency disturbance penetrates deeper into Earth’s interior, and can sense the electrical conductivity in the deeper portions. In other words, by investigating the time-variation of the electromagnetic field with various frequencies one can determine the depth variation of conductivity. In some cases, when the anisotropy in the electromagnetic field is determined, one will obtain some constraint on the anisotropy in conductivity. One

Remote Sensing of Hydrogen in the Earth’s Mantle

345

limitation with this approach is that because the amplitude of electromagnetic disturbance decays with depth exponentially, the depth resolution of this technique is not high. Similar to surface wave studies, sharp discontinuities are difficult to detect by electromagnetic induction. Since the depths of discontinuities are not well constrained from the geomagnetic studies, these depths are usually assumed and the magnitude of conductivity jumps and a smooth variation in conductivity between these discontinuities are determined by the inversion. Therefore it is critical to know the depth at which a discontinuity in conductivity may occur. An important progress in this connection was the experimental study by Xu et al. (1998b) who showed that the electrical conductivity of wadsleyite and ringwoodite are much higher than that of olivine and comparable to that of silicate perovskite2. These studies motivated inversion of electromagnetic induction data assuming jumps in conductivity at 410 km and 660 km. If one does not assume any discontinuities in conductivity, the best-fit models tend to be biased towards those with a smooth depth variation. So in short, among a range of models of electrical conductivity, those with assumed discontinuities at 410-km (and 660-km) best resolve any jumps in conductivity caused by a discontinuity in water content. Technical issues of inverting for the electromagnetic sounding are discussed by Parker (1980), Constable et al. (1987) and de Groot-Hedlin and Constable (2004). Figure 1 illustrates one example of such a study showing a marked increase in electrical conductivity at 410-km discontinuity in the mantle beneath the Pacific ocean (from Utada et al. 2003). Note also that the electrical conductivity also has a large lateral variation. Utada et al. (2005) reported a higher conductivity in the transition zone beneath Hawaii, as well as some regions of the mantle beneath Philippine Sea. A large jump in electrical conductivity at ~410-km is frequently observed in both oceanic and continental upper mantle (e.g., Olsen 1999; Tarits et al. 2004).

Figure 1. Electrical conductivity-depth profiles inferred from electromagnetic sounding in the Pacific ocean. Three profiles correspond to the results of inversion based on different assumptions about the discontinuity (after Utada et al. 2003). The results show a jump in electrical conductivity of a factor of ~10. Electrical conductivity shows a large regional variation, particularly in the upper mantle. However, a jump of a factor of ~10 is often observed at 410-km discontinuity (e.g., Olsen 1999; Tarits et al. 2004). 2

As I will explain later, the interpretation of the data by Xu et al. (1998) turned out to be incorrect. But ironically this incorrect interpretation promoted detailed studies on the depth variation in conductivity across the transition zone.

346

Karato

Seismic wave velocities Variation of seismic wave velocities in Earth can be determined by various techniques. There are excellent textbooks on this topic, but I need to summarize some of the crucial points for our purpose. The key question that I address in this chapter is the distribution of hydrogen as inferred from the geophysical observations on the upper mantle and the transition zone. In this context, it is important to appreciate the difference in the nature of seismological observations inferred from short-period body-waves and relatively long-period surface waves. The use of surface waves is critical to resolve the depth-dependent structure of the upper mantle where body-waves have little sensitivity. However, by its very nature, surface wave data are not sensitive to sharp but small discontinuities. Therefore the surface wave data are good data source to determine a gross depth-dependent structure, but are not a good data set to identify small but sharp discontinuities. The latter type of structures is better constrained by body-wave observations, particularly those using data of converted waves at discontinuities. An important case to illustrate this point is the inference of upper mantle structure of the oceanic upper mantle. A majority of studies on upper mantle structure is based on surface wave data, and these studies on the oceanic upper mantle show a gradual increase in the thickness of oceanic lithosphere associated with a gradual increase in the wave velocity in the asthenosphere (at the same depth) (e.g., Yoshii 1973; Forsyth 1975). In contrast, some recent studies using short wavelength body waves show a rather constant depth of the discontinuity at around ~60-70 km depth (e.g., Gaherty et al. 1996; Evans et al. 2005). Karato and Jung (1998) interpreted this sharp, age-independent discontinuity as a result of stiffening due to dewatering as originally proposed by Karato (1986). (Hirth and Kohlstedt (1996) adopted the same model and added some petrological details.) One of the common difficulties in inverting geophysical observations in terms of physical and chemical state of Earth is non-uniqueness. The same low velocity anomaly can be attributed to high temperature, high iron content, partial melting or a high water content. To help reduce the non-uniqueness, simultaneous inversion of multiple data set is a useful approach. Not only P-wave but also S-wave velocity anomalies should be used. However, one must make sure that the resolutions of each technique are similar in order to combine the results of different data sets. Some technical issues on simultaneous inversion of P and S waves are discussed by Masters et al. (2000).

Seismic wave attenuation Velocity anomalies are the primary data from seismology that we can get with reasonably high resolution. Velocities of seismic waves reflect the “stiffness” (i.e., elastic constants) (and density) of Earth materials. However, in some cases, we can also obtain some information on “softness” of materials from seismological data. This is seismic wave attenuation that reflects non-elastic deformation of Earth materials which is very sensitive to temperature and a small amount of hydrogen. Reviews on seismic wave attenuation from seismological view-points include Bhattacharya et al. (1996) and Romanowicz and Durek (2000). Seismic wave attenuation, or the degree to which energy is lost during the wave propagation, is characterized by a Q factor defined by Q −1 ≡

∆E 2πE

( 4)

where ∆E is the energy dissipated during one cycle of wave propagation and E is energy stored in the system. The Q factor can be determined by the amplitude decay of seismic waves that can be determined by the frequency dependence of amplitude, viz., ⎛ ω ⎞ A ( x, ω) = A0 exp ⎜ − x⎟ ⎝ 2 vQ ⎠

(5)

Remote Sensing of Hydrogen in the Earth’s Mantle

347

where v is the velocity of seismic waves, x is distance and ω is the (angular) frequency of seismic waves. Usually, determining the attenuation is more difficult than seismic wave velocities because a number of other factors can contribute to the amplitude including geometrical focusing or defocusing, and scattering. In particular, the effects of geometrical focusing or defocusing are large for surface wave studies and consequently careful analysis of seismic records is required including waveform inversion (e.g., Gung and Romanowicz 2004). Nevertheless, when highdensity seismic sources and receivers are available, one can calculate a decent three-dimensional mapping of seismic wave attenuation. An example of results of attenuation tomography is shown together with velocity tomography (Shito et al. 2006). In this special case, attenuation factor, Q, was determined from the ratio of amplitude of P- and S-waves from the same data set assuming that QP/QS is constant throughout the study region. This procedure minimizes the influence of source and geometrical effects and if one has a dense array of seismic stations and sources (earthquakes), then one can obtain high-resolution tomographic maps (Fig. 2).

Seismic anisotropy Seismic anisotropy can provide additional constraints on the distribution of hydrogen. The possible mechanism for hydrogen to alter seismic anisotropy in the upper mantle was proposed by Karato (1995) based on experimental observations showing the high anisotropy in hydrogen weakening effects in olivine single crystals (Mackwell et al. 1985). Seismic anisotropy can be determined by a range of methods, but two methods are most widely used. These are shear-wave splitting measurements, and the measurements of vSH /vSV polarization anomalies. Both of these methods use polarization anisotropy as opposed to

Figure 2. Tomographic images of a cross section of the upper mantle beneath the Philippine Sea. The left hand side figures show anomalies in attenuation, and the right hand side figures show anomalies in P-wave velocities (from Shito et al. 2006). See Plate 1 for color figure.

348

Karato

azimuthal anisotropy. The polarization anisotropy is a type of anisotropy that describes the difference in the velocity of two types of shear waves with different polarizations. The major advantage of using this method is that one can determine seismic anisotropy from a single seismic record, and therefore the results are free from the influence of lateral heterogeneity. Azimuthal anisotropy can also be determined if a dense coverage of seismic rays is available for a given region, but in general, the errors in azimuthal anisotropy are larger than those for polarization anisotropy. An example of high-resolution shear wave splitting measurements is shown in Figure 3.

Topography of discontinuity The topography of some of the seismic discontinuities may also depend on hydrogen content. This is the case when a seismic discontinuity is caused by a phase transformation and when there is a large partitioning of hydrogen between the two minerals. An example is the olivine to wadsleyite transformation that occurs at ~410-km. The topography of a discontinuity can be measured by using seismic waves converted or reflected at a boundary (see e.g., Shearer 2000). Topography on a given discontinuity could be caused by several reasons including temperature anomalies and water content anomalies. Therefore, similar to velocity anomalies, if only topographic anomalies are known, it is difficult to obtain unique conclusion about the cause. Therefore it is important to obtain not only the topography on the discontinuity but also some other parameters from the same region. For example, as will be explained in the next section, topography on the discontinuity could be due to a temperature anomaly or to an anomaly in hydrogen content. Both factors affect the topography and velocity in different ways, so the simultaneous inversion of these two parameters will provide a tight constraint on hydrogen content. Since inferring the depth of a discontinuity relies on the knowledge of seismic wave velocities, there is a trade-off between the determination of topography and velocity anomalies. Gu et al. (2003) discussed the methods to reduce the uncertainties caused by this trade-off. By combining the results of topography on the “410-km” discontinuity and velocity anomalies

(a)

(b)

Figure 3. An example of spatially varying pattern of shear wave splitting using regional seismic sources and dense stations (from Nakajima and Hasegawa 2004). (a) A map of study area (Tohoku Japan) where plate subduction occurs from the east. (b) The results of shear wave splitting measurements. An abrupt change in the pattern occurs near the volcanic front. Near the trench the direction of polarization of fast Swave is near trench parallel, whereas it becomes nearly normal to the trench at a far distance. Such a trend is commonly observed in many subduction zones (e.g., Smith et al. 2001; Long and van der Hilst 2005).

Remote Sensing of Hydrogen in the Earth’s Mantle

349

in the transition zone, Blum and Sheng (2004) inferred the hydrogen content in the transition zone. Suetsugu et al. (2006) conducted a similar study for the western Pacific. However, in these studies, topography and velocity anomalies were not determined in a consistent way, but rather determined separately. Therefore the results of these earlier studies have large uncertainties.

Sharpness of discontinuities The sharpness of discontinuities may depend on the hydrogen content if the sharpness is related to a phase transformation loop involving two minerals with a large contrast in hydrogen solubility (Wood 1995). The sharpness can be inferred from a range of seismic observations particularly those relying on the amplitude of converted phases at a discontinuity (e.g., Benz and Vidale 1993; van der Meijde et al. 2003). In essence the inference of sharpness is based on the amplitude of converted waves or reflected waves as a function of wavelength. However, because these methods use the information on the amplitude of waves, the uncertainties are large as compared to the measurements of the depth of seismic discontinuities. In addition, the mineral physics background for interpreting the sharpness is more complicated than the interpretation of the topography of a discontinuity as I will discuss later.

PHYSICAL BASIS FOR INFERRING HYDROGEN CONTENT FROM GEOPHYSICAL OBSERVATIONS Electrical conductivity The first quantitative work on the influence of hydrogen on electrical conductivity is Karato (1990) who pointed out that a high diffusivity and solubility of hydrogen in olivine implies that hydrogen contributes to electrical conductivity. In this earlier suggestion I used the Nernst-Einstein relation for electrical conductivity that allows us to calculate electrical conductivity from the concentration and mobility of charged species, viz., σ cond = f

Dcq 2 RT

(6)

where σcond is electrical conductivity, f is a constant about unity, D is the diffusion coefficient of a charged species, c is the concentration of the charged species, q is the electric charge of the species, R is the gas constant and T is temperature. Given this relation, one can calculate electrical conductivity from the known concentration of the charged species and its diffusion coefficient. The simplest assumption that I made in that paper was (i) all the hydrogen atoms dissolved in mineral (olivine) contribute to electrical conductivity equally (i.e., c is the total concentration of hydrogen), and (ii) the (chemical) diffusion coefficient reported by Mackwell and Kohlstedt (1990) can be used to calculate the electrical conductivity. This hypothesis was used to interpret geophysical observations by many scientists without experimental tests (e.g., Lizarrale et al. 1995; Hirth et al. 2000; Simpson 2002; Tarits et al. 2004; Evans et al. 2005; Simpson and Tommasi 2005). The first experimental test of this hypothesis was reported by Huang et al. (2005) for wadsleyite and ringwoodite (Fig. 4) and a similar study has been performed by Wang et al. (2006) for olivine. These experimental studies have shown the basic validity of Karato (1990)’s hypothesis that hydrogen enhances electrical conductivity, but in detail some of the assumptions of Karato (1990) are not supported by the experimental observations. Two points must be noted. First the hydrogen content exponent, i.e., the value of r in σ cond ∝ CWr is found to be ~0.6-0.7 for olivine, wadsleyite and ringwoodite, whereas the Karato (1990) model predicts r = 1. Second, for olivine and wadsleyite, where we have data on electrical conductivity and diffusion, the activation enthalpy of diffusion is considerably higher than that of electrical conductivity (Wang et al. 2006). These observations show that the charge carrier

350

Karato

Figure 4. The influence of water content on electrical conductivity in wadsleyite (after Huang et al. 2005). The numbers next to each line are the temperatures (K).

is not the majority of hydrogen atoms (protons) dissolved in minerals, i.e., protons trapped at M-site vacancies, (2H)×M , but rather some other species with higher mobility (lower activation enthalpy). An immediate conclusion is that one cannot calculate the electrical conductivity of a mineral from the solubility and diffusion coefficient of “hydrogen” as originally proposed by Karato (1990). Huang et al. (2005) proposed that electrical conductivity is due to the motion of free protons that can be produced by the ionization reaction (2H)×M ↔ H′M + H i

(7)

where H′M is a M-site vacancy that contains one proton, and H• is a free proton. If free proton is easier to move, as expected, then the activation enthalpy for conduction will be smaller than that of diffusion of (2H)×M . Therefore I consider that a better hypothesis for the hydrogenenhanced electrical conductivity is that the charge carrier is free proton, viz., σ cond ∝ [H i ] ⋅ µ H i ∝ CWr ∝ fHr 2 O

(8)

where CW is the water content, [H•] is the concentration of free protons and µH• is their mobility and r is a constant that depends on the types of the dominant charged defects. Here I made a relation that CW ≈ [(2H)×M ] ∝ fH 2 O . When the charge neutrality condition is [H′M ] = [ Fe iM ] as suggested by Karato (1989a) (see also Mei and Kohlstedt 2000a), then this model predicts r = ¾, whereas if the charge neutrality condition is the same as in the water-poor conditions, i.e., ′′ ] = [ Fe iM ], r = ½. These predictions are consistent with the experimental observations of 2[ VM r = 0.6-0.7. An immediate implication of this model is that anisotropy in conductivity can be different from the anisotropy of chemical diffusion of hydrogen because anisotropy of diffusion of a free proton may be different from that of (2H)×M. In many previous studies, anisotropy in conductivity predicted from the Karato (1990) model was used to interpret geophysical observations (e.g., Lizarrale et al. 1995; Hirth et al. 2000; Simpson 2002; Evans et al. 2005; Simpson and Tommasi 2005). I conclude that there is no strong mineral physics basis to support this assumption. Direct experimental studies on electrical conductivity under hydrous conditions are needed to determine the possible anisotropy in electrical conductivity.

Remote Sensing of Hydrogen in the Earth’s Mantle

351

I should also note that there are several issues that also need more attention before electrical conductivity can be used as a sensor for hydrogen content with high confidence. First, so far we have made a simple assumption that the electrical conductivity is dominated by volumetrically dominant minerals such as olivine (in the upper mantle) and wadsleyite or ringwoodite (in the transition zone). The validity of this assumption needs further examination. In the shallow upper mantle, the hydrogen solubility in aluminous orthopyroxene is significantly higher than that of olivine (e.g., Rauch and Keppler 2002). In this case, the electrical conductivity of the upper mantle could be controlled by orthopyroxene. Similarly, the influence of garnet must be examined for the deep upper mantle and the transition zone. Second, so far little attention was paid to the influence of grain-size on electrical conductivity. ten Grotenhus et al. (2004) showed that for hydrogen-poor iron-free olivine (forsterite), the electrical conductivity increases with decreasing grain-size. If the grain-boundary effect is important, then the anisotropy of conductivity will be weaker than expected from single crystal data. Our preliminary data on olivine indicate that grain-boundary effect is weak for iron-bearing olivine for the grain-size of ~ 10 µm to ~ 1 mm, but the details are not known yet. Electrical conductivity may also be enhanced by hydrogen through the enhancement of diffusion of some ionic species (e.g., Karato 1990). This is an obvious possibility since Karato et al. (1986) inferred the enhancement of diffusion by hydrogen (see also Mei and Kohlstedt 2000a). Hier-Majumder et al. (2005) investigated this through an experimental study, and concluded that this effect is minor compared to the direct influence by proton conduction. I should also mention an important technical issue in the experimental study of hydrogenrelated properties. In a paper by Xu et al. (1998b), they reported that wadsleyite and ringwoodite have much higher electrical conductivity than olivine compared at similar pressure and temperature. Similar studies were made on cation diffusion (e.g., Farber et al. 1994; Chakraborty et al. 1999). It was considered (incorrectly) that the electrical conductivity and diffusion in wadsleyite (and ringwoodite) are intrinsically higher (faster) than those processes in olivine. However, later studies in my lab clearly showed that the contrast in electrical conductivity between these minerals is almost entirely due to the (then unrecognized) difference in hydrogen content in these minerals: compared at the same pressure, temperature and hydrogen content, electrical conductivity in olivine is similar to that of wadsleyite (or ringwoodite). Since the affinity of hydrogen to wadsleyite (or ringwoodite) is much higher than that of olivine, wadsleyite and ringwoodite tend to dissolve more hydrogen from the surrounding medium during high-pressure experiments than olivine, which enhances electrical conductivity and other defect-related properties enormously. In any experimental study on hydrogen-related properties, one must determine the hydrogen content in the sample both before and after each experiment.

Seismic properties Seismic wave velocities and attenuation. The most obvious effect of hydrogen on seismic wave velocities is that the addition of hydrogen reduces seismic wave velocities by reducing the bond strength. The magnitude of such an effect was calculated by Karato (1995) assuming that the incorporation of hydrogen in a crystal will create a region with zero elastic modulus, and the average elastic modulus of a mineral containing these weak regions is calculated by a homogeneous strain model (i.e., Voigt model). This model predicts ~1% reduction for 1 wt% addition of water, which roughly agrees with the later experimental observations (e.g., Inoue et al. 1998; Jacobsen et al. 2004), see also Jacobsen, this volume for more detail of this effect). This is an effect on bond strength at an elastic limit. In other words, this is an effect on seismic wave velocities at infinite frequency, viz., v ∞ = v0∞ (1 − A ⋅ CW )

( 9)

where CW is water content (in wt%), v0∞ is the unrelaxed velocity for hydrogen-free mineral,

352

Karato

and A is a constant (~1-2). An immediate conclusion from this analysis is that this direct effect is important only for very large values of water content, ~0.1-1 wt%. If the water content is less than 0.1 wt%, the influence of water on seismic wave velocity is less than ~0.5%, so is not important. The water content in the upper mantle is less than ~0.1 wt%. Consequently, Karato (1995) concluded that this effect is not very important in the upper mantle, although this effect is important in the transition zone if water content is as high as ~1 wt%. In addition to the effects on unrelaxed velocity, hydrogen may affect seismic wave velocities through its effect on anelasticity. The importance of anelasticity in seismic wave propagation was pointed out by Gueguen and Mercier (1973), Karato (1993), and Minster and Anderson (1980). Karato (1995, 2003) suggested that the effect of hydrogen on seismic wave velocity is mostly due to the enhancement of anelasticity (and through changes in anisotropy, see later part of this paper). Experimental evidence for this is provided by Jackson et al. (1992) who showed that anelasticity of a dunite sample (Åheim dunite) is reduced significantly after the majority of water is removed from the sample by heat treatment. Since the heat treatment used to remove hydrogen resulted in other complications, Jackson et al. (1992) did not conclude that the change in anelasticity is due to the difference in hydrogen content. However the observations by Jackson et al. (1992) are similar to those by Chopra and Paterson (1984) on exactly the same dunite who found that the creep strength was significantly increased after the heat treatment. The results of Chopra and Paterson (1984) were supported by later works on synthetic olivine polycrystals (e.g., Karato et al. 1986; Mei and Kohlstedt 2000a,b), and are interpreted as a result of change in hydrogen content. Therefore the simplest explanation for the observation by Jackson et al. (1992) is to attribute it to the change in hydrogen content. As such the experimental evidence on the influence of hydrogen on anelasticity is preliminary, and the exact functional form by which seismic wave attenuation depends on hydrogen content is not known. However, based on the frequency dependence of seismic wave attenuation, and the experimental results on the relationship between hydrogen content and various kinetic processes, one can propose a plausible model. Let us recall that in most of solid-state elastic wave attenuation at low frequencies and high-temperatures, attenuation will follow the following form (e.g., Karato and Spetzler 1990), Q −1 ∝ ω− α

(10)

with α = 0.3±0.1 (e.g., Jackson 2000). Now since attenuation is a non-dimensional quantity, such frequency dependence must imply that the seismic wave attenuation depends on relaxation time as Q −1 ∝ ( ωτ )

−α

∝ τ−α

(11)

The relaxation time depends on the rate of some kinetic processes such as the motion of dislocations or grain-boundary sliding. Hydrogen is known to enhance all of these processes, and in most cases the characteristic time of these processes depend on hydrogen content as τ ∝ CW−r

(12)

where r is a constant that depends on the process (Table 1). Consequently, I assume that Q −1 ∝ CWαr

(13)

So in summary, the functional form for the dependence of seismic wave velocity on hydrogen content is πα −1 ⎡ 1 ⎤ v = v0∞ (1 − A ⋅ CW ) ⎢1 − cot ⋅ Q ( CW ) ⎥ 2 ⎣ 2 ⎦

(14)

Remote Sensing of Hydrogen in the Earth’s Mantle

353

Table 1. The dependence of various kinetics processes on hydrogen content. The values of r for various properties are shown where a power-law relation R ∝ CWr is assumed (R: the property e.g., strain-rate, diffusion coefficient). property

r

diffusion coefficient

~1.0 (oxygen, quartz)1

diffusion creep

~0.7-1.0 (olivine)2

dislocation creep

~1.0-1.2 (olivine, quartz)3,4,5

grain-boundary mobility

~2.0-2.3 (for wadsleyite)6 (r > 1 for olivine)7

electrical conductivity

~0.7-0.9 (for olivine, wadsleyite, ringwoodite)8

1

2

Farver and Yund (1991); Mei and Kohlstedt (2000a); 3 Kohlstedt et al. (1995); 4 Mei and Kohlstedt (2000b); 5 Karato and Jung (2003); 6 from grain-growth (Nishihara et al. 2006); 7 from dynamically recrystallized grain-size (Jung and Karato 2001a); 8 Huang et al. (2005), Wang et al. (2006)

The relations (13) and (14) provide a basis for inferring hydrogen content from seismic wave velocity and attenuation. Figure 5 illustrates how hydrogen content affects seismic attenuation for a range of parameters. I should also discuss the issue of grain-size sensitivity of seismic wave attenuation. In contrast to rather preliminary observations on the influence of hydrogen, there have been solid results on the grain-size sensitivity of seismic wave attenuation in olivine aggregates (e.g, Tan et al. 1997, 2001; Gribb and Cooper 1998; Cooper 2002; Jackson et al. 2002). These studies showed Q −1 ∝ L− αs

(15)

with α~0.25 and s~1 where L is the grain-size. I note that the functional form of grain-size dependence is the same as that of water content dependence, both cases are represented by a power-law formula, Q−1 ∝ Xβ where X is water content or inverse grain-size with β = 0.25-0.50. Therefore the influence of grain-size cannot be distinguished from that of hydrogen from the observations on attenuation. Therefore the interpretation of the attenuation tomography needs to be made on the basis of some other considerations such as geodynamic plausibility (e.g., Shito et al. 2006). I should emphasize that although more experimental observations are available

Figure 5. The relation between seismic wave attenuation and water content for a power-law relation (Eqn. 13) for α = 0.25 and r = 1-2. Jackson et al. (1992)’s data indicate that dried and undried (“wet”) dunite have Q values that are different by a factor of ~2. For exactly the same sample, this drying procedure changes the strainrate by a factor of ~10-20 (Chopra and Paterson 1984). Using the water content dependence of strain-rate, this can be translated into the difference in water content of a factor of ~10-20. Therefore I conclude that the results by Jackson et al. (1992) are consistent with the relation (13) with r = 1-2.

354

Karato

for the grain-size sensitivity than for hydrogen sensitivity, the degree to which grain-size influences attenuation in Earth’s interior is likely much smaller than that of a likely effect of water. This is simply due to the fact that grain-size in a typical upper mantle has a narrow range, ~3-10 mm (e.g., Mercier 1980; Karato 1984) whereas water content has a broad range, ~10−4-10−1 wt%, depending on the geological processes such as melting (Karato 1986; Hirth and Kohlstedt 1996). The expected change in Q corresponding to these changes will be a factor of ~1.3 for grain-size and a factor of ~6-30 for hydrogen (water) content. Therefore, as a first approximation, it is safe to ignore the influence of grain-size compared to the influence of hydrogen. A key issue in applying these relations to infer the hydrogen content in Earth’s mantle is how to distinguish the effects of hydrogen from other effects. The issue of partial melting is separately discussed in a later section, and the main conclusion is that in most of Earth’s upper mantle, there is no clear evidence so far to suggest any significant effects of partial melting on seismic wave velocities or attenuation. Here I will briefly review the methodology to distinguish the influence of hydrogen from that of temperatures and major element chemistry (for detail, see Shito et al. 2006). Key points are (i) the major element chemistry has relatively small influence on seismic wave attenuation, but has important effects on seismic wave velocities, whereas (ii) hydrogen has a large effect on seismic wave attenuation, and (iii) temperature has effects on both seismic wave velocity and attenuation. Consequently, when one plots the velocity and attenuation anomalies in a certain region, anomalies due to major element chemistry will show large variation in velocities but not in attenuation. If anomalies are due mainly to the variation in hydrogen content, then there will be large lateral variation in attenuation with relatively small velocity anomalies. Finally, if anomalies are due to temperature variation, both attenuation and velocities will show some variations (Fig. 6). To determine the hydrogen content as well as other variables from tomographic data, we need to perform a formal inversion of the observed data in terms of unknown parameters. Because there are a large number of unknowns, it is important to use as many independent observations as possible to constrain unknowns. Generally, if anomalies in seismic wave velocities, attenuation and density are obtained from one region, then one can write a general equation, m

m

j =1

j =1

δ log X i = ∑ Aij ⋅ δY j = ∑

Figure 6. A schematic diagram showing the difference in the influence of temperature, water (hydrogen) content and major element chemistry on seismic wave velocities and attenuation (after Shito et al. 2006).

∂ log X i δY j ; ∂Y j

i = 1… n

(16)

Remote Sensing of Hydrogen in the Earth’s Mantle

355

which can be inverted for five unknowns. Here δlogXi (I = 1…n) is a set of observed data such as (δ log VP δ log VS δ log QP−1 δ log QS−1 δ log ρ) and δYj (j = 1…m) is a set of unknowns such as (δT δ log X1 … δ log X n δξ), and Aij ≡ ∂logXi/∂Yj is a matrix made of partial derivatives of seismological observations with respect to physical/chemical variables (e.g., ∂logVP/∂T). This matrix is made of elements that need to be evaluated based on mineral physics. Note that in general, one has a large number of unknowns particularly because there are a large number of elements to specify the chemical composition, and the solution is non-unique. Matsukage et al. (2005) analyzed the compositional data and elasticity of constituent minerals in mantle peridotites and concluded that in most cases the chemical variation of mantle peridotite can be specified in terms of a small number of parameters. In the simplest case of peridotites in the oceanic environment, a single parameter, i.e., Mg# (mole fraction of Mg relative to Fe) is enough to specify the compositional dependence of seismic wave velocities of peridotite. With this simplification (and the assumed null effects of partial melting), one can invert seismic anomalies in terms of anomalies in water content, temperature and major element chemistry if at least three data are obtained for each point. The details of the inversion scheme are described in Shito et al. (2006). The quality of such an inversion depends strongly on the quality of seismological data and of the mineral physics-derived partial derivatives. In general, the inversion is non-linear because the values of matrix elements depend on the values of unknowns such as hydrogen content and temperature. Seismic anisotropy is in most cases caused by the lattice-preferred orientation (LPO) of elastically anisotropic minerals such as olivine (e.g., Nicolas and Christensen 1987; Chapter 21 of Karato 2006a). The possible influence of hydrogen on LPO of olivine was suggested by Karato (1995). This hypothesis was proposed based on the experimental study by Mackwell et al. (1985) who showed that the effect of hydrogen to enhance plastic deformation of olivine is anisotropic: deformation by slip systems with b = [001] is more enhanced by hydrogen than deformation by b = [100] slip systems. Consequently, Karato (1995) postulated that at high water fugacity conditions, slip systems with b = [001] (e.g., [001](010), [001](100)) might become the dominant (easiest) slip system, and consequently the LPO will be different from that usually observed at low water fugacity conditions. This hypothesis has been tested by experimental studies in my lab (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006). Based on these results, a variety of olivine LPOs have been identified (see Fig. 7), and each LPO has its own anisotropic signature (Table 2). A fabric transition is a commonly observed phenomenon in a material where multiple slip systems operate that have relatively small contrast in strength (e.g., quartz, Lister 1979). A fabric transition will occur when the relative easiness of two slip systems change, i.e., σ σ ⎛ T ⎞ ⎛ T ⎞ , ,C ⎟ = ε2 ⎜ , ,C ⎟ ε1 ⎜ W W ( ) ( , ) ( ) ( , ) T P T P T P T P µ µ ⎝ m ⎠ ⎝ m ⎠

(17)

where ε1,2 is strain-rate with the slip system 1, 2, T is temperature, Tm(P) is melting temperature, σ is stress, and µ(T,P) is shear modulus. Consequently, the conditions for a fabric transition will be given by σ ⎛ T ⎞ , , CW ⎟ = 0 F⎜ ( ) µ ( , ) T P T P ⎝ m ⎠

(18)

A boundary between different fabric types is characterized by a hyper-surface in the space defined by three variables, [T/(Tm(P)), σ/(µ(T,P)), CW]. Several points may be noted on the nature of fabric transitions. First, because the fabric boundary is defined by the relative easiness of two slip systems, strain-rate does not explicitly enter the equation for a fabric boundary. In other words, the fabric diagram determined for

356

Karato Table 2. Seismological signature of various olivine LPOs corresponding to the horizontal shear. fabric

fast S-wave polarization

vSH/vSV

A-type

parallel to flow

>1

B-type

normal to flow*

>1

C-type

parallel to flow

<1

E-type

parallel to flow

>1 (weak)

* This relation holds also for the vertical shear.

Figure 7. Various olivine deformation fabrics found in the experimental studies by (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006); pole figures on the equal area projection on the lower hemisphere. See Plate 2 for color figure.

a certain range of strain-rates can be applied to slower strain-rates without any (explicit) problems. Second, I note that a conventional power-law formula as applied to olivine does not predict stress-induced fabric transformation. The most detailed study on dislocation creep in olivine single crystal is the work by Bai et al. (1991) who showed a highly complicated creep laws for olivine single crystals, but a common feature they reported is that the stress exponent is common to all slip systems, n~3.5. In this case, Equation (18) will not contain stress as a variable and one should not expect stress-induced fabric transformations.

Remote Sensing of Hydrogen in the Earth’s Mantle

357

To see these points, let us consider the following simple power-law creep constitutive relationship, ⎛ H* ⎞ n ε1,2 = A1,2 exp ⎜ − 1,2 ⎟ ⋅ σ 1,2 ⎜ RT ⎟ ⎝ ⎠

(19)

where A1,2 are the pre-exponential factors, H1*,2 are the activation enthalpy, and n1,2 are the stress exponent for the 1,2 slip systems respectively. Equating ε1 = ε 2, one gets the conditions for the fabric boundary, viz.,

( n1 − n2 ) log σ =

H1* − H 2* A − log 1 A2 RT

(20)

This equation does not contain strain-rate, so the fabric boundary does not explicitly depend on strain-rate. Also if n1 = n2 as Bai et al. (1991) showed, then the boundary will be given by [( H1* − H 2* ) / RT ] − log( A1 / A2 ) = 0 and would not depend on stress. The latter point is inconsistent with some of the experimental results including Carter and Avé Lallemant (1970) and the B- to C-type, B- to E-type or B- to A-type transition observed in our study (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006). Therefore there is a need to go beyond a simple power-law constitutive relation to interpret the observed fabric transitions. One way is to incorporate a subtle deviation from the power-law formula observed in some of the experimental studies at high stress levels. This power-law breakdown occurs beyond a certain stress (>100-200 MPa), and can be explained by the stress dependence of activation enthalpy, H*(σ) ⎛ H * ( σ ) ⎞ n1,2 ε1,2 = A1,2 exp ⎜ − 1,2 ⎟⋅σ ⎜ RT ⎟⎠ ⎝

(21)

A simple case is a linear stress dependence, viz., H1*,2 = H1*,02 − B1,2 ⋅ σ

(22)

where B1,2 is a constant related to the resistance for dislocation motion (e.g., the Peierls stress). With this formula, Equation (20) becomes

( n1 − n2 ) log σ =

H1*0 − H 2*0 B1 − B2 A − σ − log 1 A2 RT RT

(23)

In most cases, n1 = n2 = 2, so that one has

( H1*0 − H 2*0 ) − ( B1 − B2 ) σ − RT log A1 = 0 A

(24)

2

A formula similar to (24) has been shown to be consistent with the observations on B- to C-type fabric transition observed in the lab as well as in naturally deformed peridotites (Katayama and Karato 2006). This type of transition may be classified as a stress- (and temperature-) induced fabric transition. How about the hydrogen effect? To include the effect of hydrogen on deformation in addition to the stress-dependence of activation enthalpy, the relation (19) can be extended to

(

)

⎛ H* ( ) ⎞ n ε1,2 = A1,2 + C1,2 ⋅ f Hr1,2O ⋅ exp ⎜ − 1,2 σ ⎟ ⋅ σ 1,2 2 RT ⎠ ⎝

(25)

where fH2O is the fugacity of water, C1,2 is a constant, and r1,2 is a non-dimensional constant that

358

Karato

depends on the mechanisms of hydrogen weakening. From this formula, one obtains

( n1 − n2 ) log σ =

A1 + C1 f Hr12O H1* ( σ ) − H 2* ( σ ) − log RT A2 + C2 f Hr2 O

(26)

2

This is a more general form for a fabric boundary that depends on temperature, stress and hydrogen content. If n1 = n2, and the stress dependence of activation enthalpy is weak, then we will have a hydrogen-induced fabric transition for which the boundary will follow 0=

A1 + C1 f Hr12O H1*0 − H 2*0 − log RT r A2 + C 2 f H22O

(27)

Both stress-induced transitions and hydrogen-induced transitions have been identified in olivine. The B- to C-type (and the B- to E-type, B- to A-type) transition is the stress- (and temperature-) induced transition, and the A- to E-type (and the E- to C-type) is the hydrogeninduced transition. The [T/(Tm(P)), σ/(µ(T,P)), CW] space under which various fabrics dominate is shown in Figure 8. Five types of olivine fabrics have so far been identified (see Fig. 7 that shows four of them). Among them A-, B-, C- and E-type fabrics are particularly relevant for Earth. Given the data on the distribution of crystallographic orientation, one can calculate the macroscopic elastic constants of aggregates. Seismic anisotropy resulting from these fabrics can readily be calculated from these elastic constants. Among many aspects of seismic anisotropy, those frequently used in seismology are summarized in Table 2. Notable points are:

Figure 8. A three-dimensional fabric diagram of olivine in the [T/(Tm(P)), σ/(µ(T,P)), CW] space. Dots represent the data from our lab (P = 0.5-2.0 GPa). There are a large number of data for A-type fabric that are not shown. The boundaries between A- and E-type, E-and C-type fabrics are water (hydrogen) content sensitive but not sensitive to stress nor temperature. In contrast, the boundary between B- and C-type (also B- and Eor A-type) is sensitive to stress and temperature. At low water content conditions (i.e., depleted lithosphere), A-type fabric dominates. As water (hydrogen) content increases, E-type and then C-type fabric dominates at relatively high temperatures. When temperature decreases, the domain of B-type becomes important. At low temperature (T/Tm < 0.5), B-type dominates in most cases. The results by Couvy et al. (2004) obtained at P = 11 GPa are also plotted after normalization by T/(Tm(P)), σ/(µ(T,P) (a rectangular normal to the T/Tm axis in high water (hydrogen) content region). It is seen that their results lie in the region for the C-type fabric.

