Advanced Inorganic Fibers: Processes - Structures - Properties - Applications

ADVANCED INORGANIC FIBERS Processes - Structures - Properties - Applications

MATERIALS TECHNOLOGY SERIES Series editor: Renee G. Ford The Materials Technology series is dedicated to state-of-the-art areas of materials synthesis and processing as related to the applications of the technology. By thorough presentation of the science underlying the technology, it is anticipated that these books will be of practical value both for materials scientists and engineers in industry and for engineering students to acquaint them with developments at the forefront of materials technology that have potential commercial significance. Ceramic Injection Molding Beebhas C. Mutsuddy and Renee G. Ford Hardbound (0 412 538105) Cryochemical Technology of Advanced Materials Yu. D. Tretyakov, N.N. Oleynikov and O.A. Shlyakhtin Hardbound (0412 639807) Modelling of Materials Processing Gregory C. Stangle Hardbound (041271120 6) Porous Materials Kozo Ishizaki, Sridhar Komarneni, Makota Nanko Hardbound (0412711109) Functionally Graded Materials Yoshinari Miyamoto, Wolfgang A. Kaysser, Barry H. Rabin, Akira Kawasaki, Renee G. Ford Hardbound (0412 607603)

ADVANCED INORGANIC FIBERS Processes - Structures - Properties - Applications Contributors:

FREDERICK T. WALLENBERGER Manager, Advanced Technology PPG Fiber Glass Research Center, Pittsburgh, Pennsylvania

ROGER NASLAIN Professor, University of Bordeaux Director, High Temperature Structural Composites Laboratory Pessac, France

JOHN B. MACCHESNEY Fellow, Bell Laboratories Lucent Technologies, Murray Hill, New Jersey

HAROLD D. ACKLER Lawrence Livermore National Laboratory Livermore, California

Editor: FREDERICK T. WALLENBERGER

KLUWER ACADEMIC PUBLISHERS Boston / Dordrecht / London

Distributors for North, Central and South America: Kluwer Academic Publishers 101 Philip Drive Assinippi Park Norwell , Massachusetts 02061 USA Telephone (781) 871-6600 Fax (781) 871-6528 E-Mail

Distributors for all other countries: Kluwer Academic Publishers Group Distribution Centre Post Office Box 322 3300 AH Dordrecht, THE NETHERLANDS Telephone 31 78 6392 392 Fax3l 786546474 E-Mail "

Electronic Services

Library of Congress Cataloging-in-Publication Data Advanced inorganic fibers : processes--structures --properties--applicationsl contributors, Frederick T.Wallenberger. . .[et al.]; editor,Frederick T. Wallenberger. p. em. -- (Materials technology series) ISBN 0-412-60790-5 1. Inorganic fibers I. Wallenberger, Frederick T., 1930TA418 .9.F5 A38 1999 620.1921--dc21

99-046026

Copyright @) 2000 by Kluwer Academic Publishers . All rights reserved . No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, mechanical, photocopying , recording , or otherwise, without the prior written permission of the publisher, Kluwer Academic Publishers, 101 Philip Drive, Assinippi Park, Norwell, Massachusetts 02061 Printed on acid-free paper. Printed in the United States of America

TABLE OF CONTENTS SECTION I. INTRODUCTION F. T. Wallenberger 1 1.1 1.2 1.3 1.4

FIBERS FROM THE VAPOUR PHASE The most important phase isthe liquidphase Afibre by any name isstill afiber Biographic Sketches of the authors Acknowledgements

3 3

4 6 7

SECTION II. FIBERS FORM THE VAPOUR PHASE F. T. Wallenberger

2

2.1

2.2

2.3 2.4

SHORT FIBERS, WHISKERS AND NANOTUBES Advanced vapor phase processes 2.1 .1 Evolution ofa technology 2.1 .2 Crystal growth and phase transformations (a) Vapor-liquid-solid (VLS) growth (b) Vapor-solid (VS) growth 2.1 .3 Metal catalyzed chemical vapor deposition (a) Reaction chemistry (~ Controlled whisker growth (c Whisker morphology (~ Generic whisker properties 2.1.4 Laser ablation ofwhisker precursor alloys 2.1 .5 Hot fiber chemical vapor deposition 2.1 .6 Chemical vapor infiltration 2.1 .7 Carbothermal reduction (a) Pyrolytic processes (b) Chemical mixing processes (c) Self-propagating high-temperature synthesis 2.1 .8 Plasma and related processes (a) Arc discharge processes (b) Laser vaporization and ion bombardment Advanced liquid phase processes 2.2.1 Self-assembly ofsilver nanowires 2.2.2 Whiskers from organic solvents 2.2.3 Whiskers from mesopitch Advanced solid phase processes 2.3.1 Micropillars by lithography and etching Selected fiber structures and properties 2.4.1 Silicon whiskers and nanowhiskers 2.4.2 Silicon carbide whiskers and nanowhiskers

11 11 11 12 12 13

15 15 18

19

20 20

21 22

23 23 23 24 24 24

25

26 26

27 28

29 29 30 30 34

vi

2.4.3 2.4.4

2.5

Short graphite fibers Carbon nanotubes (a) Structures (b) Properties Selected fiber products and applications 2.5.1 Silicon whiskers and nanowhiskers 2.5.2 Silicon carbide whiskers and nanowhiskers 2.5.3 Short carbon and diamond fibers Short carbon fiber composites Diamond/carbon fiber composites 2.5.4 Carbon nanotubes

36 36 37 39 39 40 41 41 41 42

CONTINUOUS OR ENDLESS INORGANIC FIBERS Continuous vapor phase processes 3.1 .1 Laser assisted chemical vapor deposition (a) The rceneric process concept (~ The ow pressure process (c The high pressure process (~ Automatic process control 3.1.2 Conventional chemical vapor deposition (a) Commercial hot filament CVD process (b) Experimental CVD and PVD processes 3.1 .3 Chemical vapor infiltration processes (a) CVI ofcarbon fibers with silicon oxide (bj CVI ofboron oxide fibers with ammonia (c CVI ofpolyborazine fibers with ammonia 3.1.4 Laser vaporization ofcarbon-metal mixtures Selected structures and properties 3.2.1 Hit and low pressure LCVD fibers (a Reactor pressure vs. growth rate (~ Tip temperature vs. ~roperties (c Side growth versus ipgrowth ~ ~ Versatility versus whIsker processes 3.2.2 Commercial hot filament CVD fibers (a) Sheath/core boron/tungsten fibers (bj Sheath/core versus pure boron fibers (c Sheath/core silicon carbide/carbon fibers Important CVI and PVD fibers 3.2.3 3.2.4 Structure - property commonalties (a) Straight, coiled and tubular structures (b) Fiber strength, modulus and toughness Selected products and applications 3.3.1 BIW and SiC/C fiber reinforced composites 3.3.2 Rapid evaluation ofnew fibers by LCVD (a) Ultrahigh temperature fibers (b) High temperature sensor fibers 3.3.3 Rapid PrototyPinp ofmicroparts by LCVD (a) Evolution 0 rapid prototyping (~ Laser chemical vapor deposition (c Photonic band-gap microstructure ~ ~ The future ofvapor phase processing

47 47 47 47 49 53 54 55 55 56 59 59 59 60 60 60 61 61 62 63 63 65 66 66 67 68 69 69 69 70 70 71 71 73 73 74 74 75 75

f~

3

3.1

3.2

3.3

34

vii

SECTION III. FIBERS FROM THE LIQUID PHASE F. T. Wallenberger with a chapter by H. Ackler and J. MacChesney

4 4.1

4.2

4.3

4.4

CONTINUOUS MELT SPINNING PROCESSES Important melt forming processes 4.1 .1 Princ~les offiber formation (a) ehavior ofviscous melts (li] Behavior ofinviscid melts (c Generic fiber forming processes Structure ofmelts and fibers 4.1 .2 (a) From melts tofibers (li] Fiber structure versus modulus (c Fiber structure versus strength Forming glass fibers from strong melts 4.2.1 Downdrawing from solid preforms (a) Structural silica fibers (li) Optical silica fibers Melt spinning from strong silicate melts 4.2.2 4.2.3 Structural silicate glass fibers (a) Product design parameters (li) General and special purpose fibers Forming glass fibers from fragile melts 4.3.1 Glass fibers from fragile silicate melts 4.3.2 Melt spinning from supercooled melts (a) Single and double crucible processes (li) Single and bicomponent fluoride fibers 4.3.3 Updrawing from supercooled melts (a) P/odrawing oftellurite ~/ass fibers (li) pdrawing ofalumina e glass fibers Hybrid fiber forming processes 4.3.4 4.3.5 Quaternary calcium aluminate fibers (a) Fiber properties (b) Potential applications Forming amorphous fibers from inviscid liquids 4.4.1 Attainment offiber forming viscosities 4.4.2 Rapid solidification (RS) processes (a) Amorphous metal ribbons (b) Products and applications 4.4.3 Inviscid melt spinning (IMS processes (a) Principles ofjetand fi erformation (b) Principles ofincreasing the jet lifetime 4.4.4 Oxide fibers from containerless, laser heated melts 4.4.5 Metal fibers in a reactive environment 4.4.6 Oxide glass fibers inareactive environment 4.4.7 Mechanism ofjetsolidification 4.4.8 Cryogenic fibers from liquefied gasses Growing single crystal fibers from inviscid melts 4.5.1 Edge defined film fed growth (a) Growth ofsapphire fibers (b) Process versatility 4.5.2 Laser heated float zone growth (a) Growth ofsingle crystal fibers

6

4.5

81 81 81 81 84 85 87 87 88 91 92 92 92 92 92 93 94 94 95 95 97 97 97 97 98 98 100 101 101 102 103 103 103 104 105 105 106 107 107 108 110 111 113 113 113 114 114 115 115

viii

4.5.3 5 5.1

5.2 6

6.1

6.2

6.3

6.4

(b) High T« superconducting fibers The future ofsingle crystal oxide fibers (a) Single crystal sapphire fibers (b) Other single crystal oxide fibers

116 118 118 119

CONTINUOUS SOLVENT SPINNING PROCESSES Dry spinning ofsilica glass fibers 5.1.1 Process concepts 5.1 .2 Pure silica fibers from water glass solutions 5.1 .3 Ultrapure silica fibers from sol-gels Silica fibers by other processes

123 123 123 124 126 128

STRUCTURAL SILICATE AND SILICA GLASS FIBERS General purpose silicate glass fibers 6.1 .1 Commercial fiber forming processes 6.1 .2 Commercial commodity 91ass fibers (a) Evolution ofborosil/cate E.glass fibers (b) Boron- and fluorine-free E.glass fibers 6.1 .3 Structures and properties (a) Mechanical properties (b) Other fiber properties 6.1.4 Commercial products and applications Special purpose silicate glass fibers 6.2.1 Hir strength - h~h temperature fibers (a Process an products (b) Properties and applications 6.2.2 High modulus - high temperature fibers 6.2.3 Ultrahigh modulus glass-ceramic fibers (a) Process and products (b) Properties and applications 6.2.4 Fibers with high chemical stability (a) Chemical resistance ofglass fibers (bJ Alkali resistant ~/ass fibers (c Acid resistant g ass fibers 6.2.5 Other special purpose fibers (a) Fibers with low dielectric constants (~ Fibers with high densities and high dielectric constants (c Fibers with very high dielectric constants (~ Fibers with super- and semiconducting properties (e) Fibers with bone bioactive glass compositions Non-round, bicomponent and hollow fibers 6.3.1 Silicate glass fibers with non-round cross sections (a) Processes and structures (b) Products and applications 6.3.2 Structural bicomponent silicate 91ass fibers (a) Sheath/core and side-by-s/de bicomponent fibers (~ HoI/ow sheath/core silicate glass fibers (c HoI/ow porous sheath/core silicate glass fibers (~ HoI/ow superconducting sheath/core glass fibers (e) Solid side-by-side bicomponent glass fibers High temperature silica glass fibers Value-in-use ofsilica glass fibers 6.4.1 6.4.2 Ultrapure silica fibers from solid preforms 6.4.3 Ultrapure and pure silica fibers from solutions 6.4.4 High silica fibers by leaching ofborosilicate fibers

129 129 129 130 130 131 132 132 133 134 136 136 136 139 140 141 142 144 145 145 146 148 149 149 150 151 152 153 153 154 154 155 156 156 156 158 158 160 162 162 163 164 165

ix

7 7.1 7.2

7.3

7.4 7.5 7.6

7.7

OPTICAL SILICA FIBERS (H. Ackler and J. MacChesney) Introduction Principles ofoptical transmission Wave guide physics 7.2.1 (a) Step index fibers (b) Graded index fibers 7.2.2 Optical loss (a) Scattering (b) Absorption 7.2.3 Dispersion Birefringence 7.2.4 Fabrication ofoptical fibers 7.3.1 Fabrication ofpreforms Doublecrucible method 7.3.2 Outside vapor deposition (OVD and VAD) 7.3.3 7.3.4 Modified chemical vapor deposition (MCVD) (a) Chem~alequmbria (b) Thermophoretic deposition and sintering 7.3.5 Plasma chemical vapor deposition (PCVD) Fiber drawing process 7.4.1 The drawing tower 7.4.2 Protective fiber coatings Sol-gel processing Applications ofoptical fiber devices 7.6.1 Optical amplifiers 7.6.2 Fiber gratings as mirrors and filters 7.6.3 Strainsensor and other applications Summary and outlook

169 169 169 169 171 172 172 173 173 174 179 180 181 181 183 185 187 189 190 191 191 192 193 194 195 196 197 198

SECTION IV. FIBERS FROM SOLID PRECURSOR FIBERS R. Naslain

8 8.1

8.2

CERAMIC OXIDE FIBERS FROM SOL·GELS AND SLURRIES General considerations 8.1 .1 The generic sol-gel process (a) The starting materials (~ The gelation step (c The drying step ~ ~ The calcination and sintering steps Alumina and alumina based fibers 8.2.1 General considerations 8.2.2 Processing ofalumina based fibers (a) Polycrystalline alumina fibers (~ Transition alumina fibers (c Mullite and related fibers (~ Alumina-zirconia fibers 8.2.3 Structure and microstructure (a) Transition alumina fibers (b) Mullite and related fibers

205 205 205 205 206 207 207 207 207 209 210 211 212 215 216 216 216

x

(c) Corundum and related fibers Mechanical properties (a) Atroom temperature (b) At high temperature 8.2.5 Physical properties Applications 8.2.6 Zirconia based fibers 8.3.1 General considerations 8.3.2 Processing of zirconiabased fibers (a) Fibers from zirconia sols (b) Fibers from polyzirconoxanes 8.3.3 Properties and applications Yttrium aluminumgamet (YAG) fibers 8.4.1 General considerations 8.4.2 Processing of YAG fibers (a) From diphasicgels fbJ From polymer precursors (c From YAG powders Properties and applications 8.4.3 8.4.4 Applications

218 21 9 219 222 224 225 225 225 226 226 226 227 227 227 228 228 228 228 228 229

CARBON FIBERS FROM PAN AND PITCH General considerations 9. 1.1 History ofcarbon fibers 9.1.2 Elemental carbon 9.1.3 Classification ofcarbon fibers Processing ofcarbon fibers 9.2.1 Principles offiber formation 9.2.2 From polyacrylonitrilebased precursor fibers (a) Nature of theprecursor (~ Spinning of PAN based precursor (c Stretchm~ (~ Stabiliza Ion (~ Carbonization ( Post heat lfeatment 9.2.3 From pitch based precursor fibers (a) Nature ofpitches (~ The carbonaceous mesophase stage fc ~inning and stabilization ~ ~ arbomzation and graphitization Structure of carbon fibers 9.3.1 Structural parameters 9.3.2 Microtexture (a) PAN based high tenacity carbon fibers fbJ PAN based hi~h modulus carbon fibers (c Mesopitch (M ?based carbon fibers Properties of carbon fibers 9.4.1 Mechanical Properties (a) Youn~ 's modulus (~ Tensl e strength fc Compressive strength ~ ~ High temperature properties 9.4.2 Thermal and electrical properties (a) Thennalexpans~n (b) Transportproperties

233 233 233 233 235 235 235 237 237 237 237 237 238 239 239 239 240 243 245 245 245 247 247 247 247 250 252 250 253 256 256 257 257 258

8.2.4

8.3

8.4

9

9.1

9.2

9.3

9.4

xi

9.5

9.4.3 9.4.4

Oxidation ofcarbon fibers Coated carbon fibers Applications

259 261 261

10

SILICON CARBIDE AND OXYCARBIDE FIBERS

265 265 266 267 269 269 270 272 272 272 275 275 275 275 276 276 276 279 280 281 283 284 284 287 287 288 291 291 293 295

10.1 General considerations 10.2 Preparation ofSi-C-O fibers 10.2.1 The Yajima process 10.2.2 Melt spinning ofPCS fibers 10.2.3 Stabilization and curing 10.2.4 Pyrolysisof PCS fibers 10.2.5 Related Si-C-O (Ti) fibers 10.3 Preparation ofoxygen-free Si-C fibers 10.3.1 From radiation cured PCS precursor fibers 10.3.2 From infusible PCS precursor fibers 10.4 Preparation ofquasi-stoichiometric SiC fibers 10.4.1 Pyrolysis of PCS precursor fibers under hydrogen 10.4.2 Pyrolysisofboron doped PCS precursor fibers 10.4.3 From extruded SiC powder/polymer mixtures 10.5 Structure ofsilicon carbide fibers 10.5.1 Silicon oxycarbide fibers 10.5.2 Silicon carbide fibers 10.6 Thermal stability ofsilicon fibers 10.6.1 Silicon oxycarbide fibers 10.6.2 Silicon carbide fibers 10.7 Mechanical properties ofSiC fibers 10.7.1 Atroom temperature 10.7.2 Athigh temfeeratures (a) Tensi e tests (b) Creep tests (c) Bend stress relaxation test 10.8 Oxidation ofsilicon carbide fibers 10.9 Transport properties ofSiC fibers 10.10 Applications 11

11.1 11.2

11.3

SILICON NITRIDE AND BORIDE BASED FIBERS General considerations Si-C-N-O and Si-C-N fibers 11 .2.1 Processing (a) From porvsilazane (PSZ) fibers (b) From poycarbosilazane (PCSZ) fibers 11.2.2 Structure and properties (a) Fiber structure (~ Thermal stability (c Mechanical pro erties ( ~ Oxidation resisrance (e) Other properties Si-N-Oand Si-N fibers 11 .3.1 Processing (a) From Yajima type polycarbosilane reS) fibers {bj From pemydrOtO%SilaZane (PHPS Jfibers {c From other poysiazane fibers

299 299 299 299 299 300 301 302 302 304 304 306 306 306 306 307 308

xii

11.3.2

11.5

Structure and properties (a) Thermal properties (bJ Mechanical properties (c Other properties Si-B-O-N, Si-B-N and Si-B-N-C fibers 11.4.1 Processing (a) From perhrodropogsilazane (PHPSZ) fibers (b) From trich orosilyamino-dichloroborane (TADB) fibers 11.4.2 Structures and properties (a) Structure and thermal stability (b) Mechanical properties Applications

308 308 309 309 309 309 310 311 311 311 311 311

12

APPLICATIONS OF CARBON AND CERAMIC FIBERS

315 315 316 316 320 322

ACRONYMS GLOSSARY

331 335

INDEX

341

11.4

12.1 Fiber applications 12.2 Composite applications 12.2.1 Polymer matrix composites 12.2.2 Metal matrix composites 12.2.3 Carbon and ceramic matrix composites

SECTION I INTRODUCTION F. T. Wallenberger This book serves as an introduction to advanced inorganic fibers and aims to support fundamental research, assist applied scientists and designers in industry, and facilitate materials science instruction in universities and colleges. Its three main sections deal with fibers which are derived from the vapor phase such as single crystal silicon whiskers or carbon nanotubes, from the liquid phase such as advanced glass and single crystal oxide fibers, and from solid precursor fibers such as carbon and ceramic fibers.

Contents FIBERS FROM THE VAPOR, LIQUID AND SOLID PHASE

1.1 1.2 1.3 1.4

The most important phase isthe liquid phase Afiber by any name is still a fiber Biographic sketches ofthe authors Acknowledgments

CHAPTER 1 FIBERS FROM THE VAPOR, LIQUID AND SOLID PHASE F. T. Wallenberger The book describes advanced inorganic fibers, focuses on principles and concepts, analyzes experimental and commercial processes, and relates process variables to structures, structures to fiber properties and fiber properties to end-use performance. In principle, there are discontinuous or inherently short, and continuous or potentially endless, fibers. Short fibers range from asbestos fibers, which were described as early as 300 BC to carbon nanotubes which were discovered in 1991 [1] and have been fully described in 1999 [2]. Continuous inorganic fibers range from silicate glass fibers which were reported in 1630, to vapor grown boron fibers which were reported in 1995 [3], single crystal germanium fibers [4] and amorphous yttrium aluminum garnet fibers [5] which were reported in 1998. Even continuous cryogenic hydrogen and argon fibers [6] were recently reported. 1.1 The most important phase is theliquid phase Some short as well as discontinuous fibers can be grown from the vapor phase, some formed from the liquid phase, either a viscous melt or a viscous solution, and others yet are derived from a solid precursor or green fiber. The present book subordinates the discussion of individual orgroups of fibers tothe functional hierarchy of these process concepts. Almost all short fibers, which are derived from the vapor phase, grow by a vapor-liquid-solid (VLS) phase transformation, including single crystal silicon whiskers and carbon nanotubes. Only rarely does the growth of short fibers or whiskers occur by a vapor-solid (VS) phase transformation, and the evidence for this type of phase transformations is often difficult to obtain experimentally. Silicate glass fibers, which command by far the largest sales volume in the market, are derived from the liquid phase, a viscous melt. Ultrapure silica fibers are either derived from a melt, which isdowndrawn from a preform, orthey are dry spun from a viscous solution. While a high melt or solution viscosity seems to be a general prerequisite for fiber formation, it is possible to form glass fibers, such as YAG oraluminate glass fibers, from melts. Carbon fibers, as well as ceramic oxide and carbide fibers, which have a combined sales volume well below that ofglass fibers, are derived from solid precursor orgreen fibers. These precursor fibers are in turn derived from a liquid phase, e.g., from a viscous solution orfrom a viscous polymer by dry or melt spinning, respectively. By virtue of its organization, this book is uniquely able to pay equal attention to the formation, structures and properties of the

4

Chapter 1

functional carbon and ceramic fibers and to those of the nonfunctional green or precursor fibers from which they are derived. Not only is a liquid phase a key process element during the formation of nearly all inorganic fibers, but it must have a solution ormelt viscosity oflog 2.5(or 316) to log 3.0(or 1000) poise atthe forming temperature. Even the viscosities of YAG oraluminate compositions which are only <1 poise above their melt temperature, must be raised to, and stabilized between, log 2.5 and 3.0 poise, before continuous amorphous glass fibers can be melt spun. This commonalty among almost all fiber forming processes further underscores the importance of the liquid phase among the other phases. In exceptional cases, growth of a inorganic whiskers from the vapor phase may occur by a direct vapor-solid phase transformation, and the formation of continuous cryogenic hydrogen, deuterium, and argon fibers from liquefied gases seems to occur by a direct solid-solid phase transformation. In this case, a solid phase is extruded through the tiny spinneret orifices, and not a viscous liquid phase. Nonetheless, the most important phase in the formation of inorganic fibers seems to be the liquid phase. Technical books in the field of inorganic fibers include monographs, which cover single fibers [6], encyclopedic books, which cover individual fibers alphabetically starting with alumina and ending with zirconia fibers [7], and applications related books, which cover fiber reinforced composites by market segment starting with aerospace composites and ending with wood composites [8-9]. This book provides a timely update ofthe rapidly advancing literature inthis field, adds a new perspective by focusing on the phases from which the fibers were derived, and thereby complements other excellent books in the field using either encyclopedic or applications related hierarchies orprinciples oforganization. 1.2 A fiber by any name is still a fiber

Each chapter addresses a self-contained and extensively cross-referenced topic that is supported by the most up-to-date literature. The aspiring materials science student may find the book to be a useful introduction to the entire range of inorganic fibers, and the expert in one field of fiber science and technology may gain up-to-date technical insights from other and rapidly advancing fields offiber science and technology. Importantly, this book addresses the fundamentals for designing the types of ceramic fibers which, according to recent recommendations by the National Research Council, will be needed for the twenty-first century [10]. Most fibers are solid materials having a round cross section and a uniformly cylindrical shape. Fiber diameter, fiber length and fiber composition will almost always adequately describe a given fiber (Table I). In practice, the situation is more complicated. With regard to fiber diameter, cylindrical structures with a diameter >200 IJm are generally known as rods. Fibers with diameters ranging from 100 to 200 IJm are either known as very large diameter structural fibers (e.g., boron/tungsten fibers) or optical fibers (e.g., silica fibers). Fibers with diameters ranging from 1 to 25 IJm are generally known as microfibers, and those with diameters generally ranging from 1to 25 nm are known as nanofibers.

Chapter 1

5

Table I. Inorganic fibers from the vapor phase, the liquid phase and from solid precursor fibers Fiber Compo- Process Precursor Fiber length form nents status phase Rods> 200 um diameter Silicate glass rod endless solid one comm . liquid/melt Silicate glass tube endless hollow two comm . liquid/melt exper . liquid/melt rod short solid one Superconductor Coarse fibers> 100 urn diameter comm . vapor phase filament endless solid two Boron/tungsten waveguide endless solid two comm . liquid/melt Silica - optical fiber endless solid two comm . vapor phase Silicon carbide/C fiber endless solid one exper. liquid/melt Superconductor Fine fibers >1 urn diameter fiber endless solid one comm . liquid/disp. Alumina - cryst. fiber endless solid one comm . liquid/melt Aluminate - glass Aluminum fiber endless solid one exper . invsicid melt fiber short solid one exper . liquefied gas Argon - cryogenic Boron fiber endl ess solid one exper. vapor phase one comm. vapor phase Carbon - structural fiber short solid Carbon - structural fiber short solid one exper . liquid/sol'n Carbon - structural fiber endless solid one exper . vapor phase Carbon - structural fiber endless solid one comm . solid fiber Diamond / tungsten fiber endless solid one exper . vapor phase Fluoride fiber endless solid two exper . liquid/melt fiber endless solid one exper. vapor phase Germanium - s.c. Hafnium carbide whisker short solid one exper. vapor phase ribbon endless solid one comm. liquid/melt Metal alloys Oxide - s. crystal fiber endless solid one comm . liquid/melt Oxide - ceramic fiber endless solid one comm . solid fiber Oxide - glass fiber endless solid one exper. inviscid melt Silica - structural fiber endless solid one comm . liquid/melt Silica - structural fiber endless solid comm . liquid/sol'n one Silica - structural fiber endless solid one comm . solid fiber Silica - structural fiber endless solid one comm . solid fiber Silicate - glass fiber endless solid one comm. liquid/melt hollow fiber endless hollow two comm . liquid/melt Silicate - glass Silicate - glass porous fiber endless porous two exper. liquid/melt whisker one comm . vapor phase Silicon - s. crystal short solid Silicon - s. crystal micropillar short one exper . bulk solid solid Silicon - amorphous fiber endless solid one exper . solid fiber Silicon carbide whisker short solid one exper. vapor phase Silicon carbide fiber endless solid one exper. vapor phase Silicon carbide fiber endless solid one comm. solid fiber Steel - melt spun fiber endless solid one exper. inviscid melt Superconductor fiber endless hollow three exper . liquid/melt YAG - s. crystal fiber endless solid one comm . liquid/melt fiber YAG - polycryst. endless solid one exper. solid fiber YAG - amorphous fiber endless solid one exper . liquid/melt Zirconia - cryst. fiber endless solid one ~.~uid / sol'n Nanofibers >1 nm diameter Boron nitride nanotube short hollow two exper . vapor phase Carbon fullerene pipe endless hollow two exper. vapor phase Carbon nanotube short hollow two exper. vapor phase Carbon fullerene rope endless hollow two exper . vapor phase Carbon nanotubule hollow two short exper . vapor phase Silicon nanowire short solid one exper . vapor phase Silicon quantum wire short one solid exper . vapor phase Silicon carbide nanowhisker short solid one exper. vapor phase Silver/DNA nanowire short two solid exper. liquid/sol'n Composition (alphabetical)

Trival fiber name

Chap . #

2.2.2 3.2.2 7.3.2 3.2.2 4.5.2 8.4.1 4.3.3 4.4.6 4.4.6 3.2.1 2.5.4 2.4.3 3.2.1 9.2.1 2.4.6 4.3.2 3.2.1 2.6.2 4.4.2 4.5.1 8.1.2 4.4.7 6.4.1 6.4.3 6.4.2 6.4.4 6.1.2 6.3.2 6.3.2 2.6.2 2.3.2 3.2.1 2.3.3 3.2.1 10.2.1 4.4.6 6.3.2 4.5.2 8.5.2 4.4.4 8.4.2 2.2.7 2.2.7 2.2.7 2.2.7 2.2.7 2.2.7 2.2.7 2.6.3 2.3.1

6

Chapter 1

Specialty fibers have a non-round, e.g., ribbon, dumbbell, or trilobal cross section, and bicomponent fibers often consist of two concentrically arranged materials, a core of one material or a hollow core, and a sheath or cladding ofa second material. Furthermore, fibers with round and non-round cross-sections, and fibers with bicomponent structures, can either be continuous (practically endless) or discontinuous (short). In general, the latter have high aspect ratios (length divided by diameter ratios >1000). Experimental fibers, which are made by potentially continuous processes, have been counted as potentially continuous fibers in groups of chemically related fibers are known by their generic names, e.g., boron fibers, silicon whiskers or carbon nanotubes. Fiber producers often protect the commercial identity, quality and reliability of their products with tradenames; e.g., Dacron is a registered trademark owned by Du Pont toprotect its polyester fibers. Individual mountain climbers name the peak which they scaled for the first time, and scientists often give trivial names to fibers they synthesized for the first time, for example, fullerene pipes, buckytubes, or quantum wires. Such nicknames reflect the sense of discovery that always prevails in a new and rapidly growing field . In the scientific context of this book, generic and specific compositions are more meaningful than trade names. Therefore, compositional descriptions have been used throughout the book to characterize a given fiber, notably indiscussions which relate structures to properties. Trade names have been used only when absolutely necessary. No attempt has been made to suppress the diversity of trivial names as they appear inthe literature. A fiber byany name is still a fiber. In summary, the book introduces a unified view of advanced inorganic fibers to the aspiring materials science student and attempts to foster cross-fertilization among the experts inthe field .

1.3 Biographic sketches ofthe authors Fred Wallenberger is an expert in the fields of inorganic, polymer-organic and natural fibers. He got his Ph.D. degree from Fordham University, was a Research Fellow at Harvard, and joined the staff of Pioneering Research Laboratory, Du Pont Fibers, where he contributed for over three decades to the commercialization of new fibers through intrapreneurial research, project management and technology transfer. Subsequently, he became a Research Professor at the University of Illinois in Urbana-Champaign and a Visiting Professor at the University ofCalifornia in Davis, and assisted entrepreneurial high technology businesses with organizational advice, technical assistance and license negotiations. He has published over 100 papers, several in Science, and recently joined the staff of the Fiber Glass Research Center, PPG Incorporated as Manager, Advanced Technology. Roger Naslain received his Ph.D. degree at the University of Bordeaux and spent one year at the General Electric Corporate Research Center in Schenectady, NY. Since then he pursued research in composite materials, first as a group leader at CNRS Laboratory of Solid State Chemistry, and then as a manager of the Institute of Composite Materials, a technology transfer center. He is now manager of the Laboratory for Thermostructural Composites located in the Bordeaux area, and professor of materials science at the University of Bordeaux. He has published more than 200 papers, has received fifteen patents and has edited several books inthe field offiber and composites technology.

Chapter 1

7

John MacChesney is a materials scientist with a BA degree from Bowdoin College and a Ph.D. degree from the Pennsylvania State University. He has spent his professional career at Bell Laboratories working on glasses for electronic oroptical use. A pioneer infiber optics, he is credited with the invention of the MCVD process to make fibers, and is a principal in the development of sol-gel silica for fiber use. He has published about 100 papers and an equal number of patents. He is a Fellow of Bell Laboratories, and a member of the National Academy of Engineering. Harold Ackler, Lawrence Livermore National Laboratory, got a MS degree in Materials Science and Engineering at the University of California in Berkeley, and a Ph.D. degree in Ceramics at the Massachusetts Institute of Technology. His research, while at Bell Laboratories, Lucent Technologies, has focused on the processing of optical fiber preforms via sol-gel methods, planar photonic devices, and glasses with non-linear optical properties.

1.4 Acknowledgments Dr. Wallenberger gratefully acknowledges the initial encouragement for writing the book from the late Norman Kreidl, pioneer and teacher; James Nottke, formerly Director, Pioneering Research Laboratory, Du Pont Fibers; and Zhao Jiashiang, Director, Beijing Research Institute for Materials and Technology. He enjoyed the searching discussions with Eugene Givargizov, head of the Crystallography Laboratory, Russian Academy of Sciences in Moscow; Paul Nordine, President, CRI in Evanston IL; Gary Tibbetts, Staff Scientist, General Motors Research in Warren, MI; Robert Feigelson, Stanford University; Austen Angell, Arizona State University; and with Norman Weston, Consultant, Lewes, DE. The completion of the book would not have been possible without the support from Jaap van der Woude, formerly Director, PPG Fiber Glass Research Center, and with that of Norman Weston who reviewed the manuscript and prepared the appendix. Professor Naslain is grateful for valuable advice, data and illustrations from I. Mochida, Kyushu University; A Oberlin, J. B. Donnet and X Bourrat, CNRS; J. L. White, University of California, San Diego; R. J. Diefendorf and D. D. Edie, Clemson University; J. C. Lewis, Union Carbide; J. Economy, University of Illinois; H. Ichikawa, Nippon Carbon; 1. Yamamura, Ube Industries; J. Lipowitz, Dow Corning; K. Okamura, Osaka Prefecture University; R. M. Laine, University ofMichigan; M. D. Sacks, Universityof Florida; H. P. Baldus, Bayer AG; J. DiCarlo, NASA-Lewis; R. Tressler, Penn State University; A. R. Bunsell, Ecole des Mines, Paris; P. Olry, SEP/SNECMA; J. Dunogues , University ofBordeaux and R. Pailler, CNRS. REFERENCES [1} [2J [3J [4} [5} [6J

S.lijima, Carbon nanotubes, Nature, 354, 56(1991). L. C. Venema, J. W. C. Wildoer, J. W. Janssen , S. J. Tans, H. L. J. Temminick Tuinstra, L. P. Kouwenhoven and C. Dekker, Imaging electron wave functions as quantized energy levels in carbon nanotubes, Science, 283, 52-55 ([999). F. T. Wallenberger, Rapidprotoptying directly from the vapor phase, Science, 276, 1274-1275 (1995). F. T. Wallenberger and P. C. Nordine, Potentially continuous single crystal germanium fibers bylaser assisted chemical vapor deposition, in preparation (1998). J. K. Weber, J. J. Felton, B. Cho and P. C. Nordine, Glass fibres of pure and erbium- or neodymium-doped yttria-alumina compositions, Nature, 393, 769-771 (1998). R. Aliaga-Rossel and J. Bayley, A cryogenic fiber maker forcontinuous extrusion, Rev. Sci. Instrum., 69 [6}, 2365-2368 (1998).

8

(7] [8) [9]

Chapter 1

M. S. Dresselhaus, G. Dresselhaus and P. C. Eklund, Science of Fulferenes and Carbon Nanotubes, Academic Press, San Diego, CA (1996). A.Kelly, Editor, Concise Encyclopedia of Composite Materials, Pergamon, London (1994). V. L. Kostikov, Editor, Fibre Science and Technology, Chapman & Hall, London (1995).

[10] P. W.Johnson, Ceramic fibers and coatings, advanced materials for the twenty-first century , Public ation NMAB-494, National Academic Press, Washington, DC (1998) .

SECTION II FIBERS FROM THE VAPOR PHASE F.T. Wallenberger Advanced inorganic fibers fall into two categories: (1) discontinuous or short fibers and (2)continuous fibers or atleast potentially continuous fibers. Chapter 2 deals with short fibers from the vapor phase butalso introduces short liquidand solid-phase derived fibers. There isa better fit forthese fibers here than in achapter oncontinuous fibers.

Contents 2

SHORT FIBERS, WHISKERS AND NANOTUBES 2.1 Advanced vapor phase processes 2.2 Advanced liquid phase processes 2.3 Advanced solid phase processes 2.4 Selected fiber structures and properties 2.5 Selected fiber products and applications

3

CONTINUOUS OR ENDLESS INORGANIC FIBERS 3.1 Continuous vapor phase processes 3.2 Selected structures and properties 3.3 Selected products and applications

CHAPTER 2 SHORT FIBERS, WHISKERS, AND NANOTUBES Fred Wallenberger Short needle shaped, inorganic fibers occur in nature, or can be synthesized by a variety of experimental and commercial processes. If these fibers are filamentary single crystals, they are called whiskers. If however they are polycrystalline or amorphous, they are called short fibers. 2.1 Advanced vapor phase processes The recorded history of short fiber technology starts over two thousand years ago with asbestos fibers and reaches into the future with silicon nano-whiskers and carbon nanotubes. Asbestos is derived from the solid phase, but today, the most important short inorganic fibers are derived from the vapor phase. 2.1.1

Evolution ofa technology

The evolution of modern vapor phase processes starts with metal catalyzed chemical vapor deposition and ends with laser vaporization (see Table I). Most vapor phase processes require metal particle catalysts; some proceed without the addition of metal particles. The growth temperatures range from 100 to 4000·C. The length of silicon nanowires is <10 nm [74]; that ofcarbon nanotubes is<300 urn [76] but they can be potentially endless [81]. Metal catalyzed chemical vapor deposition has become the most versatile and therefore most important whisker growth process. This process and other vapor phase processes facilitate the formation of uniform and reproducible products for demanding applications, where they offer new premium electrical, magnetic, dielectric and near theoretical mechanical properties [1-2]. Three major breakthroughs in process technology have recently been made. These processes facilitate the growth from a liquid phase. They include the formation of (1) InP, InAs and GaAs whiskers [18] from other organic solvents by a solution-liquid-solid phase transformation, (2) short carbon fibers from liqUid pitch melts by centrifuging [19], and (3) silver nanowires by a novel self-assembly process [71]. Micro- and nanopillars (or microand nanocolumns) are anew class ofshort inorganicfibers. They are advanced inorganic fibers derived from the solid phase, e.g., from solid bulk materials such as silicon by exacting methods such as lithography and etching. As fabricated, they represent highly uniform structures and offer highly uniform properties in advanced integrated circuit (IC) and selected non-IC applications.

Chapter 2

12

In summary however, vapor phase processes offer a more effective route to advanced inorganic fibers than solid or liquid phase processes because they facilitate greater control over diameter, length, aspect ratio, and properties of the resulting whiskers, microfibers, and nanotube structures. 2.1 .2 Crystal growth and phase transformations Whiskers are short single crystal fibers. Basal growth occurs by a solid phase migration of atoms, for example that of tin whiskers, which may grow from a tin-plated surface (Figure 1, left). Tip growth is the major mode of growing whiskers from the vapor (Figure 1, center). It occurs at the whisker tip and proceeds by incorporating atoms from the vapor phase into the growing tip rather than into its base, and stops when the activity of a catalyst is exhausted. Side growth (Figure 1, right) is often observed with short carbon fibers. Initially these fibers grow by tip growth. Tip growth is often followed by side growth. Actually, in this mode a secondary overgrowth isrecognized by a thickening of the fibers and leads toa fiber structure that resembles the growth rings ofa tree orthe growth layers of an onion [2). Table I. Growth of whiskers, microfibers and nanotubes from the vapor phase Year 1961 1983 1983 1991 1991 1992 1994 1995 1996 1997 1997 1998 1998 1998

Vapor Phase Processes Metal particle catalyzed CVD Self catalyzed rice hull processes Metal cat. carbothermal reduction Metal catalyzed arc discharge Self-cat. carbothermal reduction Ion bombardment processes Metal particle catalyzed CVI Diamond deposition on fibers Metal catalyzed laser discharge Self-catalyzed arc discharge Self-propagating HT synthesis Laser ablation of metal alloys Silver ion deposit on DNA chain Laser vaporization of carbon

Fiber Semiconductor, ceramic, carbon SiC and SiN microwhiskers Ultrahigh strength SiC whiskers Multishell C and BN nanotubes SiC whiskers, in-situ composites Short carbon/graphite whiskers Short silicon nitride fibers, coils Composite fibers , diamond tubes Single shell carbon nanotubes Single shell carbon nanotubes Single crystal SiAlON whiskers Si, Ge semiconductor nanowires SilverlDNA nanowires Long & endless carbon nanotube

Chapter 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2

Ref. [I) [14) [29) [15) [30) (7) [24) [28) [52) [76) [75) [74] (71) (81)

Year 1987 1995 1998

Liquid Phase Processes LC catalyzed centrifuge process Metal catalysis - organic solvents Cooling of K,CO" TiD , fluxes

Fibers Short graphite fibers, papers, mats Short InP, !nAs, GaAs microfibers Titanate fibers with high lid ratios

Chapter 2.3 2.3 2.3

Ref. [19) [l8) [77)

Year 1998

Solid Phase Processes Microstructures by lithography

Fibers Array s of silicon micropillars

Chapter 2.4

Ref. [56)

(a) Vapor-liquid-solid (VLS) growth Crystal growth [4) requires high activityand either a reversible pathway between a liquid and a solid phase, or a high surface bulk mobility in the solid phase. These conditions facilitate a process by which atoms, ions, or molecules can adopt correct positions, microstructure, and directionality in the crystal lattices. The most common mechanism by which whiskers grow from the vapor phase involves a vapor-liquid-solid phase transformation . Gaseous atoms or species generated by a vapor phase reaction dissolve in a molten metal catalyst particle. They form an alloy and a supersaturated solution. The metal alloy reduces the operating temperature, initiates the growth ofa solid phase, and controls the growth rate (Table II).

13

Chapter 2

- - I

B,",lg~lh

L. .

Figure 1.

_

Tip growth

Side growth

Modes ofwhisker and shortfiber growth (schematicdrawing).

The supersaturated metal droplet that is located on the tip of the growing whisker or short fiber sustains the directional growth of the localized solid phase. This continues to catalyze the reaction as long as an equilibrium is maintained between the diffusion of gaseous atoms or molecules from the vapor phase into the liquid phase and their subsequent extrusion as a solid whisker or short fiber. Growth stops if the catalyst is exhausted or sufficiently contaminated byimpurities imported from the vapor phase. Metal particle catalyzed chemical vapor deposition is the most versatileVLS process (Table II) yielding a wide range of single crystal whiskers and nanowhiskers [1-2] [5-6], short amorphous or polycrystalline fibers [1] [7], and nanotubes [8]. Laser ablation of selected metal alloys is a recent VLS process used for the synthesis of semiconductor nanowire [74]. Metal particle catalyzed carbothermal reduction, another VLS process, yields single crystal whiskers [9-10]. Metal catalyzed arc discharge [11 ], metal particle catalyzed laser ablation [1 2], and metal particle catalyzed plasma arc discharge [13] yield nanotubes by a VLS mechanism.

(b) Vapor-solid (VS) growth A vapor-solid phase transformation was originally believed to govern the growth of short carbon and graphite fibers [1 ] by metal particle catalyzed chemical vapor deposition. A VLS mechanism, however, is now believed to govern carbon fiber growth [7], and a VS mechanism, the growth of SiC whiskers in the self-catalyzed rice hull process [14]. A new process [75], the self-propagating high-temperature synthesis of single crystal SiAION or oxynitride fibers, has recently been added by its authors to the category of solid fibers which are believed to be formed directly from the vapor phase bya VS mechanism. These insights, notably those with regard to carbon nanotubes [1 5-17] and SiAlON fibers, are likely to revive

14

Chapler2

the currently dormant discussion of vapor-solid mechanism, since vapor-to-solid phase transformations are difficult toprove. Table II . Phase Transformations Growth Process (see text)

Metal particle catalyzed chemical vapor deposition

Catal yst Phase

Short Discontinuous Fiber Products

Vapor-liquid-solid (VLS) transformations Liquid metal and/or liquid 0 Si, SiC and other whiskers metal alloy (An-Si, Fe-Fe,C , 0 Short SiC and Si,N. fibers 0 SiC nanowhiskers and balts other) droplets catalyze whisker growth 0 Short graphite and carbon fibers o Short carbon & TiO, microcoils o Carbon nanotubes (high yields)

Metal particle catalyzed chemical vapor infiltration

Liquid metal and/or liquid metal alloy droplets

o Silicon nitride fibers with high strength, o Uniform SiN microcoils

Laser ablation

Liqu id metal alloy nanodroplets

o Single crystal silicon nanowires o Single crystal Ge nanowires

Metal catalyzed carbothermal reduction

Liquid metal and/or alloy droplets

o SiC whiskers from SiO, and C o SiC whiskers/powder from SiN

Self-propagating high temperature synthesis

Molten silicon and aluminum

o Single crystal SiAlON whiskers o Hollow sic SiAlON microtubes

Metal particle catalyzed arc discharge

Liquid (?) metal alloy droplets

o Single/multiple shell carbon and BN nanotubes with high yields

Laser vaporization

Liquid (?) metal alloy droplets

o Short carbon nanotubes o Practically endless C nanotubes

Metal particle catalyzed chemical vapor deposition

Vapor-solid (VS) phase transformations Solid metal particles, vapor 0 Amorphous Si, SiC, Ge fibers 0 Polycrystalline Si, SiC fibers species does not dissolve o Amorphous carbon fibers ?

Chemical vapor deposition on hot filament surfaces

Solid surface of carbon, graphite or metal filaments

o Short diamond/carbon and diamond/metal sheath/core fibers o Diamond microtubes, microcoils

Self catalyzed rice hull whisker processes

Spontaneous whisker formation

o Very short SiC, Si,N. whiskers o Arco, Tateho, other technologies

Self catalyzed carbothermal reduction

Spontaneous whisker formation

o SiC whiskers wlo metal caps o SiC nanowhiskers and batts

Deposition of silver grains on DNA strand

Self-assembly of nanowire

o Nanoelectronic silver wire for future nanochips

Particle catalyzed process- organic solvent LC catalyzed centrifuge/ curing process Lithography & etching

Solution-liquid-solid (SLS) phase transformations Liquid metal droplets 0 Short amorphouslpolycrystalline InP, formed in situ InAs and GaAs fibers Liquid crystal mesop itch particles

o Short graph itic fibers and cohesive graphite fiber batts

Etching of bulk solids (no phase transformation) No catalyst 0 Short silicon micropillars

Chapter 2

15

In summary, a VLS phase transformation will occur when particulate metal catalysts are added tothe reaction, and when the whisker diameters are related tometal particle diameters. If no metal particles are added, a VS phase transformation may occur, but in either case rigorous experimental evidence would be required for a definitive assignment. An earlier book [1) and an informative review article [5) should be consulted for details. 2.1.3 Metal catalyzed chemical vapor deposition A film is deposited in a conventional chemical vapor deposition (CVD) process when the gaseous reactants are presented with a large hot support surface. Supported growth of whiskers occurs also when the gaseous reactants are presented with discrete hot metal catalyst particles located on the surface of a suitable substrate. Unsupported whisker growth occurs when hot metal catalyst particles are freely interspersed with the gaseous reactants in the vapor phase. The most common mechanism for whisker growth is a vapor-liquid-solid transformation, and the most versatile VLS process is a metal particle catalyzed chemical vapor deposition. (a) Reaction chemistry Whiskers or short fibers can be grown by this process from virtually any material that can otherwise be deposited as a film by a conventional chemical vapor deposition process. The best known materials which have been grown by this process are single crystal silicon and gallium arsenide whiskers (2) and silicon nanowires (74) for semiconductor uses [1] [5], and single crystal silicon carbide whiskers [6], hafnium carbide whiskers (22), and silicon carbide nanowhiskers (2) for structural uses. Other binary single crystal whiskers made by this process include InP, InAs, ZnSe for semiconductor uses, and TiC, Ab03, Y203, LaB6, TiB2, BN, and SbN4whiskers for structural uses (1). Single crystal silicon whiskers are an important semiconductor material. Silicon atoms are formed by a vapor phase reaction (Equation 1) above 900·C, and dissolve in molten metal droplets. Gold is the particulate metal of choice but platinum, silver, and nickel particles can also be used (5). The supersaturated silicon solution causes the deposition of solid whiskers onto a given base substrate. Unless specifically controlled, the whisker growth will be irregular (Figure 2), resulting in a forest of whiskers with irregular diameters and lengths (Figure 3). Single crystal silicon carbide whiskers represent an important structural material. The most common metal particle catalyst used in their growth (Equations 2a and 2b) is iron. Silicon monoxide and methane react when exposed to the catalyst at 1400·C. In about 10 hours, the whiskers grow toalength of 10 to 20 mm [6]. Atgrowth temperatures above 1450·C, the iron particle originally embedded in the tip of the growing fiber may sometimes evaporate, leaving behind a silicon-rich silicon carbide tip [9]. Thus [6], in the typical laboratory process, a graphite plate is coated with iron or stainless steel catalyst particles, and highly irregular growth ofsilicon carbide whiskers isobserved (Figure 4). SiCI" (g) + 2 H 2 (g)

Au - Si ) Si (whi sker) + 4 HCI (g)

2Si02 (solid) + C (solid) ~ 2SiO ( gas) + CO2 ( g)

(1)

(2a)

16

Chapter 2

Figure 2. Single crystal whisker tip consisting of a solidified gold catalyst droplet. Courtesy of Dr. E. I. Givargizov, Russian Academy ofSciences. Moscow

Figure 3. Growth of irregular arrays of silicon whiskers. Courtesy of Dr. E. I. Givargizov, Russian Academy of Sciences, Moscow.

Chapler2

CH.J

17

(g)+SiO(g)~SiC(whisk er)+H20(g)+H2

HjCl.J+CH.J +H 2

(gas)

(2b)

Ni Co V . " ) HjC ( whlsker ) + 4 HCI (g ) + H 2 (g )

(3)

Two new processes were recently reported [43-44] potentially opening the door to intensive exploration of silicon carbide whiskers in ceramic, metal and polymer matrix composites. One potentially continuous process [43] is an adaptation of the laboratory batch process [5]; the other [44] uses highly reactive amorphous forms of Si02 and Cto accelerate growth .

Figure 4.Growth ofirregular arrays ofsilicon carbide whiskers. Courtesy ofDr. J. V. Milewski, Los Alamos, NM

Alternatively [42], a gas mixture of methyltrichlorosilane and hydrogen is reacted on a Nicoated graphite and Ni foil substrate. The bulk catalyst activates VLS growth and avoids the need for a catalyst layering step in conventional processes. More recently, hafnium carbide whiskers where synthesized by metal particle catalyzed chemical vapor deposition using a variety of metal particle catalysts including nickel, cobalt and vanadium (Equation 3). The most effective of these was vanadium at 1250°C [22]. In both cases supported but irregular growth was observed. Short amorphous silicon nitride fibers were synthesized by metal particle catalyzed chemical vapor deposition [23] at an operating temperature of 1200°Cand using a gas mixture of SbCI6 + NH 3 + H2 + argon . Irregular growth of straight fibers was obtained along with coiled fibers. The coiled fibers had a constant coil pitch and coil diameter, and could be elastically extended to about 3x their original length [23]. It should be noted that straight and coiled short silicon nitride fibers [24J were also obtained by chemical vapor infiltration of silica with ammonia (see Chapter 2.2.5). Short carbon fibers with diameters of 10 IJm grow by iron particle catalyzed chemical vapor deposition from methane, natural gas [7] [25], benzene [26], or acetylene (23) as shown in

18

Chapter 2

Equations 4 to 6. Initially, tip growth occurs, yielding fibers with a tubular structure and diameters comparable to those of the catalyst particles (:=;15 nm). SUbsequent side growth adds multiple secondary growth rings and increases the diameter [7). The iron particles become encapsulated in the fiber tips after thickening. In addition, short carbon and graphite microsprings can also be grown by metal catalyzed chemical vapor deposition [23). Generally, unsupported and highly irregular growth occurs. Using a nickel plate as well as nickel particle catalysts, carbon microsprings grow at 800·C in the upper part of the apparatus near the gas inlet that supplies a gas mixture of acetylene, hydrogen and argon, while straight fibers grow elsewhere in the reactor. The average coil length increased from 100-150 urn after 30 minutes to 200-350 urn after 60 minutes of reaction time. Short graphite microsprings [23) were obtained by heat treating as-grown carbon coilsat 2800·C. f. .\ Fe or iron pentaca rbotyl C (.s h art fiber) 2H 2 ,gas; f. .} CH .f ,gas; --------'---~~ ) I er) +

C6H6 (gas)

iron carbide

CH == CH (gas)

iron

--~)

) 6 C (short fiber) + 3H 2 (gas) 2 C (fiber or nanotube) + H 2 (gas)

(4) (5) (6)

Finally, carbon nanotubes with diameters of 30 nm were obtained by metal catalyzed chemical vapor deposition using acetylene and nanoparticlulate iron catalysts which had been embedded into a mesoporous silica substrate [8) prior towhisker growth. After harvesting the nanotubes, their tips contained 1% iron (by EDX), suggesting a VLS mechanism, although some iron had evaporated during their growth. While single shell nanotubes have already been obtained by other methods, only multi shell nanotubes were obtained so far by metal particle catalyzed CVD. Single shell nano-tubes with diameters of 1-3 nm will also grow by this method. (b) Controlled whisker growth

The hot metal catalyst particle serves as the focus that facilitates the localized (versus area) deposition of a given vapor species (Figure 2). Its diameter is therefore related to that of the whiskers grown from it. Silicon whiskers grow to:=;100 umin length and have diameters of :=;1 urn. Silicon carbide whiskers, carbon fibers and carbon nanotubes grown by this process can have lengths ranging from <100 nm to >100 mm, and diameters ranging from <30 nm to >10 um. Specifically, the diameter of a given molten metal droplet will determine the diameter of the resulting whisker structure. Irregular growth (Figure 3) occurs when the molten metal catalyst particles do not have the same diameter. The resulting whiskers will have irregular diameters and lengths. With silicon whiskers for example, irregular growth will occur when a thin film of gold is deposited on the growth substrate before the reaction is initiated. Heating the reactor chamber to the reaction temperature will cause the gold film to melt and break up into small gold droplets of irregular diameters. Regular growth (Figure 5) of silicon whiskers occurs when gold droplets of equal size are deposited on the growth substratebefore the reaction isinitiated. This result isobtained when the gold droplets are deposited by lithography or similar means. In this case the catalyst particles have will equal diameters and activities and, as a result, "forests' of individual

Chapter 2

19

whiskers will grow on the support substrate with a precise whisker spacing, and the individual whiskers will have precisely equal diameters and equal lengths after the whisker growth process iscomplete. Commercial silicon whiskers are grown by a regular growth process for demanding semiconductor applications requiring a high degree of whisker-to-whisker uniformity. They have high value-in-use, meaning their product uniformity supports their relatively high cost. Silicon carbide whiskers are made by an irregular growth process for potentially low cost applications, e.g., as a potential reinforcement for ceramic matrix composites. The cost of producing iron catalyst particles with uniform diameters is apparently not cost effective. The same considerations apply to all other metal catalyzed chemical vapor deposition processes, whether they proceed by VLS phase transformations (as most do) ornot.

Figure 5. Growth of regular arrays of silicon whiskers. Courtesy of Dr. E. I. Givargizov, Russian Academy of Sciences, Moscow.

Vapor grown whiskers tend to retain the exhausted metal catalyst particles in their tips after the growth process is completed. This is equally true for supported or unsupported growth and for regular or irregular growth. However in rare cases, especially at very high operating temperatures, the metal particles may have partially or completely evaporated when the growth process has reached its completion . In these cases the evidence for a VLS process will have been lost. (c)

Whisker morphology

Chapter 2

20

The morphology of the product derived from the vapor phase (Table III) depends on concentration and diffusion factors [27], the particular phase transformation and in turn on the melting point of the metal catalyst being used, and on the liquidus temperature of the growth substrate. The highest internal order, that of a single crystal, is obtained when the tip temperature is above the liquidus of the substrate, yields a liquid tip, and proceeds by a vapor-liquid-solid phase transformation. The lowest internal order, that of an amorphous structure, is obtained when the tip temperature is below the glass transition temperature of the substrate. A solid tip yields a vapor solid phase transformation . Between the liquidus and the glass transition temperature ofa substrate, intermediate internal order is that of a polycrystalline fiber. In this case, whisker growth iseither governed by a VLS and/or by a VS phase transformation. Table III. Phase Transformations and Fiber Morphology Fiber tip and metal particle temperature Above the liquidus Liquidus to Tg BelowtheTg

Diffusion vs. catalyst temperature high intermediate low

Supersaturation versus surface concentration low intermediate high

Fiber tip and metal phase Liquid melt Soft solid Solid

Most likely phase transformation

Most likely morphology or fiber structure Vapor-liquid-solid Single crystal VLSNS (1) Polycrystalline Vapor-solid (VS) Amorphous

Specifically, single crystal silicon carbide whiskers grow at 1350-1450°C and single crystal silicon whiskers at 900-1000°C because the former has a higher liquidus and the preferred catalysts have higher melting points (Le., Fe and Fe-C versus Au and Si-Au). Polycrystalline silicon fibers grow between 600 and 900°C, amorphous silicon fibers :s:600°C[1]. Amorphous germanium [1], silicon nitride and titanium carbide fibers [23] grow at :2:400°C. The growth habits ofcarbon fibers are analogous. Graphitic carbon nanotubes grow at>3700°C, graphitic carbon fibers at >2800°C, turbostratic fibers between 1400 and 900°C [7] and amorphous carbon fibers <900°C. (d) Generic whisker properties Silicon whisker, short graphite fiber, and carbon nanotube processes are costly. They are used only for premium end-uses requiring outstanding mechanical properties combined with attractive electrical, magnetic, dielectric and/or conductive properties [1-2]. Filamentary single crystals have strength levels approaching those which can be calculated from the binding forces of adjacent atoms. Their stiffness (modulus) reflects high internal structural order, ultimate chemical purity, and crystal perfection. High strength and stiffness translate into a high elongation-at-break and therefore high mechanical toughness. A large area under the stress strain curve of single crystal fibers yields high work-to-break, high mechanical toughness and therefore high damage resistance. Polycrystalline fibers have moduli approaching those of their single crystal analogs, but their strength, elongation, and mechanical toughness are much lower. 2.1.4 Laser ablation ofwhisker precursor alloys Laser ablation [74] is a new generic method that facilitates the production of nanowhiskers or nanowires from semiconductor materials. Nanowires grown so far include silicon (Si) and

Chapter 2

21

germanium (Ge). The process uses laser ablation of a solid preformed alloy target, e.g., SiogFeo1> which contains the element desired in the product, e.g. , Si, as well as the metal catalyst required for whisker growth, e.g., Fe. Ablation produces a vapor of Si and Fe that quickly condenses into liquid Si-rich nanoclusters that become supersaturated. The Si phase grows and crystallizes as single crystal Si nanowire or nanowhisker by a vapor-liquid-solid growth mechanism. As produced, the nanowire has a concentric sheath/core structure with an amorphous Si02 sheath and a single crystal Si core. The sheath, which may be caused by residual oxygen in the reactor and the iron in the tip of the nanowire, can be removed by etching in hydrofluoric acid. The final bare nanowire shows only Si with traces ofoxygen by EDX analysis [74J. The nanowire length is >1~m, the outer nanowire diameter is 17.1 ± 0.3 nm and the average diameter of the crystalline core ofthe nanowire is7.8 ± 0.6 nm. Sio.9Feo / (solid phase)

laser ablation

9 Si + Fe (vapor phase) ----~ 9 Si + Fe nanoclusters (liquid phase)

crystal growth

)

(7)

Sheath / core SiO ] / Si nanowire with Fe tip (solid) ~ Single crystal silicon nanowire (solid phase)

Except for initially producing a sheath/core Si02/Si nanowhisker, the laser ablation process parallels the metal catalyzed chemical vapor deposition process (Chapter 2.2.3). In this process, the Si that is desired is generated by chemical vapor deposition and dissolved in molten metal droplets, e.g., Au or Fe. The molten alloy droplets, e.g., SiAu, which result in this process sequence, give rise to the growth of single crystal Si micro-whiskers by a similar overall VLS phase transformation. The laser ablation process has been demonstrated so far only for Si and Ge nanowires [74J, but it isclear that it isa new generic tool for growing crystalline nanowires. Thus, it should be possible to make nanowires or nanowhiskers of SiC, GaAs, BbTe3and BN in this way and perhaps, inthe presence ofatomic hydrogen, even diamond nanowires [74). 2.1.5 Hot fiber chemical vapor deposition Boron, silicon carbide, diamond and other materials can be deposited by chemical vapor deposition on the surface of hot wires or hot fibers. If a minimal vapor deposit is applied, the process will modify only the surface of the fiber and produce a coating, while leaving its core functionality unchanged. If, however, a thick vapor deposit is applied, the process will create a new and very large diameter fiber that has the functionality of the sheath and a sacrificial core. The hot fiber (Wire) CVD process has been commercially used for 30 years to produce continuous sheath/core bicomponent boron/tungsten and silicon carbide/carbon fibers. Since they are continuous fibers, they are discussed in Chapter 3.3. More recently, this process was used to produce discontinuous, i.e., short, experimental sheath/core diamond/carbon fibers by depositing a thick diamond sheath on short pieces ofa potentially carbon fiber.

22

Chapter 2

Figure 6.Growth ofdiscontinuous sheath/core bicomponent diamond/carbon fibers. Duration ofdiamond deposition at 1000·C - 12 hours (left) and 72 hours (right). Courtesy of J. M. Ting and M. L. Lake, Applied Sciences, Incorporated, Cedarville, OH

Short diamond/carbon whiskers (Figure 6), the first truly discontinuous sheath/core fibers [28], were made bya two step process. The short vapor grown carbon core fibers were produced by pyrolysis of H2/CH4 mixtures in the presence of iron catalysts [25]. These vapor grown carbon fibers were then ultrasonically polished, and diamond was deposited by a microwave plasma-enhanced chemical vapor deposition technique [28]. 2.1 .6 Chemical vapor infiltration A process that appears to proceed by a metal particle catalyzed chemical vapor infiltration and a vapor-liquid-solid phase transformation [24] was found to yield well-defined short amorphous and polycrystalline silicon nitride fibers reported to have very high strength. These fibers were up to 5 mm long, had smooth surfaces, diameters ranging from <1 to 50 urn, and tensile strengths ranging from 2 to 5 GPa. The process is based on a reaction of ammonia with a mixture of silicon dioxide containing 10% of a metal particle catalyst, such as Fe, Ti, Cu according to the overall Equation 8a, and it involves a high temperature nitridation step that optimally takes place above 1400°C 3 SiO] (short fiber) + 4 NH 3

(g)~

Si3N" (short fiber) +6 Hp (g)

(8a) (8b)

C (short fiber) + TiCI" (g) + 2H]

~

TiC (short fiber) + 4 HCI (g)

(9)

The decomposition of ammonia increases above 850°C, a temperature at which nitridation begins to increase. As a result active hydrogen breaks the Si-O bonds while active nitrogen reacts with unbonded silicon to form silicon nitride fibers (Equation 8b). The process does not only yield straight fibers but also well defined, coiled fibers. The formation of titanium carbide fibers by chemical vapor infiltration with titanium at 1200°C (Equation 9) can be carried out with short vapor grown carbon fibers. The flow rates of TiCI 4

Chapter 2

23

and H2 were fixed at 3.0 and 40.0 ccms, respectively, and a reaction time of 5 hours was required toachieve complete conversion of the carbon fiber totitanium carbide. 2.1 .7 Carbothermal reduction The term carbothermal reduction covers a wide range of processes, including processes which do not require metal particle catalysts, processes which do require the addition ofmetal catalyst particles, chemical mixing processes, and self-propagating high temperature synthesis. (a) Pyrolytic processes Rice hull hydrocarbons supply the carbon source and rice hull ash the silicon source. This is an early carbothermal process (Equ. 10a). At s900·C, coking (Equ. 10b and 10c) removes water and organics and yields a mixture ofsilica and carbon, and ats1700·C, pyrolysis yields silicon carbide whiskers (Equation 10d). The overall yield is 1 percent. Three variants of the rice hull process are known. Compacted rice hulls can be pyrolized at 1600·C without coking [14], thus directly yielding a mixture of graphite and SiC whiskers, 10-200 IJm in length and 0.3-1.5 IJm indiameter [14]. Rice hull

~

SiO 2 + 3 C

(10a)

SiO+C~Si+CO

(10b)

SiO}

(10c)

+C~SiO+CO

Si+C~SiC

(10d)

Cuboid and cycloid niobium monocarbide (NbC) whiskers, 0.1-2.0 IJm in diameter and 5-100 IJm in length, and having a square-shaped tip, were recently synthesized by heating mixtures ofniobium oxide (Nb203) and carbon black at temperatures over 1100·C (38). Silicon carbide nano-whiskers, 20-50 nm in diameter and 2-5 IJm in length, were carbothermally synthesized by reducing ultrafine precipitated silica powders with ultrafine carbon black by microwave heating (29). These processes proceed without addition of metal particle catalysts, and therefore bya VS phase transformation [14] [29] [38]. Silica gel, carbon furnace black and cobalt chloride yield silicon carbide whiskers, or Tokawhiskers [30], ina metal catalyzed process at>1450·C. A process variant [9] yields SiC whiskers >1350·C in a fixed bed percolated by a hydrogen flow. The addition of iron above 1450·C affords submicron whiskers ending with a silicon rich droplet. The iron seems to evaporate and condense below 1450·C leaving behind whiskers with silicon rich tip >1450·C. These processes use the same starting materials as the rice hull processes but they also use a metal particle catalyst. As a result, they are believed to proceed by a VLS phase transformation. (b) Chemical mixing processes Silicon carbide whiskers can also be synthesized by carbothermal reduction of silicon nitride [10]. Silicon nitride decomposes >1300·C, silicon melts at 1410·C, and reacts with graphite. Whisker formation in this process is initiated >1400·C and can be completed between 1550

24

Chapter 2

and 1650·C. With the addition of a metal catalyst, distinct metal droplets were found in the tips of the whiskers [10) suggesting a VLS phase transformation, but none by VS phase transformation without the addition ofparticulate catalysts. SiJN.J (s) >130(fC ) 3 Si(s)+2 N 2 (g) (11a) Si(l)+C(s)

>140(fC

(11b)

)SiC(s)

If this carbothermal process is brought to only partial completion (Equation 11a and 11b), a homogeneous mixture of silicon carbide whiskers and silicon nitride powder [10) is obtained which can be fired directly to yield whisker reinforced ceramics. Silicon carbide reinforced alumina composites and silicon carbide whisker reinforced zirconia composites [31) are also products of the "chemical mixing process". The whisker growth rate in the zirconia process can be accelerated by adding metal particle catalysts such as cobalt chloride, thus potentially facilitating a VLS phase transformation.

(c) Self-propagating high temperature synthesis Very pure single crystal SiAlON whiskers [75) were recently made by an inexpensive selfpropagating high temperature synthesis (SHS), a process that has earlier resulted in SbN4 whiskers [75). While many routes are available for the production of silicon nitride whiskers, this seems to be the first method capable of yielding single crystal oxynitride whiskers. The synthesis is performed in a pressurized water-cooled stainless steel reactor vessel that is 1 meter long and has a capacity of30 liters [75). A homogeneous powder mixture ofsilicon (86 wl.%), alumina (8%), aluminum (1%), silicon nitride (5%) and a trace of pure ammonium fluoride in nitrogen is raised to a pressure of 100 atmospheres, and ignited. Pure single crystal SiAION whiskers with diameters of 2 I.1m are formed in the reaction wave, having a temperature >2000·C (Equations 12, a-d). (12a)

NH .J F ------+ HF + NH 3 nSi+n(a I 2-b I 6)N 2 +(b l 3)NH 3

~(SiNuHh )n

(12b)

AI 203 + Al ------+ 3Ato

(12c)

(3 - z )(SiNuHh)n + z AIO ~ nSi3 _ zAI:0:N.J_: +

(12d)

[an(3 - z) -( 4 - z)]NH J + 112[bn(3- z) - 3an(3 - z) + 3(4 - z))H 2

Some whiskers have a hollow, tubular structure; none have metal droplets at their tips. This and other factors suggest [75) that whisker growth does not proceed by a VLS, but by a VS transformation whereby the ammonium fluoride catalyst plays an important role in the growth process. 2.1 .8 Plasma and related processes Carbon and graphite fibers with diameters of 0.3-3.0 I.1m, and multishell carbon and silicon nitride nanotubes with diameters of 3-20 nrn, have been shown to grow by metal catalyzed chemical vapor deposition. Carbon nanotubes will also grow by arc discharge, carbon ion bombardment and laser discharge processes. In each case, there is the option of adding metal catalysts to the process, thus facilitating a more controlled VLS phase transformation and therefore amore uniform product with higher yields. (a) Arc discharge processes

Chapter 2

25

Carbon nanotubes were discovered in 1991 by a carbon arc discharge method [11] whereby a DC current of 150 Alcm 3 is applied with a voltage set at 20 V inhelium at a pressure of 50 torr [17]. The arc discharge is generated at >3700°Cbetween two carbon rods, 1 mm apart. The positive electrode is consumed and a complex deposit forms at the negative electrode. The outer hard shell of the deposit is removed, and its soft core contains aligned bundles of 10 to 100 multishell nanotubes. The bundles are separated by sonication in alcohol [17]. The individual multishell nanotubes have outer diameters ranging from 2 to 20 nm. Carbon nanoparticles remain present after sonication and are selectively burned away in the presence of oxygen, a step that also consumes a portion of the nanotubes [17]. This basic arc discharge process gave only about a 1-% overall yield ofcarbon nanotubes. Carbon nanotubes are also formed in an arc discharge process, when carbon black [32] or graphite [33] is covered with transition metal nanoparticle catalysts. Multishell boron nitride nanotubes with inner diameters of 1-3 nm and lengths of <200 nm are formed in a plasma arc discharge apparatus [13] similar to that used for carbon nanotube production. The compound anode consisted of a boron nitride sheath and a tungsten core. The cathode was a copper electrode. After arcing, pieces of solidified tungsten were found suggesting that the anode temperature had been at least 3700 oK , the melting point of tungsten. This, and the metal particles embedded in the tip ofeach BN nanotube, suggests a VLS phase transformation. Finally, significant amounts of single shell carbon nanotubes were obtained by adding metal catalysts such as Co, Fe, or Ni [15] [34] to the plasma arc process. The addition of metal catalysts seems to facilitate a controlled process by a VLS phase transformation. For example, the addition of cobalt was achieved with a modified positive carbon rod (electrode) into which a hole had been drilled that was filled with cobalt powder [34].

(b) Laser vaporization and ion bombardment Single shell carbon nanotubes were produced in over 70 percent yields by condensation of a laser-vaporized carbon-nickel-cobalt mixture at 1200°C [12] [81]. No multishell nanotubes were detected in the VLS process. X-ray diffraction and electron microscopy showed that the single shell nanotubes have uniform diameters and self-organize into metallic "ropes" (mats or arrays) of 100-500 nanotubes having a single-rope resistivity of <10 4 ohm-ern at 300 K. The particulate mixed-metal Ni-Co catalyst exists at the live end of the growing nanotube and leaves the end by evaporation. Even more perfect single wall (or shell) carbon nanotubes were reported by a scaled-up version [81] of the laser oven method described above [12]. This process yields a nearly endless, highly tangled raw nanotube material. It is purified in nitric acid and the resulting nanotube ropes, non-woven mats, or bucky paper can be cut into individual intermediate length (100-300 nm) macromolecules for further evaluation [81]. This method, which yields potentially endless nanotubes, will also be discussed in Chapter 3, which deals with continuous orpotentially continuous fibers from the vapor phase. Figure 7 shows a multiwalled carbon nanotube produced by electron beam evaporation [72]. Carbon ion bombardment processes using either electron beam [72] or electrical heating [17] afford multi- and single shell carbon nanotubes by vaporizing carbon in a vacuum. These processes show promise but still need tobe fully developed.

26

Chapter 2

Figure 7.HRTEM of multi-walled carbon nanotube. Courtesy of Dr. E. I. Givargizov, Russian Academy of Sciences, Moscow

2.2 Advanced liquid phase processes

Needle-like crystals can readily grow on the surfaces of slowly cooling melts, but they are of no technological and practical significance. Their yield is low, they are difficult to separate from the solid mass and their variable dimensions afford variable properties. In contrast, three new processes afford new short liquid phase-derived semiconductor fibers and novel silver nanotube wires respectively.

2.2.1 Self-assembly ofsilver nanowires Self-assembly isa process that enables molecules toorganize themselves into tiny nanowires or other nanostructures. In the field of polymers, a large and 50 um long structure was recently obtained by self-assembly from large polymer macromolecules. In the field of inorganic fibers, an equally unique structure was recently created by self-assembly for nanoelectronic components using strands of DNA to assemble tiny silver particles into a nanowire that conducts electricity [71]. It is possible that future circuits can be made of

27

Chapter 2

nanowires and nanotransistors allowing close packing to yield nanochips which are much faster and more complex than today's products.

DNA bridge

9\\\\S\\SP

! ~'----------'

DNA anchors Gold electrodes

Silver ions

Figure a.Self-assembly of a nanoelectronic silver wire (schematic). Redrawn from an article by A. Fisher, Do-ityourself molecules. in Popular Science, p. 20, July 1998.

Specifically, a short strand of DNA was bonded to each oftwo gold electrodes deposited on a glass plate (Figure 8). A long strand of DNA was then allowed to attach itself to each of the short DNA anchor strands to form a continuous nanofiber connection between the gold electrodes. The DNA molecules serve as a support fiber upon which silver ions are then deposited to create a sheath/core silver/DNA bicomponent nanowire. Accordingly, this nanowire has a DNA core and a silver sheath. And although it was found to still have a high resistance to current flow, the nanowire was indeed found to carry current. 2.2.2 Whiskers from organic solvents Until 1995, micrometer scale whiskers of the III-IV semiconductors have been grown only by metal-catalyzed or metal-organic CVD processes at relatively high temperatures, e.g., ;;:500°C. Lack of useful crystallization mechanisms for highly covalent non-molecular solids has so far prevented their growth at lower process temperatures. A recent breakthrough [18]

28

Chapter 2

facilitates the growth of InP, InAs and GaAs whiskers (Table IV) at very low temperatures (111-203·C) from organic solvents such as toluene. The process isbased on a novel reaction using the otherwise well-studied metal-organic CVD process under conditions which support low temperature growth, i.e., catalysisof the reaction by protic hydrocarbon solvents and catalysis of whisker growth by metallic indium flux particles (droplets). The flux particles, in turn were produced in-situ during the reaction as a partial decomposition product oft-Bu3ln (Equation 13). (1 - BU)3M + EH 3

Hydrocarbon solvent (e.g ., toluene) ) M

= In or Ga

E

ME + (1- Bu)H

(13)

= P or As

Amorphous and polycrystalline fibers and near-single crystal whiskers were obtained having diameters of 10-150 nanometers and lengths ofseveral micrometers. Growth ofwhiskers and short fibers by this method is believed to proceed by a solution-liquid-solid (SLS) phase transformation, suggesting that similar synthesis routes may now also become available for other covalent short fibers and perhaps whiskers. Table IV. Low temperature whisker growth (after (18)) Reaction product

Elements M E

InP

In

P

Protic reaction catalyst (10 mol percent XH)

Metallic whisker growth catalyst (flux)

MeOH, PhSH, Et,NH In orPhCOOH InAs In As MeOH or PhSH In GaAs Ga As PhSH GalIn Abbreviations (Table V & Equation 13): t-Bu, ten-butyl; Me. methyl ; Ph, phenyl ; Et, ethyl.

Solution temp., ·C 111-203 203 203

The reaction between tertiary butyl indium and PH3, yielding InP fibers, is catalyzed byliquid indium metal particles, which are produced in-situ as a partial decomposition product oft-butyl indium. Since indium particles were found in the tips of InP fibers, it seems that liquid indium droplets perform the same function in this SLS process as liquid gold or other molten metal droplets ina VLS process. 2.2.3 Whiskers from mesopitch Until 1987, the only route to short carbon fibers was a metal catalyzed chemical vapor deposition. Since then, a novel process has become available [19] that facilitates the growth of discontinuous carbon fibers from mesopitch by a continuous liquid phase centrifuge process. Pitch may be considered to consist of a complex mixture of polycyclic aromatic hydrocarbons. It is a semisolid at room temperature but, depending on the composition, it melts above 100·C. Pitch has two phases, a high melting anisotropic, and a low melting isotropic, phase. The anisotropic phase, called mesopitch, ispreferred for this process. In this process, molten mesopitch is centrifuged "over a lip" at relatively low temperatures (450-525·C). High rotor speeds develop centrifugal forces of up to 15,000 gs. The resulting array or batt of short fibers is stabilized without melting between 250 and 380·C, carbonized between 800 and 1500·C, and graphitized between 1600 and 3000·C. Individual fibers in the randomly disposed batt are 2-12 IJm indiameter and 10 mm in length. Since many individual fibers made by this process have a tip (head or terminal ball) with a diameter that is greater than that of the remainder of the fiber [19], they may have been obtained by a solution-liquidsolid (SLS) phase transformation. Conventional centrifugal spinning ofpitch through confining

29

Chapter 2

spinneret orifices would limit throughput, provide undesirably large diameter fibers, and reduce spinning continuity. The novel centrifuge process [19) that yields short discontinuous carbon fibers may also proceed by a solution-liquid-solid (SLS) mechanism. Many carbon fibers made by this method have a short head after graphitization, i.e., a very short segment having a diameter that is greater than that of the remainder of the fiber. The head may be the tip of the growing fiber, formed by a mesophase nucleation, e.g., by a three-dimensional, high melting liquid crystal polycyclic aromatic hydrocarbon particle [20]. 2.3 Advanced solid phase processes

Attention is drawn to the fact that short fibers can be obtained from solid bulk precursor materials by lithography followed by selective etching. In this field , short fibers are called micropillars ormicrocolumns. 2.3.1 Micropillars by lithography and etching Regular arrays of micropillars fabricated by this route are structurally equivalent to regular arrays of whiskers (Chapter 2.2.2) grown by tip growth from the vapor phase using a regular array of molten metal catalyst droplets. The schematic representation shown in Figure 9 illustrates the steps which are involved inthe fabrication ofa regular array ofmicrocolumns. Functionally, a regular array of silicon micropillars made from the solid phase by a process involving lithography (Figure 10) iscomparable toa regulararray ofsilicon whiskers (Figure 5) prepared from the vapor phase by chemical vapor deposition by a vapor-liquid-solid transformation. It is not surprising therefore to find that some, but not all, applications are identical. For example, in the field of biomedical applications, astrocyte and fibroblast cells have been grown on textured silicon surfaces consisting of silicon whiskers [56) and silicon micropillars [80), respectively. 1

2

3

5

========j

FI

~~~~~~~~~

I i !!!

~~~~~~~~~

I i !!!

connnnnnnn==-:J

Figure 9.Formation of micropillars from a solid precursor (schematic). (1) A 175 nm thickthermal silicon dioxide film is grown ona silicon wafer. (2)A photoresist is applied byspin coating. (3) The oxide film is selectively exposed, and then (4) selectively plasma etched with CHF3. (5)The photoresist isstripped, and (6)the silicon wafer is plasma etched with CI2/BCI3. (7) The residual oxide film is removed with a HF etch, and the silicon micropillars as-formed are ready for use. Courtesy of Dr. A. Perez. Cornell University. Ithaca. NY.

30

Chapter 2

Figure 10. Regular array ofsilicon micropillars etched from a solid precursor. Courtesy ofDr. A. Perez, Dr. S. W. Turner, and Professor H. G. Craighead, Cornell University, Ithaca, NY.

2.4 Selected fiber structures and properties

The most important products, which can be synthesized from the vapor phase, are silicon whiskers and nanowhiskers, silicon carbide whiskers and nanowhiskers, short carbon and graphite fibers, and carbon nanotubes. 2.4.1

Silicon whiskers and nanowhiskers

Growth of whiskers by metal catalyzed chemical vapor deposition, tip growth, and VLS phase transformation has become a powerful tool for the synthesis of commercial products. Gaseous silicon, the vapor phase product, dissolves inthe liquid alloy droplets. There it forms a super-saturated solution first and then catalyzes the growth of whiskers. After the growth stops, solidified hemispherical globules, which were the liquid droplets, are now the whisker tips (Figure 3 and 5). Because of their uniformity, only growth of regular arrays is of interest for commercial uses. The production of a regular array of equal diameter gold particles on the surface of a silicon substrate requires lithographic techniques. A highly ordered arrangement of gold catalyst droplets with equal diameters will yield a regular array of single crystal silicon whiskers with equal diameters, equal spacing and height, and with a whisker diameter that corresponds to the diameter of the gold droplets. Thus, the selection of a desired whisker diameter depends on selecting an appropriate gold droplet diameter. Regular arrays of whiskers with sharp tips and individual whiskers with tip radii ranging from <10 to <20 nm are needed for demanding applications in vacuum microelectronics, i.e., field emission devices. To achieve the required tip radii, silicon whiskers are sharpened by successive chemical etching [35J inHF (Figure 11). The gold caps separate, leaving behind a regular array ofsharpened whiskers (Figure 12).

Chapter 2

31

JlYlJlh l L - -

Figure 11 . Sharpening of a single-crystal silicon whisker tip (schematic drawing). This illustration shows, from left toright, the effect of progressive etching inHF.

Figure 12. Field emitting structures and point sources forelectron microscope guns. Regular array of sharpened silicon whisker. Courtesy ofDr. E. I. Givargizov, Russian Academy of Sciences, Moscow.

I

32

Chapter 2

If one mechanically removes all whiskers from a regulararray of sharpened whiskers except one, a structure is formed that can be used as a point source of electrons for electron microscope guns. Diamond coatings can be applied to silicon whiskers, and sharpening by etching yields sharpened diamond coated silicon whiskers (Figure 13) with tip radii less than 20 nm for scanning tunneling microscopy (STM) tips. Additional sharpening gives ultrasharp tips with nanometer-scale radii of curvature <10 nm, which can be used as field-emitting structures, i.e., "cold cathodes" in vacuum microelectronics. An ultrasharp tip (Figure 14) is one to several atoms wide at its apex, and useful as a source ofcoherently-emitted electrons for holographic studies [35] [37]. Cells with extracting electrodes ("Spindt triodes") can also be made by this technique [36]. By controlling growth parameters, whiskers with a special shape can be prepared. Regular arrays of whiskers with conical tips are obtained when the silicon supply from the chemical reaction is stopped. The shaped whisker shown in Figure 15 has a large (2 IJm) base diameter and the same sharp tip as that shown inFigure 14. Cone shaped whiskers combine high spatial resolution with high rigidity. The sharp tip ensures high spatial resolution of the device in both vertical and horizontal dimensions. The relatively thick base (Figure 15, right) gives sufficient mechanical stability against vibrations (an advantage that is crucial in probe devices). Whiskers with such a base are useful for scanning probe devices, scanning tunneling microscopes, atomic force microscopes and scanning probe metrological devices [37] [39].

Figure 13 Sharpened, diamond-coated silicon whisker. Product bulletin, Containerless Research Inc., Evanston, IL., with permission from Dr. P. C. Nordine, President.

33

Chapter 2

I~ 1 1J

Figure 14. HRTEM of uUra-sharp single-crystal silicon whisker tip. Courtesy of Dr. E. I. Givargizov, Russian Academy ofSciences, Moscow.

Figure 15. A cone-shaped, ultrasharp silicon whisker. Courtesy of Dr. E. I. Givargizov, Russian Academy of Sciences, Moscow.

34

Chapter 2

A method combining laser ablation cluster formation and vapor-liquid-solid (VLS) growth was recently developed for the synthesis of single crystal semiconductor silicon and germanium nanowhiskers (74). Specifically, laser ablation was used to prepare clusters of molten metal catalyst particles with a nanometer diameter. The droplet diameter defines the diameter of the resulting nanowhiskers. Bulk quantities of uniform silicon and germanium nanowhiskers with diameters from 6 to 20 and from 3to 9 nanometers, respectively, and lengths from 10 to 300 nanometers were obtained. 2.4.2 Silicon carbide whiskers and nanowhiskers Silicon carbide whiskers have been made by metal catalyzed chemical vapor deposition, a VLS process (Chapter 2.1.4). Silicon carbide whiskers have also been made by variants of the carbothermal reduction of silicon (or a silicon precursor) with carbon, either by the chemical mixing orby the rice hull process (Chapter 2.1.7). VLS silicon carbide whiskers with diameters ranging from <3 to 11 IJm, and lengths ranging from 5IJm to<10 cm are readily obtained by metal catalyzed chemical vapor deposition. This process facilitates an exacting control over whisker shape and dimension in a batch or a continuous process. The diameter of the catalyst determines that of the whisker. Less welldefined whiskers have been obtained by vapor deposition, chemical mixing and carbothermal processes with diameters ranging from <3 IJm to>30 nm.

Short VLS or VS whiskers are obtained with diameters ranging from 0.6 to <1 .0 IJm and lengths ranging from 10 to <100 IJm by chemical vapor deposition, chemical mixing and carbothermal processes. Whiskers derived by the rice hull process reflect the most irregular growth habitat. Two variants have been observed, whiskers without inclusions in any part of their structure, and whiskers with nanoparticulate inclusions inthe core oftheir structure. The sheath of these sheath/core whiskers is free of inclusions. Only their core contains nanoparticulate inclusions, such as Si-O-C phases and nanocrystalline Ca. Mn, and Fe impurities. Tiny VLS nanowhiskers can be obtained by chemical vapor deposition with less than 111 OOth the diameter of ordinary VLS-CVD silicon carbide whiskers by proper selection of the metal catalyst and process conditions. Indeed, these "cobweb" (2) or nanowhiskers have been made with diameters of <20 nm. To the naked eye these whiskers look like a blue cloud created by light scattering similar to that which causes the sky to appear blue. VLS and VS wh iskers grown by chemical vapor deposition or carbothermal reduction respectively have near theoretical strength (::::16 GPa) but VLS whiskers are stiffer than VS whiskers derived from rice hulls (580 vs. 481 GPa). Both VLS and VS whiskers have up to 10x the strength and more than 2x the modulus of Nicalon, a continuous commercial silicon (oxy) carbide fiber, having an amorphous structure with a minor nanocrystalline phase. Higher strength reflects higher structural uniformity and chemical purity, and higher stiffness reflects higher internal structural order. A bibliography of the mechanical characteristics of silicon carbide (45) isavailable.

2.4.3 Short graphite fibers Carbon fibers can be made in batch processes yielding experimental quantities, or in continuous processes yielding commercial quantities. So far, no details have been published

35

Chapler2

about carbon fibers from mesopitch [19], but extensive literature is available about the structures and properties of carbon fibers derived from the vapor phase. Thus, only vapor grown carbon fibers will be discussed inthe following discussion. Supported carbon fibers are obtained in a batch process. The metal particles are deposited on the surface of a nesting substrate before the reaction is initiated and the grown fibers are harvested after the reaction is complete. The hydrocarbon reactants are natural gas [7] or benzene [49-50]. The metal catalyst particles having a diameter of ~ 1 5 nm include iron and iron-nickel powder or magnetite. In the first 30 seconds, carbon fibers grow to an average length of 1 cm by tip growth or lengthening and have diameters equal to those of the catalyst particles (~15 nm). In the next two hours side growth orthickening takes place that increases the fiber diameter -10 IJm [7]. Partially graphitized carbon fibers grow at 1000-1130°C, full graphitization occurs at 2600-3000°C. Unsupported carbon fibers are formed in a continuous process [47-48] where the metal catalyst particles are continuously mixed into the flow of the feed gases, and where fibers carrying a metal catalyst particle in their tips are continuously removed. A specific process [51] that is currently under commercial development uses iron pentacarbonyl as a catalyst and hydrocarbons such as methane, natural gas, or others which can be derived from coal, recyclate and discarded rubber tires. These fibers are only several IJm long and have much lower diameters (0.1 to 0.2 IJm). Apparently, less time is available for side growth (or thickening) ina continuous process. Tip growth occurs at the location of the metal particle catalyst by a mechanism that is not incompatible with a VLS phase transformation. Side growth may involve a VS phase transformation. The fiber cross section, before thickening, reflects that of tubes. After thickening, it resembles the growth rings of a tree [7]. The following factors exert a major effect over the resulting product properties in a metal catalyzed growth of carbon or graphite fibers by chemical vapor deposition: (1) the structure of the initial carbon tube formed by lengthening or tip growth, (2) the structure of the final partially graphitized carbon fiber obtained by thickening or side growth , and (3) the degree of graphitization of the final thickened fiber atelevated temperatures. Table V. Mechanical properties of vapor grown carbon fibers (After (5)) Type of fiber

Type of Process

Length

Short fibers (whiskers) Carbon (32) Graphite (32) Carbon [26)

batch , I 130°C batch, 2600 °C con tinuou s

1000 1000 <10

Continuous fibers Commercial T-3oo Commercial T-500

PAN-based PAN-based

continuous continuous

(urn)

Diameter (urn)

Densi\)' (g/cm)

Modulus GPa

Strength Mpa

7-10

230 360 210

2.20 3.00

0.1-0.2

1.80 2.00 2.00

7-10 7-10

1.76 1.79

228 241

3.45 3.65

In summary, fibers with a length of 1 cm and a diameter of 10 IJm are best grown in a batch process on an inert support substrate carrying particulate metal catalysts. Fibers with a length of 10 IJm and a diameter of 0.1 IJm are best grown in a continuous process on catalyst

36

Chapter 2

particles floating in the vapor phase [7] [51]. Table V compares the mechanical properties of experimental discontinuous vapor grown carbon fibers with those of continuous commercial PAN based carbon fibers. The minimum tensile strength, which increases with decreasing fiber diameter, is at least 2.2 GPa for a 7-10 IJm diameter carbon fiber thickened at 1130°C, and its minimum stiffness (modulus) is 230 GPa [25]. A partially graphitized fiber of this kind can be fully graphitized by heat treatment above 2600°C. The resulting graphite fibers had minimum strength levels of 3.0 GPa and minimum moduli of 360 GPa [25]. Short carbon fibers grown in the continuous process [51] are slightly denser than those grown in the batch process (2.0vs. 1.8 g/cm 3) and have a slightly lower modulus (210 vs. 237 GPa). Finally, the oxidation resistance of vapor grown graphite fibers is superior to that of PAN based carbon fibers. The thermal and electrical conductivities approach those ofcrystalline graphite [7]. 2.4.4 Carbon nanotubes The technology ofgrowing carbon nanotubes from the vapor phase dates back to 1991 when they were first found [11] in arc discharge experiments. Nanotubes can be obtained by chemical vapor deposition, laser evaporation, arc discharge, and carbon ion bombardment (62). Addition of particulate metal catalysts creates a more controlled growth habitat and helps growth ofwhiskers with greatly enhanced dimensional uniformity. (a) Structures

Nanotubes consist ofone ormore seamless cylindrical structures orshells, each composed of a single tubular graphite sheet (Figure 16). A nanotube is a cylindrical extension ofthe sheetlike member of the family of polycyclic aromatics that includes benzene, anthracene, pyrene and coronene. Its tip (one end) is either open orcarries a metal catalyst particle [12), and its base (other end) iseither attached to a support substrate [8] orotherwise closed by a polygon cap, hemibuckyball or graphitic hemidome [16). In some cases both ends are closed and these nanotubes resemble elongated buckyballs [11]. Early specimens were obtained as multishell structures [17). Single shell structures (Figure 16) were obtained in large amounts and high yields [52), thus aiding commercial development. Three specific nanotube structures [81-82) were identified, depending upon the alignment of individual benzene rings with respect to their main axis (Figure 16). The relationship between structures, properties and applications will be discussed in subparagraph (b). The pathways by which nanotubes grow are the same as those by which all other whiskers and fibers grow from the vapor phase. Without added metal particle catalysts, growth is seeded by trace impurities serving as nucleating sites, and proceeds by a VS transformation [16). When metal particles are added, their diameter translates into that of the resulting nanotubes, and growth proceeds by VLS transformation. The catalyst particles are always located at the tips of the growing nanotubes. The ends of the growing nanotubes are either supported by a common substrate or they are unsupported, i.e., floating freely in the gas phase. Supported growth produces an array of nanotubes perpendicular to the common base. Unsupported growth produces nanotubes with closed ends.

Chapter 2

37

Except for the difference indiameters, the structure of a single shell nanotube resembles that of a carbon fiber [25) grown by tip growth (lengthening). However, the structure of a multishell carbon fiber differs from that of a multishell carbon nanotube. The formation of a multishell carbon fiber by side growth (thickening) requires up to two hours to complete, while that of a multishell nanotube occurs too fast to facilitate side growth (thickening). It is likely that individual nanotubes comprising a concentric multishell structure grow simultaneously by VLS from the common solvated interface between the molten metal droplet and the growing multitubular solid phase. The fate of metal catalyst particles during the growth of nanotubes parallels that observed for the metal catalyst particles during the growth of single crystal whiskers. This commonalty suggests that vapor-liquid-solid phase transformations are involved in the majority ofcases. For example, boron nitride nanotubes have cobalt particles in their tips [13) as compared to silicon whiskers having gold caps in their tips [1). Some carbon nanotubes have residual trace amounts of iron in their tips [8]. Since silicon carbide whiskers may also have iron-rich tips when grown at low temperatures, or even silicon-rich tips when grown at high temperatures [9), it is possible that partial vaporization occurs of iron that was originally in both cases located in the fiber tip. Metal particles may not only vaporize but may actually break off from nanotube tips, and evidence of solidified metal droplets has been found in a recent arc discharge experiment [13). If no metal particle catalysts are added, buckyballs and elongated buckyballs may grow, ie ., nanotubes which are closed on one orboth ends [17]. (b) Properties

Carbon nanotubes have novel mechanical and electronic properties which will eventually support their commercial uses as high strength fibers, SUbmicroscopic test tubes and new semiconductor materials. The mechanical properties ofcarbon nanotubes are governed bya high strength to weight ratio. Although actual experimental values for strength and modulus are not yet available, both calculations and experimentation suggest that they have neartheoretical strength. With regard to their stiffness, nanotubes were found to buckle and deform under a load but recover fully from deformation without damage [52], thus taking advantage ofthe in-plane rigidity and strength ofgraphite sheets. Table VI. Stiffne ss carbon nanotubes Inorganic fiber Ca aluminate fiber Pure silica fiber HM glass fiber Sapphire fiber UHM carbon fiber MW carbon nanotube SW carbon nanotube

Process route from inviscid melt by preform drawing by melt spinning by pedestal growth from acrylic fiber by several processes by several processe s

Modulus

370Pa 690Pa 1300Pa 4100Pa 800 OPa 1,800 OPa 3,700 OPa

Chapter 4

6 6

4 9 2 2

38

Ghapter2

The theoretical Young's modulus has been calculated for fullerene nanotubes aligned in the direction of the tubule (Table VI). Results show that nanotubes composed entirely of singlewalled tubules have a high modulus that increases as the tubule radii decrease and as the distance between close-packed tubules in the fiber gets shorter (78). Specifically, direct experimental measurements of the moduli of single, multiwalled carbon nanotubes gave an average value of 1.8 TPa (T=tera). The tubule with the smallest inner diameter was considerably stiffer, having a modulus of 3.7 TPa (79). Theoretical modulus values were calculated to range from 5.2-5.5TPa (78) The capillarity of carbon nanotubes offers two key properties, low chemical reactivity and superior burst strength. The hollow core of a carbon nanotube has been known to suck in water despite the hydrophobic nature ofgraphite, ormolten metal such as lead (17) [52) and it has recently been filled, also by capillary forces, with molten silver nitrate [53). Only nanotubes with a minimum capillary diameter of 4 nm were required, and found to be chemically less reactive than graphite. This property has been illustrated by monitoring the decomposition of silver nitrate within nanotubes in-situ in an electron microscope. In the nanospace available, chains of silver nanobeads were produced, separated by gas pockets developing gas pressures ofup to1300 bar atroom temperature.

1 -_····i

----.

Cu rrent

now -

Conductor

Semiconductor

Diode

-

-

-

Currenlflow

- --

;

I

__ ____J

Figure 16. Schematic representation ofa conducting carbon nanotube (top, left). Scanning tunneling microscope (STM) image of a conducting carbon nanotube (top, right). Schematic representation of a semiconducting carbon nanotube (center, left). STM image of a semiconducting carbon nanotube (center, right). Schematic representation of a carbon nanotube diode (bottom). The schematic representations were redrawn from an article by M. W. Browne, A microscopic revolution - nanotubes expected to replace silicon devices, in Intemational Herald Tribune, page 10, February 19, 1998. The scanning tunneling microscope images were supplied by, and are reproduced with permission, from Dr. Gees Dekker, Department of Applied Physics, Delft University of Technology, 1 Lorenlzweg, 2628 GJ Delft. Netherlands, [email protected]. See also reference [70J.

39

Chapter 2

Different nanotube structures (Figure 16) offer different properties. (1) A carbon nanotube acts like an electrical conductor, ora metal, if the alignment ofadjacent, fused benzene rings are parallel to the main axis of the molecule. (2) A carbon nanotube acts like an semiconductor, i.e., conducts electrical current only after reaching a certain threshold, if adjacent fused benzene rings are not parallel to the main axis of the molecule but are aligned ata helical angle ortwist to the main axis. (3) A nanotube acts like a diode when a conductor and a semiconductor nanotube are joined into a single molecule, whereby the junction permits electrical current toflow only inone direction. The electronic properties ofcarbon nanotubes are therefore a function of the diameter, helicity and kinks which can be observed in their cylindrical structure, and their electronic properties vary in a periodic way between those of a metal and those of a semiconductor. Accordingly, of all the possible tubes one could make, one third would be either metallic or narrow band semiconductors, and two thirds would be moderate gap semiconductors [17]. Metallic carbon nanotubes have recently been achieved invery high yields [12], and field emission currents of 0.1 microampere at80 volts have been achieved for individually mounted nanotubes [54].

2.5 Selected fiber products and applications A discussion of applications of whiskers, short fibers, microtubes and nanotubes must start with a discussion of asbestos fibers. The blanket ban in the United States on all forms of asbestos was lifted by the Environmental Protection Agency in 1991 , but the legal climate is still complex. This holds true not only for asbestos and its composites [3] but also for all other needle-like fibers such as silicon carbide whiskers and its composites when the whisker diameters are as low as ~3 IJm [57-58]. Pertinent governmental regulations [3] [57-58] need tobe considered inthe large-scale production, use ordisposal ofshort needle-like fibers. 2.5.1

Silicon whiskers and nanowhiskers

VLS technology emerges as a powerful enabling tool for the synthesis of silicon whisker structures with a high value in use, as already described, in micro electronics, micromachining, sensors, fiber optics, semiconductor micro-p-n-junctions and nanometric quasi-one-dimensional filaments (Table VII). As our knowledge of fundamental properties of solids increases, it becomes possible to study bodies whose sizes approach atomic dimensions. While no applications have so far been identified for silicon and/or germanium nanowhiskers [74], several have been found for sharpened silicon whiskers, terminating essentially in a single atom tip (Figure 15) Table VII. Potential uses of single crystal silicon whiskers Field emission research and applications Diamond coated scanning tunneling microscopy Micro-indentation and micro-tools Dense whisker arrays as high absorption light traps Dense whisker arrays as high emissivity light sources Microsubstrates and array substrates Product Bulletin. Containerless Research Inc. Evanston IL, 60201

40

Chapler2

Thus, ultrasharp tips with nanometer-scale radii of curvature (10 to 20 nm) are used as fieldemitting structures, Le., "cold cathodes" in vacuum microelectronics. They also used as a point source ofelectrons for electron microscope guns. An ultrasharp tip (Figure 14) is one to several atoms wide at its apex, and is used as a source of coherently emitted electrons for holographic studies [35] [37]. Cells with extracting electrodes ("Spindt triodes") are also made by this technique [361. Shaped whiskers (Figure 15) have a large, -2 IJm, base diameter and an apex of a few atoms. They combine high spatial resolution with high rigidity important for such devices in order to avoid otherwise inherent vibrations. Shaped whiskers are useful for scanning probe devices such as scanning tunneling microscopes, atomic force microscopes and scanning probe metrological devices [37] [39]. Their very sharp end ensures high spatial resolution of the device in both vertical and horizontal dimensions. Solid surfaces covered by arrays (forests) of whiskers are black bodies because of multiple reflections of the photons in the labyrinth maze structure [40-41]. A spectrum selectivity related to topical properties of the material is inherent in arrays of semiconductor whiskers. The transmitted part is characteristic of silicon absorption starting at a wavelength of 1.1 IJm. The reflected part changes montonically. The calculated absorption was 93.0% at 1.00 IJm, 86.8% at 1.34 IJm, 84.5% at1.55 IJm, and 79.8at2.00 IJm [40]. 2.5.2 Silicon carbide whiskers and nanowhiskers With regard to applications, grown silicon carbide whiskers fall into two categories. Those having diameters >3 IJm have not had widespread use, but are not subject to stringent health restrictions as to their commercial and manufacture use. Those having diameters <3 IJm have found limited niche markets and are subject to restrictions itemized in ASTM standards [57-58]. This category includes both VLSNS carbothermal microwhiskers [10] and VS rice hull whiskers [14] having diameters <1 IJm, and VLS nanowhiskers [2] having diameters <20 nm. SiC whisker reinforced composites are expected to find applications in the field of aerospace, automobile, heat engine, plate armor, and energy industries [59-60]. Recent advances further enhance their commercial potential in metal matrix composites such as aluminum, nickel, and copper; ceramic matrix composites, such as alumina, zirconia and silicon nitride; and glass ceramic matrix composites such as lithium aluminosilicate. Silicon carbide whiskers increase strength, reduce crack propagation, and add structural reliability in ceramic matrix composites. Structural applications include cutting tool inserts, wear parts, and heat engine parts. They increase strength and stiffness of a metal, and support the design of metal matrix composites with thinner cross sections than those of the metal parts they replace, but with equal properties in applications such as turbine blades, boilers and reactors. Finally, a thinlayer, sheet, or batt ofsilicon carbide nanowhiskers [2] has been deposited on a suitable support substrate which is "peeled off' and coated with a plastic material. The initial target is reverse osmosis (RO) membranes having a thickness of <300 nm, reinforced with a thin layer of silicon carbide nanowhiskers having individual diameters of <20 nm. Small amounts of nanowhiskers are expected to yield a large increase in strength relative to that of the unreinforced membrane.

Chapter 2

41

2.5.3 Short carbon and diamond fibers The technology ofgrowing short carbon fibers from the vapor phase dates back to 1889 when a patent [46] issued claiming their growth by using iron catalyzed pyrolysis of methane. About a hundred years later a continuous process has become commercially available [47-48] potentially opening the door tocommercial development [51). (a) Short carbon fiber composites

Short vapor grown carbon fibers are readily incorporated into molding compounds, thermoplastics, thermosets and elastomers by injection and compression molding. A sufficiently low price would facilitate their use in low cost, high volume applications such as lightweight automotive sheet molding compounds and fiber reinforced cements [51]. Because these fibers are discontinuous they are not really expected [25) to compete with continuous carbon fibers in demanding structural applications, such as airframes, where both low cost and high performance are important factors. Their superior thermophysical properties after a heat treatment could make them a top candidate incarbon and metal matrix composites where effective electromagnetic shielding is needed [61]. Electromagnetic shielding is needed in some packaging components where these carbon fibers offer the highest thermal conductivity and the lowest electrical resistance. A greater than 50% increase in thermal conductivity was achieved by reinforcing neat aluminum with an 18% fiber volume fraction . Carbon/carbon composites with a 70% fiber volume fraction have twice the conductivity ofcopper. In summary, the highly graphitic nature of these carbon fibers affords aluminum and carbon matrix composites with values of thermal conductivity not previously achieved with any other carbon fiber. They represent a new material for thermal management in applications such as packaging for high power, high density electronicdevices [61]. (b) Diamond/carbon fiber composites

The key properties in thermal management are the thermal conductivity and the coefficient of thermal expansion ofpolymer, metal, and ceramic matrixcomposites. Highly graphiticcarbon whiskers offer major improvements in the thermal conductivity and density of a heat sink, but their thermal conductivity is highly anisotropic and their thermal expansion does not match that of circuit material such as silicon or gallium arsenide. Diamond fibers far exceed the thermal conductivity ofgraphitic carbon fibers. Diamond fibers made by depositing diamond on the surface of short graphitic carbon whiskers by chemical vapor deposition [281 was described in Chapter 2.1.4. They may aid the fabrication of a low density, thermally hyperconductive heat sink material. These bicomponent diamond carbon fibers have high thermal conductivity coupled with a positive axial coefficient of thermal expansion, i.e., properties which can be tailored to match those of silicon orgallium arsenide. They may represent the next generation material beyond graphitic carbon fibers for thermal management in packaging for high power, high density electronic devices [28).

42

Chapter 2

2.5.4 Carbon nanotubes Nanotubes were discovered in 1991, and the first patents began to issue in 1995 [65-68] pointing the way to several emerging commercial nanotube technologies [52]. Depending on their twist (see Figure 16), they may serve as conductors, semiconductors, ordiodes [70] [73], thus potentially ushering in revolutionary changes with regard to the design and use of electronic devices. Other electronic devices not based on silicon technology are needed to improve the speed and power ofcomputers. One advantage of nanotubes is their small size, another a much faster heat transfer. Carbon nanotubes are longer than 1 nm, while off-the-shelf silicon transistors are up to 1 IJm long. Because of their electrical properties, ropes of single shell carbon nanotubes may become a commercial single molecule transistor [64], both on the basis of performance and dimensions. When dissimilar nanotube molecules are joined end-to-end, the junction between them may function as a diode, commonly used to convert alternating current into direct current. Different junction types result by inserting defects into such junctions [70] [73]. A molecular recording and reproducing method has recently been described [68] for binary coded information using endohedrally doped cage-like elements, especially fullerenes and nanotubes, as storage elements. By applying a probing tip to the molecule for the read/write process, an enhanced storage density isachieved [68]. Carbon nanotubes may also become the pinning material of choice in high Tc superconductors [17] [52]. Another potential application istheir use as submicroscopic test tubes [53].

Nanotube film

-

0 0 0 0

+

-

0 0

..

0

0 0 0

0

0

0 0

0

0

0

0

0

0 0

0 0

0

0

0

0

0

0

0 0

0

0

I----

Insulator

0

0 0

Metal

0

0

Metal

I-

+

............ '0

0

0

0

:::::::::::::: <, Nanotubes ....... . . . . . ,

,

Phospho r screen

Figure 17.Carbon nanotube image display. Redrawn from a paper by R. F. Service, Nanotubes show image display talent, inScience, 270, 1119 (1995).

Chapter 2

43

In addition, the use of nanotube reinforced metal, ceramic, glass and a variety of polymer matrix composites may offer high product value [52). An array of thousands of aligned carbon nanotubes (Figure 17) was incorporated into an electrically conducting polymer matrix (63). The resulting composite had superior field emission properties and may become the basis for a new generation of atomic field emission wires [54) and light weight electron emitters functioning like the bulky cathode ray tubes in television and desk top computer displays (63). Finally, entirely new opportunities may open up for carbon, boron nitride and molybdenum sulfide nanotubes [69]. Theory that predicts that electrons flow ballistically through carbon nanotubes and that the conductance (the inverse of the resistance) is quantized, has recently been confirmed (82). Quantized conductance results from electronic wave guide properties of fine wires and constrictions. Electronic transport is ballistic when the length of the conductor is smaller than the electronic free path. In summary, both quantization and ballistic properties (82) have been independently demonstrated for multiwalled carbon nanotubes (MWNTs) consisting of 15 layers and having lengths of 1 to 10 IJm, outer diameters ranging from 5 to 25 nm and inner cavities ranging from 1to4 nm. These nanotubes were embedded in larger fibers (see Figure 17) inside the shelled deposit of the arc which was nominally 50 IJm in diameter and 1 mm. Several individual nanotubes typically were found to protrude from this fiber, and the longest of these, having a diameter of 14 nm and a length of 2.2 IJm, was used to establish contact with the liquid metal surface. It is therefore possible that nanoscopic graphitic structures will eventually be used as electronic elements. REFERENCES [1] [2] [3] [4] [5]

[61 [71 [8] [9] [10] [11J [12J [13] [14] [15J

E. I. Givargizov, Growth of whiskers from the vapor phase, in Highly Anisotropic CIYsta/s, D. Reidel Publishing Company, Dortrecht, 70-229, (1987). J. V. Milewski, Whiskers, in Concise Encyclopedia of Composite Materials, A. Kelly, (ed.), Pergamon, ElsevierScience, 281-284 (1994). J. E. Alleman and B.T. Mossman, Asbestos revisited, Scientific American, [7]70-75 (1997). A.W. Ware, Crystal growth, Principles and Progress, Chapter 1,Plenum Press, NY (1987). R. S.Wagner and W. S. Ellis, Vapor-liquid-solid mechanism ofsingle crystal growth, Appl. Phys. Lett., 4, 8990(1964). J. J. Petrovic and G. F. Hurley, Vapor-liquid-solid (VLS) SiC whiskers, in Fiber Reinforced Ceramic Composites, K. S. Mazdiayasni, (ed.), Noyes Publications, Park Ridge, New Jersey, 93-121 (1990). G. G. Tibbetts, Vapor-qrown carbon fibers, in Carbon Fibers, Filaments and Composites, J. L. Figueriredo et al. , (eds.), Kluwer Academic Publishers, Dortrechl, 73-93 (1990). W. Z. Li, S. S. Xie, L. X. Qian, B. H. Chang, B. S. Zou, W. Y. Zhou, R. A. Zhao and G. Wang, Large-scale synthesis ofaligned carbon nanotubes, Science, 274, 1701 -1703 (1996). T. Bellamonte, L. Bonnetain and J. L. Ginoux, Synthesis of silicon carbide whiskers using the vapour-liquidsolid mechanism ina silicon rich droplet, Journal ofMaterials Science, 31 [9],2367-2371 (1997). H. Wang, Y. Berta and G. S. Fishman, Microstructure of silicon carbide whiskers synthesized by carbothermal reduction ofsilicon nitride, J. Am. Ceram. Soc., 75, 0, 1080-1084 (1992). S. Iijima, Nature, Carbon nanotubes, 354, 56(1991). A. Thess, R. Lee, P. Nikolaev, H. Dai, P. Pettit, J. Robert, C. Xu, Y. H. Lee, S. G. Kim, A. G. Rinzler, D. T. Colbert, G. E. Scuseria, D. Tomanek, J. E. Fischer and R. E. Smalley, Crystalline ropes of metallic carbon nanotubes, Science, 273, 483-487 (1996). N. G. Chopra, R. J. Luyken, K. Cherrey, V. H. Crespi, M. L. Cohen, S. G. Louie, and A. Zettl, Boron nitride nanotubes, Science, 269, 996-997 (1995). R. V. Krishnarao, M. M. Godkhindi, M. Chakraborty and P. G. Mukunda, Formation of SiC whiskers from compacts of raw rice husks, Jour. ofMaterials Science, 29, 2741-2744 (1994). S.lijima and T. Ichahashi, Single-shell carbon nanotubes of 1nm diameter, Nature, 363, 603-605 (1993).

44 [16] [17] [18] [19] [20] [21] [22] [23) [24] [25] [26J [27] (28) [29] [30) [31) [32] [33] [34] [35) [36) [37] [38J [39] [40J [41] [42) [43] [44] [45] [46] [47] [48] [49)

Chapter 2

J. C. Charlier, A. De Vita, X. Blase and R. Car, Microscopic growth mechanisms for carbon nanotubes,

Science, 275, 646-649 (1997). T.W. Ebbessen, Carbon Nanotubes, Ann. Rev. Mater. Sci.,24, 235-264 (1994) T. J. Trentler, K. M. Hickman, S. C. Goel, A. M. Viano, P.C. Gibbons and W. E. Buhro, Solution-liquid-solid growth ofcrystalline III-IV semiconductors: ananalogy tovapor liquid solid growth, Science, 270, 1791-1794 (1995). R. G. Parrish, Pitch carbon fibers and batt, U. S. Patent 4,861 ,653, August 29, 1989. T. Imamura and MNakamizo, Carbon, 17, 507 (1979). F. Benissad, P. Gadalle, W. Coulon and L.Bonnetain, Formation defibres decarbone a partir dumethane I: croissance catalytique et epaississement pyrolytique, Carbon, 26, 61 (1988). S.Motojima and Y. Kawashima, Chemical vapour growth of HfC whiskers and their morphology, Journal of Materials Science, 31, 3697-3700 (1996). H. Iwanaga, M. Kawaguchi and S. Motojima, Growth mechanisms and properties ofcoiled whiskers ofsilicon nitride and carbon, J. J. Appl. Phys.,32[1], 105-115 (1993). U. Vogt, J. KObler, and G. Engeli, Synthesis of SbN4 Fibers by a Gas Phase Reaction, Ceramics International (1996). G. G. Tibbetts, M. Endo and C.P. Beetz, Jr., Carbon Fibers Grown from the Vapor Phase: A Novel Material, SAMPE Journal, 22(5), 30-35 (1986). M. Hatano, T. Ohsaki and K. Arakawa, Graphite whiskers by new process and their composites, SAMPE Symposium Series, 30,1467-1476 (1985). Information Bulletin, Production ofceramic whiskers, Battelle, Geneva Research Center (1987). J. M. Ting and M. L.Lake, Diamond coated carbon fiber, J. Mater. Res.,9,636-642 (1994). C. Dai and X. Zhang, Microwave synthesis of ultrafine silicon carbide whiskers, J. Am. Ceram. Soc., 80 (5), 1274-1276 (1997). Tokai Carbon Company, Tokawhisker and Tokamax product literature (1987) P.K. Panda, L. Mariappan, V. A.Jaleel, T. S. Kannan, A. Amroune, J. Dubois and G. Fantozzi, Preparation of zirconia and silicon carbide whisker biphasic powder mixtures by carbothermal reduction of zircon powders, Jour. of Mat. Science, 31 [16],4277-4288 (1996). M. Endo etaI., J. Phys. Chem. Solids, 54, 1841 (1993). H. Dai, E. W. Wong, Y. Z. Lu, S. Fan and C. M. Lieber, Nature, 375, 769 (1995). D. S. Bethune, C. H. Kiang, M. S. de Vries, P. Gorman, R. Savoy, J. Vasquez and R. Beyers, Cobaltcatalyzed growth ofcarbon nanotubes with single-atomiclayer walls, Nature, 363, 605-607 (1993). P. Gese, Scanning probe microscopy: materials analysis with anadded dimension, Journal of Metals, 12,810(1992). C.A. Spindt, H. E. Holland, A. Rosengreen, and I. Brodie, Field emitter arrays forvacuum microelectronics, IEEE Trans. Electron Dev., 38, 2355-2363 (1991). H. W. Fink, H. Schmid, H. J. Kreuzer, and A. Wierzbicki, Atomic resolution in lensless low-energy electron holography, Physics Review Letters, 67,1543-1546 (1991). B. Li, G. Y. Xu, E. Y. Sun, Y. Huang and P. F. Becher, Synthesis and morphology of niobium monocarbide whiskers, J. Am. Ceram. Soc., 81 [6], 1689-1691 (1998). J. E. Griffith, D. A.Grigg, M. J. Vasile, P. E.Russell, and E. A. Fitzgerald, Scanning probe metrology, J. Vac. Sci. Technology, Al0, 674-679 (1992). L. R. Chapmann and G. L. Vaneman, Solar collector based on whisker-shaped oxides grown on metallic substrates, Solar Energv, 28, 77-79 (1982). G. A. Niklasson and C. G. Granquist, Surfaces for selective absorption of solar energy: an annotated bibliography, 1955-1981 , Journal of Materials Science, 18,3475-3534 (1983). I. C. Leu, Y. M. Lu, and H. M. Hon, Substrate effect on the preparation of silicon carbide whiskers by chemical vapor deposition, Journal ofCrystal Growth, 167 [3-4], 607-611 (1996). J. V. Milewski, Method and apparatus forcontinuous growth ofsingle crystal whiskers, US Patent 5,404,836, April 11, 1995. H. J. Choi and J. H. Lee, Continuous synthesis of silicon carbide whiskers, Journal of Materials Science, 30, 1982-1986 (1995). Mechanical properties of silicon carbide, Published search, NTIS Accession number: PB-95-854485, NERAC, Inc..Tolland, CT, February 1996. T. V. Hughes and C. R. Chambers, U. S. Patent 405,480, Manufacture ofcarbon filaments (1889). G. G. Tibbetts, D. W. Gorkiewicz and. D. C. Hammond, Apparatus for forming carbon fibers, US Patent 5,024,818, June 18, 1991. G. G. Tibbetts, D. W. Gorkewicz and R. I. Alig, Method forforming carbon fibers, US Patent 5,413,773, May 9,1995. M. Endo, T. Koyama, and Y. M. Hishiyama, Structural improvement of carbon fibers prepared from benzene, Japan J. Appl. Phys., 15,2073(1976).

Chapter 2

[501 [51] [52] [53] [54]

[55] [56J [57J [58J [59]

[60J [61J [62J [63J [64J [65]

[66] [67]

[68] [69] [70] [71] [72J [73]

[74] [75]

[76J [77]

[781 [79J [80l

45

J. S. Speck and M. S. Dresselhaus, Graphitization of thin benzene derived carbon fibers, in Graphite

Intercalation Compounds: Science and Applications, M. Endo, M. D. Dresselhaus, and G. Dresselhaus, Eds, Materials Research Society, p. 196 (1988). J. A. Grande, Vapor-grown carbon fiber isa low-cost reinforcement, Modern Plastics, 41, December 1997. T. W. Ebbessen, Carbon nanotubes, Physics Today, 26-32 (1996). D. Ugarte, A Chatelain and W. A. deHeer, Nanocapillarity and chemistry incarbon nanotubes, Science, 274, 1887-1899 (1996). A. G. Rinzler, J. M. Hafner, P. Nikolaev, L. Lou, S. G. Kim, D. Tomanek, D. T. Colbert, P. Nordlander and R. E. Smalley, Unraveling nanotubes: field emission from an atomic wire, Science, 269,1550-1553 (1995). R. B. Marcus, T. S. Ravi, T. Gmitter, K. Chin, D. Liu, W. J. Orwis, D. R. Ciarlo, C. E. Hunt, and J. Trujillo, Formation ofsilicon tips with 1I-Im radius, Appl. Physics Letters, 56, 236-238 (1990). Y. A. Rovensky, A. D. Bershadsky, E. I. Givargizov, L. N. Obolenskaya and J. M. Vasiliev, Spreading of mouse fibroblasts ona substrate with multiple spikes, Experimental Cell Research, 197, 107-112 (1991). New standards make manufacturing silicon carbide whiskers safer, a review, ASTM Standardization News, 23, p. 11 (1995). G. L. Vaughn and S. A. Trently, The toxicity of silicon carbide whiskers, a review, Journal of Environmental Science and Health, Part A, 31 [8],2033-2054 (1996). Silicon carbide whisker composites, Published search, NTIF Accession number: PB-95-854667, NERAC, Inc.,Tolland, CT, February 1996. D. L.Jiang and J. H. Shea, Silicon carbide based composites, a review, Key Engineering Materials, 108-110, 67-84 (1995). G. G. Tibbetts, C. P. Beetz and G. Devour, Thin inorganic fibers, US Patent 5,424,126, June 13, 1995. M. Geand K. Sattler, How togrow buckytubes, ";cience, 260, 515-518 (1993) W. A. deHeer, A. Chatelain and D. Ugarte, A carbon nanotube field emission electron source, Science, 270, 1179-1 180 (1995). M. Bockrath, D. H. Cobden, P. L.McEuen, N. C. Chopra, A. Zettl, A. Tess and R. E. Smalley, Singleelectron transport inropes ofcarbon nanotubes, Science, 275, 1922-1925 (1997). D. S. Bethune, R. B. Beyers and C. H. Kiang, Carbon fibers and method for their production, US Patent 5,424,064, June, 13,1995. K. Uchida, M. Yumura, S. Ohshima, Y. Kuriki and F. Ikazaki, Process of isolating carbon nanotubes from a mixture containing carbon nanotubes and graphite particles, US Patent 5,560,898, October 1,1996. K. Ohta and N. Hamada, High-molecular weight carbon materials and method for forming the same, US Patent 5,4489,477, February 06, 1996. J. Gimzewski, M. Parrinello, and B. Reihl, Molecular recordinglreproducing method and recording medium, US Patent 5,547,774, August 20,1996. Y. Feldman, E. Wasserman, D. J. Srolovitz and R. Tenne, High rate, gas-phase growth of MOS2 nested inorganic fullerenes and nanotubes, Science, 267, 222-225 (1995). L.C. Venema, J. W. C. Wildiier, J. W. Janssen, S. J. Tans, H. L. J. Temminick Tuinstra, L. P.Kouwenhoven and C. Dekker, Imaging electron wave functions asquantized energy levels in carbon nanotubes, Science, 283,52-55 (1999). A. Fisher, Do-it-yourself molecules, Popular Science, p. 20, July 1998. L. A. Chemozatonskii, Z. Ja. Kosakovskaja, A. N. Kiselev, N. A. Kiselev, Carbon films of oriented multilayered nanotubes deposited on KBr and glass by electron beam evaporation, Chem. Phys. Letters, 228,94-99 (1994). M. W. Browne, A microscopic revolution - nanolubes expected to replace silicon devices, in International Herald Tribune, page 10, February 19, 1998. A. M. Morales, and C. M. Lieber, A laser ablation method for the synthesis of crystalline semiconductor nanowires, Science, 279, 208-210 (1998). M. A. Rodriguez, F. J. Limpo, J. A. Escrina, N. S. Makhonin, M. I. Osendi, M. F. Barba, J. E. Iglesias and J. S. Moya, Single crystal SiAION fibers obtained by self-propagating high-temperature synthesis, Scripta Materialia, 37[4], 405-410 (1997). M. S. Dresselhaus, G. Dresselhaus and P. C. Eklund, Science of Fulferenes and Carbon Nanotubes, Academic Press, San Diego, CA (1996). J. H. Choy and Y. S. Han, A combinative flux evaporation - slow cooling route to potassium titanate fibres, Materials Letters, 34,111 -118 (1998). S. B.Sinnott, O. A. Shenderova, C.T.White and D. W. Brenner, Mechanical properties of nanotubule fibers and composites determined from theoretical calculations and simulations, Carbon, 36[1 -2], 1-9 (1998). M. M. J.Treacy, T. W. Ebbesen and J. M. Gibson, Nature, 381, 678 (1996). H. G. Craighead, S. W. Turner, R. C Davis, C. James, A. M. Perez, L. Kam, W. Shain, N.J. Turner and G. Banker, Chemical and topographical surface modification forcontrol ofcentral nervous system cell adhesion, Biomedical Microdevices, submitted forpublication (1998).

46 [81J [82] [83J

Chapter 2 J. Liu, A. G. Rinzler, H. Dai, J. H. Hafner, R. K. Bradley, P. J. Boul, A. Lu, T. Iverson, K. Shelimov, C. B. Huffman, F. Rodriguez-Macias, Y. S. Shon, T. R. Lee, D. T. Colbert and R. E. Smalley, Fullerene pipes, Science, 280, 1253-1255 (1998). S. Frank, P. Poncharal, Z. L. Wang and W. A. de Heer, Carbon nanotube quantum resistors, Science, 280, 1744-1746 (1998). John V. Milewski and H. S. Katz, Handbook of Reinforcements for Plastics, Van Nostrand Reinhold Company, New York (1987).

CHAPTER 3 CONTINUOUS OR ENDLESS INORGANIC FIBERS F. T. Wallenberger This chapter deals with four continuous processes capable of yielding continuous, potentially continuous, and discontinuous fibers from the vapor phase. One is a commercial process; the others are experimental processes. 3.1 Continuous vapor phase processes Laser assisted chemical vapor deposition (LCVD) yields continuous or discontinuous low diameter fibers directly from the vapor phase by tip growth. Chemical vapor deposition (CVD) on the surface ofa small diameter, preferably sacrificial, precursor fiber yields a large diameter fiber or microtube. Chemical vapor infiltration (CVI) can change the chemistry of precursor fibers by infiltration of a chemically reactive vapor species. Finally, laser vaporization (LV) of carbon-metal mixtures yields highly entangled mats of nearly endless nanotube ropes. 3.1 .1 Laser assisted chemical vapor deposition Laser assisted chemical vapor deposition (LCVD) is a relatively new process. It has already shown promise of becoming a major route for the fabrication of (1) potentially continuous small diameter fibers having premium structural, thermal or optical functionality, (2) complex fiber based microparts such as microsprings and solenoids, and (3) microdevices which operate by using coupled electrical, magnetic and thermal fields. Three excellent reviews should be consulted for details [1 -3]. (a) The generic process concept

Laser assisted chemical vapor deposition is an evolutionary extension of the metal particle catalyzed chemical vapor deposition, wherein a hot laser focus takes the place of a hot solid or liquid metal particle catalyst (Figure 1). Conventional chemical vapor deposition has no "hot spot" capable ofpreferentially focusing the vapor phase deposition. The laser beam in a generic LCVD process has a focal point adjusted inside the reaction chamber tocoincide with the tip ofthe growing fiber or micropart, whether a single (Figure 2a, b) ordual laser source isused (Figure 2c). In either case, fiber growth is achieved by moving the growing fiber away from the stationary laser focus ata constant rate. Since laser assisted CVD is a containerless process, the growing fiber (or micropart) is not in contact with any reactor wall or foreign material such as a metal particle catalyst, and is therefore chemically pure and structurally uniform.

48

Chapter 3

Hot metal catalyst

Substrate

Substrate

Substrate

Figure 1. Metal particle catalyzed and laser assisted chemical vapor deposition. Left: Chemical vapor deposition causes the formation of a film orcoating on a hot surface. Center and right: Metal catalyzed and laser assisted chemical vapor deposition causes the formation of a potentially continuous fiber with a diameter corresponding tothe hot metal catalyst particle orlaser focus respectively. Redrawn from F. T.Wallenberger, P. C. Nordine and M. Boman, Inorganic fibers and microstructures directly from the vapor phase, Composites Science & Technology, 5, 193-222 (1994).

Short (or discontinuous) fibers are best prepared inabatch process, e.g., in a small cylindrical reaction chamber. The value of the technology, however, lies in its capability to facilitate the growth of continuous (potentially endless) fibers with a recently discovered automatic selfregulating growth mechanism (2). Finally, the diameter of the laser focus determines the diameter of fibers grown by laser assisted chemical vapor deposition, just as the diameter of the metal particles determines the diameter of the whiskers grown by metal catalyzed chemical vapor deposition. The fiber diameters which can be achieved with laser assisted CVD are laser dependent and the growth rates are reactor pressure dependent. A CO2 laser yields fibers with large diameters (>200 pm) while A( or Nd-YAG lasers are known to yield much smaller diameters (>4 urn). And as discussed in the next chapter, the reactor pressure determines the growth rate which can be as small as <1 ~m/s at pressures of <1 bar, or as high as 1 mm/s at pressures of >1 bar. Depending on the process conditions, laser assisted chemical vapor deposition can therefore be used to grow short or endless fibers, with diameters ranging from <10 umto >200 prn, and with growth rates ranging from <1 ~m/s to>1 mm/s (Figure 2). urn depending on the focal length. The cylindrical shape of a freestanding fiber with a micrometer diameter is a simple, yet demanding shape. It lacks three-dimensional complexity, but perfect process control is required to achieve a uniform diameter over the length of the fiber and therefore a uniformly cylindrical shape. Such a high degree of uniformity, which incidentally is the cause of high fiber strength, is not readily obtained by any other method. In addition, the new technology is capable of yielding three-dimensional, fiber based microparts including microcoils with a

Chapter 3

49

single laser system, and irregular shaped, fiber based microobjects with a complex dual laser system (Figure 2).

Lens

Overlapping laser focus

J-x

z

Figure 2. Growth of microstructures by laser assisted chemical vapor deposition (LCVD). Linear single laser system (A). Goniometer controlled single laser system (8)and complex dual laser system (C). Redrawn from F. T. Wallenberger, Rapid prototyping directly from the vapor phase, Science, 267, 1274-1275 (1995).

(b) The low pressure process The applied physics community uses the low pressure LCVD process as a time saving device first to prototype the reaction rates and deposition kinetics in this relatively small system and then to apply the results to the large scale surface deposition of films in a conventional CVD process [1]. More recently, this process was used tofabricate a wide range of microstructures directly from the vapor phase. The resulting products include low diameter carbon, boron and

50

Chapter 3

silicon fibers, boron microsprings with low coil diameters [4-5], low diameter silicon microsolenoids with tungsten lines [2] [4], and complex microobjects based on low diameter fiber elements [6]. This process serves as a rapid prototyping device but may not be useful for the commercial fabrication offibers or microparts because only very low growth rates «10 um.s) are obtained in the low-pressure regime.

Scrubber F

I; ... :l~

::: :=::::::;::;: ~

I Control I

QI

,~

Gaspanel

---- E~d"

IThermostat I Focus

Shutler

II

Laser

I

--.

- .-

Lens

'\

ILasershape

Pump

1 ~1l=lllIml

~ .en

J'.

~

~

MFC

~

l\-

"-

/'

lL

.. ~

u,

Reactor

I I

Total pressure control

r

Motor

~

Translation stage

Beam shape

Figure 3. Schematic drawing of the low-pressure LCVD process. Redrawn from F. T. Wallenberger, P. C. Nordine and M. Boman, "Inorganic fibers and microstructures directly from the vapor phase", Composites Science and Technology, 5, 193-222 (1994).