Remote Sensing of Hydrogen in the Earth’s Mantle

359

(i) The fast olivine [100] axis, is normal to the flow direction for the B-type fabric, and consequently, the direction of polarization of the faster S-wave (which is reported as a shear-wave splitting observations) is normal to the flow direction. When deformation is horizontal shear, then the direction of polarization of the faster S-wave will be orthogonal to the macroscopic shear direction. Similarly, when shear occurs in a vertical plane, then the direction of polarization of faster S-wave will be in the plane but orthogonal to the flow direction. (ii) The C-type fabric will show a similar anisotropy signature to the A-type fabric in terms of shear wave splitting but vSH /vSV anisotropy will be different because the fast olivine axis is normal to the shear plane in this fabric as opposed to the case of the A-type fabric. (iii) The E-type fabric is qualitatively similar to the A-type fabric. However, because the fastest axis ([100] axis) and the slowest axis ([010] axis) are on the shear plane, the amplitude of shear wave splitting corresponding to the horizontal flow will be large for this fabric, and the amplitude of vSH /vSV anisotropy will be small. These features can be compared with seismological observations to obtain some insights into the distribution of physical and chemical conditions in Earth particularly the distribution of hydrogen. Some of the results will be discussed in the next section. I should also comment on two recent papers in which similar fabric transitions were reported but different causes were suggested. First, Holtzman et al. (2003) reported that olivine [001] axis is subparallel to the maximum elongation and olivine [010] axis is normal to the shear plane (B-type fabric) when olivine is deformed with a small amount of melt that contains a large amount of chromite or FeS. They noted that olivine fabric in their sample has strong [010] along the direction normal to the shear plane, but olivine [100] and [001] axes assume a girdle when strong shear bands are not formed (their results are different from those by Zimmerman et al. (1999) who observed a typical A-type fabric in which olivine [100] direction has a peak at a direction subparallel to the shear direction). Clear shear bands were formed in these cases (due presumably to a smaller compaction length due to the presence of chromite or FeS). After clear shear bands are formed (at larger strains), olivine [001] peak starts to strengthen along the direction normal to the shear direction. Holtzman et al. (2003) interpreted this evolution of olivine fabric in terms of deformation geometry. In their experimental setup, a significant compression component exists and therefore extrusion of sample occurs. The “B-type fabric” they observed is likely due to the anisotropic extrusion: more extrusion normal to shear direction Holtzman et al. 2003). However, the reason for this deformation geometry is due to an artifact caused by the sample geometry. First of all, extrusion is the result of compression that would not occur in truly simple shear deformation. Furthermore, the anisotropic extrusion is likely a result of oblate shape of their sample. Otherwise there is no obvious reason for the selective extrusion normal to the shear direction. Consequently, I conclude that the results reported by Holtzman et al. (2003) on LPO of olivine in partially molten olivine are due to experimental artifacts and the relevance of their observation to seismic anisotropy is highly questionable. In fact, if such an olivine fabric develops beneath a mid-ocean ridge, one would expect an anisotropic structure in the oceanic lithosphere that is totally inconsistent with observations. I conclude that the results of Holtzman et al. (2003) are unlikely to be relevant to Earth science. Second, Mainprice et al. (2005) argued that the fabric reported by Couvy et al. (2004) is due to a pressure-induced fabric transition. However, their samples contained a large amount of hydrogen (on the order of ~2000 ppm H/Si, see Couvy et al. 2004), and a comparison of the deformation conditions of their experiments with those by our group (Jung and Karato 2001b; Katayama et al. 2004; Jung et al. 2006; Katayama and Karato 2006) shows that the samples that show C-type fabric in Couvy et al. (2004)’s experiments were deformed precisely in the

360

Karato

[T/(Tm(P)), σ/(µ(T,P)), CW] conditions where the C-type fabric is developed (see Fig. 8). There have been similar results using D-DIA apparatus showing C- or B-type fabrics under highpressures but with a low water content (Ratteron, Li and Weidner, private communication, 2005). However, in these cases, the magnitude of deviatoric stress is high (~500-900 MPa). The dominance of slip systems involving b = [001] in olivine under high-stress conditions has been well known (e.g., Carter and Avé Lallemant 1970). Therefore I conclude that the observation by Couvy et al. (2004) can be naturally attributed to the high hydrogen content, and those by Ratteron, Li and Weidner are likely due to the high stresses. Obviously, there is a possibility that pressure might change the rate of deformation by different slip systems differently than the homologous temperature and normalized stress scaling would imply. However, in order to demonstrate intrinsic pressure effects on deformation fabrics, one needs to show different fabrics for two samples deformed at different pressures but otherwise nearly identical conditions (including stress levels). Such a study has not been reported to my knowledge. I therefore consider that the hypothesis of pressure-induced fabric transition in olivine proposed by Mainprice et al. (2005) and Ratteron, Li and Weidner has little experimental support and remains highly speculative at this stage. Topography and sharpness of discontinuities. Dissolution of hydrogen reduces the free energy of a material. Experimental studies show that a significantly larger amount of hydrogen can be dissolved in wadsleyite than in olivine (e.g., Young et al. 1993; Kohlstedt et al. 1996). Therefore the dissolution of hydrogen will expand the stability field of wadsleyite relative to that of olivine (the depth of “410-km” boundary will be shallower if a large amount of hydrogen is present). A more subtle effect is the change in the width of the “410-km” discontinuity with hydrogen content. When the system is considered to be a binary system, i.e., Mg2SiO4-Fe2SiO4, there is a range of pressure (at a fixed temperature) during which the phase transformation is completed. When the upper and the lower boundaries are affected by hydrogen differently, then the width of the boundary will be modified by hydrogen. The degree to which the dissolution of hydrogen affects the free energy depends on the atomistic mechanisms of hydrogen dissolution. In his paper on this topic, Wood (1995) used a model of hydrogen dissolution in wadsleyite by Smyth (1987, 1994) and in olivine by Bai and Kohlstedt (1993). These models are not consistent with the recent experimental observations, and here I use a model that is consistent with the current experimental observations. In my model, the dissolution mechanism is identical for both olivine and wadsleyite, hydrogen is dissolved mainly as (2H )×M, and only the magnitude of solubility is different between the two phases. In calculating the phase diagram, I use a simplifying assumption that the system under consideration can be treated as an ideal mixture of three components, Mg, Fe and H for the M-sites. In this approximation, the chemical potential of each component is a function of concentration of each component as µ ij = µ ij0 + RT log xij

(28)

where µij is the chemical potential, µ ij0 is the chemical potential for a pure material and xij is the mole fraction of M-sites for a component i (Mg, Fe and H) in phase j (olivine and wadsleyite). Chemical equilibrium demands 0 µ io0 + RT log xio = µ iw + RT log xiw

for i = Mg, Fe, and H

(29)

where subscript o(w) refers to olivine (wadsleyite). The M-site is shared by three elements so that

∑ xio =∑ xiw = 1 i

(30)

i

There are six unknowns (three components in a two phase system) with five equations (Eqns. 29 and 30). Another necessary parameter is the water fugacity that determines the hydrogen

Remote Sensing of Hydrogen in the Earth’s Mantle

361

content (in olivine, xHo). In the first calculation, I fixed the water fugacity, i.e., I assumed an open system behavior. I use the various values of hydrogen content in olivine (corresponding to the various values of water fugacity) and the ratio of water content in wadsleyite to olivine is fixed to be xHw/xHo = 10, which is assumed to be independent of pressure (this is justifiable because in my model, the mechanism of hydrogen dissolution is identical between olivine and wadsleyite, but not valid for the model assumed by Wood (1995)). Thermodynamic data for Mg and Fe components of olivine and wadsleyite are from Akaogi et al. (1989). The results are shown in Figure 9a, showing that the depth of transition decreases with the increase in hydrogen content. Both the upper and the lower boundaries move nearly the same amount, and the width of the transition does not change with hydrogen in this case (open system behavior). The shift of the transformation pressure by water is given by ⎛ ∂z ∆z = ⎜ ⎝ ∂CW

⎞ ⎟ ⋅ CW ⎠

(31)

with (∂z/∂CW ~ 30 km/wt% where water content is water in wadsleyite. In general, where both water content and temperature vary with lateral position, the depth to the “410-km” discontinuity will change as ⎛ ∂z ⎞ ⎛ ∂z ⎞ ∆z = ⎜ ⎟ ⋅ CW + ⎜ ⎟ ⋅ ∆T ⎝ ∂T ⎠CW ⎝ ∂CW ⎠T

(32)

where (∂z/∂T)CW ≈ 0.13 km/K is the temperature dependence of transformation depth from olivine to wadsleyite (Akaogi et al. 1989). When a phase transformation occurs in a closed system with hydrogen-under-saturated condition, then a progressive phase transformation will change the concentrations of hydrogen in each phase according to the degree of transformation (water fugacity will change with the progress of a phase transformation). Consequently, when the pressure just reaches the

Figure 9. Influence of hydrogen on the olivine-wadsleyite phase boundary. (a) Solid lines: results for an open system. Broken curves: results for a closed system with the total water content of 50% saturation. Addition of water expands the stability field of wadsleyite relative to that of olivine, leading to a shift of the phase boundary, ∆z = (∂z/∂CW)·CW with (∂z/∂CW) ~30 km/wt%. When the transformation occurs at a fixed total hydrogen content, then the broadening of the transition occurs due to the presence of hydrogen. (b) Variation of the width of the olivine-wadsleyite boundary with hydrogen saturation (saturation in olivine corresponds to ~0.1 wt% water or 1.5 × 104 ppm H/Si).

362

Karato

minimum pressure at which wadsleyite phase appears, most of the hydrogen is in olivine, whereas the hydrogen concentration in olivine decreases as the volume fraction of wadsleyite increases. Therefore the pressure at which olivine starts to transform to wadsleyite is controlled by the initial water content in olivine, and is significantly lower than the pressure at which this transformation would start in a hydrogen-free system. In contrast, the pressure at which this transformation is completed is controlled by the hydrogen content in olivine at this stage that is significantly lower than the hydrogen content at the beginning of the olivine to wadsleyite transformation (exactly the same can be applied to a case when a phase transformation starts from wadsleyite). A phase diagram for this more realistic case can be calculated from the results for a fixed hydrogen content by incorporating the mass balance requirement (see broken curves in Fig. 9a). As a result of this gradual change in hydrogen content in each phase, there is a broadening of the binary loop as first pointed out by Wood (1995). The width of the olivine to wadsleyite binary loop calculated in this way is shown in Figure 9b. Note that although the width of the boundary increases with hydrogen content similar to the model by Wood (1995), the degree to which the width increases with hydrogen content is somewhat different between the present model and the model by Wood (1995). This difference is caused mainly by the difference in the model of hydrogen dissolution and the choice of thermodynamic parameters. A few comments are in order here. (1) The model results for an open, saturated system agree with the experimental observations by Chen et al. (2002) and Smyth and Frost (2002) or an open, water-saturated system. But the closed system behavior predicted by the model has not been tested experimentally. Chen et al. (2002) and Smyth and Frost (2002) compared their results for an open system directly with those by Wood (1995) for a closed system and discussed that their results did not agree with those by Wood (1995). This is misleading. The variation of the width of transformation by hydrogen content predicted by Wood (1995) model occurs only in a closed system but not in an open system. (2) The validity of the assumption of a closed system behavior in real Earth is not necessarily obvious. If the amount of water (hydrogen) in Earth’s transition zone exceeds a critical value (~0.05 wt%, see the next section), the phase transformation from wadsleyite to olivine in a upwelling current could cause partial melting (Bercovici and Karato 2003). In this case, hydrogen can be removed from the system during the phase transformation and the assumption of a closed system behavior will be violated. (3) A comparison of the model results on the width of the “410-km” boundary with seismological observation is not straightforward due to the fact that the actual depth variation of acoustic properties in a phase loop may not a simple function of the volume fraction of each phase (Stixrude 1997). In summary, I conclude that the use of the width of the “410-km” discontinuity to infer the hydrogen content as proposed by Wood (1995) (see also van der Meijde et al. 2003) is subject to large uncertainties. Other techniques such as the use of electrical conductivity or seismic wave velocities (or attenuation) provide more robust estimate of hydrogen contents (e.g., Karato 2003; Huang et al. 2005).

Partial melting? A frequently asked question when water content is to be inferred is what about partial melting? Partial melting may also explain the majority of geophysical anomalies (high electrical conductivity, low seismic wave velocities, high attenuation). In fact, throughout the geophysical literature, these anomalies (low seismic wave velocities, high attenuation and high electrical conductivity) have often been attributed to partial melting (e.g., (Gutenberg 1954; Shankland et al. 1981)). This classical view has been questioned on various grounds. First, based on mineral physics considerations, Gueguen and Mercier (1973) proposed a solid-state mechanism of anelasticity could explain a high attenuation and low velocity zone. The follow-up studies include Minster and Anderson (1980), Karato (1993), Karato and Jung (1998) and Faul and Jackson (2005) who quantified this notion. Although low velocity and high attenuation could be attrib-

Remote Sensing of Hydrogen in the Earth’s Mantle

363

uted to sub-solidus processes, high electrical conductivity was considered to be difficult to be attributed to sub-solidus processes. Consequently, Shankland et al. (1981) argued that it is the high electrical conductivity in the asthenosphere that provides the strongest constraint on the presence of partial melt. Karato (1990) challenged this view by showing that high conductivity can also be attributed to sub-solidus process if high diffusivity and solubility of hydrogen in minerals such as olivine is taken into account. This hypothesis has now been supported by laboratory studies (e.g., Wang et al. 2006). So in short, the current status of our understanding of the role of partial melting is that all of the observed geophysical anomalies on the asthenosphere can be attributed to subsolidus processes if the role of hydrogen is included. In other words, much of the anomalous properties of the asthenosphere can be explained by a high hydrogen content as well as high temperature without invoking the influence of partial melting. One natural question, then, is if both high water content and high temperature are needed to explain geophysical anomalies of the asthenosphere, why doesn’t partial melting occur and contribute to some of these geophysical anomalies in the asthenosphere? To answer this question, one needs to understand the fact that the partial melting in the upper mantle likely occurs as a two-stage process, first (in the deeper part) by hydrogen-assisted melting and then (in the shallow part) as dry (hydrogen-free) melting. A close look at the phase diagram of the upper mantle system shows that the conditions of the asthenosphere corresponds to hydrogen-assisted melting regime, and the degree of melting there is controlled by hydrogen content and is estimated to be ~0.1-0.2% (Plank and Langmuir 1992). A larger degree of melting occurs only in the shallow region near mid-ocean ridges where temperature exceeds the dry solidus, and even if the degree of melting is well constrained to be ~10% (from the thickness of the oceanic crust), the fraction of melt that determines the degree of change in physical properties by partial melting, can be much smaller. Indeed there are strong constraints on the fraction of melt near mid-ocean ridges to be ~0.1% or less from geochemical observations (see e.g., (Spiegelman and Kenyon 1992; Spiegelman and Elliott 1993) and the MELT experiment at the east Pacific Rise failed to detect evidence of melt from seismic anisotropy (Wolfe and Solomon 1998). Also Shito et al. (2004) found that the frequency dependence of seismic wave attenuation in the upper mantle beneath the Philippine Sea is not consistent with the presence of a significant amount of melt. I conclude that the water (hydrogen) content in the asthenosphere (~0.01 wt%) is low in a petrologic sense: with this amount of water, a significant amount of melt does not exist in the majority of the asthenosphere to cause detectable change in seismic wave propagation or electrical conductivity (melt fraction is less than ~0.2%). If the degree of melting is at this level, a large fraction of total hydrogen will stay in solid minerals. This amount of water (~0.01 wt%) is, however, large compared to the concentration of defects in solid minerals at hydrogen-free conditions. Consequently, many of the solid-state processes including electrical conductivity, seismic wave attenuation are markedly affected by this much of hydrogen in the asthenosphere. In other words, the asthenosphere has anomalous physical properties because of the absence of a large fraction of melt as opposed to the conventional model as first proposed by Karato and Jung (1998). However, it is still possible that a small amount of melt exists in the asthenosphere that causes a velocity reduction without affecting attenuation (e.g., Karato 1977). These effects appear to occur only in limited regions according to the geophysical, petrological and geochemical observations (see also Shito et al. 2006).

SOME EXAMPLES Water content in the transition zone The transition zone minerals such as wadsleyite and ringwoodite are known to have large solubility of hydrogen (to ~3 wt% as water, Kohlstedt et al. 1996). However, the actual

364

Karato

hydrogen content in the transition zone was not well constrained. In fact, based on geodynamic modeling assuming whole mantle convection, Richard et al. (2002) showed that it is hard to keep a high hydrogen content in the transition zone, and the transition zone might be an empty hydrogen reservoir. In contrast, Bercovici and Karato (2003) proposed that melting at ~410km is likely and material circulation may occur in such a way that incompatible elements including hydrogen may be sequestered in the deep mantle by melting. This latter model implies that the hydrogen content in the transition should be higher than that of the upper mantle, whereas a conventional model would imply a similar hydrogen content between the upper mantle and the transition zone. Consequently, the determination of hydrogen content in the transition zone provides a good test for the models of mantle circulation. The hydrogen content in the transition zone has been inferred using seismological observations or electric conductivity. Wood (1995) inferred that hydrogen content of the transition zone is ~0.02 wt% (global average) based on the observed width of the 410-km boundary. Van der Meijde et al. (2003) applied the same method to obtain somewhat different results for the transition zone beneath Europe. Blum and Sheng (2004) used both the velocity anomalies in the transition and the topography of the 410km to infer hydrogen content and temperature anomalies. They concluded that the hydrogen content in the transition zone beneath South Africa is ~0.1 wt%. These two studies suffer major limitations. Wood (1995)’s method is based on a model of hydrogen dissolution that is not consistent with our latest knowledge. Karato (2006b) showed that a model consistent with the latest knowledge of mechanisms of hydrogen dissolution in these minerals gives different results. In addition the inferring the thickness of the boundary is not trivial as discussed in the previous section. The inference of hydrogen content from velocity anomalies and topography of the 410-km is more straightforward although estimating these two parameters from seismology involves some uncertainties (Gu et al. 2003). One major problem with the earlier work by Blum and Sheng (2004) is the ignorance of anelasticity. Huang et al. (2005) determined the relation between electrical conductivity and hydrogen content (plus temperature) for wadsleyite and ringwoodite, and by comparing these results with geophysically inferred electrical conductivity, they inferred the water content in the transition zone to be ~0.1-0.2 wt% beneath the Pacific Ocean (Fig. 10b). Electrical conductivity of the transition zone varies from one region to another. The conductivity of the transition zone in the Philippine Sea region is significantly higher than average Pacific, and also the transition zone beneath Hawaii has a higher conductivity than surrounding regions (Utada et al. 2005). It is likely that a major cause of this regional variation in conductivity is the regional variation is hydrogen content (regional variation in temperature also causes regional variation in conductivity, but the influence of temperature is less important than hydrogen, see Fig. 10a). A more robust analysis of distribution of hydrogen was made by Huang et al. (2006) who used a jump in conductivity at 410-km to infer the jump in hydrogen content. Utada et al. (2003) presented a model of electrical conductivity of the Pacific region that is characterized by a factor of ~10 jump in conductivity across the 410-km discontinuity. A similar jump at ~410km is observed in other regions (e.g., Olsen 1999; Tarits et al. 2004). This jump in conductivity can be translated into jumps in physical and chemical conditions between the transition zone and the upper mantle as +

+

σ410 − σ410

=

3 −1 8 ⎛ 410 + ⎞ 4 ⎜ fH O ⎟ 2 ⎜ ⎟ = ⎜ 410 − ⎟ ⎜ fH O ⎟ ⎠ ⎝ 2 ⎠

⎛ ⎞ ⎟ σ wad ⎜ f O410 2 ⎜ ⎟ − ⎟ ⎜ σ oli ⎜ f 410 ⎟ ⎝ O2

+

−1 8 ⎛ 410 + ⎜ Cw ⎜ − ⎜ C 410 ⎝ w ⎠

⎛ ⎞ ⎟ σ was ⎜ f O410 2 ⎜ ⎟ − ⎟ ⎜ σ oli ⎜ f 410 ⎟ ⎝ O2

3 ⎞4 ⎟ ⎟ ⎟ ⎠

(33)

where σ410 /σ410 is the contrast in electrical conductivity across the 410-km discontinuity (~10), and any quantities with 410+ (410−) means quantities just below (above) the 410-km, i.e., the uppermost transition zone (the lowest upper mantle). In writing this I note that a possible +

−

Remote Sensing of Hydrogen in the Earth’s Mantle

365

(a) Figure 10. (a) A trade-off between temperature and hydrogen (water) effect on the electrical conductivity in wadsleyite (numbers next to each curve are electrical conductivity in Sm−1), (b) a comparison of electrical conductivity profile in the transition zone for various water contents showing ~0.1-0.2 wt% of water content is consistent with the geophysical observations (after Huang et al. 2005).

(b)

temperature jump at 410-km (~50-100 K) yields less than ~30% change in conductivity, so this effect is ignored. From the relation (33), one can calculate the combinations of jumps in water (hydrogen) content and oxygen fugacity across the 410-km that are consistent with the observed jump in electrical conductivity. O’Neill et al. (1993) suggested that if the oxygen to metal ion ratio is constant throughout the transition zone and the upper mantle, then the oxygen fugacity of the transition zone should be significantly (a factor of ~103-104) higher than that of the upper mantle. Figure 11 shows the trade-off between these two factors. In order to explain the observed jump in electrical conductivity, the variation in water (hydrogen) content and oxygen fugacity must satisfy the relations shown by thick lines. Such an estimate contains some uncertainties related to the calibration of hydrogen content based on FT-IR, so I included a range of values corresponding to this uncertainty. This analysis shows that the influence of oxygen fugacity is small (because of a weak dependence of conductivity on oxygen fugacity), and if the water content were the same between the transition zone and the upper mantle, the

2

log10

2

fO410

Karato

fO410

366

log10

410 CW 410 CW

Figure 11. A diagram showing the range of combination of a jump in water content and oxygen fugacity between the upper mantle and the transition zone that is consistent with the observed jump by a factor of ~10 jump in electrical conductivity at ~400 km (after Huang et al. 2006). Two lines correspond to the two different choices of hydrogen content calibration curves.

jump in oxygen fugacity must be on the order of 108-1012, which is unacceptable. Therefore I conclude that the current geophysical observations combined with the available experimental data on electrical conductivity in olivine and wadsleyite strongly suggest that there is a jump in water content across the 410-km discontinuity. Major remaining uncertainties in this approach include (i) the role of secondary phases such as garnets, and (ii) the role of grain-size. Our current data on grain-size indicate only a small effect (a comparison of the conductivity data for olivine with ~10 µm grain-size and ~1 mm grain-size shows less than a factor of ~3 difference). However, the dependence of electrical conductivity of garnet on hydrogen content is not known. Garnet is the second most abundant mineral in both the deep upper mantle and the transition zone, and the determination of electrical conductivity of this mineral is urgent.

Distribution of hydrogen in the upper mantle Some comments on petrologic approach. A large number of petrological or geochemical data are available to infer the distribution of hydrogen in the upper mantle. This approach uses either the hydrogen contents of minerals in mantle rocks (mostly xenoliths) or the water contents in the magmas. There have been numerous publications on this topic (see e.g., Martin and Donnay 1972; Michael 1988; Jambon and Zimmermann 1990; Thompson 1992; Bell and Rossman 1992; Stolper and Newman 1994; Hirth and Kohlstedt 1996; Kurosawa et al. 1997; Wallace 1998; Jamtveit et al. 2001; Katayama et al. 2005), so I will give only a brief review for completeness. These studies have shown the following general trend: (i) MORB source regions have generally low water content (~0.01 wt%), (ii) the source regions of arc magmas have high water content (~1 wt%), and (iii) the source regions of OIB have intermediate water content (~0.02-0.05 wt%). A major advantage of this approach is that this provides a direct measurement of water (hydrogen) content from real rocks, so there is little ambiguity as to what one obtains. However, there are three major sources of uncertainties or limitations in

Remote Sensing of Hydrogen in the Earth’s Mantle

367

this approach. (i) The region from which one can infer hydrogen content is limited. (ii) In case where magmas are used to infer the water content of the mantle, in order to estimate the hydrogen content of the source region from the hydrogen content in magmas, one needs to know the partition coefficient and the degree of melting (and the mode of melting, either batch melting or fractional melting). (iii) In all cases, there is a potential addition of loss or addition of hydrogen during the transport of these rocks from Earth’s interior. The last point is particularly relevant for xenoliths, and I will address this issue in some detail. It is well known that (chemical) diffusion of hydrogen in many minerals is fast (e.g., Kohlstedt and Mackwell 1998; Mackwell and Kohlstedt 1990). Based on this observation, it is often argued that hydrogen may have escaped (or been added) very easily from the minerals during their ascent to the surface. Although this is true, there are potential complications. (i) In cases where hydrogen dissolution is coupled with the dissolution of another species whose diffusion coefficients are low, the majority of hydrogen atoms in the mineral will be preserved. This is the case of hydrogen in orthopyroxene where dissolution of hydrogen is often coupled with dissolution of Al2O3 (e.g., Rauch and Keppler 2002). (ii) Even though hydrogen diffusion is fast, not all hydrogen will escape from a mineral, but some of it will precipitate. This was observed in a laboratory experiment on olivine (unpublished data by Karato 1984). After heating olivine single crystals that contained a large amount of hydrogen (at room pressure with a controlled oxygen fugacity), I found that hydrogen precipitated as water-filled bubbles in addition to some hydrogen loss. In such a case, some fraction of original water is preserved in a different form as the original hydrogen in the crystalline lattice as defects. The precipitated water will react with the host mineral to form hydrous minerals at low temperatures. The time scale of diffusion of hydrogen-related species during the change in P-T conditions can be analyzed based on the experimental data on diffusion coefficients. The characteristic time for the motion of hydrogen-related species with a distance d is given by τ ≈ d 2/π2D where D is the relevant diffusion coefficient. If the time-scale of a given process is much less than this time scale, then a hydrogen-related species will be kept in its original form. In contrast, if the time scale is much larger than the characteristic time, then hydrogen-related species will either escape (or to be added) to a crystal or precipitate in a crystal. Figure 12

Figure 12. A diagram showing the conditions where hydrogen-related features (hydrogen content, lattice-site where hydrogen sits) can be preserved in a process with a given time and space-scale. Thick lines show the characteristic time for hydrogen diffusion, τ, for a characteristic distance, d, using τ ≈ d 2/π2D, where D is the diffusion coefficient of hydrogen (a range corresponds to temperature of 1300 to 1800 K). Shown together are some timescales and length-scales corresponding to laboratory experiments and some geological processes. If (τ,d) for a given process falls below the thick lines, diffusion is efficient and diffusion-loss or change in the speciation will occur, whereas if (τ,d) falls above the lines, these hydrogen-related features will be preserved (after Karato 2006a).

368

Karato

summarizes this analysis. I conclude that for most of xenoliths, the characteristic time (time for xenolith transport) is similar to the characteristic time for (chemical) diffusion. Therefore some hydrogen is preserved but some hydrogen is likely lost or gained during the ascent. It must also be noted, from this diagram, that the characteristic time for diffusion among various lattice-sites is very short (less than a micro second). Therefore it is impossible to preserve the lattice sites where hydrogen occupied at high P-T during the quenching in any experimental studies. This means that the room P-T measurements such as FTIR measurements provide us with the data that reflect crystallographic sites of hydrogen at room P-T or some other P-T during quenching, but these observations do not necessarily reflect the lattice site that hydrogen occupies at high P-T. A great care needs to be exercised in interpreting the results of FTIR measurements at room P-T on a sample prepared at high P and T. Hydrogen in the upper mantle. Several geophysical observations can be used to infer the distribution of hydrogen in the upper mantle. (i) From the electrical conductivity of the asthenosphere determined by Lizarrale et al. (1995) or Evans et al. (2005) (~0.1 S/m), the hydrogen content in the asthenosphere is estimated to be ~0.01 wt% (I assumed the temperature of 1600 K and used our latest experimental results on the relation between hydrogen content and electrical conductivity in olivine (Wang et al. 2006)). This value agrees well with an estimate from the petrologic approach. Obviously, the electrical conductivity in the upper mantle (asthenosphere) varies from one region to another suggesting a regional variation in hydrogen content (as well as temperature). (ii) The onset of seismic low velocity zone as determined by high-frequency body waves using reflected (or converted) waves showed nearly age-independent depth of the onset of a low velocity zone (e.g., Gaherty et al. 1996). This is in contrast to the well-known feature of age-dependent change in lithosphere thickness and velocities as inferred from surface wave studies (e.g., Forsyth 1975; Yoshii 1973). The age-independent sharp change in velocity has been attributed to a sharp contrast in hydrogen content caused by partial melting near midocean ridges (Karato and Jung 1998). I conclude that in order to explain an age-independent velocity jump at ~60-70 km depth detected by short wavelength body wave studies, hydrogeninduced anelasticity provides a good explanation. The age-dependent smooth variation in velocity detected by surface wave studies can be attributed to the temperature effects as has been known long time (e.g., Gueguen and Mercier 1973; Karato 1977; Minster and Anderson 1980 and recent similar works with new parameters Faul and Jackson 2005; Stixrude and Lithgow-Bertelloni 2005). In order to explain both surface wave and body wave observations, one needs to invoke both hydrogen and temperature effects on anelasticity. (iii) The spatial distribution of shear wave splitting in the subduction zone can be interpreted in terms of spatial variation in hydrogen content, stress and temperature. This observation does not provide strong constraint on water content, but does require some water (>20 ppm wt) in the wedge mantle (Kneller et al. 2005). (iv) A joint inversion of velocity and attenuation tomography provides a constraint on the distribution of hydrogen in the upper mantle (Fig. 13, Shito et al. 2006). Variation in hydrogen content by a factor of ~10-100 is found. A hydrogen-rich region is identified in the deep (~300400km) upper mantle beneath the Philippine Sea. This deep hydrogen-rich region is likely caused by the deep transportation of water by hydrous minerals by fast and cold subducting slabs in this region (e.g., Rüpke et al. 2004). (v) The fabric type in the asthenosphere is likely not the A-type fabric as in the lithosphere. The fabric type in the asthenosphere is either E- or C-type according to the results summarized in Figure 8. The fabric type of the asthenosphere can be identified by a close examination of seismic anisotropy in that region, which will provide a useful constraint on the hydrogen content in that layer.

Remote Sensing of Hydrogen in the Earth’s Mantle

369

Figure 13. Results of a joint inversion of velocity and attenuation tomography using mineral physicsbased inversion scheme (after Shito et al. 2006). The tomographic maps shown in Figure 2 are inverted for anomalies in the temperature (δT), major element composition [δ(Mg/Mg+Fe)] and water content anomalies (δlnCW).

Hydrogen in the lower mantle Currently very little is known about the distribution of hydrogen in the lower mantle. Generally, the solubility of hydrogen in lower mantle minerals is lower than those in the upper mantle or transition zone minerals (e.g., Bolfan-Casanova et al. 2000, 2003). One major limitation in inferring distribution of hydrogen in the lower mantle is our lack of any mineral physics data on the relationship between hydrogen content and physical properties of lower mantle minerals. However, we can make some inferences based on our knowledge on these relationships for upper mantle and transition zone minerals. (i) The direct effects of hydrogen on seismic wave velocities, namely the effects of hydrogen on unrelaxed seismic wave velocities, will be negligibly small in the lower mantle. The direct effect of hydrogen can be roughly calculated from a simple model (Karato 1995). Applying this model to lower mantle minerals where the maximum water content is ~0.1 wt% or less, then one will conclude that the maximum degree of hydrogen to change the seismic wave velocities will be less than ~0.1% for the lower mantle. (ii) Hydrogen may enhance seismic wave attenuation. Recently, Lawrence and Wysession (2006) reported a broad high attenuation region in the lower mantle beneath Asia. This region shows only modest low velocity anomalies. In this sense, the nature of velocity and attenuation anomalies in this region is similar to those in the deep upper mantle found by Shito and Shibutani (2003) and Shito et al. (2006). This suggests that hydrogen may have an important effect of enhancing seismic wave attenuation in lower mantle minerals, but very

370

Karato

little is known about the possible mechanisms by which hydrogen enhances anelasticity in lower mantle minerals. (iii) The influence of hydrogen on electrical conductivity in lower mantle minerals is difficult to assess. However, a comparison of laboratory electrical conductivity data (e.g., Katsura et al. 1998; Xu et al. 1998a) with geophysical observations suggests that hydrogen may not have a large effect to enhance conductivity in lower mantle minerals. Electrical conductivity of silicate perovskite is high due to a high concentration of point defects (Xu and McCammon 2002) and hydrogen solubility in perovskite is low (Bolfan-Casanova et al. 2000) and consequently, hydrogen may not play an important role in electrical conductivity in the lower mantle.

SUMMARY AND OUTLOOK The remote sensing of hydrogen in Earth’s mantle is an exciting new opportunity. The key concept behind this is a notion that hydrogen has strong effects on various physical properties of minerals some of which can be detected by geophysical methods. In earlier papers (e.g., Karato 1990; Karato 1995), possible relationships between hydrogen content and geophysically measurable properties such as electrical conductivity and seismic wave propagation were proposed based on then available sketchy experimental observations and theoretical models on defect-related properties. Many of these hypotheses have now been transformed to more solid models based on detailed experimental studies. However, inferring the distribution of hydrogen from geophysical observations remains challenging, and a number of issues need to be explored in more detail. Here I will list some of the mineral physics issues that are critical to make further progress in this approach: (i) The quantitative relationships between hydrogen content and seismic wave attenuation must be determined for major mantle minerals. Preliminary data exist (e.g., Jackson et al. 1992), but quantitative studies under well-controlled chemical environment are needed. (ii) The quantitative relationships between electrical conductivity and hydrogen content must be determined for all major constituent minerals in the mantle. These studies are needed particularly for orthopyroxene and garnet for the upper mantle and transition zone, and for silicate perovskite (and post-perovskite phase) and (Mg,Fe)O for the lower mantle. (iii) The role of hydrogen on non-elastic deformation and other deformation-related processes (e.g., diffusion, grain-growth) in lower mantle minerals must be clarified. Not only silicate perovskite but also the role of hydrogen in (Mg,Fe)O needs to be investigated. As I emphasized in several places, further developments in geophysical studies are also needed. Because of the trade-off among competing factors (i.e., non-uniqueness), reliable inference of distribution of hydrogen can only be made when high-resolution data are available for several geophysical parameters from the same place. This requires not only the development of dense stations, but also the development of some theoretical approach such as the simultaneous inversion of attenuation and velocity. A combination of developments in two areas and a close conversation among scientists in these two areas are a key to make further progress in this interdisciplinary area of Earth science.

ACKNOWLEDGMENTS This article is based on an extensive set of studies that I have performed with a number of colleagues during the last ten years or so. Financial support for these studies was obtained from National Science Foundation of USA and Japan Society for Promotion of Sciences.