In a typical low pressure LCVD process, a cold or hot wall reactor is used incombination with laser heating [7]. In a hot wall process, the reactor temperature is set to a value just below the deposition temperature tofavor a homogeneous reaction. The laser beam is then used to increase locally the heat of the substrate to above the threshold temperature for the requireddeposition and growth tooccur (Figure 3). Specifically, a continuous wave (cw) argon ion laser, as shown in Figure 3 [7], is operated at a wavelength of 514.5 nm. The beam is focused on the substrate atperpendicular incidence with a spot diameter ranging from 2-42. The use oflow reaction chamber pressures «1 bar) in the generic LCVD process yields large (>200 urn) diameter carbon fibers [8] with low growth rates «10 ~m/s) when a C02 laser is used. Small «20 urn) diameter carbon, boron and silicon fibers [1-2] [4] [7] are produced with equally low growth rates at <400 mbar when an efficient Ar- laser is used. In summary, the low growth rates are due tothe low pressure regime, and the fiber diameters are a function of the individual laser capability. 2 BCI] (g)+3 H 2

(g) ~2

B(s)+6 HCl

(1)

51

Chapter 3

Amorphous and polycrystalline boron fibers (Equation 1) are obtained by chemical vapor deposition from boron trichloride [1) [5) [7). At the lowest temperatures where deposition occurs, the rate is controlled by chemical kinetics, and amorphous fibers are obtained with a growth rate of 2 um/s. At higher temperatures, the deposition process becomes limited by gas phase transport, and polycrystalline fibers with a ~-rhombohedral crystal structure grow with a growth rate 5 IJm/s. The 2.5x increase in growth rate occurs around the glass transition temperature for boron, which is -1500 K. SiH.J ( g) ~ Si ( s ) + 2 H 2 (g)

(2)

Single crystal silicon fibers [4) were grown in a closed reactor from a gas mixture consisting of 6% SiH 4 in argon according to Equation 2. The total pressure in the LP-LCVD reactor was 200 Torr (27 kPa). The single crystal silicon fibers were 5 mm long and 150 IJm in diameter. The laser intensity was 2.11x105 W/cm 2 and the growth rate was 1 prn/s. The silicon fiber was grown perpendicular to a silicon substrate and in order to achieve steady state conditions, the lens was moved perpendicularly away from the substrate surface at the same velocity as the growth rate ofthe fiber. Trimethylamin e alane(g ) +02 (g)~ Al 20 3 ( s )

(3)

Trimethylamine alane( g) + N0 2 (g)~ Al 203 ( s)

(4)

Alumina fibers and rods [9) were grown from an aluminum precursor, trimethylamine alane, which is otherwise known to facilitate the production of high quality aluminum deposits. To obtain high quality alumina deposits, either oxygen or N02was mixed into the precursor gas now in a typical low pressure LCVD process (Equations 3 and 4). Oxygen levels between 30 and 35% resulted in the deposition of clear fibers having diameters ranging from 3 to 20 IJm. With 1.3 mW of incident laser light, growth rates of 80 IJm/s were observed. A similar deposition occurs when nitrous oxide isused. Several micromechanical processes have recently been advanced, opening up new opportunities for simple microstructures in the field of sensors and actuators. Indirect micromechanical processes include the fabrication of a freestanding, three-dimensional fiber structure resembling the Eiffel Tower [6). In a three step, LP-LCVD process, a polycarbonate grid is formed first and aluminum lines are then deposited on its surface by a laser-direct write method. The polycarbonate substrate is dissolved, leaving behind the aluminum grid (3x3x7 mm) with a grid orfiber diameter of about 10 IJm. Direct micromechanical processes include fabrication of freestanding three-dimensional microstructures directly from the vapor phase. Using a laser oriented in the growth direction of the helix and a novel goniometer [2) [5) to achieve a computer controlled rotation of the substrate, a crystalline boron microspring, 360 IJm wide, 1.5 mm high and having a fiber diameter of 21 IJm was made with a growth rate of 3.5 IJm/s. A silicon microspring was also made [4) with the same rotational system. The mechanism of the goniometer consists of a stepping motor with a backlash-free reduction gear assuring good positional accuracy. With a gear ratio of 80:1 (or 16,000 steps per revolution) a practically continuous rotational motion was obtained. 2 WF6 ( g) +3 Sirs) = 2 W( s ) + 3 SiF-/

(5)

WF6 (g)+3H 2 ~W(s)+6 HF( g)

(6)

52

Chapter 3

Figure 4. Simple microstructures grown with a single laser system. Left: freestanding boron microcoil (courtesy of Prof. M. Boman, University of Uppsala, Sweden). Right: silicon/tungsten microsolenoid (courtesy of Prof. M. Boman, University ofUppsala, Sweden).

Figure 5. Complex microstructure grown with a dual laser system fabricated with the dual laser system shown in Figure 2C. Courtesy ofProfessor M. Stuke, Max-Planck-Institute fUr biophysicalische Chemie, G6ttingen, Germany.

53

Chapter 3

Using the same rotating goniometer (Figure 4, left), a novel silicon/tungsten microsolenoid (2) (4) was made in a two step process (Figure 4, right). A singlecrystal silicon rod 140 urn wide and 2.5 mm long was first grown from the vapor phase (Equation 2). The microsolenoid was made by depositing a tungsten helix (Equations 5 and 6) with a pitch of 15 turns/ mm on the rotating silicon rod. The growth rate ofthe tungsten line was 1.2 ~m/s . A recent advance (9) marks amajor leap forward inthe evolution ofLCVD. Using two crossed laser beams with an overlapping two beam LP-LCVD process (Figure 5, left), threedimensional fiber growth was achieved facilitating the direct, one step fabrication of complex freestanding micro-structures. For example, by continuously forming and joining alumina fiber segments, each having a diameter of 20 urn, a cage-like microstructure (Figure 5, left) was fabricated with linear growth rates <80 ~m/s . Further extensions of this technology have recently afforded a new method for laser rapid prototyping of photonic band-gap structures (10). Three-dimensional periodic microstructures of aluminum oxide, which are important for creating photonic band-gap structures (PBGs) were fabricated [10) by means of laser induced direct-write deposition from the vapor phase. A face centered tetragonal structure was formed that consisted ofperpendicularly arranged layers ofparallel aluminum oxide rods, each 40 urn indiameter and 3000~m in length, and with lattice constants of66 and 133 urn, respectively. A similar periodic 3-D microstructure in the form of a diamond lattice is shown in Figure 5, right. Potential applications ofthis technology will be discussed inChapter 3.3.3(d). (c) The high pressure process The use ofhigh reaction chamber pressures (1-10 bar) and a small wavelength Nd-YAG laser made it possible to fabricate small (>6 urn) diameter fibers (2) [12-23) at high (0.3-1.1 ~m/s) growth rates. Most fibers obtained so far were made in an experimental high-pressure process and were therefore 14 mm long. However a novel method for controlling the reaction rate makes it possible to achieve uniform growth over continuous lengths of fibers. The highpressure LCVD route may become useful in the commercial fabrication of continuous fibers directly from the vapor phase. In principle, any material that can be obtained as a film by conventional CVD can be obtained as a fiber by LCVD and therefore by high pressure LCVD with relatively high growth rates. To date, chemically pure and structurally uniform boron, carbon, silicon, silicon nitride, silicon carbide and germanium fibers (2) [12-23) were formed (Table I), thereby potentially facilitating the development ofa commercial process. Table I. Growth of high pressure LCVDfibers HP-LCVD Fibers Boron Carbon Germanium Silicon SiC Si,N,

Reactant gases B,HJH, CH, GeH SiH SiH>C,H, SiH,/NH,

Fiber diameter, um Low Avg. 6 19 10 63 59 70 15 45 13 120 21 45

Growth Rate, um/s Avg. High 625 1100 125 331 18 35 460 500 75 125 338 740

54

Chapter 3

The experimental high pressure process provides cwpower up to 270 mW. The beam from a one watt, TEM oo mode, cwNd-YAG laser (emission wavelength 1.064IJm) is passed through a polarizer, a Linconix laser power stabilizer, a variable neutral density filter, a beam expander, and focused with a 10 em focal length lens into the fiber growth reactor [2] [12]. Laser power ranges from 0 to 200 mW with a stability of one mW, as measured outside the reaction chamber. The laser beam was focused onto a point inside a reaction chamber, where localized heating promotes vapor deposition in the direction of the laser. The reactor could be moved parallel and orthogonal to the laser beam direction. Motion parallel to the laser beam was driven at selected rates by a computer operated by a stepper motor. Growth of the fiber at a location where the laser beam converges toits focal pointbecomes self-regulating [2]. (d) Automatic process control In the experimental high pressure process, the gas flows, reactor pressure, and laser power are set atconstant values and fiber growth is initiated by opening a shutter on the laser beam. The fibers grow from the carbonized edge of a paper substrate held inside the reactor by a removable probe. Fiber growth occurs spontaneously when the laser beam is turned on or when the edge of the paper substrate is drawn near to the laser focal point by moving the reactor. Growth of the fiber at a location where the laser beam was converging to its focal point provided an intrinsic method forrate control in vapor-solidgrowth [2). Thus, if the deposition rate exceeds the reactor translation rate, the fiber tip would grow to a point further from the beam waist, thereby cooling the tip and reducing the deposition rate. If the deposition rate is less than the growth rate, the fiber tip will bedrawn closer to the beam waist where its temperature is higher and the growth rate increases to match the pulling rate. The fiber therefore adjusts its position to achieve a temperature that made the pulling and growth rate equal. It is evident that the HP-LCVD process affords considerable latitude in the design and execution of fiber growth and kinetic studies. The discovery of the self-regulating mechanism made it possible tofabricate uniform two meter long fibers [2] bya semi-continuous version of the batch process (Figure 6). The results confirm that a continuous process is feasible in the high pressure regime, and should yield optimal growth rates exceeding 1 mm/s or about equal to the rate by which commercial sapphire fibers are produced by a flux method (see Chapter 6). Recently, a method was described for the real-time measurement of growth rates and feedback control of three-dimensional laser assisted chemical vapor deposition [11]. This method allows the accurate reproduction of high quality films, fibers, and three-dimensional structures. High aspect ratio axisymmetric forms of desired shape and microstructure were grown from vapor phase precursors by this method. Three-dimensional rods, cones, hyperboloids, and spheroids ofpyrolytic graphite, nickel, iron, and nickel-iron superalloys were obtained from ethylene, nickel tetracarbonyl, iron pentacarbonyl, and mixtures of nickel and iron carbonyls, respectively. To control the process [11), a measure of the volumetric growth rate was obtained from specific emission spectra generated during the heterogeneous reaction, and direct feedback control of the reaction rate was realized byusing this growth rate to modulate the laser power in real time. By this feedback method, layered and continuous prototyping is possible on a microscale since real time compensation for growth rate perturbations can be made. The

55

Chapter 3

study was carried out atpartial pressures with growth rates up to 45 IJm!s. While the process is potentially continuous even at partial pressures, the growth rates are too low to be of commercial significance. 3.1 .2 Conventional chemical vapor deposition Mass transfer in metal catalyzed and in laser assisted CVD processes is driven by highly localized temperature gradients. The relatively small area ofeither a hot molten metal particle or of a hot laser focus affords whiskers [4] or continuous fibers, respectively [2] [18-19]. The transfer of an equal mass from the vapor to the solid phase in a conventional chemical vapor deposition results in a thin coating over the relatively large area of a hot surface, i.e., that of a flat complex shaped composites part.

1-----Laser beampath

Centralcavity Mechanism to support fiber growth

• '-"-"-"-r'!

F'--: .;;,:._ •• ...J

"-"L_,

ii :

, I

._.._.._.. _!._~ Reactant inlet

Outlet

Figure 6. Drawing of the high pressure LCVD reaction chamber. Courtesy of Dr. P. C, Nordine, Container1ess Research Inc.. Evanston, IL.

(a) Commercial hotfilament CVD process

Conventional chemical vapor deposition produces a coated fiber when a thin coating is uniformly deposited by this method over the fiber surface. Such a coating affords an insignificant diameter increase of the resulting fiber and the functionality of the product continues to be that of the coated fiber. It merely provides an enabling, e.g., an oxidation resistant, function. Chemical vapor deposition, however, produces a large diameter sheath! core fiber, when a thick coating is uniformly deposited over the surface of a small diameter fiber (Figure 7). The fiber does not grow by lengthening (i.e., directional growth) but by thickening (i.e., side growth). The diameter of the fiber increases in this process to up to 10x without change in its length, and the functionality changes from that of the core to that of the sheath.

56

Chapler3

Continuous boron/tungsten fibers were the first high performance fibers tobe designed, about 40 years ago and commercialized about 30 years ago to meet the demanding end use requirements for resin and metal matrix composites in aircraft and sporting goods markets. Using the same process, silicon carbide/carbon fibers were also commercialized. Both fibers continue to represent important niche products. Experimental boron/carbon and silicon carbide/tungsten fibers were also developed. The structures, properties, and applications of these fibers are discussed inChapter 3.2.2. Boron/tungsten (BIW) fibers are produced by vapor deposition of boron (Equation 1) on the surface of a practically endless, electrically heated tungsten filament having a diameter of 121Jm. The reaction chamber (Figure 7) is a closed system with mercury seals on both ends and two segments [24-25]. The surface of the tungsten filament isdecontaminated by heating it to ~1350°C [30] in the first (short) segment in a reducing atmosphere. The cleaned tungsten filament and the reaction mixture of boron trichloride and hydrogen are then passed through the second (long) segment, the reactor. The electrically heated tungsten filament causes boron trichloride to decompose and deposit boron on the filament surface. Since each boron fiber requires its own reaction chamber, commercial production requires literally hundreds ofreaction chambers which are linked toa common gas supply, a gas mixer and gas regenerator. An individual reactor may be about 2 mlong. The single individual high density tungsten filament enters the reactor with a diameter of 12 IJm and a low-density boron/tungsten fiber with a diameter of 100 or 140 IJm exits. Since it passes through the reactor within a minute or two, the throughput may be as high as 1.0 m/min. The materials cost for the sacrificial tungsten wire dominates the overall cost of manufacture. A less expensive carbon filament was used for awhile as a substitute for the tungsten wire. Figure 8 contrasts a boronltungsten fiber with a pure boron fiber. Aside from the ~10x difference in diameter, the differences in surface texture are noteworthy. The surface of the pure boron fiber made by high pressure LCVD is smooth. Its strength is 7.5 GPa and its modulus is 400 GPa. In contrast, the surface texture of the boronltungsten fiber is "nubby". Its strength is3.6 GPa and its modulus is 400 GPa. In summary, the tensile strength ofboron fibers isrelated totheir surface uniformity. Silicon carbide/carbon (SiC/C) fibers are also commercially produced by side growth in a single stage hot wire CVD process, using silane or tetrachlorosilane, hydrogen and methane as the reactants [25]. A good balance between vapor phase reaction and deposition ofsilicon carbide from the vapor phase is obtained when the carbon filament is resistively heated to 1200°C. The bicomponent fiber reaches 140 IJm in this process, has a density of 3.0g/cm 3, strength of3.45 GPa, and modulus of400 GPa. Apyrolytic graphite coating may be applied to the carbon fiber to modify the interface between the core and the sheath, and various commercial coatings, designated as SCS, may be applied to modify the outer surface of the silicon carbide sheath. A tungsten wire can be used as a commercial alternative instead of the carbon fiber core. (b) Experimental CVD and PVD processes This subchapter deals with experimental sheath/core fibers made by conventional as well as plasma enhanced chemical vapor deposition, and by plasma enhanced physical vapor deposition.

57

Chapter 3

Payout Pyroolytic Graphite deposition section

H2 --+J"-+...,

o a. B E

20

Tungsten

g

:5c

Boron deposition section

40

~

tii

§ '0 C

-

~


11.

60

80

100

/

/

7

V

~

~

II /

L

II Bo~on

on carbon I

1100

1200

I

1300

Temperature. °C

Figure 7, Schematic diagram of the CVD boron/tungsten fiber process. Redrawn from M, L. Dorf, Product bulletin, Textron Specialty Materials, Lowell, MA,

Figure 8, Boron fibers made by hot filament and by laser assisted CVD, This illustration compares the fiber diameter and surface character of a sheath/core boron/tungsten fiber (A,C) with that of a pure boron fiber (B,D). Reproduced from F, T, Wallenberger and P, C, Nordine, Strong, Small Diameter Boron Fibers by Laser Assisted Chemical Vapor Deposition, Materials Letters, 14 [4] 198-202 (1992), With permission from Elsevier Publishers (1992).

58

Chapter 3

For example, a straight or helical, but potentially continuous diamond/ tungsten fiber can be made by depositing diamond on a straight tungsten wire [26]. The reaction was carried out in a hot filament reactor [24] using a 1% CHJH 2 gas mixture and a flow rate of 200 standard cubic centimeters per minute at a pressure of about 20 torr (1 torr = 1.333 x 102 Pal. A vertical Ta filament dissociated the gases at about 2000'C and deposited diamond onto a straight or coiled tungsten wire, which was radiantly heated to about 900'C and held parallel to, and about 5 mm from, the filament. A similar process [27] affords the hot wire deposition ofdiamond on carbon, silicon carbide, boron, boron carbide, zirconia and alumina. A straight hollow diamond fiber or microtube can be made by etching the tungsten core [26] from a diamond/tungsten fiber. Alternatively [26], a diamond/tungsten microspring results if diamond isdeposited on an open, orextended, tungsten microcoil whilea straight diamond/tungsten microtube results if diamond is deposited on a tightly wound, or unextended, tungsten microcoil (Figure 9). The wire reinforced diamond microtube and the diamond/tungsten microspring require the use of a coiled tungsten fiber. It can be made by winding an annealed tungsten wire around a support rod, relaxing the wire coil, and removing it from the rod. The inside diameter of the wire spring corresponds to the outside diameter of the support rod. Plasma enhanced chemical vapor deposition isan alternate route to sheath/core fibers having a diamond, metal or ceramic sheath on a polymer, metal or ceramic core [28]. Again, microtubes result when the sacrificial (e.g., polymer) core is removed [28]. For example, the use of a magnetron sputtering system offers a high degree of process fleXibility and enables the production of metal microtubes (Cu, Ni, AI, Au, Ptand Ag), ofceramic microtubes (SiC, C, ShN4 and sapphire), and of layered material combinations (carbon/nickel and silver/sapphire) ina myriad ofshapes and with a wide range ofinternal diameters ranging from 0.5to300 urn [28]. Microtubes made by this route to help revolutionize the miniaturization of electronic components and systems will be covered in alater chapter.

Figure g. Potentially continuous sheath/core diamond/tungsten bicomponent fibers. A straight diamond fiber results when astraight tungsten wire iscoated with diamond (not shown). A helical diamond fiber (orspecifically, a diamond coated, tungsten reinforced microcoil) results when a helical wire is coated with diamond (left). A hollow diamond fiber with a much larger outside diameter (i.e., a tungsten reinforced, diamond coated microtube) results when a tightly coiled tungsten wire isover-eoated with diamond (right). Courtesy of Drs. G. H. Lu, P. G. Partridge and P. W. May, University ofBristol, UK.

59

Chapler3

3.1 .3 Chemical vapor infiltration processes Chemical vapor infiltration (CVI) is a process whereby reactive chemical species are generated in the vapor phase and allowed to react with a solid substrate thus modifying its chemistry. A successful use of this process requires (1) an absolutely continuous and stoichiometric conversion of the initial to the final chemistry in the solid state, and (2) a continuous and eventually complete conversion of the morphology, density, surface tension, mechanical as well all other properties. Discontinuous or incomplete results cause a steep drop in strength, the premier measure of uniformity. By combining the complexities of chemical vapor deposition (this chapter) with those of fiber formation from a precursor fiber (Chapters 8 to 12), the process istherefore intrinsically more difficult to control than any other. (a) CVI of carbon fibers with silicon oxide

In a typical CVI reactor (29), individual carbon fibers, multi-filament yarns, or woven fabrics were exposed to silicon monoxide (Equation 7) in an inert environment and under controlled conditions at temperatures up to 1500°C. The reactor was an experimental unit capable of infiltrating carbon fiber and yarn samples with lengths up to 15 cm, as well as woven fabric swatches with dimensions ofapproximately 15 cm x 15 cm. While the reaction (Equation 7) is potentially stoichiometric in a fiber-to-fiber, yarn-to-yarn or fabric-to-fabric infiltration process [29) may remain incomplete. 2C (fiber) + SiO (gas)

~

SiC (fiber) + CO (gas)

(7)

As a result, the strength level for a stoichiometric or 100% converted silicon carbide yarn was reported to be 1.11 GPa (uncoated) and 1.60 GPa (coated), and that for a 90% converted yarn was reported to be 1.70 GPa (uncoated). The individual silicon carbide fibers, multifilament yarns and fabric samples which were obtained were suitable for evaluation of mechanical properties from room temperature to 1500°C. A commercial process for the fabrication of continuous multi-filament silicon carbide yarns by this route has not yet been reported [29). (b) CVI of boron oxide fibers with ammonia

Boron nitride fibers have been prepared in the laboratory by chemical vapor infiltration of boron oxide glass fibers with ammonia (Equation 8), a process that may alternatively be considered to be a nitridation of 8203 precursor fibers [31). The precursor fibers, in turn, are melt spun ata low temperature (480°C) from a viscous melt. Thus, the nitridation of a boron oxide fiber could alternatively be considered to be derived from a solid precursor fiber, a topic otherwise discussed in Chapters 8 to 12. In this process, the final step is the chemical conversion of a given precursor fiber at a high temperature in a highly reactive vapor phase environment. BP3 (fiber) + 2NH3 (gas)

~

2BN (fiber) + 3 H p(gas)

(8)

Due to the low liquidus of B20 3 (450°C), the vapor infiltration and chemical reaction proceeds in several stages, first at 200°C to form an infusible shell at the fiber surface, then above 350°C to give off water and to create a microporous and highly disordered structure [31). Since the conversion requires an inward diffusion of NH3 and an outward diffusion of water, it ismore readily achieved with precursor fibers having small diameters. Thisconversion step is

60

Chapter 3

complex and not fully understood but empirically feasible. A short heat treatment at 1500°C simply removes the last traces ofresidual volatile species and stabilizes the microtexture. (e) CVI of borazine fibers with ammonia

A boron nitride fiber derived from a borazine precursor fiber has recently been reported [32]. As with boron nitride fibers from B203, the final heat treatment of boron nitride fibers from borazine must be carried out ina reactive atmosphere (also ammonia) tochemically complete the conversion of the precursor fiber. This nitridation step renders the reaction sequence a chemical vapor infiltration process. 2, 4,6 trichloroborazine and methylamine, the starting materials for this BN fiber, react to yield 2, 4, 6- tris (methylamino) borazine [32], a waxy solid (m.p. = 65°C). This intermediate undergoes thermal self-condensation when heated at 140 and yields an infusible product. However, inthe presence of 10 wt.% ofa long chain (e.g., lauryl) amine, a clear and colorless thermoplastic trimer is formed at 150-250°C. An 8-25% weight loss that is observed corresponds tothe evolution of methylamine. Laurylamine is incorporated into the polymer as alkylimino and alkylamino groups. 0

e

Despite itslow molecular weight, the precursor can be melt spun at 100°Cin air, and yields a continuous precursor fiber. This green fiber is heat treated in ammonia at 1000o e. Chemical vapor infiltration and the resulting reaction convert carbonaceous materials in the fiber into volatiles. A short heat treatment at 1800°C in nitrogen removes all residual volatiles and it stabilizes the boron nitride fiber and its microtexture. The final boron nitride fiber is white and has adensity of2.05 g/cm3. 3.1.4 Laser vaporization ofcarbon-metal mixtures Single wall carbon nanotubes have been prepared with lengths of <10 nm, lengths ranging from 10 to 300 nm [50] and with lengths <1200 nm [51]. Short carbon nanotubes are extensions of spheroidal fullerenes. Short and intermediate length carbon nanotubes have been analyzed in Chapter 2, but carbon nanotubes orfullerene pipes, believed to be endless, still require methods for their disentanglement from as-produced ropes and/or mats. From extensive SEM and transmission electron microscope imaging, it is known that the fullerene rope fibers are so highly entangled with one another and so long that their ends are rarely visible [50]. In addition, the frequent occurrence of fullerene toroids in these samples suggests that many of the fullerene rope ends are difficult to find and that the ropes are therefore believed to be truly endless [51]. It appears that the driving force for their continuous self-assembly is high, given the right conditions, and that methods will eventually become available for the manufacture of freestanding, continuous and unentangled carbon nanotubes. 3.2 Selected structures and properties

Each of above processes offers new insights into the relationship between process variables and structures, and between structures and properties. Since however most of these processes are relatively new, a comprehensive comparison along these lines is still not quite possible. The discussion, which follows, highlights important structure property relationships

61

Chapter 3

for each of the major processes, and then compares the general relationships between structures and mechanical properties. 3.2.1 High and low pressure LCVD fibers The relationships between process variables and structures and properties [2] [20-21] are best illustrated with examples from the extensive literature describing important commonalties and differences between the high pressure and the low pressure iaser assisted chemical vapor deposition. (a) Reactor pressure vs. growth rate The first example (Figure 10) shows the powerful effect ofreactor pressure on the growth rate of boron fibers [2] [12]. The diameters of the boron fibers made in the high pressure process ranged from 6 to 26 IJm and their tensile strengths from 3.5 to 7.6 GPa. Their modulus was 400 GPa. The diameters and tensile properties of amorphous low pressure boron fibers are generally the same. Increasing pressure, therefore, increases the growth rate at a given tip temperature. An increase in tip temperature in either system would therefore afford a progression from nano- to polycrystalline fibers. At high tip temperatures and low reaction pressures, polycrystalline boron fibers have been reported to have a tensile strength of up to 13 GPa, and a modulus of up to 420 GPa [5]. None of the LCVD boron fibers had a tungsten core such as the commercial CVD boron/ tungsten fibers. These fibers have diameters ranging from 100 to 140 IJm, a tensile strength of 3.5 GPa and a modulus of 400 GPa. The higher strength obtained in either LCVD process seems to reflect the fact that they were grown by tip growth, while the boron/tungsten fibers were grown by side growth ("thickening"). lCVD chamber pressure

• •

1000

6.89

1020 kPa (134 psig)

5.52

800 'iii

•

0.

"'"

.£ 0> c

600

670 kPa (74 psig)

!!!

1i> ~

'iii

c ~

400

•

380 kPa (40 psig)

'"

c,

4.14

2.76

C)

.£

0> c !!! 1i>

~

'iii c

~

~

240 kPa «20 psig) 1.38

200 0

200

400

600

800

lCVD chamber pressure

Figure 10. Effect of reactor pressure on the strength of boron fibers. Redrawn from F. T. Wallenberger, P. C. Nordine and M. Boman. Inorganic fibers and microstructures directly from the vapor phase, Composites Science and Technology,S, 193-222 (1994).

62

Chapter 3

The second example (Table II) deals with LCVD silicon fibers. In the high pressure LCVD process, single crystal silicon fibers were obtained with tip temperatures above 1400°C, high growth rates (>500 IJm/s) and VLS phase transformation. And, polycrystalline silicon fibers were obtained with tip temperatures between 600 and 1400°C and intermediate growth rates (12-500 IJm/s) under conditions where VLS or VS transformations could occur. Amorphous silicon fibers were obtained with very low tip temperatures (525°C), low growth rates and VS phase transformation. In the LP-LCVD process, single crystal silicon fibers were obtained with high tip temperatures and low growth rates (~1 IJm/s) and polycrystalline silicon fibers with lower tip temperatures, low growth rates (::;1 IJm/s) and VS. Table II. Growth of High and Low Pressure LCVD Silicon Fibers [2] [4] [5] [14] [15] TipT.,OC HPandLP >1400 1400-600 >525

Rate. um/s HP LP >500 ~1 >12 :51

1

Fiber morphology HPandLP Single crystal Polycrystalline Amorphous

Strength. GPa

HP

HP

4.0

3.0-6.7

Modulus. GPa HP LP 180 164 130

The relationship between process variables (e.g., tip temperature) and structures is reminiscent of that which governs the growth of metal particle catalyzed chemical vapor deposition (Chapter 2). It follows traditional patterns. Single crystal silicon fibers have relatively high strength and relatively high stiffness. Polycrystalline fibers have lower stiffness (modulus) than single crystal silicon fibers. Single crystal silicon fibers made by the lowpressure process were occasionally found to have a polycrystalline overgrowth . The latter serves as a stress riser and is responsible for the variable strength levels which were observed. (b) Tip temperature vs. properties

The third example (Table III) deals with HP-LCVD carbon fibers and illustrates the same overall relationships [13] [16]. Depending upon growth conditions and feed gas chemistry [16], these fibers were very strong and graphitic when formed at high tip temperatures; thickened and brittle when formed atintermediate tip temperatures; orvery flexible and elastic when formed at low tip temperatures. Graphitic LCVD carbon fibers had the highest strength (3.0 GPa) and modulus (::;180 GPa), and flexible carbon fibers the lowest strength (::;0.4 GPa) and modulus « 30 GPa). Flexible high pressure LCVD carbon fibers could be readily bent toradii with curvatures of ::;1 mm. The force required was much lower than that required for (a) equal diameter HP-LCVD boron fibers having a modulus of >275 GPa [12] or (b) equal diameter intermediate modulus or 1M carbon fibers having a modulus of250 GPa [16]. These qualitative relationships parallel those observed for silicon fibers, where single crystal fibers were formed with the highest tip temperatures and had the highest strength and stiffness, and where amorphous fibers were formed atthe lowest tip temperature and had the lowest strength and modulus. Table Carbon Fiber Type Commercial 1M Fiber "Graphitic" Fibers "Brittle" Fibers "Flexible" Fibers

m.

Mechanical Properties of HP -LCVD Carbon Fibers Strength, GPa - 3.5 0.5-3.0 NA 0.2-0.4

Elongation, % - 1.5 >1.5 NA

Modulus, GPa -250.0 28-180 NA <30

Chapler3

63

The fourth example shows that some amorphous HP-LCVD fibers having binary siliconnitrogen compositions [17] were silicon rich , others were near stoichiometric Si-N compositions, and only a few were representative of stoichiometric silicon nitride. Thus, the process offers wide latitude in the design of fibers with amorphous or glassy structures. These design options facilitate the production of amorphous fibers from equilibrium and nonequilibrium melt compositions and of amorphous fibers having stoichiometric or nonstoichiometric, binary compositions. The structure property relationships observed in the Si-N system [17], are reminiscent of those observed for silicon nitride whiskers grown by metal catalyzed chemical vapor deposition (Chapter 2.2). The difficulty in obtaining binary (i.e., boron or silicon nitride) fibers with exactly required stoichiometry seems to have so far precluded the production of single crystal fibers by this route. (e) Side growth versus tip growth

Chemical vapor deposition (Chapter 3.1.2) of boron on hot tungsten fibers, or of silicon carbide on hot carbon filaments proceeds exclusively by side growth, i.e., by overgrowth of the core fiber with boron orsilicon carbide. Metal particle catalyzed chemical vapor deposition proceeds primarily, or at least initially, by tip growth, and often continues to grow by side growth after the initial tip growth is complete, for example with carbon whiskers and nanotubes (Chapter 2.3). The same holds true for the laser assisted chemical vapor deposition (Table II) where single crystal fibers grow by a VLS mechanism at high growth temperatures, amorphous fibers at low temperatures by a VS mechanism, or polycrystalline fibers atintermediate growth temperatures by a mixed VLSNS mechanism. Side growth has also been observed with single crystal germanium fibers made by high pressure CVD [22]. Itoccurred when the vapor phase was still hot enough tofacilitate surface deposition of glassy or polycrystalline germanium, but not hot enough to sustain continuing growth of the single crystal germanium fiber by tip growth. The nubby surface of a sheath/core boron/tungsten fiber (Figure 8) is also evidence of side growth. The side-grown boron deposit can be removed by etching [35], a process that produces a smooth surface and a measurable increase in strength. Side growth of amorphous carbon (Figure 11) on the surface of an amorphous HP-LCVD carbon fiber [5] yields a smooth surface that is comparable to that of side-grown carbon on vapor grown carbon whiskers. In these cases, the secondary side growth (or overgrowth) occurs by conventional chemical (surface) vapor deposition. (d) Versatility versus whisker process

The formation of a filamentary structure bymetal particle catalyzed or laser assisted chemical vapor deposition requires a localized energy source, i.e., either a hot metal particle or a hot laser focus tofacilitate the vapor phase reaction, initiate fiber growth and sustain it. Figure 12 visualizes how both the laser assisted and the metal particle catalyzed CVD process utilize a localized heat source, and how its diameter determines that of the resulting fiber. In the laser assisted chemical vapor deposition process at high or low pressures, single crystal fibers are obtained bya VLS phase transformation when the incident laser power (focal temperature) is high, and supports the growth of afiber with a molten tip.

64

Chapter 3

Figure 11 . Tip grown single-cryslal LCVD germanium fiber with secondary amorphous germanium side growth. Courtesy ofDr. P. C. Nordine. Containerless Research Incorporated, Evanston. IL

This includes single crystal silicon [15], germanium (22) and alumina (10) fibers. Polycrystalline fibers can grow either by a VLS or a VS phase transformation when the incident laser power (focal temperature) is intermediate, and supports the growth of a fiber with a semisolid tip. This includes polycrystalline silicon (15), boron (5) and silicon carbide fibers (23). Amorphous fibers are obtained by a VS phase transformation when the incident laser (focal temperature) is low, and supports the growth of a fiber with a hot but solid tip. This includes amorphous silicon [15), boron (12), carbon [13) (16). silicon carbide (23), and silicon nitride (17) fibers.

c: 0

"-5

(5

'"

'0

2

(Jl

::J

u

.2

W

(Jl

ee -'

Fiber growth by laser assisted chem ical vapor deposition

-

OJ

OJ

u

~

s U

:e

::J

ro

'"

i!!

0. '0

OJ 0.

0.

:Q

'5 ({)

::J ({)

Whisker growth by liquid metal particle catalyzed CVD _

- -----_._--_ ._------------------ -~-_ ..

-

(5

({)

Whis ker growth by solid metal part icle catalyze d CVD

---_.._._----

- _

Tubular growt h by solid metal particle catalyzed CVD

•.•. "

_ --_._

_- -'i

Figure 12. Phase transformations during fiber formation from the vapor phase (schematic). Far left: Fiber growth by laser assisted chemical vapor deposition occurs by a vapor-liquid-solid (VLS) mechanism if the tip is a melt, otherwise by a vapor-solid (VS) mechanism. Center left: Whisker growth by liquid metal particle catalyzed CVD occurs byVLS if the tipisa liquid.super-saturated metal alloy. Center right: Whisker growth bysolid metal particle catalyzed CVD occurs byVS if the tipisa solid metal particle or by VLS if itssurface is a eutectic melt Far right: Tubular growth bysolid metal particle catalyzed CVD occurs byVS. or byVLS if the solid catalyst particle forms a transient liquid melt surface.

65

Chapter 3

The differences and commonalties between the laser assisted and metal particle catalyzed CVD processes (Table IV) are readily summarized. (1) Depending on the fiber tip temperature, both the metal catalyzed and the laser-assisted process can yield amorphous, polycrystalline and/or single crystal fibers by VLS or VS phase transformation. (2) The laserassisted process can yield both continuous and discontinuous fibers and is therefore more versatile than the metal particle catalyzed process that is limited to short and discontinuous fibers. (3) The metal catalyzed process can produce freestanding microcoils, microtubes and nanotubes, while the laser assisted process can be used to yield microcoils, microtubes and complex microstructures. (4) The metal particle catalyzed process produces arrays orforests of thousands of individual whiskers or nanotubes on a common substrate. (5) At low or high pressures, the laser-assisted process yields only single filaments; the fabrication of fiber arrays ormultifilament yarns ispossible but would be difficult in practical terms. (6) The metal catalyzed. and the low pressure, laser assisted, processes are slow. (7) The high pressure, laser assisted process facilitates much higher growth rates than those possible in either the metal catalyzed orthe low-pressure laser processes. Table N . Metal catalyzed versus laser assisted chemical vapor deposition Metal Particle Catalyzed CVD Laser Chemical Vapor Deposition Molten metal Solid metal particles Above T,' of substrate Below Tg of substrate droplets VS transformation VLStransformation VLStransformation VS transformation Liquid phase (melt) Liquid No liquid phase No liquid phase (supersaturated) Single crystal Amorphous Single crystal Amorphous whiskers fibers fibers fibers Short whiskers Short fibers Continuous and Continuous and only only short fibers short fibers Spontaneous Spontaneous Designed growth of Designed growth of microtubes, coils microtubes, coils growth of coils growth of coils and tubes and tubes Arrays of Arrays of fibers, Single fibers, Single fibers, whiskers , but no but no simple / complex simple / complex microparts microparts microparts microparts Slow growth at Very slow Slow growth at low Very low growth at low reactor growth at low reactor pressures low pressures pressures pressures High pressure High pressure Much faster growth Much faster growth growth rates are growth rates are at high vs low at high vs low unkn own unknown pressure pressure 'In this table, TL denotes the liquidus temperature, and Tg the glass transition tempe rature.

3.2.2 Commercial hot filament CVD fibers The technology of manufacturing sheath/core bicomponent boron/ tungsten, boron/carbon, silicon carbidellungsten. and silicon carbide/ carbon fibers is 40 years old and the relationships between process variables, structures, and properties have been authoritatively described in important review articles. One article deals mainly with their preparation (33); another correlates process variables with structures [34). and one explores potential correlations between structures and properties (30).

66

(a)

Chapler3

Sheath/core boronltungsten fibers

Although this is mature commercial technology (see chapter 3.1 .2), several major new insights have been recently obtained [30] [34] with regard tothe relationship between surface uniformity and strength of boron/tungsten fibers, and that between the properties of large diameter boron/tungsten fibers and small diameter pure boron fibers. Commercial boron/tungsten fibers have high strength, high stiffness, and low density - in summary, high value-in-use. The average strength of a commercial product is 3.6 GPa [37]. The large fiber diameter (e.g., 140 IJm), however, is a disadvantage among inorganic reinforcing fibers. The latter have about 1/10th the diameter of boron/tungsten fibers. Using the same 15 IJm diameter tungsten core, the final diameter has been experimentally reduced from 140 to 55 IJm, whereby the density increases from 2.48 to 3.10 g/cm3 [34]. However, reducing the diameter by raising the tungsten/boron ratio has practical limits. It would increase materials cost, part weight, and further reduce specific strength (strength divided by weight density) and value-in-use. Table V. Polished and unpolished boron/tungsten fibers [after 35] Fibers As-produced boron!tllllgsten Avg. Strength, GPa COY, % Diameter, um 203 4.0 7 280 3.6 12 406 2.1 14 Average 3.2 11 Change Fracture Surface and core • COy - Coefficient of va riation

Slightly polished boron! tllllgsten Avg. Strength, Gpa COV, % 4.4 3.0 4.2 4.0 4.6 3.0 4.4 3.3 +37% - 70% Core only

Boron/tungsten fibers have a highly textured ("nubby") surface (Figure 8). The effect of surface uniformity of boron/tungsten fibers on tensile strength before and after a slight chemical polish (Table V) has recently been demonstrated [35]. The chemical polish raised the average strength by 37% from 3.2to 4.4 GPa, reduced the average coefficient ofvariation by 70% from 11 to3.3% and left the core as the only source offracture. The average strength of as-polished fibers (4.4 GPa) therefore excludes the surface initiated fracture ordinarily observed in as-produced fibers, and is representative only of the core-initiated fracture of as produced fibers.

(b) Sheath/core versus pure boron fibers A comparison of boron/tungsten fibers made by hot wire CVD and with pure boron fibers made by high pressure LCVD highlights commonalties and differences between the two processes, and establishes additional relationships between structures and properties. Both processes offer a hot surface toinitiate and sustain material deposition from the vapor phase. In the hot filament CVD process, large diameter boron/tungsten fibers grow by side growth on the surface of a hot tungsten or carbon filament. In the high pressure LCVD process, unsupported low diameter boron fibers grow by tip growth directly from the vapor phase inthe focus ofa hot laser beam. Commercial boron/tungsten fibers are, in practical terms, limited to fiber diameters of 100-140 IJm, and strength levels up to4.8 GPa. Pure boron fibers can be made with diameters of >6 IJm and a strength levels 7.6 GPa, i.e. , with 1.6x the maximum strength at 0.06-0.04x the diameter of the former. High specific properties (strength or modulus divided by density) are

Chapter 3

67

very important properties since nearly all transportation composites are weight sensitive. In these terms (Figure 13), average low diameter single component boron fibers were nearly as strong and stiff as VLS SiC whiskers. Their average specific strength was 1.2x that of commercial 1M carbon fibers, 1.7x that of commercial boron/tungsten fibers and 2.1 x that of commercial Nicalon SiC fibers. Their average specific modulus was 2.3x that of Nicalon SiC fibers, and comparable tothat of 1M carbon orboron/tungsten fibers. Finally, pure boron fibers grow by tip growth or lengthening and therefore have a smoother surface and higher strength than boron/tungsten fibers, which grow by side growth or thickening. Carbon whiskers made by metal particle catalyzed CVD, may serve as an analogy. Initially they grow by tip growth, but an additional carbon sheath is obtained by side growth, or thickening. Tip-grown carbon fibers are stronger than tip-grown carbon fibers with a secondary, side grown carbon sheath. Also, the temperature in the small laser focus is more readily controllable than that of a practically endless hot wire, i.e. another factor favoring tip over side growth. Specific modulus, 108 in 2.4

350

-

300

-

E

-'"

250 £ 0> c:

3

4

5

6

7

I

I

I

I

I Best

'0

'"0.

i

Avg.• 10Jlm 1M Carbon

•

200 l-

100

-

6

.

-

17 Jlm Nicalon SiC

• I

8

10

14

12

.S

10

~

T.

8

12

14

16

I

18

~

0. '"

4

I

o

(/J

I

£ 0> c:

en

6

I

s

CD

100Jlm Boronltungsten Avg. 140Jlm 55urn • Boronltungsten Boronltungsten

•

I

-

-

Best~

(/J

-

t

6 Jlm VLSfSiC

-

150

-

o

<25Jlm Boron

~

en .g

8

20

Specific modulus, 103x km Figure 13 Specific strength and modulus of boron fibers. Redrawn from F. T. Wallenberger and P. C. Nordine, Strong,small diameter boron fibers by laser assisted chemical vapor deposition, Materials Leiters, 14 [4J, 198-202

(1992).

(c) Sheath/core silicon carbide/carbon fibers

Advanced heat engines require a strong ceramic matrix to offer reliable performance, especially high impact. oxidation and creep resistance, at service temperatures exceeding

68

Chapter 3

1400·C. Two very different SiC fibers meet these requirements [35]. One of these fibers (see Chapter 10) is Nicalon, an amorphous fiber having a fiber diameter of 10-20lJm. It is commercially fabricated by thermal decomposition of a polycarbosilane precursor fiber. The other fiber, which isdiscussed here, isa polycrystalline ~-SiC fiber made by Textron. It has a sheath/core structure and afiber diameter of 1421Jm. The core consists ofa 37 IJm diameter graphite coated carbon fiber. The polymer-derived and essentially amorphous Nicalon silicon carbide fiber can be made with optimally small diameter but it lacks the desired stiffness, strength and strength retention at elevated temperatures. The CVD-derived and essentially polycrystalline silicon carbide/carbon fiber offers a belter balance between structural integrity, strength retention, and oxidation resistance above 1200·C [35], but it cannot be produced with small diameters. A new LCVD based single crystal silicon carbide fiber might, as proposed, offer a viable alternative [23]. As a single crystal fiber, itsmodulus of578 GPa and near theoretical strength would by far exceed the properties of polycrystalline and amorphous SiC fibers and facilitate excellent property retention above 1500·C. 3.2.3 Important CVI and PVD fibers Chemical vapor infiltration (CVI) of carbon fibers with silicon monoxide provides a textbook example (Figure 14) about the relationship between percent conversion of carbon to silicon carbide (aprocess variable) and fiber strength atelevated temperatures (aproduct variable).

4.50

o MER (100% converted, tested in air)

4.00

• MER (90% converted , tested in air) • ~.vCO (tested in argon) NICALON (tested in argon)

3.50

o

3.00
a. 2.50 (!)

----

£ "5> 2.00 c l!! Cii 1.50 .!11 'wc 1.00 ~ 0.50

0.00

~

"o,

.

.-----o------- --~ .D_. -

~~O~

-I---.---.-----r-~-_.-__,r__-r___;

800

900

1000 1100 1200 1300 1400

1500 1600

Test temperature. °C

Figure 14. Thermal stability of CVI-SiC fibers. Courtesy of Dr. M. Loutffy, Materials and Electrochemicals Research Corporation (MER), Tucson, AR.

A stoichiometric CVI silicon carbide fiber made by MER had a diameter of 10 IJm. Its tensile strength was 1.1 GPa both at room temperature and at 1400·C, and the fiber was stable to

Chapter 3

69

1500·C. A 90% converted CVI silicon carbide fiber had about the same diameter. Its tensile strength was higher than that ofthe 100% CVI SiC fiber (1.7 GPa) atroom temperature and at 1300·C, but the fiber lost strength above 1300·C. By way of comparison, a CVD-derived sheath/core SiC/C fiber made by Textron had a diameter of 140 urn. Its strength was 4.0 GPa at room temperature and 1.0GPa at 1400·C. It was weaker than the 100% CVI fiber at 1500·C. Nicalon, the fourth SiC fiber shown in Figure 14, was stable to about 1250°C, but lost strength above 1300°C. Boron nitride fibers made by chemical vapor infiltration of boron oxide with ammonia had low tensile properties at room temperature. The modulus ranged from 27 to 83 GPa, the failure strain was 1-2%, and the average strength ranged from 350 to 900 MPa. The fibers were amorphous or turbostratic boron nitride. Thermal cycling in an inert atmosphere to 2500°C did not affect the room temperature strength [31). Fibers produced from a borazine precursor were also amorphous orturbostratic, and had a little higher strength (1 .0GPa) and a modulus of78 GPa (30). The low strength and modulus of boron nitride fibers may be the result of microporosity. The presence ofmicroporosity has been indirectly verified (31) by demonstrating that the observed densities of individual turbostratic boron fibers were 5-20% lower than the densities calculated from electron diffraction do02 spacings ranging from 3.50 to 3.36 A. No direct evidence is available that porous boron nitride fibers are formed by CVI. A reference to LCVD-derived silicon nitride fibers (17) may be appropriate. There, solid fibers are formed with silicon-rich and stoichiometric compositions, and porous fibers with nitrogen-rich compositions. On the nitrogen-rich side of the Si-N phase diagram, ammonia is soluble in the melt, and rapid solidification of such a fiber after it is formed causes the development of a microporous structure. 3.2.4 Structure: property commonalties Many commonalties among structures and properties have already been discussed (Chapter 3.1). What remains to be discussed is a comparison of physical structures and mechanical properties obtainable only by those processes which are capable offorming continuous fibers, including laser assisted and hot filament CVD, as well as physical vapor deposition. (a) Straight, coiled, and tubular structures

All processes discussed in this chapter are actually or potentially capable of yielding continuous fibers. Some fibers have diameters which are preferred in the production of fiber reinforced composite (5-15 urn). Others, i.e., the sheath/core fibers, have non-optimum diameters inthis regard (100-140 ~m) and require selective composite fabrication techniques. Two processes are capable of yielding microsprings. They are laser assisted chemical vapor deposition of boron or silicon using a rotating goniometer, and hot filament chemical vapor deposition of diamond on an open tungsten wire coil. Several processes, inclUding conventional chemical vapor deposition and plasma enhanced physical vapor deposition, are capable of yielding sheath/core fibers, as well as hollow fibers or microtubes if a sacrificial core isused and then removed. (b) Fiber strength, modulus and toughness It is well worth reemphasizing that short vapor-grown whiskers and fibers, (Chapter 2), and continuous vapor-grown fibers (Chapter 3) have higher mechanical properties than the same

70

Chapter 3

fibers obtained from the liquid phase, e.g., a melt, or from the solid phase, e.g., a precursor fiber (Chapters 8 to 12). The following discussion analyzes the root causes. The key properties of any reinforcing fiber are modulus (stiffness), strength (uniformity) and mechanical toughness (damage resistance). Modulus reflects the structural order of a material and to a lesser degree internal or surface uniformity. Strength reflects internal and surface uniformity of a material rather than structural order (stiffness). High mechanical (versus fracture) toughness depends on a high work-to-break area under the stress strain curve, i.e., on high fiber strength, high break elongation, orboth. This relationshiphas already been discussed for whiskers inChapter 2. The strength level of fibers obtained from the vapor phase is generally higher than that of equivalent fiber made from the melt or precursor fibers, reflecting the fact that they are made directly from the vapor phase and/or by a containerless process. For example, single crystal CVD-SiC whiskers (7.5 GPa) are stronger than polycrystalline CVD-SiC fibers (7.5 vs. 3.5 GPa) and both are stronger than polycrystalline SiC fibers which are made from solid precursor fibers (1 .1-3.0 GPa). The highest modulus of a given substrate is obtained with a single crystal structure. Single crystal CVD-SiC whiskers (578 GPa) have a stiffer, more highly ordered, structure than polycrystalline CVD-SiC fibers (190-400 GPa), and sapphire whiskers and fibers (415 GPa) are stiffer than slurry spun polycrystalline alumina fibers such as Fiber FP (380 GPa). Superimposed upon this relationship is a compositional factor. Fiber modulus and structural order generally also decrease with increasing compositional complexity, e.g., silicon carbide is intrinsically stiffer than silicon oxycarbide such as Nicalon, and slurry spun alumina fibers are stiffer than sol-gel ormelt spun aluminate fibers. Mechanical toughness (work-to-break) predicts the damage resistance of a material before an initial fracture occurs. Fracture toughness predicts the residual damage resistance after an initial fracture has occurred. The mechanical toughness is very high for single crystal whiskers and fibers because their high strength results in a relatively high break elongation, and therefore in a relatively large area under the stress strain curve. 3.3 Selected products and applications

The reinforcing materials of choice for making high strength, lightweight, advanced composite materials for aerospace, industrial, commercial, and sporting goods markets are boron/tungsten and silicon carbide/carbon fibers.

3.3.1 81W and SiC/C fiber reinforced composites Boron/tungsten fiber applications include the use of filaments and of boron/tungsten fiber reinforced prepreg tape, aluminum matrix composites, and boron/graphite structures. The major applications for these structures are found in the aerospace market and about 25% in sporting goods markets [36). SiC/carbon fiber reinforced products include aluminum, titanium, and ceramic matrix composites. Major applications for these structures are also found in the aerospace market, minor uses in the industrial market [37). Key boron/tungsten uber reinforced epoxy tapes are used in the aerospace market in selected applications on the F-14 and F-15 fighters, the B-1 bomber, the Blackhawk series Sea Stallion

Chapter 3

71

helicopters, the Mirage fighter and the Space Shuttle. In addition, boron/epoxy tapes are used in the repair of aircraft structures, specifically for structural damage caused by fatigue, stress corrosion, and impact. In original and in repair applications, boron epoxy tapes are fatigue resistant, tailorable, easily applied, light weight and require no fasteners. Other important products for this market are boron/aluminum preform sheets, which can be consolidated via diffusion bonding. In the sporting goods market, the most important applications are golf club shafts, tennis rackets, fishing rods, skis and bicycles. Silicon carbide/carbon fibers are used to reinforce epoxy, aluminum, silicon nitride, and titanium. Silicon carbide/aluminum composites are found on tactical missiles and advanced fighters in rocket motor cases, projectile fins, and aft fuselage parts where they provide the burst strength at the weight of aluminum. In commercial markets, they serve as structural tUbing in the production of bicycles, and as cold plates in electronic packaging, where they offer a tailored coefficient of expansion and high conductivity. Silicon carbide/titanium composites are used in aerospace markets when very high in-use temperatures are required, e.g., inthe NASA Aerospace Plane (NASP) and in advanced fighters. In commercial markets, silicon carbide/titanium composites are used, for example, in gas turbine engines. SiC/aluminum nitride composites are used in advanced engines, heat recovery equipment and waste incineration systems. 3.3.2 Rapid evaluation ofnew fibers by LCVD The need for new structural and new sensor fibers is well known, but they cannot be made without extensive process and product research. In fact, each fiber would require its own lengthy process development before it could be evaluated in a minimum adequate composite or sensor application. Thus, the staggering cost of research precludes the development of new fibers. In this context, the HP-LCVD process has become an ideal tool for the rapid fabrication, without extensive process research, of test specimens of a variety of continuous length fiber candidates. These specimens can then be rapidly evaluated against the needs of a given end use long before a decision is needed if a new fiber should be commercialized by the LCVD process itself orby an adaptation ofanother fiber process. (a) Ultrahigh temperature fibers

The major continuous high temperature fibers in the market are sapphire fibers (Saphikon), aluminum silicate fibers (Nextel), boron/tungsten fibers (Textron), silicon carbide fibers (Nicalon), silicon carbide/carbon (Textron) and various types of carbon fiber. These commercial fibers have a range ofdiameters (10-140 um), high strengths (3.0-5.7GPa), high moduli (190-415 GPa) and high service temperatures (1200-1400·C). Nonetheless a survey ofthe literature reveals an expressed need [38-39] for new continuous fibers able to withstand prolonged exposure to even higher in-use temperatures (2000-2200°C) in new advanced aerospace composites [52]. Potentially useful bulk materials are listed in Table VI in descending order of melt and service temperature range from tantalum carbide to diamond. Their conversion into continuous ultrahigh temperature inorganic fibers defies incumbent process technologies but should pose no problem for the high pressure LCVD process. The upper service temperature (Table VI) is projected tobe 70% of the melting point (or liquidus) of a given material to account for creep, volatilization, and/or degradation . In these weight sensitive end uses, the design criterion for

Chapter 3

72

part stiffness is not the modulus itself but the specific (density corrected) modulus. The fiber modulus isexpected tobe the same as that ofthe bulk material. In summary, continuous sapphire fibers are commercially available, and new YAG fibers are readily achieved with the Saphikon process, orthe LHPG process (see Chapter 6), or else by the new containerless laser melt process (Chapter 4). Currently however, there is only one route known, l.e., HP-LCVD, that might eventually be capable of yielding continuous, single crystal fibers such as SiC ortitanium carbide fibers. A single crystal SiC fiber by LCVD has projected diameters of <15 urn (like Nicalon), a strength of7.5GPa (vs. 3.0 GPa for Nicalon), a projected modulus of 427 GPa (vs. 190 GPa for Nicalon), and a projected service temperature of 2000·C (vs. 1200 to 1400·C for Nicalon). The specific modulus of single crystal SiC fibers would be >2x that ofincumbent Nicalon SiC fibers. Table VI. Properties of bulk materials suitable for rapid evaluation by LCVD in the form of continuous, low diameter, single crystal fibers for ultrahigh temperature composites Bulk Properties

Melt temp.

Service temp .

Substrate Incumbents Graphite E-Glass Single crystals TaC zrC HfN TiN HfB, TiC HfC SiC HfO,

z-o,

AlP, Diamond

Young's modulu s g/cm"i) E,GPa'o

Bulk density

Specific modulus

Service temp ., T. vs. specific modulus, Eo

Eo

comparison

2.2 2.2 2.5

824

381.1

1050 (b)

<500'd) <1650") <450'Q

75

30.0

3880 (0) 3532 (0) 3305 (0) 3290 (0) 3100 (0) 3067 (0) 3000 (0) 2830 (0) 2765 (0) 2710(0) 2054(0) 3550" )

<2715") <2472(&) <2313(&) <2303(g) <2170") <2147") -2100 (g) <1981(&) <1942{g) <1897'g) <1434(g) <10001')

13.9 6.7 13.8 5.4 10.5 4.9 12.2 3.2 9.7 5.7 4.0 3.5

496 455 476 684 345 496 414 427 296 253 443 1000

35.7 67.9 34.4 63.5 65.1 100.1 33.9 123.4 30.6 44.5 110.8 284.1

3550" )

in air or uncoated in argon or coated very low T, - low Eo ultrahigh T, - low Eo ultrahigh T, - high Eo ultrahigh T, - high Eo ultrahigh T. - high Eo very high T, - high Eo very high T, and Eo very high T, -low Eo high T, - very high Eo high T, - low Eo high T, -low Eo high T, - very high Eo low T. - ultrahigh Eo

(a) Under pressure; (b) liquidus; (c) from the "Handbook of Chemistry and Physics", 77th Edition, D. R. Lide, Editor, CRC Press, N. Y, 4, 37-98, 1996-1997; (d) burns in air >500 ·C; (e) with suitable protective coating; (f) Tg; (g) 70% of melt temperature; (h) without protective coating burns in air >800·C, with suitable coating stable to 1000·Cand turns into graphite at 1000·C; (i) from E. L. Courtright, H. C. Graham, A. P. Katz and R. J. Kerans, "Ultrahigh Temperature Assessment Study - Ceramic Matrix Composites", WL-TR-91-4061, Material Directorate, Wright Patterson Air Force Base, Ohio , September, 1992.

73

Chapter 3

(b) High temperature sensor fibers The HP-LCVD process appears to be a valuable rapid fabrication tool to design and explore new fibers inthe field ofultra high temperature sensor systems before first developing a costly conventional commercial process. Using HP-LCVD as such a tool would facilitate the rapid evaluation of new and potentially useful compositions for emerging sensor applications where current industrial and aerospace design specifications exceed the capabilities of commercial sapphire and YAG fibers.

: Quartz (reference) ! Sapphire

0.12 · 4.5 0.14 - 6.5

I Tellurium

3.50 1.00 1.20 0.50 · 1.80 · 1.00· 0.40 0.25 .

Gallium arsenide ! Silicon : Zinc selenide Germanium Amorphous selenium Silicon carbide Diamond i

8.0 15.0 15.0 20.0 23.0 30.0 50.0 80.0

Wavelength, 11m

I L.

..

. ...

,- ,

--

-

-1-

,~ I -

.1

.

1-

100

10

_

Figure 15. Transmission windows of important electrooptic materials. Redrawn from F. T. Wallenberger, P. C. Nordine and M. Boman, Inorganic fibers and microstructures directly from the vapor phase, Composites Science and Technology, 5, 193-222 (1994).

Potentially useful single crystal HP-LCVD fibers include hafnium boride and tantalum carbide and have projected service temperatures ranging from 2170 to2715°C. Presently envisioned applications include the potential use of these fibers as consumable sensors to monitor rocket exhaust temperatures. Other HP-LCVD sensor fibers, including Si, Ge and ZnSe, (Figure 15), promise to offer high value in premium automotive and medical sensor systems. Single crystal HP-LCVD germanium [20) and silicon carbide [21) fibers can now also become available for exploration. In summary, the Hp·LCVD process is an ideally suited tool for the rapid fabrication and evaluation, without extensive process research, of test samples of potentially new fiber candidates for structural and sensor uses. 3.3.3 Rapid prototyping ofmicroparts by LCVD Rapid prototyping is a concept whereby a "design" of a part in the form of a CAD file can be used to generate the finished part in net or near-net shape in a single manufacturing process within a relatively short period of time [40). These methods share a common feature, Le., a significant reduction in process and product iterations during the development phase and a major reduction in development time from inception to commercialization. A comparison of rapid prototyping methods isshown in Table VII.

74

Chapter 3

(a) Evolution ofrapid prototyping Stereolithography is the first process where a prototype is produced directly from a CAD file [41]. Once a 3-dimensional design is completed, a data file is generated using one of many solid modeling methods [42]. The prototype is polymer based; the final part can be another material such as silicon. Fabrication of the part utilizes a laser beam in question to activate the reaction of a resin pool to initiate curing [43]. Selective laser sintering (SLS) uses solid polymer powders such as polycarbonates, investment wax or nylons and creates threedimensional prototypes from conventional CAD files by a method similar to that of stereolithography, except that solid powders are used, one cross section at a time [44]. Successive sintered layers are added until the object isinhand. Table VII. Recent Rapid Prototyping Processes Rapid Prototyping Process (Examples) Stereolithography Selective Laser Sintering X-ray Lithography Mechanical Prepolymer Deposition Conventional Chemical Vapor Deposition Laser Assisted Chemical Vapor Deposition

Prototype (Material) Polymer Resins Polymer Resins Polymer Resins Preceramic Polymers Inorganic Films Fibers and Microparts

Smallest Part Dimension Centimeter Centimeter Micrometer Millimeter Micrometer Micrometer

X-ray lithography (L1GA) is a method for rapidly fabricating inorganic microstructures using xrays generated by synchrotron radiation machines [45]. The separate manufacturing steps are lithography, electroforming, and plastic forming. Microfilters and microconnectors can be fabricated with this technology. This process is limited to planar, essentially two-dimensional objects. Mechanical prepolymer deposition is another rapid prototyping concept that promises to become a tool for the fabrication of ceramic components [46]. In this process which is analogous to decorating a cake, a viscous preceramic polymer is extruded and deposited through a moving supply nozzle. The motion of the nozzle is controlled by digitized CAD files to form a complex shaped object, and heat is applied at the point of deposition to consolidate the material while preserving its shape.

(b) Laser chemical vapor deposition Laser assisted chemical vapor deposition yields freestanding low diameter fibers and complex shaped, 3-dimensional microstructures [2] [4] [9-11]. Thus, a CAD file can be developed and the gas chemistry and motion of the laser can be controlled by the digitized CAD files in the commercial fabrication of a continuous fiber, a microspring, a microsolenoid or another micropart. Numerous materials and geometries facilitate the fabrication of LCVD derived microdevices that operate by using coupled electrical, magnetic, mechanical and thermal fields. Combined structural actuation and position sensing can be performed using combinations of piezoelectric, magnetostrictive springs [18] and key shape memory alloy materials [47]. Structural thermal sensing can be performed using thermoelectric and shape memory alloys as well as micro heat pipes [48-49].

Chapter 3

75

(c) Photonic band-gap microstructures Three-dimensional periodic photonic band-gap microstructures (PSG's) of aluminum oxide represent a new class of materials capable of uniquely controlling radiation since they are able to entirely reflect electromagnetic radiation in a band of frequencies propagating in any direction [10]. The successful construction of 3-dimensional PSG materials by laser assisted chemical vapor deposition (see Chapter 3.1 .1 (d) and Figure 5) showed transition minima around 4 tetrahertz (75 IJm) and 2 tetrahertz (150 IJm) required for the precise control of the optical properties ofmaterials, including lasers without threshold [10]. (d) The future of vapor phase processing In the past decade, boron/carbon and silicon carbide/carbon fibers have consolidated their position as premium fibers in aerospace, industrial commercial and key sporting goods markets with an enviable balance of properties such as strength, modulus, density and coefficient of thermal expansion. They will continue to serve as the benchmark for outstanding performance in these markets. However, lower diameter fibers will be required with even higher performance, such as strength and stiffness, and with service temperatures between 1900 and 2200°C. These markets are neither accessible to single crystal sapphire fibers nor topolycrystalline silicon carbide fibers, including Nicalon and sheath/core fibers. In the past decade, laser assisted chemical vapor deposition has become a rapid evaluation tool for new low diameter high performance fibers such as boron. It will continue to serve in this function and therefore may afford single crystal silicon carbide and single crystal titanium carbide fibers which will have the potential for offering a next generation continuous fiber to these markets. In addition, laser assisted CVD has proven tobe capable of becoming a rapid prototyping tool for new and highly complex microparts. It will continue to serve in this function and also meaningfully contribute tothe fabrication ofmicroelectromechanical devices. Finally, in the past few years, experimental diamond fibers have become available. They point the way toward the development ofa completely new and extremely valuable technology in the foreseeable future. The resulting materials would be aimed at structural and sensor applications, i.e., markets requiring ultrahigh strength, ultrahigh specific modulus, ultrahigh hardness, unmatched thermal conductivity and ultrahigh acoustical wave velocity. REFERENCES [1] [2] [3] [4] [5] [6]

17] [8]

D. Bauerle, Chemical Processing With Lasers, Berlin, Springer Verlag (1986). F. T. Wallenberger, P. C. Nordine and M. Boman, Inorganicfibers and microstructures directlyfrom the vapor phase, Composites Science and Technology, 5.193-222 (1994). F. T. Wallenberger, Rapid prototyping directly from the vapor phase, Science, Vol. 267,1274-1275 (1995). H. Westberg, M. Boman, S. Johansson and J. A. Schweitz, Freestanding silicon microstructures fabricated by laser chemical processing, Physics, 73 [11], 7864-7871 (1993). S. Johansson, J. A. Schweitz, H. Westberg and M Boman, Microfabrication of three-dimensional boron structures bylaser chemical processing, J.Appl. Phys. 72 [12], 5956-5973 (1993). O. Lehmann and M. Stuke, Generation of three-dimensional, freestanding metal microobjects by laser chemical processing, Applied Physics, A53, 343-345 (1991). H. Westberg, Thermal laser assisted chemical vapor deposition, Acta Universitatis Upsaliensis, Comprehensive summary ofdissertations, No. 375, FaCUlty ofScience, University ofUppsala, Sweden (1992). L. S. Nelson and N. L. Richardson, Formation ofthin rods of pyrolytic carbon by heating with a focused carbon dioxide laser, Materials Research Bulletin, 7, 971 -976 (1972).