Remote Sensing of Hydrogen in the Earth’s Mantle

371

REFERENCES Akaogi M, Ito E, Navrotsky A (1989) Olivine-modified spinel-spinel transitions in the system Mg2SiO4Fe2SiO4: calorimetric measurements, thermochemical calculation, and geophysical application. J Geophys Res 94:15671-15685 Bai Q, Kohlstedt DL (1993) Effects of chemical environment on the solubility and incorporation mechanism for hydrogen in olivine. Phys Chem Mineral 19:460-471 Bell DR, Rossman GR (1992) Water in Earth’s mantle: The role of nominally anhydrous minerals. Science 255:1391-1397 Benz H, Vidale JE (1993) Sharpness of upper-mantle discontinuities determined from high-frequency reflections. Nature 365:147-150 Bercovici D, Karato S (2003) Whole mantle convection and transition-zone water filter. Nature 425:39-44 Bhattacharya BB, Masters G, Shearer PM (1996) Global lateral variations of shear wave attenuation in the upper mantle. J Geophys Res 101:22273-22289 Blum J, Shen Y (2004) Thermal, hydrous, and mechanical states of the mantle transition zone beneath southern Africa. Earth Planet Sci Lett 217: 367-378 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett182:209-221 Bolfan-Casanova N, Keppler H, Rubie DC (2003) Water partitioning at 660 km depth and evidence for very low water solubility in magnesium silicate perovskite. Geophys Res Lett 30, doi:10.1029/2003GL017182 Carter NL, Avé Lallemant HG (1970) High temperature deformation of dunite and peridotite. Geol Soc Am Bull 81:2181-2202 Chakraborty S, Knoche R, Schulze H, Rubie DC, Dobson D, Ross NL, Angel RJ (1999) Enhancement of cation diffusion rates across the 410-kilometer discontinuity in Earth’s mantle. Science 283:362-365 Chen J, Inoue T, Yurimoto H, Weidner DJ (2002) Effect of water on olivine-wadsleyite phase boundary in the (Mg,Fe)2SiO4 system. Geophys Res Lett 29, 10.1029/2001GRL014429 Chopra PN, Paterson MS (1984) The role of water in the deformation of dunite. J Geophys Res 89:78617876 Constable SC, Parker RL, Constable CG (1987) Occam’s inversion: A practical algorithm for generating smooth models from electromagnetic sounding data. Geophysics 52:289-300 Cooper RF (2002) Seismic wave attenuation: energy dissipation in viscoelastic crystalline solids. Rev Mineral Geochem 51:253-290 Couvy H, Frost DJ, Heidelbach F, Nyilas K, Ungar T, Mackwell SJ, Cordier P (2004) Shear deformation experiments of forsterite at 11 GPa-1400°C in the multianvil apparatus. Eur J Mineral 16:877-889 de Groot-Hedlin C, Constable SC (2004) Inversion of magnetotelluric data for 2D structure with sharp resistivity contrasts. Geophysics 69:78-86 Evans RL, Hirth G, Baba K, Forsyth DW, Chave A, Makie R (2005) Geophysical evidence from the MELT area for compositional control on oceanic plates. Nature 437:249-252 Farber DL, Williams Q, Ryerson FJ (1994) Diffusion in Mg2SiO4 polymorphs and chemical heterogeneity in the mantle transition zone. Nature 371:693-695 Farver JR, Yund RA (1991) Oxygen diffusion in quartz: dependence on temperature and water fugacity. Chem Geol 90:55-70 Faul UH, Jackson I (2005) The seismological signature of temperature and grain size variations in the upper mantle. Earth Planet Sci Lett234:119-134 Forsyth WD (1975) The early structural evolution and anisotropy of the oceanic upper mantle. Geophys J Royal Astro Soc 43:103-162 Gaherty JB, Jordan TH, Gee LS (1996) Seismic structure of the upper mantle in a central Pacific corridor. J Geophys Res 101:22291-22309 Gribb TT, Cooper RF (1998) Low-frequency shear attenuation in polycrystalline olivine: grain boundary diffusion and the physical significance of the Andrade model for viscoelastic rheology. J Geophys Res 103:27267-27279 Gu YJ, Dziewonski AM, Ekström G (2003) Simultaneous inversion for mantle velocity and topography of transition zone discontinuities. Geophys J Inter 154:559-583 Gueguen Y, Mercier JM (1973) High attenuation and low velocity zone. Phys Earth Planet Interior 7:39-46 Gung Y, Romanowicz B (2004) Q tomography of the upper mantle using three-component long-period waveforms. Geophys J Int 157:813-830 Gutenberg B (1954) Low-velocity layers in the earth’s mantle. Bull Seismo Soc Am 65:337-348 Hier-Majumder S, Anderson IM, Kohlstedt DL (2005) Influence of protons on Fe-Mg interdiffusion in olivine. J Geophys Res 110, doi:10.1029/2004JB003292 Hirth G, Evans RL, Chave AD (2000) Comparison of continental and oceanic mantle electrical conductivity: Is Archean lithosphere dry? Geochem Geophys Geosyst 1, doi:10.1029/2000GC000048

372

Karato

Hirth G, Kohlstedt DL (1996) Water in the oceanic upper mantle - implications for rheology, melt extraction and the evolution of the lithosphere. Earth Planet Sci Lett 144:93-108 Huang X, Xu Y, Karato S (2005) Water content of the mantle transition zone from the electrical conductivity of wadsleyite and ringwoodite. Nature 434:746-749 Huang X, Xu Y, Karato S (2006) A wet mantle conductor? (Reply). Nature 439:E3-E4 Inoue T, Weidner DJ, Northrup PA, Parise JB (1998) Elastic properties of hydrous ringwoodite (γ-phase) of Mg2SiO4. Earth Planet Sci Lett160:107-113 Jackson I (2000) Laboratory measurements of seismic wave dispersion and attenuation: recent progress. In: Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, p 265-289 Jackson I, Fitz Gerald JD, Faul UH, Tan BH (2002) Grain-size sensitive seismic-wave attenuation in polycrystalline olivine. J Geophys Res 107(B12), doi: 10.1029/2001JB001225 Jackson I, Paterson MS, Fitz Gerald JD (1992) Seismic wave dispersion and attenuation in Åheim dunite. Geophys J Int 108:517-534 Jacobsen SD, Smyth JR, Spetzler HA, Holl CM, Frost DJ (2004) Sound velocities and elastic constants of ironbearing hydrous ringwoodite. Phys Earth Planet Interior 143/144:47-56 Jambon A, Zimmermann JL (1990) Water in oceanic basalts: evidence for dehydration of recycled crust. Earth Planet Sci Lett101:323-331 Jamtveit B, Brooker R, Brooks K, Larsen LM, Pedersen T (2001) The water content of olivines from the North Atlantic volcanic province. Earth Planet Sci Lett186:401-425 Jung H, Karato S (2001a) Effect of water on the size of dynamically recrystallized grains in olivine. J Struct Geol 23:1337-1344 Jung H, Karato S (2001b) Water-induced fabric transitions in olivine. Science 293:1460-1463 Jung H, Katayama I, Jiang Z, Hiraga T, Karato S (2006) Effects of water and stress on the lattice preferred orientation in olivine. Tectonophysics 421:1-22 Karato S (1977) Rheological Properties of Materials Composing the Earth’s Mantle. Ph. D, University of Tokyo Karato S (1984) Grain-size distribution and rheology of the upper mantle. Tectonophysics 104:155-176 Karato S (1986) Does partial melting reduce the creep strength of the upper mantle? Nature 319:309-310 Karato S (1989a) Defects and plastic deformation in olivine. In: Rheology of Solids and of the Earth. Karato S, Toriumi M (eds) Oxford University Press, p 176-208 Karato S (1989b) Plasticity-crystal structure systematics in dense oxides and its implications for creep strength of the Earth’s deep interior: a preliminary result. Phys Earth Planet Interior 55:234-240 Karato S (1990) The role of hydrogen in the electrical conductivity of the upper mantle. Nature 347:272-273 Karato S (1993) Importance of anelasticity in the interpretation of seismic tomography. Geophys Res Lett 20: 1623-1626 Karato S (1995) Effects of water on seismic wave velocities in the upper mantle. Proc Japan Academy 71: 61-66 Karato S (2003) Mapping water content in Earth’s upper mantle. In: Inside the Subduction Factory. Eiler JE (ed) American Geophysical Union, p 135-152 Karato S (2006a) Hydrogen-related defects and their influence on the electrical conductivity and plastic deformation of mantle minerals: A critical review. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (eds), American Geophysical Union, in press Karato S (2006b) Microscopic models for the influence of hydrogen on physical and chemical properties of minerals. In: Superplume: Beyond Plate Tectonics. Yuen DA, Maruyama S, Karato S, Windley BF (ed), Springer, in press Karato S, Jung H (1998) Water, partial melting and the origin of seismic low velocity and high attenuation zone in the upper mantle. Earth Planet Sci Lett157:193-207 Karato S, Jung H (2003) Effects of pressure on high-temperature dislocation creep in olivine polycrystals. Philos Mag A 83:401-414 Karato S, Paterson MS, Fitz Gerald JD (1986) Rheology of synthetic olivine aggregates: influence of grain-size and water. J Geophys Res 91:8151-8176 Karato S, Spetzler HA (1990) Defect microdynamics in minerals and solid state mechanisms of seismic wave attenuation and velocity dispersion in the mantle. Rev Geophys 28:399-421 Katayama I, Jung H, Karato S (2004) New type of olivine fabric at modest water content and low stress. Geology 32:1045-1048 Katayama I, Karato S (2006) Effects of temperature on the B- to C-type fabric transition in olivine. Phys Earth Planetary Interior 157:33-45 Katayama I, Karato S, Brandon M (2005) Evidence of high water content in the deep upper mantle inferred from deformation microstructures. Geology 33:623-616

Remote Sensing of Hydrogen in the Earth’s Mantle

373

Katsura T, Sato K, Ito E (1998) Electrical conductivity of silicate perovskite at lower-mantle conditions. Nature 395:493-495 Kneller EA, van Keken PE, Karato S, Park J (2005) B-type olivine fabric in the mantle wedge: Insights from high-resolution non-Newtonian subduction zone models. Earth Planet Sci Lett 237:781-797 Kohlstedt DL, Evans B, Mackwell SJ (1995) Strength of the lithosphere: constraints imposed by laboratory measurements. J Geophys Res 100:17587-17602 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohlstedt DL, Mackwell SJ (1998) Diffusion of hydrogen and intrinsic point defects in olivine. Z Phys Chem 207:147-162 Kurosawa M, Yurimoto H, Sueno S (1997) Patterns in the hydrogen and trace element compositions of mantle olivines. Phys Chem Minerals 24:385-395 Lawrence JF, Wysession ME (2006) Seismic evidence for subduction-transported water in the lower mantle. In: Earth’s Deep Water Cycle. Jacobsen SD, Lee Svd (ed) American Geophysical Union, in press Lister GS (1979) Fabric transitions in plastically deformed quartzites: competition between basal, prism and rhomb systems. Bull Mineral 102:232-241 Lizarrale D, Chave A, Hirth G, Schultz A (1995) Northeastern Pacific mantle conductivity profile from longperiod magnetotelluric sounding using Hawaii-to-California cable data. J Geophys Res 100:1783717854 Long MD, van der Hilst RD (2005) Upper mantle anisotropy beneath Japan from shear wave splitting. Phys Earth Planet Interior 151:206-222 Mackwell SJ, Kohlstedt DL (1990) Diffusion of hydrogen in olivine: implications for water in the mantle. J Geophys Res 95:5079-5088 Mackwell SJ, Kohlstedt DL, Paterson MS (1985) The role of water in the deformation of olivine single crystals. J Geophys Res 90:11319-11333 Mainprice D, Tommasi A, Couvy H, Cordier P, Frost DJ (2005) Pressure sensitivity of olivine slip systems and seismic anisotropy of Earth’s upper mantle. Nature 433:731-733 Martin RF, Donnay G (1972) Hydroxyl in the mantle. Am Mineral 57:554-570 Masters G, Laske G, Bolton H, Dziewonski AM (2000) The relative behavior of shear velocity, bulk sound speed, and compressional velocity in the mantle: implications for chemical and thermal structure. In: Earth’s Deep Interior. Karato AMFS, Liebermann RC, Masters G, Stixrude L (eds), American Geophysical Union, p 63-87 Matsukage KN, Nishihara Y, Karato S (2005) Seismological signature of chemical evolution of Earth’s upper mantle. J Geophys Res 110, doi:10.1029/2004JB003504 Mei S, Kohlstedt DL (2000a) Influence of water on plastic deformation of olivine aggregates, 1. Diffusion creep regime. J Geophys Res 105:21457-21469 Mei S, Kohlstedt DL (2000b) Influence of water on plastic deformation of olivine aggregates, 2. Dislocation creep regime. J Geophys Res 105:21471-21481 Mercier J-CC (1980) Magnitude of the continental lithospheric stresses inferred from rheomorphic petrology. J Geophys Res 85:6293-6303 Michael PJ (1988) The concentration, behavior and storage of H2O in the suboceanic upper mantle: implications for mantle metasomatism. Geochim Cosmochim Acta 52:555-566 Minster JB, Anderson DL (1980) Dislocations and nonelastic processes in the mantle. J Geophys Res 85: 6347-6352 Nakajima J, Hasegawa A (2004) Shear-wave polarization anisotropy and subduction-induced flow in the mantle wedge of northern Japan. Earth Planet Sci Lett 225:365-377 Nicolas A, Christensen NI (1987) Formation of anisotropy in upper mantle peridotite: a review. In: Composition, Structure and Dynamics of the Lithosphere-Asthenosphere System. Fuchs K, Foridevaux C (eds) American Geophysical Union, p 111-123 Nishihara Y, Shinmei T, Karato S (2006) Grain-growth kinetics in wadsleyite: effects of chemical environment. Phys Earth Planet Interior 154:30-43 O’Neill HStC, McCammon CA, Canil D, Rubie DC, Ross CRI, Seifert F (1993) Mössbauer spectroscopy of mantle transition zone phases and determination of minimum Fe3+ content. Am Mineral 78:456-460 Olsen N (1999) Long-period (30 days - 1 year) electromagnetic sounding and the electrical conductivity of the lower mantle beneath Europe. Geophys J Int 138:179-187 Parker RL (1980) The inverse problem of electromagnetic induction: Existence and construction of solutions based upon incomplete data. J Geophys Res 85:4421-4428 Plank T, Langmuir AH (1992) Effects of melting regime on the composition of the oceanic crust. J Geophys Res 97:19749-19770 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536

374

Karato

Richard G, Monnereau M, Ingrin J (2002) Is the transition zone an empty water reservoir? Inference from numerical model of mantle dynamics. Earth Planet Sci Lett 205:37-51 Romanowicz B, Durek JJ (2000) Seismological constraints on attenuation in the Earth: a review. In: Earth’s Deep Interior. Karato AMFS, Liebermann RC, Masters G, Stixrude L (eds), American Geophysical Union, p 161-179 Rüpke LH, Phipps Morgan J, Hort M, Connolly JAD (2004) Serpentine and the subduction zone water cycle. Earth Planet Sci Lett223:17-34 Shankland TJ, O’Connell RJ, Waff HS (1981) Geophysical constraints on partial melt in the upper mantle. Rev Geophys Space Phys 19:394-406 Shearer PM (2000) Upper mantle discontinuities. In: Earth’s Deep Interior: Mineral Physics and Tomography from the Atomic to the Global Scale. Karato S, Forte AM, Liebermann RC, Masters G, Stixrude L (eds) American Geophysical Union, p 115-131 Shito A, Karato S, Matsukage KN, Nishihara Y (2006) Toward mapping water content, temperature and major element chemistry in Earth’s upper mantle from seismic tomography. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (eds), American Geophysical Union, in press Shito A, Karato S, Park J (2004) Frequency dependence of Q in Earth’s upper mantle inferred from continuous spectra of body wave. Geophys Res Lett 31, doi:10.1029/2004GL019582 Shito A, Shibutani T (2003a) Anelastic structure of the upper mantle beneath the northern Philippine Sea. Phys Earth Planet Inter 140:319-329 Shito A, Shibutani T (2003b) Nature of heterogeneity of the upper mantle beneath the northern Philippine Sea as inferred from attenuation and velocity tomography. Phys Earth Planet Interior 140:331-341 Simpson F (2001) Resistance to mantle flow inferred from the electromagnetic strike of the Australian upper mantle. Nature 412:632-635 Simpson F (2002) Intensity and direction of lattice-preferred orientation of olivine: are electrical and seismic anisotropies of the Australian mantle reconcilable? Earth Planet Sci Lett 203:535-547 Simpson F, Tommasi A (2005) Hydrogen diffusivity and electrical anisotropy of a peridotite mantle. Geophys J Int 160:1092-1102 Smith GP, Wiens DA, Fischer KM, Dorman LM, Hildebrand JA (2001) A complex pattern of mantle flow in the Lau back-arc. Science 292:713-716 Smyth JR (1994) A crystallographic model for hydrous wadsleyite (β-Mg2SiO4): An ocean in the Earth’s interior. Am Mineral 79:1021-1024 Smyth JR (1987) β−Mg2SiO4: a potential host for water in the mantle? Am Mineral 75: 1051-1055 Smyth JR, Frost DJ (2002) The effect of water on the 410-km discontinuity: An experimental study. Geophys Res Lett 29: 10.129/2001GL014418 Spiegelman M, Elliott T (1993) Consequences of melt transport for uranium series disequilibrium in young lavas. Earth Planet Sci Lett 118:1-20 Spiegelman M, Kenyon PM (1992) The requirement of chemical disequilibrium during magma migration. Earth Planet Sci Lett 109:611-620 Stixrude L (1997) Structure and sharpness of phase transitions and mantle discontinuities. J Geophys Res 102: 14835-14852 Stixrude L, Lithgow-Bertelloni C (2005) Mineralogy and elasticity of the oceanic upper mantle: Origin of the low-velocity zone. J Geophys Res 110, doi:10.1029/2004JB002965 Stolper EM, Newman S (1994) The role of water in the petrogenesis of Mariana trough magmas. Earth Planet Sci Lett121:293-325 Suetsugu D, Inoue T, Yamada A, Zhao D, Obayashi M (2006) Towards mapping three-dimensional distribution of water in the transition zone from P-wave velocity tomography and 660-km discontinuity depths. In: Earth’s Deep Water Cycle. Jacobsen SD, van der Lee S (ed) American Geophysical Union, in press. Tan B, Jackson I, Fitz Gerald JD (1997) Shear wave dispersion and attenuation in fine-grained synthetic olivine aggregates: preliminary results. Geophys Res Lett 24:1055-1058 Tan B, Jackson I, Fitz Gerald JD (2001) High-temperature viscoelasticity of fine-grained polycrystalline olivine. Phys Chem Minerals 28:641-664 Tarits P, Hautot S, Perrier F (2004) Water in the mantle: Results from electrical conductivity beneath the French Alps. Geophys Res Lett 31, doi:10.1029/2003GL019277 ten Grotenhuis SM, Drury MR, Peach CJ, Spiers CJ (2004) Electrical properties of fine-grained olivine: evidence for grain-boundary transport. J Geophys Res 109, doi:10.1029/2003JB002799 Thompson AB (1992) Water in the Earth’s upper mantle. Nature 358:295-302 Utada H, Goto T, Koyama T, Shimizu H, Obayashi M, Fukao Y (2005) Electrical conductivity in the transition zone beneath the North Pacific and its implications for the presence of water. EOS, Trans Am Geophys Union 86:DI41A-1260 Utada H, Koyama T, Shimizu H, Chave AD (2003) A semi-global reference model for electrical conductivity in the mid-mantle beneath the north Pacific region. Geophys Res Lett 30, doi:10.1029/2002GL016092

Remote Sensing of Hydrogen in the Earth’s Mantle

375

van der Meijde M, Marone F, Giardini D, van der Lee S (2003) Seismic evidence for water deep in Earth’s upper mantle. Science 300:1556-1558 Wallace P (1998) Water and partial melting in mantle plumes: inferences from the dissolved H2O concentrations of Hawaii basaltic magmas. Geophys Res Lett 25:3639-3642 Wang D, Mookherjee M, Xu Y, Karato S (2006) The effect of hydrogen on the electrical conductivity in olivine. Nature, submitted. Wolfe CJ, Solomon SC (1998) Shear-wave splitting and implications for mantle flow beneath the MELT region of the East Pacific. Science 280:1230-1232 Wood BJ (1995) The effect of H2O on the 410-kilometer seismic discontinuity. Science 268:74-76 Xu Y, McCammon C (2002) Evidence for ionic conductivity in lower mantle (Mg,Fe)(Si,Al)O3 perovskite. J Geophys Res 107, doi:10.1029/2001JB000677 Xu Y, McCammon C, Poe BT (1998a) Effect of alumina on the electrical conductivity of silicate perovskite. Science 282:922-924 Xu Y, Poe BT, Shankland TJ, Rubie DC (1998b) Electrical conductivity of olivine, wadsleyite, and ringwoodite under upper-mantle conditions. Science 280:1415-1418 Yoshii T (1973) Upper mantle structure beneath the north Pacific and marginal seas. J Phys Earth 21:313-328 Young TE, Green HW II, Hofmeister AM, Walker D (1993) Infrared spectroscopic investigation of hydroxyl in β-(Mg,Fe)2SiO4 and coexisting olivine: implications for mantle evolution and dynamics. Phys Chem Mineral 19:409-422 Zimmerman MR, Zhang S, Kohlstedt DL, Karato S (1999) Melt distribution in mantle rocks deformed in shear. Geophys Res Lett 26:1505-1508

3

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 53-66, 2006 Copyright © Mineralogical Society of America

Structural Studies of OH in Nominally Anhydrous Minerals Using NMR Simon C. Kohn Department of Earth Sciences University of Bristol Bristol, BS8 1RJ, United Kingdom e-mail: [email protected]

INTRODUCTION Nuclear magnetic resonance (NMR) is a technique that is used very widely throughout science and medicine. There are approaching 20 journals devoted exclusively to NMR and magnetic resonance imaging, and well in excess of 10,000 papers published annually which involve this family of techniques. It has found extraordinarily diverse applications, and its applications to inorganic solids such as minerals represent a very small part of NMR as a whole. There had been occasional NMR studies of minerals since the discovery of the NMR effect in 1946, but until the early 1980s, the large width of NMR resonances in the solid state precluded widespread application of the technique. The most important factor in the application of NMR to minerals was the development of magic angle spinning (MAS), a technique for narrowing lines in solid state NMR. More recently the development of more complex sample spinning arrangements and multiple pulse methods together with the availability of ever higher magnetic fields opens up more and more possibilities for NMR in mineralogy and geochemistry. NMR is far too diverse and complex for the whole subject to be covered here, so this review will be tightly focused on aspects which relate to understanding the structural role of water in nominally anhydrous minerals. Two review papers (Kirkpatrick 1988; Stebbins 1988) in the “Reviews in Mineralogy” volume on Spectroscopic Methods in Mineralogy and Geology provide an excellent starting point for Earth scientists wishing to learn more about NMR in general. More recent reviews aimed at Earth scientists include those by Fechtelkord (2004) and Kohn (2004), more general reviews of NMR of inorganic solids include Engelhardt and Michel (1987) and MacKenzie and Smith (2002). The web page maintained by J.P. Hornak (http://www.cis.rit.edu/htbooks/nmr/) is also an extremely good resource. NMR is a multinuclear technique, which means that a sample containing several different NMR active nuclei can be studied independently using NMR of each of the nuclei. Thus, for example, the zeolite natrolite with a composition Na2[Al2Si3O10]·2H2O could be studied using 23Na, 27Al, 29Si and 1H NMR, and each would provide different and complementary information. If suitably isotopically enriched samples were available, 17O and 2H spectra would provide additional distinct information. However, if we are interested in the structural aspects of H dissolved in nominally anhydrous minerals the situation is rather different. In this case the concentration of water is usually too low to have a significant influence on the local environments of the major components. For example the H/Si ratio in enstatite containing 200 ppm H2O is 2.2 × 10−3. 1H NMR is therefore the most promising NMR technique for NAMs except in particular circumstances which will be described later. There are a number of problems that have limited the use of NMR in studying NAMs. In this chapter the principles of solid state NMR will be briefly outlined, then both the benefits and problems of 1H MAS NMR applied to the problem of water in NAMs will be discussed. A review of published studies on 1529-6466/06/0062-0003$05.00

DOI: 10.2138/rmg.2006.62.3

54

Kohn

1 H MAS NMR will follow and finally other NMR methods, such as static 1H NMR and use of other NMR active nuclei, and prospects for future developments will be discussed.

PRINCIPLES OF SOLID STATE NMR The references given in the introduction give a much more complete introduction to NMR spectroscopy, so the following description should only be considered to be a highly simplified outline of the most important concepts which are relevant for NMR studies of NAMs. Some nuclei (depending on the numbers of protons and neutrons in the nucleus) have a quantized property known as spin. The NMR effect is based on the nuclear spin, and can be conceptualized by either classical or quantum mechanical descriptions. In the quantum mechanical view the spin energy levels become non-degenerate in an external magnetic field, and the difference between the energy levels is given by ∆E =

µB0 I

(1)

where B0 is the applied magnetic field and I is the nuclear spin. The nuclear magnetic moment, µ, is given by µ=

γ hI 2π

(2)

where γ is the magnetogyric ratio (a constant which is specific for each nucleus) and h is Planck’s constant. The energy differences, ∆E, are typically in the radio frequency range for the magnetic fields used in NMR spectroscopy, so transitions between the energy levels can be stimulated by irradiating with the appropriate tuned RF frequency. The relaxation of the nuclear spins back to equilibrium is then measured by emission of the same RF frequency. In the classical description, the nuclei are considered to behave like tiny bar magnets in a magnetic field. At equilibrium, the torque exerted on the magnetic moments causes them to precess around the direction of the B0 magnetic field with a frequency known as the Larmor frequency; this leads to a net bulk magnetization along the same direction as the applied B0 magnetic field. Applying a relatively small additional B1 magnetic field, by irradiation with a radio frequency field oscillating at close to the Larmor frequency, applies an additional torque to this bulk magnetization. The result is that the bulk magnetization moves away from the direction of the B0 field. When the B1 field is turned off, two things happen i) local differences in the B0 field mean that individual spins experience slightly different local fields. Hence when the magnetization is in the transverse plane phase coherence is lost as the differing Larmor precession frequencies cause the spins to fan out. The timescale of this process is known as the spin-spin relaxation rate, with a time constant T2. ii) the bulk magnetization is built up to re-attain equilibrium. This process is known as spin-lattice relaxation, with a time constant T1. Spin lattice relaxation is usually much slower than spin-spin relaxation in solids, i.e., T1 > T2. NMR of liquids usually gives narrow resonances such that chemically inequivalent sites in a molecule can be identified. The resonances are narrow, because the rate of the tumbling motion of the molecules is fast compared with the strength of all the line broadening interactions, so anisotropy in the NMR interactions, and coupling between nuclei are averaged. In solids, however, there are a range of interactions that cause line broadening, and obscure most of the useful information in the spectra. Most NMR measurements that have been reported in the mineralogical literature have, therefore, used the magic angle spinning (MAS) technique. In this technique the sample is packed into a ceramic rotor (typically 2-7 mm in diameter), and a polymer or ceramic cap is inserted into the end of the rotor to contain the

Structural Studies of OH in NAM’s Using NMR

55

sample. The rotor is then placed into an NMR probe and physically spun at very high speeds using a compressed gas system. The angle between the axis of rotation of the sample and the direction of the B0 field is set to be 54.7°, the magic angle. At this angle, the term (3cos2θ−1), which appears in many of the line broadening functions, becomes zero. Magic angle spinning at sufficiently high speeds narrows most of the possible broadening interactions. A single NMR resonance can potentially provide information from its position (frequency or chemical shift), width and intensity. The nature of the information contained in these three parameters will vary from nucleus to nucleus, but the case of 1H MAS NMR is reviewed below.

Positions of 1H MAS NMR resonances

Isotropic chemical shift (ppm)

One of the most useful features of NMR is that small differences in the chemical environment around a nucleus result in a slightly modified magnetic field, and hence the position of a 1H NMR peak. These differences in resonance frequency are expressed as a chemical shift with units of ppm and reflect the chemical environment of the hydrogen. 1H chemical shifts can be measured most easily using fast magic angle spinning. In nominally anhydrous silicate minerals one would expect H to always be strongly bonded to an adjacent oxygen, and several studies on hydroxyl containing minerals and other materials have shown that there is a strong correlation between 1H chemical shift and O-H distance, rOH (Brunner and 18 Sternberg 1998). Hydroxyl groups in silicates may be oriented towards other oxygens in the 16 structure and form hydrogen bonds of variable 14 strength. It has been shown that the O-H..O distance (rO..O), which reflects the strength 12 of a hydrogen bond, is also correlated with 10 both rOH and 1H chemical shift (Eckert et al. 1988). The correlation between rO..O and 1H 8 shift shown in Figure 1 is particularly useful as 6 rO..O is measurable from X-ray diffraction data, whereas rOH requires neutron diffraction data. If 4 several peaks are observable in a spectrum, rO..O can be deduced for each hydrogen environment. 2 These correlations between 1H NMR shift 0 and structure are analogous to those for O-H stretching frequency (Libowitzky 1999), so in 2.4 2.5 2.6 2.7 2.8 2.8 3.0 3.1 principle NMR can give the same structural d(O-H..O)/angstroms information as FTIR spectroscopy, even though 1 Figure 1. H NMR chemical shift as a function the latter is much more frequently used. The of O-H...O distance for various crystalline key to extracting this structural information is compounds (after Eckert et al. 1988). that the width of the resonances must be small 1 compared with the shift range of H in NAMs.

Widths of 1H MAS NMR resonances Magic angle spinning dramatically reduces the widths of 1H NMR spectra of solids because it averages homonuclear and heteronuclear dipolar coupling between protons and also chemical shift anisotropy. Nonetheless a variety of mechanisms are involved in determining the linewidth of a 1H MAS NMR resonance, and these are important in interpreting the spectra.

i)

Chemical shift dispersion. This is the range of chemical shifts present, and reflects the actual distribution of H environments in the sample. The linewidth in ppm resulting from chemical shift dispersion is independent of magnetic field.

56

Kohn ii) Residual

iii) Dipolar coupling to quadrupolar

2000 1800

Full width at half maximum (Hz)

homonuclear dipolar broadening. If protons are close together in the structure of a material, magic angle spinning at conventional speeds of 5-15 kHz is insufficient to completely remove the dipolar coupling. This effect can be seen in Figure 2 where the residual linewidth decreases as a function of spinning speed for two minerals with high H-density. For NAMs, dipolar coupling should be completely removed by magic angle spinning unless hydrogens are clustered, for example in water molecules, as pairs of H charge balancing a divalent cation vacancy or as a hydrogarnet substitution. If dipolar coupling is not completely removed, spinning sidebands will be observed.

1600

Datolite

1400 1200 1000

Pyrophyllite

800 600 400 200 1

2

6 4 5 3 Spinning speed (Hz)

7

8

Figure 2. The full width at half maximum (FWHM) of 1H MAS spectra of datolite and pyrophyllite as a function of spinning speed (after Yesinowski et al. 1988). The decrease in width with increasing spinning speed for these minerals indicates that the homonuclear dipolar coupling is incompletely averaged by the spinning speeds that were available at that time.

nuclei. Quadrupolar nuclei (those with I > ½) are subject to additional line broadening mechanisms compared with dipolar nuclei, and these are not completely averaged by MAS. However, quadrupolar interactions are reduced relative to the chemical shift interaction at higher magnetic fields. Protons which undergo strong dipolar coupling to quadrupolar nuclei can be broadened, but in this case the linewidth (in Hz) will also decrease with increasing magnetic field.

iv) Paramagnetic samples. Coupling between the nuclear dipole and the much larger dipole of unpaired electrons in paramagnetic samples (such as iron-bearing, natural mantle minerals) can cause enormous broadening of NMR signals. Although there are some circumstances where NMR of paramagnetic samples is possible, NMR of natural iron-bearing samples does not generally give useable signals.

v) Motional narrowing. Protons that undergo rapid isotropic motion (such as those in macroscopic fluid inclusions) will give very narrow lines. Even certain restricted motions, such as rotation of water molecules about a single axis can be effective in narrowing lines.

Intensity of 1H MAS NMR resonances In favorable circumstances, and if the NMR experiment is performed using the correct conditions, the area of an NMR resonance is proportional to the number of resonating nuclei. Thus if a 1H spectrum consists of several peaks, the relative areas of the peaks can be used directly to determine the relative abundances of the different H environments which give rise to the different peaks. Furthermore if the area of the resonance for a standard with a known concentration of H is measured, a comparison with the area of the resonance for an unknown will yield the absolute number of H nuclei in the sample, as long as the masses of the standard and sample are taken into account. Obviously the standard that is used should be very well characterized, as any uncertainty in the water concentration of the standard will be reflected in the uncertainty in the water concentrations of the unknowns. The standard should also have a similar magnetic

Structural Studies of OH in NAM’s Using NMR

57

susceptibility to the samples, and have a relatively low water concentration if possible. These two factors are not thought to be critical compared with the other difficulties outlined below, but more work on the effect of using different standard materials should probably be undertaken. In practice there are a number of reasons why the observed NMR signal may not be fully quantitative. i)

Resonance may be broadened, by residual dipolar coupling or interaction with paramagnetic centers, to the point where it is unobservable. A broad line in an NMR spectrum (frequency domain) corresponds to a rapidly decaying signal in the time domain. Even with fast digitization, the signal of a broad line may only correspond to a few data points, and these may be obscured by ringdown of the probe, which occurs for a few µs after the pulse.

ii) The relaxation delay between pulses may not be sufficiently long. To obtain quantitative spectra it is crucial to allow a sufficiently long time between pulses to allow complete spin-lattice relaxation for all the nuclei. Materials with high concentrations of hydrogen usually have quite short values of T1, and since the signal is extremely large, only a few repetitions of the pulse and acquisition cycle are required. In the case of dilute hydrogen, T1 can be much longer, and since many more cycles are required to get an acceptable signal, it can be difficult, but nonetheless crucial to ensure that the relaxation delay is sufficiently long. This can be done either by a rigorous T1 determination, or simply by increasing the relaxation delay (while keeping other parameters constant) until the signal intensity per pulse becomes constant. iii) The B1 field may not be homogeneous. The B1 field is the magnetic field generated by applying the pulse of RF to the NMR coil. Over a long distance this cannot be homogeneous, especially at the ends of the coil, and the best that can be aimed for is that it is homogeneous over the sample volume. The true homogeneity can be tested by checking whether all parts of the sample have the same 90° pulse length, but for practical purposes the effect of B1 inhomogeneity on signal size can be tested by quantifying spectra collected with vary amounts of sample in a rotor. Figure 3 shows the integrated signal (after background subtraction) plotted as a function of sample mass for the specialized 1H MAS NMR probe used in our laboratory. This probe is based around a DOTY Si3N4 stator, and gives an excellent linear relationship between sample mass and integrated signal.

Figure 3. The integrated area of 1H MAS NMR spectra of a hydrous glass standard plotted as a function of sample mass within a 5 mm MAS rotor. The linear plot illustrates the quantitative nature of NMR, and that for this NMR probe the B1 field is effectively constant over the sample volume.

!REAOFSPECTRUMARBITRARYUNITS

-ASSOFSAMPLEMG

58

Kohn APPLICATION OF 1H MAS NMR TO NOMINALLY ANHYDROUS MINERALS

Attractive features of 1H MAS NMR for studies of NAMs There are a number of characteristics of 1H NMR in particular, which, in principle, can be exploited to determine both the concentration and environment of H in NAMs. 1H has the highest magnetogyric ratio of all nuclei and is nearly 100% abundant. Furthermore, it has a spin of ½, and is therefore generally free of all the complications and line-broadening experienced by quadrupolar nuclei. Therefore, of all the NMR active nuclei, it is the most suitable for measurement at low concentrations, and should be a valuable tool for studying the concentration and environment of low concentrations of H in NAMs. Concentrations of less than 1 ppm H2O can be detected in favorable circumstances. As described earlier, NMR is an intrinsically quantitative technique, so as long as one suitable standard is available, no further calibration for different bulk compositions is required. This a major advantage over FTIR. Of course, NMR is also an element specific technique, so all the intensity in a 1H NMR spectrum corresponds to H in the sample. This contrasts with vibrational spectroscopy, where a peak in a spectrum could be result from an OH vibration, or a combination of other vibrations of the structure. NMR will also provide information on all dissolved H, unlike FTIR, which is only sensitive to O-H species. This point could be important if it turns out that other species such as Si-H or other hydrides, organic molecules or H2 are significant for the H budget in NAMs, as proposed by Freund et al. (2002). The chemical shift for gaseous H2 is 4.45 ppm (Raynes 1977) and H2 intercalated into different materials shows a significant range of shifts, so molecular H2 would not necessarily be easy to distinguish from H2O molecules or OH groups on the basis of chemical shift. There is a well known correlation between 1H MAS NMR shift and structure and information on H-H distances and clustering can be obtained by exploiting the dipole-dipole interactions between H nuclei, so abundant structural information is accessible.

Problems and difficulties in applying 1H MAS NMR to NAMs Despite all the potentially attractive features of 1H NMR for studying NAMs, there have been very few published studies because of the practical difficulties of working with the small signal from the low concentration of H. The specific problems are

i)

The NMR rotor, the caps on the rotor and the stator (the assembly in which the rotor is contained) can be made of H-containing materials or contain adsorbed water. This problem can be minimized by drying all components before use, by using caps made from Kel-f (a fluorine based polymer) and an NMR probe with a stator containing minimal H containing materials, and by avoiding the use of porous materials in the stator (as they are difficult to dry). Even if all these precautions are taken, it is impossible to completely eliminate background signals. A spectrum of an empty NMR rotor should therefore be taken, and this spectrum should be subtracted from all sample and standard spectra prior to further processing and integration. An example from the work of Johnson and Rossman (2003) is shown in Figure 4. The bottom spectrum (A) is that of an empty rotor, while spectrum (B) is for anhydrous labradorite powder, and (C) is an uncorrected spectrum of an OH bearing sample of andesine. The top spectrum (D) is the final background subtracted data, obtained by subtracting (B) from (C), and even with careful background subtraction there are still artifacts in the spectrum around 1.5 ppm. The difficulty of performing an effective background subtraction is compounded for experimental equipment with a large proton background and samples with small signals. In the case of samples containing paramagnetic ions a normal background subtraction is not applicable, because the paramagnetic susceptibility of the sample influences the local B0 and broadens the background (M. Fechtelkord, personal communication).

Structural Studies of OH in NAM’s Using NMR

59

Figure 4. Illustration of the procedure for subtracting a background signal from the 1H MAS NMR spectrum of a nominally anhydrous mineral (after Johnson and Rossman 2003).

ii) The signal from H dissolved in NAMs can easily be confused with signals from small concentrations of contaminant H-bearing phases such as melt inclusions, fluid inclusions and hydrous minerals as well as adsorbed water on the surfaces of grains and trapped water on grain boundaries. In our experience it is this problem which is the most limiting for NMR studies of NAMs. It is hard to obtain sufficiently large samples, either natural or experimentally produced, which are free of hydrous impurities. It is well known that narrow peaks from surface contamination by organic species have a shift of around 1-2 ppm, and thus overlap with the expected shift from non hydrogen bonded OH. Great care has therefore to be taken in interpreting any features in this region of the spectrum.

iii) The spin lattice relaxation time, T1, for 1H can be long. This can be a particular problem for materials where protons are distant from each other and for chemical pure systems where there are no paramagnetic cations to help relaxation.

iv) NMR studies are generally restricted to samples that are free of significant concentrations of paramagnetic ions. Even if all the other problems can be overcome, 1 H NMR is probably only suitable for studies of iron-free synthetic analogues of mantle minerals rather than natural mantle mineral samples. 1

H MAS NMR studies of orthopyroxene

A 1H MAS NMR spectrum of synthetic enstatite prepared at 1.5 GPa and 1050 °C was presented by Kohn (1996). The spectrum consisted of a broad feature, with narrower peaks at 7.9 and 5.9 ppm, together with peaks which were interpreted as either fluid inclusions or organic surface contaminants (Fig. 5). The shift and relative intensities of the two narrow peaks are entirely consistent with FTIR spectra of pure enstatite which has peaks at 3062 and 3360 cm−1 (e.g., Rauch and Keppler 2002; Stalder and Skogby 2002; Grant et al. 2006). The absolute concentrations of water in the sample was calculated to be 870 ppm H2O. However, it was noted that of the broad components could be related to hydrous mineral or glass impurities or water molecules at grain boundaries and that if only the narrow components are considered the solubility in enstatite would be 240 ppm H2O. Keppler and Rauch (2000)

60

Kohn

suggested that the solubilities in the study of Kohn (1996) could be overestimated because of hydrous inclusions, growth defects and surface water. Subsequent studies in our laboratory (Najorka and Kohn, in prep) suggest that the lower figure of 240 ppm is close to the true solubility and that the broad features in the spectra result mainly from a hydrous phase formed upon quenching the coexisting aqueous fluid at the end of the synthesis run. 1

H MAS NMR studies of clinopyroxene

Three clinopyroxene samples in the system diopside-CaTs were studied by Kohn (1996). These samples had much higher dissolved water concentrations than the enstatite sample, with the total spectral areas corresponding to 2615-4100 ppm H2O. If only the narrow parts of the spectra are considered to be dissolved water the solubilities are in the range 1160-2430 ppm for synthesis conditions of 1.5 GPa and 1000-1150 °C. These values are comparable with water concentrations in other aluminous Figure 5. The 1H MAS NMR centrebands of clinopyroxenes determined using FTIR experimentally synthesised forsterite (Fo-10), (Skogby et al. 1990; Smyth et al. 1991; enstatite (En-2) and three clinopyroxenes on the Di-CaTs join (Dicat1-9, Dicat2-9 and Dicat3-1) Bromiley and Keppler 2004). The spectra (Kohn 1996). The prominent peak at 4.7 ppm is have different shapes, with the main peak due to water in fluid inclusions and those at 1.3 becoming broader and more asymmetric and 0.8 ppm are probably due to contamination by with increasing CaTs component (Fig. 5). organic compounds. The spectra also have prominent spinning sidebands (Fig. 6), suggesting that some of the H has strong H-H dipolar coupling because of clustering of hydrogen in species such as hydrogarnet substitution or included water molecules. 1

H MAS NMR studies of olivine

Kohn (1996) presented a 1H MAS NMR spectrum of a synthetic forsterite sample. This spectrum contained a broad resonance at 4.3 ppm, together with a small feature at 6.9 ppm and an intense peak at 1 ppm (Fig. 5). Although peaks from surface contamination are known to be near 1 ppm the peak in the olivine sample was anomalously large, so it was interpreted as being a possible peak for dissolved OH in the forsterite. The total area of the spectrum corresponded to 1790 ppm and even the narrow part alone give a value of 560 ppm. These values were much higher than expected, and should be treated with caution. A much improved spectrum of forsterite synthesised at 2.1 GPa and 1100 °C, and in equilibrium with a small amount of enstatite to buffer silica activity, was subsequently presented in a published abstract (Kohn 1999). This spectrum contained a broad resonance at 1 ppm (in addition to the background signal), and a distinct peak at 6.7 ppm, corresponding to relatively strongly hydrogen bonded OH. Although there is still some ambiguity in the intensity of features around 1 ppm, because of adsorbed organic contaminants, this spectrum does not have the very broad feature that dominates the earlier spectrum. Quantification suggests a solubility of 400-500 ppm H2O. This is still higher than the solubility determined on different samples using FTIR, although it is closer to solubili-

Structural Studies of OH in NAM’s Using NMR

61

ties calculated with the FTIR calibration of Bell et al. (2003) than those calculated using alternative calibrations. More work needs to be done to resolve this important issue, although one possibility is that elevated nonequilibrium concentrations of OH can be incorporated in forsterite under certain crystal growth conditions (Lemaire et al. 2004). The peak at 6.7 ppm is an interesting feature of the spectrum. The correlation shown in Figure 1 would suggest that this corresponds to an O-H…O distance of 0.284 nm (Eckert et al. 1988), which in turn would be predicted to give an O-H stretching frequency around 3450 cm−1 (Libowitzky 1999). However, there is a large spread in the data on which the Libowitzky (1999) correlation is based (at this distance), and the corresponding stretching frequency could be as low as 3200 cm−1. The peak at 6.7 ppm is therefore consistent with the low frequency OH stretching peaks which have been observed in the FTIR spectra of forsterite when synthesized under conditions of high silica activity (Lemaire et al. 2004; Grant et al. 2006). 1

H MAS NMR studies of garnet

Figure 6. The same 1H MAS NMR spectra as shown in Figure 5, but displayed over a wider frequency range to show the spinning sidebands.

A sample of grossular garnet that contained 3.6 wt% H2O was included in the early 1H MAS NMR study of hydrous minerals by Yesinowski et al. (1988). The MAS spectrum was broader than would be expected based on the average H density, suggesting either that proton-proton dipolar coupling is larger than would be expected based on a homogeneous distribution of H in the sample or that an alternative line-broadening mechanism was operating. It was suggested that the width was too large to be explained by chemical shift dispersion, and that the width was not related to paramagnetic impurities because the width of the spinning sideband envelope did not increase significantly with increasing field. It was therefore concluded that the width was due to proton-proton dipolar coupling, and hence that protons are clustered within the structure. This conclusion was not unexpected as there is known to be a solid solution series between grossular and hydrogrossular and the latter contains the clustered (OH)4 defect. 1

H MAS NMR studies of SiO2 polymorphs

Yesinowski et al. (1988) reported a 1H MAS NMR spectrum for a natural quartz sample. The only peaks that were observed were attributed to the organic contaminant at 1.5 ppm and fluid inclusions at 4.7 ppm. No features for dissolved hydroxyl were observed. No NMR study of hydrogen in coesite or stishovite has yet been published. 1

H MAS NMR studies of feldspars and other aluminosilicate framework minerals

The earliest application of 1H MAS NMR to NAMs was the study of feldspars by Yesinowski et al. (1988). Three different resonances were observed, termed A, B and C. Resonance A was a narrow peak at 1.5 ppm, and was attributed to contamination from organic species because its intensity was reduced by refluxing the samples in CCl4, then packing the sample into

62

Kohn

hexane-washed rotors using gloves. Peak B was assigned to fluid inclusions because it was very narrow even at slow spinning speeds, had no spinning sidebands, and had a shift close to that for liquid H2O. In contrast peak C had an extensive, field-independent, spinning sideband pattern that was similar to that for water molecules in analcite. This resonance was therefore attributed to isolated water molecules which could be experiencing 180° flips or librational motion, but for which any motion was strongly anisotropic. The spin-lattice relaxation time, T1, was also measured for A, B and C, and found to be much longer for C than A or B, consistent with the assignments of A and B to surface or included components which cannot effectively interact with the dissolved C component. The Yesinowski et al. (1988) study also included the ammonium feldspar, buddingtonite. 1H MAS NMR of this sample gave a narrow peak at 6.8 ppm with moderate sideband intensities. This shift is consistent with ammonium in other compounds and the narrow spectrum indicates that the strong dipolar coupling between the protons of the NH4 group must be partially averaged by molecular motion. More recently Johnson and Rossman (2003) have used both FTIR and 1H MAS NMR to characterize the hydrous components in feldspars and to calibrate the infrared absorption by both OH and H2O. The NMR data showed that microcline samples contained structural water molecules at a level of 1000-1400 ppm H2O, whereas a sanidine sample contained 170 ppm H2O as hydroxyl. Plagioclase samples were also studied and dissolved OH concentrations were reported, but difficulties were encountered in unambiguously assigning intensity to dissolved or adsorbed OH. As in previous studies, the spectra were also complicated by the presence of organic molecules at the surface and the presence of fluid inclusions. Xia et al. (2000) also used 1H MAS NMR to determine the water concentrations of several anorthoclase megacrysts from Cenozoic alkalic basalts from China. The total water contents of three of the samples are 405 ppm, 915 ppm and 365 ppm, but more work would be needed to determine the distribution of water of different types in these samples. 1

H MAS NMR studies of wadsleyite

Wadsleyite was produced unintentionally in an NMR study of high-pressure hydrous silicates (Phillips et al. 1997). In one sample of superhydrous phase B, a significant proportion of the sample was actually either phase B or wadsleyite. This enabled the spectra of both superhydrous phase B and wadsleyite to be obtained. The 1H MAS spectrum of wadsleyite consisted of a narrow peak (indicating isolated hydroxyl groups) at 1.5 ppm. The intensity of the signal, when coupled to the proportions of the phases was used to calculate a water concentration in the wadsleyite of 0.2 wt%. Kohn et al. (2002) measured the 1H MAS NMR spectrum of a wadsleyite sample with a much higher water concentration of 1.5%. This sample had a complex and asymmetric spectrum that was fitted with six peaks in the range 1.4-11.0 ppm (Fig. 7). The 1H MAS NMR data were consistent with FTIR spectra measured on the same samples, and quantification of both data sets suggested that the Libowitzky and Rossman (1997) calibration of IR intensities works well. The resolution between the sites was inferior to that for the FTIR spectra for the same sample. One could speculate that the linewidth results from residual dipolar coupling between protons, but subsequent studies at higher magnetic fields and with faster magic spinning (unpublished data) did not show a significant improvement in resolution, so the controls on linewidths in this system are not yet clear.

NON-SPINNING 1H NMR EXPERIMENTS In an earlier section it was explained that a variety of interactions cause broadening of NMR lines in the solid state. Many of these interactions are averaged by magic angle spinning, leaving the spectrum as a measure of the distribution of chemical shifts in the sample. MAS enables much useful information to be obtained, and by reducing linewidths, also increases sensitivity. This reduction in sensitivity is a major drawback in applying static NMR to NAMs,

Structural Studies of OH in NAM’s Using NMR

63

Figure 7. The 1H MAS NMR centerbands of experimentally synthesised hydrous wadsleyite. (a) is the experimental spectrum, (b) is a fit the spectrum with the Lorentzian peaks shown in (c), and (d) is the residual. This NMR spectrum is less well resolved than the FTIR spectrum of the same sample, so this fit is not intended to be unique, but does provide a basis for comparison (see Kohn et al. 2002 for details).

a b c d 25

20

15

10

5

0

-5

-10

-15

PPM

and all the problems of accidental contamination by H-bearing compounds or materials are, if anything, more severe for static measurements, so application of static NMR to NAMs is of limited applicability. However, some of the information which is lost under MAS is useful in its own right, and a variety of strategies can, in principle, be employed to exploit this information. The simplest option is to perform static (non-spinning) experiments on powdered samples. In this case characteristic lineshapes are obtained which result from summing the contributions of all directional interactions over all orientations. The classic case is that for pairs of protons, such as immobile water molecules in a structure. The strong dipolar interaction between the two protons leads to a characteristic broad doublet feature, known as the Pake doublet. In this case, the strength of dipole-dipole coupling can be measured, and the H-H distance calculated (e.g., Phillips et al. 1997). If large single crystals are available, additional information can be obtained by orienting the crystal, and collecting spectra with the crystal fixed at varying angles to the magnetic field. The three dimensional nature of the different interactions can be explored in this way. A more sophisticated method of obtaining data on H environments using static samples was applied to garnet samples by Cho and Rossman (1993). Grossular samples with a water concentration of 0.2-0.3 wt% H2O were studied using non-spinning NMR experiments, and the data were compared with data for hydrogrossular. A multiple quantum technique was used to deduce that one sample contained mostly pairs of H, whereas the other sample contained clusters of two and four protons. As expected the hydrogrossular contained clusters of four protons. In addition Cho and Rossman (1993) analyzed the static lineshapes in detail and obtained an average interproton distance of 0.169 nm in grossular, compared with 0.216 nm in hydrogrossular.

STUDIES OF OTHER NUCLEI IN NAMS 1 H is the most sensitive nucleus for NMR, so it is the obvious candidate for NMR studies of the dissolution of water in NAMs. 2H could potentially be used for deuterated samples, but 2H is very much less sensitive than 1H and it is a quadrupolar nucleus (i.e., I > ½) so is subject to additional line-broadening mechanisms. Hydrous glasses with water concentrations of 1.6-4.8

64

Kohn

wt% have been studied with 2H (Eckert et al. 1987) and speciation and dynamic information were obtained. It is unlikely that 2H NMR of NAMs will be successful in the near future. In most cases NAMs will not contain enough H to have a significant effect on the average environment of any of the other atoms in the mineral. There are a few exceptions however, and in these cases NMR studies of other nuclei may be useful. Two important examples are the high pressure polymorphs of Mg2SiO4, wadsleyite and ringwoodite. These phases, although nominally anhydrous, can contain up to several wt% H2O, and there will therefore be significant differences in the environments of 29Si and 17O environments between hydrous and anhydrous samples. The relevant 17O NMR data for hydrous wadsleyite and ringwoodite have not been published, but comparison with 17O NMR for dry samples (Ashbrook et al. 2005) could provide an important constraint on the dissolution mechanism. In some cases the molar concentration of H becomes significant compared to that of other important minor components in NAMs. A recent study of hydrous and anhydrous aluminous orthopyroxene (Kohn et al. 2005) used 27Al NMR to test the proposition (Rauch and Keppler 2002) that dissolution of H is coupled to tetrahedral Al3+, with a consequent increase in the tetrahedral Al: octahedral Al ratio. Rauch and Keppler (2002) suggested that this ratio could potentially be used as the basis of a geohygrometer, even for mantle xenoliths which had dehydrated on ascent. In contrast, the 27Al NMR data of Kohn et al. (2005) suggest that the tetrahedral Al: octahedral Al ratio is the same in both dry and hydrous samples. It should be noted, however, that 27Al is a quadrupolar nucleus with I = 5/2, so interpretation of 27Al MAS NMR spectra is not straightforward and requires an understanding of the nature of the quadrupole interaction (MacKenzie and Smith 2002). Figure 8 shows 27Al spectra for dry and hydrous orthopyroxenes, together with a subtraction that emphasizes the difference between them. Although the [Al]4:[Al]6 is constant the wet sample contains new Al environments which are not present in the dry sample. The concentrations of these new sites imply that each dissolved H modifies the environment of one tetrahedral and one octahedral Al, consistent with protonation of O21 and O22 sites (Kohn et al. 2005).

Figure 8. 27Al MAS NMR spectra of dry and hydrous aluminous enstatite, obtained at a magnetic field of 18.8 T. The ratio of tetrahedral Al : octahedral Al is (within error) the same for the two samples, and after correction for the magnitude of the quadrupole coupling constants, is very close to 50:50 (Kohn et al. 2005). The difference spectrum shows that new Al[4] and Al[6] sites are present in the wet sample, and this information can be used to make deductions about the mechanism of water incorporation. [Used with permission of Elsevier from Kohn et al. (2005) Earth Planet. Sci. Lett., Vol. 238, Fig. 1, p. 342-350.]

Structural Studies of OH in NAM’s Using NMR

65

PROSPECTS FOR FUTURE DEVELOPMENT OF NMR FOR STUDIES OF NAMS In the context of the whole body of research on nominally anhydrous minerals, NMR data have played a relatively minor part. This is certainly because the measurements of 1H are not easy, and special precautions have to be made to overcome the problems of background H signal. More work on obtaining H-free probe materials, drying and purging etc., could potentially reduce backgrounds, but it is debatable whether the improvement would justify the required effort. Various pulse sequences have been used for background suppression in NMR e.g., (Cory and Ritchey 1988; Chen et al. 2004); these have not yet been applied widely to NAMs. The availability of higher magnetic fields could potentially increase the intrinsically small signal for 1H NMR of NAMs. However, the improvement would not be dramatic, and faster spinning would also be required in order to avoid problems resulting from increased chemical shift anisotropy. For some samples, increasing the field could increase the resolution of 1H MAS spectra, depending on the line broadening mechanisms. Studies of quadrupolar nuclei, such as 27Al, 17O and 23Na are more likely to be successful at high fields, and some progress in studying dissolution mechanisms via the effect of dissolved H on other nuclei should be expected. The use of cross-polarization from 1H to other nuclei (e.g., Farnan et al. 1987) could have a major effect on the sensitivity of NMR of other nuclei to hydrous species. If hydrogen is clustered in NAMs, the H-H distances could potentially be accessed using double quantum MAS techniques (e.g., Schnell and Spiess 2001). In summary, despite several partially successful attempts to introduce NMR as a major tool for studying the dissolution of water in NAMs, the full potential of the technique has not yet been realized. The problem of paramagnetic cations in most naturally occurring NAMs will remain a major limitation of NMR, so its major contribution is likely to be in simplified synthetic systems. As with most spectroscopic techniques, the most crucial factor in obtaining high quality 1H NMR spectra is probably sample selection. Samples should be entirely free of fluid or melt inclusions or hydrous phases, and should be only coarsely crushed to minimize the surface area. The full potential of NMR in this field may ultimately be limited more by experimental methodologies for producing such clean samples than by the methods of obtaining NMR spectra.

ACKNOWLEDGMENTS I would like to thank NERC for funding, Michael Fechtelkord, Mark Smith and an anonymous reviewer for helpful comments on the manuscript and Hans Keppler for his efforts in editing this volume.

REFERENCES Ashbrook SE, Berry AJ, Hibberson WO, Steuernagel S, Wimperis S (2005) High-resolution 17O MAS NMR spectroscopy of forsterite (α-Mg2SiO4) wadsleyite (β-Mg2SiO4), and ringwoodite (γ-Mg2SiO4). Am Mineral 90:1861-1870 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: A quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res-Solid Earth 108(B2) Art. No. 2105 Bromiley GD, Keppler H (2004) An experimental investigation of hydroxyl solubility in jadeite and Na-rich clinopyroxenes. Contrib Mineral Petrol 147:189-200 Brunner E, Sternberg U (1998) Solid-state NMR investigations on the nature of hydrogen bonds. Prog Nucl Magn Reson Spectrosc 32:21-57 Chen Q, Hou SS, Schmidt-Rohr K (2004) A simple scheme for probehead background suppression in one-pulse 1 H NMR. Solid State Nucl Magn Reson 26:11-15

66

Kohn

Cho H, Rossman GR (1993) Single crystal NMR studies of low-concentration hydrous species in minerals grossular garnet. Am Mineral 78:1149-1164 Cory DG, Ritchey WM (1988) Suppression of signals from the probe in Bloch decay spectra. J Magn Reson 80:128-132 Eckert H, Yesinowski JP, Silver LA, Stolper EM (1988) Water in silicate glasses - quantitation and structural studies by 1H solid echo and MAS NMR methods. J Phys Chem 92:2055-2064 Eckert H, Yesinowski JP, Stolper EM, Stanton TR, Holloway J (1987) The state of water in rhyolitic glasses a deuterium NMR study. J Non-Cryst Solids 93:93-114 Engelhardt G, Michel D (1987) High-Resolution Solid-State NMR of Silicates and Zeolites. Wiley Farnan I, Kohn SC, Dupree R (1987) A study of the structural role of water in hydrous silica glass using crosspolarization Magic Angle Spinning NMR. Geochim Cosmochim Acta 51:2869-2873 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 Freund F, Dickinson JT, Cash M (2002) Hydrogen in rocks: an energy source for deep microbial communities. Astrobiology 2:83-92 Grant KJ, Kohn SC, Brooker RA (2006) Solubility and partitioning of water in synthetic forsterite and enstatite in the system MgO-SiO2-H2O±Al2O3. Contrib Mineral Petrol 151:651-664 Johnson EA, Rossman GR (2003) The concentration and speciation of hydrogen in feldspars using FTIR and 1H MAS NMR spectroscopy. Am Mineral 88:901-911 Keppler H, Rauch M (2000) Water solubility in nominally anhydrous minerals measured by FTIR and 1H MAS NMR: the effect of sample preparation. Phys Chem Mineral 27:371-376 Kirkpatrick RJ (1988) MAS NMR spectroscopy of minerals and glasses. Rev Mineral 18:341-403 Kohn SC (1996) Solubility of H2O in nominally anhydrous mantle minerals using 1H MAS NMR. Am Mineral 81:1523-1526 Kohn SC (1999) Partitioning of water between nominally anhydrous minerals in the upper mantle. In: Processes and Consequences of Deep Subduction. Vol 99/7. Mysen B, Rubie D, Ulmer P, Walter M (eds) Terra Nostra, Alfred Wegener Stifung, p 58-59 Kohn SC (2004) NMR studies of silicate glasses. EMU Notes Mineral 6:399-419 Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ (2002) Ordering of hydroxyl defects in hydrous wadsleyite (β-Mg2SiO4). Am Mineral 87:293-301 Kohn SC, Roome BM, Smith ME, Howes AP (2005) Testing a potential mantle geohygrometer; the effect of dissolved water on the intracrystalline partitioning of Al in orthopyroxene. Earth Planet Sci Lett 238: 342-350 Lemaire C, Kohn SC, Brooker RA (2004) The effect of silica activity on the incorporation mechanisms of water in synthetic forsterite: a polarised infrared spectroscopic study. Contrib Mineral Petrol 147:48-57 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H..O hydrogen bond lengths in minerals. Monatshefte für Chemie 130:1047-1059 Libowitzky E, and Rossman GR (1997) An IR absorption calibration for water in minerals. Am Mineral 82: 1111-1115 MacKenzie KJD, Smith ME (2002) Multinuclear Solid State Nuclear Magnetic Resonance of Inorganic Materials. Pergamon Phillips BL, Burnley PC, Worminghaus K, Navrotsky A (1997) 29Si and 1H NMR spectroscopy of high-pressure hydrous magnesium silicates. Phys Chem Mineral 24:179-190 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Raynes WT (1977) Theoretical and physical aspects of nuclear shielding. NMR Spectrosc Period Rep Chem Soc 7:1-25 Schnell I, Spiess HW (2001) High-resolution 1H NMR spectroscopy in the solid state: very fast sample rotation and multiple quantum coherences. J Magn Reson 151:153-227 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene - variations in the natural environment. Am Mineral 75:764-774 Smyth JR, Bell DR, Rossman GR (1991) Incorporation of hydroxyl in upper-mantle clinopyroxenes. Nature 351:732-735 Stalder R, Skogby H (2002) Hydrogen incorporation in enstatite. Eur J Mineral 14:1139-1144 Stebbins JF (1988) NMR spectroscopy and dynamic processes in mineralogy and geochemistry. Rev Mineral 18:405-429 Xia QK, Pan YJ, Chen DG, Kohn S, Zhi XC, Guo LH, Cheng H, Wu YB (2000) Structural water in anorthoclase megacrysts from alkalic basalts: FTIR and NMR study. Acta Petrologica Sinica 16:485-491 Yesinowski JP, Eckert H, Rossman GR (1988) Characterization of hydrous species in minerals by high-speed 1 H MAS NMR. J Am Chem Soc110:1367-1375

4

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 67-83, 2006 Copyright © Mineralogical Society of America

Atomistic Models of OH Defects in Nominally Anhydrous Minerals Kate Wright Nanochemistry Research Institute Curtin University of Technology GPO Box 1987 Perth, Western Australia 6845, Australia e-mail: [email protected]

INTRODUCTION The Earth’s upper mantle may contain substantial amounts of water dissolved in nominally anhydrous minerals (NAMs) such as the Mg2SiO4 polymorphs, pyroxenes and garnets. This water, incorporated into the crystal lattice as hydrogen defects, can have a profound influence on the physical properties of the mantle, even when present at low concentrations. An understanding of these defects at the atomic level is therefore of fundamental importance for the development of models of the evolution and dynamics of the Earth’s mantle. The incorporation of hydrogen and its influence on the properties of NAMs has been an active area of research for almost three decades. High pressure synthesis of hydrous phases, analyzed using a range of spectroscopic techniques (see Kohn 2006; Libowitzky and Beran 2006; Rossman 2006), have yielded a wealth of information that allow us to determine the concentration of hydrogen that can be accommodated by various NAMs, and provide information on the mechanisms of uptake. However, these data are often complex, and difficult to interpret unambiguously. Computer simulation methods can offer real insights at the atomic level, often not accessible by experiment, and provide an alternative way to explore hydrogen defects in minerals. The past 20 years have seen an explosion in the use of computational modeling to study a range of phenomena in minerals. These include the high-pressure behavior of mantle (Oganov and Price 2005) and core (Vocadlo et al. 2003) phases, diffusion (Walker et al. 2003), dislocation structures (Walker et al. 2005), and mineral surface reactivity (Kerisit et al. 2005). A broad introduction to the methods and applications to the geosciences is given in the recent MSA volume edited by Cygan and Kubicki (2001). In this chapter we explore the contribution of computational methods to the development of atomistic models of hydrogen defects in NAMs of the Earth’s upper mantle and transition zone. We begin with an introduction to defects in solids, followed by a brief overview of the computational methods used to model them, and then review the results of studies on the most important upper mantle NAMs.

POINT DEFECTS IN MINERALS All crystalline materials contain imperfections, or defects, that disrupt their long-range ordering. These can be point defects, line defects (dislocations) or planar defects (stacking faults, crystallographic shear planes), as well as interfaces such as grain boundaries and twin planes. A whole range of different defect types can be present in the same crystal, depending on the conditions of formation and subsequent history, and can interact in a variety of ways. 1529-6466/06/0062-0004$05.00

DOI: 10.2138/rmg.2006.62.4

68

Wright

Point defects are the simplest type of imperfection that can occur in a crystal, and involve the removal, inclusion or replacement of an atom or ion at specific sites in the crystal. They are important, since they are the means by which atomic migration takes place and can influence color, electrical conductivity and reactivity. In general, we define three types of point defect: vacancies, where an ion is removed from its normal lattice site; interstitials, where an ion is present at a non-lattice site; and impurities, where dopants are present either at lattice or interstitial sites (Fig. 1). In ionic and semi-ionic crystals, these point Figure 1. Schematic representation of point defects defects will typically be charged species in a crystal of composition MX. and so must occur in balanced defect populations to maintain charge neutrality. Within pure crystals, defects made up of balanced populations of cations and anions are termed Schottky defects, while those composed of a vacancy and interstitial of the same species are known as Frenkel defects. In the strictest sense, Schottky disorder requires that charge neutrality and stoichiometry be maintained, however, the term is fairly loosely applied in the literature and we will use Schottky to refer to any charge neutral group of vacancies. The formation energy of a Schottky defect (ESch) is the sum of the individual vacancy energies (Ex) plus the lattice energy (U) of the phase removed: ESch = EV1 + EV 2 + .......EVn + U

(1)

For a Frenkel defect, the energy is simply the sum of the corresponding vacancy and interstitial energies. Point defects occur in all crystals at temperatures above 0 K and, in pure crystals, there will be a finite population of these intrinsic defects in thermodynamic equilibrium with the system. The change in free energy (∆G) associated with the introduction of a defect is expressed in the usual way as: ∆G = ∆H − T ∆S

(2)

∆H is the enthalpy, associated with changes in nearest neighbor interactions, and ∆S the entropy increase due to the introduction of disorder into an otherwise perfect crystal. The entropy term includes vibrational disorder in the atoms around the defect as well as configurational terms related to way in which the defects are distributed within the crystal. The equilibrium concentration of defects in a stoichiometric material of composition MX (e.g., MgO, NaCl) can be approximated by the following: ndef = Ne−∆H / 2 RT

(3)

where N is the number of sites, ∆H is the enthalpy required to form the defect, R the universal gas constant, and T temperature. The full derivation of this formula can be found in Putnis (1992). Populations of intrinsic defects are generally small; for a typical alkali halide at room temperature less than 1 site per million will be vacant (Tilley 1987). Normally, we would expect one type of point defect to dominate, and this will be the one with the lowest value of ∆H. Generally speaking, Schottky defects tend to dominate in close packed solids, while Frenkel defects are more common in framework and layered structure materials.

Atomistic Models of OH in NAM’s

69

In many crystalline materials, non-intrinsic vacancies and interstitials can be created in response to the presence of impurities and are thus termed extrinsic. These impurities may occur as neutral species (e.g., Mn2+ replacing Mg2+), or as charged species (e.g., Al3+ replacing Si4+) that must be charge balanced by another impurity (coupled substitution) or by an accompanying vacancy or interstitial. Since our defects are all charged species, we might expect them to interact strongly and form discrete defect clusters with significant binding energies. When defects form, the ions around them must relax to accommodate the new configuration, and in some cases, the energy will be lower for a cluster than for the same isolated defects. At this stage it’s a good idea to introduce the notation used for point defects, so that they can be easily identified and written down. The Kröger-Vink (Kröger 1972) notation is widely used to describe point defects. Vacancies (V) are defined in terms of species (subscript) and charge (superscript), where the charge may be neutral (x), positive (•) or negative ( / ). For // example, VMe describes a metal (Me) vacancy with an effective 2− charge. A Me2+ interstitial is x written Mei•• , while a neutral impurity (A) at a Me site is given as AMe . In the case of hydrogen, − 2− we are generally referring to an (OH) group replacing an O , which is written (OH)•O.

THEORETICAL BACKGROUND This section presents a brief introduction to the different computational approaches used for the study of defects. Technical details of the different theories are not included since there are many excellent texts available that cover computational methods (e.g., Foresman and Frisch 1997; Leach 2001; Gale and Rohl 2003; Griffiths 2005) and readers are referred to these for a more detailed and rigorous treatment. Computer simulation methods aim to determine the energy of a solid as a function of the interaction of all particles within that system, with varying degrees of approximation depending on the level of theory used. Simulations can be static, where the system is essentially at 0 K, or dynamic, where the free energy is calculated by molecular dynamics (MD) or lattice dynamics techniques. This can be obtained either by quantum mechanics or by atomistic techniques, based on classical mechanics, in which the details of the electronic structure are subsumed into a series of effective interactions that depend only on the nuclear positions.

Quantum mechanical methods Within quantum mechanical (QM) theory, both the electrons and nuclei are explicitly considered and their interactions are generally calculated using either Hartree-Fock (HF) or density functional theory (DFT). In both cases the Born-Oppenheimer approximation, that the motion of electrons can be separated from that of the nuclei, is assumed to hold true. However, the difficulty arises when trying to calculate interactions between electrons, since the potential experienced by one electron depends on the position of all other electrons in the system. These exchange interactions, between electrons of like spin, and correlation interactions, between electrons of opposite spin, are treated in different ways depending on the approach used. HF theory calculates the exchange energy explicitly but ignores correlation, although post-HF techniques, such as Moller-Plesset (Leininger et al. 2000) and Coupled Cluster theory, can overcome this to some extent at the price of significantly increased computational cost. DFT, while being in principle an exact theory, in practice has to approximate both the exchange and correlation potentials, using either the Local Density Approximation (LDA) or Generalized Gradient Approximation (GGA). Hybrid functionals, such as B3LYP (Becke 1993), that combine GGA or LDA with exact HF exchange, are also available. The wave function is generally described by a linear combination of basis functions that can be atom centered Gaussian type functions, or plane waves. In many cases, only the valence electrons need

70

Wright

to be explicitly considered, as it is these that are responsible for bonding. The electrostatic potential due to the frozen core electrons and the nucleus are commonly represented by a pseudopotential, and can lead to substantial computational savings, particularly when used in conjunction with plane waves. Quantum mechanical calculations, are by their very nature, the most accurate and reliable approach, although they require significant computational resources. Early studies using these techniques were limited to the use of clusters of atoms representing a solid, or very small unit cells. Recent developments in both hardware and software mean that it is now possible to calculate the properties of complex mineral phases using these methods and thus their use is increasing. DFT is by far most commonly used technique within the Earth Sciences, with particular success being enjoyed using the planewave, pseudopotential codes such as CASTEP (Segall et al. 2002) and VASP (Kresse and Furthmuller 1996a,b). DFT does, however, have its limitations; band gaps are typically underestimated by about 50%, and while LDA overestimates binding, GGA underestimates it leading to calculated cell parameters that are normally 1-2% too large. In addition, long-range van de Waals interactions are not well described, so that layered structures can prove difficult to model accurately.

Classical methods Classical atomistic, or molecular mechanics (MM), simulation techniques employ interatomic potential functions to describe the total energy of the system in terms of atomic positions. These potentials include long-range electrostatic effects as well as short-range interactions produced by the overlap of nearest neighbor electron clouds. Terms to describe oxygen ion polarizibility and directionality of bonding are also available. The effective potential parameters are derived either by fitting to experimental data (structure, elastic constants, dielectric constants, etc.), or by fitting to potential energy surfaces generated by high level QM calculations. The equilibrium positions of the ions are then evaluated by minimizing the lattice energy until all forces acting on the crystal are close or equal to zero. The majority of defect calculations carried out using these methods are performed at 0 K and 0 GPa and so the energies obtained are enthalpies rather than free energies of defect formation, although free energies can be obtained by the use of lattice dynamical techniques. A comprehensive overview of interatomic potential methods can be found in Gale and Rohl (2003). Interactions between closed shell ionic species are well modeled by standard two-body potentials of the Born-Meyer type but bonded molecules, such as (OH), need to be treated differently. The hydroxyl molecule is generally described using a Morse potential of the form: 2 ⎡ ⎤ − α(r −r ) UijMorse = D ⎢⎛⎜ 1 − e ij 0 ⎞⎟ − 1⎥ ⎠ ⎢⎣⎝ ⎥⎦

( 4)

where D corresponds to the dissociation energy of the bond, r0 is the equilibrium bond length and α, in combination with D, is related to the vibrational frequency of the stretching mode. Traditionally, studies of hydrogen defects in minerals use the Morse potential parameters originally derived from HF calculations on NaOH (Saul et al. 1985). These have been used extensively in the study of hydroxyl in zeolites (e.g., Schröder et al. 1992), and in a whole range of other minerals including muscovite (Collins and Catlow 1990) and goethite (Steele 2002). In the Saul et al. model, polarizibility of the hydroxyl oxygen ion is not included, and the ions have fractional charges (qO = −1.412, qH = +0.412) such that the (OH) unit has −1 overall charge. Despite its success in calculating defect structures and energies, the Saul et al. potential does not give O-H vibrational frequencies that agree with experiment. This is a reflection on the fact that HF calculations were used in the original fitting procedure. Gatzemeier and Wright (2006) modified the α parameter by fitting it to O-H stretching frequencies in forsterite obtained using QM/MM methods (Braithwaite et al. 2003). Other,

Atomistic Models of OH in NAM’s

71

more sophisticated, Morse potentials have been developed for OH, such as that by Baram and Parker (1996). The model explicitly includes oxygen polarizibility via the use of a shell model (Dick and Overhauser 1956), where the ion is divided into a core containing all of the mass, and a shell, coupled by an harmonic spring. Other potential forms used to model water and hydroxyl groups include simple LennardJones type models as well as more sophisticated potentials, such as that of Stillinger-David (Stillinger and David 1980) that allow the dissociation of the water molecule. However, these models have not been used in the study of hydrogen in NAMs and so will not be discussed further.