76

Chaprer3

(9)

O. Lehmann and M. Stuke, Three-dimensional laser direct writing of electrically conducting and isolating microstructures, Materials Letters, 21 ,131-136 (1994). M. C. Wanke, O. Lehmann, K. Muller, Q. Wen and M. Stuke, Laser rapid prototyping of photonic band-gap microstructures, Science, 275,1284-1286 (1997). J. L. Maxwell, J. Pegna, and D. V. Messia, Real-time growth rate measurements and feedback control of three-dimensional laser chemical vapor deposition, Applied Physics A,inprint (1998). F. T. Wallenberger and P. C. Nordine, Strong, small diameter boron fibers by laser assisted chemical vapor deposition, Materials Letters, 14[4]198-202 (1992). F. T. Wallenberger and P. C. Nordine, Pure, strong and uniform carbon fibers directly from the vapor phase, Science, 260, 66-68 (1993). F. T. Wallenberger and P. C. Nordine, New process grows high quality single-crystal fibers at high rates, Materials Technology, 8 [7/8),136-1 38(1993). P. C.Nordine, S.dela Veaux and F. T. Wallenberger, Silicon fibers produced byhigh pressure laser assisted chemical vapor deposition, Applied Physics, A57, 96-100 (1993). F. T. Wallenberger, R. J. Diefendorf, K. D. Frischknecht and P. C. Nordine, Novel carbon fibers by laser assisted chemical vapor deposition (LCVD) in Mat. Res. Soc. Symp. Proc., Materials Research Society, Vol. 349,51-59 (1994). F. T. Wallenberger and P. C. Nordine, Silicon nitride fibers produced by high pressure LCVD, Journal of Materials Research, 9 [3], 527-530 (1994). F. T. Wallenberger, Inorganic fibers and microfabricated parts by laser assisted CVD: synthesis and phase transformations, in Novel Techniques in Synthesis and Processing ofAdvanced Materials, J. Singh and S. M. Copley, eds.,The Metals and Materials Society, pages 251-260 (1995). F. T. Wallenberger and P. C. Nordine, Inorganic fibers and microstructures by laser assisted chemical vapor deposition, Materials Technology, 8 [9/10], 198-292 (1993). F. T. Wallenberger, Inorganic fibers and microfabricated parts by laser assisted chemical vapour deposition (LCVD): structures and properties, Cer. Intemational, 23, 119-126 (1997). F. T. Wallenberger, Inorganic fibers and microfabricated parts by laser assisted chemical vapor deposition: a new rapid prototyping tool, Journal ofMaterials Processing and Manufacturing Science, 3, 196-213 (1994). F. T. Wallenberger and Paul C. Nordine, Single crystal germanium fibers by laser assisted chemical vapor deposition, inpreparation (1998). F. T. Wallenberger, P. C. Nordine and S. C.dela Veaux, Strong silicon carbide fibers directly from the vapor phase: structures and properties, inpreparation (1998). R. L. Crane and V. J. Krukonis, Strength and fracture properties ofsilicon carbide filament, American Ceramic Society Bulletin, 54[2], 185 (1975). A. R. Bunsell and J.-O. Carlsson, Silicon carbide fibers, in Concise Encyclopedia of Composites Materials, Revised Edition, Anthony Kelly, editor, pages 253-257, Pergamon, Elsevier Science, New York (1994). G. H. Lu, P. G. Partridge and P. W. May, A Technique forthe manufacture of long hollow diamond fibres by chemical vapour deposition, Journal ofMaterials Science Letters, 14, 1448-1450 (1995). R. Gat, Composite fibers having a diamond surface, US Patent 5,439,740, August 8, 1995. K. Upadhya and W. P. Hoffman, Advanced composite microtubes for microelectro mechanical systems, Journal ofMetals, 46[5],54-56 (1994) Loutlly, private communication, MER Corporation, Tucson, AR. F. E. Wawner, Boron and silicon carbidelcarbon fibers, in Fiber Reinforcements forComposite Materials, A.R, Bunsell, Editor, Elsevier, Amsterdam, pp. 371-425 (1988). J. Economy and R. V. Anderson, Boron nitride fibers, Journal ofPolymer Science, Part C,19,283-297 (1967). Y. Kimura, Y. Kubo and N. Hayashi, High-performance boron nttride fibers from poly(borazine) preceramics, Composites Science and Technology, 51 ,173-179 (1994). J.O. Carlsson, Techniques for the preparation of boron fibers - a review, J. Mater. Sci., 14255 (1979). A. M. Tsirlin, Boron filaments, in Strong Fibers, W. Watt and B. V. Perov, editors, pp. 155-200, North-Holland, Amsterdam (1985). J. A. DiCarlo, Fibers forstructurally reliable metal and ceramic composites, Journal of Metals, 37 [6]. 44-49 (1985). M. A. Mittnick, SCS-6 - Product Bulletin, Textron Specialty Materials, Lowell, MA. M. Mittnick, Boron Fibers - Product Brochure, Textron Specialty Materials, Lowell, MA. W. B. Hillig, in: Prospects for Ultrahigh-Temperature Ceramics, eds., R. E.Tressler et al.Plenum Press, New York, NY, 697-712 (1986). E. L. Courtright, H. C. Graham, A. P. Katz and R. J. Kerans, Ultrahigh temperature assessment study ceramic matrix composites, Final Report, WL-TR-91-4061, Materials Directorate, Wright Laboratory, Wright Air Force Base, OH, September 1992. P. Eyerer, P. Eisner, B. Wiedemann, F. Baumann and B. Keller, Rapid prototyping, new methods, Kunststoffe, 83[1 2], 949-955 (1993).

[10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27) [28) [29] [30] [31) [32) [33) [34] [35] [36] [37) [38) [39) [40]

Chapter 3

77

[41] H. Heinzmann, Stereo-lithography - the fast way from 3D CAD model to prototypes, Kautschuk & Gummi Kunststoffe, 46[1], 19-21 (1993). [42] A. Manthiram, D. Bourell and H. Marcus, Nanophase materials in solid freeform fabrication, JaM, 45[11], 6670(1993). [43] K. Bartels, R. Crawford, S. Das, S. Gudur, A. Bovik, K. Diller and S. Aggarwal, Fabrication of macroscopic solid models of 3-dimensional microscopic data byselective laser sintering, Journal of Microscopy, Oxford, 196,383-389 (1993). [44] N. Vail, J. Barlow, J. Beaman, H. Marcus and D. Bourell, Development of a poly(methyl methacrylate co Nbutyl methacrylate) copolymer binder system, Journal ofApplied Polymer Science, 52[6], 789-812 (1994). [45) E. P. Beyeler and S. I. GUgeri, Thermal analysis of laser-assisted thermoplastic matrix composite tape consolidation, J. HeatTransfer, 113,424-430 (1991). [46) C. Decker, Photo-initiated curing ofmultifunctional monomers, CHIMIA, 47 [10], 378-382 (1993). [47] R. E, Newnham, D. P. Skinner and L. E. Crass, Connectivity and piezoelectric-pyroelectric composites, Materials Research Bulletin, 13, 525-536 (1978). [48] A.B. Duncan and G. P.Peterson, A review ofmicroscale heat transfer, Applied Mechanics Review (1994). [49] G. P. Peterson, A. B. Duncan, and M. H. Weichold, An experimental investigation of etched micro heat pipes fabricated insilicon wafers, ASME Jour.ofHeat Trans.,115,751-756 (1993). [501 J. Liu, A. G. Rinzler, H. Dai, J. H. Hafner, R. K. Bradley, P. J. Boul, A. Lu, T. Iverson, K. Shelimov, C. B. Huffman, F. Rodriguez-Macias, Y. S. Shon, T. R. Lee, D. T. Colbert and R. E. Smalley, Fullerene pipes, Science, 280,1253-1255 (1998). [51J J. Liu etaI., Nature, 385,780 (1997). [52J D. W. Johnson, Ceramic fibers and coatings, advanced materials for the twenty-first century, Pubication NMAB-494, National Academic Press, Washington DC (1998).

SECTION III FIBERS FROM THE LIQUID PHASE F. T. Wallenberger with achapter by H. Ackler and MacChesney Many continuous inorganic fibers can be formed directly from the liquid phase including melts (Chapter 4) and solutions (Chapter 5). Ceramic aluminate fibers are indirectly derived from a viscous liquid phase, which includes dispersions and sol-gels. These fibers aswell as carbon fibers and silicon carbide fibers are initially obtained asnonfunctional precursor fibers. Since the final functional fibers are actually derived from solid precursor fibers, they will be covered inChapters 8to10dealing with advanced inorganic fibers derived from the solidphase.

Contents 4

CONTINUOUS MELT SPINNING PROCESSES 4.1 Important melt forming processes 4.2 Forming glass fibers from strong melts 4.3 Forming glass fibers from fragile melts 4.4 Forming amorphous fibers from inviscidliquids 4.5 Growing single crystal fibers from inviscidmelts

5

CONTINUOUS SOLVENT SPINNING PROCESSES 5.1 Dry spinning ofsilica glass fibers 5.2 Silica fibers byother processes

6

STRUCTURAL SILICATE AND SILICA GLASS FIBERS 6.1 General purpose silicate glass fibers 6.2 Special purpose silicate glass fibers 6.3 Non-round, bicomponent and hollow fibers 6.4 High temperature silica glass fibers

7

OPTICAL SILICA FIBERS (H. Ackler and J. MacChesney) 7.1 Introduction 7.2 Principles ofopticallransmission 7.3 Fabrication ofoptical fibers 7.4 Fiber drawing process 7.5 Sol-gel processing 7.6 Applications ofoptical fiber devices 7.7 Summary and outlook

CHAPTER 4 CONTINUOUS MELT SPINNING PROCESSES F. T. Wallen berger The behavior of fiber forming inorganic melts is well understood [1-2]. They are viscous or inviscid [3-4], i.e., have high orlow viscosities, and fiber forming processes are either very fast (>1000 m/min,) when continuous amorphous glass fibers are desired, or slow «0.1 m/min) when continuous single crystal fibers are desired.

4.1 Important meltforming processes Continuous fibers can be formed from viscous (high viscosity) melts [2] [17], and from inviscid (low viscosity) melts [11] [19]. The design of a viable fiberizing process from either melt depends primarily on three important factors: (1) the relationship between melt viscosity and temperature, above and below the fiber forming temperature, (2) the liquidus temperature, the highest temperature at which crystals can form , (3) the nature of the crystalline phase at and below the liquidus and the crystal growth rate.

4.1 .1 Principles offiber formation Continuous glass fibers can be formed when they have (1) a melt viscosity of 316 (orlog 2.5) to 1000 (orlog 3.0) poise, (2) a relatively high process speed, (3) a fiber forming temperature that is at least 50°C above the liquidus temperature, and (4) a low nucleation and crystal growth rate when crystallization occurs [19]. (5) The temperature difference (L1T) between the forming and liquidus or crystallization temperatures provides a safety margin assuring so that cold spots in the process will not induce localized crystallization. (6) Continuous glass fibers can be formed below their liquidus temperature from super-cooled melts. Continuous oxide glass fibers can be formed from high and low viscosity melts and benefit from high process speeds, low crystallization rates, or both. Continuous single crystal oxide fibers can be formed only from low viscosity melts, and their formation requires a highly directional crystallization process and very slow crystal growth. Continuous cryogenic fibers can be formed byextruding solidified gasses, e.g., hydrogen, argon, nitrogen, and deuterium at 10 to30K through suitable spinneret orifices.

(a) Behavior of viscous melts In the technical literature, the relationship between melt viscosity and temperature is traditionally expressed bycomparing the increase in log viscosityofa fiber forming melt with a linear decrease of its temperature. In the typical fiber forming range between log 2.5 and log 3.0 poise, this relationship yields a nearly straight line for melts of commercial interest, and readily facilitates a comparison of melts from different compositions. This representation

82

Chapter 4

unfortunately minimizes important differences at viscosities below log 2.5 poise, or for those temperatures which are ordinarily just beyond those of immediate interest in a commercial process.

Table I. Fiber formation from viscous and inviscid melts Type of melt

Melt viscosity LT" >MP"

Glass &:lZllass ceramic . fibers ..... Strong high high

Fiber substrate

Strong

high

high

.....

Fragile

high

low

.....

Fragile

high

low

.....

Inviscid

.....

.....

Inviscid

.....

.....

very low very low

pure solid silica silicate melts yttriamod. silicates fluorides & aluminates YAG, yttriaaluminates binary aluminates

Metal ribbons & metal fibers ..... ..... very Inviscid low ..... ..... very Inviscid low

metal alloy ribbons steel, other metal fibers

Generic fiber forming processes

Environment

downdrawing >LT downdrawing >LT downdrawing MP

from preforms from bushings from precision

bushinzs/ tins from supercooled melts levitated laser heated melt by jet surtace stabilization

ambient air ambient air amb ient air ambient air containerless reactive gas

extrusion >MP melt spinning >MP

by rap id solidification by jet surtace stabilization

cold quench wheel reactive gas

staI ceramic fibers

very low very , other fibers >MP fiber owth low "Liquidus temperature (LT), melting point (MP), rapid solidification (RS)process

In the scientific literature, the relationship between melt viscosity and temperature is often expressed by comparing the increase in log viscosity of a fiber forming melt with the glass transition temperature (Tg) divided by melt temperature (T), both in Kelvin (K). Among the viscous mells, this approach distinguishes between strong and fragile melts (2). Silica, the ideal strong melt, exhibits a perfect straight line relationship. Other strong melts, such as those of silicates exhibit a near straight line relationship. Fragile melts of certain aluminates or fluorides deviate noticeably from the ideal straight line behavior. The glass transition temperature, however, is nota process-related factor and this relationship is not useable in this context. The most important factor in the analysis or design of a process is the liquidus temperature (TL). The original concept that distinguishes strong from fragile melts (2) is valuable, and it can be modified by focusing on the liquidus temperature (TL), rather than on the glass transition temperature (Tg) . Thus, a linear change in mell viscosity (in poise or Pa.s) is compared, without normalizing for reciprocal temperature, with a linear change in melt temperature (in ·C orOF). Accordingly, a strong fiberforming melt has a high viscosityabove

Chapter 4

83

and below the liquidus temperature, specifically eXhibiting only small changes in viscosity above the liquidus temperature. And a fragile melt has a relatively high viscosity below the liquidus temperature [3-4] [9] [17] but a low viscosityabove the liquidus temperature. E-glass, is a commercial general-purpose fiber. It isderived from a strong silicate melt, has a log 3.0 (poise) fiber forming temperature of 1185°C, a liquidus temperature of 1050°C, and a ~ T between forming and liquidus temperatures of 135°C(Figure 1, #1). Because of its high ~ T, this melt isconsidered to be not only a strong but also a long melt. S-glass (Figure 1, #5), a commercial specialty fiber in the familyof high strength (HS) fibers is derived from a strong but short silicate melt. It has a log 2.95 (poise) fiber forming temperature of 1565°C, a liquidus temperature of 1500°C, and a ~ T between forming and liquidus temperatures ofonly 65°C(Figure 1, #2).

-

-----

10,000

LT FT FV .---

9900 9800 \ 1100-7000 C;;

cil

0-

II Q) til

1000 800 700

~

600

Q) til

500

0.

400

'0

~ 'iii 0

u

s

til

.

0 0

900

'0

0. 0

Liquidus temperature Fiber forming temperature Fiber forming viscosity Conventional melt Supercooled melt

.

0 0

1

FT\\FT ..

.. 2l.. .\ 3

j

.0

·· 6 ·.

~

0

4

0

FT

3.0113.886 3.000

0.

FT\

o'

3.991

'0

0 0

0

3.995

Q) til

0

~

FV

. ··

4.000

~ 'iii 0 o

til .:;;

01 0

...J

2.500

300 200 100

2.000

0

0.000

1100

1200

1500

1600

Temperature, °C

Figure 1. Relationshipbetween linear viscosity (10 poise = 1Pa.s) and temperature. Glass fibers requirea fiber forming viscosity (FV) ranging from log 2.5to log 3.0poise. The corresponding temperaturesare the fiber forming temperatures (FT). Glass fiberswhich are formed from strong viscous melts include Eijlass(#1) and Sij lass (#5). Their liquidus temperatures (LT), the highest temperatures atwhich crystalscan form are ::::50°Cabove their forming temperature. Glass fibers which are formed from frag ile viscous melts which have useful forming viscosities only below their liquidus temperatures include an yttria modified silicate fiber (#2) and a quaternary aluminate fiber (#3). The viscosity of these melts must be raised by supercooling before fiber formation can commence. Glass fibers which can be formed frominviscid melts include a binary aluminate (#4) and a YAG melt (110), but special viscositybuilding fiber forming processes are required toraise theirviscosity from <1poise to>300 poise.

84

Chapler4

Only experimental fibers have been produced from fragile melts. Yttria modified silicate glass fibers (Figure 1, #2) have fragile and short melts [19]. They have a log 3.0poise fiber forming temperature of 1280°C, a liquidus temperature of 1390°C, and fibers can be formed 110°C below the liquidus temperature of the melt, but the molten jet must be efficiently cooled in an especially designed bushing and cooling tower. Quaternary aluminate glass fibers hsve also fragile and short melts [8-9). These melts have a liquidus temperature of 1350°C, and fibers can be formed at 1300°C, i.e., below the liquidus temperature, by updrawing them from a super-cooled and carefully stabilized melt (Figure 1, #2). When crystallization occurs with E-glass melts, the rate ofcrystal growth is low (Figure 2, #1). The crystal growth rate of S-glass mells is somewhat faster (Figure 2, #2). When crystallization occurs with a modified silicate melt (Figure 2, #3), the crystal growth rate is higher than that ofS-glass mells and with yttria modified melts it isnearly uncontrollable. Thus, fibers are best formed from strong silica and silicate melts by down-drawing them from preforms or from conventional bushings. Glass fibers can be formed from fragile melts, but they must be super-cooled to a viscosity in the log 2.5 and log 3.0poise range. Glass fibers 8.0 6.0

~ 4.0 :1.

J 2.0

o/

1

1100

1200

1300

1400

1500

Temperature, °c Figure 2. CrystallizaUon rates offiber forming melts. (1) Commercial E-glass melt, (2) commercial S-glass melt, (3) experimental silicate glass melt, (4) yttria modified silicate glass melt. Redrawn from V. E. Khazanov, Y. I. Kolesov and N. N. Trofimov, Glass fibers in Fibre science and technology, pages 15·230, Chapman and Hall, London (1995).

(b) Behavior ofinviscid melts The most common inorganic materials including oxides and metals are polycrystalline, and have high and well-defined melting points. Their melts are inviscid above the melt temperature and their melt viscosities are comparable to that of light to heavy motor oil at

Chapler4

85

room temperature. When they solidify they will almost instantly revert topolycrystalline solids. Crystallization can be prevented by increasing the quench rate without increasing the viscosity or by increasing the viscosity without increasing the quench rate. Some oxides and metals can be rapidly solidified to yield glasses by increasing their quench rate. For example, continuous metal glass ribbons can be rapidly quenched on the surface of a cold quench wheel [59), but no self-supporting glass or metal fibers have as yet been made by rapid solidification by increasing the quench rate from 104 to 1061<1s. However, selfsupporting continuous binary aluminate glass fibers [11) and YAG glass fibers [73) can be formed from their inviscid melts by increasing the viscosity of the respective inviscid jet (Figure 2, #4 and #6)) without increasing the quench rate. Whether the process for forming solid fibers (or ribbons) from inviscid melts is fast or slow, i.e., whether amorphous or single crystal fibers are formed, it proceeds through a transient viscous range. The inviscid melt will solidify as a glass fiber when a transient viscosity of log 2.5-log 3.0 is reached as required for formation of any fiber that is obtained from the liquid phase ormelt [12).

(c) Generic fiber forming processes In principle, five generic fiber-forming processes are known to yield glass or glass ceramic fibers from strong, fragile and inviscid melts. They are (1) downdrawing fibers from preforms, (2) drawing fibers from bushing tips, (3) updrawing fibers from their melts (4) extruding fibers into a chemically reactive environment and (5) forming fibers in a containerless laser melt process. These processes are schematically shown inFigure 3. The fiberglass and fiber optics industries are based on strong silica and silicate melts. Strong melts such as these have a relatively flat temperature-viscosityrelationship above and below the liquidus. Structural silicate glass fibers having fiber forming viscosities at temperatures ranging from 1200 to 1500°C are spun from the melt through bushing tips (or orifices) and wound on a winder (Figure 3, top left). Structural and optical silica fibers having very high melt viscosities and fiber forming temperatures >1500°C are downdrawn from the surface melt of an appropriate preform (Figure 3, top right). For details regarding fibers from strong melts see Chapter 4.2. Alumino-silicate glass fibers can be drawn from their fragile melts in an otherwise conventional fiber drawing process (Figure 3, top, left), but with extremely careful control of the cooling temperature [19]. Bicomponent heavy metal fluoride glass fibers can be drawn from their supercooled fragile melts (Figure 3, top, left) which are contained in a double crucible bushing [7). Quaternary calcium aluminate glass fibers can be updrawn from their supercooled fragile melts (Figure 3, bottom, left) which are maintained ata carefully controlled temperature [8-9). These fibers can also be downdrawn from preforms (Figure 3, top right) but stringent process control is required [41]. For detailsregarding fibers formed from fragile melts see Chapter 4.3 Continuous glassy metal ribbons can be formed with high quench rates from their inviscid melts by a rapid solidification process [60) that is akin to a generic bushing process (Figure 3, top left), except that the extruded ribbon must be rapidly cooled on the surface of a cold quench wheel. Continuous aluminate glass fibers and metal wires [10-12) and continuous amorphous YAG fibers [731 can be melt spun from inviscid melts by increasing the jetlifetime

86

Chapter 4

without increasing the quench rate. The former can be formed from the inviscid jet by chemically stabilizing its surface (Figure 3, bottom right) and the latter can be formed from the inviscid, argon-levitated, laser-heated melt in a containerless process (Figure 3, center). For details regarding these fibers see Chapter 4.4. Continuous single crystal oxide fibers can be grown from their inviscid melts byrelatively slow updrawing processes (Figure 3, bottom left) such as laser assisted [13] or flux assisted [14] crystal growth. For details regarding growth ofsingle crystal fibers see Chapter 4.5.

B

Preform Ring Heater

Windup

D

E

Melt

+

I-

NOZZI~bbon I

Inert gas

F

Laser \ I Focus- - \ I

I

Melt - -

Quench I \ wheel: I

.

,

(Enlarged)

,

I

.

\

'Ribbon

N~"'~

Fiber- -

Windup

tJ

Quench wheel

Figure 3. Generic fiber forming processes. Drawing fibers from a bushing (A), downdrawing fibers from a preform (B), updrawing fibers from a supercooled melt (C), rapid solidification of a metal ribbon on a quench wheel (D), extruding fibers from a bushing into a chemically reactive environment (E), and drawing fibers from an acoustically levitated, therefore, containerless melt (F). Redrawn and expanded from F. T. Wallenberger and S. D. Brown, High modulus glass fibers for new transportation and infrastructure composites and for new infrared uses, Composites Science and Technology, 51 , 243-263 (1994).

87

Chapter 4

4.1 .2 Structure ofmelts and fibers Fiber forming melts may appear to be homogeneous, but are often immiscible mixtures of a primary major phase with at least one secondary nano-, rnicro-, or even macrodomain [5J. The resulting fibers may appear to be amorphous (literally featureless) when examined by xray analysis, but sensitive tools, e.g., small angle x-ray scattering [57J or NMR [15J often reveal ordered domains from ambient temperature to 2000 K [15] in solid fibers and melts. A description of immiscible domains [5J in terms of coordination numbers, network formers, network modifiers and intermediates is possible, but remains devoid of physical meaning as long as their physical dimensions cannot be tied tomechanical properties. (a) From melts tofibers

This lack of technical information is regrettable, especially since the structure of a melt not only predetermines the structure ofthe resulting fiber [4] but also its mechanical properties. A conceptual overview is inorder. In a strong melt, any secondary domain, nanodomain or any other form of order such as nano-or microcrystals, will dissolve gradually with increasing temperature and decreasing viscosity. In a fragile melt any secondary domain seems to dissolve gradually below the liquidus, but then more rapidly above the liquidus temperature, thereby causing the discontinuous change in melt viscosity with temperature above the liquidus temperature that isobserved. In an inviscid melt, a highly crystalline material is instantly transformed into a highly fluid melt with a very low viscosity, t.e., generally <1 poise. In such a melt, there is no, or almost no, secondary nanodomain left at the melting point, which incidentally coincides with the liquidus temperature, since the viscosity remains Virtually unchanged when the melt temperature is further increased. On cooling, an inviscid melt crystallizes instantly when it reaches the melting point or liquidus temperature, and transforms toapolycrystalline solid.

C=>c=:tc=o<=JoC=>C:=:=>C=»c:::=>c==>C=> C:::=>C::>C=>C::::=>C=>C:::=:>c::::::>c::::::>C:::=> c::::::>c:::>c=>~c=>c::::::>c::>c=:oc::::::>c::::::>

<::>c::>c::>c::::::>c:::>c::>c=>c::>c=> A c=:o c=> c::::::> c:::=> c::::::> c=> c::::> c=o c=:> c::::::> C=>C::>C:::=>C=>C:::=:>C::::::>c:=>c=><=>

c:::>c:::=>c::::::>c=>c::>c:=>c:::=>c:=>c:=> C::::::>C:=:::OC::>C:::=>C=::>C::>C:=:=>c=>C:::> c=»c:::=>c=>c:=><=>e::::::-c:::=>c:=>c:=>

Melt flow

Figure 4. Schematic relationship between the melt uniformity and the structure of resulting fibers. (A) Represents a melt with a uniform second phase flowing horizontally in a fumace toward the bushing, and a glass fiber formed from this melt. Itsrelatively uniform structure can beenvisioned to have high strength. (B) Is a melt with a non-uniform second phase, and a resulting fiber which can bereadily envisioned to possess low strength.

88

Chapter 4

Alternatively, an inviscid melt can be rapidly quenched or solidified to yield an amorphous glass. Analogously, an inviscid jet, the molten precursor ofa fiber, can be rapidly supercooled and stabilized, under certain conditions, atafiber forming viscosity oflog 2.5 tolog 3.0 poise. In a stationary strong, fragile or inviscid melt, the nanodomain(s) are most likely spherical, and have dimensions ranging from relatively small to relatively large. In a dynamic melt, as for example that in a commercial fiber forming process, the nanodomains will most likely directionally elongate as the melt flows toward a bushing (Figure 4). When fibers are pulled from the tiny orifices orbushing tips, the lId ratio of the second phase domains will most likely further attenuate and, upon solidification, form a directional, second solid phase. A composite fiber structure that results may even appear as if it were homogeneous and amorphous (featureless) by older instrumental analysis including x-ray diffraction. Newer experimental methods, such as low angle x-ray scattering, will yield quantitative and confirmative answers In any event, a uniform second phase in a typically inhomogeneous oxide glass melt is expected to yield a more uniform, Le., stronger fiber than a non-uniform second phase, and the lId ratioofthe secondary domain may additionally also affect the fiber modulus. Specifically, silica melts and fibers have a uniform anisotropic network structure. The uniformity of this network structure causes a linear increase of the log melt viscosity with decreasing melt temperature. The anisotropy of the network structure, however, affords a relatively low fiber modulus. Two examples, admittedly extreme cases, may conceptually document the effect of disrupting the uniform anisotropic silica network structure by the addition ofother oxides. For example: (1) the uniformly anisotropic silica structure can disrupted by the addition of alumina and magnesia, i.e., by the formation of a ternary eutectic, Si02-AbOrMgO. Melt viscosity, liquidus temperature and forming temperature drop, the crystallization potential at the liquidus and fiber modulus increase, and a high strength, high modulus HS-glass fiber solidifies. (2) The relatively uniform structure of a ternary high strength SiOrAb03-MgO fiber can be further disrupted by the addition of CaO and either 5-20% boron oxide or 16-17% lithium oxide In both cases, the formation of a secondary, liquid or nanocrystalline, nanophase, is most likely "the initial stage ofstructural ordering" [57). This ordering process stops when the fiber solidifies. No pertinent information is available for melts ofhigh strength (HS) glass fibers, but the formation ofa submlcro-heteroqeneous [57) melt structure has been documented by small angle x-ray scattering for a specific borosilicate composition. It increases with decreasing temperature or increasing viscosity. In summary, the structure of a melt predetermines the structure of a given fiber as well as of itsproperties. Amorphous glass fibers, nanocrystalline glass-ceramic fibers, polycrystalline ceramic fibers, and single crystal fibers possess different levels ofstructural order and uniformity, therefore vastly different mechanical properties. (b) Fiber structure versus modulus

The modulus or stiffness of a fiber reflects its structural or internal order [3]. Figure 5 compares the fiber modulus of silica-alumina and calcia-alumina based composition ranging from 100% silica (or calcia) to 100% alumina. The fiber modulus increases with increasing structural order, Le., from 41 to 125 GPa for amorphous glass fibers to 125 to 250 GPa for nanocrystalline glass fibers, 250 to 400 GPa for polycrystalline ceramic fibers, and 405 to 410

89

Chapter 4

GPa for single crystal alumina fibers. The overall trend reflects the weight percent increase in alumina. Differences in each category reflect super-imposed effects of other oxides (not shown) orprocess-induced differences. Squares represent fibers drawn from strong silica and silicate melts having 0-25% alumina. Triangles represent fibers drawn from fragile aluminate melts having 30-45% alumina, as well as a fragile yttria-modified silicate melt. Inverted triangles represent aluminate fibers spun from inviscid melts having 54-81% alumina. Circles represent sol-gel orslurry spun fibers with 64-99% alumina and single crystal sapphire fibers (100% alumina) made by slow crystal growing processes. Oxide fibers 500-,.---------------------, Single crystal fibers

Saphikon

()

400-+-----------------------1

Fiber-FP ()

Polycrystalline fibers

fu

CIl

en

A/max

300-

Safimax

()

0

:J

"5 'C

o

E

.9l 'w c: ~

o Nexte/480

DSialon-2 Nanocrystalline fibers

200-

Nexte/440

0

OSialon-2

m

100

[]AO

\l 0

Sialon-1

61MS-81

1\ RIMS-54

Y-10

_ .JZ-OO

OA/tex

6Z-32 60-46

OS-Glass DE-Glass

6IMS-54

.6,IMS-80

Amorphous fibers

0 - + - - - -I - , - - - - - , - - - - -I , - - - - - ,I - - - - - 1 o 20 60 80 100 Alumina, wt. %

Figure 5. Fiber composition versus modulus. This illustration correlates the increase in alumina content in the silica-alumina and/or calcia-alumina system with the measured increase in fiber modulus, and observed form of crystallinity. Redrawn and enlarged from F. T. Wallenberger, The structure ofglasses, Sdence, 267, 1549 (1995).

90

Chapter 4

E-glass and S-glass are generic names for two commercial glass fibers. Astroquartz (AO), Nextel, Altex, Safimax, Fiber FP and Saphikon are tradenames for commercial fibers with compositions ranging from 100% silica to 100% alumina. Among the experimental fibers, Z refers to zinc oxide modified silicate and aluminate glass fibers. Z-OO has no alumina, Z-32 has 32% alumina. Sialon refers tooxynitride or nitride modified silicate glass fibers, 0-46 toa quaternary non-silica aluminate glass fiber with 46% alumina, Y-10 toa silicate glass fiber with 10% yttria. IMS refers to inviscid melt spun aluminate glass fibers with 54, 80 and 81% alumina, respectively; and RIMS refers toa redrawn IMS fiber with 54% alumina. AQ, a silica glass fiber, has the lowest modulus (69 GPa) among the silicate fibers derived from strong melts (Figure 5) because it has a highly uniform but anisotropic network structure. Incorporating 10% alumina into itsnetwork structure yields E-glass (aborosilicate fiber) with a modulus of 72 GPa, and incorporating 15% alumina yields S-glass (a magnesium aluminum silicate fiber) with a modulus of 84 GPa. While deferring an analysis of nitride (Sialon), zinc oxide (Z) modified compositions, and inviscid melt spun (IMS and related) glass fibers, the modulus of the other fibers shown in Figure 5 continue to increase with increasing alumina content. This can be seen from 0.46, a quaternary aluminate glass fiber with 46% alumina (110 GPa) which is based on a fragile melt, to those of ceramic fibers ranging from Nextel (140 GPa) toSaphikon (410 GPa) which can no longer be melt spun. Generically, alumina is a powerful modulus modifier. Specifically, alumina increases the crystallization potential of a given melt and therefore the internal order and the modulus, and that of the resulting fibers. Glass fibers with very high moduli become glass ceramic fibers. In ceramic fibers, alumina continues to increase the modulus by the same mechanism and the morphology changes from polycrystalline to single crystal. Superimposed on the effect of alumina are the contributions of other modifiers as shown in Figure 5, i.e., that of nitride (Sialon), yllria (Y), and zinc oxide (Z), and that of inviscid melt spinning (IMS) and related processes. Yttria is a more powerful modulus modifier than ZnO, whereby 10% yttria, as in Y-10, raises the modulus ofa typical S-glass composition from 85 to 130 GPa. All Sialon fibers (Figure 5) were melt spun. Those having moduli ranging from 125 to 140 GPa are known to have an amorphous, glassy structure. Sialon fibers with moduli ranging from 140 to 170 GPa are known to have a nanocrystalline, glass ceramic structure. The dramatic increase in modulus (from 125 to 248 GPa) at about the same silica and alumina level is known to be due to a proportionate increase in surface tension rather than crystallinity. Sialon fibers with the moduli ranging from 170 to 248 GPa are therefore most likely glass ceramic, and not polycrystalline ceramic fibers Inviscid melt spinning (IMS) from aluminate melts having viscosities of <1 poise gave amorphous glass fibers with compositions ranging from 54 to 80% alumina (Figure 5). These fibers, therefore, retained the random order of the inviscid melt from which they had been spun. A slight increase in alumina to 81% (IMS-81) increased the internal order, and gave nanocrystalline glass fibers with a modulus of 170 GPa. Compositions with >81% alumina could no longer be processed by inviscid melt spinning [17]. Specifically, IMS-54, a fiber with 54% alumina had the lowest modulus (44 GPa). Redrawing IMS-54 at a high temperature [171 yielded a nanocrystalline glass fiber (RIMS-54) with twice the modulus. In summary, the crystallization potential ofalumina can be suppressed by rapid solidification and recovered by heat treatment

91

Chapter 4

(c) Fiber structure versus strength

Strength is a measure of the structural (internal) uniformity and surface uniformity. Table II shows the average tensile strength of sixteen glass fibers. Their strengths range from 5.57 GPa for a military optical glass fiber (FOG-M) to0.37 GPa for a highly porous, high silica fiber obtained byleaching E-glass with hydrochloric acid. All fibers have diameters ranging from 4 to20 urn (19), except FOG-M (62) and the binary calcium aluminate fibers [17], both of which have a diameter of>1 00 urn. Surface flaws or non-uniformities tend to reduce the strength of individual fibers, and directionality or spin orientation, as inferred from birefringence measurements, tends to increase the strength of individual fibers. No study, however, is known that relates spin orientation and/or surface uniformity of a wide range of fiber types, such as those shown in Table II, totheir relative fiber strengths. Surface uniformity and spin orientation are important factors but, even if their relative relationships were fully documented, the up to 15-fold differences instrength among the sixteen generic fibers shown inTable II cannot be attributed to these factors alone.

Table II. Yam tensile strength versus composition Generic glass fiber Military fiber optics glass fiber, FOG-M Magnesium aluminosilicate, 10% MgO Magnesium aluminosilicate, 15% MgO Ultrapure silica fiber, Astroquartz Zn/Ti magnesium aluminosilicate Sodium calcium aluminosilicate E-type aluminum borosilicate Copper aluminum borosilicate Borate glass fiber Lead silicate glass fiber Phosphate glass fiber Sodium silicate glass fiber Calcium aluminate, 54% CaO Pure silica fiber from waterglass Calcium aluminate, 80% CaO Porous silica fiber from E-glass

GPa 5.57 4.80 4.00 3.50 3.20 2.75 2.70 2.70 1.90 1.55 1.50 1.10

0.95 0.85 0.50 0.37

Reference [62] [19] [19] [19] [19] [19] [19] [19] [19] [19) [19] [19] [11)

[4) [11)

[4]

A major reason for the observed differences in fiber-to-fiber strength must be therefore be sought in differences in the uniformity of the internal structures. A uniform network structure such as that of FOG-M silica fibers translates into high tensile strength. A highly nondirectional (random) arrangement of calcium oxide and aluminum oxide in a rapidly solidified binary calcium aluminate fiber translates into very low tensile strength. The ultralow strength ofporous high silica glass isdue to its porosity. Furthermore, the differences in the strengths of amorphous glass. nanocrystalline glass ceramic and polycrystalline ceramic fibers are attributable to differences in the internal order of the fiber structure. For example, the presence of a uniformly distributed minor nanocrystalline (second) phase in a major amorphous (primary) phase of a glass-ceramic

92

Chapler4

fiber may not affect its strength while the weak interface bonding between the crystals in a polycrystalline ceramic fiber is certainly responsible for its very low strength. 4.2 Forming glass fibers fromstrong melts

Two generic fiber-forming processes are generally used to fabricate glass fibers from strong melts; downdrawing from solid preforms and conventional melt spinning, a process that consists ofdrawing fibers from bushings. 4.2.1 Downdrawing from solid preforms Small diameter, structural silica glass fibers and large diameter, bicomponent optical silica fibers are downdrawn from the surface melt of a solid preform. The melt temperatures needed to contain the melt exceed the capability ofpractical ceramic and bushing materials. (a) Structural silica fibers

For structural applications, pure silica fibers are pulled from high purity silica rods as shown in Figure 3, top right. Gas flame or electrical furnaces are used to soften and melt the ends of the preform rods sufficiently for drawing. Each individual silica filament obtained in this downdrawing process has a diameter of 9 urn, and yarns and/or rovings with up to 4800 filaments are then made available to the trade, where they are used either as woven reinforcement fabrics, oras yarns, rovings, and threads for various specialty uses. Pure silica fibers are amorphous glass fibers despite the fact that they are often, and quite incorrectly, called "quartz fibers". They are suitable for high temperature applications. Silica glass fibers have a moderately high tensile strength (3.4 GPa), low modulus (69 GPa) and very high continuous service temperature (1050°C), softening point (1670°C) and volatilization temperature (>2000°C). Chemical analysis shows them to be 99.95+ % pure. The technology, properties, and applications ofsilica glass fibers are analyzed inChapter 6. (b) Optical silica fibers

For optical applications, bicomponent fibers are pulled from the ends of >2 meter long and >0.6 meter diameter preform rods having a silica sheath (clad) and a waveguide core. The preform is fabricated by depositing the waveguide substrate by chemical vapor deposition within a tube of a high silica glass oron a removable rod known as bait. The windup of each fiber is individually controlled by a double laser interferometer and therefore by its diameter, not by a constant windup speed. Individual fibers are covered with protective organic coatings to preserve their inherent strength and to prevent abrasion that might occur during handling. Fiber optics technology and applications ofoptical glass fibers are discussed inChapter 7. 4.2.2 Melt spinning from strong silicate melts Glass-forming silicate melts approach the ideal requirements of a strong melt such as that of silica, but they do so at much lower temperatures. As a result they can be formed by a conventional melt spinning process. Thisdiscussion includes only continuous glass fibers and excludes discontinuous glass fibers such as glass wool made by a centrifuge process.

Chapter 4

93

Structural silicate glass fibers have high melt and forming viscosities (1000 or log 3.0poise) at temperatures ranging from >1100 to <1600·C [20-21). Important fibers in this category include general purpose E-glass fibers and specialty high strength S-glass fibers. In the conventional bushing process, an array of fibers is (a) spun by pulling the melt through orifices or bushing tips, (b) coated with a suitable finish , and (c) wound on a winder at high process speeds. A multifilament process (Figure 3, top left) produces these fibers by using bushings having less than one hundred tomore than a thousand tips. As shown inTable III: (1) a typical silicate melt is formed ata viscosity of log 1.7-log 2.0poise, (2) it flows through the furnace ata viscosity of log 2.0-log 2.S poise, and (3) fibers are formed ata viscosity of log 2.S-log 3.S. (4) The temperature atthe fiber-forming viscosity is the fiberforming temperature. (S) The liquidus temperature is the highest temperature at which crystals can form over a given period of time. (6) It is a practical requirement that the fiberforming temperature at a viscosity of log 3.0 poise exceeds the liquidus temperature by at least SO·C toavoiddevitrification inthe melt delivery and bushing system. Table III. Viscosity-temperature relationship for strong fiber forming melts Melt formation Melting and fining Flow in furnace parts Fiber formation Fiber formation Softening point Annealing po int Strain point

Log viscosity range

Representative log viscosity

1.7-2.0 2.0-2.5 2.5-3.5

2.0 2.5 3.0

7.65 13.0 14.5

Two examples will suffice. Conventional E-glass, one of two generic, general purpose glass E-glass fibers, has a fiber forming temperature of 1200·C and a liquidus temperature of 1064·C. Commercial S-glass, a commercial high strength fiber, has a fiber forming temperature of 1S6S·C and a liquidus temperature of 1S00·C. The fiber forming temperature of S-glass is 26S·C higher than that of E-glass, thus requiring considerably higher forming temperatures. The liquidus temperature of S-glass lies closer to its forming temperature (~T =6S·C) than that of E-glass (~T =136·C), a fact that would suggest that the S-glass process is more crystallization prone and therefore less tolerant toward excursions in operating temperature than the E-glass process. Further checkpoints as shown in Table III include the following: (1) the softening point represents the temperature atwhich the viscosity islog 7.6S. Atthis temperature a glass fiber deforms under its own weight. (2) The annealing point represents a temperature atwhich the viscosity is log 13.0, and (3) the strain point representing a temperature atwhich the viscosity is log 14.0. The strain point is near the glass transition temperature, and represents approximately the highest tolerable in-use temperature. 4.2.3 Structural silicate glass fibers General-purpose glass fibers, i.e., boron-containing [20-21) and boron-free E-glass fibers [24], represent nearly 99% of all continuous commercial glass fibers. Special purpose glass fibers, i.e., high strength and hollow glass fibers, represent about 1% ofall continuous glass fibers.

94

Chapler4

(a) Product design parameters

Table IVsummarizes the effect ofvarious oxides on the resulting fiber properties [30-31]. The addition of increasing amounts of alumina, beryllia, yttria or nitride has already been discussed in Chapter 4.1 .2. These modifiers tend to increase the modulus or stiffness of a glass fiber. In addition, alumina and baria tend to increase the density, alumina and strontia tend to increase the refractive index, and zinc oxide and zirconia tend to increase the alkali resistance ofa fiber. Ofthe compositional modifiers, 8203is unique. As a flux, it reduces the liquidus temperature of a melt and the fiber forming temperature, while not materially affecting the mechanical properties ofthe fiber. Addition offluorine (CaF2) has a similar effect. Table IV. Structure-properties map. After reference (30): P. F. Aubourg and W. W. Wolf, "Glass Fibers", in Advances in Ceramics, Vol. 18, Commercial Glasses, pages 51-63, D. C. Boyd and J. F. MacDowell, editors, American Ceramic Society, Westerville OH (1986). Tensile strength Glass formers SiO, +++ B,O, Intermediates TiO, 0 Al,O, ++ ZIO, 0 Zno 0 BeO

Modifiers MgO CaO srO BaO Y,O,

Up

Nap

K,0

F,

Fiber YO~'s Mod us Density

Refrac. Dielectric Resistivity Chem. durability Inde x Constant Water Acid Base ++

+ ++

+

++

+

+ +

+ +

+

+

+++

+

+ ++ + +

+

+

0 0

+

0

0

0 +++ 0

+ ++ +++ ++

0 0

0

+ + + ++

+

0

+

+

0 + +

0

+ ++ ++

0

+ +

+ ++

+ +++' + ++' +++

0

0

0

+++ ++

+

++

++ +

Others AlN' ++ +++ ++ +++ +++ The addition of a given oxide has a pos itive (+), strong positive (++) or very strong positive (+++) effect on the specific property, or alternatively, a nega tive (-), strong nega tive (- ), or very strong negative (- ) effect, while (0) indicates that the specific property remains unaffected (30). An asterisk (') indicates a ma jor addition to the original structure-property map (30) which d id not cover the addition of nitrides to oxide glass melts.

(b) General and special purpose fibers

The first commercial general-purpose glass fiber was described in 1943. This E-glass variant (Table V) [20] [34] contained - 10% boron oxide and had excellent processability, low electrical conductivity (hence the name E-glass), excellent water durability, good mechanical

Chapter 4

95

properties, and an acceptable dielectric constant Another E-glass variant was described in 1951 . It contained only - 5% boron oxide (Chapter 6), and boron-free E-glass variants were described in1985 (25) and in 1995 [24). A combination of 2.5% Ti02 and 2.5% lnO also reduces the melt viscosity and fiber forming temperature, but the combined contribution is not as effective as -5% B203 (Table V). Boronand fluorine-free compositions having high levels of lnO and Ti02 are special purpose corrosion resistant (CR) glass fibers which command a premium price. They are commonly designated as E-CR glass. As part of the same technology [30) (34), the addition of lnO , MgO, BaO and SrO offers higher acid resistance than E-glass while retaining its electrical properties. Some compositional modifiers affect the fiber structure and therefore the strength and modulus, and afford high strength (HS), high modulus (HM) glass fibers. Others affect the fiber chemistry, and afford glass fibers with high alkali and/or acid resistance, high or low electrical conductivity, semi- or superconducting properties, and high or low densities. This flexibility in composition demonstrates the versatility of the generic melt spinning process whereby a variety ofglass fibers can be formed between 1185 and 1590°C. An entire chapter, Chapter 6, is devoted to general and special purpose glass fibers derived from strong silica and silicate melts. Two excellent reviews of commercial general-purpose fibers [30-31) are also available. 4.3 Forming glass fibers from fragile melts

Most silicate melts such as E-glass and S-glass melts have high viscosities above and below their liquidus temperature. These are considered to be strong melts and are preferred for the development of new fibers. However, fibers from fragile silicate melts may be required to meet special product requirements such as the attainment of very high moduli; fibers from fragile aluminate [9) [14), fluoride, tellurite [17), fluorophosphate (66), and certain chalcogenide [65) melts may be required for optical and other specialty uses. 4.3.1 Glass fibers from fragile silicate melts Several new high modulus magnesium aluminosilicate fibers (19) have recently been produced from fragile melts having relatively low melt viscosities above the liquidus temperature « 140 poise). The motivation for this development was the desire to produce high modulus silicate glass fibers (-130 GPa) with a high degree of internal order (crystallization potential), i.e., properties which would ordinarily require much higher melt viscosities. Appropriate magnesium aluminosilicate compositions were selected which have not only relatively low melt viscosities above their liquidus temperatures but also high crystallization rates when crystallization occurs. Addition of yttria to these compositions reduces the crystallization rate (Table V) and increases the modulus. In this process, the fiber forming cones under the bushing tips consist of two sections. The first, just below the bushing tips, has a much lower viscositythan the second, where crystallization may occur before the fibers solidify. To assure adequate process continuity, a precision process was designed that also relies on forced cooling of the fiber cones.

1185-1200 1050-1064 2.53-2.55 <15 3.4-3.5 72-74

amorphous glass [21]

Structure and Morphology References

. ..... . .. . . .

tr . 0-1 0.1-0.4 0-0.5

. ..... . .... .

52-56 12-16 ...... 5-10 16-26 0-6

B-containing silicate

amorph. glass [24]

1260 1200 2.62 <15 3.5 81

22.6 3.4

22.13 3.11 .... ... 0.55 0.14 0.63 0.25 0.04 . ...... .......

amorph. glass (25)

1250 ...... 2.67 <15 3.5 ......

...... ......

.. .. . .

1.5 0.9 0.2

. . . ... . .... .

59.0 12.1 ......

60.01 12.99 .......

Boron- and Zn-free silicate compositions

amorph. glass (34) [20]

1230 1163 2.68 <15 3.5 80

1.0 0.1 tr . ...... ......

21.7 2.0 2.9 2.5

58.2 11.6 .... ..

B-free Zn silicate

E-CR glass

amorph. glass [21]

1560 1500 2.49 <15 4.6 88

...... ... ...

. . . .. .

... ... 10.0

am./n.-cr. gl. cer. (?) (26)

>20 <9.0 213

3.50 <15 4.5 120 am./n.-cr gl. cer. (?) [19]

1580

17.0 ......

. . . . ..

.... ..

50.3 7.6 25.1

Y-Si-AI-O-N

n.-cryst. gl. ceram. [27] [28]

<15 <4.0 248

1590

23.4

62.2 0.7

ILl 2.2

Ca-Si-Al-O-N

Oxvnitr ide fibers

1450

. ..... ..... .

10.0

65.0 12.0 13.0

Y-Mg-Al silicate

HM-g.lass

65.0 25.0 ......

Mg-Al silicate

HS-glass

v. Compositions and properties of structural silicate fibers

Typ ical E-glass variants

Log3 poise,T °C Liquidus T, °C Density, g/ccm Diameter, urn Strength, GPa Modulus, GPa

Si,N.

A~N,

F,

Y,O, B,O, CaO MgO ZnO, TiO, K,O Nap Fe,O,

A~O,

SiO,

Compositional Features

Fibers

Table

to

'"

~ ....

C'> :::r

0>

Chapter 4

97

The resulting precision process is subject to forced cooling in the forming zone and yields glass and glass ceramic fibers with lower strength than high strength fibers, such as S-glass (3.5vs. 4.6GPa), but with a considerably higher modulus (130 vs. 84-95 GPa). A modulus of -130 GPa is the highest that has been observed for a silicate based glass fiber without incurring severe strength loss. It should be noted however that this modulus level can be surpassed by Sialon ornitrogen modified silicate glass fibers (Chapter 6). 4.3.2 Melt spinning from supercooled melts Melt spinning from supercooled fluoride melts affords optical single and bicomponent glass fibers [7]. The latter have a concentric fluoride core and a fluoride sheath or clad with a slightly different composition. (a) Single and double crucible process The double crucible apparatus, which yields concentric bicomponent fluoride fibers, is a generic melt spinning process except that the dual melt is separately maintained under carefully controlled conditions in a supercooled state well below the liquidus temperature of the fluoride glasses [6-7]. The inner and outer crucibles form the concentric tip system through which the glass melts flow. In this process [7], the inner and outer crucibles are heated to a temperature of about 750 to 800·C, and held at that temperature for approximately 30 minutes to allow the melts to cure. The melts are then rapidly quenched until the core melt reaches about 350·C. The outer crucible is then rapidly raised to a fiberizing temperature of about 315·C, and fibers are drawn, depending on the pressure in the crucibles, at speeds of 900 to 3300 m/s. Low pressures are adequate tofacilitate high drawing speeds. This transient operating cycle [7] minimizes devitrification as the melt passes from refining to fiberizing conditions. The low pressures which are independently applied to each crucible facilitate the formation of solid concentric sheath/core fibers and prevent the formation of hollow fibers. The core diameters can range from 8.4 to 75.1 IJm and the overall fiber diameter, including sheath orclad, can range from 140 to 149 IJm. (b) Single and bicomponent fluoride fibers

Core/clad bicomponent fibers were obtained with core diameters ranging from 8.4 to75.1 IJm, and single component fluoride fibers having these diameters were obtained by a single crucible version of the double crucible process [71. Single component calcia-alumina and alumina-telluria glass fibers could also be made by this process, and the process could be used tofabricate concentric sheath/core fibers from fragile dual calcia-alumina or fragile dual alumina-telluria melts, if there were demand for such fibers. 4.3.3 Updrawing from supercooled melts A narrow region exists near the eutectic of alumina-telluria and calcia-alumina melts which has fiber forming viscosities (log 2.5-log 3.0poise) below, but not above the liquidus. All other alumina-telluria and calcia-alumina melts have inviscid melts with low melt viscosities «10 poise).

98

Chapter 4

(a) Updrawing oftellurite glass fibers

The alumina-telluria phase diagram offers a glass forming region between 5.0and 11 .0weight (or 7.8-16.8 mol) percent of alumina, where high viscosity compositions can be found surrounded by low viscosity compositions. Within this region, alumina-telluria glass fibers could be manually updrawn from supercooled melts [35] at 800-850°C, yielding the first selfsupporting tellurite glass fibers on record. These fibers had tensile strengths ranging from 0.4 to 1.1 GPa, moduli ranging from 69 to 83 GPa, and diameters ranging from 40 to 150 IJm. They were x-ray amorphous, clear but pale yellow, and are of interest because of their potentially valuable spectral transmission properties [351. (b) Updrawing of aluminate glass fibers

Over 500 quaternary aluminate glass fibers were manually updrawn from fragile melts supercooled about 50 e below their liquidus while carefully controlling the temperature of the melt [8] [36-37]. The glass compositions in this narrow region ofthe phase diagram (Figure 7) contained 30-55% calcia, 45-55% alumina and 0-10% silica. Their ratio of oxygen ions to network forming ions (AI>3 and Si", and an occasional minor component) was between 2.35 and 2.6 [17], a ratio that is a necessary but not sufficient condition for the formation offragile fiber forming aluminate glass melts. 0

A continuous updrawing process (Figure 8) was designed [38] [17] using a quaternary low silica composition (44.3%CaO - 48.7%Ab03 - 3.5%MgO - 3.5%Si02). The melt properties of this composition (Table VI) confirm the steep drop in mell viscosity from 182 poise at a temperature 20 0 e above the liquidus, to278 poise atthe liquidus temperature, to720 poise at the fiber forming temperature which is50 0 e below the liquidus.

Air-

Figure 6. Double crucible melt spinning process depicting the outer and inner crucible, the airflow and the void orhollow fiber formation. Redrawn from M. L. Nice, Apparatus and process forfiberizing fluoride glasses using a double crucible and the compositions produced thereby, US Patent 4,897, 100, Jan. 20, 1990.

Chapter 4

99

Thiscalcium aluminate fiber was evaluated in structural applications but it was not suitable for the evaluation ofinfrared optical properties because it contained bound water as evidenced by a strong hydroxyl band at2.9IJm in the IR transmission spectra. Hydroxyl-free compositions were made in carbon crucibles [17Jby the Davy process [39J, i.e., by a procedure [40J by which disposable optical calcium aluminate bulk glasses are prepared for commercial applications in optical windows. Table VI. Melt properties of a fragile aluminate melt [8, 31) 44.3% Al,o, - 48.7% CaD- 3.5 % MgO - 3.5% SiO, (Weight) Viscos ity, poise LT + 100°C LT + 50°C LT LT -50°C LT - 2000C Modulus, GPa

58.0 182.0 278.0 720.0 10000.0 110.3

Quaternary calcium aluminate glass fibers were also downdrawn from preforms on a laboratory [9J and a development unit [41 J. Downdrawing is a valid, but costly, method for fiberizing fragile melts. In addition, only one of >500 compositions could be fiberized on a conventional melt fiberizing unit [17J. Thisprocess is not suitablefor fragile melts. 4

3

LT-200

-----.().-.. Selectedternary and quaternary calcia-alumina melts

2 (J)

til

...

Q.

*

--

LT __-0••

... ...

Cl

0 ....J

0 LT-200 -1

-2 30

LT LT+100

40

50

60

70

80

90

100

Alumina concentration, wt.% Figure 7. Calcia-alumina-silica system. This illustration compares the melt viscosity of the inviscid binary aluminate melts from 60%CaO to 75% alumina. Addition of MgO and CaO produces quatemary aluminate melts, raises theviscositysignificantly andproduces a fragile melt with high viscositiesbelow the liquidus temperature and low viscosities above theliquidus temperature. Redrawn from F. T. Wallenberger and S. D. Brown, High modulus glass fibersfor newtransportation and infrastructurecompositesand fornew infrared uses, Composites Science and Technology. 51 , 243-263 (1994).

100

Chapter 4

Monofilament

Hepurge Water-cooled

_\~OO_C_OI __---, "

1

G

Graphite guide spool

eQ G e c e e

Furnace no. 1

Hepurge Me"ed glass charge in platinum boat

o e e Q e e

Alumina support ring ~ Zirconia support ring

1/2 RPM

Figure 8. Continuous updrawing process. Redrawn from T. F. Schroeder, H.W. Carpenter and S. C. Camiglia, High modulus glasses based onceramic oxides, Technical Report R-8079, Contract N00019-69-C-Q150, US Navy Department, Naval Air Systems Command, Washington, DC, December 1969.

4.3.4 Hybrid fiber forming processes The most important glass forming systems contain elements from the sixth, or chalcogenide, column of the periodic table which includes oxygen, sulfur, selenium, and tellurium. Oxygen containing (i.e., oxide) glasses are insulators; the others tend to be semiconductors. Some melts are inviscid and/or fragile such as the tellurite melts described in the previous chapter. Others are viscous, and whether they represent strong or fragile melts, they are difficult to fiberize. A hybrid process for forming chalcogenide glass fibers [65] has been described that uses elements of downdrawing from preforms and fiberizing through bushings. Specifically, a cylindrical chalcogenide preform isvertically inserted into acylindrical crucible furnished with a nozzle inits bottom plate. The crucible is heated only in the vicinity of the nozzle, and a fiber is continuously drawn from the nozzle at a forming temperature that corresponds to a melt viscosity oflog 3 poise.

101

Chapter 4

Heavy metal fluoride fibers require a fiber forming process that relies on a supercooled melt, while certain fluorophosphate melts can be fiberized by pulling fibers from the melt using a conventional bushing process [66]. Fluorophosphate glass fibers are difficult to pull from the melt because the temperature range around the fiber forming viscosity oflog 3 poise isnarrow and the crystallization rate of the melt is high. Nonetheless, glass fibers were obtained with a diameter of 27 IJm and a tensile strength of 334 MPa for producing reinforced visible-IRtransparent poly (chloro-trifluoroethylene) composites. 4.3.5 Quaternary calcium aluminate fibers Quaternary calcium aluminate glass fibers made by updrawing from a supercooled fragile melt offer superior mechanical properties and sapphire-like infrared transmission spectra. (a) Fiber properties The highest reported pristine strength (8.3 GPa) was obtained (Table VII) with low silica 44.3% Ab03'48.7% CaO - 3.5% MgO - 3.5% Si02 glass fibers when they were updrawn from supercooled melts in an induction furnace [37]. When updrawn in an oxyacetylene furnace, they had a somewhat lower strength (4.2GPa), but it was still higher than that of E-glass (3.5 GPa). The modulus (110.3 GPa) of these fibers was 1.5x that of E-glass. The lowest modulus of any calcium aluminate fiber shown in Table VII was 10% higher than that of Eglass. The stiffest fibers shown in Table VII, a group of zinc oxide modified calcium aluminate glass fibers, had moduli ofup to 122.7GPa, or1.7x the stiffness ofE-glass. With two exceptions, all fibers shown in Table VII had strong hydroxyl bands in their infrared transmission spectra. The exceptions are two recent hydroxyl-free compositions made bythe Davy process [39]. One is a low silica composition and the other is a non-silica composition (46.2% Ab03- 36.0% CaO - 4.0% MgO - 13.8% BaO). An in-depth analysis of the physical properties offibers shown in Table VII isavailable [36-37]. Table VII. Modulus of updrawn calcium aluminate fibers [4, 6-8,31) Modulus GPa 72.0 79.3 103.4 108.9 109.6 109.6 110.3 110.3 110.3 111.7 115.8 122.7 248.0

5.2 5.3 4.0

Na,o,K,O 13.8% Baa 30.0%ZnO

3.5

5.6

25.7% PbO 25.0%ZnO 20.0%ZnO

Process Route and Comments Melt spun through bushing [21J Updrawn low-silica fiber [17] Melt spun through bushing [17] Updrawn low-silica fiber (17) Updrawn non-silica fiber [17] Updrawn high Zno fiber (17) Updrawn low-silica fiber (17) Updrawn low silica fiber (17) Updrawn high PbO fiber (17) Updrawn high Zno fiber (17) Updrawn high ZnO fiber [17] U drawn ZnO/L' 0 fiber 17

102

Chapter 4

100 90

-(-;--_. . I.

80 ;,!! 0

c:

0

'iii en

70 60

'E

50

c:::

40

en

III

t=

., ·· ·,·· ··,

, ..

" " "'. ' ..'

·· ···.'

30 20

"

10 0 0

~

··, ., .. . .,

345

2

6

7

Wavelength, 11m _ . - Ca aluminate

-

-

Sapphire

•••••• Quartz

Figure 9. Spectral transmission of calcia-alumina glass fibers. Redrawn from F. T. Wallenberger, N. E. Weston and S. A.Dunn, Melt spun calcia-alumina fibers: infrared transmission, J. Non-Cryst. Solids, 12[1),116-119 (1990).

Their resistance to alkaline media exceeds that of commercially available AR silicate glass fibers (Chapter 6) having a zirconia content of up to 15% [20). Hydroxyl-free quaternary calcium aluminate glass fibers (Figure 9), e.g., non-silca fibers containing 46.2% Ab03 36.0% CaO - 4.0% MgO - 13.8% BaO, afford sapphire-like infrared transmission properties. (b) Potential applications

The commercial potential of updrawn quaternary calcium aluminate glass fibers was tested in two stages. In the early 1960s, they were evaluated because they yielded higher moduli than those which could then be achieved with silicate glass fibers [36-37]. Timing for this development coincided with the onset ofthe commercial development ofcarbon fibers and no new aluminate orsilicate glass fiber was commercialized until 1995. In the early 1990s, a renewed evaluation of quaternary calcium aluminate glass fibers was triggered by their sapphire-like optical properties [17] [42-43]. Properly coated aluminate fibers might afford sapphire-like sensor performance applications at a more affordable cost than single crystal sapphire fibers, but unlike the latter, they would be limited to ambient and moderately high temperatures [431. Calcium aluminate fibers are not commercially available, but so far they offer valuable models for important structure-property relationships [4] regarding strength, modulus, and optical properties.

Chapler4

103

4.4 Amorphous fibers from inviscid liquids Melts of metals as well as crystalline ceramic oxides have low viscosities. All solidify at a sharp melting point and their viscosity above the melting point increases rapidly and then reaches a viscosity comparable to that of motor oil at room temperature, i.e., log <0.2 poise. Yet, fiber formation from the liquid phase requires a viscosity oflog 2.5 to log 3.0 poise. Only liquid droplets are formed if a low viscosity liquid is extruded ata normal quench rate of -104 Kls through an orifice, spinneret hole or bushing tip. To facilitate fiber formation, the viscosity must be raised from log <0.2 poise to log 2.5tolog 3.0poise. This can be achieved by one of two generic routes, i.e., by a rapid solidification (RS) or an inviscid melt spinning (IMS) process. 4.4.1 Attainment offiber forming viscosities In the rapid solidification (RS) process, the quench rate increases from-ttl' Kls to >106 Kls, and the liquid or molten jet almost instantly traverses its narrow, high viscosity range and almost instantly solidifies before it can crystallize. The attainment of self-supporting metal and/or oxide glass fibers by a true rapid solidification process remains a challenge for the future, but amorphous metal glass ribbons can be made by this process. It is possible tolook upon ribbons as ribbon shaped fibers but, in this process, they not are produced as selfsupporting structures. The high quench rate that is needed to obtain them is achieved by casting them on a support structure, a cold quench wheel. In the inviscid melt spinning (IMS) process the quench rate is -104 Kls, but the inviscid liquid (melt) or resulting jet, or at least its surface, is carefully under- or supercooled to reach its narrow fiber forming range (log 2.5 to log 3.0 poise) to facilitate fiber formation. Thus, continuous cryogenic hydrogen, deuterium, nitrogen and argon fibers, and continuous optical yttrium aluminum garnet (YAG) glass fibers are formed by this process at <88°Kor >1600°C, respectively. Continuous aluminate and metal fibers can be formed by raising only the surface viscosity of the molten jet during fiber formation, i.e., by chemically modifying it in order tofacilitate its solidification. Rapid solidification and inviscid melt spinning suppress crystallization, which would otherwise occur with aluminate and YAG melts which are derived from highly crystalline materials. Fibers from liquefied gases, fibers from inviscid melt spun oxides and ribbons from rapidly solidified metal alloy melts are amorphous. Fibers spun from inviscid metal melts are predominantly amorphous but contain a minor crystalline phase. 4.4.2 Rapid solidification (RS) processes The planar flow casting process is the rapid solidification process of choice for forming continuous ribbons of metallic glasses or metal alloys from their inviscid melts. Successful ribbon formation by this ribbon casting process requires the molten metal to be extruded with quench rates of 105 to 106 Kls through a ribbon shaped orifice onto a cold rotating quench wheel. The alloy ribbon is formed without losing its precision shape, and is collected on a continuous windup roll. Recent review articles and bulletins (16) [58-60) can be consulted for details. Rapid solidification of self-supporting metal fibers having round fiber cross sections remains a challenge for the future.

104

Chapter 4

(a) Amorphous metal ribbons

Glass forming metal alloys [16] include late transition metal-metalloids (e.g., FelOo-xB,), early transition metal-metalloids (e.g., TilOo-xSi,), early transition-late transitions metal alloys (e.g., NblOo-xNi,), aluminum based alloys (e.g., AI63CU1 6V7), lanthanum based alloys (e.g., Lal00.,Au,), alkaline earth based alloys (e.g., Mgloo2n,), and actinide based alloys (e.g., U100.'CO,). The techniques for producing amorphous metal alloys [16] include rapid liquid cooling, supercooling of liquids, physical vapor deposition, chemical methods, irradiation, and mechanical methods [16].

Stainless steel vessel

Molten alloy --It--U-.cc!!Z~::::;::::n-tt-1 Copper levitation coil -#---1~-''fp.=« Supporting stand of specimen ---tt--it---;

-:

o

Copper roll

Stainless steel vessel

Figure 10. Rapid solidification process. Redrawn from D. E. Polk and B. C. Giessen, Amorphous or glassy materials in Rapid solidification technology source book, edited byR. L.Ashbrook, pages 213-247, American Society forMetals. Metals Park, Ohio (1983).

The industrial fabrication of amorphous metal ribbons dates back to the late 1970s and a process known as planar flow casting [61] as shown in Figure 10. Specifically, the low viscosity melt of a metal or metal alloy is extruded through an orifice or ejection nozzle and deposited onto the flat cold surface or rim of a rotating wheel (typically copper) where it is rapidly quenched. The quench rates may range from 105 to 106110 urn, and a uniform rectangular cross section.

105

Chapter 4

(b) Products and applications Amorphous magnetic glass ribbons based on alloys of iron, nickel, or cobalt are among the softest magnetic materials known. Their tensile strengths range from 1.0 to 1.7 GPa, their moduli from 100-110 GPa and their service temperatures from 90 to 150°C. Their high degree of softness combined with excellent mechanical properties has benefited applications ranging from microscopic recording heads to large architectural EMI shielding. Other applications include electronic and power cores, and anti-theft sensors [60].

Figure 11 . Amorphous metal ribbons. Parsippany, NJ(1993).

From H. H. Liebermann, Metglas® product bulletin, Allied Signal,

Tin-lead alloy Metglas ribbons, having moduli as low as 18 GPa (Figure 11), are used as solders for diebonding applications. Their liquidus ranges from 190-314°C, and their solidus from 182-310°C. Metglas brazing foils are made from nickel, cobalt, palladium, and copper based alloys as melting point depressors, and they come in ribbons 10 cm wide and >25 urn in thickness. Solder and brazing ribbons have solidus temperatures ranging from 770 to 1130°C and liquidus temperatures ranging from 925 to 1150°C. Applications include aerospace components, electric motors, and brake pads [60].

4.4.3 lnviscid melt spinning (IMS) processes In processes with conventional quench rates of -104 K1s, the inviscid liquid (melt) or resulting jet is carefully under- or supercooled to its narrow fiber forming range (log 2.5 to log 3.0 poise). Three types of fibers have been made by variants of this process: continuous optical yttrium aluminum garnet (YAG) glass fibers, continuous aluminate glass fibers and steel fibers.

106

Chapter 4

(a) Principles of jet and fiber formation

The key principle governing the formation and breakup of a liquid jet is well known [12]. A liquid jet is unstable with respect to viscosity, diameter, and surface tension. A jet has a tendency to break up due to axisymmetric surface pressures, and produces Rayleigh waves, orperiodicvariations ofincreasing amplitude in the jetdiameter (Figure 12). Ultimately, these diameter variations cause the jet to break up into separate droplets, and they, in turn, crystallize and form shot. For any process to yield a continuous fiber from a low viscosity liquid, the transient viscosity ofthe melt exiting the bushing tip orspinneret orifice must quickly reach the fiber forming level oflog 2.5-log 3.0poise before jetinstability (formation of Rayleigh waves) and potentially disruptive crystal growth can occur.

Figure 12. Straight fiber and frozen Rayleigh waves. The straight, cylindrical fiber represents an inviscid melt spun calcium aluminate glass fiber that had been surface stabilized with particulate carbon. The frozen Rayleigh wave structure represents a calcium aluminate fiber that was not surface stabilized and solidified while it was in the process ofbreaking upinto droplets and shot.

For fiber formation (Equation 1), a jet must have a lifetime (t) sufficient for the viscosity to reach log 2.5-log 3.0poise before the onset of turbulence orbreakup [12]. The viscosity (11) of the jet is the key factor since it depends exponentially on the temperature. The diameter (D) ofthe jetisthe next most important factor. Thus, a reduction of the viscosity (11) shortens the jet lifetime (t) and limits the attainment of low diameter fibers. Surface tension (y) and density (p) are less important, i.e., less sensitive totemperature. t

= J4 [( pD 3 / r)' 12 + [31]D / r ))

dn / d t: TL < T"

L dVc / dr

(1 ) (2) (3)

Chapter 4

107

The conditions (Equation 2) under which oxide or metal fibers can be formed from inviscid melts is also defined by the dynamics of the relationship inthe forming zone between the rate of solidification or the change of the viscosity (11) and the rate of crystal growth (Vel with time (T). Any change in melt viscosity, even if only of the surface viscosity, of a mol en jet will disproportionately affect its lifetime. If the lifetime is short, the jet will, on cooling, break up into Raleigh waves, liquid droplets and crystalline shot, and if it is long, the jet will instead solidify into a continuous fiber on cooling. (b) Principles ofincreasing the jet lifetime One inviscid melt spinning process, the containerless laser heated melt process (Chapter 4.4.4) is believed tofacilitate the formation of fibers by increasing the viscosity of the inviscid melt (and jet lifetime) at a normal quench rate of 104 Kls, i.e., without increasing the quench rate to -108 Kls. Two older inviscid melt spinning processes require a chemically reactive environment and are believed to facilitate fiber formation only by increasing the surface viscosity of the jet. They are a metal fiber forming process (Chapter 4.4.6) and an oxide glass fiber forming process (Chapter 4.4.7). The theory governing inviscid melt spinning by chemically modifying the jet and fiber surface surface isdiscussed in Chapter 4.4.8. 4.4.4 Oxide fibers from containerless, laser heated melts Containerless liquid phase processing has been successfully used to achieve deep undercooling of molten oxides to a temperature significantly below their liquidus temperatures where the viscosity is sufficiently high for fiber pulling. Containerless conditions eliminate heterogeneous nucleation by containers (crucibles, spinning cells or bushings, and precious metal tips) and help deep undercooling [73). One may assume the fiber forming viscosity, although not reported, was between log 2.5and log 3.0poise.

Optical pyrometer

Fiber _ Stepper motor

---- Levitated molten oxide _ Stinger

Stepper motor controller

Computer

Figure 13. Container1ess laser-heated melt process. Redrawn from J. K. Weber, J. J. Felton, B. Cho and P. C. Nordine, Glassfibres of pure and erbium- or neodymium-doped yttria-alumina compositions, Nature, 393, 769-771 (1998).

108

Chapter 4

In this process (Figure 13), molten samples of yttrium aluminum garnet (YAG: YJAb01Z), 3 mm in diameter, were levitated in a flow of argon gas. The levitated material was completely melted ina continuous wave COz laser heating beam. The viscosity ofthe inviscid melt above the melting point was 0.05 Pa.s (0.5 poise). The laser beam was then blocked, resulting in a cooling rate of -250 K1s. At a pre-selected fiber forming temperature between 1600 and 1660· C, i.e.,a temperature that iswell below the liquidus temperature, fibers were pulled from the levitated droplet by rapidly introducing and withdrawing a 100 IJm diameter tungsten wire "stinger". At higher temperatures the stinger pulled out of the melt without forming a fiber, while crystallization was likely tooccur atlower temperatures [73]. Glass fibers, up to 0.5minlength and 5 to 30 IJm in diameter were pulled from the apparatus at rates of 1.0-1 .5 m/s before crystallization terminated the process. The fibers had a homogeneous appearance and smooth surfaces. They were transparent, highly flexible, and x-ray amorphous. Glass fibers of this kind, which were also made from erbium- or neodynium-doped ytlria alumina compositions, would therefore expand the range of fiber lasers into the mid-infrared. The economics and scalability of the new process are not known. The materials cost and the cost ofoperating a laser process are probably about the same for an amorphous YAG sensor fiber made by the containerless laser heated melt process and a for single crystal YAG sensor fiber made by laser heated pedestal growth (Chapter 4.5.2). And both are containerless processes. However the higher process speed may favor the laser heated melt process (1 .5 m/s) over the laser heated pedestal process (1 mm/s). 4.4.5 Metal fibers ina reactive environment Commercial wire drawing processes produce metal wires with round cross sections but they are highly energy and labor intensive. Wire drawing falls outside the scope of this book. Commercial rapid solidification processes yield amorphous metallic ribbons. lnviscid melt spinning yields metal fibers by a chemically assisted jetstabilization process. In the inviscid melt spinning process [10) [51), steel wires are formed by the same mechanism as glass fibers. In this case, the process shown in Figure 14 depends on the presence of silicon in the steel formulation and on the presence of carbon dioxide in the process environment. A commercial pilot production unit based on the schematic process diagram shown in Figure 14 consisted of a 0.9 m long furnace, a 1.5 m long cooling column and a windup [51). The furnace contained a crucible holding 50 kg of steel, and was heated with a 4 kHz power unit supplying 70 KW. An essential ingredient in the cooling medium was carbon dioxide. Wires with diameters of 100-200 IJm can be made at 1500·C with speeds of 10-20 mIs, respectively. At a rate of 15 mIs, the spinning of a 165 IJm diameter wire lasted 4 hours. In this pilot process, unbroken wire was obtained for periods in excess of 1 hour which represents a continuous length ofabout 80 km. As shown in Equations 4 to 6, carbon dioxide diffuses into the surface of the inviscid molten jet. Its concentration decreases with increasing distance from the surface, but wherever it finds silicon that is evenly distributed throughout the melt (and therefore the surface of the liquid jet), it forms silicon dioxide and carbon which cause a steep increase in the surface

Chapler4

109

viscosity. Carbon may further react to form carbides. The surface skin is neither a sheath nor a film. (4)

Si+20~Si02

2 Si + 2 CO ~~ SiO

2

+ 2C

(5) (6)

Si+C~SiC

ESCA analysis confirms the presence of oxidized silicon and oxidized iron in the wire surface or skin [51). The silica peak gradually disappears at a depth of 100 nm, giving way to the peaks ofiron and silicon. Analysiswith a CAMECA ion analyzer shows that the intensity of Si peaks decreases to naught between 17 and 55 nm from the surface of a 165 IJm diameter steel wire. The results parallel those noted for carbon with aluminate fibers. Silica is the most viscous inorganic material known, especially at 1500°C. It is the ideal surface viscosity builder to increase the lifetime of a hot inviscid steel jet long enough to prevent formation of Rayleigh waves and shot. Other viscosity builders can be formed in-situ at the spinning temperature to stabilize a given molten metal jet [10). They must have a higher melting point than the metal, and be insoluble in the molten jet[10).

-

_

Reactive gas

Figure 14. Inviscid melt spinning process (schematic drawing). Redrawn from F. T. Wallenberger, N. E. Weston and S. A. Dunn, Inviscid melt spinning: as-spun amorphous alumina fibers, Materials Leiters, 2[4]121-127 (1990).

Inviscid melt spinning is considered to be a potentially viable alternative to wire drawing [51) for making steel wires for radial automobile tires, but a prior product development did not reach beyond the pilot plant level. Using silica steels, the complex chemistry (Equations 4-6) produce also minute amounts of iron oxides which were detected by ESCA [51), and are a potentially undesirable trace byproduct. The challenge [4) remains to fine tune the chemistry ofthis process before commercial development.

Chapler4

110

4.4.6 Oxide glass fibers ina reactive environment Continuous binary calcium aluminate glass fibers can also be formed by inviscid melt spinning. In this case, carbon particles which are formed by the decomposition of propane enter into the surface of the molten jetand raise its surface viscosity, a process that lengthens the lifetime of the jetand prevents itsbreakup. In this process [8) [10), alumina and calcia are placed in a tungsten crucible having an orifice in itsbottom plate. The crucible isplaced ina furnace and the surrounding air is replaced with argon. The oxide powder is melted and the mell is maintained 100°C above its melting point under a mild vacuum (Figure 3, bottom right, and Figure 14). The fibers are spun from the oxide mell after introducing propane below the crucible and by increasing the argon pressure above the crucible. Propane decomposes on the hot surface of the mollen jet, forms carbon particles which enter into the surface of the jet, and sometimes deposits an additional secondary carbon sheath on the fiber surface after it solidifies. These fibers contained 51 .5- 80.2% alumina and 43.5 to 19.8% calcia, and occasionally 3.8 to4.0% silica and/or 0.1 to 7.5% magnesia (Table VIII). Pristine strengths ranged from 0.16 to 1.05 GPa (vs. 3.5GPa for E-glass) and moduli ranged from 41 .1 to61.2 GPa (vs. 72.5 GPa for E-glass). The moduli [11] were much lower than those of updrawn high viscosity calcium aluminate fibers, suggesting much lower internal order. By SEM, the fracture patterns were typical ofglass fibers, and fibers with up to 80% (!) alumina were amorphous.

Table Vlll, Properties of inviscid melt spun calcium aluminate glass fibers (11) Al,O, 54.0 54.6 54.8 59.0 60.8 66.8 67.3 80.2

Fiber Composition (Wt % ) CaO MgO 39.0 3.0 39.0 2.5 38.9 2.5 40.8 0.2 39.1 0.1 33.2 34.2 19.8

Pure silica glass fiber Borosilicate E-glass

SiO, 4.0 3.9 3.8

100.0 54.0

Spin T. °C 1700 1500 1500 1500 1700 1800 1700 1800

Diam . J.1IIl 170 190 216 190 225 102 167 117

Strength GPa 0.5 0.7 0.6 0.7 0.5 0.9 0.8 1.1

Modulus GPa 50.8-68.9

1750 1200

100 10

3.5 3.5

69.0 72.0

46.6-61,2 45.9-56.4 41.0-54.3

Some fibers have a secondary carbon sheath which may be up to 600 nm thick [17). Not surprisingly, these fibers are black. The secondary overgrowth is not an integral part of the fiber [11-12]. It does not affect the fiber properties, and can be peeled or burned off. Other fibers have no carbon sheath. These fibers are translucent. Carbon that is present in the fiber surface or skin does not affect its transparency, whether the fiber had originally no carbon sheath as-spun, orwhether the sheath had been removed from the as-spun fiber [12). Sputtered neutral mass spectrometry (SNMS) depth profiles document that carbon is present in the skin of all fibers to a depth of about 50 nm (Figure 15), whether a given fiber has a secondary carbon sheath overgrowth or not [11). X-ray photoelectron spectroscopy (XPS)

Chapter 4

111

showed that carbon in the surface or skin [12) consists of carbide (51%) carbon (41%) and carbonate (8%). Infrared depth profiling by diffuse reflectance infrared spectroscopy (DRIFT) provided further important insights [52). It showed that carbon alters the oxygen environment of the aluminum atoms near the fiber surface from octahedral to tetrahedral coordination and promotes the generation of carbonaceous species such as ethers and esters in addition to carbonates and carbides [52) which have also been found with XPS [12). In summary, a typical aluminate fiber is spun between 1500 and 1700·C (Table VIII). Carbon enters into the skin of the still liquid jet. A secondary carbon sheath is obtained when more carbon is present in the reactive propane environment than needed to stabilize the liquid jet, but after the fiber is solidified «500·C). At higher temperatures it would oxidize (burn off). Thus, only some fibers have a secondary carbon sheath. If it is formed, it results from side growth, a secondary growth mode, that has already been discussed with regard to the growth ofcarbon whiskers in acarbon vapor environment (Chapter 2.1 .1). 100.------:::::=========j 90

80

70

s!:!. 5060 ~

:; LL

oe

40 30

20 10 50

100

150

200

Depth, nm

Figure 15. SNMS depth profile ofa translucent calcium aluminate fiber. Redrawn from F. T. Wallenberger and S. D. Brown, Highmodulus glass fibers fornew transportation and infrastructure composites and fornew infrared uses. Composites Science and Technology, 51 , 243-263 (1994).

Inviscid melt spun calcium aluminate glass fibers have low strength (0.5-1.1 GPa) and moduli (46-58 GPa). Low strength and low stiffness can be attributed to the random structure frozen into the fibers during rapid solidification. As a result, they are not likely to become composite reinforcing fibers, despite their excellent alkali resistance which they share with quaternary calcium-aluminate fibers [9). 4.4.7 Mechanism ofjetsolidification Continuous aluminate glass fibers are formed in the presence of propane, but not in its absence. A viable mechanism of jet stabilization must therefore explain (1) the function of carbon which enters into the surface orskin of the molten jet, (2) the function of carbides and carbonates which are instantly formed in the molten jet surface, and (3) the increase in tetrahedral from octahedral coordination of aluminum atoms in the surface [51) before the fiber solidifies and secondary overgrowth with carbon can occur.

112

Chapter 4

The principle governing jet formation and jet breakup [12] has already been discussed in Chapter 4.2.2. A liquid jet is unstable and will degrade into Rayleigh waves (Figure 12, right) and then droplets. The stability of a jet (Equation 1)depends on liquid density, jetdiameter, melt viscosity, and surface tension. The lifetime of a jet is the time required for the melt to traverse the continuous length of the jet before the onset of Rayleigh waves (12). A jet of a silicate glass oran organic polymer melt has a high viscosity (>W Pa.s) and a lifetime greater than 10° seconds; it can be spun or drawn from the melt by conventional methods, and it solidifies well before itcan form Rayleigh waves. Table IX. Properties of inviscid calcium aluminate jets [12] Alum. content 51.5% 54.6% 66.8% 80.2%

Fiber structure amorph amorph amorph amorph

Melt temp.

Spin temp.

1415 1390 1650 1830

1500 1500 1700 1900

0c.

0c.

Jet/Fiber Diam . urn 375 190 105 118

Melt density

g/cm 2}0 2.70 2.68 2.68

Surface tension mN/m 680 680 625 575

Melt viscosity Pa.s 0.34 0.55 0.14 0.06

Unassisted jet life, sec 1.4x10" 8.7x10·' 2.0xlO'3 1.7xlO'3

A jet of an aluminate melt with >50% alumina has a low viscosity «1 Pa.s) and a calculated lifetime less than 10.2 seconds (Table IX). If ejected into an ambient, neutral environment it will form Rayleigh waves and droplets (or shot when they freeze) rather than uniform continuous fibers. Equation 1 shows that viscosity is the major factor in determining jet lifetime; surface tension is a secondary factor [12). Any increase in the surface viscosity of the molten oxide will disproportionately increase the jet lifetime from that calculated for an unassisted jet (2.0x10·3 seconds) to that calculated for an assisted jet (2.0x10·1 seconds) which must have been obtained since continuous fibers were obtained. Particles, especially shaped particles are known to increase the viscosity of a suspension, following Mooney-Einstein [53). Thus, carbon particles (12) may enter into the surface or skin of a liquid inviscid jet and increase its viscosity sufficiently and long enough to facilitate its solidification and fiber formation. For this mechanism to be viable, three conditions must be fulfilled. (1) The increase in the jet surface viscosity must afford a stabilized (assisted) jet lifetime that at least matches the jetcooling time. (2) The assisted lifetime resulting from the viscosity increase must be comparable to the actual (unassisted) lifetime of a typical silicate fiber such as E-glass. (3) The surface viscosity increase needed to achieve this lifetime must be realistically achievable by carbon insertion inthe jetsurface. The pyrolytic production of carbon has been said to create a "snowstorm of large, flat molecules containing the hexagonal ring structure of graphite" [54), Le., flakes or flat aggregates of smaller particles. They enter into the surface of the molten jet and act as viscosity builders, where they and their instant reaction products such as carbides can be detected by ESCA. Since jetgeometry and surface forces tend toconstrain particle formation into planar structures parallel to the jet surface, the rheological treatment for flakes is appropriate. The viscosity (11) of a Newtonian fluid containing solid, suspended particles, relative tothat (110) of the suspending fluid, isgoverned by Equation 7 [53].

In (71 1710) = kEf:z l [1- f2 1 f ilii

(7)

The Einstein coefficient (kE) for incorporated particles depends on particle shape, f2 is the volume fraction offiller, and fm isthe maximum packing fraction for the flakes. The viscosity of

Chapler4

113

the suspending fluid (bulk molten) oxide is 0.14 Pa-s, and the required 36.6 Pa-s viscosity for the surface layer of the jetcan be attained with flake shapes having UDbetween 4 and 9, and volume fractions of solids between 0.32 and 0.45. The combined volume fraction of the solids in the oxide skin detected by SNMS was calculated (12) to be 0.508, a value well above that needed forrheological jetstabilization. E-glass jets with a diameter comparable to that of the 66.8% alumina jet can bemelt spun at temperatures ranging from 1100 to 1480°C(11) where they have a viscosity between 6.1 and Table X. Chemical stabilization of inviscid calcium aluminate jets Jet/Fiber Compo (21) (105 }lIlt Diameter) CaO-A},O, (66.8%) E-glass

Melt Behavior Inviscid Strong

Temp. °C 1700 1360

Viscosity, Pa.s Bulk Surface 0.14 36.6 31.7 31.7

Jet Lifetime, sec Unassisted Assisted 1 2.0x10·' 2.6xlO· 2.6x10"

816 Pa-s and a calculated lifetime between 8.8x10·2 and 1.2x10' seconds (Table X). The assisted lifetime of the 66.8% alumina jet (2.6X1 0 1 seconds) is therefore well within the range of the unassisted E-glass jet lifetimes (12). In summary, the stabilization of inviscid aluminate jets can be attributed to the increase in surface viscosity due to suspension of solid carbon particles (as well ascarbides and carbonates) in the molten oxide surface (12). The viscosity (rheology) controlled jetstabilization appears to be accompanied bythe observed change [52) from octrahedral totetrahedral coordination ofaluminum atoms near the surface. 4.4.8 Cryogenic fibers from liquefied gasses

A viable process forthe formation ofcontinuous, self-supporting fibers such ashydrogen from

liquefied gases has emerged over the past two decades (74). Like all prior process iterations (74), it appears to be an inviscid melt spinning process (IMS) and not a rapid solidification (RS) process. The first step in this process consists of increasing the viscosity, presumably to log 2.5 to log 3.0 poise, bydecreasing the temperature, e.g., forliquid hydrogen, to 29 K. The second step consists of extruding a liquid jet into a closed system and tosolidify the resulting fiber at a 10 K lower chamber temperature, l.e., at 19 K.

4.5 Growing single crystal fibers from inviscid melts Continuous single crystal fibers can be grown from inviscid melts by two relatively slow processes: the edge defined film fed growth (EFG) process (13) and the laser heated float zone (LHFZ) or laser heated pedestal growth (LHPG) process (14). Both offer growth rates of to0.3-0.7 mm/s [13-14). 4.5.1

Edge defined film fed growth

Edge defined film fed growth (EFG) is a commercial process (13) that facilitates the fabrication of continuous void free single crystal oxide fibers (Figure 16) from tungsten or other growth orifices.

114

Chapter 4

(a) Growth of sapphire fibers

This process yields commercial single crystal sapphire fibers. A liquid pool from which the continually growing filamentary crystal iswithdrawn isformed on top ofa planar surface of the orifice and fed by capillaries which extend down through the orifice into a liquid reservoir. The crystal shaping or edge definition is maintained by the geometry of the top surface of the orifice and the fulfillment of a contact angle of <90· between the liquid and material from which the orifice isfabricated.

Sapphire

crystal

Growth orifice

l__

- -- - -' Figure 16. Edge defined film fed growth process (schematic drawing). Redrawn from J. Monbleau, Single-crystal technology, Product BUlletin, Saphikon Inc., Milford, NH (1994). Single crystal alumina (sapphire) fibers grow in this process in the <0001> growth direction. These fibers are void free, and have a density of3.97 g/cm 3, a tensile strength of 2.1-3.4 GPa at room temperature, a Young's modulus of 453 GPa, superior electrical and optical properties, and superior chemical stability. Although individual fibers can be obtained with diameters ranging from 64 to 350 urn, their commercial diameters range from 150 to 250 urn. Sapphire is produced commercially throughout the world and is used in virtually every industry. The optical, electrical, chemical, mechanical, and nuclear properties of sapphire fibers, as described in the literature [13), make them an ideal material for many applications other than their use as sensor or reinforcing fibers for metal and ceramic matrix composites. Frequently, the combination of two or more of its properties make sapphire the only material available tosolve complex engineering design problems. (b) Process versatility

The edge defined film fed growth method (EFG) is a "near-net-shape" process for growing continuous sapphire fibers and for prototyping a wide variety of other shaped parts. Unlike other crystal growth methods used to make sapphire structures, only the edge defined film growth method can be used to also produce grown-to-shape tubes, rods, ribbons, and three-

Chapter 4

115

dimensional shapes. All other processes form ingots of various size, which must be cut to shape byhighly skilled workers using costly diamond impregnated tools. 4.5.2 Laser heated float zone growth In the laser heated float zone (LHFZ) or pedestal (LHPG) growth process, a circumferential laser is placed around a preform rod (e.g. polycrystalline alumina) to zone refine a segment of the material while at the same time updrawing a single crystal fiber (e.g., sapphire). (a) Growth of single crystal fibers Materials may mell congruently or incongruently in the float zone [14] [67-68]. The singular function of the laser in this process (Figure 17) is to provide uniform circumferential heating. The tip of the preform rod is melted, and a seed rod is introduced into the melt and slowly withdrawn to initiate the growth of a potentially endless filamentary single crystal. Sapphire fibers [14] [47] updrawn from sintered polycrystalline alumina preform rods are the best researched material made by this process. They offer a superior value in end-uses requiring prolonged exposure to ultrahigh temperatures. The properties of these sapphire fibers are comparable to those made by edge defined film fed growth, except that potentially lower diameters are possible at about equal growth rates [14]. Growth rates of 0.7 mm/s are possible, but growth rates below 0.3 mm/s are typically used with pure materials [14]. For comparison, sapphire fibers grown by edge defined film fed growth were reported to have growth rates up to 0.5 mm/s. The laser heated float zone process is not limited to sapphire fibers, oreven tooxides. Table XI lists fibers other than sapphire which have been grown by this process. This includes filamentary single salts, eutectics, oxides, ceramics, and semiconductors, as well as superconductor and metallic fibers. Sapphire, LiNb03 LbO-NaF, Nd:YAG and Cr:Ab03fibers have been grown with diameters as low as 55, 40, 20,'6 and 3 IJm, respectively. Some of the more important fibers grown by this method are Ab03-YJAIS012 (YAG) eutectic fibers [48] and Y203stabilized, cubic Zr02single crystal fibers [49]. A major motivation for current research is the potential ofsingle crystal fibers in optoelectronic applications [71]. Single crystal fibers of potassium lithium niobate are known to be useful materials for electrooptic applications. They can be grown with a growth rate of 11 mm/h with diameters of 0.9 mm by the halogen lamp assisted laser heated pedestal growth method, alternatively known as the laser float zone method. The single crystal fibers, like the ceramic feed rods from which they were grown, consisted of30K20·17LbO·53 Nb203 [46]. Core clad waveguide structures of the c-axis, 200 11m, single crystal lithium niobate fiber have potential applications as second harmonic generators, optical modulators, lasers, and optical oscillators. They are obtained by amagnesium ion diffusion process [69]. Filamentary eutectic ZrO/CaO crystals are also easily grown by the laser heated float zone process. They have a high degree ofion conduction and are aimed atadvanced applications such as solid electrolytes, where their conductivity at high temperatures may be exploited in heating elements and in-situ gas sensors [70). The starting materials are Zr02/CaO powders with a eutectic composition. Single crystal fibers are prepared by a polymer matrix route or a ceramic route. The phase formation inboth routes issimilar.

116

Chapter 4

I I

Push

I

_

Oriented seed _

Growing fiber crystal " "

1

Pull

Circumferential C02 laser beam

/

_

"'"

Molten zone _

Source

material--~'

1

Feed

Figure 17. Laser heated float zone or pedestal growth process (schematic drawing). Redrawn from R. S. Feigelson, Growth of fiber crystals, in Crystal growth of electronic materials, E. Kaddis, pages 127-145, Elsevier Science Publishers, London (1985).

The laser float zone process has also been used successfully to synthesize phosphorescent single crystals of SrAb04:Eu 2+,Dy3+ and CaAb04:Eu 2+,Nd3+ [72]. These single crystal fibers may have important applications inoptoelectronic technology. The crystals show bright green and purple phosphorecence, respectively, with persistence times exceeding 16 hours. Their brightness is 10 times stronger than that of traditional sulfide phosphors. Sulfide phosphors vaporize near their melting points, and single crystals are therefore difficult to grow. Alkaline earth aluminates have low vapor pressures near their melting points, and single crystals can be grown directly from their inviscid melts [72]. Typically, fibers having diameters up to 100 urn are readily possible; fibers with diameters above 200 urn are technically speaking no longer fibers but rods. This includes the niobate fibers [46] and the superconducting fibers [14) [50) [33] [63) discussed below. In principle, the float zone method is a containerless process. There is a nominal diameter reduction in the float zone from that of the preform rod to that of the final fiber. Zone (or overall process) stability for the growth ofoxide and fluoride fibers isusually achieved with diameter reductions inthe range of 1/2 to 1/3 [14). (b) High Tcsuperconducting fibers

These materials may also melt congruently or incongruently in the float zone [14]. Congruently melting materials do not undergo a phase transformation above room temperature, have a low vapor pressure attheir melting, and are easy to grow insingle crystal

Chapler4

117

form by this method. Incongruently melting preform materials first yield a phase that is different from that of the source rod and the melt [67J. As growth proceeds, the melt composition changes, and the liquidus drops until it reaches the peritectic decomposition temperature. At this point, the desired phase crystallizes and steady state growth will be achieved [14J. Table XI. Materials grown by the pedestal growth or float zone method' Material Fluorides: BaF, CaF, LiF-NaF NaF-NaCl Oxides: Nd:YAG YAG

AlP,

Ti:A~O, Cr:A~O,

LiNbO, Nd:LiNbO, LiTaO , L~GeO,

Gd ,(MoOJ , CaSc,O, Nd :CaSc,O, SrSc,O, YlG Eu:~O, Ti:MgA~O,

Ti:YAlO, BaB,O, Nd ,SiO, SrTiO , Nb,O, BaTiO,

SrBaT~O.

ScTaO, ScNbO, SrBaNb,O. Cr :Y,O, Cr :Sc,O, Cr:Lu,O, L~O-GeO,

Semiconductors LaB, B,C Superconductors: Nb (B~.srLBCa'2

Cu,P,)

Meltpt., OC

Orientation

Diameter, llIIl

Application

1280 1360 676 640

[110] [lll]

200-600 600

IRguide IRguide Eutectics Eutectics

1940 1940 2045 2045 2045 1260 1260 1650 1170 1157 2200 2200 2200 1555 2410 2105 1875 1095 1980 1860 1495 1618 1700 2300 2100 1500 2400 2400 2400 1106

[lll], [110] [111] . [110] a,c c c a, c c [110] a, c [1101 a,b,c c

4-500

[110] c

700-1700

c a, c a a, c c, [110) c, [110] c, [110)

2715 2468 1420 900

6-1000 100-1000 55-600 200-800 3-170 20-800 800 600 100-600 200-600 100-600 600 600 100-600 500-800 1000 1000 500 750 600

[lll]

300-800 200-600 200-600 200-600 600-1700 600 600 600

Laser Laser Remote sensor Laser Laser Acoustics Laser SAW device Raman device Ferroelastic Model Laser Model lnsu1ator Laser Laser Laser Nonlinear optics Laser Model Model Ferroelectric Ferroelectric Ferroelectric Ferroelectric IRguide Laser Laser Laser Eutectics

200 200

Cathode filament Thermoelectric.

200 250-1000

High T, sIc wires High T, sIc wires

Metallics: Si 960 200 Model Ge 2400 200 IR guiding Co 1495 100-600 Magnetics Fe-Co 1500 100-600 Magnetics Fe 1535 100-400 Magnetics "Private Communication from Professor Feigelson, Stanford University, California

118

Chapter 4

The laser float zone process is an effective method for growing oriented BbSr2CaCU20afibers [50]. Single crystal Y3Fe501 5 (YIG) and polycrystalline high T, superconducting Bh.aSrl.sCal.2CU2.20a fibers [50] (Figure 16) were first grown by the float zone process with diameters of 0.7and 1 mm, respectively, from incongruently melting compounds. The critical current measurements on the high Te fiber were 5x10-4 Alcm 2 at 68 K with zero resistance near 85 K. More recently, careful processing afforded Pb-substituted Bi-Sr-Cu-Ca-O fibers with critical temperatures (Te) of 106 K and critical current densities (Jc) of 1015 Alcm 2(77 K, 0 T) after annealing [63]. Magnetic susceptibility and criticai current density ofsuperconducting fibers of the Bi-Sr-Ca-Cu-O system grown by the laser float zone method are strongly dependent on the pull rates. A favorable grain orientation for current transport and highest critical currents was obtained athigh pulling rates, e.g., >4 ~m/s [67]. In summary, the laser heated float zone (LHFZ) method [67], inparticular the traveling solvent zone melting (TSZM) configuration [68], is a highly effective technique to grow centimeter long crystals of high Te and other low dimensional cuprates. High temperature superconducting fibers, wires, tapes, and ribbons have also been made by the powder-in-tube method. These are sheath/core bicomponent fibers or ribbons having a protective metal sheath and a functional core consisting of an appropriate multiple oxide material. In this method [32] the super conducting powder iscontinuously introduced into a metal tube and the filled tube isdrawn by conventional wire drawing methods. Although the powder-in-tube method did yield the highest current density[64] reported so far (>1,000,000 Alcm 2, 77 K, 0 T), it is not reviewed here in detail because it is a metal drawing, not a melt forming process. High temperature superconducting sheath/core bicomponent fibers have also been made by introducing the superconducting material into the core of hollow glass fibers as they are formed under the bushing (Chapter 6.32). 4.5.3 The future ofsingle crystal oxide fibers Continuous sapphire fibers (Chapter 4) and continuous sheath/core bicomponent silicon carbide/carbon fibers (Chapter 3) offer impressive performance as reinforcing fibers and in ceramic and metal matrix composites. Here are some noteworthy commonalties and differences. (a) Single crystal sapphire fibers

Sapphire fibers are hard, strong and scratch resistant to most materials and provide excellent wear surfaces. They can withstand higher pressures than polycrystalline alumina since they lack the grain boundary interface breakdown of the latter. Sapphire fibers transmit ultraviolet, visible, infrared and microwaves and serve as excellent wave guides between 10.6 and 17 microns, and offer durable and reliable IR transmission. By virtue of their high thermal conductivity they can be rapidly heated and cooled. EFG sapphire fibers melt sharply at 2050°C and maintain measurable strength at extreme temperatures [13]. Table XII shows tensilestrength as a function of test temperature from 25 to 1500°C. The room temperature strength, 3.57 GPa, is low for a single crystal fiber but typical for sapphire fibers, irrespective ofprocess. For a single crystal fiber, room temperature strength should be approaching the theoretical value, which is>10 GPa. In fact, its strength is only 40% of an about equal diameter polycrystalline sheath/core bicomponent silicon

Chapter 4

119

carbide/carbon fiber (Chapter 3). This deficiency ofsingle crystal sapphire fibers still needs to be corrected. The tensile strength ofsapphire fibers at 1500·C in this example (Table XII) is 0.55 GPa. In addition, isolated literature values report strength levels of0.40 GPa up to1900·C. These are impressive results since they refer to an oxidative environment. Strength levels of0.80 GPa Table XII. Strength of EFG sapphire fibers at elevated temperatures Fiber test temperature (0C) 25 400

800

1094 1500

Average tensile strength (GPa)

3.57

2.08 1.85 1.03

0.55

Standard deviation (GPa) 0.66 0.49

0.34 0.14

0.09

were observed for silicon carbide/carbon fibers at 1600°C, but the thermal stability of silicon carbide, except for single crystals, is lower than that of single crystal oxides in an oxidative environment especially inprolonged use above 1400°C. (b) Other single crystal oxide fibers

Continuous single crystal oxide fibers, including sapphire, have a number of property advantages over comparable polycrystalline oxide fibers (see Chapter 8). They include microstructural stability at high temperatures, retention of high elastic moduli at high temperatures, and creep resistance. But because of the high diameters, single crystal oxide fibers made by today's processes cannot be woven and must either be wound or used as inserts. Further improvements of the high temperature creep behavior are therefore being sought. The goal is an optimum continuous single crystal oxide, irrespective of process [13] [14] [4849], or new continuous single crystal silicon carbide fibers by laser assisted chemical vapor deposition (Chapter 3) having low diameters «15IJm), near theoretical strength at room temperature, and low creep, high strength and high strength retention at 1600 to 2000°C in oxidative environments. Key opportunities also exist for single crystal fibers for high T, superconductor [67] and for optoelectronic applications [71]. REFERENCES [1) [2J [3] [4] [5] [6]

[7]

D.R. Uhlmann, A kinetic treatment ofglass formation, J Non-crysl. Solids, 7,337-348 (1971). A. Angell, Relaxation in liquids, polymers, and plastic crystals - stronglfragile patterns and problems, J. NonCrystalline Solids, 131-133, 13-31 (1991). F.T.Wallenberger, The structure ofglasses, Science, 267,1549 (1995). F. T. Wallenberger, Melt viscosity and modulus ofbulk glasses and fibers - challenges forthe next decade, in "Present state and future prospects of glass science and technology", Kreidl Symposium, Triesenberg, Liechtenstein, July 3-8, 1994, Glasstech. Ber. Glass Sci.Technology 70C, 63-78 (1997). A. K. Vareshneya, Fundamentals of inorganicglasses, Acad. Press, Boston (1994). H. Tokiwa, Y. Mimura, T. Nakai and O. Shinbori, Fabrication of long single-mode and multi-mode fluoride glass fibers bythe double crucible technique, Electronics Letters, 21 [24], 1130-1131 (1985). M. L. Nice, Apparatus and process forfiberizing fluoride glasses using a double crucible and the compositions produced thereby. US Patent 4,897,100, Jan. 20, 1990.

120 [8J [9J [10) [11] [12] [13] [14] [15] [16] [17J [18) [19] [20] [21] [22] [23) [24) [25) [26] [27] (28) [29J (30) [31] [32] [33] (34) [35) [36] [37] [38]

Chapter 4 G. Y. Onoda, Jr. and S. D. Brown, Low silica glasses based on calcia-aluminas, Journal of the American Ceramic Society, 53(6), 311-316 (1970). F. T. Wallenberger, N. E. Weston and S. D. Brown, Calcia-alumina glass fibers: drawing from super-cooled melts versus inviscidmelt spinning, Mat. Letters, 11 (89), 229-235 (1991). R. E.Cunningham, L. F. Rakestraw and S. A. Dunn, Inviscid mett spinning of filaments, in Spinning wire from motten metal, J. Mottern and W. J. Privott, ed.; AIChE Symposium Series, 74(180), 20-32 (1978). F. T. Wallenberger, N. E. Weston and S.A. Dunn, Inviscid melt spinning: as-spun amorphous alumina fibers, Materials Letters, 2 [4]121-127 (1990). F.T.Wallenberger, N.E.Weston, K. Motzfeldt, and D. G.Swartzfager, Inviscid melt spinning ofalumina fibers: chemical jetstabilization, Journal oftheAmerican Ceramic Society, 75[3], 629-639 (1992). R. S, Feigelson, Growth of fiber crystals, in Crystal growth of electronic materials, E. Kaddis, Ed.,127-145, ElsevierScience Publishers, London (1985). J. Monbleau, Single crystal technology, Product Bulletin, Saphikon Inc., Milford, NH (1994). B. Cote, D. Massiot, F. Tantelle and J. P, Coutures, 27AI NMR spectroscopy of aluminosilicate melts and glasses, Chemical Geology, 96, 367-370 (1992). A. L. Greer, Metallic glasses, Science, 267,1947(1995). F.T.Wallenberger and S.D. Brown, High modulus glass fibers fornew transportation and infrastructure composites and fornew infrared uses, Composites Science and Technology, 51, 243-263 (1994). F. T. Wallenberger, New melt spun glass and glass-ceramic fibers forpolymer and metal matrix composites, in High performance composites: commonalty of phenomena, K. K. Chawla, P. K. Law and S. G. Fishman, eds., The Minerals, Metals and Materials Soc., 85-92 (1994). V. E. Khazanov, Yu, I Kolesov and N. N. Trofimov, Glass fibers, in Fibre science and technology, V. I. Kostikov, Editor, 15-230, Chapman and Hall, London (1995). K. L. Loewenstein, The manufacturing technology of continuous glass fibres, third, completely revised edition, Elsevier, Amsterdam (1993). P. K. Gupta, Glass fibers forcomposite materials, Chapter 2 in Fibre reinforcements forcomposites materials, A.R. Bunsell, ed., Composite Materials Series 2,Elsevier, Amsterdam, 19-71 (1988), R. A. Houpt, R. M. Potter, T. D. Green, D. P. Aschenbeck and C. Berdan, II, Dual-glass fibers and insulation products therefrom, US Patent 5,431,992, July 11, 1995. J. E. Loftus, C. R. Strauss and R. Houston, Method for making fibers by causing one glass to flow around another asthey arespun, US Patent 5,539,596, June 25,1996. W. L. Eastes, D. A. Hofmann, J. W. Wingert, Boron-free glass fibers, International Patent Application, W096/39362, December 12, 1996. J. F. Sproull, Fiberglass composition, US Patent 4,542,106, September 17, 1985. H. Kaplan-Diedrich and G. H. Frischat, Properties of some oxynitride glass fibers, Journal of Non-Crystalline Solids, 184,133-136 (1993). J. Kobayashi, M. Oota, K. Kada and H. Minakuchi, Oxynitride glass and the fiber thereof, US Patent 4,957,883. September 18, 1990. K. Suganuma, H. Minakuchi, K. Kada, H. Osafune and H. Fujii, Properties and micro-structure of continuous oxynitride glass fiber and itsapplication toaluminum matrix composite, J. Mater. Res., 8 [1], 178-186 (1993). F. T. Wallenberger, S. D. Brown and G.Y. Onoda, ZnO-modified high modulus glass fibers, Journal of NonCrystalline Solids, 152,279-283 (1993). P. F.Aubourg and W. W. Wolf, Glass Fibers, in Advances in Ceramics, Vol. 18, Commercial Glasses, pages 51-63, D. C. Boyd and J.F. MacDowell, editors, American CeramicSociety, Westerville OH (1986). F. Rossi and G. Williams, A new erainglass fiber composites, Paper presented at the 28 th AVK Conference, Baden-Baden, Germany, pages 1-10, October 1-2,1997. U. Balchandran, A. N. Iyer, P. Haldar and L. R. Motowidlo, The powder-in-tube processing and properties of Bi-223, J. Metals, 45[9], 54-67 (1993) F. M. Costa, R. F. Silva andJ. M. Vieira, Influence onepitaxial growth ofsuperconducting properties of LFZ BiSr-Ca-Cu-O fibres, Part I.,Physica C, 289, 161 -170 (1997) and Part II., Physica C, 289, 171-176 (1997). T. D. Erickson and W. W. Wolf, Glass composition, fibers, and methods for making same, US Patent 4,026,715, May 31 ,1977 F. T. Wallenberger, N. E. Weston and S. D. Brown, Infrared optical tellurite glass fibers, Journal of NonCrystalline Solids, 144 (1),107-110 (1992). S. D. Brown and G. Y. Onoda, Jr., High modulus glasses based onceramic OXides, Report R-6692, Contract NOw·65-Q426-d, US Department ofthe Navy, October 1966. G. Y. Onoda, Jr. and S. D. Brown, High modulus glasses based onceramic oxides, Report R-7363, Contract N00019-67-C·301, US Department ofthe Navy, February 1968. T. F. Schroeder, H. W. Carpenter and S. C. Carniglia, High modulus glasses based on ceramic oxides, Technical Report R-8079, Contract N00019·69·C-Q150, US Navy Dept., Naval Air Systems Command, Washington, DC, December 1969.

Chapter 4

121

[39] J. R. Davy. Development of calcia-alumina glasses for use in the infrared spectrum, U. S. Patent No. 3,338.694 (1967), Glass Technology. 19 [2], 32-36 (1978). (40) R. Maddison, Calcia-aluminas, Product Bulletins WB37A and WB39B, Sassoon Advanced Materials LTD, Dumbarton. U.K. (1994). [41] P. R. Foy, T Stockert, J. Bonja. G. H. Sigel, jr., R. McCauley, E. Snitzer and G. Merberg, Meeting Abstracts, American CeramicSociety, 94th Annual Meeting, Presentation 7-JXV-92, Minneapolis, MN, April 12-16, 1992. [42] F. T. Wallenberger, N. E. Weston and S. A. Dunn, Melt spun calcia-alumina fibers: infrared transmission. J. Non-Cryst. Solids. 12[1),116-119 (1990). [43] F. T. Wallenberger, N. E. Weston and S. D. Brown, Melt processed calcia-alumina fibers: optical and structural properties. inGrowth of materials for infrared detectors, R. E. Longshore & J. Baars. Eds., Proceedings ofthe SPIE, Society ofPhoto-Optical Instrumentation Engineers, Bellington, WA, Volume 1484, 116-124 (1991). [44] F. T. Wallenberger, Design factors affecting the fabrication of fiber reinforced infrastructure composites, Annual Wilson Forum. Santa Ana, CA, March 20-21, 1995; in Applications of Composite Materials in the Infrastructure, 1-10 (1995). [45] F. T. Wallenberger, High modulus glass fiber reinforced composites for currently emerging infrastructure applications. Proceedings, ASCE Materials Engineering Conference, San Diego, California, November 14-16, 1994. [46] M. Matsukura, Z. Chen, M. Adachi and A. Kawabata, Growth of potassium lithium niobate single-crystal fibers bythe laser-heated pedestal growth method, Jpn. J. Appl. Phys., 36. Part 1, No. 9B. 5947-5949 (1997). [47] J. T. A. Pollock, Filamentary sapphire - The growth of void-free sapphire filament at rates upto 3.0 em/min, Journal ofMaterials Science, 7, 786-792 (1972). [48] T. Mah, T. A. Parthasarathy, M. D. Petry and L. E. Matson, Processing, micro-structure, and properties of AI203-YJAlsO,2 (YAG) eutectic fibers, Ceramic Engineering and Science Proceedings, 622-638, 17th Ann. Conference onComposites and Advanced CeramicMaterials, Am. Ceram. Soc.,Westerville OH (1993). [49] K. J. McClellan, H. Sayir, A. H. Heuer. A. Sayir, J, S. Haggerty and J. Sigalovsky, High strength, creep resistant Y203-stabilized cubic Zr02single-crystal fibers, Ceramic Engineering and Science Proceedings, 651 659, 17th Ann. Conf. on Composites and Advanced Ceramic Materials, Am. Ceram. Soc., Westerville, OH (1993). [50] R. S. Feigelson, D. Gazit, D. K. Fork and T. H. Geballe, Super-conducting Si-Ca-Sr-Cu-O fibers grown bythe laser-heated pedestal growth method, Science, 240,1642-1645 (1988). [51] J. M. Massoubre and B. F. Pflieger, Small diameter wire making through solidification of silicon steel jet, in Spinning wire frommolten metal, J. Mottern and W. J. Privott, eds.:AIChE Symposium Series, 74 (180), 48-57 (1978). [52] F. Fodeur and B. S. Mitchell, Infrared studies of calcia-alumina fibers, J.Am. Ceram. Soc., 79 [9] 2469-2473 (1996). [53] M. Mooney, The viscosity of a concentrated suspension of spherical particles, J. Colloid Science, 6 (2), 162170 (1951). [54] R. J. Diefendorf and E. R. Stover, Pyrolytic graphites: how structure affects properties, Metal Progress. 81 [5J. 103-108 (1962). [55J S. A. Dunn and E. G. Paquette, Redrawn inviscid melt-spun fibers, Advanced Ceramic Materials 2. 804 (1987). [56] B. S. Mitchell. K. Y. Yon. S. A. Dunn and J. A. Koutsky, Phase identification incalcia-alumina fibers crystallized from amorphous precursors, Journal ofNon-crystalline Solids, 152, 143-149 (1949). [57J V. V. Golubkov, A. P. Titov and E. A. Porai-Koshits, On the structure of lithium borate glasses according to small angle scattering data. Soviet Journal ofGlass Physics and Chemistry, 18 [2], 122-129 (1992). [58] D. E. Polk and B. C. Giessen, Amorphous orglassy materials, inRapidsolidification technology source book, edited byR. L.Ashbrook, 213-247, American Society forMetals, Metals Park, Ohio (1983). [59] H. H. Liebermann, Rapidly solidified al/oys. Marcel Decker, Inc.,New York (1993). [60] H. H. Liebermann, MetglaS® product bulletin, Allied Signal, Parsippany, NJ (1993). [611 D. M. C. Narashima, Planar flow casting ofalloys, US Patent 4,142,571 (1979). [62J J. E. Ritter and J. D. Helfinstine, Atougher fiber for the FOG-M, Photonics Spectra, 8. 90-93 (1967). [63] H. Miao, J. C. Dietz, L.A. Angurel, J. I. Pena and G. F. dela Fuente, Phase formation and micro-structure of laser floating-zone grown BSCCO fibers: reactivityaspects, SolidState lonics, 101-103, 1025-1032 (1997). [64J G. Geiger, New record for super-conducting wire, The American CeramicSociety BUlletin, 74(12), 19 (1995). [65] J. Nishii, I. Inagawa, T. Yamagishi. S. Morimoto and R. lizuka, Process for producing chalcogenide glass fiber, US Patent, 4,908,053, March 13, 1990. [66) H. Lin. W. L.Dechent, D. E. Day and J. O. Stoffer, Preparation and properties of mid-infrared glass fibers and poly(chlorotrifluoroethylene) composites, J. Mater. Sci., 32, 6573-6578 (1997). [67] A. Revcoleschi and J. Jegoudez, Growth oflarge high-Tcsingle crystals by the floating-zone method: a review. Progress inMaterials Science, Vol. 42, 321-339 (1997). [68J P. H. Keck and M.J. E. Golay, Phys. Rev.,39, 1297 (1953).

122

Chapter 4

[69] W. Que and S. Lim, Evaluation ofmicro-structure characteristics oflithium niobate single-crystal fiber with magnesium-ion-indiffused cladding, J. Am. Ceram. Soc., 80, (11) 294548 (1997). (70] J. I. Peiia, H. Miao, R. I. Merino, G. F. de la Fuente and V. M. Orera, Polymer matrix synthesis of zirconia eutectics for directional solidification into single-crystal fibers, Solid State lonics, 101-103, 143-147 (1997). (71] W. M. Yen, Preparation of single-crystal fibers, in Insulating maferials foropto-electronics, F. AguIl6-Lopez, Editor, World Scientific, Singapore (1995). (72] W. Jia, H.Yuan, L. Lu, H. Liu and W. M. Yen, Phosphorescent dynamics in SrAI204:Eu 2',Dy3+ single-crystal fibers, J. Luminescence, 76 &77, 424428 (1998). (73] J. K. Weber, J. J. Felton, B. Cho and P. C. Nordine, Glass fibres of pure and erbium- or neodymium-
CHAPTER 5 CONTINUOUS SOLVENT SPINNING PROCESSES F. 1. Wallenberger A liquid phase, as opposed to a vapor or solid phase, includes dispersions, solutions and melts. Several processes, which yield continuous inorganic fibers directly from the melt, have been discussed in Chapter 4. Only one generic process, dry spinning, is known to yield one specific amorphous oxide fiber directly from a liquid phase other than that of a melt. All other processes which start with a liquid phase (see Chapters 8-12) yield first a solid, non-functional precursor or green fiber, and then a functional, nano- or polycrystalline ceramic fiber. Such refractory ceramic fibers are therefore directly derived from a solid phase, a precursor or a green fiber, and only indirectly from a liquid phase. 5.1 Dry spinning of silica glass fibers Silica glass fibers have higher glass transition temperatures (1150-1200·C vs. 550-600·C) and lower coefficients of thermal expansion (0.5 x 10.6 vs. 5 X 10.6 "C:') than general purpose borosilicate and boron free E-glass fibers. In addition, they also have lower dielectric constants, (3.4 vs. 6.1) and loss tangents (2 x 10..\ vs. 30 X 10..\). This balance of properties translates into significant product advantages (Chapter 6) but demands a premium price. 5.1 .1

Process concepts

Only dry spinning facilitates the formation of amorphous silica glass fibers from viscous solutions. Two process variants are known. One uses a viscous water glass or sodium silicate solution, wherein water is the solvent, and produces a pure and relatively low cost fiber. The other uses a viscous tetraethylorthosilicate solution wherein alcohol is the solvent, and produces an ultrapure premium fiber. Ineither process, the viscous solution is spun from, or extruded through, multiple spinneret orifices into a hot column that removes the solvent. Before the as-spun fibers are wound onto a package, they are solidified in a final, high temperature curing step. Ineither process, the final curing step may bean integral part ofthe entire process sequence, or it may be a separate step. In the first process, the as-spun precursor fiber is already an amorphous silica glass fiber, except that it is not yet entirely free of solvent and not yet fully consolidated. Inthe second process, the as-spun precursor fiber still has a different chemical composition than the final silica fiber, but it has the same morphology, i.e., an amorphous structure. In these examples, a glass fiber is therefore directly derived from the liquid phase, i.e., from a viscous solution.

124

Chapter 5

The same generic dry spinning process can be used to fabricate the precursor fiber for a carbon fiber from an infusible polymer, such as polyacrylonitrile, for a polycrystalline aluminate fiber from a sol-gel or for a polycrystalline alumina fiber from a slurry. Again, a high temperature curing step isrequired toconvert the as-spun, amorphous precursor fiber into the final functional fiber. In these cases, however, an amorphous polycrylonitrile precursor fiber changes into a carbon fiber, and an amorphous aluminate precursor fiber into a crystalline aluminate fiber. The final functional fiber is therefore directly derived from a solid and amorphous precursor fiber and only indirectly from a liquid phase, i.e., a melt or sol-gel, respectively. Glass fibers, which are derived from a melt, were discussed in Chapters 4 and 5. Glass fibers, which are derived from a viscous solution, will be described in this chapter. The most important applications of glass fibers made by any process will be discussed in Chapter 6. Carbon fibers and ceramic carbide and oxide fibers, which are derived from solid precursor fibers, will be discussed, along with their applications, in Chapters8 to 12. 5.1 .2 Pure silica fibers from water glass solutions The one step process for the fabrication of pure silica sliver from sodium silicate [1-3J is an adaptation ofthe generic dry spinning process that has been practiced for over 50 years in the fabrication of polymer organic textile fibers. Acrylic polymers, unlike polyesters or nylons, are infusible and cannot be melt spun. They can be dry orwet spun from viscous solutions. The model for inorganic dry spinning processes is the dry spinning process by means of which acrylicfibers such as Orion, are spun using dimethyl formamide as the solvent (Table I).

Table I. Sol-gel processing and dry spinning Fiber liquid phase Polyacrylonitrile solution acrylic polymer Glass solute dimethyl formamide Solvent removeDMF Solidification removeDMF Wet treatment remove water Drying step Precursor fiber Heat consolidation one or two Process steps commercial fiber Final product

Pure silica solution waterglass water remove water removeNa+ remove water is not isola ted in-line step one commercial fiber

Ultrapure silica sol-gel tetraethoxylsilane remove alcohol remove alcohol

is wound up

separate step one or two commercial fiber

The inorganic dry spinning process yields a continuous silica sliver yarn but not a continuous multifilament product directly in a single process sequence, starting with a viscous water glass solution in water. To wit, a sliver yarn is a continuous assembly of cohesive, slightly bonded staple fibers in an essentially parallel arrangement. In general, discontinuous fibers, such as staple fibers, sliver yarns or insulation wool are not covered in this book; they neither represent advanced technology or products, nor do they serve advanced markets. Because ofits high value, silica sliver isan exception. The commercial process (Figure 1) utilizes a viscous solution of water glass in water. Conceptually, the process can be described by pointing to Figure 2, i.e., the ternary Na20 ·Si02"H20 phase diagram. Water is removed by distillation and polysilicate anions grow

Chapter 5

125

by a polycondensation reaction until a fiber forming viscosity is attained. A molar ratio of Si02/Na20 equaling 2.48 has a viscosity of 240 poise at30·C [4J, and this solution isextruded through 120-hole spinnerets at 750-800 m/min. The drying column removes excess water and enables the formation ofsolid but metastable water glass fibers still containing about 20% water.

3-

...

4I 6

Figure 1. Process forthe fabrication ofsilica sliver from water glass. A viscous waterglass solution isextruded (1) through tiny spinneret holes (orifices) into a drying chimney (2). A spin finish is applied to the resulting sodium silicate filament yam by a kiss roll (3). The yam is taken upby a drawing drum and scraped offprior to complete rotation onthe drum, intermingled ina conical chamber and drawn laterally (4). The resulting sodium silicate sliver (acohesive staple yam) is transported over agodet (5), placed on a conveyer belt and passed through an acid bath (6), a washing station (7) and adrying zone (8). Silica sliver isformed inthe calcining zone (9), and a textile sizing is applied (10) before it isdried (11), further intermingled inair(12) and wound onto a bobbin (13). Redrawn from H. D. Achtsnit, Textile silica sliver, its manufacture and use, U. S. Patent, 5,567,516, October 22, 1996.

These fibers are immediately taken up by a drawing drum, scraped offthe drum, intermingled in a conical chamber, and drawn off[3] vertically. The resulting sliver yarn is then fed atabout 94 m/min into a bath containing salt solutions of hydrogen ion bearing acids. Within a residence time of0.5 to 1 min, the acid bath removes almost all Na'ions. The drying step in the continuous process sequence is carried out at about 150·C. It removes free water and produces sliver yarn containing weak nanoporous silica fibers with a highly hydroxylated surface. Consolidation of the nanoporous fibers at 800·C still within the continuous process sequence causes the fibers todehydroxylate and shrink, and also to increase instrength. This process facilitates the formation of pure silica sliver having linear densities of 135 to 330 g/1000 mof yarn [5J. The individual staple fibers have a length ranging from 50 to 1000 mm, and an average diameter of 10 urn. Individual silica fibers have a lower density (1.8 to 2.0 g/cm 3) than that of E-glass (2.5 g/cm 3) and polycrystalline ceramic fibers (3.5 g/cm 3) . Unconsolidated fibers having a density of 1.8 g/cm 3 are nanoporous and weak, but consolidated fibers, having a density of 2.0 g/cm 3, are solid, moderately strong, and retain their strength to 1000·C. Sliver, yarns, braids, and wovens are commercially available [5]. Silica fibers made by this route have higher strength than those made by the sol-gel route

126

Chapter 5

(see below) but lower strength than that of commercial E-glass fibers. For a discussion of properties as well as applications see Chapter 6.

80

40

60

20

Figure 2. Ternary Na20-Si02-H20 phase diagram. The process path leads from a viscous waterglass solution (1) to sodium silicate filaments and sodium silicate sliver (2), tohydrated silica sliver (3) and finally towater-free silica sliver (4). Redrawn from G.H. Vitzhum, H. U. Herwig, A. Wegerhoff, and H. D. Achtsnit, Silica fiber forhigh temperature applications, Chemiefaseml Textilindustrte, 36/88, E-126-127 (1986).

5.1 .3 Ultrapure silica fibers from viscous sol-gels The first process step is a polycondensation reaction . Tetraethylortho-silicate (TEOS) is dissolved in pure ethanol (C2HsOH), and a dilute solution of the catalyst, hydrochloric acid (HCI), is added <25°C. The resulting reaction mixture is heated to, and maintained at 70ao°c,until the polymerization has reached a fiber forming viscosity that facilitates dry spinning (Equations 1-5). The second step in the process is fiber formation by dry spinning. The process is very similar tothat shown inFigure 1, except that the drying and curing protocol is less complicated. Fibers can be drawn from gels near the gel point, Le., when the viscosity is greater than log 2.5 poise.

(ROh Si-OR+HOH

~(ROh

(ROh Si-OH +OH -Si(ORh (ROJJ Si-OR+OH -Si(OR))

~(ROh ~

fora chain-like polysiloxane polymer: nSi(OR).+HP

OR~ I

~RO -Si

6R

Si-OH +ROH (R=C2Hs) Si-O-Si(OR)) +H 20

(RO)) Si-O-Si(OR)) + ROH

0t

I O -Si

6R

n-l

(1 )

(2) (3)

OR I 0 -Si-OH+(2n-l)ROH 6R

(4)

Chapter 5

127

for complete hydrolysis: n Si(OR).J + 4n H]O nSi(OH).J

~

(5)

n Si(OH).J + 4n ROH

(6)

~nSiO]+2nH]0

n Si(OR).J +2n H]O

~

(7)

n SiO ] +4n ROH

The third step in the process (Equations 6-7) involves the conversion of the precursor, green, orgel fiber into a silica glass fiber. This is achieved by slowly heating the silica precursor fiber to 800-900°C in air with low drawing tension [6]. The gel fiber contains large amounts of solvent and pendant C2HsO groups (Equation 5). Most of the solvent is eliminated below 200·C. A fiber with a porous structure and a high specific surface area is obtained [7]. The pendant C2H sO groups are eliminated between 200-400°C. Additional weight loss and additional fiber shrinkage take place [16] above 400·C. The silica fiber is sintered to full density (2.2g/cm 3) at900°C[7-8]. The phase diagram shown in Figure 3 relates composition tospinning performance. Solutions with compositions in region (A) form gels and fibers with optimum process conditions. Solutions with compositions in region (B) form gels but are unspinnable, and solutions in region (D) are dominated by immiscibility. Optimum dry spinning compositions and fiber forming conditions yield unbranched polymer chains and avoid microcracking of the gel [812]. Depending on the spinning conditions, the diameter of continuous silica filaments ranges from 5 to 10 IJm [7].

TEOS (%)

H20 (%) L...-_-"T"""--"","--"""T'"--~--~ ETOH (%)

20

40

60

80

Figure 3. Ternary TEOS-H20-C2HsOH phase diagram. Gels are formed in two regions of the phase diagram (A and B). Fibers can bespun from gels in region (A) butnot from gels in region (B). Region (D) is dominated by immisdbility. When fibers can bespun, (0) indicates that more than two thirds have a circular cross section, and (0) indicates formation offibers with non-circular cross sections [12]. Reproduced bypermission.

128

Chapter 5

Sol-gel derived and melt spun silica fibers are amorphous [13]. Due to the disordered character of the Si04 tetrahedra skeleton, the density ofglassy silica and silica glass fibers is lower than that of crystalline silicas (i.e., 2.2 g/cm 3 versus 2.33 g/cm 3 for cristobalite and 2.65 g/cm 3 for quartz). Sol-gel derived silica glass fibers with a 20 urn diameter have very low room temperature strength (800 MPa) [12] when compared with that of E-glass fibers (>3000 MPa) and melt spun silica glass fibers (>5900 MPa). A further strength loss occurs when amorphous silica fibers are crystallized [14]. For a discussion of properties and applications ofsilica fibers see Chapter 6. 5.2 Silica fibers by other processes Two of the four known silica glass fiber processes are dry spinning processes. They are continuous processes, which afford either a continuous sliver or staple product. i.e., a cohesive array of discontinuous fibers, or a continuous multifilament yarn, respectively. The other two processes are continuous processes yielding continuous fibers and multifilament yarns. Ofthese, the preform process (Chapters 4 and 6) relies on the use of silica preforms, and produces continuous, ultrapure silica glass fibers, or quartz fibers. This process yields the strongest silica glass fibers on record. The fourth process which yields high silica glass fibers relies on acid leaching of borosilicate or aluminosilicate precursor fibers in fabric form (Chapters 4 and 6). This is the oldest and least expensive process. Acid leaching removes most of the compositional oxides other than silica from a precursor fabric. Individual fibers can be leached also, but they are not satisfactorily converted into sliver, braids of woven fabrics. Process details and properties of melt spun and acid leached silica substrate processes have been discussed in Chapter 4. Commercial applications ofall four silica glass fibers will be discussed in Chapter 6. REFERENCES [1)

A. Wegerhoff, H. Zengel, H. Brodowski, H. Beck, E. Seeberger, G. Steenken and K. Hillermeyer, Watercontaining water glass fibers, US Patent, 4,471 ,019, September 11, 1984. (2) A. Wegerhoff and H. D. Achtsnit, High temperature resistant fibrous silicon dioxide material, US Patent, 4,786,017, November 22, 1988. [3) G. H. Vitzhum, H. U.Herwig, A. Wegerhoff, and H. D. Achtsnit, Silica fiber forhigh temperature applications, ChemiefasemfTextilindustrie, 36/88, E-126-127 (1986). [4) H. D. Achtsnit, Textilesilica sliver, itsmanufacture and use, U. S. Patent 5,567,516, October 23,1996. [5) ProductBulletin, Silfa Silica Yams, Ametek, Haveg Div., Wilmington, DE (1996). [6) K. Matsuzaki, D. Arai, N.Taneda, T. MUkaiyama and M. Ikemura, Continuous silica glass fiber produced by sol-qel process, J. Non-crystalline Solids, 112,437-441 (1989). [7J T. Hashimoto, K. Kamiya and H. Nasu, Strengthening of sol-qel derived Si02glass fibers by incorporating colloidal silica particles, J. Non-crystalline Solids, 143,31-39 (1992). [8J E. M. Rabinovitch, SO/ile/ processing: general principles, in So/-Gel Optics, Processing and Applications, L.C. Klein, ed., 1-37, Kluwer Academic Publishers, Boston (1994). [9] M. Pozo de Femandez, C. Kang and P. L. Mangonon, Process ceramic fibers by solilel, Chemical Engineering Progress, 49-53, Sept. 1993. [10) S. Sakka andK. Kamiya, The sol-qel transition inthe hydrolysisofmetal alkoxides in relation tothe formation of glass fibers and films, J. Non-crystalline Solids, 48, 31 46(1982). [11) S. Sakka, Solilel synthesisofglasses: present and future, Ceram. Bull., 64(11)1463 (1985). (12) K. Kamiya and T. Yoko, Synthesis of Si02glass fibres from Si (OC2Hs)4 - H20-C2HsOH -HCI solutions through solilel method, J. Mater. Sci., 21, 842-848 (1986). [13JK. Kamiya, R. Uemura, J. Matsuoka and H. Nasu, Effect ofpreheat treatment onthe tensile strength of solile l derived Si02glass fibers, J. Ceram. Soc. Japan, 103 [3J245 (1995). [14) W. Zhou, Y. Xu, L. Zhang, X. Sun, J. Ma and S. She, Crystallization of silica fibers made from metal alkoxide, Mater. Letters, 11 [1 0-1 2)352-354 (1991).