Treatment of defects The choice of theoretical method to use for defect calculations depends on the level of accuracy required and the actual quantity to be calculated. This, along with the limitations imposed by available computational resources, determine which level of theory to use. Classical MM methods have been extremely successful at predicting defect behavior in a whole range of solids from complex ionic materials, such as zeolites (Schröder et al. 1992), clays (Cygan et al. 2004) and carbonates (Austen et al. 2005), to battery materials (Islam et al. 2005) and semi-conductors (Wright and Gale 2004). They need minimal computer time and memory so that large numbers of possible configurations can be easily sampled. However, the quality of the results will only be as good as the potential parameters available for the system under consideration. Interatomic potential methods do have their limitations, as they are generally unable to model bond breaking and bond forming processes, although there are some exceptions to this such as the reactive empirical bond order (REBO) type models (e.g., Brenner et al. 2002). Equally importantly, they cannot be used to assess the influence of defects on those properties that explicitly depend on the electronic structure, such as JahnTeller distortions associated with transition metal ions. QM methods usually give a much more accurate description of a system, but use far more resources. There are essentially two approaches to calculating the structure and energetics of defects in solids, namely the supercell (SC) and cluster methods. The supercell approach, illustrated in Figure 2, has the defect within in a large supercell and the system is modeled in any code, QM (e.g., CASTEP, VASP) or classical (e.g., GULP, PARAPOCS), using periodic boundary conditions. The cell should be large enough that the defect does not interact with those in the periodic images as this will introduce an additional component into the total energy obtained. Defect-defect interactions can be corrected for in terms of electrostatic multipoles, and, in the case of charged defects, a neutralizing background needs to be applied (Leslie and Gillan 1985). Although this first consideration is not an issue for force field calculations, which can handle cells containing thousands of atoms, it can be a problem when performing QM calculations. For periodic DFT codes (PDFT) such as CASTEP and VASP, the CPU time required to run a calculation goes up in Figure 2. Illustration of a supercell used in defect a non-linear fashion as the number of atoms calculations. The cell defined by a dark line is increases. With the advances in parallel the unit cell which is periodically repeated in the supercell. computing, and increases in efficiency, large-

72

Wright

scale simulations of cells containing hundreds of atoms are possible, although most calculations use much less than this. The cluster method involves cutting out a fragment of a crystal that has the defect at its centre, and embedding it in some representation of the bulk material (Fig. 3). The embedding approach is most commonly implemented in MM calculations, where the polarization caused by introducing the defect is handled using the formulation of Mott and Littleton (Mott and Littleton, 1938). In this approach the crystal is divided into two concentric spherical regions (Fig. 3). In region 1, which contains the defect at its centre, an explicit atomistic Figure 3. The embedded cluster has a central area containing simulation is carried out to adjust the defect that is embedded in a representation of the bulk the coordinates of all ions within material. the region until they are at positions at which no net forces act on them, i.e. they are relaxed around the defect. In region 2, the effects of the defect are relatively weak and the relaxation can be calculated essentially as the polarization response to the effective charge of the defect. In practice an interface region between regions 1 and 2, referred to as 2A, is normally used. The resulting defect energy is a measure of the perturbation by the defect of the static lattice energy of the crystal. As with supercells, size matters, and region 1 should be large enough that the defect energy is converged with respect to region size. The Mott-Littleton method is implemented in codes such as GULP (Gale and Rohl 2003). Although the cluster approach works well within classical calculations, its use in quantum mechanical simulations is more problematic, primarily due to edge effects and the limited number of atoms that can be included in the cluster. Hybrid, so called QM/MM embedded cluster methods (see for example Braithwaite et al. 2002) overcome these problems, by having a quantum region surrounded by a classical one. The central region contains the defect of interest and is treated at the quantum mechanical level of theory, using either HF or DFT. This QM region is normally terminated with atoms described by effective core pseudopotentials and is embedded in a large (>50 Å) array of point charges which represent the potential due to the bulk crystal that acts on the embedded cluster. Between the two, is a sphere of classical atoms that are described by interatomic potentials. The embedded cluster approach overcomes many of the problems associated with studying charged defects using periodic supercell methods, including the problem of the energy term produced by defect – defect interactions.

OH DEFECTS IN MANTLE SILICATES Water can be incorporated into NAMs via a number of different mechanisms, depending on the chemistry and defect structure of the host mineral. Equation (5) describes the formation of a hydrogarnet defect, where two molecules of water interact with a silicon ion to form a silicon vacancy charge balanced by four (OH) groups, and a unit of SiO2: 2H 2O + SiSix + 4OOx → [ VSi⋅ 4(OH)O ] + SiO2 x

(5)

Atomistic Models of OH in NAM’s

73

The energy of the above reaction is found by summing together the energies of the different terms, including the self-energy of the water molecule, calculated using the same level of theory. In some MM calculations, this self-energy is substituted by a proton transfer term representing the energy of the H2O + O2− → 2(OH)− reaction (see Wright et al. 1994 for details). Other reaction pathways (Eqns. 6-8) involve the creation of other vacancies, or reactions with impurities and their energetics are calculated in a similar manner. In Equation (6), water // is incorporated via the formation of VMe , two (OH) groups and a unit of metal oxide. x H 2O + MeMe + 2OOx → [ VMe⋅ 2(OH)O ] + MeO x

(6)

Similar reactions can occur for metal cations of different charges, with corresponding numbers of (OH) and different oxide products. Reactions, such as those in Equations (7) and (8), involve interactions of water with impurity cations and vacancies on the oxygen sub-lattice. Within clinopyroxenes, Al Si/ and Na /Me substitutions can be charge balanced by the inclusion of (OH) as: / H 2O + 2 AlSi + 2OOx + VO•• → 2 [ AlSi⋅ (OH)O ] + OOx

( 7)

/ H 2O + 2 Na Me + VO•• + OOx → 2 [ Na Me⋅ (OH)O ]

(8)

x

x

The identity of the charge compensating defects in NAMs had been the subject of considerable debate in the literature, as has the extent to which these defects bind with the hydrogen. This is the sort of problem that can readily be addressed by computational methods, as the relative stabilities of known defect configurations can be assessed by calculating their formation energies, both bound and unbound. In addition, the O-H stretching frequency can be determined for each configuration and compared with experiment. In this way, the results obtained from the calculations can be used to both constrain models, and to aid in the interpretation of experimental data. In the following sections we consider the literature on hydrogen defects in the major mineral phases of the Earth’s upper mantle and transition zone; i.e., the Mg2SiO4 polymorphs and the clinopyroxenes diopside and jadeite. Hydrogen defects in a number of other important nominally anhydrous minerals, have been studied using computational methods including garnets (Wright et al. 1994; Nobes et al. 2000), quartz (Lin et al. 1994; Purton et al. 1992) and feldspar (Wright et al. 1996) although these will not be covered here.

The Mg2SiO4 polymorphs Forsterite. There is considerable experimental evidence for the presence of hydrogen in all three of the Mg2SiO4 polymorphs, as discussed in various chapters in this volume. Of the three, forsterite is by far the most well studied both experimentally (Kohlstedt et al. 1996; Matveev et al. 2001; Demouchy and Mackwell 2003) and computationally (see Table 1 for references). Olivine [(Fe,Mg)2SiO4] is the dominant mineral in the Earth’s upper mantle and thus will exert a major control on the rheological behavior. Natural samples show levels of hydrogen in the range 1 to 140 ppm H2O, where hydrogen is expected to be incorporated into the olivine lattice in association with both silicon and magnesium vacancies (e.g., Kohlstedt et al. 1996; Kohn 1996). Concentrations of OH in natural samples appear to show some correlation with the geological setting and composition suggesting that P/T history, as well as local stoichiometry, can affect the uptake of hydrogen. There is some evidence (Bell and Rossman 1992) suggesting that iron rich olivines contain a greater proportion of hydrogen than those with low iron content; however, this relationship has not been quantified in any way. Calculations, based on both QM and classical methods (Table 1), have been used to investigate the relative stability of hydrogen defects at different positions in the forsterite

74

Wright Table 1. Summary of calculations carried out on the Mg2SiO4 polymorphs. All calculations with the exception of those marked with *, were carried out at 0 K and 0 GPa. Mineral

Method

Reference

Forsterite

P-DFT P-DFT QM/MM MM MM MM

Haiber et al. (1997) Brodholt and Refson (2000) Braithwaite et al. (2003) Wright and Catlow (1994) de Leeuw et al. (2000) Walker et al. (2006)

Wadsleyite

P-DFT MM MM MM

Haiber et al. (1997)* Wright and Catlow (1996) Parker et al. (2004)* Walker et al. (2006)

Ringwoodite

MM P-DFT

Blanchard et al. (2005) Haiber et al. (1997)

lattice and to calculate energies of the reactions in Equations (5) and (6). Forsterite has an orthorhombic unit cell with isolated SiO4 tetrahedra separated by magnesium ions octahedrally coordinated by oxygens. There are two symmetry inequivalent magnesium positions, and three different oxygen sites, as shown in Figure 4. Looking at the structure of forsterite, we can identify a number of possible environments for hydrogen: (i) interstitial hydrogen bound to any one of the three oxygen sites but isolated from any cation vacancies; (ii) hydrogen bound to oxygen adjacent to either M1 or M2 vacancies; and (iii) hydrogen bound to oxygen adjacent to silicon vacancies. Of the three oxygen sites, all calculations agree that O3 is the most easily protonated, and that the M1 vacancy has a lower formation energy that M2. The most favorable defect configuration involving magnesium vacancies has hydrogen bound to two O2 oxygens around the vacant Mg1 site, as shown in Figure 5a. The third possibility, of hydrogen surrounding a silicon vacancy is shown in Figure 5b. Calculated binding energies (Brodholt and Refson 2000; Braithwaite et al. 2003; Walker et al. 2006) for [VMg⋅2(OH) ]x and [VSi⋅4(OH) ]x and associated clusters, given in Table 2, are ° ° substantial, and hence there is a strong driving force for hydroxyl groups to combine with cation vacancies. Haiber et al. (1997) have suggested that under mantle conditions entropy would cancel out any vacancy-hydrogen binding and therefore only isolated interstitial hydrogen defects would be present. However, the magnitude of the binding energies is sufficient to overcome the activation energy for hydrogen diffusion in olivine, estimated as 130 kJ·mol−1 (Mackwell and Kohlstedt 1990), so that cation vacancies will act as a ‘sink’ for unassociated hydroxyl species. Brodholt and Refson (2000) suggest that reactions with water could actively promote the formation of metal vacancies, particularly silicon vacancies, leading to a form of hydrolytic weakening. Calculated energies for the dissolution reactions in Equations (5) and (6) are given in Table 3 and show considerable variation depending on the methodology used. The P-DFT and QM/MM calculations compare well, with both methods showing that reactions of water with silicon vacancies will be exothermic. The error on the QM/MM values comes from uncertainties in the value for the lattice energy of oxide phases produced on formation of the defect, which by necessity had to be calculated using a periodic QM code. Calculated values from MM calculations are much higher in energy than either of the QM values, although reaction with

Atomistic Models of OH in NAM’s

75

Figure 4. The unit cell of forsterite (Mg2SiO4) viewed along the [100] direction.

(a)

(b)

Figure 5. Structure of hydrogen defect complexes in forsterite produced from the data of Braithwaite et al. (2003). (a) [VMg⋅2(OH) ]x cluster, and (b) hydrogarnet ° defect [VSi⋅4(OH) ]x cluster. °

Table 2. Defect binding energies in forsterite calculated using periodic DFT (Brodholt and Refson 2000) and QM/MM DFT (Braithwaite et al. 2003). Binding energy (kJ·mol−1) Reaction

1

// / + H•I → H Mg VMg

2

/ + H•I → ( 2H ) H Mg

3

VSi////

4

( 3H )

5

VSi////

6

( 3H )

/ // + H Mg → ( 4H ) + VMg

7

( 3H )

// + VMg

8

( 4H )

+ H•I / Si

/// → HSi

+ H•I

/ + H Mg / Si / Si x Si

X Mg

→ ( 4H )

X Si

/// → HSi

// + VMg X Si

→ ( 2H )

X Si

/ + H Mg

/ x + H Mg → ( 3H ) + (2H)Mg / Si

P-DFT

QM/MM DFT

−239

−245

−157

−203

−546

−519

−155

−202

−3.18 0.84 117 −2.89

76

Wright Table 3. Comparison of solution reaction energies in forsterite calculated using different techniques. P-DFT, Brodholt and Refson (2000), QM/MM, Braithwaite et al. (2003), MM Walker et al. (2006). Equation #

Reaction energy per OH (kJ·mol−1) P-DFT

QM/MM

MM

(5)

−23

−7 (± 30)

43

(6)

58

5 (± 30)

145

water leading to formation of the hydrogarnet defect is still most favorable. These energy differences most likely arise form the overestimation of polarization effects induced by the defect. All of the studies discussed above considered only defects in otherwise pure, perfect crystals of forsterite. However, other defects such as dislocations and grain boundaries will be sites for high concentrations of point defects and thus could be sinks for hydrogen. De Leeuw et al. (2000) investigated the formation of [VMg⋅2(OH)]x defect complexes in the bulk and along {010} tilt grain boundaries of forsterite. Their calculations showed this process to be more favorable along the grain boundary by over 100 kJ·mol −1 compared to the bulk. Infra-red (IR) spectroscopy has been used to measure the concentration of water in both natural olivine and synthetic forsterite. Polarized IR in particular can provide information on defect structures, as individual O-H vectors can be resolved. However, these analyses are complex with multiple peaks in the O-H stretching region, particularly in the case of mantle derived olivines that can make them difficult to resolve. Therefore the unambiguous assignment of specific frequencies to any one defect can be problematic. In general, two distinct groups of frequencies can be identified in the IR spectra of olivine, designated as group 1 for higher bands (3450 – 3650 cm−1), and group 2 for those in the lower frequency range of 3200-3450 (Bai and Kohlstedt 1993). IR frequencies can be calculated, within the harmonic approximation, from the second derivatives of the total energy with respect to each bond length. Braithwaite et al. (2003) identified two distinct bands analogous to those seen experimentally; the first, at around 3200 cm−1, was associated with the [VMg⋅2(OH) ]x defect cluster, and the second, higher band ° (3266-3478 cm −1) with the hydrogarnet [VSi⋅4(OH) ]x cluster. IR frequencies calculated using ° interatomic potential methods are consistently around 300 wavenumbers higher than those from the QM/MM embedded clusters (Walker et al. 2006). The assignment of OH bands to defects made by Braitewaite et al. (2003) is supported by the recent experimental work of Lemaire et al. (2004), who synthesized hydrous forsterite over a range of silica activity conditions. These authors found that samples with low silica activity (Si vacancies dominant) exhibited OH bands at 3620-3450 cm−1, while the high silica activity samples (Mg vacancies dominant) showed bands at 3160, 3220 and 3600 cm−1. Comparison of experimental and computational results is more difficult in the case of olivine, where the oxidation state of Fe and its influence on defect chemistry must be considered. Studies by Matveev et al. (2001) on synthetic olivine also attribute group 1 bands to OH at Si vacancies and group 2 to OH at Mg vacancies. However, these authors note that the signature of OH in mantle derived olivines is quite different to that of synthetic samples, which has been interpreted by Berry et al. (2005) as being due to the presence of trace elements. Wadsleyite. Wadsleyite is believed to be the dominant mineral in the upper part of Earth’s transition zone. The wadsleyite structure (Fig. 6) is based on a nearly perfect cubic closepacking of oxygen atoms with silicon atoms in tetrahedral sites. The structure is orthorhombic

Atomistic Models of OH in NAM’s

77

Figure 6. The unit cell of wadsleyite (Mg2SiO4) viewed along [001].

(space group Imma), but it also has a monoclinic polymorph (space group I2/m) recently identified by Smyth et al. (1997). Wadsleyite differs from forsterite in that it contains Si2O7 groups, so that there are four distinct oxygen sites as well as three magnesium sites. Smyth (1987), using analyses of bond strengths and electrostatic site potential, predicted that O1 would be a very favorable site to attach a proton, charge-balanced by metal vacancies. If all O1 sites were protonated then wadsleyite would be able to incorporate up to 3 wt% H2O (Smyth 1994) making it an enormous reservoir for water in the Earth’s mantle. A later study by Ross et al. (2003) carried out a similar investigation using the Laplacian of the electron density, as calculated from HF theory, to determine possible sites for protonation in wadsleyite and a number of other high pressure silicates. Their results also suggested that O1 would be the most likely site for protonation. However, Downs (1989), suggested that the O2 site could also be important. Calculations using P-DFT (Haiber et al. 1997) and interatomic potentials (Wright and Catlow 1996; Parker et al. 2004; Walker et al. 2006) methods confirm that O1 is the most favorable site for protonation, followed by O4, O3, and lastly by O2. The O1 site is quite isolated from silicon, and from other oxygens and is thus underbonded. Mulliken population analysis (Haiber et al. 1997) shows that O1 has an anomalously low charge that becomes more like normal oxygen when protonated. Magnesium vacancies, silicon vacancies and iron impurities have all been investigated with interatomic potential methods as possible charge compensating defects in wadsleyite. Mg1 has the lowest vacancy energy of the three magnesium positions, followed by Mg3 and Mg2, with Mg1 being approximately 95 kJ·mol−1 more stable than Mg2 (Walker et al. 2006). The solution reaction energies of wadsleyite and water are summarized in Table 4. The values of Parker et al. (2004) are for free energies calculated at simulated mantle conditions (1900 K and 15 GPa). This is in contrast to the other MM calculations discussed here, where it is the enthalpy that is calculated for 0 K and 0 GPa, so that the two are not directly comparable. However, the calculations all show that reactions leading to formation of the defect complex at the Mg3 site are most favorable. Table 4. Calculated solution reaction energies in wadsleyite. Equation # (5) (6)

Reaction energy per OH (kJ·mol−1) >5001 60

1

662

5883

2

1063

32

[1] Wright and Catlow (1996), [2] Parker et al. (2004), [3] Walker et al. 2006.

78

Wright

A further reaction considered by Wright and Catlow (1996) involved the incorporation of hydrogen via the reduction of ferric iron as: x H 2O + 2 Fe•Mg + 2 OOx → 2 ( OH )O + 2 Fe Mg + •

1 O2( g ) 2

( 9)

Here, Fe3+ is present as an impurity replacing magnesium at a magnesium site and reacts with a water molecule leading to the formation of two OH groups and Fe2+. This reaction was predicted to be exothermic, and therefore a highly favorable mechanism for water incorporation in wadsleyite. Haiber et al. (1997) carried out QM molecular dynamics calculations on the optimized O1-H defect, predicting an O-H stretching frequency of 3180 cm−1 (± 30 cm−1), at a simulated temperature of 400 K. The MM calculations of Walker et al. (2006) give a value of 3506 cm−1 for the same defect at a temperature of 0 K, the difference between the two values being similar to that found between MM and QM/MM calculation on forsterite. Walker et al. (2006) also calculated OH frequencies for defects associated with magnesium and silicon vacancies and found that, as with forsterite, lower frequency vibrations were related to OH at magnesium vacancies, and higher ones to OH associated with silicon vacancies. These results are broadly supported by the experimental work of Jacobsen et al. (2005) and with that of Kohn et al. (2002). Ringwoodite. The final polymorph in this family, ringwoodite, γ-Mg2SiO4, is the major constituent of the mantle between 520 and 660 km depth, i.e., lower part of the transition zone. It has a cubic spinel structure, as illustrated on Figure 6. The oxygens in this structure are close-packed with silicon in tetrahedral sites (isolated) and magnesium in octahedral sites (Fig. 7). Some disorder over the cation sites is considered likely. High-pressure experiments have shown that this mineral can incorporate up to 2.7 wt% H2O in its structure in the form of OH groups (Kohlstedt et al. 1996; Bolfan-Casanova et al. 2000). To date, only two computational studies of OH in ringwoodite have appeared in the Figure 7. The unit cell of ringwoodite literature. The DFT study of Haiber et al. (1987) (Mg2SiO4), which has the spinel structure. quotes a protonation energy for the oxygen that is comparable to the O2 and O3 values for forsterite, but less favorable than any of the oxygen sites in wadsleyite. No further information on the spinel structure is given. The second study is based on interatomic potentials (Blanchard et al. 2005) and investigates a range of different hydrogen defect positions. In addition to OH associated with magnesium and silicon vacancies, the authors consider the influence of iron (Eqn. 9) and of cation disorder. This last reaction is described by (Blanchard et al. 2005): x // H 2O + Mg Mg + SiSix + 3OOx → ⎡⎣ MgSi ( OH )2 ⎤⎦ + VMg + VO•• + SiO2 x

(10)

The calculated defect energies indicate that the most favorable mechanisms for hydrogen incorporation are coupled with reduction of iron (Eqn. 9) or with the creation of silicon vacancies. As with the other Mg2SiO4 polymorphs, binding energies between OH and cation vacancies are large and negative, so that isolated hydroxyls are not expected to occur in significant amounts. The solution reaction energies for Equations (5), (6), (9) and (10) calculated by Blanchard et al. (2005) are shown in Table 5 and indicate that substantial amounts of water could be incorporated in ringwoodite via reaction with ferric iron, and with silicon and magnesium vacancies.

Atomistic Models of OH in NAM’s

79

Summary of Mg2SiO4 polymorphs. InteTable 5. Calculated solution reaction grating the information obtained on hydrogen energies per OH in ringwoodite from defects by computational methods, some clear Blanchard et al. (2005). trends emerge. Firstly, it seems that the majority of hydrogen defects in forsterite, wadsleyite Equation # Energy (kJ·mol−1) and ringwoodite will be closely bound to cation vacancies and that the formation of isolated (5) −297 OH groups is not expected to be an important (6) −626 mechanism for the uptake of water in these (9) −680 minerals. In pure, iron free forsterite and ringwoodite, hydrogen can be easily incorporated (10) 170 through the formation of silicon defects. It has been suggested (Brodholt and Refson 2000) that reactions with water will drive the creation of silicon vacancies in olivine. A mechanism involving magnesium vacancies will be important in wadsleyite, especially the Mg3 site, and also in ringwoodite. Hydrogen incorporation via redox reactions with iron is exothermic in wadsleyite and ringwoodite and thus energetically favored. Grain boundaries and dislocations in forsterite are likely to contain high concentrations of hydrogen compared to the bulk thus deformation history could influence the ability of minerals to incorporate water. Finally, OHvibrational spectra for forsterite and wadsleyite indicate that higher frequencies (~3500-3800) are associated with silicon defect complexes, and low frequency vibrations (~3000-3300) with hydrogen at magnesium defects. In general, the results show excellent qualitative agreement between the different methods. Indeed, where results are comparable, as in the case of forsterite (Table 3), agreement is good quantitatively as well. However, it is worth pointing out that the majority of the above calculations have been performed at pressures of 0 GPa and thus the conclusions based on the results are only strictly valid for low pressure regimes.

Pyroxene Concentrations of OH in natural clinopyroxene samples range from 100-1300 ppm H2O (Skogby et al. 1990) with diopside, augite and omphacite showing the highest water concentrations of all NAMs, greater than olivine and pyrope garnet, which contain only trace amounts. Diopside (CaMgSi2O6) and jadeite (NaAlSi2O6) are monoclinic pyroxenes with space group C2/c. Each has two distinct cation sites; M1 is a regular 6-fold octahedral site, while M2, is distorted such that it becomes 8-coordinate, and three distinct oxygen sites. In diopside, the M1 site is generally occupied by Mg, and by Al in jadeite, while the M2 site is occupied by Ca and Na in diopside and jadeite respectively. There are also three different oxygen sites as shown on Figure 8. Gatzemeier and Wright (2006) have recently published the first study of hydrogen defects in clinopyroxenes. They use MM methods with interatomic potentials similar to those used by Wright and Catlow (1996) and Walker et al. (2006). Their calculations of intrinsic Schottky energies indicate that cation vacancies in both diopside and jadeite are more favored on the M2 than the M1 site, while the lowest energy oxygen vacancies are at different sites in the two phases—O2 in diopside and O3 sites in jadeite—although O2 is most easily

Figure 8. The unit cell of diopside and jadeite viewed along [001].

80

Wright

protonated in both. Reported solution reaction energies for various defect complexes are given in Table 6. In the pure phases hydrogen is most easily incorporated via the formation of [VSi(OH)4]x hydrogarnet type defects. When components of the two phases are mixed, then solution energies can become exothermic. The substitution of Al for Si and Na for Ca or Mg in diopside, provides favorable routes for hydrogen incorporation. In jadeite, Al rich compositions, with Al at Si sites, and the presence of both Ca and Mg at Al sites, also favor hydrogen incorporation, with exothermic values of solution energy. Thus the amount of water present in these minerals in the Earth’s upper mantle will vary with composition. Analysis of IR frequencies associated with O-H stretching at specific defect clusters in diopside and jadeite (Gatzemeier and Wright 2006) give hydrogen-oxygen bond lengths in good agreement with those correlated by Libowitzky (1999). Comparison of experimental and calculated IR frequencies were problematic, partly due to the complexity of experimental spectra, but also due to possible deficiencies in the ability of the model to accurately describe the O-H stretching frequency. Table 6. Calculated solution reaction energies per OH in diopside and jadeite from Gatzemeier and Wright (2006). Defect complex Equation #

Energy (kJ·mol−1)

Diopside (5) (6) (6) (7) (8) (8)

[VSi⋅4(OH) ] ° [VCa⋅2(OH) ]x ° [VMg⋅2(OH) ]x ° [AlSi⋅(OH) ]x °x [NaCa(OH) ] ° [NaMg⋅(OH) ]x ° x

Defect complex

Energy (kJ·mol−1)

Jadeite 87 227 203 −281 −105 −103

[VSi⋅4(OH) ]x ° [VAl⋅3(OH) ]x ° x [VNa⋅2(OH) ] ° [AlSi⋅(OH) ]x ° x [MgAl⋅(OH) ] ° [CaAl⋅(OH) ]x °

61 142 338 −335 −307 −277

GENERAL REMARKS AND FUTURE DIRECTIONS This review has concentrated on studies of the more important minerals in the Earth’s upper mantle and transition zone that contain hydrogen defects. For forsterite and wadsleyite there is good general agreement between the different computational methods used and with experiment. Thus a comprehensive atomistic model for OH in these minerals is emerging. For all of the phases considered, it appears that: (a) hydrogen is closely bound with metal vacancies and with charged impurities; (b) the magnitude of binding energies is such that water could facilitate the formation of point defects; and (c) the concentration of hydrogen is closely linked to chemistry and to the populations of other defects, such as dislocation, present in NAMs. However, the effects of pressure on defect energetics and on calculated frequencies must be considered before the results can be applied realistically to minerals under mantle conditions. Clearly simulations will continue to play a critical role in understanding the incorporation of water into NAMs and its influence on their properties. Of particular importance is the calculation of IR frequencies, and of the interaction of hydrogen with extended defects such as dislocations. The accurate prediction of IR frequencies associated with specific defect complexes can only be obtained using QM methods, and with advances in hardware and software, such calculations on minerals with large unit cells now become possible. Modeling dislocations in complex ionic materials is still in its infancy however, and requires a simulation

Atomistic Models of OH in NAM’s

81

cell containing many thousands of ions (Walker et al. 2005), making it more suited to MM approaches. Therefore both QM and MM techniques will continue to provide insights into OH defect behavior in nominally anhydrous minerals at the atomic level.

ACKNOWLEDGMENTS The bulk of the work reviewed here was carried out in the UK and funded by the Natural Environmental Research Council, the Engineering and Physical Sciences Research Council, the Royal Society, and the European Union. The author is grateful for funding from all of these agencies during the last 15 years and would also like to acknowledge very fruitful collaborations with Richard Catlow over this time. In addition, thanks to input from former students and post-docs Andrew Walker, Alex Gatzemeier, Marc Blanchard and Spencer Braithwaite. Finally, thanks to Julian Gale for discussions and comments on the manuscript.

REFERENCES Austen KF, Wright K, Slater B, Gale JD (2005) The interaction of dolomite surfaces with impurities: A computer simulation study. Phys Chem Chem Phys 7:4150-4156 Bai Q, Kohlstedt DL (1993) Effects of chemical environment on the solubility and incorporation mechanism for hydrogen in olivine. Phys Chem Minerals 19:460-471 Baram PS, Parker SC (1996) Atomistic simulation of hydroxide ions in inorganic solids. Phil Mag B 73:49-58 Becke AD (1993) A missing of Hartree-Fock and local density-functional theories. J Chem Phys 98:1372-1377 Bell DR, Rossman GR (1992) Water in the Earth’s mantle: the role of nominally anhydrous minerals. Science 255:1391-1397. Berry AJ, Hermann J, O’Neill HSC, Foran GJ (2005) Fingerprinting the water site in mantle olivine. Geology 33:869-872 Blanchard M, Gale JD, Wright K (2005) A computer simulations of OH defects in Mg2SiO4 and Mg2GeO4 spinels. Phys Chem Minerals 32(8-9):585-593 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Plan Sci Lett 182:209-221 Braithwaite JS, Sushko PV, Wright K, Catlow CRA (2002) Hydrogen defects in Forsterite: A test case for the embedded cluster method. J Chem Phys 116(6):2628-2635 Braithwaite JS, Wright K, Catlow CRA (2003) A theoretical study of the energetics and IR frequencies of hydroxyl defects in forsterite. J Geophys Res Solid Earth 108(B6) article 2284 Brenner DW, Shenderova OA, Harrison JA, Stuart SJ, Ni B, Sinnott SB (2002) A second-generation reactive empirical bond order (REBO) potential energy expression for hydrocarbons. J Phys Condens Matter 14: 783-802 Brodholt JP, Refson K (2000) An ab initio study of hydrogen in forsterite and a possible mechanism for hydrolytic weakening. J Geophys Res 105:18977-18982 Collins R, Catlow CRA (1992) Computer simulation of the structure and cohesive properties of micas. Am Mineral 77:1172-1181 Cygan RT, Kubicki JD (eds) (2001) Molecular Modeling Theory: Applications to the Geosciences. Reviews in Mineralogy and Geochemistry 42. Mineralogical Society of America. Cygan RT, Guggenheim S, van Groos AFK (2004) Molecular models for the intercalation of methane hydrate complexes in montmorillonite clay. J Phys Chem B 108:15141-15149 de Leeuw NH, Parker SC, Catlow CRA, Price GD (2000) Proton-containing defects at forsterite {010} tilt grain boundaries and stepped surfaces. Am Mineral 85:1143-1154 de Leeuw NH, Parker SC (2001) Density functional theory calculations of proton-containing defects in forsterite. Rad Effects Defects Solids 154:255-259 Demouchy S, Mackwell SL (2003) Water diffusion in synthetic iron-free forsterite. Phys Chem Minerals 30: 486-494 Dick BG Jr, Overhauser AW (1958) Theory of the dielectric constants of alkali halide crystals. Phys Rev 112: 90-103 Downs JW (1989) Possible sites for protonation in β-Mg2SiO4 from an experimentally derived electrostatic potential. Am Mineral 74:1124-1129 Foresman JB, Frisch A (1996) Exploring Chemistry with Electronic Structure Methods. Gaussian Inc.

82

Wright

Gale JD, Rohl AL (2003) The General Utility Lattice Program (GULP). Molecular Simulation 29:291-341 Gatzemeier A, Wright K (2006) Computer modeling of hydrogen defects in the clinopyroxenes diopside and jadeite. Phys Chem Minerals 33:115-125 Griffiths DJ (2005) Introduction to Quantum Mechanics. Pearson Prentice Hall Haiber M, Ballone P, Parrinello M (1977) Structure and dynamics of protonated Mg2SiO4: An ab initio molecular dynamics study. Am Mineral 82:913-922 Islam MS, Driscoll DJ, Fisher CAJ, Slater PR (2005) Atomic-scale investigations of defects, dopants and lithium transport in the LiFePO4 olivine type battery material. Chem Mater 17:5085-5092 Jacobsen SD, Demouchy S, Frost DJ, Boffa-Ballaran T, Kung J (2005) A systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single-crystal X-ray diffraction: oxygen sites for hydrogen storage in the Earth’s interior. Am Mineral 90:61-70 Kerisit S, Marmier A, Parker SC (2005) Ab initio surface phase diagram of the {10-14} calcite surface. J Phys Chem B 109(39):18211-18213 Kohlstedt DL, Keppler H, Ruby DC (1996) Solubility of water in the α, β and γ phases of (Mg,Fe)2SiO4. Contrib Mineral Petrol 123:345-357 Kohn SC (1996) Solubility of H2O in nominally anhydrous minerals using 1H MAS NMR. Am Mineral 81: 1523-1526 Kohn SC (2006) Structural studies of OH in nominally anhydrous minerals using NMR. Rev Mineral Geochem 62:53-66 Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ (2002) Ordering of hydroxyl defects in hydrous wadsleyite (beta-Mg2SiO4). Am Mineral 87:293-301 Kresse G, Furthmuller J (1996a) Efficient iterative schemes for ab initio total energy calculations using a planewave basis set. Phys Rev B 54:11169-11186 Kresse G, Furthmuller J (1996b) Efficiency of ab initio total energy calculations for metals and semiconductors using a planewave basis set. Comput Mater Sci 6:15-50 Kröger FA (1972) The Chemistry of Imperfect Crystals. North-Holland Leach AR (2001) Molecular Modeling Principles and Applications. Prentice Hall Leininger ML, Allen WD, Schaefer HF, Sherrill CD (2000) Is Moller-Plesset perturbation theory a convergent ab initio method? J Chem Phys 112(21):9213-9222 Lemaire C, Kohn SC, Brooker RA (2004) The effect of silica activity on the incorporation mechanisms of water in synthetic forsterite: a polarized infrared spectroscopic study. Contrib Mineral Petrol 147:48-57 Leslie M, Gillan MJ (1985) The energy and elastic dipole tensor of defects in ionic crystals calculated by the supercell method. J Phys Chem 18:973-982 Libowitzky E (1999) Correlation of O–H stretching frequencies and O–H···O hydrogen bond length in minerals. Monatshefte für Chemie 130:1047-1059 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52 Lin JS, Payne MC, Heine V, McConnell JDC (1994) Ab initio calculations of (OH)4 defects in alpha-quartz. Phys Chem Minerals 21:150-155 Mackwell SL, Kohlstedt DL (1990) Diffusion of hydrogen in olivines: Implications for water in the mantle. J Geophys Res 95:5079-5088 Matveev S, O’Neill HSC, Ballhaus C, Taylor WR, Green DH (2001) Effect of silica activity on OH− IR spectra of olivine: implications for low-αSiO2 mantle metasomatism. J Petrol 42:721-729 Mott NF, Littleton MJ (1938) Conduction in polar crystals. I. Electrolytic conduction in solid salts. Trans Faraday Soc 34:485 Nobes RH, Akhmatskaya EV, Milman V, White JA, Winkler B, Pickard CJ (2000) An ab initio study of hydrogarnets. Am Mineral 85:1706-1715 Oganov AR, Price GD (2005) Ab initio thermodynamics of MgSiO3 perovskite at high pressure and temperature. J Chem Phys 122(12) art. No. 124501 Parker SC, Cooke DJ, Kerisit S, Marmier AS, Taylor SL, Taylor SN (2004) From HADES to PARADISE – atomistic simulations of defects in minerals. J Phys Cond Matter 16:S2735-S2749 Purton J, Jones R, Heggie M, Oberg S, Catlow CRA (1992) Pseudopotential calculations of the alpha-quartz structure and hydrogarnet defect. Phys Chem Minerals 18:389-392 Putnis A (1992) Introduction to Mineral Sciences. Cambridge University Press Ross NL, Gibbs GV, Rosso KM (2003) Potential docking sites and positions of hydrogen in high-pressure silicates Am Mineral 88(10):1452-1459 Rossman GR (2006) Analytical methods for measuring water in nominally anhydrous minerals. Rev Mineral Geochem 62:1-28 Saul P, Catlow CRA, Kendrick J (1985) Theoretical studies of protons in sodium hydroxide. Philos Mag B 51: 107-117

Atomistic Models of OH in NAM’s

83

Schröder KP, Sauer J, Leslie M, Catlow CRA, Thomas JM (1992) Bridging hydroxyl groups in zeolitic catalysts: a computer simulation study. Chem Phys Lett 188:320-325 Segall MD, Lindan PJD, Probert MJ, Pickard CJ, Hasnip PJ, Clarke SJ, Payne MC (2002) First-principles simulation: ideas, illustrations and the CASTEP code. J Phys Cond Matter 14:2717-2743 Skogby H, Bell DR, Rossman GR (1990) Hydroxide in pyroxene: variations in the natural environment. Am Mineral 75:764–774 Smyth JR (1987) β-Mg2SiO4: a potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR (1994) A crystallographic model for hydrous wadsleyite (β-Mg2SiO4): an ocean in the Earth’s interior? Am Mineral 79:1021-1024 Smyth JR, Kawamoto T, Jacobson SD, Swope RJ, Hervig RL, Holloway JR (1997) Crystal structure of monoclinic hydrous wadsleyite [β-(Mg,Fe)2SiO4]. Am Mineral 82:270-275 Steele HM, Wright K, Hillier IH (2002) Modeling the adsorption of uranyl on the surface of goethite. Geochimica Cosmochim Acta 66:1305-1310 Stillinger FH, David CW (1980) Study of the water octamer using the polarization model of molecular interactions. J Chem Phys 73:3384-3389 Tilley RJD (1987) Defect Crystal Chemistry and its Applications. Blackie and Sons Ltd. Vocadlo L, Alfe D, Gillan MJ, Price GD (2003) The properties of iron under core conditions from first principles calculations. Phys Earth Planet Ints 140:101-125 Walker AM, Wright K, Slater B (2003) A computational study of oxygen diffusion in olivine. Phys Chem Minerals 30:536-545 Walker AM, Slater B, Gale JD, Wright K (2005) Atomic scale modeling of the cores of dislocations in complex materials part 2: applications. Phys Chem Chem Phys 7(17):3235-3242 Walker AM, Demouchy S, Wright K (2006) Computer simulation of hydroxyl groups in α- and β-Mg2SiO4. Euro J Mineral (in press) Wright K, Freer R, Catlow CRA (1994) Energetics and structure of the hydrogarnet defect in grossular: A computer simulation study. Phys Chem Minerals 20:500-504 Wright K, Catlow CRA (1994) A computer simulation study of OH defects in olivine. Phys Chem Minerals 20: 515-518 Wright K, Catlow CRA (1996) Calculations on the energetics of water dissolution in wadsleyite. Phys Chem Minerals 23:38-41 Wright K, Catlow CRA, Freer R (1996) Water related defects and oxygen diffusion in albite. Contrib Mineral Petrol 125:161-166 Wright K, Cygan RT, Slater B (2002) Impurities and non-stoichiometry in the bulk and on the (10-14) surface of dolomite. Geochimica Cosmochim Acta 66:2541-2546 Wright K, Gale JD (2004) Interatomic potentials for the simulation of the zinc-blende and Wurtzite forms of ZnS and CdS: Bulk structure, properties and phase stability. Phys Rev B 70:03521

5

Reviews in Mineralogy & Geochemistry Vol. 62, pp. 85-115, 2006 Copyright © Mineralogical Society of America

Hydrogen in High Pressure Silicate and Oxide Mineral Structures Joseph R. Smyth Department of Geological Sciences University of Colorado Boulder, Colorado, 80309, U.S.A. e-mail: [email protected]

INTRODUCTION Earth is the water planet. Liquid water covers more than 70% of the surface and dominates all surface processes, geological, meteorological, and biological. However the hydrosphere composes only about 0.025% of the planet’s mass, so that small amounts of H incorporated into the oxygen minerals of the interior may constitute the majority of Earth’s total water. The Earth is thought to be generally similar in composition to the chondrite meteorites which average about 0.10% by weight H2O. So if the Earth were strictly chondritic in its H content, about 75% of that H as water would have either been tied up in the minerals of the interior or lost to space. Understanding how H behaves at the atomic scale in these materials will help us to understand how the Earth balances and retains its water and may help us to understand how water planets develop and how common they might be. In addition to the surface processes, water also controls the processes of the interior. Water dramatically reduces the melting temperature of rocks controlling igneous processes. Even trace amounts of hydrogen have a major effect on some physical properties such as deformation strength and electrical conductivity (Karato 1990). The nominally anhydrous minerals of the Earth’s interior are capable of incorporating many times the amount of water in the hydrosphere, and these phases would need to be saturated before stoichiometrically hydrous minerals could be stable. Hydrogen in amounts reported in olivine, wadsleyite, and ringwoodite by Kohlstedt et al. (1996) as recalibrated by Bell et al. (2003), if present in the Earth, would constitute a significant fraction of the total water budget of the planet. The amounts that can be incorporated into the nominally anhydrous minerals of the Transition Zone (410-660 km depth) may constitute the largest reservoir of water in the planet and may have controlled the chemical evolution and interior processes of the planet. Hirschmann et al. (2005) have estimated the storage capacities of the various mineral reservoirs in the mantle. These volumes of water imply that there may be a deep water cycle in the Earth whereby some of the water in subducted slabs may be returned to the large deep interior reservoir and then be released in mid-ocean ridge basalts so that the amount of water in the Earth’s oceans would represent a dynamic balance between these processes. This process would depend on the ability of the nominally anhydrous phases of the upper mantle to incorporate the water released by the breakdown of the hydrous phases on increasing pressure and temperature with subduction.

GEOCHEMISTRY OF H Hydrogen is the most abundant element in the cosmos, and the geochemical behavior of hydrogen is unlike that of any other element. Because the proton does not behave like other cations in the crystal, it is generally inappropriate to treat H as an incompatible element 1529-6466/06/0062-0005$05.00

DOI: 10.2138/rmg.2006.62.5

86

Smyth

or to compare its compatibility with other cations. In the highly reducing conditions of the condensing solar nebula, H was primarily atmophile, as the diatomic gas H2, but also as methane, ammonia, and water. However, in the Earth’s crust and mantle, hydrogen in its ionic state, H+, is strongly lithophile. It substitutes readily in silicates and other oxygen minerals in both trace and stoichiometric amounts. Because H does not occupy a normal cation site in a mineral, it does not have an effective ionic radius that controls its geochemical behavior. Its compatibility is therefore not systematic as other trace cations are, but strongly dependent on temperature, pressure, and the chemical activity of possible charge-balancing cations. The chalcophile nature of H is not well known, and its substitution in sulfide minerals in trace amounts is difficult to measure and poorly studied. H2S is an abundant volatile in mafic to silicic volcanic systems, and there are a few OH-bearing sulfide minerals such as tochilinite [6(Fe0.9S)·5(Fe,Mg)(OH)2)] (Beard 2000), but I was unable to identify a single H-bearing sulfide mineral that does not also contain oxygen. Under reducing conditions, neutral H is highly soluble in metallic liquids and forms solid metal hydrides. However, very little is known about H partitioning between silicate and metallic liquids, and the amount of H in the core is unknown as is its effect on liquid metal densities under conditions of the core. The objectives of this review are to examine the various structural substitution mechanisms whereby H enters major high pressure silicate and oxide minerals in stoichiometric amounts and then use this information to look at H substitution in nominally anhydrous minerals of the Earth’s mantle.

CRYSTAL CHEMISTRY OF H Because oxygen is the only anionic species of significant abundance in the crust and mantle, we think of hydrogen and water as synonymous. At low pressure, water can enter silicates either as molecular water or as hydroxyl, or both. In low-density silicates such as zeolites and clays, the water molecules are located in large cavities or interlayer sites and freely flow into and out of the crystals. In other low-temperature minerals such as gypsum, the molecular water is tightly bound structurally and does not exchange. Hydrogen also enters low temperature and pressure minerals as structural hydroxyl. At higher pressures, hydrogen occurs in the solid minerals of the mantle in several forms, but generally does not exchange. It can be present as discrete, structurally bound water molecules as in lawsonite, K-cymrite, or 10 Å phase, but most often it is present as hydroxyl, OH−. The hydroxyl can be stoichiometric, part of the nominal mineral formula, or it can be a minor constituent, where the hydrogen may substitute ionically for other cations in the structure. In nominally hydrous silicates, the hydroxyl rarely bonds directly to the Si cation. This is also true of other small, high-fieldstrength cations such as B, C, P, and S6+. The proton position is difficult to locate by X-ray diffraction, but neutron single-crystal or powder diffraction can give proton positions with high precision. Additionally, the protonated oxygen is relatively easy to identify from X-ray data by a simple Pauling bond strength calculation. When H enters nominally anhydrous minerals, the proton does not occupy the normal cation position, but attaches to one or more of the oxygens. The usual proton to oxygen nucleus distance (0.95 to 1.2 Å) is less than the nominal oxygen radius (1.32 to 1.4 Å). The oxygen atoms that can be protonated in a stoichiometric, fully occupied structure are those that are most underbonded. The degree of underbonding can be calculated on the basis of Pauling bond strength or a Madelung site potential calculation. Pauling bond strength at the oxygen is calculated as the sum of the bond strengths (nominal cation valence divided by coordination number) around an oxygen atom. The Madelung site potential (Smyth 1987, 1989) is the nominal valence charge divided by distance and summed to convergence. These methods may identify the oxygen most likely to be protonated if there are several non-equivalent oxygen positions in a structure, but does not identify the proton location. Libowitzky (1999) reports a correlation of O-H-O distance with O-H stretching frequency. This has been used together

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

87

with polarization vectors to deduce proton positions in nominally anhydrous structures such as wadsleyite (Kohn et al. 2002) and akimotoite (Bolfan-Casanova et al. 2002). Ross et al. (2003) propose a computational method to identify non-bonding electron-pairs on oxygens in order to locate potential docking sites for protons in high pressure silicates. Extensive protonation of an oxygen site in a nominally anhydrous mineral generally requires a charge balancing substitution or a cation vacancy. Cation vacancies normally result in a significant expansion of the vacant coordination polyhedron, however such vacant polyhedra are typically large and highly compressible (Jacobsen 2006). Tetrahedral cation vacancies charge-balanced by protons are well documented. This is the so-called hydrogarnet substitution because the H4O4 tetrahedron can completely replace the silicate tetrahedron in hydrogarnets (Lager et al. 2005). The H4O4 tetrahedron is larger than the silicate tetrahedron with the Si-O distance in silicate garnets being 1.60 to 1.64 Å whereas the equivalent distance (4-O) in hydrogarnet is over 2.0 Å. This means that pressure will inhibit this substitution mechanism so that garnets from natural high pressure (2-5 GPa) environments generally contain less than about 50 ppmw H2O (Bell and Rossman 1992). It may be possible that this substitution mechanism may again become viable at pressures above about 7 GPa, as it has been proposed to be present in hydrous coesite above this pressure (Koch-Mueller et al. 2003). Oxygen-oxygen edges are typically 2.6 to 2.8 Å for tetrahedral silicon, whereas edges of Mg octahedra typically are 2.8 to 3.0 Å. These distances have been used to infer proton positions from infrared spectra based on the calibration of Libowizky (1999) (e.g., Kohn et al. 2002), however the correlation curve is quite flat in this region and cation vacancy may result in local distortion of the coordination polyhedra. Octahedral cation vacancy charge-balanced by protons appears to be more common at pressures of the upper mantle. Protonated octahedral cation vacancy appears to become a very significant substitution mechanism in olivine (Smyth et al. 2006a) and wadsleyite (Smyth et al. 1997; Ross et al. 2003; Jacobsen et al. 2005). Wadsleyite (β-Mg2SiO4) can contain more than 3 wt% H2O (Inoue et al. 1995; Kohlstedt et al. 1996), where the charge balance mechanism is octahedral site vacancy, principally at M3 (Smyth et al. 1997; Kohn et al. 2003). Even trace hydration (10 to 1000 ppmw H2O) can have very large effects of physical properties of nominally anhydrous phases such as mechanical strength (Kavner 2003), effective viscosity (Karato et al. 1986), and electrical conductivity (Karato 1990; Huang et al. 2005). Minor hydration (1000 to 10000 ppmw H2O) can have a major effect on density, compressibility (Smyth et al. 2003; 2004), seismic velocity (Jacobsen et al. 2005; 2006), and pressure-temperature conditions of phase transitions (Wood 1995; Smyth and Frost 2002). In order to understand the crystal chemistry of H at high pressure, it is necessary to first look briefly at the nominally hydrous phases on the Earth’s mantle and then to examine the mechanisms for minor and trace substitution of H in the nominally anhydrous silicates and oxides that compose the mantle. The dense hydrous magnesium and aluminum silicate phases covered here are listed in Table 1 along with formulae, cell parameters and calculated densities. The dense anhydrous magnesium and aluminum silicate phases covered here are listed in Table 2.

NOMINALLY HYDROUS HIGH-PRESSURE SILICATE PHASES Compositions of the dense hydrous magnesium silicate (DHMS) phases can be displayed in the magnesia-silica-brucite (MgO-SiO2-Mg(OH)2) ternary (Fig. 1). Along the anhydrous edge fall periclase (MgO), anhydrous phase B, forsterite and its polymorphs (wadsleyite and ringwoodite), enstatite and its polymorphs (akimotoite, and perovskite-type MgSiO3) and quartz and its polymorphs (coesite, and stishovite). On the brucite-forsterite join lie phase A and the humites (norbergite, chondrodite, humite and clinohumite). Near the bruciteanhydrous phase B join, lie phase B and super-hydrous phase B (Fig. 1).

Formula

Mg(OH)2 Mg3Si2O5(OH)4 Mg3Si4O10(OH)2 Mg5Al2Si3O10(OH)8

KAl2AlSi3O10(OH)2 KMg3AlSi3O10(OH)2 KMgAlSi4O10(OH)2

(Mg,Fe)7Si8O22(OH)2 Ca2Mg5Si8O22(OH)2

CaAl2Si2O7(OH)2·H2O

Ca2Al3Si3O12(OH) Ca2Al3Si3O12(OH) Ca2FeAl2Si3O12(OH)

Mg9Si4O16(OH)2 Mg5Si2O8(OH)2

Mg7Si2O8(OH)6 Mg12Si4O19(OH)2 Mg10Si3O14(OH)4 MgSi2O4(OH)2 Mg2SiO2(OH)4

AlSiO3(OH) Al3Si2O7(OH)3 Al2SiO4(OH)2

KAlSi3O8 H2O

Mineral

Brucite Serpentine Talc Chlorite

Mica Group Muscovite 2M1 Phlogopite1M Phengite 2M1

Amphibole Group Cummingtonite Tremolite

Lawsonite

Epidote Zoisite Clinozoisite Epidote

Humite Group Clinohumite Chondrodite

Phase A Phase B Suphyd. Phase B Phase D Phase E

Phase Egg Phase Pi Topaz-OH

K-Cymrite 296.356

120.074 300.137 180.063

456.398 742.096 619.403 178.499 176.739

621.162 339.744

454.366 454.366 483.232

314.243

843.953 812.419

398.317 417.290 396.753

58.327 277.137 379.294 555.838

F.W. (g)

4.3346 7.2832 8.9207 5.3348

7.868 10.588 5.089 4.745 2.967 7.1441 6.0885 4.7203

P63 P21/c Pnnm P 31m R 3m P21/n P1 Pbnm P6/mmm 5.3348

7.868 14.097 13.968 4.775 2.967

4.741 4.7459

10.275 10.3480

5.550 5.583 5.628

5.847

18.1833 18.048

9.015 9.190 9.037

3.142 5.332 9.173 9.227

b (Å)

P21/b P21/b

16.188 8.861 8.888

8.795

Ccmm

Pnma P21/m P21/m

9.5220 9.863

C2/m C2/m

5.192 5.308 5.205

3.142 5.332 5.290 5.327

P 3 m1 P31m C1 C2/m

C2/c C2/m C2/c

a (Å)

S.G.

7.7057

6.9525 7.7234 8.4189

9.577 10.073 8.696 4.345 13.886

13.704 7.9002

10.034 10.141 10.152

13.142

5.3184 5.285

20.046 10.166 19.886

4.766 7.223 9.460 14.327

c (Å)

90

90 115.71 90

90 90 90 90 90

100.1 108.70

90 90 90

90

90 90

90 90 90

90 90 90.46 90

α (°)

90

98.40 88.85 90

90 104.1 90 90 90

90 90

90 115.46 115.38

90

102.020 104.79

95.74 100.10 95.62

90 90 98.68 96.81

β (°)

2 4 2 1 1

2 2

4 2 2

4

2 2

4 2 4

1 1 2 2

Z

120

1

90 4 92.89 2 90 4

120 90 90 120 120

90 90

90 90 90

90

90 90

90 90 90

120 120 90.09 90

γ (°)

114.372

32.066 92.795 53.371

154.569 219.530 186.122 59.287 73.613

201.006 116.094

135.719 136.388 138.147

101.745

271.634 273.88

140.546 147.012 140.133

24.54 107.24 136.63 210.540

MVol (cm3)

Table 1. Physical properties of nominally hydrous high pressure silicate phases.

2.591

3.744 3.234 3.373

2.952 3.380 3.327 3.010 2.401

3.089 2.926

3.347 3.331 3.497

3.088

2.879 2.966

2.834 2.838 2.831

2.377 2.584 2.776 2.639

Fasshauer et al. (1997)

Schmidt et al. (1998) Wunder et al. (1993) Northrup et al. (1994)

Kagi et al. (2000) Finger et al. (1993) Pacalo and Parise (1992) Yang et al. (1997) Shieh et al. (2000)

Ross and Crichton (2001) Ross and Crichton (2001)

Grevel et al. (2000) Pawley et al. (1996) Gabe et al. (1973)

Baur (1978)

Yang et al. (1998) Hawthorne et al. (1976)

Rothbauer (1971) Hazen and Burnham (1973) Smyth et al. (2000)

Zigan and Rothbauer (1967) Mellini (1982) Perdikatsis and Burzlaff (1981) Smyth et al. 1997

ρ STP References (g/cm3)

88 Smyth

MgO FeO Al2O3 TiO2

SiO2 SiO2 SiO2

Mg2SiO4 Fe2SiO4

Mg2SiO4 Mg2SiO4 Mg2SiO4 Fe2SiO4

Mg14Si5O24

Mg3Al2Si3O12 Fe3Al2Si3O12 Mg3(MgSi)Si3O12

Mg2Si2O6 Fe2Si2O6 Mg2Si2O6 NaAlSi2O6 CaMgSi2O6

Mg2Si2O6 Al2SiO5

MgSiO3 MgSiO3 ZrSiO4 CaTiSiO5

Periclase Wüstite Corundum Rutile

Quartz Coesite Stishovite

Olivine Forsterite Fayalite

Wadsleyite Wadsleyite II Ringwoodite γ-Fe2SiO4

Anhyd. Phase B

Garnets Pyrope Almandine Majorite

Pyroxenes Orthoenstatite Orthoferrosilite Clinoenstatite Jadeite Diopside

Akimotoite Kyanite

Perovskite Post-perovskite* Zircon Titanite 100.397 100.397 183.304 196.063

200.795 162.047

200.795 263.865 200.795 202.139 232.330

403.153 497.758 401.590

864.789

140.709 140.709 140.709 203.779

140.709 203.779

60.086 60.086 60.086

40.311 71.846 101.961 79.899

F.W. (g)

5.908

Pmcb

18.227 18.427 9.618 9.423 9.750 4.728 7.126 4.775 2.65 6.6042 7.069

Pbca Pbca P21/c C2/c C2/c R3 P1 Pbnm Cmcm I41/amd P21/a

11.452 11.531 11.501

5.711 5.6884 8.092 8.234

Imma Imma Fd 3 m Fd 3 m

Ia 3 d Ia 3 d I41/a

4.753 4.820

4.929 8.69 6.6042 8.722

4.728 7.852

8.819 9.076 8.815 8.564 8.926

11.452 11.531 11.501

14.241

11.467 28.924 8.092 8.234

10.190 10.479

4.914 12.370 4.179

P3121 4.914 C2/c 7.137 P42/mnm 4.179 Pbnm Pbnm

4.211 4.311 4.7589 4.5845

4.211 4.311 4.7589 4.5845

Fm 3 m Fm 3 m R 3c P42/mnm

b (Å)

a (Å)

S.G.

* decompressed from Murakami et al. (2004) assuming K = 300 GPa and K′ = 5

Formula

Mineral

6.908 6.59 5.9796 6.566

13.559 5.575

5.179 5.237 5.176 5.223 5.251

11.452 11.531 11.480

10.069

8.256 8.2382 8.092 8.234

5.978 6.087

5.405 7.174 2.665

4.211 4.311 12.9912 2.9533

c (Å)

90 90 90 90

90 89.99

90 90 90 90 90

90 90 90

90

90 90 90 90

90 90

90 90 90

90 90 90 90

α (°)

90 90 90 113.86

90 101.11

90 90 108.37 107.56 105.90

90 90 90

90

90 90 90 90

90 90

90 119 90

90 90 90 90

β (°)

90 90 90 90

120 106.03

90 90 90 90 90

90 90 90

90

90 90 90 90

90 90

120 90 90

90 90 120 90

γ (°) 11.242 12.060 25.577 18.693

MVol (cm3)

4 4 4 4

3 4

8 8 4 4 4

8 8 8

2

8 20 8 8

4 4

24.482 22.89 39.264 55.739

52.700 44.219

62.666 65.930 62.704 60.498 66.167

113.056 115.412 114.305

255.081

40.699 40.812 39.886 42.023

43.596 46.283

3 22.685 16 20.573 2 14.014

4 4 6 2

Z

Table 2. Physical properties of nominally anhydrous high pressure silicate phases.

4.100 4.386 4.668 3.517

3.810 3.664

3.204 4.002 3.202 3.341 3.511

3.565 4.312 3.513

3.390

3.457 3.447 3.527 4.848

3.227 4.402

2.648 2.920 4.287

3.585 5.956 3.986 4.274

Horiuchi et al. (1987) Murakami et al. (2004) Hazen and Finger (1976) Kek et al. (1995)

Horiuchi et al. (1982) Winter and Ghose (1979)

Yang and Ghose (1995) Sueno et al. (1976) Pannhorst (1984) Cameron et al. (1973) Cameron et al. (1973)

Armbruster et al. (1992) Armbruster et al. (1992) Angel et al. (1989)

Hazen et al. (1992)

Finger et al. (1993) Smyth et al. (2005) Smyth et al. (2004) Yagi et al. (1974)

Smyth et al. (2006) Fujino et al. (1981)

Kihara (1990) Smyth et al. (1987) Ross et al. (1990)

Hazen (1976) Hazen (1981) Newnham and DeHaan (1962) Shintani et al. (1975)

ρ STP References (g/cm3)

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures 89

90

Smyth

Figure 1. The compositions of the dense, hydrous and anhydrous magnesium silicate phases displayed on the MgO-SiO2-Mg(OH)2 ternary. W denotes the field for wadsleyite, olivine, and ringwoodite; aB is anhydrous phase B; B is phase B, sB is superhydrous phase B; CH clinohumite; H humite, Chd chondrodite; Nb norbergite; A phase A; E phase E; and D phase D.

Brucite Brucite, Mg(OH)2 is the first phase discussed among the nominally hydrous minerals of the mantle. Although brucite is not a silicate, it forms a prominent structural component in many silicate minerals. Because of its very high water content, more than 30% by weight or about 75% water by volume, it is not a likely mantle mineral. Brucite forms the most hydrous end member in our systems and is a common ingredient in starting compositions to experimentally produce hydrous high pressure phases. This component, with some Al substitution also occurs in the chlorite structure. The brucite structure (Fig. 2) is trigonal, P 3 m1, and consists of tri-octahedral layers of Mg(OH)6 octahedra parallel to (001). All oxygens in the structure are equivalent and protonated, so that each oxygen is bonded to three Mg atoms and one proton. The layers are bonded by relatively weak hydroxyl bonds giving the mineral its perfect basal cleavage. Gibbsite, Al(OH)3, is isostructural with brucite except that one third of the octahedra are vacant.

Serpentine Serpentine, ideally Mg3Si2O5(OH)4, is a major alteration phase in ultramafic rocks. It is stable at ambient pressure and to depths of roughly 250 km in a cool, subducting slab (Kawamoto et al. 1996; Schmidt and Poli 1998). It contains roughly 13% H2O by weight which corresponds to more than 30% by volume. The structure consists of a tri-octahedral brucite-like layer attached to a single pure-silica tetrahedral layer (Fig. 3). The structure has several stacking polytypes, but most are similar in composition and density. In its absestiform habit known as chrysotile, the sheets are rolled into tubes, so that the actual space groups and structure are not well defined. Several stacking polytypes have been described with differing degrees of order each having a density of about 2.58 g/cm3. Lizardite 1H is trigonal P31m (Table 1) (Mellini 1987). The well crystallized massive form is known as antigorite, the space group is triclinic, P1. There are no Si-OH bonds in the structure, so that each oxygen is bonded either to three Mg and one Si or to three Mg and one proton.

Talc Talc, Mg3Si4O10(OH)2, is also a major alteration phase of mafic rocks. It contains more silica than serpentine and may occur in more siliceous rock compositions than serpentine. The water content is a bit less than 5% by weight. Its triclinic, C 1 structure (Fig. 4) is that of a T-O-T layer silicate like mica, but without interlayer cations. The bonding between layers is just the weak Van der Waals bonds resulting in a very soft and easily deformable structure. There are no Si-OH bonds in the structure so that one sixth of the oxygen atoms are protonated and bonded to one proton and three Mg atoms. The remaining oxygen atoms are each bonded to one Si and three Mg atoms.

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

91

Figure 2. The structure of brucite (Mg(OH)2) is trigonal, P 3 m1. All oxygen atoms are equivalent and bonded to one H and three Mg atoms. The Mg octahedra are arranged in a sheet parallel to (001). The sheets are H-bonded together giving the mineral its perfect basal cleavage.

Figure 3. The simple trigonal structure of the serpentine mineral lizardite, Mg3Si2O5(OH)4, is P31m. The octahedral Mg atoms are arranged in a trioctahedral sheet as in brucite. All nonsilicate oxygens are protonated.

Figure 4. The structure of talc, Mg3Si4O10(OH)2, is triclinic, C1. The octahedral Mg atoms are arranged in a trioctahedral sheet as in brucite and serpentine except tat there are tetrahedral sheets on both sides of the octahedral sheet. Again, all non-silicate oxygens are protonated.

True micas The true micas (Fig. 5) have a T-O-T layer like that of talc, but one fourth of the Si cations are replaced by Al and charge balanced by an interlayer alkali cation, dominantly K. Like talc, the micas contain 4.5 to 5% H2O by weight. In muscovite, KAl2AlSi3O10(OH)2, the octahedral layer is dioctahedral with two Al cations, whereas in biotite and phlogopite, KMg3AlSi3O10(OH)2, it is trioctahedral with three divalent cations, Mg or Fe, per formula unit. The phengite substitution into the dioctahedral micas puts additional silicon into the tetrahedral layer in place of Al which is charge-balanced by Mg in the dioctahedral layer. This substitution is stabilized by pressure, and high-silica phengites have been synthesized at pressures as high as 11 GPa (Domanik and Holloway 1996; Smyth et al. 2000). Phengite is stable in a mafic composition to over 300 km depth if K is present and temperatures are low as in a subducting slab. The micas exist in several polytypes, that is, different stacking sequences,

92

Smyth

Figure 5. The structure of phlogopite 1M, KMg3AlSi3O10(OH)2 is monoclinic C2/m. One third of the Si atoms in the tetrahedral layer are replaced by Al and charge-balanced by the interlayer K atoms (gray sphere). Muscovite, KMg3AlSi3O10(OH)2, is similar except that one third of the octahedra are vacant and the rest replaced by Al. There are several distinct stacking arrangements called polytypes.

predominantly 2M1 (C2/c) and 3T (P3112) in dioctahedral micas, and 1M (C2/m) and 2M1 in trioctahedral micas. The different polytypes commonly coexist in natural samples and are so close in physical properties that separate stability fields for the different polytypes have not been documented. Again, there are no Si-OH bonds and protons coordinate the non-silicate oxygens in the octahedral layer. The 10 Å phase, Mg3Si4O10(OH)·H2O, is a mica-like dense hydrous magnesium silicate phase that occurs at 3-5 GPa as a breakdown product of serpentine and chlorite (Yamamoto and Akimoto 1977). It is structurally similar to talc and phlogopite, but has neutral molecular water in the inter-layer (Fumagalli et al. 2001; Comodi et al. 2005). It is likely to be an important host phase for H in subducting hydrated lithosphere (Fumagalli and Poli 2005).

Chlorite Chlorite, Mg3AlSi3O10(OH)2·Mg2Al(OH)6, is another low pressure alteration phase of mafic and ultramafic rocks. Like talc, it is stable to about 100 km depth, but is distinct from talc in its Al-content. The structure (Fig. 6) is monoclinic C2/m or triclinic, C 1, and consists of trioctahedral talc-like layer, but with one fourth of the tetrahedral sites occupied by Al instead of Si, giving the layer a net negative charge. Instead of an interlayer cation as in micas, there is a trioctahedral brucite-like layer with one third of the octahedra occupied by Al instead of Mg giving the layer a net positive charge. In the brucite-like layer, all of the oxygen atoms are protonated, whereas in the talc-like later one-sixth of the oxygens are protonated. Again, there are no Si-OH bonds in the structure. Chlorite, like serpentine, contains about 13% H2O by weight.

Amphiboles The amphiboles A0-1X7Y8O22(OH)2, are complex hydrous chain silicate minerals of high grade metamorphic and igneous rocks in which A is an alkali cation, X an octahedral divalent or trivalent cation, and Y is tetrahedral Si or Al. The structure (Fig. 7) is based on a double tetrahedral chain parallel to c. Again, there are no Si-OH bonds and all non-silicate oxygens (one in 12) are protonated. Amphiboles are stable in subducting lithosphere to about 3 GPa (Kawamoto et al. 1996; Schmidt and Poli 1998), so they are not expected to be major hosts for H in the sub-lithospheric mantle. Amphibole-like double chain defects are relatively common in pyroxenes at low pressures and so may be a water-carrying defect in mantle pyroxenes.

Lawsonite Lawsonite, CaAl2Si2O7(OH)2·H2O, contains molecular water as well as hydroxyl. The structure (Fig. 8) is orthorhombic, Ccmm, and is a sorosilicate with Si2O7 groups, Al in octahedral coordination, and Ca in 8-coordination. Again, there are no Si-OH bonds, and the two of the non-silicate oxygen atoms coordinating Al are hydroxyls, and two of the oxygens coordinating Ca are water molecules. Lawsonite is a common hydrous alteration product of mafic igneous rocks, replacing calcic plagioclase feldspar. The total water content of lawsonite

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

93

Figure 6. The structure of chlorite, Mg3AlSi3O10(OH)2 Mg2Al(OH)6 is monoclinic, C2/m, or triclinic, C1. There are talc-like layers interspersed with brucitelike layers. One third of the Si atoms are replaced by Al giving the talc layer a net negative charge, and one third of the Mg atoms in the brucite-like layer are replaced by Al to give that layer a net positive charge. There are no Si-O-H bonds and all non silicate oxygens are protonated.

Figure 7. The structure of the amphibole tremolite, Ca2Mg5Si8O22(OH)2, is monoclinic C2/m. The only non-silicate oxygen is protonated (black).

Figure 8. The structure of lawsonite, CaAl2Si2O7(OH)2·H2O is orthorhombic, Ccmm. Lawsonite is a sorosilicate containing Si2O7 groups. The structure contains molecular water as well as hydroxyl. The Ca atom (gray sphere) is 8-coordinated, whereas the Al is octahedral, and the Si tetrahedral.

is high at about 11.5% by weight, and it is stable to relatively high pressures (~10 GPa) and low temperatures (Pawley 1994). Despite its high water content, it is about 10% denser than anorthite, and relatively incompressible with an isothermal bulk modulus of 122 GPa (Boffa-Balaran and Angel 2003). Being stable to depths of 300 km in the crustal portion of a subducting slab, lawsonite may act as a major conduit for water in the crustal portion of the slab to depths approaching those of the transition zone.

Epidote The epidote group comprises epidote (Ca2(Al,Fe)3Si3O12(OH)), zoisite, and clinozoisite (Ca2Al3Si3O12(OH)). Epidote is a very common metamorphic alteration product of mafic igneous rocks, whereas zoisite and clinozoisite are more restricted in composition and occurrence to aluminous and peraluminous rocks. The pressure stability ranges from less than 0.1 GPa to near 7 GPa (Poli and Schmidt 2004). There are also Mn-rich varieties (piemontite), and rare-earth-rich (allanite) varieties as well as several more named chemical variants (Franz and Liebscher 2004). The structure (Fig. 9) is monoclinic, P21/m (Z = 2), and has both isolated SiO4 tetrahedra as well as Si2O7 groups, so it is classed as a sorosilicate. Zoisite is orthorhombic, Pnma, with a nearly identical structure, but twice the unit cell volume (Z = 4). Most of the iron

94

Smyth

Figure 9. The structure of epidote, CaAl2FeSi3O12(OH), and clinozoisite, CaAl3Si3O12(OH), is monoclinic P21/m. The only non-silicate oxygen has a proton (black).

is ferric, and epidote has all of its trivalent cations in octahedral coordination, so it is also denser than anorthite. The O10 position is a non-silicate oxygen and is protonated. There is another non-silicate oxygen, O4, which is bonded to three trivalent metal octahedra. This oxygen is not protonated directly but shares a longer hydrogen bond to the proton on O10.

Humite The humite group comprises norbergite (Mg3SiO4(F,OH)2), chondrodite (Mg5Si2O8 (F,OH)2), humite (Mg7Si3O12(F,OH)2), and clinohumite (Mg9Si4O16(F,OH)2). Humites are relatively rare components of hydrothermally altered ultramafic rocks. They also occur in other silica-undersaturated rocks such as skarns and carbonatites. Natural humites almost always contain more F than hydroxyl. All of the humites lie on the join forsterite-brucite (Fig. 1). The formulas can be thought of as n·(Mg2SiO4)·Mg(OH)2 where n is one for norbergite, two for chondrodite, three for humite, and four for clinohumite, so that all have a higher (Mg+Fe)/Si ratio than does olivine. Because of this they are not thought to be major hydrous components of the mantle, which is generally considered to have a lower (Mg+Fe)/Si ratio than olivine. The F-free pure Mg clinohumite and chondrodite are stable to pressures greater than 14 GPa and temperatures greater than 1250 °C, but are not known to coexist with enstatite. The structures of chondrodite and clinohumite are illustrated in Figures 10 and 11. Although humite and norbergite have not been reported from pressure higher than about 3 GPa, hydroxy-chondrodite and hydroxy-clinohumite are stable at pressures and temperatures well into the transition zone (Yamamoto and Akimoto 1977; Burnley and Navrotsky 1996; Wunder 1998).

Clinohumite Clinohumite (Mg9Si4O16(OH)2) can coexist with chondrodite or with olivine at high pressure but not with phase A or enstatite (Fig. 1). The structure (Fig. 10) is monoclinic, P21/b (a-unique). The odd setting of the space group is chosen to preserve the olivine axial relation (Table 1). The c-axis is greater than that of chondrodite by approximately 6 Å, and the α-angle is reduced to about 100°. Hydroxy-clinohumite has the problem of protonating two identical oxygens symmetrically disposed about the inversion (Friedrich et al. 2001). But again, there are no Si-OH bonds and all non-silicate oxygens are protonated in pure hydroxy-clinohumite. Berry and James (2001) report a second partially occupied deuteron position in pure hydroxyl clinohumite located on the hydroxyl oxygen approximately 180° away from the position near the inversion on the O9-O9 edge.