CHAPTER 6 STRUCTURAL SILICATE AND SILICA GLASS FIBERS F. T. Wallenberger Fiberglass is by far the world's most important continuous inorganic composite reinforcing fiber. It outsells continuous carbon fibers, boron fibers, and ceramic fibers by a wide margin. Typical general and special purpose fiber compositions contain >50% silica. Commodity or general purpose products are characterized by universal applicability, large sales volumes, and low unit cost. They represent nearly 99% of the commercial fiber glass market. The remaining glass fibers are niche or special purpose products, characterized by special, and therefore premium, properties, small sales volumes, and high unit costs. They represent 1% of the fiber glass market. 6.1 General purpose silicate glass fibers Currently there are two commercial general purpose fibers on the market (see Chapter 4). Both are variants of products within the broad category of E-glass fibers. One is a generic boron-containing, orcalcium aluminum borosilicate, glass fiber [1 -2] and the other isa generic boron-free, orcalcium magnesium aluminum silicate, glass fiber [3-4]. 6.1 .1 Commercial fiber forming processes The use of continuous glass fibers dates back to antiquity. Some of the earliest Egyptian glass vessels were made around 1630 BC by Winding glass fibers by hand around a core of shaped clay [5]. Interestingly, filament winding precedes glass blowing. The commercial production of glass "silk" started nearly a century ago in Germany [5], whereby fibers were drawn out of holes, which had been drilled into refractory furnaces, and wound onto wheels and drums. The production of continuous glass fibers as they are known today started in 1936 with a process whereby streams ofmolten glass were pulled from small furnaces, called bushings, and attenuated [2]. The production of discontinuous or staple fibers is accomplished either by chopping continuous fiber strands into uniform lengths ranging from 3-40 mm, or by producing a loose mat orwool made up ofindividual fibers ina centrifuge process. This rotary process generally yields short fibers which are not straight but have discrete and often non-uniform lengths. With one exception (Le., a bicomponent insulation product, Chapter 6.4.4), only continuous glass fibers will be discussed in this chapter. Continuous fibers can be formed in a direct melt process (Figure 1), where the melt is formed from mixed raw materials, or in a remelt process, where the melt is formed from marbles which had previously been made from batch ina separate process step. Continuous fibers

130

Chapter 6

made byeither process can be direct drawn, wound, and sold, or wound, redrawn and sold. In either process, it is useful to coat strands or bundles of -100 filaments rather than individual fibers, so as to protect them from abrasion fram the point of their formation to the point oftheir incorporation into a composite matrix.

1

------ ---- - - -- - -- ------,

Batch silos

/0 ,...----------, Furnace

,

1--

I I

I

I I I

Weighing and I mixing

I

I

I I I

Melting

, , I

Batch charging

Forehearth

...,

Bushing Fiber formation Sizing application Strand formation Traversing Winding (or chopping)

Figure 1. Direct melt process tor the production of continuous glass fibers. Redrawn from J. F. Dockum, Fiberglass, in Handbook of reinforcements for plastics, J. V. Milewski and H. S. Katz, Editors. 233·286, Van Nostrand Reinhold, New York, NY (1987).

Chopped strands are short, coated fiber bundles and are used to reinforce thermoset or thermoplastic matrix materials, or else to form, with additional binder, a chopped strand mat on a conveyor. Rovings consist ofa number ofstrands which are wound without twisting on a cylindrical package, and are suitable for weaving, filament winding, or for spraying with additional resin onto a mold. Yarns consist of a number of strands or ravings which are twisted before they are wound; they are also used in weaving processes. 6.1 .2 Commercial commodity glass fibers The age of the modern glass fiber started in 1940 with a patent claiming new compositions and the production ofcontinuous fibers (6) based on the dominating quaternary eutectic (60% Si02- 9% Ab03- 27% CaO - 4% MgO). (a) Evo/ution ofborosificate E-g/ass fibers The first composition is often called the original E-glass (Table I). It had a liquidus temperature of 1233°C, a log 3 forming temperature of 1288°C, and a ~ T of 55°C between the forming and liquidus temperatures [3). Fluxes included 6.6-9.5% B203, 2.0-3.0% CaF2and

131

Chapter 6

0.0-3.1 % Na20 . The second generic E-glass variant was disclosed in 1943. It was also a derivative of the quaternary eutectic [8) but specifically covered 9-11 % 8203as an auxiliary flux. These adjustments reduced the liquidus temperature to 1120°C and the 1000 (log 3) poise forming temperature to 1200°C, thus offering a ~ T of 80°C.

Table 1. Com position of commercial general purpose glass fibers Ch emi cal composition in Weight percen t Year SiO, B,O, A~O,

CaO MgO TiO, ZnO Nap K,O Fe,O, F,

Original E-glass Compositions 1940 1943 60.0 54.0 10.0 15.0 14.0 20.0 17.5 5.0 4.5

Standard E-glass 1951 54.5 6.6 14.0 22.1 0.6 0.5 0.8 0.2 0.2 0.5

Fiber forming T. (Log3), °C Liquidus temperature, °C Pristin e strength, GPa Elastic modulus, GPa Softening point, °C Uppe r in-use limit, °C Emission

1288 1233 3.4

References

[1) [6)

72

840 620 No

B/F

1200 1120 3.4

1200 1064 3.4

72

72

840 840 620 620 Emission control needed [1) (8) [1)[7)[9)

Boron- free fibers ZnO-rree w ith ZnO 1977 1996 58.2 60.01 11.6 21.7 2.0 2.5 2.9 1.0 0.1 tr,

12.99 22.13 3.11 0.55 0.63 0.14 tr.

1230 1260 1163 1200 3.5 3.5 80 81 880 916 660 690 No boron /fluorine emission (12) [13) [3] [4]

In 1951 , the E-glass composition was simplified (Table I) by nearly eliminating MgO, previously a key ingredient. (2) The batch cost was reduced by reducing the amount of 8203, the most costly ingredient. (3) The process stability was enhanced by a further reduction of the liquidus temperature [7) from 1120 to 1064°Cand by a further increase in the ~T to 136°C. This significant reformulation of the E-glass composition meant a fundamental change from a modification of the quaternary eutectic, 60% Si02- 9% Ab03- 27% CaO - 4% MgO, to a modification ofthe ternary eutectic, 62.2Si02- 14.5% Ab03- 23.3% CaO. The new composition has since become the generic borosilicate E-glass standard. (b) Boron- andfluorine-free E-g/ass fibers By the mid-1960s, multiple concerns arose about the use of boron and fluorine. (1) Boron oxide is a costly batch ingredient and up to 15% of this ingredient in a given melt can volatilize. (2) Fluorine presents a similar problem [4). (3) The atmospheric temperature above the glass melt in most E-glass production furnaces is between 1400 and 1500°C and the exhaust can react atthese temperatures with potassium, sodium and sulfur oxides to form

132

Chapler6

particulates which are then emitted [4]. The emission levels of these particulates are regulated in both North America and Europe, requiring aftertreatment with pollution control devices such as scrubbers. The first boron-free glass fiber was commercialized in 1977, but large amounts of ZnO and Ti02 (>2.5% each) were required to reduce the liquidus temperature to 1163°C (9). This fiber was generically known as ECR glass (CR stands for corrosion resistance) but because of its costly ingredients, it remained a premium, special purpose, niche product. More recently, a general purpose glass E-glass fiber was commercialized that is not only boron- but also zincfree [3]. Eliminating boron and zinc oxide from a composition eliminates two costly batch ingredients and addresses the emission problem, but it also raises the liquidus, forming and operating temperatures (Table I). 6.1 .3 Structures and properties E-glass is byfar the most important inorganic fiber. Like the other inorganic fibers which will be discussed in Chapters 8 to 10, it is mostly used to reinforce composite structures and, as such, is the lowest cost candidate with the widest range of applications. Thus, E-glass has become the benchmark for the product designer against which the value-in-use of other organic and inorganic fibers isbeing measured. (a) Mechanical properties The mechanical properties of the two general purpose glass E-glass fibers are comparable (Table III of Chapter 4.2.1). Specifically, single filament strength (3.1-3.8 GPa) and elongation-at-break (4.5-4.9%) are about the same for both fibers, but the modulus of borosilicate E-glass is lower (72 vs. 81 GPa) than that of the boron-free E-glass [3-4). Tensile strength and break elongations were determined by ASTM method D2101 at 23°C (2-inch gauge). The modulus was obtained by the sonic method (4). Fibers are always purchased on a cost-per-weight basis but they are generally tested on a function-per-weight basis, and are generally used on a length (volume) per shape (volume) basis. Two derived properties that are important to the product designer are specific fiber strength and specific fiber modulus, Le.,fiber strength and fiber modulus corrected for density, or weight at equal volume. For composites aimed at weight sensitive transportation uses, a similar correction is often applied to compare the weight of different reinforcing fibers at an equal fiber volume fraction . Figure 2 compares the specific tensile properties of important reinforcing fibers. Borosilicate E-glass serves as the reference point for the following discussion since it isa general purpose fiber offering a universally acceptable balance of properties at low cost. The other inorganic fibers shown in Figure 2 are specialty fibers offering specific mechanical properties which are not attainable with E-glass. However, 8-glass, a generic high strength fiber, costs -5x as much as E-glass. Carbon oraramid fibers cost -1Ox as much as E-glass and boron orsilicon carbide fibers cost >100x as much as E-glass. In summary, the other inorganic reinforcing fibers offer higher value-in-use athigher cost. E-glass will have to be replaced by a specialty reinforcing fiber if the design specifications for a new composite part exceed the capabilities of E-glass. For example, higher part stiffness may be required atequal part size, or equal part stiffness at reduced part size. Alternatively,

133

Chapter 6

special purpose glass fibers may be needed to achieve high or low electrical conductivity, superior alkali resistance, orhigh service temperatures. Still, E-glass will continue toserve as the material of choice to reinforce the largest number of new composites having the highest sales volumes and lowest costs.

I

I

Carbon . TORAYCA

rtooo

I I

I

• Polyethylene

I

SPECTRA 1000

3 Aramid

I

Carbon

KEVLAR 29

E :::!:

Aramid

. Aramid

0, ~

I

KEVLAR 149

c

1ii

MAGNAMITE IM7

• KEVLAR49

•

.r:

1

2

$

•

S·Glass

"(;;

c

.!!!

.g

Silicon Carbide

'iiia.

•

NICALON

•

C/)

E·Glass

ISAFFIL

• TITANIUM 6A 1·4V • ALUMINUM L65

o

Carbon

. Alumina

THORNEL P120S

• Beryllium

• Steel (bulk)

100

200

300

400

Specific tensile modulus, Mm

Figure 2. Specific tensile properties of advanced reinforcing fibers. Redrawn from T. Starr, Carbon and high performance fibers, directory and databook, edition 6, page viii, Chapman & Hall, London (1995).

(b) Other fiber properties

Among the general purpose fibers, the boron-free E-glass fibers exhibit higher acid resistance than borosilicate E-glass fibers (Table II). In this test [4], single filament fibers were weighed, immersed ina 10% sulfuric acid solution for a period of24 hours, and re-weighed. The weight loss for borosilicate E-glass exceeded 40% ; that for boron-free E·glass was less than 5%. After 499 days in a 5% sulfuric acid solution, a spray gun laminate made from boron-free Eglass lost half as much flexural strength as a comparable laminate made from a boroncontaining E-glass fiber. Borosilicate and boron-free E-glass exhibit equivalent corrosion resistance and modulus retention in water tests.

134

Chapler6

Table II. Physical properties of commercial general purpose glass fibers [3] [4] Test method

Property

Weight loss • Bare fibers Softening point ASTM C 338 Refractive index Oil immersion Thermal lin. expansion ASTM D 696 References: • In 10% H,SO,

Unit

%/24 hours °C Bulk glass Ppmrc

Borosilicate E-glass 40 830-860 1.547-1.562 5.4

Boron-free E-glass >5

916 1.560-1.562 6.0

Boron-free E-glass fibers [4] have a higher softening point and estimated upper service temperature. The softening point of a fiber is the temperature at which a fiber will deform under its own weight. The estimated upper service temperature of a given fiber is the temperature at which it retains useful strength. The softening point of boron-free E-glass is 56-86°C higher than that of borosilicate E-glass (Table II), and affords a higher useful service temperature «700 vs. <600°C). This product advantage is the result of a process penalty, te., its fiber forming temperature [3]. Table III. Electrical bulk properties of general purpose glass fibers [3] [4] Property

Test method

Unit

Dielectric constant Dissipation factor Dielectric breakdown Volume resistivity

ASTMDl50 ASTMD150 ASTMDl49 ASTMD257

23°C/lMHz

23°C/1 MHz volts/mil logI0/23°C

Borosilicate E-glass 6.9-7.1 0.0001 262 22.7-28.6

Boron-free E-glass 7.0 0.0001 258 28.1

The electrical properties of both general purpose glass fibers are comparable (see Table III). The refractive index is a property ofconsiderable importance with regard to the appearance of a glass fiber in a laminate or composite. Electrical properties and the coefficient of linear expansion of borosilicate and boron-free E-glass were measured on bulk annealed samples. The differences intest results between both fibers are considered tobe insignificant [4]. 6.1.4 Commercial products and applications General purpose E-glass fibers are the dominant reinforcing fiber of composites in today's market. Continuous glass fibers can be cut into staple or chopped strand. Continuous fiberglass structures which are used to service a variety of markets include (1) rovings, yarns, chopped strands and milled fibers, (2) woven rovings, weaver yarns and braids, and (3) chopped strand, continuous and combination mats. Several generic commercial applications offiberglass reinforced composites are shown in Figure 3. Rovings are essentially untwisted bundles of fiberglass strands wound up in parallel on cylindrically shaped packages. They are used in open lay-up moldings, woven fabrics, rods, and tubes. Woven rovings are used for open lay and press moldings. Yarns [1] consist of twisted yarn bundles. Weaver yarns are used for electrical and aircraft laminates, insulating tape, window shades, and filtration applications. Their major use is in printed circuit boards. Chopped strand serves as a reinforcement in a variety of markets. For example, it is used to

Chapter 6

135

reinforce thermoplastic composites whereby it facilitates part consolidation in the automotive industry.

Figure 3. Commercial applications offiberglass reinforced polymer matrix composite structures. Manifold intake (top, left): shower and tub unit (center, left); printed wiring board (bottom, left); boat hull (top, right): pipes (center, right); recreation vehicle body (bottom, right). Reproduced with permission from PPG Industries, Incorporated, Pittsburgh, PA.

136

Chapter 6

Mats are non-wovens which consist of specific arrays of glass fibers held together by a suitable binder. A continuous strand mat consists of fine strands of glass fibers either held together by a resin binder or by stitch bonding, and it is used in sheet molding and reinforced thermoplastic sheeting. A chopped strand mat is a non-woven fabric consisting of a random array ofchopped glass fiber strands. It is used to reinforce composites deployed in boat hulls and decks, car bodies, sheeting and tanks. A roofing mat is used to manufacture roofing shingles. Reference [1] should be consulted for details.

6.2 Special purpose silicate glass fibers Specialty glass fibers, whether commercial or experimental, have superior mechanical, electrical, orother functional properties relative tothose exhibited by general purpose E-glass fibers. Generic specialty fibers like E-glass are known byletter designations. "HS" stands for high strength, "A" for high alkali content, "L" for high lead content, "HM" for high modulus, "AR" for high alkali resistance [21], and "ECR" for high acid corrosion resistance. These designations are far from consistent. "Z" has been used for zirconia [211 as well as ZnO modified compositions [29], and trade names have been applied to specific general purpose and specialty fibers. 6.2.1 High strength - high temperature fibers This chapter deals with specialty glass fibers having a commercially useful balance of high strength (HS), high temperature resistance (HT) and intermediate modulus (1M). Higher service temperatures are a more important design factor than higher strength or stiffness. Higher in-use temperatures require a fiber composition with higher thermal stability than that ofE-glass, Le., higher forming temperature, higher process energy and higher overall cost. For a fiber to be usable athigher service temperatures than that afforded by a typical general purpose glass fiber, it must have a higher softening or glass transition temperature, and will therefore have a higher forming and liquidus temperature. Although these fibers are made in a conventional melt spinning process, the ceramic furnace linings and the precious metal bushings must withstand higher process temperatures. Higher materials and process cost translate into a premium price for these fibers. Higher strength orhigher stiffness alone does not necessarily require a more costly fiber. In a composite, the same result can be obtained by increasing the amount (weight) of a general purpose fiber, except in aircraft composites where high specific strength and high specific stiffness may be required toachieve a low part weight. (a) Process and products

Experimental glass fibers are known to possess individually either over 1.3x the strength of general purpose E-glass fibers (4.5vs. 3.4 GPa), over 100% higher softening points orservice temperatures (1250 vs. 600°C), or over 3 times the stiffness or modulus (248 vs. 72 GPa). Specifically, intermediate modulus (1M) glass fibers have a modulus of 85-95 GPa, high modulus (HM) glass fibers have a modulus of up to 131 .5 GPa and ultrahigh modulus (UHM) glass fibers can have a modulus up to 248 GPa. Three of the HS-glass fibers shown inTable IVare commercial products. They are S- and Teglass, Le., derivatives of the ternary SiOrAb03-MgO eutectic, as well as R-glass, a derivative

137

Chapter 6

of the quaternary Si02-Ab03-CaO-MgO eutectic (Table IV) . The generic letter designation "S" applies also tosimilar ternary compositions. Table IV. Highstrength (HS-) silicate glass fibers Fiber Modified Eutectic Composition, wt. %

sio,

Alp, MgO CaO Zr,O, B,O, TiO, Fe,O,

Nap

Pristine Properties Strength, GPa Elongation, % Modulus, GPa Density,g/ cc Sp. Strength, Mrn Sp. Modulus, Mrn Softening point,OC Practical UseT., °C Structure Morphology References:

5-glass ternary

Te-glass

65.5 25.0 9.5

65.0 23.0 11.0

ternary

R-glass quaternary 60.0 25.0 6.0 9.0

Experimental ternary ternary 64.5 24.6 9.4 0.5

67.5 15.3 15.5

1.0 1.0 1.5

tr. tr ,

0.1 <0.1

4.6 5.2 88-95 2.53 1.85 35.0-37.5 1056 <1000

4.7 5.5 84.3 2.49 1.90 34.0 975 <900

4.4 5.2 86.0 2.52 1.74 34.0 975 <900

Properties like 5-glass but improved melt and forming uniformity

1.2 95.3 2.70 0.45 35.2 <1295

amorph. glass

amorph. glass

amorph. glass

amorph. glass

nanocryst. glass -cer,

(12) (14)

(12)

(12)

(10) (14)

(16) (18)

Zr02 and B203are fluxes (see Table IV) . The addition of 1% Zr02, to the ternary Si02-Ab03MgO eutectic (l.e., S-glass) composition causes a major drop in melt viscosity [14], and the resulting Te-glass fibers have an >80°Clower softening and service temperature than S-glass fibers. The addition of 1% 8203has a similar, if not more pronounced effect. It reduces both softening point and service temperature. In contrast, addition of 5% Ti02 to a ternary composition was found to raise its modulus, density, and service temperature [16] [18]. The resulting experimental fibers have a higher service temperature (1295°C) than commercial aluminate fibers (1200°C), silica fibers (1090°C), and S-glass fibers «1 OOO°C), but the development ofa major nanocrystalline phase reduces its overall strength by over 50% . The commercial HS-glass fibers such as S-glass have higher strengths and higher moduli than E-glass, and importantly also higher elongations at break (Table IV). Thus, they have higher mechanical (not to be confused with fracture) toughness which translates into higher damage, including ballistic impact, resistance. To wit. the energy required tobreak a relatively elastic fiber is proportionate to the area under the stress-strain curve (Figure 4). The impact damage resistance of S-glass is therefore higher than that than of E-glass, and of nominally stronger, but more brittle, standard modulus (SM) carbon fibers having -1/10th the elongation at break. For very different reasons, the damage resistance of amorphous HS-glass fibers appears tobe the same as that ofsinglecrystal sapphire fibers. Specific properties (properties at equal density) are an important design consideration for transportation, especially aircraft, composites, where higher functionanty at lower weight translates into higher value-in-use. At room temperature, HS-glass fibers (Table IV) have

138

Chapter 6

about equal specific strel}9th (174-190 Mm) and specific moduli (34-35 Mm). General purpose glass fibers, e.g., E-glass, afford a specific modulus of20 Mm at-1/5th t~e cost, and standard modulus carbon fibers offer a specific modulus of 131 Mm at-2x the cost.

Toughness versusmorphology

<0

a.

e

12.0

£

C,

c

l1!

u;

10.0

Singlecrystal sapphire fiber

~

'w c: ~

8.0

6.0

4.0

2.0 4 2 3 Percent elongation (strain)

5

6

Figure 4. Toughness and impact damage resistance of inorganic fibers. High strength (HS) glass fibers, e.g., Sglass, have higher mechanical toughness than borosilicate E-glass fibers.

The utility of HS-glass fibers is ultimately determined by their softening point, the temperature above which a fiber will rapidly deform under its own weight. In practical terms, R- and Teglass retain usable functionality to about 900°C, S-glass and the B203-modified S-glass to about 1000°C, and the Ti02 modified glass ceramic fiber to 1295°C,while E-glass is limited to 620°C (Table IV). In contrast, the upper service temperature limit of unprotected carbon fibers is 350-500°C in an oxidative environment, that of protected or coated carbon fibers is -1000°C in an oxidative environment, and that of unprotected or coated carbon fibers is »1500°Cinan inert environment such as helium. HS-glass fibers retain useful properties at high temperatures. In the strand tensile test, a bundle of fibers is briefly exposed to a given temperature and broken. The strand tensile strength of R-glass, a typical HS-glass fiber shown in Figure 5, is higher than that of E-glass

Chapter 6

139

from room temperature to 750·e (and beyond). E-glass fails at -600·e. The modulus of this fiber is also higher than that of E-glass throughout the entire temperature range tested, from room temperature to 750·e (and above).

50

40

-

ttl

a.

o

£ '5>

30

-. ~ r-. -,

v-R-glass

r-. G r-, r-,

c: ~

in .9! '0; 20 c:

~

E-glass--"

~

10

1\ ,1\ ,

\ \

o

100

200

300

400

500

,

600

700

Temperature Co

Figure 5. Strength and modulus of R-glass, a typical HS-glass fiber, atelevated temperatures, Drawn from data contained in French patent 1,435,0739, issued in 1963 to Saint Gobain Company, Chambery, France,

In another tensiletest, R-glass is exposed for prolonged periods of time to a specific elevated temperature before its strength is tested. At 750· itstensile strength drops within the first 400 hours to 1 GPa (or1/4th its room temperature strength), but then remains nearly constant for at least the next 600 hours of exposure. A temperature of 750·e is well beyond the reach of either ofthe two general purpose E-glass fibers. In summary, high strength fibers have superior high temperature resistance, superior impact damage resistance, high strength retention at elevated temperatures, and intermediate moduli. The formation of a nanocrystalline structure which facilitates the attainment of high service temperatures tends toreduce strength, but in most cases the strength loss is minimal.

(b) Properties and applications HS-glass fibers fill an important niche in the composite reinforcement market between E-glass and the lowest cost carbon fibers. They are used in a great variety of applications in the aerospace and aircraft industry, automotive industry, electrical and electronics industry, sporting goods industry, and in military markets. The higher cost of other inorganic high

140

Chapter 6

strength fibers can be reduced by hybridization, Le., by designing composites reinforced with mixed HS-glass/carbon, HS-glass/aramid, or HS-glass/boron fibers. In the aerospace and aircraft market HS-glass fibers are used because of their high specific properties and relatively low cost, e.g.; in satellite components, motor cases, nose cones, aircraft flooring, cargo liners, and radome skin sheets. Their use in helicopter rotor blades benefits from the high damage resistance (toughness) of HS-glass fibers. In automotive composites, HS-glass fibers are used because of their high specific strength in pressure vessels and their superior toughness in leaf springs. Their use in compressed natural gas cylinders isone ofthe fastest growing applications. In the electrical and electronics market, HS-glass fibers are used as strength members for optical fiber cables, printed circuit boards, and other cables, but mostly when high temperatures are involved. In the sporting goods market, they are used for product differentiation and because of their toughness in tennis racquets, squash racquets, fishing rods, surfboards, windsurfing masts, skis, archery bows, arrow shafts, and racing yachts. In military markets, HS-glass fibers are used in aircraft fuel tanks because of their specific strength, and inrigid armor and helmets because oftheir superior ballistic impact resistance. 6.2.2 High modulus - high temperature fibers HM-glass fibers have modulus ranging from 100 to 132 GPa, and otherwise offer about the same properties as HS-glass fibers, such as high strength and superior resistance to high temperature. Some experimental HM-glass or glass ceramic fibers are known to have not only higher stiffness but also higher strength than typical HS-glass fibers. Their superior stiffness and strength, however, is often accompanied by a significant increase in their density, afact that requires special consideration oftheir specific strength and stiffness. In addition to alumina (see Chapter 4), beryllia [2) [16], ZnO [17], titania [4), lanthana [16), yttria [101 and CuO [10] (16) are known as modulus modifiers for silicate glass compositions (16). Lanthana, for example, can raise the fiber modulus of a typical HS-glass composition from 84-95 to100 GPa (Table V) , but because ofitshigher density, the specific modulus (37.7 Mm) remained unchanged. Zinc oxide, another effective modulus modifier, can raise the modulus to 104 GPa, and the specific modulus to 41.1 Mm. BeO yielded the first glass fiber (YM-31) that became officially known as a generic HM-fiber. It had a complex BeO-modified composition [2], but a simpler, statistically designed BeO-modified HM-glass fiber (Table V) had [15-16] a modulus of 112 GPa and a specific modulus of43 Mm. The effectiveness of BeO as a modulus builder is the result of the high field strength of the Be-2 ion and itsability tocoordinate four oxygen ions tightly toit [17]. The structure of BeO is ofthe wurtzite type. The only other wurtzite structure with oxygen is ZnO. In addition, ZnSi04 is isomorphous with Be2Si04• On a molar basis, the ZnO-modified and the original BeOmodified YM-31A compositions are identical [17], and on this basis ZnO is a less effective modulus builder than BeO (Table V) . The modulus of Y203-modified fibers (Table V) increases to 132 G [10]. At the same time, however, their density increases to 4.0 g/cc, their specific modulus drops to 30 Mm, their crystallization potential increases and their strength drops to 3.5 GPa [10]. Highly Y203modified HM-fibers possess a clearly observable nanocrystalline phase [10]. Upon heat treatment, large microcrystals «2 IJm) grow and significantly reduce the strength of these

141

Chapter 6

fibers. The value-in-use or cost considered performance of highly Y203-modified HM glass fibers is therefore lower than that oftypical HS-glass fibers (Table IV). Table V. High Modulus silicate glass fibers Modulus Modifier Composition, wt. % SiO, A~O,

CaO MgO La,O,

Zno

La,0,

ZnO

BeO

Y,O,

Y,O,

BeO/Y,o,

50.0 32.5

45.8

50.0 35.0

52.6 35.8

41.7 27.2

36.2 20.5

7.5

5.4

4.3

8.1

6.2

26.8

5.0 30.2

4.5 100 3.0 1.36 33.0

3.5 130 4.3 0.81 30.0

5.4 132 3.29 1.64 40.0

12.5 5.0

22.4

BeO

Y,O,

z-o, Li,o c-o,

Fe,O,

Structure Morphology References:

7.5

1.8 6.8 1.7 2.5 0.3

TiO,

Pristine Properties Strength, GPa Modulus, GPa Density, glee Sp. Strength, MIn Sp. Modulus, MIn

11.0 7.7

4.8 100 2.69 1.78 37.2

104 2.77 37.5

5.1 112 2.60 1.96 43.0

amorph. glass

amorph. glass

amorph. glass

(16) (18]

(17)

(15)

amorph. microcryst. glass glass cer. [10]

(10)

amorph. Glass [15] [18]

Y20:JBeO-modified HM-glass fibers (Table V) offers a better balance of mechanical properties than either Y203- or BeO-modified fibers. Their specific strength (1 .64 Mm) approaches that of the Y203-modified fiber (1.94 Mm) and their specific modulus (40 Mm) approaches that of the BeO-modified fiber (43 Mm). These fibers have a 30% higher specific stiffness than Sglass and would facilitate a 15% weight reduction of aerospace composites assuming the use of a 50% fiber volume fraction . Y20:JBeO-modified fibers are commercially unattractive because ofthe suspected toxicityof BeO.

6.2.3 Ultrahigh modulus glass-ceramic fibers By modifying the ternary Si02-Ab03-MgO or the quaternary Si02-Ab03-MgO-CaO eutectic compositions with appropriate oxide modifiers one can raise the fiber strength to 5.4 GPa (vs. 3.4 GPa for E-glass) and the fiber modulus to 132 GPa (vs. 72 GPa for E-glass). The increase in modulus iscaused byan increase ininternal order as evidenced bya change from an amorphous to a nanocrystalline structure. An increase in fiber modulus to 248 GPa can be achieved by inserting nitrogen into the oxide network [19-231, thereby creating a nitridemodified silicate, Si-AI-O-N, oroxynitride glass fibers. This increase is caused byan increase in surface tension as evidenced bymicrohardness. Thus, one mechanism seems to depend on increasing structural order, the other on crosslinking.

142

Chapter 6

(aJ Process and products Oxynitride fibers are formed by an adaptation of the conventional bushing process. Since nitrides would corrode precious metals in an oxidative environment, the bushings with up to 200 tips are therefore made from boron nitride-coated carbon or from molybdenum. The melts are formed under nitrogen at 1600-1750 ·C and refined at lower temperature. Fibers with diameters ranging from 12to 20 urn are continuously drawn from the melt -200·Cbelow the melt temperature, and mechanically wound at 1000-2000 m/min [24].

Silicon formation :

Si0 2 +Si1N"

~2Si+2SiO+2

N2

(1 ) (2)

Oxygen formation :

Si1N" · Si0 2 +12

Mo~4

Mo1Si+2 N 2 +0 2

(3) (4)

Silicon oxidation:

Si+0 2 ~Si02

(5)

The melt process may cause the reduction of silica yielding particulate silicon. Increasing numbers of silicon defects, when formed, impart a blue-gray (hazy) to dark-brown (opaque) appearance to the glass (Equations 1 and 2), and these defects may proportionately reduce the strength of the resulting fibers [23]. The formation of silicon defects can however be reversed byusing molybdenum (but not boron nitride) asthe bushing material (Equations 3 to 5), and by inserting an 8 hour refining cycle midway between the melt and fiber forming temperatures (as shown for the second entry inTable VI). A refined Si-AI-O-N glass is colorless and clear [23] and the resulting defect-free fibers [20] [24] are much stronger (>4.0 GPa) than those obtained from unrefined melts (2.0-3.0 GPa). Si-AI-O-N compositions with <15% nitrogen yield glass fibers with moduli ranging from 100 to 140 GPa, and microhardnesses ranging from 660 to 700 kg·mm·2 [19] as shown in Figure 6. Compositions with >15% nitrogen yield glass ceramic fibers with moduli ranging from 140 to 248 GPa, and microhardnesses ranging from 900 to 1200 kg'mm'2 [19]. The substitution of nitride for25% of a given oxide composition under the same conditions of synthesis no longer yields glasses. The respective melts will foam and crystallize [25]. Selected glass and/or glass ceramic Si-AI-O-N fibers are shown inTable VI. The first [23] is Sglass, and the second is a Mg-Si-AI-O-N composition that is similar tothe S-glass composition

143

Chapter 6

but modified to contain 7.5 wt.% SbN. [23). Since the latter had been properly melt refined before it was fiberized [24), it yielded high strength fibers. lis low nitrogen content produced a significant modulus increase beyond that of the S-glass control. The third example in Table VI, a Ca-Si-AI-O-N, has a still higher nitrogen content [19) and therefore a higher absolute and specificmodulus. lis low strength suggests that the melt was not or could not be adequately refined .

.-

26.0

a. '"

(!)

vi ::3 :; "0 0

E 0

~ til

24.0

I-

22.0

l-

20.0

-

18.0

l-

16.0

I-

14.0

I-

12.0

I-

10.0

I-

8.0

l-

.,

I

iii

6.0

,~

,...

_V'

.,..",,'

.....

,f.

• / A

i

I

I

20

1150

-

1100

-

1050

-

1000

'"vi

-

950

l:

-

900

-

10

1200

')I

E

~

l/)

~

700

-

600

-

500

-

400

Ql

"E

'"

s:

eQl '"

5

30

Nitrogencontent. weight % Figure 6. Modulus and microhardness of Sialon or oxynitride fibers versus nitrogen content. Redrawn from J. Kobayashi, M. Oota, K. Kada and H. Minakuchi, Oxynitride Glass and the Fiber Thereof, US Patent 4,957,883, September 18, 1990.

The remaining examples in Table VI are ultrahigh modulus (UHM) glass ceramic fibers. The commercial Ca-Mg-Si-AI-O-N development fiber [20) reflects proper melt refining and has a moderately high nitrogen content. Accordingly, it has a measured strength of 4.0 GPa and a measured modulus of 180 GPa, as well as a specific strength of 1.38 Mm and a specific modulus of 62.5 Mm [20]. The second example, a Y-Si-AI-O-N fiber, adds the predictable effect of yttria which increases measured modulus and density, but minimizes the specific modulus. The third example, a Ca-Mg-Si-AI-O-N fiber has the highest measured modulus (248.0GPa) and the highest specific modulus (75 Mm) ofany known oxide glass fiber.

144

Chapler6

In summary, nitrides offer only a modest crystallization potential but instead seem to act as modulus builders by crosslinking the oxide structure they modify. And the presence of nitrogen in the network structure, once introduced, may restrict the dimensions of crystals which can be formed . A selected oxynitride melt having a high nitrogen content may yield nanocrystalline fibers with an ultrahigh modulus, while a nitrogen-free oxide melt with the same cation ratio may already yield microcrystalline fibers with lower moduli. Thus, the modulus of oxynitride fibers can be increased to much higher levels than that of fibers from oxide melts. Crystallinity ofoxynitride glasses can be correlated with their nitrogen content by infrared methods [25]. HM-oxynitride glass fibers are x-ray amorphous. UHM-glass ceramic fibers [20] are nanocrystalline (as shown in Chapter 4). Table VI. Ultra-high modulus Si-Al-0-N glass fibers Fiber Modifiers Composition, wt. % SiO, Alp, CaO MgO Y,O, Si.,N. A~N,

N-content

Melting temp., °C Refining temp., °C Fiber forming T, °C Bushing tips Pristine properties Strength, GPa Modulus, GPa

Density, glee Sp . Strength, Mm Sp . Modulus, Mm

Structure Morphology References:

5-glass Control

Commercial Si-AI-0-N glass fibers yCaMgCa-Mg-

65.50 25.00

56.75 25.33

9.50

ID.42

0.00

7.51

32.38 50.78 16.84

Unspecified commercial 48SiO,' 43CaO· 5MgO· 4Al,o, based fiber 12.00

50.3 7.6

Si-Al-0-N Ca-Mg 11.1 2.2 62.6 0.7

25.1 23.4 17.0 4.37

13.5

0.00

2.54

10.00

1650

1750

1600

1790

1565 200

1720 1600 1500 200

1380

1560 1

1590

4.60 88-95 2.53 1.85 35-37

4.50 115.00 2.80 1.61 41.10

213.00 3.94

248.0 (3.3)

54.00

(75)

amorph. glass [23]

amorph. glass [23] [24]

137.0 2.8 48.9 amorph. glass (19]

4.00 180.00 2.89 1.38 62.30 nanocr. glass cer . [20] [23]

nanocr. glass cer. [22]

1

nanocr. glass cer. (19]

(b) Properties and applications Glass and glass ceramic oxynitride fibers can be produced with high strength by properly refining melts while they are formed, and with ultrahigh moduli by inserting nitrogen into a suitable oxide network structure. The highest measured modulus (248 GPa) that has been reported lies between those of standard modulus (8M) and intermediate modulus (1M) carbon fibers (230 and 303 GPa, respectively), and the highest specific modulus (75 Mm) lies midway between those of 8M carbon fibers (131 Mm) and E-glass fibers (27 Mm). Two types of applications are being pursued with oxynitride fibers.

Chapter 6

145

Sialons are reinforcing fibers for metal matrix composites. An aluminum alloy 6601 matrix reinforced with a development fiber (Table VI) had a strength of 4.0 GPa and a modulus of 180 GPa. The bending strength of the MMC was 25% higher than that of an alumina fiber reinforced control, and almost the same as that of a silicon carbide fiber reinforced control. A high value-in-use may result if this fiber were to cost less than the incumbents did. Oxynitride fibers have high alkali durabilityand may be useful as a diaphragm material in the electrolytic production of chlorine from aqueous NaOH [26] and as a reinforcement for cementicious composites [26]. 6.2.4 Fibers with high chemical stability This discussion of chemical stability refers to bare (uncoated) glass fibers, i.e., fibers having neither a specific acid or alkali resistant finish nor a secondary coating. Accordingly, the relationship between fiber composition and chemical stability in water, acids, and bases is complex. It depends on the interaction between (1) the chemical agent [27] to which the glass fiber surface is exposed, (2) the pH of the glass composition [33] in the fiber surface, and (3) the internal microstructure ofthe fiber [27].

(a) Chemical resistance ofglass fibers The first step in the attack of water on the bare surface of an alkali-free or near-alkali-free glass fiber is its adsorption. The adsorbed water molecules hydrolyze the siloxane bond by protonation ofthe oxygen atom, and yield a highly hydroxylated fiber surface. = Si - 0 - Si = ~ = Si - 0 ·· ·· · -Si = ~ =S-OH +HO-S

=H+ +OH -

(6)

With high alkali-glass fiber surfaces, the reaction of water represents an electrophilic attack by the addition of the proton (W) to the negatively charged oxygen atom of the =Si-O-M bond. It proceeds in the same fashion and results, practically speaking, in the ion exchange between W (or H30+) and either alkali ions and/or network modifying alkaline earth ions, and leads to the formation of SiOH groups [27]. The reaction products of water with a highly alkaline fiber surface are NaOH, Ca(OHh and hydrated sodium silicate. = S-OM +H + ----+=S-OH + M+

(7)

However, as soon as the supply of H+ions is exhausted, the corrosive attack ofH20 turns into a nucleophilic attack by alkali ions on the fiber surface. In other words, the reaction of a glass fiber with water turns into a reaction of a glass fiber with alkali. Siloxane bonds are broken and Si-O-Na groups are formed until the glass fiber is completely dissolved in the highly alkaline medium that initially consisted only ofwater. The reaction of alkaline media starts with a nucleophilic attack by hydroxyl ions on silicon atoms in the bare surface of a glass fiber (-OH' + =Si-O-S=) where it forms new bonds (=SiOH, and/or =Si-OM). Monovalent cations (e.g., sodium) are removed from the glass fiber, leaving behind a hydroxylated surface. Bivalent cations (e.g. calcium) remain attached to the glass surface and form a crystalline sheath growing in thickness. Such a sheath develops when bare silicate glass fibers (>50% Si02) including E-glass [27], AR-glass [30], basalt [29] and oxynitride fibers [26] or when bare aluminate glass fibers (>50% Ab03), e.g. calcium aluminate fibers [18] as discussed inChapter 4, are exposed toalkaline media.

146

Chapter 6

The crystalline sheath, mostly consisting of Ca(OH)2[26), increases the alkali resistance, but also limits the practical utility of the resulting fibers since it drastically reduces their strength. Zr02, SnO, La203, Ti02, Fe203 [10), Y203[26) and Na20 enhance the alkali resistance perhaps by delaying sheath formation, but rather large amounts of Na20 (>10%), ZrO (>15%) or Y203 (>30%) and combinations of Na20 (11%) and Zr02 (16%) are often employed. These fibers are not really alkali resistant, only more alkali resistant than E-glass. In the end, all lose their physical integrity and are destroyed as evidenced from the accelerated leach test shown for E-glass inTable VII. Table VII. Composition of borosilicate E-glass and its solutes after leaching (8 hours/95°C) Composition E-glass (%)

SiO, 53.8

AI,O, 14.9

Fe,O, 0.3

CaO 17.1

MgO 4.7

B,O, 8.7

Solute (%), in H,O in2nH,SO. in2nNaOH

0.2 1.7 25.6

0.2 14.6 7.3

0.3 0.2

0.2 16.6 8.4

0.1 4.6 1.3

0.7 8.6 6.3

While alkaline media are known to create a crystalline deposit or sheath on the surface of glass fibers, mineral acids selectively dissolve specific components of the glass, first of all the ions of network modifiers [10]. Silanol bonds, as a rule, are not broken, and Si~ is not dissolved. If however the amount of Si02is not sufficient to create a continuous network structure, cations can selectively dissolve inacid media. The addition ofZr, Ti, and Fe oxides, even to the high alkali oxide glasses shown in Table VII, substantially increases the acid resistance [10J, but in the end, the entire fiber, whether it is compositionally an E-glass or Aglass, isconverted into a porous high silica fiber (see Table VII and Chapter 6.4.4). In summary, the effect on the pH of the bare fiber surface and the effect of the interaction between a chemical agent and a bare fiber surface are predictable. Zr02 seems to increase both acid and base resistance. The effect of the internal microstructure [27J ofa fiber ishighly process dependent and not predictable without a thorough prior investigation of its microstructure. Importantly however, all fibers, except experimental single fibers, have a primary finish; some have an additional secondary coating. These modifications further reduce the predictability of their chemical resistance from their compositional make-up alone.

(b) Alkali resistant glass fibers CemFil, a commercial AR glass fiber [1-2) [30), an experimental AR glass fiber [28J, and an AR glass-ceramic basalt fiber [29J, commercial only in Russia, are shown in Table VIII. CemFil and the experimental AR fiber are highly Na20- and Zr02-modified glass fibers. Basalt fibers are derived from volcanic rock with high Fe 203+FeO levels. All have higher alkali resistance than E-glass. That of basalt fibers lies between E- and AR-glass. All require higher fiber forming temperatures than E-glass. Basalt fibers require an energy intensive, nonstandard process. Experimental AR fibers offer a 100°Clower fiber forming temperature than that ofCemFil and basalt fibers. In a test that simulates their suitability as a cement reinforcement [30J, bare AR-glass and bare E-glass fibers were immersed at25, 50 and 80°Cina solution (NaOH 0.88 gIl, KOH 3.45 gIl, Ca(OH)l 0.48 gIl, pH 12.5) that simulates the aqueous phase of Portland cement. Both

Chapler6

147

fibers lost strength between 24 and 96 hours, i.e., in the time frame during which the alkalinity of cement reaches its peak as it cures. At 25°C, the strength of bare E-glass dropped to 2/3 of its original value in 24 hours, and that of AR-glass in 96 hours. In the accelerated test at BO°C, the strength of E-glass dropped to 1/3 of its original value in 24 hours and that of ARglass in 96 hours (Figure 7). Thus, neither fiber is usable to structurally reinforce cement without having an effective secondary alkali resistant coating. The application of an alkali resistant finish or secondary coating is known to render even Eglass suitable for continued use as a durable reinforcement of cement structures, whether they are composite wraps for bridge columns, or net-like structures aimed at roadbed construction. But since even AR-glass loses strength in <96 hours in a solution simulating an aqueous cement phase, it too would require a costly secondary alkali resistant coating to be acceptable in continuous structural use. In the Western world and Japan, AR-glass such as CemFIL costs 2x as much as E-glass. Thus, AR-coated E-glass rather than AR-coated CemFIL isthe preferred reinforcement ofstructural, load bearing cementicious composites. In CIS countries, e.g., Russia, AR-coated basalt fibers are apparently as economically viable as AR-coated E-glass as a replacement for asbestos in the fiber reinforcement ofcement pipes. Table VIII. Glass Fibers with High Chemical Stability Fiber Type Composition, wt. % SiO, A~03

B,O, CaO MgO BaO TiO, ZnO

z-o, L~O

Nap 1(,0

Alkali resistantllass fibers CemFil Exp. A -G Basalt 71.0 1.0

68.11 0.78

49.06 15.70

71.8 1.0

4,86 3.04 2.43

8.95 6.17

8.8 3.8

16.0 1.0 11.0

65.15 6.70 9.23 4.33

0.20

6.92 13.85

FeD

Structure Morphology References:

65.0 4.0 5 .0 14 .0 3.0

1.36

Fe,O, MnO P,O, H,O F, Forming temp ., °C Liquidus temp ., °C ~T (F-L), °C

Acid resistant glass fibers A-glass C-glass CC-glass

tr.

3.11 1.52 6.37 5 .38 0.31 0.45 1.62

1300 1200 100

1212 1074 138

amorph. glass

amorph. glass

microcr . gl. cer.

[2] [28]

[28]

[29]

1300 1220 100

13.6 0.6

8.5

12 .92

0.4

0.3

0.27

tr.

0.96

1280 10lD 270

1200 1135 65

1192 1068 124

amorph. glass [I]

amorph. glass

amorph. glass

[1]

[32]

The cost considered value of AR-glass, e.g., CemFil, without a secondary AR-coating lies elsewhere. It is used as a reinforcement of non-structural and decorative cement composites

Chapter 6

148

ranging from non-structural ready-mix cement and decorative architectural components to non-load bearing, building cladding systems such as balustrades, string courses and cornice elements, corbels and arch units, mullions and window surrounds, and consoles and copings [31]. In these uses, AR-glass fibers offer a simple solution to the problem of shrinkage cracking that often occurs during the initial curing phase «96 hours) of cement or concrete mixes, without requiring a costly secondary AR-coating. They retain useful strength long enough tosurvive the alkaline curing cycle and therefore prevent shrinkage cracking. E-glass without a costly AR-coating loses its strength and totally disintegrates before the initial, alkaline cement curing cycle iscomplete.

3.0

co

a,

(!)

~

C, c:

2.0

l!?

1ii

.!!! 'wc:

~

1.0

o

24

48

72

96

Time, hours

Figure 7. Tensile strength ofglass fibers inthe aqueous phase ofcement. Redrawn from A. J. Majumdar, Alkaliresistant glass fibres, in Strong Fibers, W. Watt and B. V. Perov, ed~ors, pages 61-85, Elsevier Science Publishers, North-Holland (1985).

(c) Acid resistant glass fibers Very high acid resistance in glass fibers is synonymous with the behavior of glass fibers having a high silica content including pure silica fibers, ora high alkali content. Four groups of glass fibers yield higher acid resistance than borosilicate E-glass, the current trade standard for fiberglass performance. The first group consists ofgeneric boron-free E-glass fibers (Table I). The second consists ofhigh strength (HS) glass fibers (Table V), high modulus (HM) glass fibers (Table VI) and pure silica glass orquartz fibers (Chapter 6.4). The third group consists of high silica fibers obtained from E-glass by leaching in mineral acids (Chapter 6.4). The fourth consists ofhigh alkali silicate fibers.

Chapler6

149

The fourth group ofacid resistant fibers (Table VIII) includes A-(or high alkali) glass fibers, C(or corrosion resistant) glass fibers and CC- (or Chinese Co) glass fibers. All are high alkali glass fibers, but A-glass is boron and fluorine free, C-glass is fluorine free but contains boron, and CC-glass [32) is boron free but contains fluorine. Designations such as A-glass, C-glass, and CC-glass (Table VIII) represent generic industry specifications. All have a much higher alkali content and offer much higher acid resistance than borosilicate E-glass. The forming and liquidus temperatures of A- and C-glass are higher than those of E-glass. C-glass has a more desirable (Le., lower) forming temperature than A-glass and a less desirable (i.e., lower) differential (~T) between its forming and liquidus temperatures. CC-glass has the lowest liquidus and forming temperatures of this group of fibers. In fact, its liquidus and forming temperatures are comparable tothose of borosilicate E-glass. Low alkali, boron-free fibers come tomind when an end use requires high acid resistance ata moderate cost. Alkali-free, high strength, high modulus and/or high silica glass fibers are premium fibers used for their high temperatures properties and seldom, if ever, for their acid resistance. High alkali glass fibers (Table VIII) offer high acid resistance at low cost. They also have lower strength, and lower alkali and water resistance than E-glass, whether they contain boron and/or fluorine ornot. A recent technical evaluation of an experimental A-glass fiber highlights the potential value of A-glass in automotive composites [34) especially when made from waste glass. At present, however, continuous high alkali glass fibers are commercially available only in China as CC-glass [32). 6.2.5 Other special purpose glass fibers The need in the market for glass fibers based on costly specialty glass formulations is low. Nonetheless, glass fibers have been made with a wide ranges of properties including glass fibers having low and high dielectric constants, high and low densities, high radiation resistance, and conducting, semiconducting and superconducting properties. (a) Fibers with lowdielectric constants

The electrical properties of glass fibers are characterized by volume resistivity, surface conductivity, dielectric loss, and dielectric constant. Glass fibers, designated as D- (or dielectric) glass, have a lower dielectric constant than borosilicate E-glass and were developed for the computer industry for use as a reinforcement for high density composite printed wiring (or circuit) boards having a higher strength and faster response time than borosilicate E-glass reinforced wire boards. The most effective compositions are shown in Table IX. Most glass fibers having low dielectric constants have low densities. The value of these fibers resides intheir low dielectric properties. D-glass has been specifically designed [2) for wire board applications (Table IX). Pure silica fibers (Chapter 6.4) happen to have a low dielectric constant, and are suitable although very costly for these applications. Hollow E-glass fibers are also known to have low dielectric constants [52), but for use in printed circuit boards, a method of sealing cut ends in drilled holes must be used to prevent water from entering the fiber cores. The presence of water would reduce the dielectric constant. A recent composition isalso ofmerit [35). S-glass [2) is another candidate fiber for specific circuit board applications. Solid E-glass fibers are the reference material in Table IX. Large mainframe computers are probably the predominant target application in terms of product requirements and product cost. Industrial applications,

150

Chapter 6

e.g., instrumentation, workstations, and general purpose minicomputers, may be more price sensitive. Table IX. Glass fibers with low d ielectric constant Fiber type Fibe r core

E-glass hollow

Com po sition, wt. % SiO, Alp,

s,o, CaO

MgO

TiO,

Li,0 Nap 1<,0 Fe,O, F,

D-glas s solid

Silica solid

Exper. soli d

5-glass solid

E-glass solid

54.0 14.0 10.0 17.5 4.5

74.5 0.3 22.0 0.5

99.9999

55.7 13.7 26.5 2.8 1.0

65.5 25.0

54.4 14.0 6.6 22.1 0.6 0.5

1.0

1.0 1.3

tr. 0.5

Fiber properties Die!. Constant " Density, gl ee Reference. ·21°C/l0 Hz

2.98 1.80 [52]

9.5

0.1 0.1 0.1

3.56 2.16 [2)

3.78 2.15 [10)

4.10 [35)

0.8 0.2 0.2 0.6 4.53 2.48 [2]

6.86 2.54 Table I

(b) Fibers with high densities and high dielectric constants Glass fibers with high dielectric constants tend to have high densities also, and their value depends on one or the other property, rarely on both. Three generic applications are known for these fibers. The oldest application, which requires radiation resistance, e.g., absorption of gamma rays, relies on their high density [10] [36]. Applications which require electrical insulation, e.g., in capacitors, rely on their electrical properties [40]. The newest application, which requires concentration of electromagnetic energy, e.g., in circuit boards for high frequency uses, also relies on their electrical properties [37-38]. Glass fibers with high densities (and high dielectric constants) are known as protective glass fibers [52] since they can absorb gamma rays as well as fast and slow neutrons. For example, lead, bismuth andlor barium containing glass fibers have high densities (4.0 - 4.8 glee) and high dielectric constants (8-13) as shown in Table X. They can be used to absorb gamma rays [10] in x-ray equipment and radiation protective composites. The ability of lead (or L-) glass fibers to absorb gamma rays increases with increasing density [10], but these fibers have low strengths «2 .0 GPa), low moduli «55 GPa), and very low resistance to light, humidity andlor elevated temperatures. Table X. Glass fibers with high densities and/ or high d ielectric constants

Strength , GPa Modulus, Gpa Density, glee Dielectric constan t References:

Generic borosilicate E-glass fibers 3.40 72.00

2.54 6.80 [1,2]

Alwninosilieate fibers

Lead silicate glass fibers

« 30% CeOJ

(<30% PbO)

N iobium silicate fibers «15% NiP,)

4.5 95.0 3.G-3.8 7.0-8.0 [36]

1.7 51.0 3.0-4.0 8.0 -10 [10] (37)

10 - 15 [37] [38]

Chapler6

151

Glass fibers containing boron, cadmium andlor cerium can be used for protection against neutrons [52]. Their densities (3.0-3.5 g/cc) are not as high as those needed for gamma ray protection (4.0-4.8 g/cc). However, glass fibers in the Si02-Ab03-Ce02system [16] [36] with 10-28% Ce02 offer very high strengths (4.5 GPa) and very high moduli (95.0 GPa). In fact they could be considered to be high strength fibers (see Table IV), but unlike the thermal stability of the other high strength fibers (Chapter 6.2.1), theirs is very low. For example, a major strength loss occurs after heat treatment at 100-200'C due to pre-crystallization and microseparation ofa second solidphase, which is revealed in x-ray diffraction patterns [52]. (c) Fibers with very high dielectric constants

Glass fibers with high dielectric constants [37-38] differ fundamentally in their applications from those requiring glass fibers with low dielectric constants [36]. They are aimed at circuit boards being designed for use in high frequency applications. Again, using the example of Eglass as a control, Table XI shows three different approaches to glass fibers with very high dielectric constants, i.e., potentially suitable for the presently emerging market need [37-38]. Lead (or L) glass fibers with extremely high levels of PbO (>70%) as well as silicate glass fiberscontaining very high combined levels of BaO and Ti02offer significantly higher dielectric constants, i.e. 13.0and 13.5, respectively. Table XI. Glass fibers wit h very high dielectric constants Glass fibers Composition, wt .% SiO, A~O,

B,O, PbO CaO MgO

srO

BaO TiO,

z-o,

Na,0 K,O Nb,O,

DieJ. constant, £r, @ 1 MHz Forming tem p ., TIog2.5, ·C Liquidus temp., T,, ·C ~T (F-L), ·C Fiber forming abilit y Referen ces:

E-glass cont rol

L-glass w/>70% PbO

BaD-TiO, silicate

54.3 14.0 6.6

26.0

40.0

55.0 2.5

47.16

7.5

9.0

7.03

7.5 15.0 23.0 7.0

6.0 15.0 7.8 1.7

7.03 14.05 13.95 3.27

3.0

7.51

1.5 72.0

22.1 0.6 0.5

Nb,0,-BaO-TiO, silicates

0.8 0.2

0.5

6.8

13.0

13.5

10.1

12.3

1299

850

1077

1199

1136

1063 +236 excellent Table I

650 +100 very poor [37J [38J

1214 -137 infeasible [37J [38J

1085 +114 very good [37J [38J

1089 +51 good [37J [38J

Lead (L) glass compositions possess a dielectric constant of 13.0, but glass fibers from lead glass compositions with such high PbO levels are difficult to form. They are very long melts which have a very high !1T (Tlog25 - Ti) between forming and liquidus temperatures and are therefore quite crystallization resistant. However PbO will evaporate violently during the melt process, thus affording highly non-uniform fiber compositions and frequent process discontinuities and fiber breaks [37-38). In addition, the use of PbO poses significant environmental concerns.

152

Chapter 6

The silicate glasses based on high levels of BaO and Ti02 have an equally high dielectric constant of 13.5, but fibers cannot be formed from their melts because the liquidus temperature is much higher than the preferred forming temperature. Crystallization would occur long before the melt reaches the preferred fiber forming viscosity of log 2.5 poise. In summary, glass fibers with very high levels ofPbO are sublimation prone and glass fibers with high combined levels BaO and Ti02are crystallization prone. In contrast, addition of 0.5to 15.0 mol% of Nb205toglass compositions having high levels of BaO and Ti02 slightly reduces the dielectric constant, and dramatically reduces the liquidus temperature, thereby reversing the ~T (TI09 25 - TL) from -137"C to +117"C (or 242°F). The ~T quoted here [37-38] describes the difference between Tl09 2 5 and TL, while the ~T customarily quoted in the technical literature refers to the difference between T109 3 and TL. This difference does however not affect the conclusions. High speed and high frequency information transmission is becoming increasingly important with the recent development of advanced information systems including mobile communication by car telephones and personal radios, as well as satellite broadcasting and cable television. As a result, there is an increasing demand for miniaturizing electronic devices and also microwave circuit elements such as dielectric resonators in conjunction with the electronic devices. Nb205 containing glass fibers are embedded for these applications in a resin such as polyethylene oxide that has a low dielectric tangent (tan 0) loss to obtain the desired high frequency performance of the resulting circuit boards for microwave applications [37-38]. In summary, microwave circuit elements can be made more compact when using a circuit board having a high dielectric constant. It acts toconcentrate the electromagnetic energy within the board and thereby minimizes the leakage ofelectromagnetic waves.

(d) Fibers with super- and semiconducting properties Superconducting glass fibers are obtained by incorporating a suitable ceramic material in the fiber core, yielding superconducting bicomponent sheath/core glass fibers (Chapter 6.3.3). Superconducting bicomponent metal ribbons are obtained by incorporating a suitable ceramic material inthe core ofcontinuous metal tubes. Superconducting single component fibers can be made drawing single superconducting ceramic preform rods bythe laser heated float zone orpedestal growth process (Chapter 4.4.2). Semiconducting glass fibers have been known for over 30 years [52] but have never attracted commercial interest because significant technical problems have never been solved. In principle, such fibers (e.g., CuO-CaO-Ab03-Si02) can be made by adding oxides of monovalent metals such as copper or silver to a suitable base glass and by subsequently reducing the glass fibers in various gaseous media, e.g., hydrogen. However these fibers are moisture sensitive, and prolonged storage leads to increased glass conductance (10). The electrical properties of glass fibers are best modified by (1) applying a permanent chemical coating to the fiber surface, (2) adding a suitable material to the binder or finish formulations, or (3) modifying the composite matrix that is being reinforced with a glass fiber. A semiconducting or conductive coating is applied by vacuum deposition, metallization from metal salts, decomposition of organometallic compounds or chemical metallization. Also, carbon black can be added tothe composite matrix.

Chapter 6

153

(e) Fibers with bone bioactive glass compositions Bioactive glasses have been developed since 1969 [67] and continuous bioactive glass fibers since 1983 [68-69]. Very recently (70), bone bioactive glass fibers were found tobond tobone tissue and help bone tissue growth when they are preferentially placed on the surface of a thermoplastic composite while carbon fibers are used tostiffen the core ofthe structure. A glass composition which was found particularly suitable for these applications (67) contained 52% Si02, 30% Na20, 15% CaO and <3% P20S(in mole %). Fiber bundles or tow of up to 5000 filaments were drawn from a melt of this composition and were interwoven with carbon fiber tow into a cylindrically braided sheath/core textile preform. The carbon fibers formed the core of the braided structure and provided the required stiffness and load support, and the bone bioactive fibers formed the sheath and functionality. To create a practical reinforcing structure of this nature, two braids are in effect braided simultaneously, one forming the carbon fiber core, the other the bioactive fiber surface layer or sheath, and both are suitably interwoven, overlaid or otherwise intermingled. The carbon fibers in the core are first co-mingled with a suitable polymer such as a polysulfone, and coarse fibers of the same polymer are intermingled with the bone bioactive glass fibers. The hybrid preform is then processed ina closed die in a hot press. The combined amount of polymer iscalculated to give the final total volume fraction ; no additional polymer nor injection molding is required. In summary, superior composite materials can now be designed and manufactured for use in practical prosthetic devices, facilitating bone tissue growth, that have a structural stiffness matching that ofhuman bone. Table XII. Properties of ultrapure silica glass fibers Astroguartz and Quartzel [l 0I Mechanical properties Filament diameter, yam and fabric, um Specific gravity (density) , g/cc Hardness (Mohs scale) Pristine filament strength, GPa, RT Yam strand tensile strength, GPa, RT Yam tensile modulus, GPa, RT Physical properties Liquidus temperature, °C Temperature at max . crystallization, °C Coeff. of thermal exp., 0-1000°C, °C Thermal stability (short term), to °C Thermal stabilit y (long term), to °C Thermal conductivity at 20°C, CGS Dielectric constant at 20°C, 1MHz Refractive index at 15°C, n.,

9

2.2 5-6 6.0 3.4 69 1670 1630 7

5.4x1O·

2000

1200 0.0033 3.78

1.4585

6.3 Non·round, bicomponent and hollow fibers A typical continuous glass fiber has a solid cylindrical shape, i.e., a round fiber cross section with a uniform diameter along its length. Special applications may require glass fibers having a non-round, e.g., a ribbon shaped or a trilobal cross section. Dual or bicomponent glass

154

Chapter 6

fibers, including hollow fibers, are composed of two different compositions. Hollow fibers are bicomponent fibers with a continuous void along the length surrounded by a tubular glass surface. 6.3.1 Silicate glass fibers with non-round cross sections A circular cross section has the smallest circumference for a given area. Any increase in circumference ofa fiber while retaining the cross sectional area will increase the fiber surface. Any increase in the fiber surface offers significant product advantages including higher adhesion and strength in composites, but also creates increased process complexity.

(a) Processes and structures The effect of the surface tension of a molten jet is the driving force toward the formation of fibers with round cross sections. Thus, a fiber with a round cross section is obtained from a bushing tip with a round cross section. With a bushing tip having a non-round cross section, the effect ofsurface tension tends toforce the melt into a round cross section. This tendency isopposed by the effect of the quench rate which would otherwise stabilize a non-round fiber cross section. Since glass melts have higher surface tensions than polymer organic melts, it is more difficult form glass fibers with non-round cross sections than polymer organic fibers with non-round cross sections (41). The basic fiber cross section technology was developed in the early 1950s for nylon yarns. Not surprisingly, the development of glass fibers with non-round cross sections was more demanding and the required technology is still not being practiced on a commercial scale. Two slow processes were developed in the late 1960s for forming continuous single glass ribbons [39-40]. Processes were claimed in the late 1980s for forming discontinuous, tapered, trilobal fibers (41). and continuous oval shaped and trilobal multifilament glass fibers (42).

Figure 8. Tipdesign forthe fabrication of non-round glass fibers at high process speeds. A bushing that has projected edges (1) extending downward from the distal end of a nozzle tip (2) can overcome the effect of the surface tension that tends to force a any hotjet into a round fiber cross section at commercial process speeds. Redrawn from H. Taguchi, K. Shioura and M. Sugeno, Nozzle tip for spinning glass fiber having deformed cross section and a plurality of projections, U. S.Patent 5,462,571, October 31 ,1995.

Chapter 6

155

A recent redesign of nozzle tips [43] is said to afford non-round multifilament E-glass at commercial production rates (Figure 8). For example, ribbon, oval, trilobal triangular or rectangular cross sections can be produced with 400 tip bushings and windup speeds of 1800-3000 m/min under otherwise conventional E-glass process conditions. This critical technological advance uses an entirely new bushing tip design. An edge, shield. or baffle protrudes from the surface of each tip at critical points to induce preferential cooling of the fibers and thereby toassist the formation ofwell-defined non-round cross sections. Non-round fiber cross sections are characterized by their modification or mod ratio. For ribbons, the mod ratio is their width-to-thickness ratio. For trilobal fibers it is the ratio of the diameter B of the outer circle around the lobes of the fiber cross section to the diameter A of the inner circle within the core of the fiber cross section (Figure 9). The term mod ratio has been used since the early development of trilobal nylon fibers. The newer term, "deformation ratio" [43], isless intuitive and therefore less desirable. (b) Products and applications Several potentially desirable properties increase with increasing mod ratio. Nylon fibers with a high mod ratio offer a high degree ofexternal light reflection and. therefore, the appearance of a silk-like sheen or luster in a yarn or fabric. Nylon ribbons with a high mod ratio offer the appearance of a metallic sparkle in a fabric or garment. They offer large reflective surfaces which constantly shift with the motion of the garment. The visual effect of a trilobal or ribbon shaped glass fiber does not seem to posses practical value, but the structural effect appears tobe useful inlimited applications.

Figure 9. Trilobal glass fibers with high modification ratio. Left: redrawn from L. J. Huey, Method and apparatus formaking tapered mineral and organic fibers. U. S. Patent 4.666,485. May 19. 1987. Right: Courtesy of Owens Coming, Circleville. Ohio.

The technology for forming non-round glass fibers was developed between 1967 [40] and 1987 [41-42], and afforded a premium application for lead glass ribbons in ribbon wound capacitors [40]. The process, however, is slow and was apparently not sufficiently cost effective to facilitate the growth of large volume applications in the composites market. The newer technology, which was developed since 1987 [43], offers higher process speeds and therefore a more cost effective route to ribbon shaped and trilobal glass fibers. The potential value of glass fibers with modified fiber cross sections has long been understood [39-40]. They offer a higher surface area as composite reinforcing fibers, higher matrix adhesion, and higher composite strength.

Chapter 6

156

6.3.2 Structural bicomponent silicate glass fibers Commercially significant inorganic bicomponent fibers have either a concentric sheath/core, or a side-by-side fiber structure. Polymer organic fibers with an additional excentric sheath/core structure are known, but inorganic fibers with this structure are not known. (a) Sheath/core and side-by-side bicomponent fibers Structural bicomponent ceramic fibers (Chapter 3) and optical bicomponent glass fibers (Chapters 4 and 7) have a concentric sheath/core structure (Figure 10). A sheath of one composition surrounds a core of another composition. Either the core or the sheath is responsible for the functionality. For example, the boron sheath of boron/tungsten fibers is responsible for the functionality while the core is sacrificial. The core (or wave guide) of optical fibers is responsible for the functionality while the silica sheath provides strength and load support. Side-by-side bicomponent structures afford non-straight, crimped orbulky fibers since the components experience differential shrinkage during processing. This chapter deals with structural silicate glass fibers consisting of two or even three components. Solid sheath/core (s/c) and side-by-side (s-b-s) bicomponent glass fibers have two different inorganic compositions. Hollow glass fibers are sheath/core bicomponent fibers having a continuous glass sheath that surrounds a continuous void or hollow core. Hollow porous glass fibers have a microporous instead of a solid glass sheath. Hollow superconducting glass fibers have three different compositions, an outer silicate sheath that surrounds a functional ceramic sheath, that inturn surrounds ahollow core.

Concentric

L

Excentric

Side-by-side

_

Figure 10. Schematic drawing ofcross-sections oftypical concentric and excentric sheath/core, and ofa side-byside. bicomponent fiber.

(b) Hollow sheathlcore silicate glass fibers Hollow fibers have a concentric sIc bicomponent structure with a solid sheath and a void- or air-filled core. Flame drawing from glass tubes yielded the earliest examples of hollow fibers. Commercial technology relies on introducing air into the core of a molten glass jet, the fiber precursor, and dates to 1966 [47-48]. The simplest design isa laboratory bushing [55]. Three

Chapter 6

157

processes are suitable for the production of hollow glass fibers. In a commercial process (Figure 11 a), the bushing tips aspirate air from the forming zone into the core of the forming cone [51J. In another process, air is injected through feeder tubes into the core of the melt in each tip [49-50]. Finally, a double crucible design (Figure 11b) is required to manufacture three component superconducting glass fibers whereby air is aspirated into the core of the forming cone from the inner crucible [46J. Hollow glass fibers have a lower weight at equal volume than solid glass fibers, greater thermal insulation, a lower dielectric constant, better long term fatigue characteristics, but also lower strength and modulus. Hollow E-glass fibers and hollow S-glass fibers are known. Hollow S-glass fibers have short melts and a much greater crystallization potential than hollow E-glass fibers which have long melts and a forgiving crystallization behavior. Because ofcost, both are specialty fibers, but since hollow E-glass fibers are much easier to fabricate, they have found a wider range ofapplications than hollow S-glass fibers [10J [52J.

Outer crucible Outer crucible Inner crucible

Glass melt

.-.a-n

+

Air ~~=~====~=~

Hollowfiber

Hollow fiber

Figure 11. Bushing tips forprodudng hollow glass fibers. Left: Redrawn from L. J. Huey, Method and apparatus forproducing hollow glass filaments,U.S. Patent 4,846,864, July 11, 1989. Right: redrawn fromJ. Huang, Hollow high temperature ceramic superconducting fibers, International Patent Application WO 97/22128, June 19, 1997.

Commercial hollow borosilicate E-glass fibers with an outside diameter (0. D.) of 13 IJm and an inside diameter (I. D.) of 8 IJm have a coefficient of capillarity (K = I. D.lO. D.) of 0.62 and therefore a nearly 40% lower specific gravity than solid E-glass fibers. These fibers are commercially available in the form of yarns, ravings, fabrics, mat, chopped strand, and tape. By definition, hollow E-glass fibers with these dimensions have lower tensile strength (2.5-2.8 GPa) than solid E-glass fibers (3.3 GPa). Their modulus is lower than that of solid E-glass fibers, and increases with decreasing capillarity [6J. However, the specific modulus of hollow E-glass fibers (modulus divided by density) ishigher than that ofsolid fibers [10].

158

Chapler6

Hollow E-glass fibers with a capillarity (K) of 0.5-0.6 are more sensitive to moisture andlor heat treatment than solid E-glass fibers. Upon exposure to high humidity (>90%), the residual strength drops to about 80-85% of the low original strength. Upon exposure to temperatures between 300 and 500°C, surface crystallization occurs and residual tensile strength drops to about 35-50% of the original strength. 8ecause of their thin walls, hollow fibers are more fragile than solid fibers and need to be handled more carefully. They will not however crush under pressure inalaminating press. Hollow E-glass fibers can produce either composite parts with equal thickness and up to 25% lower part weight than E-glass, or composite parts with equal weight and up to 25% higher thickness. In equal weight filament wound cylinders, the hydrostatic collapse pressure is increased by 30% and the dielectric constant is reduced from 6.8 to 4.0. In equal thickness laminates, acoustic transmission is increased as the result of lower mass, dynamic fatigue is significantly increased, and thermal conductivity is reduced by up to 40% . In a hybrid laminate, partial substitution for graphite does not reduce the high specific modulus while removing up to one-half ofthe more expensive graphite fiber. Weight reduction is a powerful incentive in aircraft manufacturing. Every kilogram ofstructure that can be eliminated facilitates a commensurate reduction infuel consumption and therefore cost, ora commensurate increase inpayload atequal fuel cost. Hollow glass fibers qualify for use in non-load bearing interior aircraft applications such as sidewall and ceiling panels. The use of hollow glass fibers for such applications are more prevalent in Russia and CIS countries [53) than inthe United States and the Western world. (c) Hollow porous sheathlcore silicate glass fibers Hollow glass fibers with a porous sheath were made but are not commercially available. They were first formed on laboratory [55) orcommercial equipment [54) and then acid leached by a process otherwise used in the production of porous, solid glass fibers (see Chapter 6.4.4). For example, hollow porous glass fibers with >95% Si02were produced by acid leaching of hollow glass fibers including borosilicate E-glass [54). Acid leaching of hollow glass fibers based compositions having 35 to62% Si02, 1 to 11 %Ab03, 0 to54% 8203, 3 to9% Zr02 and I to 29% Na20 (but no CaO) gave silica-rich, porous hollow glass fibers with good alkali resistance. Acid leaching of hollow glass fibers containing 54.0% Si~, 22.4% CaO, 14.3% Ab03, 7.2% 8203and 1.0% Na20 (but no Zr02) gave porous, hollow glass fibers with low alkali resistance [54). One of the most challenging tasks is to produce porous, hollow glass fibers with controlled pore size as required for inorganic membranes having uniform mechanical properties including strength and stiffness, thermal and chemical stability, photochemical and biochemical durability, and superior resistance to compaction under high membrane pressures. The required control over the desired pore size in fibers has recently been demonstrated [55) with a glass composition primarily consisting of 57.2 mol%, 22.8% 8203, 9.2% CaO. This study serves as a model for the evaluation of hollow porous glass fibers in reverse osmosis, phase separations, salt extraction, and biochemical research. (d)

Hollow superconducting sheath/core glass fibers

High temperature superconductor materials are superconductive when cooled below their respective critical temperatures (Te). This temperature should be higher than 77°K so that

ChaplerG

159

they can yield functional performance in liquid nitrogen. Low temperature, metallic superconductor alloys perform only in liquid helium, and represent a far more costly process. Known high r, materials include Y-Ba-Cu-O (YBCO) and Bi-Sr-Ca-Cu-O (BISCO). High Tc superconducting fibers have been made by three processes: The first is the powder-in-tube or Taylor process that relies on encapsulating superconducting powders in a metal tube. The second is the laser heated float zone orlaser heated pedestal growth process and the third is the fiberglass process except that a different bushing isrequired. So far the most practical route is the powder-in-tube process that consists of continuously filling metal tubes with high Tc ceramic powders, reducing their diameter :n a metal drawing process, and pressing the resulting bicomponent sheath/core metal/ceramic wire into a ribbon shaped product. Wire drawing falls outside the scope of this book. The laser heated float zone or laser heated pedestal growth is a slow crystal growth process. It is useful for experimental exploration but not commercial production. Forming sheath/core fibers having a glass (rather than a metal) sheath surrounding the high Tc superconducting ceramic material goes back to1990 [44]. In the earliest process for making superconductive glass fibers [44·45], the raw material powder needed for Y-Ba-Cu-O type fibers was filled into a large, thick walled, closed end glass tube. The tube was first heated until the superconductor powder melted and was further heated until a solid concentric bicomponent sheath/core fiber could be downdrawn. One component, the sheath, was glass, and the second component, the core, was a superconducting material. Essentially, solid fibers were downdrawn from a preform having a molten core. Upon solidification, the fiber was proven to possess moderately effective superconducting properties. The most recent version of this process [46] seems to have overcome the deficiencies of earlier processes [44-45]. It yields a hollow, three component, superconducting glass fiber, having an outer glass sheath and an inner superconducting sheath, both surrounding a hollow core (see Figure 11 b). These fibers possess the requisite density and microstructure to effectively carry an electrical current while also possessing the flexibility and strength required for commercial uses. The first step in this process involves melting a superconducting composition, e.g., Bi-Sr-CaCu-O, and a leachable glass composition, e.g., 75% Si02, 20% B203, and 5% Na20 (by weight), in a double bushing (Figure 11b). The gas outlet of the double bushing is centrally located within the superconducting core portion. Concentric streams of molten glass and superconductor are discharged from the bushing, converge below the tips, and are drawn through tension applied from the winder. The resulting fiber is a hollow preform with an amorphous superconductor core but it does not yet possess superconducting properties. The as-produced fiber preform is unwound from the package and is first annealed at 700950°C, depending upon the superconducting system used to convert the core material from its amorphous state into a crystalline state capable of superconducting properties. The superconducting fiber that emerges from this heat treatment is then coated to prevent microcracking and moisture degradation. Alternatively, the fiber preform can first be pretreated at a temperature ranging from 400 to 650°C to promote a phase separation in its sheath into two separate phases. Phase separation converts the glass composition into a chemically active boron-rich phase and a

160

Chapter 6

chemically inert silica-rich phase. The boron-rich phase can be removed by chemically leaching in an acid bath, leaving behind a fiber preform with a microporous silica-rich surface. As before, however, the amorphous core ofthis fiber must still be converted into a crystalline, and therefore superconducting, core between 700 and 950°C. Hollow superconducting glass fibers made by this process can be 5 to10km long. They have high current carrying capacity (>10 4 amps/em'), high mechanical strength, substantially circular cross sections in these lengths, and diameters ranging from 10 to 100 IJm. Accordingly they can be easily wrapped about small cylinders to form coils for motors. (e) Solid side-by-side bicomponent glass fibers

A centrifuge process for making dual or bicomponent glass staple fibers has recently been described [56]. Two separate glass melts, each having a different composition and therefore viscosity, are supplied toeach round cross section tip in a typical multifilament bushing. The melt streams meet under the tips, fuse, and as they cool, yield side-by-side bicomponent glass fibers with essentially round cross sections. The resulting fibers develop an irregular crimp since a differential stress develops at the interface of the components. A related process [9) yields a mixture of single and dualglass insulation staple by simultaneously centrifuging single and dual melt streams. Backscattered electron images (BEl) show that the individual bicomponent fibers have cross sections ranging from round to oval and low, but widely variable, fiber diameters (3-10 IJm). Individual fibers can split at their dual glass interface either during manufacture, or in subsequent processing and handling, thus producing fine, low diameter chaff considered tobe undesirable in use. An individual cross section is shown in Figure 12. Component (A) has a different composition than component (8), and the interface between both components is clearly discernable.

Figure 12. Backscattered electron image (BEl) of a polished cross-section of a commercial bicomponent glass fiber (4000x). Courtesy ofMicron Inc., Analytical SelVices, Wilmington, DE.

Chapter 6

161

Successive secondary electron image (SEI) profiles can be used to determine differences in composition between composition (A) and composition (8). This isaccomplished by traversing a polished cross section of the fiber at a right angle to the interface between the two components, as illustrated for Ca and Mg (Figure 13). Accordingly, the Mg (or MgO level) in this sample is higher in component (A) than in component (8), and the Ca (or CaO) level is higher in component (8) than in component (A). In addition, the Si02and Na20 levels (not shown in Figure 13) were also higher in component (A) than in component (8) and the Ab03 levels were lower in component (A) than in component (8); only component (A) analyzed for potassium, and only component (8)exhibited boron.

9.00 E-02 7.20 E·02 c 0

~

Ca

5.40

~ E·02

~

E 3.60 Ql E E-02

...... . . .

~

Ql

tn

Mg

1.80 E·20 0.00 E-010.00

"....... ..........

4.00

8.00 12.00 Distance, urn

16.00

20.00

Figure 13. secondary electron image (SEI) profiles of Ca and Mg - determined by traversing the cross section of the two fiber components at a right angle totheir interface, indicate that the MgO level is higher in component (A) than incomponent (B), and that the CaO level ishigher incomponent (B) than incomponent (A). Courtesy ofMicron Inc., Analytical Services, Wilmington, DE.

In summary, the individual components of the dual component fiber differ with regard to their Si02, MgO, CaO, Ab03, 8203, K20, and Na20 content. The differential stress between the two components during fiber formation produces a three-dimensional, crimped, non-straight fiber geometry. In thermal insulation batts, a main application, arrays of crimped bicomponent glass fibers offer higher bulk and loft than possible with a comparable array of straight fibers. Arrays of crimped fibers can trap more air in their dead spaces and therefore offer higher thermal insulation than comparable arrays ofstraight fibers having the same basis weight.

162

Chapter 6

Arrays of crimped staple fibers can be processed on conventional textile staple process equipment [58]. In the carding process, crimped staple fibers are mechanically opened, aligned and combed and converted into non-woven webs. In the needling orneedle punching process, arrays of crimped staple fibers are mechanically interlocked, densified and consolidated by repeatedly punching through the fiber batt with barbed needles. In the air lay process, crimped staple fibers are separated, aligned and consolidated in a non-woven web with a high velocity air flow. In these conventional textile processes, crimped bicomponent glass staple fibers aim to compete with commodity fibers, e.g., cotton and polyester staple, and premium high performance fibers, e.g., Nomex staple [58].

6.4 High temperature silica glass fibers Silica glass fibers are made by one of three generic processes. Ultrapure silica fibers (99.99 99.999% Si02) are formed either by downdrawing from preform rods (Chapter 4) or dry spinning from a sol-gel (Chapter 5). Pure silica fibers (99.5% Si02) are dry spun from viscous water glass solutions (Chapter 5) and high silica fibers are made by acid leaching ofselected silicate fibers (this chapter). Silica fibers, irrespective of process, are amorphous, including those called quartz fibers. Ultrapure and pure silica fibers retain useful strength up to 10001100°C. In contrast, high strength (HS) glass fibers such as S-glass (Chapters 6.2) retain useful strength to800-900°C. Leached high silica fibers are limited intheir use to<600°C. 6.4.1 Value-in-use ofsilica fibers For a very good reason, silica fibers are known as ultrahigh temperature (UHT) fibers. They can be continuously used atmuch higher temperatures than typical high strength and/or high modulus glass fibers and retain usable strength at much higher temperatures than the latter. The pristine strength of a range of fibers starting with borosilicate E-glass fibers and ending with aluminosilicate Nextel ceramic fibers isshown inTable XIII both atroom temperature and atvarious elevated temperatures. At room temperature the strongest fiber is S-glass, followed by E-glass and Nextel. Borosilicate E-glass fibers support continuous use to GOO°C, S-glass to 815°C, high silica glass fibers to 1040°C, pure and ultrapure silica glass fibers to 1090°C, and Nextel to 1200°C. The estimated cost per kilogram of fiber increases one-hundred-fifty-fold as the estimated continuous use temperature doubles. Table XIII. Strength of high and ultrahigh temperature fibers Pristine Strength T,oC RT 400 600 800 1000 Cont. Use ReI. Cost

Silicate E-glass 5-glass GPa GPa

High GPa

Silica

Pure GPa

Ultrapure silica Preform Sol-gel GPa GPa

Nextel aluminate GPa

3.5 1.8 0.9

4.6 3.8 2.4 0.7

0.4 0.2 0.1 0.1 0.1

1.7 1.6 1.4 1.2 0.9

2.4 1.8 1.4 1.0 0.9

0.9 0.9 0.9 0.1 0.1

2.6 2.4 2.4 2.1 1.9

600°C 1

815°C 6

1040°C 10

1090°C 30

1090°C 68

1090°C 80

1200°C 150

163

Chapter 6

Another interesting property of silica glass fibers is their low coefficient of thermal expansion, a"" 0.5x 1O.6·C' . This value can be lowered and even rendered negative byadding Ti02 to silica [13J [16-17J. Silica glass fibers also display excellent dielectric properties. Finally, they are known for their high resistance tocorrosion in neutral oracid chemical environments. 6.4.2 Ultrapure silica fibers from preforms Temperatures above 1800·C would be required to contain a fiber forming silica melt in a bushing. A requirement like this would by far exceed the practical capability of most precious metal and/or other practical alloy materials. As a result, pure silica glass fibers are downdrawn from solid silica preform rods (Figure 14). Gas flame orelectrical furnaces soften the ends of the quartz rods, and thereby facilitate the formation of continuous silica fibers. Preforms made from natural silica contain 99.99% Si02 and impurities include 20-50 ppm AI, <5 ppm OW, and <4 ppm Na. Preforms made by oxidation ofSiCI 4 ina plasma flame [60) afford even purer silica fibers «1 ppm AI, <0.1 ppm OH-, <1 ppm Na, and <50 ppm CI).

\? \ 2

Figure 14. Schematic rod drawing process from quartz preforms. A large diameter silica preform rod (1) isheated inacircumferential heater, i.e.,agas bumer, an electrical fumace orlaser source, (2) the molten jetsolidifies and the resulting quartz filament (3) is passed over a sizing device (4) and collected on a winding drum (5). Redrawn from M. S. Aslanova, Glass fibers, page 34, Khimia Publisher (in Russian), Moscow (1979).

Structural silica or quartz fibers made by downdrawing have diameters ranging from 7 to 14 IJm; rovings are made with up to4800 filaments. These fibers offer superior heat resistance since they retain useful strength at very high temperatures. They also possess the high ablation resistance and the dielectric, acoustic, optical and chemical properties of quartz from

164

Chapter 6

which they were made. Their pristine modulus is low (69 GPa), their pristine strength is as high as 4.8 G - 6.0 Pa [10], and because of their low density (2.2g/cc), their specific pristine strength (2.2 -2.7 Mm) is the highest ofany pristine glass fiber on record. Quartz fibers offer superior IR and UV transmittance, but contain water, and therefore exhibit a strong hydroxyl group in their IR transmission spectra. The presence of water blocks IR transmission in a critical region of the spectra, but it can be avoided by selecting a synthetic route that yields water-free quartz for the preform fabrication. Mined quartz may contain traces of uranium or thorium, which emit alpha particles and can disturb delicate electronic circuits or signals. Uranium- and thorium-free quartz fibers afford quiet HT circuit boards and/or electronicpackaging. Ultrapure silica or quartz fibers are used in fabrics, yarns, rovings and threads. Fabrics are used to reinforce radomes, antenna windows for missiles, high temperature circuit boards, and rocket nose cones. Braided yarns provide high temperature electrical insulation, e.g., for coaxial cables, thermocouple wires, and space separators. Rovings are used to reinforce polymer matrix composites for ablative and electrical uses, as well as high performance sporting goods, e.g., tennis racquets and skis, especially when hybridized with carbon fibers. Threads are used to stitch cable tray insulation for nuclear power plants. 6.4.3 Ultrapure and pure silica fibers from solutions One dry spinning process is capable of yielding ultrapure silica glass fibers from sol-gels and another dry spinning process is capable of yielding pure silica glass fibers from tetraethylorthosilicate (TEOS). Both processes have already been discussed in Chapter 5. The following discussion deals with their properties and uses in relation to those of the other silica fibers. Experimental silica fibers derived from a TEOS gel have higher purity than that of silica fibers derived from quartz preforms [61]. In fact, they contain 99.999% Si02and only very minimal impurities (<1 ppm AI, OH', and Na each, and no CI). The cause for the utmost purity isselfevident. An experimenter can use reagent grade Si{OC2Hs)4 or tetraethoxysilane in the experimental dry spinning process, while commercial grade sodium silicate is used in the commercial dry spinning process. The commercial process [62-64] for dry spinning pure silica fibers (95-99% Si02) resembles the sol-gel process for making ultra pure silica fibers (Chapter 5). The impurity level is considerably higher than that of downdrawn silica fibers, consisting of 450 ppm AI, 206 ppm Fe, 120 ppm Na,and <0.5% water. At room temperature, the pristine strength of pure silica fibers derived from water glass is one-half that ofultrapure silica fibers derived from preforms (1.7 GPa vs. 3.4 GPa) as shown in Figure 15. At 600°C, the pristine strength of both silica fibers is the same (1 .4GPa). Between 600 and 1000°C, it is also identical but drops from 1.4 to 0.9GPa. In summary, the purity and cost ofdry spun silica fibers ismuch lower than that ofdowndrawn silica fibers and, unlike the latter, they are not available as continuous filaments, only as sliver. In high temperature applications the performance of dry spun and downdrawn silica glass fibers is nearly the same. Dry spun silica glass fibers are less costly than downdrawn silica glass fibers, and therefore find increasing use in all high temperature markets except those requiring high specific strength at high temperatures, e.g., in aircraft or aerospace

Chapter 6

165

applications, or in those high temperature markets requiring high radiation resistance or electromagnetic shielding. Selected pure silica products include plied yarns for weaving narrow and broad fabrics, braided and twisted ropes, and sealings and sleevings for high temperature uses, including turbine insulation, removable flexible insulation, and compensators.

4.0

~

_

IIIIIIJ NEXTEL 312 rs:ssI a & A Silica tzzzJ E-Glass !QQQI Refrasil y-457 c:::J ASAHI Silica

3.0

o

.£ en

c:

...

Ql

Silfa Silica

4.0 3.0

2.0

2.0

1.0

1.0

I II

~

'iii

c ~

0

K 20

400

600

800

1000

0

2 hour exposure to air °C

Figure 15. Strength retention ofcommercial silica yams atelevated temperatures. Silfa and O&A fibers arederived from the liquid phase. Silfa is a pure silica yam made from a viscous waterglass solution. O&A isan ultrapure silica fiber made from a high viscosity melt. Asahi, Nextel and Refrasil are derived from solid phase precursor fibers. Asahi is an ultrapure silica fiber. Nextel is an aluminate fiber. and both arederived from a sol-gel precursor fiber. Refrasil is a high silica fiber; it is derived from borosilicate E-glass by acid leaching. Redrawn from product information supplied byAmetek Corporation inWilmington, DE

6.4.4 High silica fibers by leaching ofborosilicate fibers High silica glass fibers (95-99.5% 8i02) are the lowest cost silica fibers in today's market. In the United States they are made by leaching near-borosilicate (E-glass) fibers with aqueous HCI solutions, and in Russia and Japan they are made by leaching sodium silicate (A-glass) fibers, which have Si02/Na20 ratios of 3:1 to4:1, with aqueous H2S04 and/or HN03 solutions [10]. Almost all alumina, magnesia. calcia, iron oxide, and boron oxide [27] can be extracted from E-glass fibers in 8 hours at 95°C in dilute H2S04. A porous high silica glass fiber is obtained which contains 95-98% Si02(see Chapter 4). Leaching proceeds by ion diffusion. The chemical nature ofthe acids does not affect leaching kinetics [10]. These fibers can be used from room temperature to 1000·C, but because of their ultralow strength, only in non-load bearing end uses. Pristine fiber strength decreases from room temperature (0.4 GPa) to 1000·C (0.1 GPa). Application of a chromium oxide coating

166

Chapter 6

increases the upper in-use temperature to 1200·C for both long term and multiple cycle heat loads [10]. The fiber strength is so low that rovings or yarns are too fragile to be converted into fabrics. High silica fabrics are therefore best obtained by leaching woven, braided or other pre-designed E-glass fabrics. With increasing temperature, the decrease in fabric strength is accompanied by removal of water from the hydroxylated fiber surfaces, and by increasing nanocrystallinitythat may cause a slight but increasing fabricshrinkage. High silica fabrics obtained by acid leaching of E-glass fabrics were introduced in the early 1960s for aircraft and aerospace applications. Their use inthese end uses has been eclipsed by the advent ofultrapure and pure silica fibers which are not only stronger and can be woven and braided, but are also more expensive. In today's market, high silica fabrics fill a very important market niche. They are very desirable and cost effective heat insulating materials, but for less demanding uses than those which require pure or ultrapure silica fibers. Among others, they are used as high temperature filtration media for ferrous, non-ferrous and corrosive metals, as reinforcement for polymer composites and as heat resistant electrical insulation in nuclear reactors [10]. REFERENCES [1] [2] [3] (4) (5) (6) (7)

[8] [9] [10] [11] [12] (13) [14] [15) [16] [17] (18) [19]

K. L. Loewenstein, The Manufacturing Technology of Continuous Glass Fibres, third, completely revised edition, Elsevier, Amsterdam (1993). P. K. Gupta, Glass Fibers for Composite Materials, Chapter 2 in Fibre Reinforcements for Composites Materials, A. R. Bunsell, ed., Composite Materials Series 2,Elsevier, Amsterdam, 19-71 (1988). W. L. Eastes, D. A. Hofmann, J. W. Wingert, Boron-free glass fibers. Intemational Patent Application, W096/39362, December 12, 1996. F. Rossi and G. Williams, A new era inglass fiber composites, Paper presented at the 28th AVK Conference, Baden-Baden, Germany, pages 1-10, October 1-2, 1997. B. A. Proctor, Continuous filament glass fibers, in COncise Encyclopedia of COmposite Materials. A. Kelly, editor, 62-67, ElsevierScience Inc., New York (1994). Naamlooze Vennootschap Maatschapij tot Beheer en Exploitatie van Octoorien, New and improved glass compositions forthe production ofglass fibers. Brit. Patent, GB 520,247, April 18, 1939. P. F. Aubourg and W. W. Wolf, Glass fibers, in Advances in Ceramics, Vol. 18, Commercial Glasses, pages 51-63, D. C. Boyd and J.F.MacDowell, editors, American Ceramic Society, Westerville OH (1986). R. A. Schoenlaub, Glass compositions, U. S. Patent 2,334,961 , November 23,1943. R. L.Tiede and F. V. Tooley, Glass composition, U. S. Patent 2,334,961 , November 2, 1951 . V. E.Khazanov, Yu., I. Kolesov, and N. N. Trofimov, Glass fibers, in Fibre Science and Technology, 15-230, V.I.Kostikov, Ed., Chapman and Hall, London (1995). Trevor Starr, Glass-Fibre Databook, Edition 1, Chapman &Hall, London (1995). Trevor Starr, Carbon and High Performance Fibers, Directory and Databook, Edition 6, Chapman & Hall, London (1995). T. D. Erickson and W. W. Wolf, Glass compositions, fibers, and methods of making same, U. S. Patent 4,026,715, May 31,1977. S. Tamura, M. Mori, and S. Saito, Compositions for the production of high-strength glass fiber, Japanese Patent, 8[1996)-231-240, September 10,1996. J. F. Bacon, High modulus, high temperature glass fibers, Applied Polymer Symposium No. 21, 179-200 (1973). A. Lewis and D. L.Robbins, High-strength, high modulus glass fibers, Journal ofPolymer Science, Part C, 19, 117-150 (1967). F. T. Wallenberger, S. D. Brown and G. Y. Onoda, ZnO-modified high modulus glass fibers, Journal of NonCrystalline Solids, 152,279-283 (1993). F. T. Wallenberger, Melt viscosity and modulus ofbulkglasses and fibers - challenges forthe next decade, in Present State and Future Prospects of Glass Science and Technology, Kreidl Symposium, Triesenberg, Liechtenstein, July 3-8, 1994, Glasstech. Ber. Glass Sci. Technology 70C, 63-78 (1997). J. Kobayashi, M. Oota, K. Kada and H. Minakuchi, Oxynitride Glass and the Fiber Thereof, US Patent 4,957,883, September 18, 1990.

Chapter 6

167

[20] K. Suganuma, H. Minakuchi, K. Kada, H. Osafune and H. Fujii, Properties and microstructure of continuous oxynitride glass fiber and itsapplication toaluminum matrix composite, J. Mater. Res., 8 (1),178-186 (1993). [21] D. R. Messier and P. Patel, High modulus glass fibers, Joumal ofNon-cryst Solids, 182,271-277 (1995). [22] H. Kaplan-Diedrich and G. H. Frischat, Properties ofsome oxynitride fibers, Journal of Non-crystalline Solids, 184, 1343-136 (1995). [23] M. Oota, T. Kanamori, S. Kitamura, H. Fujii, T. Kawasaki, K. Sekine and C. Manabe, Decrease of silicon defects inoxynitride glass, J. Non-cryst Solids, 209, 69-75 (1997). [24] H. Minakuchi, H. Osafune, K. Kada, K Kanamaru and H. Fujii, Development ofthe Oxynitride Glass Fiber and its Properties, inScience and Technology ofNew Glasses, S. Sakka and N. Soga, editors, Proceedings ofthe Int Conference on Science and Technology of New Glasses, Zenkyoren Building, Tokyo, October 16-17, 1991 . [25] V. Budov, P. Sarkisov, K. Bormotunov, N. Trofimov, V. Khazanov, and Z. Shaina, Oxynitride glasses- the material for prospective glass fibers, Bolet Socied. Espan. Ceram. Vidr., 31 -C, 2, 391-396 (1992). [26] R. Wusirika, Alkali durability ofoxynitride glass fibers, J. Am. Ceram. Soc. 74 [2]454-456 (1991). [27] G. Wiedermann and N. Frenzel, Untersuchungen zur chemischen Bestandigkeit der Glasseide, Faserforschung und Textiltechnik, 24 [9], 335-340 (1973) [28) P. Simurka, M. Liska, A. Plsko and K. Forkel, Development of a composition suitable for the production of alkali-resistant glass fibres with a low fiberising temperature, Glass Technology, 33[4), 130-135 (1992). [29) V. I. Kostikov, M. F. Makhova, V. P. Sergeev and V. I. Trefilov, Ceramic fibres, in Fibre Science and Technology, V. I.Kostikov, Ed.,581-606, Chapman and Hall, London (1995). [30J A. J. Majumdar, Alkali-resistant glass fibres, in Strong Fibers, W. Watt and B. V. Perov, editors, pages 61-85, Elsevier Science Publishers, North-Holland (1985). [31) Steve Loud, Composites News, Infrastructure Newsletter, #11 , Solana Beach, CA, page 3, September 1994, and Infrastructure Newsletter #28, pages 5,June 30, 1995. [32] Chinese Cillass, Government Specification (1997). [33] A. Carre, F. Roger, and C. Variot, Study of acidlbase properties of oxide, oxide glass, and glass-ceramic surfaces, Journal ofColloidand Interface Science, 154 [1], 31-40 (1992). [34] D. A. Steenhammer and J. L. Sullivan, Recyded content of polymer matrix composites through the use ofAglass fibers, PolymerComp., 18 [3], 300-312 (1997). [35] Nitto Boseki, Glass fibers having low dielectric loss tangent - composed of silica, alumina, boria, calcia, opt magnesia etc., Japanese Patent 9002839,January 7, 1997. [36J J. F. Bacon, Composfion ofglasses with high modulus ofelasticity, U, S. Patent 3,573,078, March 30,1971 . [37] K. Komori, S. Yamakawa, S. Yamamoto, J, Naka and T. Kokubo, Substrate for circuit board including the glass fibers asreinforcing material, U. S. Patent 5,334,645, August 2, 1994. [38J K. Komori, S. Yamakawa, S. Yamamoto, J.Naka and T. Kokubo, Glass fiber torming composition, glass fibers obtained from the composition and substrate for circuit board including the glass fibers asreinforcing material, U. S. Patent 5,407,872, April 18, 1995. [39] R. A. Humphrey, Shaped glass fibers, in Modern Composite Materials, L. J. Brautman and R. H. Knock, eds., Addison-Wesley, Reading, MA (1967). [40] S. T. Gulati, Ribbons, in Handbook of Reinforcements forPlastics, J. V. Milewski and H. S. Katz, editors, 78124, Van Nostrand Reinhold Co.,New York (1987). [41] L. J. Huey, Method and apparatus for making tapered mineral and organic fibers, U, S. Patent 4,666,485, May 19,1987. [42J K. Shioura, S. Yamazaki and H. Shono, Method for producing glass fibers having non-circular cross sections, U. S. Patent 4,698,083, October 6,1987. [43J H. Taguchi, K. Shioura and M. Sugeno, Nozzle tip for spinning glass fiber having deformed cross section and a plurality ofprojections, U. S. Patent 5,462,571, October 31 ,1995. [44] A. Urano, K. Takahashi, K. Ohmatsu and M. Onishi, Method for the production ofceramic superconductor filaments, US Patent 4,968,662, November 6, 1990. [45J A. Urano, K. Takahashi, K. Ohmatsu and M. Onishi, Method for making superconductor filaments, US Patent 5,215,565, June 1,1993. [46J J. Huang, Hollow high temperature ceramic superconducting fibers, International Patent Application WO 97/22128, June 19,1997. [47] J. A. Burgman and L. L. Margason, Method and apparatus for forming hollow glass fibers, U. S. Patent 3.268,313, August 23, 1966. [48] J.A. Burgman and L.L.Margason, Hollow glass article, U. S. Patent 3,510,393, May 5,1970. [49] T. H. Jensen, Hollow fiber bushing and hollow fiber tip construction, U. S. Patent 4,698,082, October 6, 1987. [50] T. H. Jensen, Hollow glass fiber bushing, method of making hollow fibers and the hollow glass fibers made by that method, U. S. Patent 4,758,259, July 19, 1988. [51] L. J. Huey, Method and apparatus for producing hollow glass filaments, U. S. Patent 4,846,864, July 11 , 1989.

168

Chapter 6

[52) M. S. Aslanova, Glass fibers, 3-60, in Strong Fibers, Handbook of Composites, Volume 1,W. Watt and B. V. Perov, Editors, North-Holland, Amsterdam (1985). 9,19(1995). [53) G. Demidov, Hollow fibres make light and strong reinforcements, Reinf. Plas., [54) R. P. Beaver, Method for producing porous hollow silica rich fibers, U. S.Patent 4,778,499, October 18, 1988. [55) T. Yazawa, H. Tanaka and K. Eguchi, Preparation ofporous hollow fibre from glass based onSi02-B20:rROZr02 (R= Ca, Zn) system, J. Mater. Sci. Letters, 13,494-495 (1994). [56) J. E. Loftus, C. R. Strauss and R. L. Houston, Method for making dual-glass fibers by causing one glass to flow around another asthey are spun from a rotating spinner, US Patent 5,529,596, June 25, 1996. [57] J. L. Bemard and G. Berthier, Mineral wool products and method and device forproducing them, European Patent Application, EP 0 801 038 A2, October 4,1997. [58) N.T. Huff,lnnovative technology can create products, Glass Researcher, 5 [lJ, 1-9(1995). [59) M. C. Kenny, S. K. Barlow and S. L. Eikleberry, New g;ass-fiber geometry - a stUdy of non-woven processability, TAPPI Joumal, 30, 169-177 (1997). [60) R. Bruckner, Silicon Dioxide, in Encyclopedia of Applied Physics, VCH Publishers, Inc., Vol. 18, 102-131 (1997). [61) Asahi, Product Bulletin (1986). [62) A. Wegerhoff, H. Zengel, H. Brodowski, H. Beck, E. Seeberger, G. Steenken and K. Hillermeyer, Watercontaining water glass fibers, US Patent, 4,471,019, September 11,1984. [63] A. Wegerhoff and H. D. Achtsnit, High temperature resistant fibrous silicon dioxide material, US Patent, 4,786,017, November 22, 1988. [64] H. D. Achtsnit, Textile silica sliver, itsmanufacture and use, U. S. Patent 5,567,516, October 23, 1996. [65] G. H. Vilzhum, H. U.Herwig, A. Wegerhoff, and H. D. Achtsnit, Silica fiber for high temperature applications, ChemiefasemfTextilinduslrie, 36/88, E-126-127 (1986). [66] Product Bulletin, Silfa Silica Yams, Ametek, Haveg Div.,Wilmington, DE (1996). [67] L.L. Hench, Bioactive glasses and glass ceramics,Handbook ofBioactive ceramics, Vol. I,T. Yamamuro, L.L. Hench, and J. Wilson, eds.,7-23, CRC Press, Boca Raton (1990). [68] H. Tagai, et aI., Preparation of apatite glass fiber for applications as biomaterials, Ceramics in Surgery, P. Vincenzini, ed.,page 387, ElsevierSci. Pub. Co.,Amsterdam (1983). [69] U. Pazzaglia, etaI., Study of the osteoconductive properties of bioactive glass fibers, J. Biomed. Mater. Res., 23,1289-1297 (1989). [70) M. S. Marcolongo, P. Ducheyne, F. Ko and W. La Course, Composite materials using bone bioactive glass and ceramic fibers, U. S. Patent 5,721 ,049, February 24, 1998.

CHAPTER 7 OPTICICAL SILICA FIBERS H. D. Ackler and J. B. MacChesney This chapter analyzes the background of data transmission, explores the principles of light guidance, optical materials properties and materials processing which are important factors in making these systems function ; it concludes by discussing fiber devices which have generated a new generation of optical communication systems.

7.1lntrodcution A primary characteristic influencing a system's information capacity is its bandwidth. This is the number of discrete channels within a finite range of frequency. Bandwidth increases with frequency of the carrying signal. The higher the frequency used for transmission, the higher is the theoretical capacity of the system. Traditional copper pairs operate at 1 GHz, which permits transmission of more than 10,000 voice channels, but loss of 50 dB/km limits the transmission span to <1 km. Coaxial cables offer higher signal attenuation but require amplifiers every mile or two. Microwave transmission systems operate at up to 10 GHz and 14 fiber optic systems at up to -10 GHz. Systems operation at terahertz (THz) frequencies [1] permits transmission of the contents of 30sets of an encyclopedia in one second. The compounded annual growth rate of single mode fiber is predicted to be about 10%-20% through 2001 [2]. The North American market alone is predicted toreach $2.3billion by2001, up from about $1 .5 billion in 1996. The use of fiber in CATV is expected to almost double from 1996 to 2001 , and is the technology of choice for new CATV system construction . Also, the number of users on the Internet is growing at around 20% per month, thus implying that this market will soon be demanding a great deal of fiber.

7.2 Principles of optical transmission The ability to guide light through space over long distances depends on the geometry of the waveguide in which the light isconfined and also on the properties of the materials from which it is made. 7.2.1

Waveguide physics

The optical index of refraction, n, of a material is a measure of the speed of light, given by n = c/vwhere c is the speed of light and v is the speed of light in the material. This property causes the direction of light to be bent or refracted when it encounters an interface between materials of different index. The refraction of light at such interfaces is governed by Snell's law (1 )

170

Chapter?

where nj isthe index ofmaterial i, and as well as 8 are the angles shown inFigure 1. As the incident angle of the light 81 gets smaller, so does the angle of the refracted light, 82. At the critical incident angle 8e, the exit angle 82 ~O. For all incident angles less than 8e, the light will be totally internally reflected. It is the principle of total internal reflection upon which waveguides are based, shown schematically in Figure 2. The exact details of waveguiding depend on the specific fiber core and cladding structure. These structures may be divided into two main types, step index and graded index fibers, shown in Figure 3. The step index structure may be either single or multimode, depending on the core diameter. The graded index structure ismultimode. These types will be discussed separately.

Refracted ray Cladding

n2

Cladding

Figure 1. Meridional ray optic representation of the propagation mechanism in an ideal step-index optical waveguide. From Figure 2-11. G. Keiser [3]. Optical Fiber Communications. McGraw-Hili Book Company. New York, NY (1983); with permission.

Figure 2. Waveguide structure showing the total internal reflection oflight. The diameter. d. is50IJm fora standard multimode system. 62.51Jm fora large core multimode. and from 4 to 10 IJm fora single mode. From Figure 2-12. G. Keiser [3), Optical Fiber Communications. McGraw-Hili Book Company. New York, NY (1983); with permission.

Chapter?

I?I

(a) Step Index Fibers One parameter common to both multimode and single mode fibers is the numerical aperture (NA), a measure ofthe light collecting ability of a fiber, given by

NA =

{n; -nJ 0,., nJIU

(2)

for step index fibers where n2 = nl (1-~). This represents the cone of angle eo shown in Figure 2 within which all incident light is guided by the fiber. This term also influences the number of modes, which may propagate in the fiber, through its relation to a dimensionless quantity called the normalized frequency, V, given by 21ra V = -NA = ka (2 A

r:': nJ -n22\~ f" ,., ka,,2~

(3)

where k=2111'A is the wavenumber of light of wavelength A and a is the core radius. For values of V<2.405 in a step index fiber, only a single mode, the fundamental mode, may propagate. For V>2.405, numerous modes may propagate proportional innumber to V. The multimode core structures shown in Figure 3 are much larger than the wavelength of the light, typically on the order of 50 urn in dlerneter. This permits many different paths, or modes, along which the light may travel, represented by the various rays or arrows in the figure. While the number of allowed modes may be large, it is in fact limited because the wave will experience a phase shift upon reflection at the core/cladding interface. For a mode to propagate, reflected waves must interfere constructively which imposes limitations on the angles at which a wave may be incident on the core/cladding interface, resulting in a set of allowed modes whose size isdetermined by the value ofV (3)15).

Cro,ss sectIOn

Multimode stepped index

Multimode graded

Indffix pro ile

Inp.ut pulse

mode s epped index

Cladding

put OUl pu se

A

r>:

A

(\

A

~

index

Sin ~ le

Light path

Figure 3. Illustration ofsingle and multimode optical fibers showing ray paths and the pulse spreading caused by each structure. Lower right diagram shows glass structure comprising high-silica glass fibers. From Figure 3, S. Kosinski and J. B. MacChesney [4), Fiber optics, Kirk-Othmar Encyclopedia ofChemical Technology, 4th Edition 10, John Wiley & Sons, Inc. New York. NY (1994); with permission.

172

Chapter?

The single mode core structure is not much larger than the wavelength of light and will only allow the fundamental mode to propagate. In this case, the wave is not reflecting off the core/cladding interface, bouncing back and forth along the waveguide. Rather, the wave is confined to the core region. Wave theory [3J [5J is required to describe the propagation of such a mode. The evanescent field lies outside the core in the cladding region ; it decays exponentially with distance from the core/cladding interface. In single mode fibers, the evanescent field may extend into the cladding a distance greater than a core diameter. This point is very important. It requires the cladding material to have optical quality comparable to that of the core material toprevent signal degradation. Other modes, which may exist in an optical fiber, are so-called radiation modes, encountering the fiber end at angles greater than So and propagating in the cladding. These modes generally decay with distance traveled due to interactions with defects on the exterior surface ofthe cladding. They may undergo mode coupling with the evanescent field of guided modes and, asa result, intensity may be transferred between the guided and radiation modes. This will lead to loss of the guided mode intensity, so the cladding is generally coated with an attenuating or absorbing material to exfinquish the radiation modes quickly, thereby preventing them from drawing power from that propagating in the core [3J [5]. (b) Graded Index Fibers From Figure 3 one may observe that the many modes in a multimode fiber travel different paths along the fiber axis. Lower order modes travel closer to the center, while higher order modes are further from it. This results in a path length difference between modes, which creates a time lag between them. This time lag causes different modes to arrive at the end of the fiber atdifferent times, and is known as intermodal dispersion. By varying the index ofthe core as a function of radius, the effective differences in path length may be reduced. Typically, the core index profile isfabricated as

for Os;r

~

a

(4)

=n,(1-2tl)YJ ""n/(J-tl)=n2 for r s;a where a defines the profile shape, nl is the core index atthe axis and n2is the cladding index. A step index profile is represented by a --+ 00 . 7.2.2 Optical loss Loss is probably the most important characteristic offiber materials. Loss, or attenuation, is a measure ofthe reduction insignal intensityasit travels through a material and isgiven by loss (dB) =

lOlog~ fa

(5)

where 10 is the input intensity and I the output intensity. For the loss after a unit length traveled, the loss per unit length isgiven bythe expression

173

Chapter 7 -a

I(x) = I ol0

lOx

(6)

where x is the distance or length traveled and a is the lossnength. The lower the loss per unit distance traveled, the farther a signal may travel along a fiber before becoming so weak that it must be amplified again. Current commercial silica fibers have losses between approximately 0.35-0.2 dB/km at telecommunications wavelengths (1 .3 and 1.5 urn). In fact, differences in loss of a few hundredths ofa dB/km can determine whether or not a given fiber iscompetitive inthe marketplace. Loss is due to a number of mechanisms, some of which are wavelength dependent. These include scattering, fiber defects, and absorption. The total loss, a(A), isgiven by

a("l) = ~+C2 +A("l) "l4

(7)

where Ais the wavelength oflight. (a) Scattering The first term inthis expression describes Rayleigh scattering which iscaused by microscopic (i.e. small compared to the wavelength) fluctuations in the refractive index of the material. These fluctuations may be caused byfluctuations incomposition ordensity ofthe glass. The second term in Equation 7 is due to defects in the fiber, such as particles or bubbles, defects between the core and cladding, or distortion from bending. A particle or bubble will cause light to be scattered out of the fiber, and these are to be avoided at all cost. The scattering due to core/cladding effects may be caused by geometrical irregularities in the core/cladding interface or localized stresses, such as residual stresses from drawing or externally imposed stress. These losses are sometimes referred to as microbending losses, and are typically generated by problems indrawing or cabling offibers. Macrobending) losses occur whenever afiber isbent. Recall that for any mode, some of the power travels in the cladding as the evanescent field . When a fiber is bent, the evanescent field in the outer portion of the bend must travel faster than the field inthe core toremain inphase. As the distance from the fiber axis increases, the velocity ofthe evanescent field inthat region must also increase until, atsome critical distance from the axis, it would be required to exceed the speed of light. Such a condition is not possible, and the intensity is radiated away. Another form of bending loss is associated with changes in bend curvature, either on the micro- or macroscale. Changes in curvature permit mode coupling between guided modes and modes in the cladding which are attenuated by loss at the cladding surface. Bending losses are reduced for larger ~, which confines the modes inthe core toa greater degree. (b) Absorption The last term represents absorption losses due to impurities and intrinsic UV (electronic) and IR (Vibrational) absorption. Absorption due to molecular vibrations in the IR tends to occur in bands over which the frequency of light resonates with the vibrational frequency of molecular bonds. For Si02, this does not become significant until wavelengths are longer than about 1.5

Chapter?

174

um, Impurities exhibit absorption peaks, generally with narrow widths, corresponding to discrete transitions. Transmission losses for silica based glass are shown in Figure 4. Extremely small quantities of transition metals, on the order of 1 ppm Fe+2, can raise losses due to absorption by an order of magnitude or more. Similarly, vibrational absorption due to hydroxyl groups (water) can be very detrimental [3] [5].

105 104

e

ill

103

': ,02

~

10 1.0

'\\t!

~

\Otr

'I'l

LL \

0.1 . 01

10. 1

B

1.0

c

10.

2

10-3 10-4

i

e ~

.~

10-5 ~ 10-6

7 10'0 -

"

~

8

Wavelength, IJfTl

Figure 4. Transmission loss vs.wavelength forasilica-based glass fiber. Region A represents electronic transitions; B, the transmission window; and C, molecular vibrations. Point LL is the lowest loss observed in an optical fiber. Absorption profiles for1 ppm OH and Fe+2 are also shown. From Figure 4, S. Kosinski and J. B. MacChesney [4], Fiber Optics, Kirk-Othmar Encyclopedia of Chemical Technology, 4th Edition 10, John Wiley & Sons, Inc.. New York, NY (1994); with pennission.

The reduction of these impurities tothe parts per billion level enabled a dramatic reduction in loss and emergence of optical communications. In the 1970s, losses were reduced to below 20 dB/km [7], thus enabling the production of optical fibers, which were practical for data transmission. Significant processing breakthroughs of silica have essentially eliminated transition metal impurities, dramatically reduced fiber losses and facilitated the production of glass with absorption low enough tomake a window a mile thick with essentially no reduction inbrightness. 7.2.3 Dispersion Dispersion is the wavelength dependence of the speed oflight traveling ina medium. It is the material property responsible for the separation of white light into a rainbow of colors by a prism. This isan extremely important fiber parameter, perhaps as important as loss, because it leads to signal degradation over long distances. It causes parts of the signal to travel at different speeds resulting in pulse broadening which limits data transmission rates. As pulses broaden it becomes difficult to distinguish adjacent pulses due to overlap, requiring the time between pulses to increase. It is due to the mechanisms of material dispersion, waveguide dispersion and, inthe case ofmultimode fiber, modal dispersion. For single mode fiber, material and waveguide dispersion can be tailored to produce zero chromatic dispersion at a particular wavelength. A parameter which is used to quantify dispersion is the time delay or group delay, tg, between components of a signal having different wavelengths (sources are not purely monochromatic) given by

175

Chapter 7

tg

A2 dfJ

L

2!ee dA

(8)

-=---

where L is length traveled, c is the speed of light, and

fJ = (27rn(A» / A is the propagation constant for a given wavelength. For a source of spectral width (rms value) of 0')... the spreading of a pulse isapproximately given by

r

dt

g

(9)

=~(j dA ..t

and the dispersion, 0, is represented by 1 dt g D=-L .u

(10)

which is typically given in units of pslkm·nm (ps=picoseconds). These expressions apply toall of the dispersive mechanisms mentioned above. Material dispersion. a fundamental property of a glass, originates in its electronic structure. This type of dispersion may be varied with changes in composition, but does not vary greatly for typical commercial applications and is essentially independent of waveguide design. Commercial fiber cores are primarily Si02with minor additions of Ge02 and perhaps P20S. Cladding is typically pure Si02orperhaps F-doped Si02. Some useful approximations for material dispersion are given in reference (6). For the wavelength range, 800-1700 nm, Equation 11 is claimed to beaccurate tobetter than 10%.

D(A) == 2.66 X /0-2 A_ 6.985 : A;

/0'11

(11)

Here D{A) is in psnm-krn, and A is in nanometers. For the range of interest to telecommunications, 1260-1700 nm, this is claimed to be accurate to better than 1%. With material dispersion essentially fixed, the profile of the core and cladding refractive indices may be designed to take advantage of the properties of waveguide dispersion to partially cancel material dispersion toproduce fiber with optimized total dispersion. Waveguide dispersion results from how waves traveling along a fiber interact with the profile in refractive index. Let us first discuss single mode fiber, since multimode fiber is treated somewhat differently. Various index profiles forsingle mode fibers are shown in Figure 5. For the sake of simplicity, the discussion of waveguide dispersion for such structures often proceeds with the approximation that the index does not vary with wavelength. For small .t\, a normalized propagation constant, b, may begiven as

b=

!j{k
(12)

176

Chapter?

Step

_~ru_

V = ak (2n 6n)1/2

T-OIC

-2a'N-DICorW ~n(1+o)


o=6'nlAn
=

~n

2a '

Figure 5. Fiber structures and parameter definitions for chromatic dispersion studies. From Figure 4.2, L. B. Jeunhomme [6], Single-Mode Fiber Optics, Principles andApplications, Marcel Dekker, Inc., New York. NY(1983); with permission.

Waveguide dispersion may thus be controlled by appropriate index profile design through its influence on V, a parameter dependent on the specific index profile chosen, as well as b, and subsequently

v( d 2 (Vb)/dV 2 ). The total dispersion isthen = -A. d

D(..1)=D +D //I

wg

C

2n(..1)

d 2(Vb) _ n2t1 v d..1 2 d dV2

(13)

The total dispersion will be zero when

~d c

2n(..1) d..1 2

= _ n2t1 V d 2(Vb) d

dV 2

(14)

From knowledge of the dependence of V and b on the fiber core diameter, A, and the dimensions of the depressed index cladding (if applicable), one may design a fiber whose wavelength ofzero dispersion has been shifted toa desired value [3] [6]. Not only will the wavelength of zero dispersion be shifted, but at a local maximum, so introduced, the dispersion will vary little over a window of wavelengths. The dispersion behavior for actual fibers of varying designs is shown in Figure 6. Note that for seemingly similar index profiles, the behavior varies widely. This places strict requirements on the precision offiber manufacture.

Chapter?

177

Total dispersion is extremely sensitive to variations in the index profile of both the core and cladding parameters as illustrated by the behavior offibers denoted W2 and W3. Though the diameters of their cores differ by only 0.1 urn, and the diameter of the depressed inner cladding even less, their dispersion behavior shows significant difference. Not only are the wavelengths of zero dispersion different, but W3 exhibits almost no dispersion over a window approximately 50 nm wide, centered at about 1530 nm. Clearly, this configuration is more desirable for operation at1.5urn than the fiber designated W2. Intermodal dispersion inmultimode fibers causes each mode totravel ata different speed due to different group delays, caused by different path lengths between modes. For step index multimode fibers, the maximum spread in delay time between the fastest and slowest modes isgiven by M

= ~(n, -n] C

)(l_lJ

(15)

VIII

where Vm is an integer corresponding to the highest mode allowed for a given value of V, approximately Vm = 2 V/'" This spread in delay time will cause pulses consisting of many modes tobe broadened.

10

. E

:.<

E

~ 0-

0

g

1.4 ,/ ,/

.

0

-10

/,

/

:/

,/

W

3

.'

A. (~m)

" ,.:, 8 2

.: / Wo

<,

.:

.

2aH

""LJ ---4a-

~-R ~63%

IJ lJr3%

to.58%

L.AO.19%

WO:

2a=5.7~m

-3a-

W1: W2: W3:

2a = 7 .7~m 2a

=6.6~m

2a = 6 .5~m

Figure 6. Dispersion of various fiber designs, showing sensitivity to variations inprofile parameters. Profiles S, and S2 illustrate wavelength sensitivity to design. Behaviorof W type fibers are shown by Wl,W2 and W3. From Figure 4.6,L. B. Jeunhomme [6J. Single-Mode Fiber Optics, Principles and Applications, Marcel Dekker, Inc., New York, NY (1983); with permission.

178

Chapter?

Graded index multimode fibers are far more complicated than their single mode counterparts. The relevant expressions are stated to permit insight into the relationships between material properties and fiber performance. The index profile as a function of radius r is given by Equation 4. For such a profile, the intermodal group delay may be approximated by

(3)j

_N,L[ 1+ a-2-c ~(rn) %+2 + 3a-2-2c 2(rn)2%+2 +O~ ( ) ~ c a+2 M 2 a+2 M

T;t11 - -

(16)

where

On, 2n,k o~ N, =n, + k - c=----

a .

N,~

a '

m is the number of modes with propagation constants between ~ and nlk, and M is the maximum possible number of bound modes allowed. It may be shown that Equation 16 goes to zero (to first order in ~) when a = 2 + s . Since I; is typically small, this gives an approximate ideal index profile for a == 2. The intermodal pulse broadening, Gint, may be described by (17) where (X) is the average of X over the power distribution of the modes. Equation 16 and Equation 17 may be combined to give an approximation to the graded index, intermodal pulse broadening

= LN,~

0" .

2c

11/1

--!!.-( a+2 )YJ[ cf 4c,c2(a+1)~ 16~2ci(a+1)2]YJ a+l 3a+2

+

2a+1

+

(Sa+2)(3a+2)

(18)

where

c,

=

a-2 -c a + 2 • c2

=

3a-2-2c 2(a + 2) .

Ignoring material dispersion (i.e. 1:=0, dn J dJ., = 0), Equation 18gives n,~2L

O"m;1I

= 205

(19)

As pointed out inreference [3], setting a ~ OC! and I: =0 allows us todetermine the intermodal pulse broadening for a step index, multimode fiber to be approximately (20) Therefore, the ratio ofgraded index to step index intermodal pulse broadening is O",,'ep _

10

~=~ mill

(21)

For common ~ values around 0.01, this indicates that intermodal pulse broadening is a factor of one thousand less than for step index fiber. Put in terms of information carrying capacity, the bandwidth of graded index fiber is a thousand times greater than for step index. This

Chapter?

179

seemingly minor change in fiber profile has enormous impact on fiber performance. However, the manufacture of such a fiber is not easily accomplished. The intermodal dispersion is extremely sensitive to deviations from the optimum a. [9]. By varying a. from 2 to 2.1 , the bandwidth decreases by an order of magnitude. This point is very important for it places severe requirements on the processing of these fibers. Producing such precise profiles in material properties is very difficult, as will be discussed later regarding the fabrication of optical fibers. 7.2.4 Birefringence Modern lightwave systems are influenced by polarization effects. One recent development is the extensive use of optical amplifiers, which dramatically increases the optical path, and the other development is the use of high bitrates, which has pushed the capacity of optical fiber tothe limit. Under these circumstances, the fiber's birefringence must be considered. So-called single mode fibers with nominal circular symmetry are in fact bimodal in that they can propagate two nearly degenerate normal modes with orthogonal polarizations. Minor perturbations along the fiber can couple these two modes so that a linearly polarized wave launched into one mode will be transformed into a general elliptically polarized wave, representing an admixture of the two modes. These perturbations result from manufacture: core ellipticity, stress produced by cabling, or by ambient conditions, such as temperature variation orcable bending. Figure 7a illustrates a slightly elliptical core and the resulting non-symmetric stress produced, among other things by the mismatch in coefficients of expansion of core and cladding. In addition to the intrinsic stress produced in manufacture, Figure 7b also shows environmental forces contributing to fiber birefringence. The sum of these perturbations causes the fiber to become bimodal and can be expressed [10J as the difference in local propagation constants: (22) where ()) = 2nc/"A. isthe angUlar frequency ofthe light, c isthe speed oflight in vacuum, n, and n, are the refractive indices of the slow and fast modes, respectively, and M .ff is their difference. By definition, n, > nt. The differential phase velocity is accompanied by a difference in local group velocities for the two polarization modes. This is responsible for broadening pulses as seen by a group delay time L'H per unit length between the modes. M L

=~ ({3" _ {3f) = linefj dco

c

_ wdlinefj cdoi

(23)

MIL is referred to as pulse mode dispersion (PMD). The length dependence for this dispersion islinear for short lengths and increases as the square root oflength for long spans.

Various strategies have been implemented for the elimination and reduction of PMD in fiber manufacture and cabling [11]. In addition to tailoring preform dimensions to improve geometry, a strategy of twisting the fiber during draw to introduce mode mixing has been successful in reducing PMD. For certain sensor and network uses, polarization-maintaining fibers are produced [12]. These use birefringences produced by asymmetric stress on the

Chapter 7

180

core either by deforming the core or by introducing stress-producing rods beside the core in the preform before draw.

--

/Ideal

-,

Elliptical cladding

Non symmetric stress

@

~:®

(1)8

He~l

(a)

Hd;l

8

Geometrical (2)(;;\

Stress

=0=~ ~ Lateral stress

Bend

Twist

(b)

Figure 7a. Anatomy ofa real fiber, b.Intrinsic and extrinsic mechanisms offiber birefringence. From C. P. Poole and S. Nagel, [10j, Polarization effects inlightwave systems. in Optical Fiber Telecommunications, IliA. pp. 115-161, I.P. Kaminow and T. L.Koch, Eds., Academic Press, San Diego, CA (1997).

7.3 Fabrication ofoptical fibers The earliest work on optical fibers began in the 1960s to produce multimode fiber with acceptably low loss. To then, typical optical quality glass exhibited losses on the order of 1000 dB/km. Losses <20 dB/km were deemed necessary to compete with copper coaxial transmission systems, and eventually losses one hundredth ofthis were achieved. It isclear that processing must address both loss (with impurities in the parts per billion range) and fiber structure designed to minimize dispersion. Figure 4 shows the window of transparency for high silica fiber. At short wavelengths, 1055 is bounded by ultraviolet

Chapter?

181

absorption resulting from electronic vibrations and in the infrared by molecular motion. Between these two, the floor is formed by Rayleigh scattering resulting from quenched-in density and composition inhomogeneities ofthe glass which occur atdistances on the order of the wavelength oflight. 7.3.1

Fabrication ofpreforms

Initial processing used existing technology: that ofdrawing fiber from conventional glass melts ofsoda-lime-silica orsodium borosilicate made from purified constituents. This benefited from comparatively low processing temperatures with resulting low scattering loss but suffered from contamination during processing. This traditional practice was soon replaced by vapor deposition methods. Here the inherent purity of vapor sources yielded lower loss as well as transparency to longer wavelengths. Compositions were limited by the availability of compounds having a high vapor pressure and resulted in glasses ofhigh silica composition. By their nature these compositions require drawing from preforms at temperatures above 2000°C. Such fibers have inhomogeneities quenched at higher temperature and have higher scattering loss than conventional glass, but overall loss is lower and dispersion can be better controlled. What has emerged from this work is a high silica fiber of 125 IJm 0.0. The outer jacket or cladding is generally fused silica and the core is germanium-doped silica. In the single mode fiber, the core is nominally 6 IJm in diameter and the doping concentration is about 6 wt.% Ge02. Graded multimode fiber has a core of 50 or 62.5 IJm diameter and dopant concentrations up to 15 wt.% . Today ultrapure silica fibers are used for optical communications, but multicomponent fiber is used for medical and other applications. Alkali oxides, an essential ingredient of common glasses, are rigorously excluded from high silica fibers because they cause devitrification at concentrations as low as 10 ppm. The refractive index of a glass, which is seen to rise with density and core indices, can be raised by appropriate substitutions, i.e., Ge02 for Si02. The differences between thermal expansion coefficients and transition temperatures make combination of these very different glasses in one fiber or splicing of different fibers impossible. 7.3.2 Double crucible method The first semi-commercial process to produce optical fiber was called the double crucible technique. This proceeded along the lines of conventional glass melting but used specially purified constituents [13-14]. Soda-lime-silicate and sodium borosilicate glasses were made from materials purified to parts per billion levels oftransition metal impurities byion exchange, electrolysis, recrystallization, orsolvent extraction. Starting glasses were melted, fined, drawn as cane and fed into an ingenious continuous casting system composed of concentric platinum crucibles, shown in Figure 8. A thin stream of core glass flowed from the upper crucible, passed through the reservoir of cladding glass and was concentrically surrounded by the cladding as it flowed through the orifice of the lower crucible. The time and temperature of contact between molten glass and the cladding reservoir were controlled toenable diffusion to produce the index-graded preform needed tominimize intermodal dispersion ofthe resulting fiber.