Chondrite Chondrodite (Mg5Si2O8(OH)2) can coexist with phase A or with hydroxy-clinohumite at pressures to about 14 GPa. Its structure (Fig 11) resembles olivine with a and b axes nearly the same as olivine, but c different and the space group is monoclinic, P21/b. The O5 is the only non-silicate oxygen and is protonated. As with clinohumite, there is a problem with protonation

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

Figure 10. The structure of clinohumite, Mg9Si4O16(OH)2, is monoclinic P21/b. The odd setting of the space group is chosen to maintain the structural relation to olivine.

95

Figure 11. The structure of chondrodite, Mg5Si2O8(OH)2, is monoclinic P21/b. The odd setting of the space group is chosen to maintain the structural relation to olivine.

of every O5 in that this position is close to the inversion center and putting the proton on the O5-O5 edge would put the protons too close to each other. For synthetic deuterated chondrodite (Mg5Si2O8(OD)2), Lager et al. (2001) identified a second partially occupied deuteron position located approximately 180° away from the primary deuteron position on O5. Again, there are no Si-OH bonds and all non-silicate oxygens are protonated in pure hydroxy-chondrodite.

Phase A Phase A (Mg7Si2O8(OH)6) is stable under very hydrous conditions at pressures of 3 to about 8 GPa and temperatures of 550 to about 1250 °C (Yamamoto and Akimoto 1977). The structure (Fig. 12) is hexagonal, P63, and consists of slightly distorted closepacked layers of oxygen atoms and hydroxyl groups repeating along the caxis in an ABCB sequence (Horiuchi et al. 1979). This contrasts with the hexagonal close-packed sequence of Figure 12. The structure of Phase A, Mg7Si2O8(OH)6, ABAB in olivine and the humites. is acentric hexagonal, P63. Mg occupies one special position on the 3-fold axis (M3) and two general positions, M1 and M2. Si occupies two special positions, one each on the 3-fold and on the 63 axes, so that there is one in each layer of cations. All tetrahedra point in the same direction along c, so that the structure is acentric. The O2 and O4 oxygen sites are hydroxyls (Kagi et al. 2000), so all non-silicate oxygens are protonated and there are no Si-OH bonds in the structure. The density is relatively low (2.95 g/cm3) consistent with its high water content (~12% by weight) and limited pressure stability range. Phase A is a possible phase in the mantle as a breakdown product of serpentine, and may coexist with brucite or chondrodite, but probably not with olivine (Luth 1995).

96

Smyth

Phase B Phase B (Mg12Si4O19(OH)2) along with superhydrous phase B (SHyB) and anhydrous phase B (AHyB), contains Si in both octahedral and tetrahedral coordination and has a Mg/Si ration greater than two (Finger et al. 1989). Phase B is stable under pressure and temperature conditions of the Transition Zone. The density is greater that that of forsterite, but less than that of wadsleyite or ringwoodite, despite the presence of octahedral silicon. The structure (Fig. 13) is monoclinic, P21/c, and all atoms except M1 and M3 are in general positions. There are four Si sites, three of which are tetrahedral and one octahedral. There are 13 distinct Mg octahedral sites and 21 distinct oxygen sites of which two are hydroxyls. All non-silicate oxygens are protonated and there are no Si-OH bonds in the structure.

Superhydrous Phase B Superhydrous Phase B (Mg10Si3O14(OH)4) (SHyB) is similar to phase B in having both octahedral and tetrahedral silicon, a stability range within the transition zone, and a Mg/Si ratio greater than two. The density is slightly less than that of phase B. The structure (Fig. 14) is orthorhombic, Pnnm (Pacalo and Parise 1992), and half of the Si atoms are octahedral and half tetrahedral. There are four distinct Mg octahedra and six oxygen sites. Although all non-silicate oxygens are protonated and there are no Si-OH bonds in the structure, one of the silicate oxygens is under-bonded (O3) and one is over bonded (O6) which leads to the distortions of the coordination polyhedra. As with the other B-phases, its Mg/Si ratio is greater than two, so it is not a likely phase in an enstatite or majorite bearing mantle assemblage. Koch-Müller et al. (2005) report polymorphic inversion in superhydrous phase B with an ordered low temperature polymorph having symmetry Pnn2. The reduction in symmetry with ordering causes splitting of the Mg positions, but not the Si positions.

Phase D Phase D (MgSi2O4(OH)2) is stable into the lower mantle at pressures of 17 to 50 GPa (Frost and Fei 1999) and has both Mg and Si in octahedral coordination (Fig. 15). The structure is

Figure 13. The structure of Phase B, Mg12Si4O19(OH)2, is monoclinic, P21/c.

Figure 14. The structure of superhydrous Phase B (ShyB), Mg10Si3O14(OH)4, is orthorhombic, Pnnm.

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

97

highly disordered with variable Mg/Si ratios and water contents ranging from 10 to 18% by weight (Yang et al. 1997). The density of the ideal trigonal structure (P 31m) is about 3.01 g/cm3. All oxygens are equivalent and bonded to two Si, one Mg, and one proton, although only about one third of the proton positions can be occupied. Although the density is only about 75% that of the lower-mantle anhydrous assemblage, phase D is the likely host phase for H in the lower mantle.

Phase E

Figure 15. The structure of Phase D, MgSi2O4(OH)2, is trigonal, P 31m.

Phase E (Mg2SiO2(OH)4) is a highly disordered structure with Si in tetrahedral and Mg in octahedral coordination. The structure (Fig. 16) is trigonal R 3 m with variable Mg/Si ratio and H content (Kudoh et al. 1993). In the structure the M2 site occurs in an octahedral site adjacent to the Si tetrahedron, so that either one or the other can be occupied but not both. There is no long range order and charge balance is made up by protonation. The structure occurs in very hydrous compositions as a breakdown product of serpentine at pressures of 13 to 17 GPa and temperatures of 800 to 1300 °C (Kanzaki 1991).

Phase Pi Phase Pi (Al3Si2O7(OH)3) is so called because was formerly thought to be the poorly described synthetic mineral piezotite (Coes 1962). The mineral has been synthesized at low temperatures and moderate pressures (500-650 °C and 4-5.5 GPa) (Wunder et al. 1993) in the hydrous aluminosilicate system. The structure (Fig. 17), is acentric triclinic, P1, with Al in octahedral and Si in tetrahedral coordination (Daniels and Wunder 1993, 1996). There are 20 distinct oxygen atoms in the unit cell, of which six should be hydroxyls if the formula is correct. Four of the oxygens (O9, O10, O19, O20) are bonded to just two Al atoms and so

Figure 16. The structure of Phase E, Mg2SiO2(OH)4 is trigonal R 3 m. The structure is highly disordered. The silicate layer (dark) can have tetrahedral voids occupied by Si or octahedral voids occupied by Mg.

Figure 17. The structure of phase Pi (Al3Si2O7(OH)3) is acentric, triclinic P1. Although the proton positions have not been determined for this phase, the oxygen atoms shown as black spheres are protonated nonsilicate oxygens. The remaining two underbonded oxygens shown as white spheres are likely protonated silicate oxygen atoms.

98

Smyth

are certainly hydroxyls. The remaining oxygens all bond to Si. Of these, O4 and O14 bond to one Si and one Al, and so are also underbonded. They each have very long Al-O distances so are apparently hydroxyls, but unusual in that they may be protonated silicate oxygens.

Topaz-OH Topaz-OH (Al2SiO4(OH)2) also occurs in the hydrous aluminosilicate system at temperatures of 600-1000 °C and pressures up to about 12 GPa (Pawley 1994; Schmidt et al. 1998; Wunder et al. 1999). The structure (Fig. 18) is orthorhombic, Pbnm, with Al in octahedral and Si in tetrahedral coordination. The structure is relatively dense (3.37 g/cm3), more dense than phase Pi, but less dense than phase Egg or kyanite. Curiously, it is significantly less dense than fluoro-topaz. This may be because the protons are disordered over two distinct positions (Northrup et al. 1994).

Figure 18. The structure of topaz-OH, Al2SiO4(OH)2 is orthorhombic Pbnm. Despite its stability to quite high pressures (~12 GPa) it is significantly less dense at zero pressure than fluorotopaz.

Phase Egg Phase Egg AlSiO3(OH) is named after the first author to describe the phase (Eggleton 1978) and has a 1:1 Al:Si ratio. It occurs at pressure ranges into the transition zone at 11-18 GPa and temperatures of 700-1300 °C as a high pressure breakdown product of hydroxyl-topaz. The structure was solved and proton positions located to high precision by neutron powder diffraction (Schmidt et al. 1998). The structure is monoclinic, P21/n, and has both Si and Al in octahedral coordination (Fig. 19). There are four distinct oxygens in the structure. The O1 and O2 oxygens are bonded to two Si and one Al positions, whereas O4 is the hydroxyl, but it is also bonded to two Al and one Si, as is O3. The long H bond extends to O3. With octahedral silica and a single hydroxyl, the structure is relatively incompressible with a bulk modulus of 157 GPa and K′ of 6.5 (Vanpeteghem et al. 2003).

Figure 19. The structure of Phase Egg, AlSiO3(OH), is monoclinic P21/n. It has both Al (light) and Si (dark) in octahedral coordination.

K-cymrite K-cymrite (KAlSi3O8·H2O) occurs as a hydration product of sanidine at pressures above 3GPa and temperatures of 350-750 °C. It is isostructural with cymrite (BaAl2Si2O8·H2O) and has a layered structure of a double tetrahedral sheet (Fig. 20) with molecular water within the layer and K between the sheets (Fasshauer et al. 1997). The symmetry is P6/mmm so that the tetrahedral Al and Si are disordered over the sheet. There are bridging

Figure 20. The structure of K-cymrite (KAlSi3O8·H2O) is hexagonal P63/mmm and is composed of a double hexagonal layer of disordered Al-Si tetrahedra. K atoms (black) form the interlayer, and molecular water (gray) is in the tetrahedral layer.

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

99

oxygens, non-bridging silicate oxygens, as well as molecular water, which lies within the tetrahedral layer (Fig. 20). The proton positions have not been determined, but are likely to be locally determined by the Al occupancy of the nearest tetrahedra. None of the silicate oxygens are protonated. K-cymrite is slightly denser than sanidine (Table 1).

NOMINALLY ANHYDROUS HIGH-PRESSURE SILICATE AND OXIDE PHASES Periclase-wüstite Periclase-wüstite ((Mg,Fe)O) is isometric, Fm 3 m, with the rock-salt structure (Fig. 21). Pure MgO is stable at low pressures and not known to undergo any high pressure phase transformations, whereas wüstite (FeO) is known to undergo a rhombohedral distortion of this structure at pressures above 20 GPa (Shu et al. 1998; Jacobsen et al. 2005). At low to modest pressures the structure can accommodate significant ferric iron in tetrahedral voids associated with octahedral vacancies. The oxygen site potentials of the nominally anhydrous mantle phases are given in Table 3. Periclase and wüstite have some of the shallowest oxygen potentials of any mantle minerals, which make these phases likely hosts for H if charge balance can be achieved. Murakami et al. (2002) report up to 2000 ppmw H2O in (Mg,Fe)O ferro-periclase synthesized from a hydrous peridotite composition at 25.5 GPa and 1650 °C. However their FTIR spectra show pleochroism unexpected for a cubic phase raising the possibility of an included hydrous phase of power symmetry. Bolfan Casanova et al. (2000) report only about 2 ppmw H2O in periclase at 24 GPa and 1500 °C in a pure MgO-SiO2-H2O system. Bolfan-Casanova et al. (2003) also report very low H contents in ferro-periclase up to 10 Gpa, so it appears that H2O solubility in pure MgO and in ferro-periclase of possible lower mantle composition is quite limited.

Figure 21. The structure of periclase (MgO) and wüstite (FeO) is the cubic rock salt structure, Fm 3 m.

Corundum Corundum (Al2O3) (Fig. 22) is rhombohedral, R 3 c, and isostructural with hematite (Fe2O3), eskolaite (Cr2O3), karelianite (V2O3), and synthetic Ti2O3. Ilmenite (FeTiO3) and akimotoite (MgSiO3) are also closely related structures with subgroup symmetry, R3. Natural corundum has not been reported with appreciable H contents, but it is not a common mineral in high pressure assemblages. It occurs in high grade peraluminous rocks with zoisite or in peraluminous eclogites. Rossman and Smyth (1990) report no observable OH stretch features in the FTIR spectrum of a natural corundum from a

Figure 22 The structure of corundum (Al2O3) and hematite (Fe2O3) is trigonal, R 3 c.

Mg2SiO4

Wadsleyite II

O O

Mg2SiO4

Wadsleyite

Mg3Al2Si3O12 Fe3Al2Si3O12

Fe2SiO4

Fayalite

Garnets Pyrope Almandine

Mg2SiO4

Mg2SiO4 Fe2SiO4 Mg14Si5O24

SiO2

Stishovite Olivine Forsterite

Ringwoodite γ-Fe2SiO4 Anhyd. Phase B

O1 O2 O3 O1 O2 O3 O1 O2 O3 O4 O1 O2 O3 O4 O5 O6 O7 O8 O O O1 O2 O3 O4 O5 O6 O7 O8 O9

MgO FeO Al2O3 TiO2 SiO2 SiO2

Periclase Wüstite Corundum Rutile Quartz Coesite

Site O O O O O O1 O2 O3 O4 O5 O

Formula

Mineral

4Mg,1Al,1Si 4Fe,1Al,1Si

3Mg,1Si 3Mg,1Si 3Mg,1Si 3Fe,1Si 3Fe,1Si 3Fe,1Si 5Mg 1Mg,2Si 3Mg,1Si 3Mg,1Si 3Mg,1Si 5Mg 3Mg,1Si 3Mg,1Si 3Mg,1Si 3Mg,1Si 1Mg,2Si 3Mg,1Si 3Mg,1Si 3Fe,1Si 3Mg,1SiIV 4Mg,1SiVI 3Mg,1SiIV 6Mg 3Mg,1SiIV 3Mg,1SiIV 3Mg,1SiIV 4Mg,1SiVI 3Mg,1SiIV

6Mg 6Fe 4Al 3Ti 2Si 2Si 2Si 2Si 2Si 2Si 3Si

Coordination

27.06 27.08

27.69 27.53 26.35 27.38 27.20 26.12 21.28 30.94 26.78 26.97 26.83 20.03 26.41 26.68 26.71 27.34 30.51 27.50 26.57 26.28 27.09 25.62 26.32 22.70 28.77 27.84 26.66 25.09 26.91

23.92 23.36 26.40 25.94 30.82 29.07 29.64 30.67 30.41 30.97 28.61

Potential (V)

MgSiO3 MgSiO3 ZrSiO4 CaTiSiO5

Perovskite Post-perovskite* Zircon Titanite

Mg2Si2O6 Al2SiO5

CaMgSi2O6

Diopside Akimotoite Kyanite

NaAlSi2O6

Mg2Si2O6

MgAlAlSiO6

Mg2Si2O6

Formula

Jadeite

Clinoenstatite

Mg-Tschermaks

Pyroxenes Orthoenstatite

Mineral O1a O2a O3a O1b O2b O3b O1a O2a O3a O1b O2b O3b O1a O2a O3a O1b O2b O3b O1 O2 O3 O1 O2 O3 O O1 O2 O3 O4 O5 O6 O7 O8 O9 O10 O1 O2 O1 O2 O O1 O2a O2b O3a O3b

Site 3Mg,1Si 2Mg,1Si 1Mg,2Si 3Mg,1Si 2Mg,1Si 1Mg,2Si 1Mg,2Al,1Si 1Mg,1Al,1Si 1Mg,2Si 3Al,1Mg 2Al,1Mg 2Al 3Mg,1Si 2Mg,1Si 1Mg,2Si 3Mg,1Si 2Mg,1Si 1Mg,2Si 1Na,2Al,1Si 1Na,1Al,1Si 2Na,2Si 1Ca,2Mg,1Si 1Ca,1Mg,1Si 2Ca,2Si 2Mg, 2SiVI 2Al,1Si 4Al 2Al,1Si 2Al,1Si 2Al,1Si 4Al 2Al,1Si 2Al,1Si 2Al,1Si 2Al,1Si 2Mg,2SiVI 3Mg,2SiVI 2Mg,2SiVI 3Mg,2SiVI 2Zr,1Si 1Ca,2Ti 1Ca,1Ti,1Si 1Ca,1Ti,1Si 2Ca,1Ti,1Si 2Ca,1Ti,1Si

Coordination

Table 3. Cation coordinations and electrostatic potentials of oxygen sites in high pressure silicate and oxide phases.

26.22 26.38 30.89 26.39 26.48 30.57 32.09 30.25 32.49 24.59 22.39 21.99 26.33 26.34 30.90 26.39 26.34 30.59 27.54 27.15 30.34 25.53 25.78 30.74 27.38 28.60 25.91 27.73 28.01 28.60 25.81 27.70 27.97 28.28 28.36 26.91 26.86 27.76 26.77 31.49 24.89 26.86 26.96 26.98 26.87

Potential (V)

100 Smyth

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

101

high pressure corundum-kyanite eclogite. There is a single oxygen site in the structure. The site potential is significantly deeper than that of periclase but might allow minor protonation if charge balance can be achieved. However, significant protonation of the isostructural akimotoite (MgSiO3) does occur as discussed below.

Coesite Coesite (SiO2) is the high pressure polymorph of SiO2 stable between about 3 and 8 GPa. The structure (Fig. 23) is a relatively dense tetrahedral framework with monoclinic C2/c symmetry. Natural coesite is normally quite pure SiO2 with only trace levels of other elements. All oxygens are bridging oxygens bonded only to two Si atoms. There are five distinct oxygen sites in the structure, all with deep potentials similar to quartz (Table 3). Of these O1 has the shallowest potential and the most likely one to be protonated if there were a small amount of B or Al substitution in the tetrahedra. Rossman and Smyth (1990) report no observable OH in a natural coesite in a relatively hydrous coesite-kyanite eclogite. Koch-Mueller et al. (2001) and Mosenfelder (2000) however report up to 200 ppmw H2O in coesite synthesized at pressures of 7.5 GPa and 1100 °C, but undetectable amounts in coesite synthesized at pressures below 5 GPa. Koch-Müller et al. (2003) report that the major substitution mechanism in coesite is by the hydrogarnettype (H4O4) with relatively minor amounts of H being associated with B and Al substitution. In a low-symmetry tetrahedral framework structure such as coesite, any Si vacancy would result in protonation of the terminating oxygens, but there would be nothing to constrain these oxygens to maintain a tetrahedral configuration, as there is in garnet. Koch-Müller et al. (2001) propose several possible proton locations for coesite on the oxygens coordinating a vacant Si2 position consistent with O-H dipoles observed in polarized infrared spectra. They further suggest that vacancy at Si1 is unlikely because of difficulty in accounting for the pleochroism of one of the major O-H vibrations.

Stishovite and rutile

Figure 23. The structure of coesite, SiO2, is monoclinic C2/c. All oxygens are bridging oxygens bonded to two tetrahedral Si atoms. Trace hydration of this structure has only been observed in samples quenched from pressures above 5 GPa.

Figure 24. The structure of stishovite (SiO2) and rutile (TiO2) is tetragonal, P42/mnm. All oxygens are equivalent. Protonation of these compounds can accompany trivalent ion substitution in the octahedra. Proton positions determined by neutron single crystal diffraction for rutile (Swope et al. 1995) are illustrated.

Stishovite (SiO2) and rutile (TiO2) (Fig. 24) are isostructural and both may incorporate considerably more H than either coesite or quartz. The stishovite structure is tetragonal P42/mnm with all cations in octahedral coordination, and the octahedra share edges in the c-direction. All oxygens in the structure are equivalent, and protonation of the oxygens can accompany Al for Si substitution in the octahedra (Smyth et al.

102

Smyth

1995). The oxygen site potential is substantially lower than those of quartz or coesite (Table 3). Bolfan-Casanova et al. (2000) report up to 72 ppmw H2O in stishovite in an Al-free system. Vlassopoulos et al. (1993) report up to 8000 ppmw H2O in natural rutile containing minor amounts of trivalent cations (Cr, Fe, V, Al). Principal rutile absorptions in the OH range are at 3290 and 3365 cm−1 (Rossman and Smyth 1990; Vlassopoulos et al. 1993) and are strongly polarized normal to the c-axis. Swope et al. (1995) report a proton position on the shared octahedral edge for hydrous rutile at x/a = 0.4176; y/b = .5033, and z/c = 0, based on neutron single crystal diffraction of a natural sample. This position is consistent with the strong IR pleochroism and is illustrated in Figure 24.

Pyroxenes Pyroxenes of major importance to mantle dynamics include enstatite (Mg2Si2O6), diopside (CaMgSi2O6), and jadeite (NaAlSi2O6), which are all significant components of the upper mantle. For a recent review of pyroxene structures at temperature and pressure see Yang and Prewitt (2000). Enstatite is an orthopyroxene, orthorhombic, Pbca (Fig. 25), at pressures to about 7 GPa, whereas enstatite quenched from higher pressures is monoclinic P21/c. Clinoenstatite transforms to majorite garnet at about 15 GPa, in a pyrolite composition and gradually dissolves into the garnet phase through the upper Transition Zone. Mantle peridotites and lherzolites contain up to about 15 modal percent clinopyroxene that is typically a Cr-diopside with very minor amounts of Na, Al or Fe3+. In eclogites, however, diopside and jadeite form a complete crystalline solution known as omphacite, which is monoclinic C2/c at high temperatures. Omphacite composes 50% or more of eclogites that form from subducting basalt at pressures of 3 to 13 GPa. Eclogites are quite distinct from peridotites and lherzolites, so that rocks of intermediate composition are virtually unknown among rocks of high pressure origin.

Figure 25. The structure of ortho-

enstatite, Mg2Si2O6, is orthorhombic Orthoenstatite can be a major host for water in the Pbca. This view down c with ashallow (lithospheric) upper mantle. Rauch and Keppler vertical, shows the alternating layers (2002) report that the solubility of H2O in enstatite of T1 and T2 tetrahedra. The likely increases to a maximum of about 850 ppmw at 1100 °C sites of protonation are on the O2b at 7.5 GPa and decreases slightly at higher pressures in and O1b oxygens (spheres) with the O-H vectors lying in the b-c plane. the clinoenstatite field. In pure Mg enstatite, the strongest OH absorptions in the infrared spectra are polarized parallel to c. However, Al has a dramatic effect on the water solubility and on the FTIR spectra of orthoenstatite, especially at pressures of 1 to 2 GPa at which Al substitution in the tetrahedral site can be extensive (Mierdel et al. 2006). In aluminous enstatite, H2O solubilities can approach 9000 ppmw at 900 °C and 1.5 GPa. In these enstatites, the O-H polarizations are strongest perpendicular to c (Mierdel et al. 2006).

Orthoenstatite is orthorhombic, Pbca, with two distinct tetrahedral sites, T1 and T2, arranged is separate layers of tetrahedral chains (Fig. 25). Al enters the structure as a coupled substitution where the Al is in both an M1 octahedron and one of the tetrahedral sites. Tetrahedral Al is known to strongly order in the structure with a very strong preference for T2 (Takeda 1973). There are six distinct oxygen sites in the structure, O1a, O1b, O2a, O2b, O3a,

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

103

and O3b, with the ‘a’ oxygens in the T1 chains and the ‘b’ oxygens in the T2 chains. The O3 atoms are the bridging oxygens in the chains. Electrostatic site potentials for the oxygens for pure Mg orthoenstatite are given in Table 3, and the O2b has the shallowest potential and is therefore the most likely site for protonation. Structure refinement of a hydrous, aluminous orthopyroxene shows up to 5% cation vacancy at M2 with nearly equal amounts of Al substitution in both M1 and T2 sites, based on chemical analysis and volumes of coordination polyhedra (Smyth et al. 2006b). Also reported in Table 3 is an oxygen site potential calculation for a hypothetical fully “Mg-Tschermaks” orthoenstatite of composition MgAlAlSiO6, fully ordered with all tetrahedral Al in T2. In this structure both O2b and O1b are substantially underbonded and likely sites for protonation. The O3b oxygen is also underbonded, but Al-avoidance would not allow Al in T2 to exceed 50%, so O3b is not as likely to protonate as O2b or O1b. It appears then that the major hydrous components are MgAlAlSiO6 and H2AlAlSiO6 (“hydro-Tschermaks”), with a cation vacancy at M2 and protons on the O1b-O2b edges of the vacant M2 polyhedron, consistent with the observed O-H polarization in the a-b plane. This substitution mechanism achieves a net volume reduction of the unit cell, and nearly 1% H2O by weight (Mierdel et al. 2006), but because it requires tetrahedral Al, H solubility decreases sharply with increasing pressure. The O2b and O1b oxygen sites are indicated by spheres in Figure 25. This “hydro-Tschermaks” substitution appears to be strongly abetted by the ordering of tetrahedral Al in T2 which can only happen in the Pbca structure. At pressures near the 410 km discontinuity, enstatite is monoclinic, P21/c, after quenching to low temperature, but C2/c at relevant mantle temperatures. The solubility of H is much less than that in aluminous orthopyroxene at lower crustal pressure, so that clinoenstatite in equilibrium with forsterite containing >8000 ppmw H2O contains less than 1000 ppmw (Smyth et al. 2006a) and somewhat less (~650 ppmw) in equilibrium with wadsleyite (Bolfan-Casanova et al. 2000). The principal substitution mechanism appears to be divalent cation vacancies, principally at M2. Natural omphacites can contain up to about 3000 ppmw H2O (Katayama and Nakashima 2003; Smyth et al. 1991). Bromiley and Keppler (2004) experimentally investigated water solubility in jadeite and found a maximum H2O content of about 450 ppmw at 2 GPa, but dramatically higher solubilities in more complex solid solutions. Natural omphacites are very complex chemically containing 10% or more of up to eight chemical end members (Smyth 1980), but crystallographically relatively simple, having space group C2/c at mantle conditions of temperature and pressure. The hydrous component referred to as Ca-Eskola pyroxene Ca0.50.5AlSi2O6, may be better described as HAlSi2O6. Crystal structure refinements of natural H-rich omphacites indicate significant M2 site vacancy (Smyth 1980). Textural evidence of kyanite and garnet exsolution from omphacite suggests that H2O solubility in these pyroxenes may approach 1% by weight (Smyth et al. 1991). Bromiley et al. (2004) have experimentally hydrated natural Cr-diopside crystals at 1100 °C and pressures of 1.5 to 4 GPa. They report up to about 450 ppmw at 1.5GPa and infer proton positions on the O2-O1 and O2-O3 edges of the M2 polyhedron based on polarizations of the O-H vector in the a-b plane, which are similar to those reported for orthopyroxene by Mierdel et al. (2006).

Akimotoite Akimotoite (MgSiO3) is the ilmenite-type polymorph of enstatite stable at pressures of the lower transition zone (18-22 GPa). The structure is trigonal R3 and has alternating layers of Si and Mg octahedra (Fig. 26). Bolfan-Casanova et al. (2000) report up to about 450 ppmw H2O in pure Mg akimotoite at 21 GPa and 1500 °C coexisting with stishovite and melt. BolfanCasanova et al. (2000, 2002) report strongly pleochroic FTIR spectra for the O-H stretching vibration in this phase with strong absorptions at 3390 cm−1 parallel to c and 3320 and 3300 cm−1 perpendicular to c. Based on the polarizations and the relation of frequency to O-H-O

104

Figure 26. The structure of akimotoite (ilmenite-type MgSiO3) is trigonal R3 and closely related to that of corundum.

Smyth

Figure 27. the structure of garnet is cubic Ia 3 d . All oxygen atoms are identical and the tetrahedra and octahedra form a corner-sharing framework structure.

distance (Libowitzky 1999), they deduce two proton positions, both likely associated with Mg vacancies. Inasmuch as the structure is essentially isostructural with corundum, possible Al substitution for octahedral Si might have a significant impact on the H solubility in this phase.

Garnet Garnet (X3Y2Z3O12) (Fig. 27) is isometric, Ia 3 d, with Si (Z) in tetrahedral coordination forming a framework by sharing oxygens with Al (Y) in octahedral coordination. Interstitial to the framework is the dodecahedral divalent cation site, which may be occupied by Mg, Fe, or Ca (X). In this high-symmetry structure, all oxygens are equivalent and in a general position. At pressures of the transition zone, garnet can accept equal amounts of Si and Mg into the octahedral site in place of a trivalent cation. The Mg3(MgSi)2Si3O12 (MgSiO3) end-member is majorite. Majorite quenches to tetragonal, I41/a, by ordering of Mg and Si in the octahedral site, although it is likely disordered Ia 3 d at mantle conditions (Angel et al. 1989). Hydrogen is accommodated in the garnet structure by Si vacancies so that the terminating octahedral oxygens are protonated. The tetrahedral site has 4 point symmetry, so symmetry constrains the oxygens to maintain a tetrahedral configuration, but the distance from the 4 point position to the oxygen increases from about 1.63 Å for the occupied site to about 1.95 Å for the vacant site (Lager and von Dreele 1996). This means that pressure inhibits the substitution so that garnets from high pressure environments generally contain less than 50 ppmw H2O (Bell and Rossman 1992). Lager et al. (1987) and Lager and von Dreele (1996) report deuteron positions for a deuterated hydrogarnet (Ca3Al2D12O12) on the edges of the vacant tetrahedra based on neutron single crystal diffraction.

Olivine Olivine ((Mg,Fe)2SiO4) is generally believed to be the most abundant phase in the upper mantle from the Moho to 410 km discontinuity. Natural olivines as reviewed in the current volume (Beran and Libowitzky 2006) contain up to about 400 ppm by weight (ppmw) H2O, but typically less than 100 ppmw (Bell et al. 2004). Olivine synthesized at high pressures and quenched can contain much more H. Kohlstedt et al. (1996) report up to 1510 ppmw in olivine equilibrated at 1100 °C and 12 GPa. Recalculating this amount based on Bell et al. (2003) one

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

105

gets about 4000 ppmw (Hirschmann et al. 2005). Mosenfelder et al. (2006) report up to 6400 ppm H2O in olivine quenched from 12 GPa and 1100 °C. Smyth et al. (2006a) report up to 8900 ppmw in olivine synthesized at 1250 °C and 12 GPa in equilibrium with either enstatite or clinohumite, but decreasing at higher temperatures with the onset of melting. Water contents approaching one per cent by weight would make olivine a major host for water in the upper mantle. The olivine structure (Fig. 28) is orthorhombic, Pbnm, with two distinct octahedra, M1 and M2, and one silicate tetrahedron. There are three distinct oxygen sites in the structure, with O1 and O2 lying on the mirror, and O3 being in a general position. All oxygens are bonded to three Mg and one Si atom (Table 3) and site potentials range from 26.3 V for O3 to 27.7 V for O1. Smyth et al. (2006a) report that the major H substitution mechanism in olivine is protonation of the O1-O2 edges of vacant M1 octahedra. The proton position suggested by Smyth et al. (2006a) at x/a = 0.95; y/b = 0.04; z/c = 0.25 is illustrated in Figure 28. They further report a volume of hydration at ambient conditions: V = 290.107 + 5.5×10−5 *cH2O Å3 where V is cell volume in Å3, and H2O is the ppm by weight H2O as determined from the calibration of Bell et al. (2003).

Wadsleyite Wadsleyite is the first high pressure polymorph of Mg2SiO4, and the olivine-wadsleyite transition at about 13 GPa is thought be responsible for the 410 km discontinuity. The wadsleyite structure (Fig. 29) is usually orthorhombic, Imma, with three distinct divalent metal octahedra, M1, M2 and M3. The structure is similar to that of spinelloid III in the Nialuminosilicate system (Ma and Sahl 1975). Unlike olivine which is based on a hexagonal close-packed array of oxygens, wadsleyite and the other spinels and spinelloids are based on a cubic close-packed oxygen array. Unlike olivine and ringwoodite, wadsleyite is a sorosilicate

Figure 28. The structure of forsterite, Mg2SiO4, and fayalite Fe2SiO4, is orthorhombic Pbnm. Hydration appears to be compensated by octahedral cation vacancies principally at M1. The proton position inferred from polarized FTIR spectroscopy on the O1-O2 shared edge of the M1 octahedron is illustrated.

Figure 29. The structure of wadsleyite, (Mg,Fe)2SiO4, is orthorhombic Imma. Hydrous wadsleyite may deviate slightly from orthorhombic symmetry as monoclinic, I2/m, due to ordered cation vacancies in M3 in violation of the mirror perpendicular to a. The structure has a non-silicate oxygen which is readily protonated. Charge balance is maintained by Mg vacancies at M3.

106

Smyth

with Si2O7 groups, a bridging oxygen (O2) and a non-silicate oxygen (O1). Smyth (1987) calculated oxygen site potentials and predicted that the under-bonded non-silicate oxygen would be a potential site for protonation. Wadsleyites with up to 3% by weight H2O have been reported (Inoue et al. 1995). The major hydrogen substitution mechanism appears to be protonation of the vacant M3 octahedral edges and ordering of the vacancies so that hydrous wadsleyites with more than about 1% H2O are monoclinic, I2/m (a subgroup of Imma). Beta angles up to 90.4° have been reported (Smyth et al. 1997; Jacobsen et al. 2005). Wadsleyite shows a significant zero-pressure volume expansion that is similar in magnitude to that of olivine. Holl (2006) reports the volume expansion as: V = 538.64 + 9.4 × 10−5 *cH2O Å3 Hydrous wadsleyite shows a strong O-H stretching absorption at about 3325 cm−1 which shows minimal pleochroism. A potential proton location on the O1-O4 edge of a vacant M3 octahedron at about x/a = 0.11; y/b = 0.20; z/c = 0.36 would be consistent with the observed frequency and pleochroism of this polarization and is illustrated in Figure 29. The complexity of the infrared absorption spectrum, however, indicates that there are multiple possible proton locations in the structure (Kohn et al. 2002).

Wadsleyite II Wadsleyite II is isostructural with spinelloid IV (Smyth and Kawamoto 1997; Smyth et al. 2005). It has only been reported from long-duration hydrous peridotite composition runs at 17.5 to 18 GPa, between the wadsleyite and ringwoodite fields. It is a well-ordered phase with a- and c-axes similar to wadsleyite but with a b-axis 2.5 times that of wadsleyite at about 30 Å. The structure is very difficult to distinguish from wadsleyite by powder diffraction or by Raman spectroscopy. The structure (Fig. 30) contains both isolated SiO4 tetrahedra as well as Si2O7 groups in three distinct tetrahedral sites. It also contains six distinct octahedral sites and eight distinct oxygens, of which O2 is a non-silicate oxygen and a potential protonation site. Analogous to wadsleyite, a possible proton location would be near the O2-O4 edge of the M6 octahedron or the O2-O5 edge of the M5 octahedron. Wadsleyite II in the high pressure peridotite system is only known with about 2.8 wt% H2O, whereas spinelloid IV in the Ni aluminosilicate system is thought to be anhydrous (Akaogi et al. 1982; Horioka et al. 1981).

Ringwoodite Ringwoodite is the true spinel polymorph of forsterite and is stable as the dominant phase in a pyrolite composition mantle from about 525 to 670 km depth. The ringwoodite to perovskite plus periclase transition is thought to be responsible for the 670 km discontinuity. The structure (Fig. 31) is cubic, Fd 3 m with octahedral Mg and tetrahedral Si. Kohlstedt et al. (1996) report up to about 2.4 wt% H2O in ringwoodite. The FTIR spectrum shows a

Figure 30. The structure of wadsleyite II, (Mg,Fe)2SiO4, is orthorhombic Imma. This structure, like wadsleyite is a spinelloid, but contains both isolated SiO4 groups as well as Si2O7 groups.

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

107

Figure 31. The structure of ringwoodite is a true spinel and is cubic, Fd 3 m. Si is in tetrahedral (dark) and Mg in octahedral (light) coordination. All oxygens are equivalent and bonded to one Si and three Mg atoms. There are no bridging or nonsilicate oxygens. Hydration is compensated by octahedral site vacancies.

broad absorption feature in the range 2600 to 3600 cm−1 (Smyth et al. 2003; Keppler and Smyth 2005). Although there is no IR pleochroism in the cubic system, the OH does appear to be structural because OH concentration computed from the FTIR spectrum correlates with a zero-pressure unit cell volume increase (Smyth et al. 2003) that is similar in magnitude to those observed for forsterite and wadsleyite cited above. Peaks in the spectra correlate with protonation of both the octahedral and tetrahedral edges (Libowitzky 1999) and crystal structure refinements indicate both octahedral and tetrahedral vacancies (Kudoh et al. 2000; Smyth et al. 2003).

Anhydrous phase B Anhydrous phase B (Mg14Si5O24) lies on the anhydrous edge of the DHMS ternary between forsterite and periclase. As with the other B-phases, anhydrous phase B has Mg/Si ratio greater than two, and so is not expected to coexist with either enstatite or majorite. It is therefore not expected to be a significant phase in the transition zone. The structure (Fig. 32) is orthorhombic, Pmcb (Hazen et al. 1992) and has Si in both octahedral and tetrahedral coordination. Little is known about its trace H content, but its oxygen sites are all electrostatically balanced according to Pauling bond strength sums, bonded to either three octahedral Mg and a tetrahedral Si, six Mg, or four Mg and one octahedral Si. Of these, the O4 is the non-silicate oxygen, has the lowest electrostatic potential and is thus a potential protonation site (Table 3). The density (3.39 g/cm3) lies between that of forsterite and periclase, but less than either wadsleyite or ringwoodite, despite its octahedral silicon.

Kyanite Kyanite (Al2SiO5) is triclinic P1, with Al in octahedral and Si in tetrahedral coordination. There are ten distinct oxygen sites in the structure (Fig. 33) most of which are bonded to two octahedral Al and one tetrahedral Si. The O2 and O6 positions are non-silicate oxygens and

Figure 32. The structure of anhydrous Phase B (AnHB), Mg14Si5O24, is orthorhombic Pmcb, and had Si in both octahedral and tetrahedral coordination.

108

Smyth

bonded to only four Al atoms (Table 3). These are potential hydration sites if charge balance can be achieved by divalent cation substitution for Al. Although Beran and Goetzinger (1987) and Rossman and Smyth (1990) report relatively large amounts of OH in kyanite up to about 4000 ppmw H2O, Bell et al. (2004) report a new calibration for kyanite, greatly reducing this amount and reporting a maximum H2O content for kyanite of about 230 ppmw.

Perovskite Perovskite-type (Mg,Fe)SiO3 is believed to be the major phase in the lower mantle, so small amounts of H in this phase can have a large effect on the total water budget of the planet. The strucFigure 33. The structure of kyanite, Al2SiO5, ture (Fig. 34) is orthorhombic, Pbnm, with Mg in is triclinic P1. eight coordination, Si in octahedral coordination, and two distinct oxygen sites. Both oxygen sites have relatively deep electrostatic potentials near 27 V (Table 3). The structure is dense (4.1 g/cm3). Meade et al. (1994) report only minor amounts of H in MgSiO3 perovskite. Bolfan-Casanova et al. (2000) report no detectable H by FTIR spectroscopy in pure MgSiO3 perovskite in equilibrium with hydrous akimotoite in an Al-free composition, however Higo et al. (2001) report up to 500 ppmw H2O by SIMS analysis of similar samples. Murakami et al. (2002) report up to 2000 ppmw H2O in (Mg,Fe)SiO3 perovskite synthesized at 25.5 GPa and 1600 °C in an Al-bearing peridotite composition. Litasov et al. (2003) observed only Figure 34. The structure of perovskite-type about 100 ppm in pure MgSiO3 perovskite, but MgSiO3 is orthorhombic Pbnm. 1400 to 1800 ppmw H2O in Al and Fe bearing perovskites in a hydrous peridotite system. None of the FTIR spectra of silicate perovskites in pure MgSiO3 or MgSiO3-Al2O3 systems show sharp absorption bands so there has been some disagreement as to whether these features represent structurally bound hydroxyl (Bolfan-Casanova et al. 2003; Litasov et al. 2003). Perovskite samples synthesized in chemically complex systems show a consistent but broad OH absorption feature at about 3397 cm−1, but variable other features. It appears that while H2O solubility in pure MgSiO3 perovskite is likely negligible, perovskite crystallized from more chemically complex systems may incorporate significant amounts of water, but in reports of higher water contents, the possibility of hydrous inclusions within the perovskite cannot be ruled out. Perovskite-type CaSiO3 is believed to be a minor phase in the lower mantle. Although it is isostructural with MgSiO3 perovskite (orthorhombic, Pbnm), is appears to form a separate phase in lower mantle synthesis experiments. Murakami et al. (2002) report up to 4000 ppmw H2O in CaSiO3 perovskite synthesized at 25.5 GPa and 1600 °C. This phase does not appear to be quenchable so interpretation of FTIR spectra on quenched material is difficult.

Post-perovskite Post-perovskite (MgSiO3) is a new structure type reported for MgSiO3 at pressures of the lower-most lower mantle near the core-mantle boundary (Murakami et al. 2004). It is

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

109

Figure 35. The structure of post-perovskite-type MgSiO3 is orthorhombic Cmcm.

postulated that the perovskite to post perovskite transition may account for the discontinuity that defines the D′′ layer near 2600 km depth. The structure (Fig. 35) is orthorhombic, Cmcm, and has edge-sharing silicate octahedra forming chains parallel to a, which are corner-linked to form sheets in the a-c plane. The sheets are linked together with 8-coodinated Mg atoms to form a strongly anisotropic structure. There are two distinct oxygen sites in the structure. Of these, O1 is slightly underbonded, being coordinated to two Si and two Mg atoms, whereas O2 is slightly overbonded to two Si and three Mg. However the potentials are rather similar to those of MgSiO3-perovskite (Table 3).

Zircon Zircon (ZrSiO4) is a primary accessory phase in nearly all igneous rocks, and a major host phase for minor U, Th, and rare earth elements in the Earth. Though nominally anhydrous, nonmetamict zircons of mantle origin can contain up to about 100 ppmw H2O (Woodhead et al. 1991; Nasdala et al. 2001). This minor hydration is consistent with the very deep potential of the oxygen site (Table 3), and probably requires trivalent cation substitution for Zr. Additionally, metamict zircons, which have experienced radiation damage from the decay of U and Th, may contain much more H2O, more than 16% by weight H2O (Woodhead et al. 1991). The structure (Hazen and Finger 1979) is illustrated in Figure 36 and has Si in tetrahedral and Zr in eight-coordination. All O atoms are equivalent and bonded to tetrahedral Si so there are no non-silicate oxygens. Woodhead et al. (1991) report that strong absorption features at 3385 cm−1 perpendicular to c, and a weaker feature at 3420 cm−1 parallel to c, are associated with an occupied tetrahedron and trivalent cation substitution for Zr. However, if the proton is located on an O-O polyhedral edge, the only edge of the Zr polyhedron that does not have a component Figure 36. The structure of zircon, ZrSiO4, is in the c-direction is the edge shared with the tetragonal, I41/amd. In this c-axis projection, tetrahedron. This would be consistent with the the Zr is seen as eight-coordinated dipyramids suggestion of Nasdala et al. (2001) that hydra(light) and the Si (dark) is tetrahedral. All tion also appears to occur by the hydro-garnet oxygens are equivalent and bonded to two Zr and one Si. substitution involving tetrahedral vacancy.

110

Smyth

Titanite Titanite (CaTiSiO5), like zircon, is a very common primary accessory phase in igneous rocks. The structure (Fig. 37) is monoclinic, P21/a (b-unique) and has Ca in eight-coordination with Ti in octahedral and Si in tetrahedral coordination. Although it is nominally anhydrous, it can accommodate substantial amounts of both OH and F with Al substitution for Ti. There is one non-silicate oxygen in the structure (O1) which is bonded to one Ca and two Ti atoms. It is under-bonded in the Pauling sense, and its electrostatic site potential is 24.9 V which makes it the obvious candidate for protonation to accommodate Al or Fe3+ in the octahedron.

Figure 37. The structure of titanite, CaTiSiO5, is monoclinic, P21/a.

CONCLUSIONS The structure of the nominally hydrous and anhydrous phases that compose the Earth’s mantle have been reviewed and compared. Among the nominally hydrous high-pressure silicate phases, we have examples of molecular water in lawsonite and K-cymrite. We also see that for hydroxyl-bearing silicates, the hydroxyls are in general, non-silicate oxygens. We see no examples of a proton on tetrahedral silicate oxygens. There are a few examples of protonated tetrahedral silicate oxygens in nature such as in the pyroxenoids, pectolite (NaHCa2Si3O9) and serandite (NaHMn2Si3O9). In these structures the chains are so strongly kinked that two of the non-bridging oxygens approach so closely that there is a H-bond between the two (Jacobsen et al. 2000). We also see a few examples of Si-OH bonds for octahedral silica, as in the very high pressure phases D and Egg. This is consistent with the octahedral Si-O bond being longer and weaker than the tetrahedral Si-O bond. Among the nominally anhydrous phases we see that the phases that have only bridging tetrahedral silicate oxygens are able to accommodate the least amount of H, whereas phases containing non-silicate oxygens are readily hydrated. The minerals containing octahedral silica can accept up to several thousand ppmw H2O if Al is present to substitute for octahedral silica.

ACKNOWLEDGMENT The author thanks U.S. National Science Foundation for grant NSF-EAR 03-36611, the Bayerisches Geoinstitut Visitors Program, and the Alexander von Humboldt Foundation. The author also thanks H. Keppler, T. Boffa-Balaran and P. Comodi for constructive, thorough, and competent reviews.

REFERENCES Akaogi M, Akimoto S, Horioka H, Takahashi K, Horiuchi H (1982) The system NiAl2O4-Ni2SiO4 at high pressures ad high temperatures: spinelloids with spinel-related structures. J Solid State Chem 44:257-267 Angel RJ, Finger LW, Hazen RM, Kanzaki M, Weidner DJ, Liebermann RC, Veblen DR (1989) Structure and twinning of single-crystal MgSiO3 garnet synthesized at 17 GPa and 1800 C. Am Mineral 74:509-512 Armbruster T, Geiger CA, Lager G (1992) Single crystal X-ray study of synthetic pyrope almandine garnets at 100 and 293 K. Am Mineral 77:512-521

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

111

Baur WH (1978) Crystal structure refinement of lawsonite. Am Mineral 63:311-315 Beard JS (2000) Occurrence and composition of tochilinite and related minerals in Site 1068 serpentinites. In: Proceedings of the Ocean Drilling Program, Scientific Results, Return to Iberia. Beslier M-O, Whitmarsh RB, Wallace PJ, Girardeau J (eds) OPD, 173:1–9 Bell DR, Rossman GR (1992) Water in the Earth’s mantle: Role of nominally anhydrous minerals. Science 255: 1392-1396 Bell DR, Rossman GR, Moore RO (2004) Abundance and partitioning of OH in a high-pressure magmatic system: megacrysts from the Monastery Kimberlite, South Africa. J Petrol 45:1539-1564 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2004) Hydroxide in kyanite: A quantitative determination of the absolute amount and calibration for the IR spectrum. Am Mineral 89:998-1003 Bell DR, Rossman GR, Maldener J, Endisch D, Rauch F (2003) Hydroxide in olivine: a quantitative determination of the absolute amount and calibration of the IR spectrum. J Geophys Res 108 doi:10.1029/ 2001JB000679 Beran A, Goetzinger MA (1987) Quantitative IR spectroscopic determination of structural OH groups in kyanites. Mineral Petrol 36:41-49 Bercovici D, Karato S-I (2003) Whole-mantle convection and the transition-zone water filter. Nature 425:3944 Beran A, Libowitzky E (2006) Water in natural mantle minerals II: olivine, garnet and accessory minerals. Rev Mineral Geochem 62:169-191 Berry A, James M (2001) Refinement of hydrogen positions in synthetic hydroxyl clinohumite by powder neutron diffraction. Am Mineral 86:181-184 Boffa-Balaran T, Angel RJ (2003) Equation of state and high-pressure phase transitions in lawsonite. Eur J Min 15:241-246 Bolfan-Casanova N, Keppler H, Rubie DC (2002) Hydroxyl in MgSiO3 akimotoite: a polarized and high pressure IR study. Am Mineral 87:603-608 Bolfan-Casanova N, Mackwell S, Keppler H, McCammon CA, Rubie DC (2002) Pressure dependence of H solubility in magnesiowustite up to 25 GPa: Implications for the storage of water in the Earth’s lower mantle. Geophys Res Lett 29:1029-1032 Bolfan-Casanova N, Keppler H, Rubie DC (2003) Water partitioning at the 660 km discontinuity and evidence for very low water solubility in magnesium silicate perovskite. Geophys Res Lett 30:1905-1908 Bolfan-Casanova N, Keppler H, Rubie DC (2000) Water partitioning between nominally anhydrous minerals in the MgO-SiO2-H2O system up to 24 GPa: implications for the distribution of water in the Earth’s mantle. Earth Planet Sci Lett 182:209-221 Bromiley GD, Keppler H (2004) An experimental investigation of hydroxyl solubility in jadeite and Na-rich clinopyroxenes. Contrib Mineral Petrol 147:189–200 Bromiley GD, Keppler H, McCammon C, Bromiley FA, Jacobsen SD (2004) Hydrogen solubility and speciation in natural, gem-quality chromian diopside. Am Mineral 89:941–949 Burnley PC, Navrotsky A (1996) Synthesis of high-pressure hydrous magnesium silicates: observations and analysis. Am Mineral 81:317–326 Cameron M, Sueno S, Prewitt CT, Papike JJ (1973) High temperature crystal chemistry of acmite, diopside, hedenbergite, jadeite, spodumene, and ureyite. Am Mineral 58:594-618 Coes L (1962) Synthesis of minerals at high pressures. In: Modern Very High Pressure Techniques. Wentorf R (ed) Butterworths, p 137-150 Comodi P, Fumagalli P, Nazzareni S, Zanazzi PF (2005) The 10 Å phase: Crystal structure from single-crystal X-ray data. Am Mineral 90:1012-1016 Daniels P, Wunder B (1993) Crystal structure of tri-aluminium tri-hydroxo disilicate, Al3Si2O7(OH)3. Z Kristallogr 206:103-105 Daniels P, Wunder B (1996) Crystal structure of phase Pi, Al3Si2O7(OH)3. Eur J Min 8 1283 Domanik K, Holloway JR (1996): The stability and composition of phengitic muscovite and associated phases from 5.5 to 11 GPa: Implications for deeply subducted sediments. Geochim. Cosmochim Acta 64:41334150 Eggleton RA, Boland JN, Ringwood AE (1978) High-pressure synthesis of a new aluminum silicate: Al5Si5O17OH. Geochem J 12:191–194 Fasshauer DW, Chatterjee ND, Marler B (1997) Synthesis, structure, thermodynamic properties and stability relations of K-cymrite. Phys Chem Minerals 24:455-462 Finger LW, Hazen RM, Zhang J, Ko J, Navrotsky A (1993) The effect of Fe on the crystal structure of wadsleyite β-(Mg1-xFex)2SiO4, 0.00 < x < 0.40. Phys Chem Minerals 19:361-368 Finger LW, Ko J, Hazen RM, Gasparik T, Hemley RJ, Prewitt CT, Weidner DJ (1989) Crystal chemistry of Phase B and an anhydrous analogue: Implication for water storage in the mantle. Nature 341:140-142 Franz G, Liebscher A (2004) Physical and chemical properties of the epidote minerals – an introduction. Rev Mineral Geochem 56:1-81

112

Smyth

Friedrich A, Lager GA, Kunz M, Chakoumakos BC, Smyth JR, Schultz AJ (2001) Temperature-dependent single-crystal neutron diffraction study of natural chondrodite and clinohumite. Am Mineral 86:981-989 Frost DJ, Fei Y (1999) Static compression of the hydrous magnesium silicate phase D to 30 GPa at room temperature. Phys Chem Minerals 26:415-418 Fujino K, Sasaki S, Takeuchi Y, Sadanaga R (1981) X-ray determinations of electron distributions in forsterite, fayalite, and tephroite. Acta Crystallogr B 37:513-518 Fumagalli P, Poli S (2005) Experimentally determined phase relations in hydrous peridotites to 6.5 GPa and their consequences on the dynamics of subduction zones. J Petrol 46:555–578 Fumagalli P, Stixrude L, Poli S, Snyder D (2001) The 10 Å phase: a high-pressure expandable sheet silicate stable during subduction of hydrated lithosphere. Earth Planet Sci Lett 186:125-141 Gabe EJ, Portheine JC, Whitlow SH (1973) A reinvestigation of the epidote structure: confirmation of the iron location. Am Mineral 58:212-253 Grevel K-D, Nowlan, EU, Fasshauer DW, Burchard M (2000) X-ray diffraction analysis of lawsonite and zoisite at high pressures and temperatures. Am Mineral 85:206-216 Hawthorne FC, Grundy HD (1976) The crystal chemistry of the amphiboles IV. X-ray and neutron refinements of the crystal structure of tremolite. Can Mineral 14:334-345 Hazen RM (1976) Effects of temperature and pressure on the cell dimension and X-ray temperature factors of periclase. Am Mineral 61:266-271 Hazen RM (1981) Systematic variation of bulk modulus of wüstite with stoichiometry. Carnegie Inst Wash Yearb 80:277-280 Hazen RM, Burnham CW (1973) The crystal structures of one-layer muscovite and annite. Am Mineral 58: 889-900 Hazen RM, Finger LW (1979) Crystal structure and compressibility of zircon at high pressure. Am Mineral 64: 196-201 Hazen RM, Finger LW, Ko J (1992) Crystal chemistry of Fe-bearing anhydrous phase B: Implications for transition zone mineralogy. Am Mineral 77:217-220 Higo Y, Inoue T, Irifune T, Yurimoto H (2001) Effect of water on the spinel-postspinel transformation in Mg2SiO4. Geophys Res Lett 28:3505-3508 Hirschmann MM, Aubaud C, Withers AC (2005) Storage capacity of H2O in nominally anhydrous minerals in the upper mantle. Earth Planet Sci Lett 236:167-181 Holl CM (2006) Effects of hydration on the structure and compression of wadsleyite: Implications for water in the Earth’s interior. Ph.D. Dissertation Geological Sciences, University of Colorado, Boulder. Horioka K, Takahashi K, Morimoto N, Horiuchi H, Akaogi M, Akimoto S (1981) Structure of nickel aluminosilicate (Phase IV): A high-pressure phase related to spinel. Acta Crystallogr B37:635-638 Horiuchi H, Morimoto N, Yamamoto K, Akimoto S (1979) Crystal structure of 2Mg2SiO4 · 3Mg(OH)2, a new high-pressure structure type. Am Mineral 64:593-598 Horiuchi H, Hirano M, Ito E, Matsui Y (1982) MgSiO3 (ilmenite-type): Single crystal X-ray diffraction study. Am Mineral 67:788-793 Horiuchi H, Ito E, Weidner D (1987) Perovskite-type MgSiO3: Single-crystal X-ray diffraction study. Am Mineral 72:357-360 Huang X, Xu Y, Karato SI (2005) Water content of the transition zone form electrical conductivity of wadsleyite and ringwoodite. Nature 434:746-749 Inoue T, Yurimoto H, Kudoh Y (1995) Hydrous modified spinel, Mg1.75SiH0.5O4: a new water reservoir in the mantle transition region. Geophys Res Lett 22:117-120 Jacobsen SD (2006) Effect of water on the equation of state of nominally anhydrous minerals. Rev Mineral Geochem 62:321-342 Jacobsen SD, Smyth JR (2006) Effect of water on the sound velocities of ringwoodite in the transition zone. Earth’s Deep Water Cycle, AGU Monograph Series (in press) Jacobsen SD, Lin JF, Angel RJ, Shen G, Prakapenka VB, Dera P, Mao HK, and Hemley RJ (2005) Single-crystal synchrotron X-ray diffraction study of wüstite and magnesiowüstite at lower-mantle pressures. J Synchr Rad 12:577-583 Jacobsen SD, Demouchy S, Frost DJ, Boffa-Ballaran T, Kung J (2005) A systematic study of OH in hydrous wadsleyite from polarized FTIR spectroscopy and single-crystal X-ray diffraction: Oxygen sites for hydrogen storage in the Earth’s interior. Am Mineral 90:61-70 Jacobsen SD, Smyth JR, Spetzler H, Holl CM, Frost DJ, (2004) Sound velocities and elastic constants of ironbearing hydrous ringwoodite. Phys Earth Planet Int 143-144:47-56 Jacobsen SD, Smyth JR, Swope RJ, Sheldon RI (2000) Two proton positions in the very strong H-bond of serandite [Na,Mn2Si3O8OH]. Am Mineral 85:745-752 Libowitzky E, Beran A (2006) The structure of hydrous species in nominally anhydrous minerals: information from polarized IR spectroscopy. Rev Mineral Geochem 62:29-52

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

113

Kagi H, Parise JB, Cho HM, Rossman GR, Loveday JS (2000) Hydrogen bonding interactions in phase A [Mg7Si2O8(OH)6] at ambient and high pressure. Phys Chem Minerals 27:225–233 Kanzaki M (1991) Stability of hydrous magnesium silicates in the mantle transition zone. Phys Earth Planet Int 66:307–312 Karato S (1990) The role of hydrogen in the electrical conductivity of the upper mantle. Nature 347:272-273 Karato S, Paterson MS, Fitzgerald D (1986) Rheology of synthetic olivine aggregates: Influence of grain size and water. J Geophys Res 91:8151-8176 Kavner A (2003) Elasticity and strength of hydrous ringwoodite at high pressure. Earth Planet Sci Lett 214: 645-654 Katayama I, Nakashima S (2003) Hydroxyl in clinopyroxene from the deep subducted crust: Evidence for H2O transport into the mantle. Am Mineral 88:229-234 Kawamoto T, Hervig RL, Holloway JR (1996) Experimental evidence for a hydrous transition zone in the early Earth’s mantle. Earth Planet Sci Lett 142:587-592 Kek S, Aroyo M, Bismayer U, Schmidt C, Eichhorn K, Krane HG (1997) The two-step phase transition of titanite, CaTiSiO5: a synchrotron radiation study. Zeit Kristallogr 212:9–19 Keppler H, Smyth JR (2005) Optical and near infrared spectra of ringwoodite to 21.5 GPa. Am Mineral 90: 1209-1214 Kihara K (1990) An X-ray study of the temperature dependence of the quartz structure. Eur J Mineral 2:63-77 Koch-Mueller M, Dera P, Fei Y, Hellwig H, Liu Z, Van Orman J, Wirth R (2005) Polymorphic phase transition in superhydrous phase B. Phys Chem Minerals 32:349-361 Koch-Müller M, Dera P, Fei Y, Reno B, Sobolev N, Hauri E, Wysoczanski R (2003) OH- in synthetic and natural coesite. Am Mineral 88:1436-1445 Koch-Müller M, Fei Y, Hauri E, Liu Z (2001) Location and quantitative analysis of OH in coesite. Phys Chem Minerals 28:693-705 Kohlstedt DL, Keppler H, Rubie DC (1996) Solubility of water in the α, β, and γ phases of (Mg,Fe)2SiO4. Contr Mineral Petrol 123:345-357 Kohn SC, Brooker RA, Frost DJ, Slesinger AE, Wood BJ (2002) Ordering of hydroxyl defects in hydrous wadsleyite (β-Mg2SiO4). Am Mineral 87:293-301 Kudoh Y, Kuribayashi T, Mizohata H, Ohtani E (2000) Structure and cation disorder of hydrous ringwoodite, γ-Mg1.89Si0.97H0.34O4. Phys Chem Minerals 27:474-479 Kudoh Y, Finger LW, Hazen RM, Prewitt CT, Kanzaki M, Veblen DR (1993) Phase E: a high pressure hydrous silicate with unique crystal chemistry. Phys Chem Minerals 19:357-360 Lager GA, Von Dreele RB (1996) Neutron powder diffraction study of hydrogarnet to 9.0 GPa. Am Mineral 81: 1097-1104 Lager GA, Marshall WG, Liu Z, Downs RT (2005) Re-examination of the hydrogarnet structure at high pressure using neutron powder diffraction and infrared spectroscopy. Am Mineral 90:639-644 Lager GA, Ulmer P, Miletich R, Marshall WG (2001) O-D...O bond geometry in OD-chondrodite. Am Mineral 86:176-180 Lager GA, Armbruster T, Faber J (1987) Neutron and X-ray diffraction study of hydrogarnet Ca3A12(O4H4)3. Am Mineral 72:756–765 Libowitzky E (1999) Correlation of O-H stretching frequencies and O-H. . .O hydrogen bond lengths in minerals. Monatsh. Chemie 130:1047-1059 Litasov K, Ohtani E, Langenhorst F, Yurimoto H, Kubo T, Kondo T (2003) Water solubility in Mg-perovskites and wate storage capacity in the lower mantle. Earth Planet Sci Lett 211:189-203 Luth RW (1995) Is phase A relevant to the Earth’s mantle? Geochim Cosmochim Acta 59:679-682 Ma C-B, Sahl K (1975) Nickel aluminosilcate, phase III. Acta Crystallogr B31:2142-2143 Meade C, Reffner JA, Ito E (1994) Synchrotron infrared absorbance measurements of hydrogen in MgSiO3 perovskite. Science 264:1558–1560 Mellini M (1982) The crystal structure of lizardite 1T: Hydrogen bonds and polytypism. Am Mineral 67:587598 Mierdel K, Keppler H (2004) The temperature dependence of water solubility in enstatite. Contrib Mineral Petrol 148:305–311 Mierdel K, Kepler H, Smyth JR, Langenhorst F (2006) The origin of the Earth’s asthenosphere. Nature (in review) Mosenfelder JL (2000) Pressure dependence of hydroxyl solubility in coesite. Phys Chem Minerals 27:610617 Mosenfelder JL, Deligne NI, Asimow PD, Rossman GR (2006) Hydrogen incorporation into olivine at 2-12 GPa. Am Mineral 91:285-294 Murakami M, Hirose K, Yurimoto H, Nakashima S, Takafuji N (2002) Water in the Earth’s lower mantle. Science 295:1885-1887

114

Smyth

Murakami M, Hirose K, Kawamura K, Sata N, Ohishi Y (2004) Post-perovskite phase transition in MgSiO3. Science 304:855-858 Nasdala L, Beran A, Libowitzky E, Wolf D (2001) The incorporation of hydroxyl groups and molecular water in natural zircon (ZrSiO4). Am J Sci 301:831-857 Newnham RE, deHaan YM (1962) Refinement of the alpha-Al2O3, Ti2O3, V2O3, and Cr2O3 structures. Z Kristallogr 117:235-237 Northrup PA, Leinenweber K, Parise JB (1994) The location of H in the high-pressure synthetic Al2SiO4(OH)2 topaz analogue. Am Mineral 79:401-404 Pacalo REG, Parise JB (1992) Crystal structure of superhydrous B, a hydrous magnesium silicate synthesized at 1400 °C and 20 GPa. Am Mineral 77:681-684 Pannhorst W (1984) High temperature crystal structure refinements of low clinoenstatite up to 700 °C. N Jahrb Mineral Abh 150:219-228 Pawley AR (1994) The pressure and temperature stability limits of lawsonite: implications for H2O recycling in subduction zones. Contr Mineral Petrol 118:99-108 Pawley AR, Redfern SAT, Holland TJB (1996) Volume behavior of hydrous minerals at high pressure and temperature: 1. Thermal expansion of lawsonite, zoisite, clinozoisite, and diaspore. Am Mineral 81:335340 Perdikatsis B, Burzlaff H (1981) Strukturverfeinerung am Talk Mg3[(OH)2Si4O10]. Z Kristalogr 156:177-186 Poli S, Schmidt MW (2004) Experimental subsolidus studies of epidote minerals. Rev Mineral Geochem 56: 171-195 Rauch M, Keppler H (2002) Water solubility in orthopyroxene. Contrib Mineral Petrol 143:525-536 Ross NL, Crichton WA (2001) Compression of synthetic hydroxyl clinohumite [Mg9Si4O16(OH)2] and hydroxyl chondrodite [Mg5Si2O8(OH)2]. Am Mineral 86:990-996 Ross NL, Gibbs GV, Rosso KM (2003) Potential docking sites and positions of hydrogen in high-pressure silicates. Am Mineral 88:1452-1459 Ross NL, Shu JF, Hazen RM, Gasparik T (1990) High-pressure crystal chemistry of stishovite. Am Mineral 75: 739-747 Rossman GR, Smyth JR (1990) Hydroxyl contents of accessory minerals in mantle eclogites and related rocks. Am Mineral 75:775-780 Rothbauer R. (1971) Untersuchung eines 2M1-Muskovits mit Neutronenstrahlung. Neues J Mineral Monats 1971:143-154 Schilling FR, Sinogeikin SV, Bass JD (2003) Single-crystal elastic properties of lawsonite and their variation with temperature. Phys Earth Planet Inter 136:107-118 Schmidt MW, Poli S (1998) Experimentally based water budgets for dehydrating slabs and consequences for arc magma generation. Earth Planet Sci Lett 163:361-179 Schmidt MW, Finger LW, Angel RJ, Dinnebier RE (1998) Synthesis, crystal structure and phase relations of AlSiO3OH, a high pressure hydrous phase. Am Mineral 83:881-888 Shieh SR, Mao HK, Konzett J, Hemley RJ (2000) In-situ high pressure X-ray diffraction of phase E to 15 GPa. Am Mineral 85:765-769 Shintani H, Sato S, Saito Y (1975) Electron density distribution in rutile crystals. Acta Crystallogr B31:19811982 Shu J, Mao HK, Hu J, Fei Y, Hemley RJ (1998) Single-crystal X-ray diffraction of wüstite to 30 GPa hydrostatic pressure. N Jahrb Mineral Abh 172:309–323 Skogby H (2006) Water in natural mantle minerals I: pyroxenes. Rev Mineral Geochem 62:155-167 Smyth JR (1980) Cation vacancies and the crystal chemistry of breakdown reactions in kimberlitic omphacites. Am Mineral 65:1185-1191 Smyth JR (1987) β-Mg2SiO4: a potential host for water in the mantle? Am Mineral 72:1051-1055 Smyth JR (1989) Electrostatic characterization of oxygen sites in minerals. Geochim Cosmochim Acta 53: 1101-1110 Smyth JR, Artioli G, Smith JV, Kvick A (1987) Crystal structure of coesite a high pressure form of SiO2 at 15 and 298 K from single crystal neutron and X-ray diffraction data: test of bonding models. J Phys Chem 91:988-992 Smyth JR, Dyar MD, May HH, Bricker OP, and Acker HA (1997) Crystal structure refinement and Mössbauer spectroscopy of an ordered triclinic chlorite. Clays Clay Mineral 45:544-550 Smyth JR, Frost DJ (2002) The effect of water on the 410-km discontinuity: an experimental study. Geophys Res Lett 29:1485, doi:10.1029/2001GL014418 Smyth JR, Kawamoto T (1997) Wadsleyite II: A new high pressure hydrous phase in the peridotite-H2O system. Earth Planet Sci Lett 146:E9-E16 Smyth JR, Frost DJ, Nestola F, Holl CM, Bromiley G (2006a) Olivine hydration in the deep upper mantle: effects of temperature and silica activity. Geophys Res Lett 33, doi:2006GL026194 (in press)

Hydrogen in High Pressure Silicate & Oxide Mineral Strutures

115

Smyth JR, Holl CM, Frost DJ, Jacobsen SD, (2004) High pressure crystal chemistry of hydrous ringwoodite and water in the Earth’s interior. Phys Earth Planet Int 143-144:271-278 Smyth JR, Holl CM, Frost DJ, Jacobsen SD, Langenhorst F, McCammon CA (2003) Structural systematics of hydrous ringwoodite and water in the Earth’s interior. Am Mineral 88:1402-1407 Smyth JR, Holl CM, Langenhorst F, Laustsen HMS, Rossman GR, Kleppe A, McCammon CA, Kawamoto T, van Aken PA (2005) Crystal chemistry of wadsleyite II and water in the Earth’s interior. Phys Chem Minerals 31:691-705 Smyth JR, Jacobsen SD, Swope RJ, Angel RJ, Arlt T, Domanik K, Holloway JR (2000) Crystal structures and compressibilities of synthetic 2M1 and 3T phengite micas. Eur J Mineral 12:955-963 Smyth JR, Kawamoto T, Jacobsen SD, Swope RJ, Hervig RL, Holloway JR (1997) Crystal structure of monoclinic hydrous wadsleyite. Am Mineral 82:270-275 Smyth JR, Mierdel K, Keppler H, Langenhorst F, Dubrovinsky L, Nestola F (2006b) Crystal chemistry of hydration in aluminous orthopyroxene. American Mineralogist (in prep.) Smyth JR, Rossman GR, Bell DR (1991) Incorporation of hydroxyl in upper mantle clino-pyroxenes. Nature 351:732-735 Smyth JR, Swope RJ, Pawley AJ (1995) H in rutile-type compounds: II. Crystal chemistry of Al substitution in H-bearing stishovite. Am Mineral 80:454-456 Sueno S, Cameron M, Prewitt CT (1976) Orthoferrosilite: High temperature crystal chemistry. Am Mineral 61: 38-53 Swope RJ, Smyth JR, Larson AC (1995) H in rutile-type compounds: I. Single-crystal neutron and X-ray diffraction study of H in rutile. Am Mineral 80:448-453 Takeda H (1973) Tetrahedral sizes of orthopyroxenes and silicon-aluminum ordering. Am Mineral 58:10961097 Vlassopoulos D, Rossman GR, Haggerty SE (1993) Coupled substitution of H and minor elements in rutile and the implications of high OH contents in Nb- and Cr-rich rutile from the upper mantle. Am Mineral 78: 1181-1191 Vanpeteghem CB, Ohtani E, Kondo T, Takemura K, Kikegawa T (2003) Compressibility of phase Egg AlSiO3OH: Equation of state and role of water at high pressure. Am Mineral 88:1408-1411 Winter JK, Ghose S (1979) Thermal expansion and high temperature crystal chemistry of the Al2SiO5 polymorphs. Am Mineral 64:573-586 Wood BJ (1995) The effect of H2O on the 410-kilometer seismic discontinuity. Science 268:74-76 Woodhead JA, Rossman GR, Thomas AP (1991) Hydrous species in zircon. Am Mineral 76:1533-1546 Wunder B (1998) Equilibrium experiments in the system MgO-SiO2-H2O (MSH): stability fields of clinohumiteOH [Mg9Si4O16(OH)2], chondrodite-OH [Mg5Si2O8(OH)2] and phase A [Mg7Si2O16(OH)6]. Contr Mineral Petrol 132:111-120 Wunder B, Andrut M, Wirth R (1999) High pressure synthesis and properties of OH-rich topaz. Eur J Mineral 11:803-813 Wunder B, Medenbach O, Daniels P, Schreyer W (1995) First synthesis of the hydroxyl end-member of humite Mg7Si3O12(OH)2. Am Mineral 80:638-640 Wunder B, Medenbach O, Krause W, Schreyer W (1993)Synthesis, properties and stability of Al3Si2O7(OH)3 (phase Pi), a hydrous high-pressure phase in the system Al2O3-SiO2-H2O (ASH). Eur J Mineral 5:637649 Yamamoto K, Akimoto S (1977) The system MgO-SiO2-H2O at high pressure and temperatures – stability field for hydroxyl-chondrodite, hydroxyl-clinohumite, and 10 Å-phase. Am J Sci 277:288-312 Yagi T, Marumo F, Akimoto S (1974) Crystal structures of spinel polymorphs of Fe2SiO4 and Ni2SiO4. Am Mineral 59:486-490 Yang H, Prewitt CT (2000) Chain and layer silicates at high temperatures and pressures. Rev Mineral Geochem 41:211-256 Yang H, Prewitt CT, Frost DJ (1997) Crystal structure of the dense hydrous magnesium silicate, phase D. Am Mineral 82:651-654 Yang H, Hazen RM, Prewitt CT, Finger LW, Lu R, Hemley RJ (1998) High-pressure single-crystal X-ray diffraction and infrared spectroscopic studies of C2/m-P21/m phase transition in cummingtonite. Am Mineral 83:288-299 Yang H, Ghose S (1995) High temperature single crystal X-ray diffraction study of the ortho-proto phase transition in enstatite Mg2Si2O6. Phys Chem Minerals 22:300-310 Zigan F, Rothbauer R (1967) Neutronenbeugungsmessung am Brucit. N Jahrb Mineral Monats 137-143