182

Chapter 7

Despite its elegance, formidable problems beset this technique from the beginning. Contamination during processing raised the impurity level from ppb levels in the starting materials to ppm levels in the fiber. Attempts made to eliminate contamination failed, but a partial solution was found by adjusting the oxygen partial pressure of the processing atmosphere tocontrol redox equilibria in the molten glass. Absorption by iron and copper, the major contaminants, could be reduced byaltering the valence state ofthe ions. Oxidizing iron to Fe3• while retaining copper as Cu' minimized absorptions by Fe 2• and Cu 2• at near IR wavelengths. Multicomponent glass made by the double crucible technique gave way to high silica glasses produced by vapor deposition methods which became available. Such glasses exhibited lower loss at wavelengths from the visible into the infrared. Several techniques appeared in the early 1970s, categorized as inside or outside processes. Both use the oxidation of silicon tetrachloride vapor to produce submicron amorphous Si02 particles. Dopants are incorporated using other chloride vapors, such as germanium tetrachloride or phosphorus oxychloride. Outside deposition is performed on a rotating mandrel by flame hydrolysis whereby chloride vapors pass through a propane- or hydrogen-oxygen flame to produce fine glass particles or "soot". The particles partially sinter as they are deposited on the mandrel. Inside processes use the same reactants with oxygen; however the reaction now occurs inside aheated fused silica tube. This serves to protect the deposit from contamination by the environment, including the hydrogen from the torch used toheat the tube, enabling the halide vapors to react only with oxygen. The hydrogen-oxygen torch traverses along the length of the tube, which is rotated on a glassworking lathe. Particles are produced by oxidation, rather than hydrolysis, and deposit on the inside wall ofthe tube downstream ofthe torch where they sinter toform a vitreous layer as the torch moves past them.

Core glass

Inner crucible

Figure 8. Double crucible technique. Core glass flows through cladding where exchange occurs producing graded index fiber. From Figure 7, J. B. MacChesney and D. J. DiGiovanni [15], Materials Technology of Optical Fibers, Materials Science and Technology 19, VCH Verlagsgesell-schaft, FRG (1990).

Chapter?

183

7.3.3 Outside vapor deposition (OVO and VAO) Two versions of outside processes have been developed. These are Outside Vapor Deposition (OVO) [16] developed by Corning, and Vertical Axial Deposition (VAD) [17] developed by a consortium of Japanese cable makers and Nippon Telephone and Telegraph Corporation. In the initial version of OVO, soot is deposited layer by layer on a horizontal, rotating mandrel at sufficiently high temperatures to partially sinter the particles and form a porous coating. The GeOrSiOzcore composition isdeposited first, followed bySiOz cladding. Atthe end ofdeposition, the mandrel isremoved and the tube thus formed issintered at 15001600°C to vitreous silica in a furnace having an atmosphere of He, Oz, and Cb. The central hole iscollapsed either during sintering orsubsequently as the preform isdrawn into fiber. Figure 9 illustrates how OVD is practiced today. A consolidated glass core rod consisting of a vapor deposited core and primary cladding is mounted in a lathe and jacketing isdeposited on it. As preform sizes become large (rumored to be 30 Kg) deposition rates can become quite large resulting inlow cost processing.

-'1

02 + Metal Halid e vapors

I

, - ----'=-

Burner

L Soot bou le Ie ====-~-=---~=v=:

o I I I I

I I I I

(a) Soot depos ition

t

: I n I I

Glass co re rod

[

]

t r., I

I

Core rod

(b) Soot preform cross-section

i

l ...._.. .

Figure 9. drawing.

.._ _. .

_

(d) Fiber

(c). Preform slnterinq

-

------

arawlng

- - --

-_.

avo process: (a) soot deposition, (b) soot preform cross section, (c) preform sintering, and (d) fiber

The VAD process also forms a cylindrical soot body, but deposition occurs end-on as shown in Figure 10. Here a porous soot cylinder isformed without a hole bydepositing the core and cladding simultaneously using two torches. When complete, the body is sintered under

Chapter 7

184

conditions similar to those used for avo. In comparison, while the composition profile of the avo preform is determined by changing the composition of each layer, the VAO profile depends upon subtle control of the gaseous constituents in the flame and the shape and temperature distribution across the face ofthe growing soot boule. Critical tothe development ofVAO was the design ofa torch composed ofup to ten concentric silica tubes. Typically, reactant vapors pass through one ormore ofthe central passages and are protected from premature reaction by a ring of inert "shield" gas positioned between the reactants and an outer series of tubes which provide hydrogen and oxygen to the flame. Manipulation of gas flows facilitates the control of temperature and particle distributions in the flame and provides a suitable surface temperature profile and boule shape. Currently VAO is used tomake single mode fiber but was initially developed for multimode manufacture.

I I

Claddinq~

torcn

~

o

(a)

[

~]

Core

j

I I

I

Clad

(b\ Soot preform cross-section

- -

(c),Preform slntenng

(d\ Fiper

tlr'awlng

-_.. _---- - - -

Figure 10. VAD process: (a) schematic of deposition apparatus. (b) soot preform cross section. (c) preform sintering, and (d) fiber drawing.

The initial challenge to VAO was how to create an optimized index profile to minimize intermode dispersion of such fiber. At that time, it was thought that control of the Ge02 distribution across the boule would require several reactant streams, each of different GeCI 4 content. However, it was found that grading could be accomplished by control of the boule surface temperature distribution, and process development focused critically upon the shape of the growth face and the temperature profile across it. Figure 11 shows Ge02incorporation

185

Chapter?

in silica as a function of the boule end-face temperature. Below 400°C, Ge02 is lost by vaporization ofdiscrete crystalline particles during boule sintering.

Hexaoonal cry~talline torm

",,~

<,

-,

o

800

Figure 11 . Graph showing stable existence of Ge02 in soot. Conditions to the right of solid line will retain Ge02 during consolidation.

Present practice istoproduce acore rod using both acore and cladding torch as shown. This core rod consists ofa Ge02 step index core with cladding whose diameter is 4-5 times that of the core. This structure is consolidated in a chlorine-containing atmosphere to reduce OH' concentration toless than 1ppm. This is then treated in one oftwo ways. Itcan be stretched and cut to provide core rods upon which jacketing soot is deposited to make a number of preforms yielding about 100 Km of fiber each. Or, it can be overclad to yield 500-1000 Km preforms. The rationale of the latter is that the deposition rate and efficiency scales with the surface area ofthe core. 7.3.4 Modified chemical vapor deposition (MCVD) Inside processes such as modified chemical vapor deposition (MCVD) had a different origin. Chemical vapor deposition (CVD) had long been used in the electronics industry for fabrication of silicon devices and was adapted to produce silica layers inside substrate tubes [18). In CVD, the concentration of reactants is very low to inhibit gas phase reaction in favor of a heterogeneous wall reaction which produced a vitreous, particle-free deposit on the substrate. This is fine for the 1000 Afilms required for semiconductor processing, but fails to produce thick deposits required for fiber. CVD was therefore reversed, the reactant concentration was increased and large volumes of particles were produced inside the silica substrate tube. They deposited on the tube wall and were sintered to glass.

186

Chapter?

MCVD was thus developed [19] as the process shown in Figure 12. High purity gas mixtures are injected into a rotating tube mounted in a glass working lathe and heated bya traversing oxy-hydrogen torch. Homogeneous gas phase reaction occurs in the hot zone created bythe torch toproduce amorphous particles, which deposit downstream of the hot zone. Heat from the moving torch sinters this deposit toa pure glass layer. Torch temperatures are sufficiently high to sinter the deposited material, but not so high as to deform the substrate lube. The torch istraversed repeatedly tobuild up, layer by layer, the core orcladding. The composition of individual layers can be varied between traversals to build the desired fiber index structure. Typically 30-100 layers are deposited to make either single mode or graded index mullimode fiber. After the initial demonstration of feasibility, fundamental investigations established the knowledge required tocreate acommercial process, l.e., tobetter understand the chemistry of MCVD required tocontrol the incorporation ofGe02 and limit hydroxyl impurities. To increase deposition efficiency, it was necessary to understand the mechanism by which particles deposit on the substrate tube and sinter to pore-free glass. Although process development preceded quantitative understanding, this knowledge formed the basis for optimization of the commercial process.

""-

Heat source 2. Deposition

3. Collapse

""-

Heat source 4. Fiber drawing

'" ,

Heat source

Figure 12. Schematic of MCVD process.

-

Fiber

Chapter?

187

(a) Chemical equilibria

The chemistry of SiCI4 and GeCI4 oxidation was investigated by infrared spectroscopy (20). Samples of effluent gases from typical MCVD reactions showed that, as the maximum hot zone temperature reached 1300oK, SiCI4 begins to react forming SbOCls(see Figure 13). The amount of oxychloride increases to a maximum at 1423°K. At higher temperatures the SbOCls, SiCI4 and POCb content decrease. Their concentrations are insignificant above 1650°K. Atthis temperature, all are converted tooxides.

10 • SiCI4 o Gecl4 .. si2 0 c is • POCls

I1l Q.

""c:

10-1

10-2

10' S '-II-''--'--'--'--'-'-.I.-.I--'--'--'-...L..-'---'---' 1000 1200 1400 1600 1800 2000 2200

T, K

Figure 13. Partial pressure ofMCVD reactants V5. temperature.

The behavior of GeCI4 is different. Its concentration in the effluent gas stream decreases between 15000K and 1700oK, but above 17000K remains approximately 50 percent of its original value. Most of the initial species escape unreacted in the effluent, indicating that at low temperatures (T<16000K) the reaction of SiCI4, GeCI4, and POCb is controlled byreaction kinetics, while at higher temperatures thermodynamic equilibria become dominant. Rate studies show that the residence times above 17000K are sufficient to produce equilibrium. SiCI4and GeCI4concentrations athigh temperatures are controlled bythe equilibria:

SiCIAg)+o2(g)~ Si0 2(s)+2CI2(g)

(24)

and GeCIAg)+o2(g)~ Ge02(s)+ 2CIAg)

(25)

Equilibrium constants for these reactions may be written

(aSioJ (po} (PSiCiJ) (po])

(26)

Chapter?

188

(27) where A are the partial pressures of gaseous species and 8i represent the activities of the solid species. The activities can be approximated by yiXi, where Xi is the mole fraction of the particular species in the solid, and Yi is the activity coefficient. An activity coefficient of unity implies an ideal solution obeying Raoult's law. The equilibrium constants for these reactions have been determined as a function of temperature and indicate that Equation 24 strongly favors the formation of Si02 at high temperature, as verified by the experiments described above. Oxidation of GeCI4 by Equation 25, on the other hand, is incomplete since the equilibrium constant K (Ge02) approaches unity at temperatures above 1600°K. This means that only a fraction of the germanium starting composition will be present as Ge02 and decreases with increasing temperature and/or chlorine content. A significant Cilconcentration results from the complete oxidation of SiCI 4 and shifts the equilibrium toward GeCI4 by the law ofmass action. Low oxygen partial pressure has the same effect.

10-5

10-6

MCVD collapse } conditions

10-7 J:

0

en o

10-8

MCVDCI2{ collapse

10-9 10-10

MCVD deposition } & consolidation Soot con} solidation

10 ppm H20 in the starting gas

Figure 14. Hydroxyl ion concentration asa function ofoxygen and chlorine partialpressures. From S. R. Nagel, J. B. MacChesney and K. L.Walker [22], Modified chemical vapor deposition, Optical Fiber Communications, Tingye Li, Editor, Academic Press, Orlando, FL(1985).

A second important aspect of MCVD chemistry is the elimination of impurity OH' (21), since reduction of OH' to sub-ppb levels is essential for realization of low attenuation in the 1.3-1.5 urn region. Hydroxyl species originate from three sources: diffusion ofOH' from the substrate tube during processing, impurities in the starting reagents including carrier oxygen gas, and contamination from leaks in the chemical delivery system. The OH' level in the fiber is controlled during deposition by the reaction

Chapter?

H}O+Cl}

189

1

(28)

~2HCI+-O}

2

with equilibrium constant K OH =

Y(POl ~ (PHzO) (PClJ

(PHCI

(29)

The amount of OH- incorporated into the glass, COH, isdescribed by

COH

=

(PH lO'nilial )(PC/l ~

(W

(30)

POlY"'

During MCVD, Cb is typically present in the range 3-10% due to oxidation of the chloride reactants. This is sufficient to reduce OH' by a factor of about 4000. However, chlorine is typically not present during collapse and significant amounts of OH- can be incorporated by diffusion of torch byproducts through the silica tube. Figure 14 shows the dependence of the SiOH concentration in the resultant glass as a function of typical partial pressures of oxygen and chlorine gas used during MCVD deposition and collapse with 10 ppm H20 in the starting gas. The figure also shows typical contamination ofthe VAD and OVD soot processes.

(b) Thermophoretic deposition and sintering The Si02particles produced by the vapor phase reaction have diameters in the range 0.02-0.1 urn and are thus entrained in the gas flow. Without the imposition of a temperature gradient they would remain in the gas stream and be exhausted. However, the MCVD design results in temperature gradients to give rise to the phenomenon of thermophoresis [23]. Here, particles residing in a thermal gradient are bombarded byenergetic gas molecules from the hot region and less energetic molecules from the cool region. A net momentum transfer forces the particles toward the cooler region. Within an MCVD substrate tube, since the wall is cooler than the center of the gas downstream of the torch, particles are driven toward the wall where they deposit. The MCVD process is shown schematically in Figure 15 interms of: (1) heat transfer inthe hot zone, (2) reaction, (3) particle formation , (4) particle deposition beyond the hot zone where the tube wall becomes cool relative to the gas stream and (5) consolidation of previously deposited particles inthe hot zone as the torch passes tothe right. A mathematical model for thermophoretic deposition (24), experimentally verified, concluded that deposition efficiency (ratio of Si02 equivalent entering tube to that contained in the exhaust) may be expressed as e=0.8(1-T.tT,.n) where T,.n is the gas reaction temperature and T. is the temperature downstream of the torch at which the gas and the tube wall equilibrate and deposition ceases. Typically, T. is about 400°C and T,.n about 2000°C, giving an efficiency on the order of 60%. Consolidation of the soot layer formed byMCVD was determined tooccur by viscous sintering (25). By this mechanism, consolidation is controlled by the rate of viscous flow of the glass, which is a strong function of glass composition and decreases exponentially with increasing temperature. The driving force is a reduction in surface energy achieved by decreasing surface area. Sintering of soot layers approximately 100 urn thick occurs rapidly as the oxy-

Chapter?

190

hydrogen torch passes over the soot deposit. (Thermal gradients along the tube wall are typically around 300°C/em.) The torch traverse rate is adjusted to allow the soot to sinter without trapping gas toform bubbles: normally about 10 em/min.

- - - I

Quartz substrate 02

SiCI4

Thermopho retic deposit

POCI3 GeCI4

He

~

I I

Particle trajectory Substrate tube

Consolidated deposit Traverse

..

I

I Figure 15. Schematic diagram depicting deposit ofparticles generated inMCVD.

Initially, the formation and growth ofnecks occurs between adjoining particles due tothe small radius of curvature. Bridges and chains are formed as the glass begins to consolidate. The initially open pores become trapped pockets ofgas. As sintering proceeds, these voids shrink in size but remain irregular in shape, indicating that viscous flow of glass is slower than the diffusion of gas out of the void. However, if the temperature becomes too high, the vapor pressure of the most volatile species in the glass may become greater than the pressure induced by surface tension, causing a bubble to form. This can occur during sintering the layer or when the tube is collapsed. Bubble formation is determined by initial pore size, temperature, gas composition, and the diffusion rate ofgas from the void. 7.3.5 Plasma chemical vapor deposition A second inside process, plasma chemical vapor deposition (PCVD) (26), is similar to MCVD. It uses the same reactants inside a silica tube, which is collapsed after deposition and drawn to fiber, but reaction inside the tube is initiated by a non-isothermal microwave plasma traversing down the tube, rather than by heating the exterior of the tube, as shown in Figure 16. The plasma requires a pressure of only a few torr and is generated by a microwave cavity operating at 2.45 GHz. A particulate soot is not produced but instead, deposition occurs directly on the tube wall in the form of a thin glass layer. Reaction and deposition of both Ge02 and Si02 are much more efficient than in MCVD, approaching 100%. Another advantage, especially for multimode preforms, is that since the plasma involves no latent heat,

Chapter?

191

it can be traversed very rapidly to produce hundreds of layers. The resulting deposit thus has a very smooth and precise index profile essential for minimizing intermodal dispersion.

2.45 GHz

microwave cavity

1200°C

furnace

Reactor

Plasma

'II

Traverse

.,

Figure 16. Diagram illustrating operation ofPCVD reactor.

7.4 Fiber drawing process The final step in the fabrication of optical fiber is drawing the preform [27]. Typical preforms produced by the above methods are about a meter in length and between 2 and 15 cm in diameter. These are drawn into standard 125 urn diameter fiber by holding the preform vertically and heating the end above the glass softening temperature until a gob of glass falls. Thisforms a neck-down region which provides transition toa small diameter filament. Uniform traction on this filament results in a continuous length of fiber up to several hundred kilometers. Before this fiber contacts a solid surface, a polymer coating is applied to protect the fiber from abrasion and preserve the intrinsic strength of the pristine silica. The fiber is then wound on a drum. Fiber can now be drawn without inducing excess loss while maintaining high strength and dimensional precision and uniformity. 7.4.1

The drawing tower

The essential components of a draw tower (Figure 17) are a preform feed mechanism, a furnace capable of 1950-2200°C, a diameter monitor, a polymer coating applicator, a coating curing unit, a traction capstan and a take-up unit. The furnace is typically a graphite resistance type or an inductively-coupled radio frequency zirconia furnace [28]. The former requires an inert atmosphere to prevent oxidation of the graphite element. The zirconia furnace may be operated in air but must be held above 1600°C even when not in use, since the volume change associated with the cubic to tetragonal crystallographic transition causes cracking ofthe susceptor, producing Zr02 particles and resulting infiber breaks. Maintaining uniformity infiber dimensions iscritical toperformance. Most obvious is the effect on splicing of fibers. Here significant loss occurs when eccentricity of the cores exceeds a

192

Chapter?

fraction of a micron. Further minor variation of the index profile causes increased dispersion and coupling between guided modes and cladding modes which increase loss. To attain required dimensions, optical sensors are placed below the furnace. High-speed, noncontacting, diameter measurement operates by feedback to the capstan controlling the draw [29] and maintains the diameter to0.1 urn. Draw conditions affect fiber performance beyond diameter control. Fiber loss occurs with inappropriate draw tension. Generally, moderately high tension (approximately 100 gms on the fiber) isused. This permits draw ata relatively low temperature (""2100°C) atnormal draw speeds. High temperature (low tension) creates Ge-O defects and other color centers. Too high tension resulting from too low temperature can decrease fiber strength and lower the fiber'score !l when the core bears the strain ofthe draw. 7.4.2 Protective fiber coatings Despite the high strength of optical silica glass fibers «10 GPa) and their extremely smooth surface upon exiting the furnace, they are vulnerable to damage from the environment. Due to the very low toughness of glass (less than 1 MPalm l12) , even the tiniest nick in its surface will produce adequate stress concentration toinitiate fracture under service conditions. Direct contact with the capstan or winding drum, where fiber is accumulated must be avoided. It would cause surface abrasion. Polymeric coatings are therefore applied immediately below the diameter monitor after the fiber has been able to cool. Coatings are applied by draWing the fiber through a cup of liquid polymer, then through a UV curing chamber or curing oven. Coatings are typically composed of urethane acrylates (UV cure) or silicone rubbers (thermal cure). A coating diameter monitor is used to keep the overall diameters at 250 urn. The coating must be free ofparticles and bubbles, and must be concentrictothe fiber. Another failure mechanism isdue towater, which isknown to cause fatigue. This causes long term failure by attacking free surfaces of glass at points of high stress, resulting in stress corrosion cracking [30]. Since polymeric coatings are permeated by moisture, they provide no long term protection. Hermetic coatings have been developed toprotect the surface. Metal or carbon coatings [311 are applied during the drawing process via vapor deposition. However, cost and adequate performance of polymer coated fibers has stalled wide use of the added protective coating. Fiber coatings not only provide protection from surface abrasion, but also provide additional benefits. Low modulus coatings reduce stresses at bends, thereby reducing bending losses. High modulus polymer coatings also provide strength, reducing the load on the fiber and thus reducing stress-induced loss or birefringence. Compound coatings, consisting of a soft inner layer and a strong outer layer, may be used for more critical applications. While the optical properties of fiber are typically the topic of primary concern, the strength and resistance to environmental hazards are of critical importance from the perspective of long term reliability. If afiber breaks, its optical performance is irrelevant. Advancing draw technology developed during the two decades offiber existence has resolved most ofthe problems associated with draw. Now attention isturned toward speed. During the first decade ofmanufacture, small preforms «100 Km offiber length) could be drawn ata few meters/second. Currently large preforms (up to 1000 Km offiber length) require higher rates 10-30 m/s to be competitive. Such rates require high draw towers (about 20 m) to allow adequate cooling of the fiber before entering the polymer cup. Redesign of the coating

Chapter?

193

system was also required to prevent cavitation and coating eccentricity. These improvements have been implemented toaccommodate draw speeds nearing 60 miles per hour. 7.5 Sol·Gel processing

Sol-gel is a processing method whose origins date back more than a century. It has been rediscovered during the last 15 years and has sparked intense interest by the world's leading laboratories. The goal is tostart from a liquid precursor and to engineer the properties of the eventual solid phase. The preferred pursuit ofthis goal involves alkoxide precursor materials. Typically, tetraethylorthosilicate, Si(OC2Hs)4, is hydrolyzed with water in the presence of ethanol and an acid catalyst. In the resulting sol, the particle size of the silica is controlled by the hydrolysis conditions (see Chapter 6). The sol is cast into molds orformed into films and fibers, dried and consolidated by viscous sintering to glass. In an alternate process, colloidal silica is dispersed in water and formed into desired shapes by a number of mechanical compaction or ceramic forming processes. Ofthese, only isostatic compaction [32) and gel casting [33) have been pursued to demonstrate large cylinders suitable for fiber production. The alkoxide approach suffers because the volumetric fraction of solids in the sol is low. Thus, shrinkage during drying is high, limiting the thickness of films formed from it or the size of monolithic bodies which can be made. Nevertheless, there has been considerable effort to produce all-gel preforms by adding up-dopants (dopants which increase the index) to the precursor alkoxides forming the core glass. Additions of germania precursors such as Ge(OC2Hs)4 have rarely produced up-doped silica and never fiber equivalent to that produced byconventional means. The only encouraging result is the production of a silica core, down-doped cladding prefom [34). The hydrolysis/polycondensation ofSi(OC2Hs)JF produces a gel which, when dried, yields a porous body having a high surface area (200-650 m2/g) . When a tube of this material is sintered in a fluorine containing atmosphere the fluorine incorporated into the glass reduces the index (L'. = -0.62%). Collapse of the tube in a stream of oxygen flowing in the center produces a preform with a silica core which gives fiber whose loss isas low as 0.4 dB/km. The more viable approach is to use gel-silica to form overcladding, which comprises 90% or more of the eventual fiber. A process is known [35) that facilitates the production of overcladding tubes large enough for 150 km preforms when used to overclad MCVD preforms. This yields fiber meeting commercial requirements (loss below 0.2 dB/km at 1.55 11m). In this process, Aerosil OX-50 (Degussa AG) is dispersed in water and the sol is stabilized byadding quaternary ammonium hydroxide, centrifuged toremove particulates, and de-aired. An ester is added to reduce the pH and the sol is poured into a mold where it gels. After aging, the tube formed in the mold is transferred to rollers upon which the tube is dried by rolling in a chamber of controlled temperature and humidity. Finally, the tube is purified in achlorine-containing atmosphere and sintered to glass. Overclad preforms are designed to reduce the evanescent field of light propagating in a fiber by 104-10s at the overclad/core-rod boundary. Still, transition metal ions, hydroxyl ions, and refractory particles must be removed. The former is accomplished byequilibrating the porous dried bodies in chlorine at temperatures in the vicinity of 900°C. More severe treatment is needed to remove refractory oxide particles present in the starting material or introduced during processing. Most of these particles are two to three times denser than silica and are

Chapter?

194

removed by centrifugation. However, a number of smaller particles remain and must be removed if low strength fiber breaks are to be avoided. In fact, to produce commercially acceptable fiber they must be removed to one part in 1015, a detection level unachievable by anything other than examination of fiber breaks to identify the particle which caused it. Particle removal is performed with thionyl chloride [361 to react with the oxides such as Zr02, for example:

Zr0 2(S )+sOCIAg)~ ZrOCI 2(g)+S02(g)

(31)

Zirconia and chromia particles are removed and a means of producing silica tubes satisfactory for commercial quality fiber is demonstrated.

1. Main frame (-12-25M tall) 2. Preform feed, support, and centering mechanism 3. Furnace 4. Diameter monitor 5. Coating applicator, reservoir, and centering mechanism 6. Coating concentricity monitor 7. Coating curing station 8. Fiber pulling capstan 9. Fiber take-up synchronizer 10. Winding mechanism 11. Control console

0------, 3 o

o

o'"

Figure 17. Components ofadraw tower.

7.6 Optical fiber devices The properties and geometries of optical fiber have been manipulated in many varied and clever ways to realize numerous devices for functions other than mere signal transmission. While some of these devices, such as fiber amplifiers or filters, are directly related to the transmission and routing ofoptical signals, others are not. An example ofthe latter issensors of mechanical or electrical conditions. This section highlights several such applications of

Chapter?

195

optical fiber, some ofwhich are still quite new. They are byno means intended to be inclusive of the entire field, which is beyond the scope of this chapter, but rather point out some interesting applications and points from which the interested reader may investigate further. 7.6.1

Optical amplifiers

The development of optical amplifiers fundamentally changed the practice and perception of optical communications. During the first two decades oftheir existence, optical fibers served a passive role, that oftransmitting optical signals to be detected and amplified electronically and rebroadcast optically. The cost of optical to electronic amplification led to research seeking longer transmission spans without amplification. Development of fluoride fibers [37] in the 1980s sought to lower optical loss by one or more orders of magnitude by extending glass transparency further into the infrared. This effort was abandoned with the discovery of the optical amplifier [38]. It not only allowed amplification without optical to electronic transition but also transmission of 16, and eventually as many as 100 channels on a single fiber. The device works this way: a section ofsingle mode fiber whose core isdoped with hundreds of ppm erbium ions is spliced to the transmission fiber. Pump light at 980 or 1480 nm is coupled to the fiber doped with Er3+ions. This light excites inner shell electrons to higher energy levels where they persist for about 10 milliseconds. When these face signal photons ofa given energy, they are triggered tofall to the ground state amplifying the signal. An important aspect of Er3+ doped glass fiber is that ligands tothese ions vary in strength. In turn, these perturb the inner 4f electron levels and result in a range of amplified wavelengths from 1530 to1580 nm. This allows a number ofclosely spaced channels to be amplified. The actual implementation of the erbium doped fiber amplifier (EDFA) is shown by Figure 18. Pump light from a semiconductor laser is multiplexed with the signal through a coupler and acts as a wavelength division multiplexer. The combined signal passes through some tens of meters of doped fiber spliced to the coupler. Here it is amplified and passes on to the transmission system through an isolator toprevent back reflections. Amplifiers in the 1.511m range [39-41] are available as commercial products. The significance of these fiber amplifiers is that they produce greater than 20 dB amplification of signals and may be pumped atwave lengths readily available from semiconductor laser sources. There is no need for optical-electronic-optical signal transformations to electronically amplify signals, for it is an all-optical process with the exception ofthe power to the pump source. In addition, by using a series offiber amplifiers with different gain spectra, typically achieved with various codopants, overall gain spectra with variations less than 1dB over a 35 nm window have been achieved [42-43]. Such performance is particularly useful for systems transmitting more than one wavelength. The consequence ofthese devices isthat many signals whose wavelengths are separated by as little as a fraction of a nanometer propagate along the same fiber. Under these circumstances, fiber non-linearities arising from stimulated scattering (Brillouin and Raman) and refractive index fluctuations affect transmission. The former are minimized by achieving zero chromatic dispersion at the operating wavelength. Refractive index fluctuations arise from self-phase modulations, cross-phase modulation, and four wave mixing. All three factors can be minimized by management ofdispersion.

196

Chapter?

¥-9

Pump .... laser

~

Input

si~1 =c~~~:::::=::;:-....-==t:~= . Pump/signal Optical multiplexer Isolator

Optical isolator

Energy level diag ram

\ 980nm 1480nm

1520 - 1570nm

Erbium amplifier gain pertormance 40 r - - - - - - - - , III

". 20 c 'iii

Cl

(ij c;

0

Cl

(jj -20

10

20

30

Pump power. mW

Figure 18. Er-doped fiber amplifier. Top f'!lure: amplifier components; middle figure: energy diagram for Er+3 ion; bottom figure: amplifier performance.

Four wave mixing (FWM) involves the mixing oftwo ormore wavelengths (traveling in a fiber) to generate new, unwanted wavelengths. Generating these new wavelengths impairs transmission by transferring power from the original optical signals or adding to the original signal where the new wavelength coincides with it. The electric fields of the generated waves interfere, degrading performance. Minimizing FWM requires a small (non-zero) dispersion in the 1,500 nm wavelength band. This is accomplished (see Figure 6) by tailoring the index profile in an appropriate way [44). 7.6.2 Fiber gratings as mirrors and filters A fiber grating is a structure generated by periodically varying the index of the fiber core. These gratings are fabricated byphoto-induced changes in the core index. When this period of the grating is half that of the wavelength of light, complete coupling into the counter propagating mode is achieved and that wavelength is reflected [45). This can be used in devices like wavelength selective mirrors [46), resonant cavities, and bandpass filters [47), and to shape optical spectra [48). Photosensitivity results from high temperature drawing of germanium doped fiber preforms. Oxygen deficient centers are present which, when irradiated with UV light, cause an increase in the core index. This effect can be increased by hydrogen doping of the fiber prior to

Chapter?

197

irradiation. The grating is created [49] when two coherent UV laser beams are directed laterally on to the fiber so as to generate a pattern on the fiber whose intensity distribution depends on the relative angle between the beams. By adjusting the relative beam angle to achieve a spacing or wavelength desired for the grating, the high intensity portions of the pattern induce alocalized increase in core index, thereby writing the grating into the fiber. One example ofthe function of Bragg gratings is the tap/combiner, which isalso known as the add/drop filter. This device is shown in Figure 19 where seven signals (11.1-11.7) enter a 3 dB coupler. They are split onto the two arms and encounter identical Bragg gratings whose pitch (spacing) isexactly half ofthe wavelength 11.4. That signal isthen reflected toTap 2. All other wavelengths exit through Port 4. Long-period gratings which are longer than Bragg gratings are chirped and primarily used as spectral filters [50]. These devices operate by selectively coupling specific wavelengths into forward propagating cladding modes. Since cladding modes are lossy (loss/scattering atouter cladding surface), they are a wavelength selective loss element. Again, the grating period determines the wavelength removed with high selectivity. Such devices may be fabricated not only to remove various wavelengths, but also to adjust the strength of a desired wavelength, i.e., they may be tuned to be more or less lossy. By combining a number of precisely designed filters, a spectrum may be shaped orsmoothed by reducing the intensity ofselected wavelengths. These devices have been used to produce flat, broad optical spectra needed for fiber amplifiers, ortoremove amplified spontaneous emission (ASE) from chains of erbium doped fiber amplifiers. 7.6.3 Strain sensors and other applications There are a few different ways to detect strain with optical fiber, e.g., an interferometer in which only one arm is strained and produces a change in path length between the two. Another, perhaps simpler method involves a fiber Bragg grating. The wavelength reflected by a fiber Bragg grating isextremely sensitive tothe grating wavelength. Thus, when strain alters the grating wavelength, the degree of strain will be indicated by the wavelength of light reflected by the grating [51]. The range of strain which can be measured is up to about 36,000 microstrain. These systems are also capable of measuring dynamic strains at frequencies over 100 Hz. Due to the inherent immunity of fiber optics to electromagnetic interference (EMI) and their durability, such systems may find use in numerous noisy or hazardous environments, including high radiation environments. One problem with measuring electric phenomena is electromagnetic interference (EMI) induced in the measurement equipment in certain surroundings. Fiber optic systems are generally immune to such problems; however, special fiber devices may be fabricated to probe such phenomena. One recently demonstrated system involves the interference of two counter propagating Brillouin fiber lasers in a coil of fiber around a conductor carrying current [52]. The Faraday effect from the magnetic field around the conductor induces different frequency shifts in each laser, dependent on the wave direction. When the counter propagating lasers interfere, the beat frequency between them is a measure of the magnetic field , and hence the current in the conductor. This device was shown to be linear for currents from 30A to 200A. With appropriate modifications, such a device could be used to measure either AC orDC current.

198

EOn

Chapter?

!II

-c

..:-25

~

·50

1

4 5

7

UVtrimming 3dBfused ~ 3dBfused

Input-l~3

Tap-2~4-0utPut Identical Bragg gratings resonanf at "'4

Figure 19. Bragg-Grating-based tap/combiner (add/drop filter). Input signals (M-M) are coupled to Mach-Zender interferometers with Bragg grating. M isreflected 10 tap 2. All others exit through port 4.

A sensor has been fabricated using a membrane, permeable tothe species ofinterest, to form a cell inwhich a reference solution was contained and the end of a fiber bundle immersed init [53). The solution contained species becomes chemiluminescent when it reacts with the species to be detected. Light emitted by this reaction is collected by the fiber bundle and analyzed on a spectrometer. By immersing the cell into fluid to be sampled, the species of interest is transported across the membrane and its concentration in the sample fluid may be determined. This device is claimed to be capable of measuring concentrations as low as about 111m, with a dynamic range of5-40 11m. Sampling times are claimed tobe on the order of 10 s. A recent fiber optic device makes use of a piezoelectric layer on the fiber cladding to impose strains on the fiber, thereby altering the index through the piezooptic effect to create a phase modulation dependent on the strain [54]. This device produces phase shifts of 1.559 11m light on the order of 0.1 rad for devices 2 mm long, driven at frequencies between 20 and 1000 MHz with less than 100 mW power. Similar devices have been fabricated using fiber Bragg gratings, whose transmission window may be shifted with such piezooptic effects, or by resistive coatings used toheat the grating [55).

7.7 Summary and outlook The evolution ofoptical fibers started with pure silica glass fibers as the transmission medium for light communication. First, the light absorbing transition metal impurities were reduced to

Chapter 7

199

the part per billion level. Then, hundreds of parts per million of Er3+ ions were incorporated into the glass tocreate optical amplifiers with a band ofwavelengths extending over about 400 nm. Within this spectral region, many optical channels could be multiplexed, allowing one fiber to do the work of many. In turn this created a need to shift the zero in chromatic dispersion to this spectral region and at the same time to introduce controlled and low dispersion tocounter nonlinear effects observed in the long unrepeated spans now used. Finally, intrinsic defects at the germania sites of the silica lattice were found to confer photosensitivity tothe glass when irradiated with short wave length (UV) light. These allowed narrowly spaced lines to be written across the fiber core (Bragg gratings) to reflect specific wave lengths from a number of wavelengths comprising telephone and data traffic, thus allowing separation and distribution ofthese signals. Optical fibers are expected to reach approximately two billion dollars by turn of the century. Growth is expected to come from expansion of long distance service, construction of new CATV systems, the use of the Internet, and introduction of fiber to the home. A new system [561 is expected to transmit 400 Gbits/s of information over a single fiber by multiplexing 40 channels, each operating at 10 Gbits/s. Research work will also concentrate on devices such as amplifiers and routers to enhance performance of the wavelength division multiplexer as the low cost way toincrease capacity. REFERENCES [1) [2) [3) [4) [5] [6] [7) [8] [9]

[10J [1 1] [12J [13] [14) [15J [16] [17) [18] [19]

A. E.Wilner, Mining the optical bandwidth fora terabit per second, IEEE Spectrum, 4, 32-41 (1997). R. Cunningham, in Focus, Fiberoptic Product News, February 1997. G. Keiser, Optical Fiber Communications, McGraw-Hili Book Company, New York, NY (1983). S. Kosinski and J. B. MacChesney, Fiber Optics, in Kirk-Othmar Encyclopedia of Chemical Technology, 4th Edition 10©John Wiley & Sons, Inc. New York, NY (1994). D. Marcuse, Theory of Dielectric Optical Waveguides, Acad. Press, Inc. San Diego, CA (1974). L.B. Jeunhomme, Single-Mode Fiber Optics, Principles and Applications, Marcel Dekker, Inc., New York, NY (1983). R. D. Maurer, Fibers foroptical communication, Proceedings IEEE, 61 , 452-62 (1973). S. Nagel, Optical fibers: the expanding medium, IEEE Circuitsand Devices 3,36-45 (1989). D. Marcuse and H. M. Presby,Effects of profile deformations on fiber bandwidth, Appl. Opt. 18, 3758-3763 (1979). C. P. Poole and S. Nagel, Polarization effects inlightwave systems, in Optical Fiber Telecommunications, lilA, 115-161, I. P. Kaminow and T. L.Koch, Eds. Academic Press, San Diego, CA (1997). A. F. Judy, Improved PMD stability inoptical fibers and cables, Proc. 43rd Int. Wire and Cable Symp., 658-669 (1994). J. Noda, K. Okamoto and Y. Sasaki, Polarization-maintaining fibers and their applications, J. Lightwave Technology, Vol. LT-4, No.8, 1071-1089 (1986). A. D. Pearson and W. G. French, Low-loss glass fibers foroptical transmission, Bell Labs Rec., 50, 103-106 (1972). K. J. Beals and C. R. Day, A review of glass fibers foroptical communication, Phys. Chem. Glass, 21 , 5-19 (1980). J. B. MacChesney and D. J. DiGiovanni, Materials Technology of Optical Fibers, Materials Science and Technology, 19,VCH Verlagsgellschaft, FRG (1990). D. B. Keck, P. C. Schultz and F. Zimar, Method of forming optical waveguide fibers, U.S. Patent No. 3,737,292 (1973). T. Izawa and N. Inagaki, Materials and processes forfiber preform fabrication-vapor phase final deposition, Proc. IEEE, 68, [10]1184-87 (1980). W. G. French, J. B. MacChesney, P. B. O'Connor and G. W. Tasker, Fabrication of graded index and single mode fibers with silica cores, Bell System Technical Joumal, 53, 951 (1974). J. B.MacChesney, P. G. O'Connor, F. V. DiMarcello, J. R. Simpson and P. D. Lazay, Preparation of low-loss optical fibers using simultaneous vapor-phase deposition and fusion, Proceedings of Tenth Intemational Congress on Glass, Kyoto, Japan, Vol. 6, 40-45, Ceramics Society, Japan (1974).

200

Chapter 7

[20J D. L. Wood, K. L. Walker, J. B.MacChesney, J. R. Simpson and R. Csencits, The germanium chemistry in the MCVD process foroptical fiber fabrication, J. Lightwave Technol., LT-5, 277-83 (1987). [21] K. L. Walker, J. B. MacChesney and J. R. Simpson, Reduction of hydroxyl contamination in optical fibers, Technical Digest 3'd International Conference Integrated Optics and Optical Fiber Communications, San Francisco, CA, 86-89 (1981 ). [22J S. R. Nagel, J. B. MacChesney and K. L. Walker, Modified chemical vapor deposition, in Optical Fiber Communications, Tingye Li, Ed ~or, AcademicPress, Orlando, FL(1985). [23) P. G. Simpkins, S. G. Kosinski, and J. B. MacChesney, Thermophoresis: The mass transfer mechanism in modified chemical vapor deposition, J. Appl. Phys., 50, 5676-81 (1979). [24) K. L. Walker, F. T. Geyling and S. R. Nagel, Thermophoretic deposition of small particles in the modified chemical vapor deposition process, J. Am. Ceram. Soc., 63, 96-102 (1980). [25J C. J. Brinker and G. W. Scherer, Sol-Gel Science, Academic Press, New York, (1990). [26J D. Kuppers and H. Lydtin, Preparation ofoptical waveguides with the aid of plasma-activated chemical vapor deposition atlow pressures, in Topicsin Chemistry, 89, 108-30 Springer-Verlag, Berlin(1980). [27] F. V. DiMarcello, C. R. Kurkjian, and J. C. Williams, Fiber drawing and strength properties, in OpticalFiber Communications, Vol. 1,Chapter 4, T. Li, Ed., AcademicPress, N.Y. (1985). [28] R. E.Jaeger, A. D. Pearson, J. C. Williams and H. M. Presby, 401-433, in Optical Fiber Telecommunications, S. E. Miller and A. G. Cheyfloweth, Eds. Acad. Press, NY (1979). [29] D. H. Smithgall and R. E. Frazee, Bell Syst.Tech. J. 60, 2065 (1981). [30J C. R. Kurkjian, J. T. Krause and M. J. Matthewson, Strength and fatigue of silica optical fibers, J. Lightwave Technol. 7 [9], 1360-1370 (1989). [31J F. V. DiMarcello, R. G. Huff, P. J. Lemaire and K. L. Walker, Hermetically sealed optical fibers, U.S. Patent 5,000,541 , March 19,1991 . [32] K. Yoshida, T. Satoh, N. Enomoto, T. Yagi, H. Hihara and M. Oku, Fabrication of large preforms for low loss singlemode optical fibers, Glastechnische Berichte (1 996). [33] E. A. Chandross, D. W. Johnson, Jr. and J. B. MacChesney, Vitreous silica product via solijel using polymeric additive, U.S. Patent 5,379,364 (1995). [34] S. Shibata, T. K~gawa, and M. Horiguchi, Wholly-synthesized fluorine-doped silica optical fibers bythe sol-qel method, J. Non-Crysl Sol., 100,269 (1988). [35J J. B. MacChesney, D. W. Johnson, Jr., S. Bhandarkar, M. P. Bohrer, J. W. Fleming, E. M. Monberg and D. J. Trevor, Optical Fibers Using Sol-Gel Silica Overcladding Tubes, Elec. Let., 33(1 8), 1573-74 (1997). [36J S. D. Bhandarkar, H. L. Chandan, D. W. Johnson,Jr. and J. B. MacChesney,U.S. Patent 5,344,375(1994). [37) M. Poulin, M. Chanthansin and J. Lucas, Noureaux verres fluores, Material Research Bulletin, 12 [3], 131 (1997). [38) E. Desurvire, J. R. Simpson P. C. Becker, Highijainerbium-doped traveling-wave fiberamplifier, Opt. Lett., 12 [11 ),888-890(1987). [39] G. R. Walker, N. G. Walker, R. C. Steele, M. J. Creaner and M. C. Brain, Erbium-doped fiber amplifier cascade formmulti-channel coherent optical transmission, J. LightwaveTech. 9 (2) 182-1 93(1991). [40) M. Zimgibl, An optical power equalizer based on Er-doped fiber amplifier, IEEE Trans., Photonics Tech. lett., 4 [4],357-359 (1992). [41] G. Nykolak, S. A. Kramer, J. R. Simpson, D. J. DiGiovanni, C. R. Giles and H. M. Presby, An erbium-doped multi-mode optical fiber amplifier, IEEE Trans., Photonics Tech. Lett. 3 (12) 1079-1081 (1991). [42] C. G. Atkins, J. F. Massicott, J. R. Armitage, R. Wyatt, B. J. Ainslie and S. P. Craig-Ryan, Highijain, broad spectral bandwidtherbium-doped fibre amplifierpumped near 1.5li m, Electron. Lett., 25 [14], 910-911 July 6, 1989. [43J P. F. Wysocki, N. Park and D. DiGiovanni, Dual-stage erbium-doped, erbiumlytterbium-codoped fiber amplifier withupto+25 dBm output power and a 17nm flat spectrum, Opt. Lett. 21 [21], 1744-1 746 (1996). [44] R. W. Tkach, A. R. Chraplyvy, F. Forghieri, A. H. Gnauck and R. M. Derosier, Four-photon mixing and highspeed WDM systems, Joumal ofLightwave Tech., 13, No.5, May 1995. [45] K. O. Hill, Y. Fujii, D. C. Johnson and B.S. Kawasaki, Photosensitivity in optical fiber wave guides: Application toreflection filter fabrication, Applied Physics Letters, 32, 647 (1978). [46] F. Bilodeau, K. O. Hill, B. Malo, D. C. Johnson and J. Albert, High-retum-Ioss narrowband all-fiber bandpass Bragg transmissionfilter, IEEEPhotonics Tech. Lett., 6 [1],80-82 (1994). [47) T. Erdogan and V. Mizrahi, Fiber phase gratingsreflect advances in lightwave technology, Laser Focus World, 73-80, February 1994. [48) A.M. Vengsarkar, Long-period fiber gratings shape optical spectra, Laser FocusWorld, 243-248, June 1996. [49) W. W. Morey, G. A. Ball and G. Meltz, Photoinduced Bragg gratings in optical fibers, Optics and Photonics News,8-14,February 1994. [50) A. M. Vengsarkar, J. R. Pedrazzani, J. B. Judkins, P. J. Lemaire, N. S. Berganoand C. R. Davidson, Longperiod fiberijrating-based gainequalizers, Opt.l ett.,21 [5J, 336-338 (1996).

Chapter 7

201

[51) S. M. Melle, K. Liu, R. M. Measures, Practical fiber-optic Bragg grating strain gauge system, Appl. Opt, 32 [19],3601-3609 (1993). [52] A. Kung, P. A. Nicati and P. A. Robert. Reciprocal and quasi-reciprocal Brillouin fiber-optic current sensor, IEEE Photonics Tech. Lett., 8 [12],1680-1682 (1996). [53] D. S. Blair, L. W. Burgess and A. M. Brodsky, Study ofanalyte diffusion into a silicone-clad fiber-optic chemical sensor byevanescent wave spectroscopy,Appl. Spectr., 49 (11), 1636-1645 (1995). [54] A. Gusarov, N. H. Ky, H. G. Limberger, R. P. Salathe and G. R. Fox, High-performance optical phase modulation using piezoelectric ZnO-coated standard telecommunication fiber, J. Lightwave Technol., 14 [12], 2771-2777 (1996). [55] G. R. Fox, C. A. P. Muller, N. Setter, D. M. Constantini, N. H. Ky and H. G. Limberger, Wavelength tunable fiber Bragg grating devices based on sputter deposited resistive and piezoelectric coatings, J. Vac. Sci. Technol. A,15[3), (1997). [56] S. Hardy, Lucent unveils terabit DWDM system, Lightwave , March 1998.

SECTION IV FIBERS FROM SOLID PRECURSOR FIBERS R Naslain Commercial boron/tungsten fibers are derived directly from the vapor phase. Commercial silicate glass fibers and most commercial silica glass fibers are derived from their melts. but some silica fibers, asdiscussed in Chapter 5, can be derived from viscous aqueous solutions. Glass fibers are therefore derived directly from a liquid phase. Commercial ceramic and carbon fibers are produced from solid precursor, or green, fibers which, in turn, are derived from a melt, dispersion, orviscous solution. This section of the book deals with fibers which are derived from solid precursor fibers.

Contents 8

CERAMIC OXIDE FIBERS FROM SOL-GELS AND SLURRIES 8.1 General considerations 8.2 Alumin a and alumina based fibers 8.3 Zirconia based fibers 8.4 Yttriumaluminum garnet (YAG) fibers

11

SILICON NITRIDE AND BORIDE BASED FIBERS 11 .1 General considerations 11.2 Si-C-N-O and Si-C-N fibers 11.3 Si-N-Oand Si-N fibers 11,4 Si-B-O-N, Si-B-Nand Si-B-N-C fibers 11 .5 Applications

9

CARBON FIBERS FROM PAN AND PITCH

12

APPLICATIONS OFCARBON AND CERAMIC FIBERS 12.1 Fiber applications 12.2 Composite applications

9.1 9.3 9.4 9.5 10

General considerations 9.2 Processing ofcarbon fibers Structure ofcarbon fibers Properties ofcarbon fibers Applications

SILICON CARBIDE AND OXYCARBIDE FIBERS 10.1 Generalconsiderations 10.2 Preparation ofSi-C-O fibers 10.3 Preparation ofoxygen-free Si-C fibers 10.4 Preparation ofquasi-stoichiometric SiC fibers 10.5 Structureofsilicon carbide fibers 10.6 Thermal stability ofsilicon fibers 10.7 Mechanical properties ofSiC fibers 10,8 Oxidation ofsilicon carbide fibers 10.9 Transport properties ofSiC fibers 10,10 Applications

CHAPTERS CERAMIC OXIDE FIBERS FROM SOL-GELS AND SLURRIES R. Naslain Ceramic aluminate and zirconate fibers have higher melting temperatures, moduli, service temperatures, corrosion resistance, and lower dielectric constants and strength than amorphous silica and silicate glass fibers as discussed inChapters 5, 6 and 7. 8.1 General considerations Table 1. Properties of selected refractory oxides Oxides SiD, Alp, 3Al,O,.2SiO, (Mullite) ZrO, Y,O, Y,Al,O,,(yAG)

Mp,oC 1710 2050 1850 2700 2410 1940

p,g/em' 2.2-2.6 3.96 3.2 5.8-6.1 5.0 4.55

E,GPa 72 430-460 230 • 200

IX,lO_60C' 0.5 7.5-8.0 5.1-5.6 8-11

283

Continuous ceramic oxide fibers have low melt viscosities and very high crystallization propensities, and therefore cannot be readily obtained directly from a melt. Solid green or precursor fibers are obtained from solutions or dispersions, respectively, and they are converted into ceramic oxide fibers (Table I). In contrast, continuous silica glass fibers can be dry spun directly from a high viscosity melt orindirectly from a viscous sol-gel (Chapter 5). 8.1 .1 The generic sol-gel process Sol-gel is a generic term covering processing routes which differ from one another mainly by the nature of the starting chemicals, the most commonly used being sols or solutions of organometallic species such as alkoxides. Gelation of a sol (or of such solutions) gives a viscous product which can be shaped as fibers byextrusion and mechanical stretching. After drying and calcining, a sintering treatment at relatively low temperatures yields ceramic fibers with a very fine microstructure. The fundamentals of the sol-gel routes have been the SUbject ofmany review articles [1-7) and will be briefly recalled here for a better comprehension ofthe next sections.

(a) The starting materials A colloidal solution orsol is a suspension of nanometer size particles in a liquid. Sols can be formed by dispersing ultrafine particles in a liquid or by precipitation of fine particles from a solution followed by peptization. The stability of the sol can simply result from Brownian

206

Chapter 8

motion when the particles are small enough. However, the stabilization of the sol, l.e., peptization, is usually achieved through the addition of an appropriate electrolyte such as a mineral acid. Under such conditions, solvated protons are adsorbed on the particle surface, generating a net positive surface charge which is balanced by the negative charges of the anions from the acid, all within a region referred to as the electrical double layer. The electrostatic interactions preclude interparticle contact, resulting ina stable sol (2) [5]. Sols are available for the most common oxides and their equivalent oxide concentrations can be very high (e.g., 300-400 gIl). The term sol-gel route is still used when the starting chemicals are organometallic species such as alkoxides (9), even though a sol, as defined above, might not be formed during the conversion of the alkoxide into a gel. Alkoxides, M(OR)n where M is a metal and R an organic group (e.g., an alkyl group), have the important property of undergoing hydrolysis and polycondensation reactions corresponding to the following equations, written for partial hydrolysis with formation of a skeleton of M-O-M bonds and an increase in the molecular weight ofthe polymer. M(OR)II+HP~M(OR)II_J

(OH)+ROH

M(OR)II_J (OH) + M(OR),,_J (OH) ~ MOM(ORh"_2 + Hp M(OR)II + M(OR)II _J (OH) ~ MOM(ORhll_2 + ROH

(1)

(2) (3)

The alkoxides of interest here, such as tetraethylorthosilicate or TEOS, Si(OC zH s)4, aluminum isopropoxide, AI(OC3H7, or aluminum sec-butoxide, AI(OC4H9)3, are usually liquid and can be purified by distillation yielding high purity oxides. Alkoxides and water are generally immiscible. Hence alkoxides are used in solution in mutual solvents, such as the corresponding alcohols. Further, homogeneous solutions of several alkoxides of different metals can be prepared, with a view to obtaining homogeneous binary or ternary oxides as well as mixtures ofoxides (all alkoxide route). Metal salts, such as nitrates or acetates, can also be added to the solution, increasing the flexibility of the process. The kinetics of the hydrolysisl polycondensation reactions and the nature ofthe polymer can be tailored by using catalysts andlor by changing the functionality of the alkoxide species and the H20/M(OR)n ratio [6). Acid catalysts, e.g., mineral acids, afford chain polymers with little branching. In contrast, basic catalysts, e.g., ammonia, yield highly branched polymer networks and subsequently dense colloidal particles. Further, when working with a solution of several alkoxides, inhomogeneous products can result from different hydrolysisorcondensation reaction rates. (b) The gelation step A gel is a soft form of matter, Le., a green body capable of maintaining its shape without a mold. It is a continuous network of particles or polymeric species, swollen by a liquid. The network prevents the liquid from flowing while it impedes the solid from collapsing in a compact mass [1] (4). Gelation is the conversion of a sol into a gel, the gel point being defined as the time when an abrupt change inthe viscosity isobserved (7). Gelation of a sol can be achieved by forcing the particles together to allow interaction between them, either by physical orchemical removal ofthe solvent orby reaction of the ionic species (e.g., the mineral acid) which have caused the electrostatic double layer to form

Chapter 8

207

during peptization [5]. Vaporization of the solvent or pH change induces gelation of sols. Gelation of an alkoxide solution occurs when the hydrolysis/polycondensation reactions (Equations 1 to3)are sufficiently advanced. Gels are amorphous bodies with high porosity. They are semi-rigid and can beshaped, e.g., asfibers, while still viscoelastic or plastic. Fibers can bedrawn when the viscosity of the solgel precursor is in the range 30-100 Pa.s (300-1 000 poise). When the starting product is an alkoxide, only those precursors are spinnable which consist of chain polymers having little branching. The latter can be improved by adding spinning aids.

(c) The drying step Gels contain large amounts ofsolvent; extraction occurs with substantial shrinkage and has to becarried out with care to avoid microcracking of the body. Such microcracks would act as flaws infibers and would decrease their tensilestrength. Xerogels are produced through the extraction of the solvent by simple vaporization. Under such conditions, microcracking may occur asthe result of capillary forces appearing when the solid part ofthe gel comes incontact with the vapor [9]. Such microcracking can beavoided if the liquid vaporization rate is slow enough, but the drying process becomes extremely long. Drying control chemical additives (DCCA), such asformamide, N, N-dimethylformanide, oxalic acid, or glycerol, favor the solvent extraction. Another efficient way to avoid capillary forces and hence microcracking, is to remove the solvent by supercritical drying, the related dried gels being referred toasaerogels.

(d) The calcination and sintering steps Dried gels still contain some volatile species. They are eliminated by volatilization, decomposition, or oxidation during a calcination treatment. Depending on temperature, the residue, a porous solid, is amorphous or nanocrystalline. Finally, the calcined gel is sintered to achieve a high density and to stabilize the microstructure. The sintering of oxides produced by the sol-gel route can be performed at much lower temperatures than necessary forclassical powders, owing to their extremely fine g~ain size. This isone of the key features of sol-gel processed ceramics.

8.2 Alumina and alumina based fibers Commercial single crystal sapphire and other alumina based single crystal (e.g., YAG) fibers are obtained byedge defined film fed growth or laser heated float zone growth (Chapter 4.5). These processes are slow but yield costly fibers with premium properties. Experimental aluminate glass fibers with >50% alumina, including YAG glass fibers, have recently been reported (Chapter 4.4). In contrast, polycrystalline alumina, YAG and zirconia fibers are accessible bya sol-gel route orbya slurry process. 8.2.1

General considerations

At high temperatures, a-Ab03, i.e., corundum or sapphire, is the stable form of alumina. Transition aluminas (Figure 1)are formed between 300 and 1100°Cthrough heat treatment of

Chapter 8

208

aluminum hydrates, for example, gibbsite (a-Ab03·3H 20) , bayerite (13-Ab03·3H 20), boehmite (a-Ab03·H20) and/or diaspore (13-Ab03·H 20) . The sequence of formation of transition aluminas depends on the starting material and is affected by crystallinity, grain size, heating rate, impurities and/or additives.

I

I

Gibbsite

~[

I

I

~

Eta

.1 'I

I

I

o

I

Kappa

I Delta

Gamma

Diaspore

I

I

I

Chi

Boehmite Bayerite

I

I

Alpha

I I Alpha

ITheta Alpha

Theta

Alpha alumina

I

I

I

I

I

I

200

400

600

800

1000

1200

Temperature, °C

Figure 1. Dehydration sequences ofalumina hydrates inair. Enclosed area indicates range ofstability and open area range oftransition [20]. Path (b) isfavored bymoisture, alkalinity and coarse particle size (100 IJm) and path (a) by fine crystal size « 10 IJm); reproduced with permission of the American Ceramic Society, PO Box 6136, Westerville, Ohio 43086-6136.

The low temperature transition aluminas including the chi, eta and gamma phases are poorly crystallized and contain residual water. The eta and gamma phases display a spinel structure with vacancies. The high temperature transition aluminas, comprising the kappa, delta and theta phases, are formed at 800-1000°C. Finally, the transformation of fine theta (or kappa) transition alumina at 1100·C into a-alumina is reconstructive and occurs with an increase in grain size (10). Transition aluminas can be stabilized by addition ofsmall amounts ofdifferent oxides, the most commonly used in alumina based fibers being silica, in order to shift the formation ofa-alumina toward higher temperatures (11) (12) [13]. Mullite (3Ab03·2 Si~) contains 71 .8wt.% Ab03 and 28.20 wt.% Si02, and displays a range of homogeneity assigned to a substitution of AI:l+ for Si 4+ in tetrahedral sites with formation of oxygen vacancies (Figure 2a) (14) (15). Assuming a generic formula with an atomic structure of AbVl [Ab-+<xSb"2'I'V 010-'. the value x = 0.25 corresponds to the nominal formula 3Ab03·2Si02 (with 71.80 wt.% Ab03) and the value x = 0.40 to the formula 2AbCh·Si~ (with 77.24 wt.% AbCh). The actual homogeneity range of mullile and the nature of its melting (congruent or incongruent) are still controversial areas.

209

Chapter 8

Wt. %

20

0

60

40

80

Liquid

Mullite 55 + Liq ...

2000

~

!i

e :>

1850 0

/

"

/

/

/

Mullite + Liq

/

Cor + Liq

/

18400

Cor + mullite 55

15950

1600

I I

5i02+ mullite 1400

/

5i02+ Liq

1800

c.

E ~

100

-{ -MullneS5 (a)

I 0

60

40

20

80

100

Mol, %

5i02

AI203

1500 .--------------,,~-------,

Mullite

SOOL-

o

'-10

-'--

(b) -'-_.....J

20

30

% 5i0 2

Figure2. The binary Ab03-Si02 system. (a) Equilibrium phase diagram by S. Aramaki and A. Roy, J. Amer. Ceram. Soc., 45, 229-242 (1962): reproduced with permission of The American Ceramic Society, Westerville, OH. (b) Non-equlibrium phase sequences from heating sol-gel silica aluminas [44): reproduced with permission of the Institute ofMaterials, London.

Crystallization of mullite from stoichiometric alumina-silica mixtures, referred to as mullitization, occurs at different temperatures depending on the scale at which the constituents are actually mixed [16]. For conventional powders (grain size ",,1 I-Im), the temperature for complete mullitization is 1600-1750°C, whereas that for alumina-silica gels is 1300-1450°C and even lower (1000-1100°C). Further, the formation of mullite from gels is considerably favored by addition of boron oxide. The addition of a second phase at grain boundaries impedes grain growth and creep in a-alumina. Examples are Nextel 720, a corundum/mullite fiber [17] [18] and PRD-166, a corundum/zirconia fiber [19] [20]. 8.2.2 Processing of alumina based fibers The processes used to prepare alumina based fibers by the solution route start with precursors which are either an aqueous solution of an aluminum salt or a solution of an organoaluminum compound in an organicsolvent. The level of viscosity required for spinning is achieved by properly controlling the degree of hydrolysis/polycondensation of the precursor

210

Chapter 8

orland by adding spinning aids such as polymers with high molecular weights. In the related slurry route, a fine orultrafine alumina powder is dispersed in the liquid precursor, in order to increase the equivalent oxide content and to lower fiber shrinkage during the dryinglfiring steps. (a) Polycrystafline alumina fibers Aqueous solutions of basic aluminum salts, especially basic aluminum chloride, are the most commonly used aluminum precursors. Depending on the synthesis conditions, different species are observed in the solutions [11) [21). The monomeric hexahydrated cation, [AI (H20)6j3+ isstable only inhighly acidic solution. Atan OH/AI molar ratio of 3, AI(OHh is formed and precipitated whereas atlower ratios, i.e., 2 to 2.5, polymeric species such as the complex pseudo-spherical cation [AI O. AI,2(OHh.(H20),2t, also referred to as the AlB cation, and larger polymeric species are formed. Aqueous solutions of aluminum chloride have relatively low viscosities up to a concentration threshold [11) [21). Above this threshold, the viscosity increases dramatically (Figure 3) and a glass-like solid is then formed. Concentrated AICh solutions undergo aging accompanied by a further viscosity increase. However, adding a high molecular weight organic polymer such as poly(ethylene oxide) or partially hydrolyzed poly(vinyl alcohol) is a much more effective way ofincreasing the viscosity.

60

3

4

<Jl

50

ai

a.. 2 ;E-

'iii <Jl

40

ai

a..

;E- 30

'w 0 Co)

s

<Jl

81

s

5

<Jl

0 24

26 28 30 Oxide content, %

32

20 10 0

10

14

18

22

2Q

30

Oxide content, %

Figure 3. Relationship of viscosity and concentration ofAI20J spinning dopes prepared from aqueous solution of basicaluminum chloride containing poly(vinyl alcohol) with PVAJAI20J mass ratios of(1)30/70 ; (2)25/75 ; (3)20180 ; (4) 18/82 and (5) 15/85 (shear rate = 40 s'), The inset shows the viscosity/concentration curve for an aqueous solution of basic aluminum chloride without addition of PVA (31); reproduced with permission of Huthig and Wepf Verlag.

Chapter 8

211

Continuous, polycrystalline a-alumina fibers, e.g., Fiber FP or Almax, are produced by the slurry process [22-26]. The spin dope iseither a dispersion ofa fine a-Ab03 powder (e.g., 0.5 urn mean size) [23-24] oran ultrafine y-transition alumina powder (0.02 urn mean grain size) [25] in an aqueous solution of a basic aluminum chloride, for example, the water soluble Ab(OH)sCI·2H20 salt containing various additives, including (a) high molecular weight polymers acting as spinning aids, (b) sintering aids and (c) a grain growth inhibitor such as MgO (as magnesium chloride). The dope isconcentrated by heating under vacuum and isthen dry spun through a spinneret. The green fibers are partially dehydrated in a drying column and are taken up on a collapsible reel [23-24] or further processed continuously on line [251. They are prefired at 600-800°C and subsequently sintered at 1350 - 1500°C in a furnace or/and a propane/oxygen flame. Fiber FP is 18-20 ~m in diameter whereas the Almax fiber obtained from ultrafine y-alumina powder displays a smaller diameter (",,10 pm). In both fibers, alumina is present as corundum. a-Alumina fibers can also be prepared, at least on a laboratory scale, via pure sol or solution routes from different precursors including aqueous colloidal sols [26-27J or organoaluminum polymers [28-30], as illustrated by the two following examples. (1) Colloidal sols can be prepared from aluminum chloride (or nitrate) by dissolving AICh (or AI(N03h-9H20) and aluminum pellets in water under reflux, followed by filtration [27]. The hydrolysis/polycondensation reaction is continued until the viscosity of the precursor is suitable for dry spinning. The green fibers are dried, prefired to 800°C and sintered at 1300°C in air. Crystallization of the gel occurs above 700°C. First transition aluminas are formed and then converted to a-alumina [27]. (2) Polymeric organoaluminum precursors can be prepared by the addition of glacial acetic acid to either pure ethyl 3-oxobutanoatodiisopropoxy-aluminum (EOPA), or to a mixture of EOPA and tri-sec-butoxyaluminum (SBA) without solvent and water [28-29]. The reaction mixture is refluxed at190°Cand isthought toproceed by the following equations, where EOB isthe ethyl 3-oxobutanoato didentate ligand, CH3COCHCOOC2Hs. AI(EOB) (0 ipr)] + A cO ff _ _ Ro f1 1-O} + 'r-on + AcOipr

(4a)

EOB AI(EOB)(O jP r)] + A1(0 '''BU)3+ AcOH - -

RoBl-~n +ROH + AcOR

(4b)

EOB

After the volatile species are vaporized under vacuum, a pale yellow polymeric precursor is obtained whose viscosity is time independent and could be controlled, without adding spinning aids, by the amount of added acetic acid, the EOPAISBA molar ratio and the temperature. The green fiber, calcined at500°C, is amorphous and undergoes crystallization within the temperature range 800-1300°Ctoa-alumina. (b)

Jransition alumina fibers

Transition alumina fibers are obtained by the sol-gel route, as long as the firing temperature is

212

Chapter 8

low enough to avoid the transition to a-alumina, i.e., 1000-1100°C for pure alumina (Figure 2b) [11-12) [31). Calcination yields a fiber that is initially amorphous and porous. The crystallization of this fiber occurs through two exothermic main transformations, the first being the formation of the transition aluminas starting with the n-phase and the second that of aalumina (Figure 1). Simultaneously, the fiber shrinks. If a shrinkage resistant transition alumina fiber is desired, sintering of the porous fiber should take place in a temperature range that is actually quite narrow for pure alumina. Thus, most commercially available transition alumina fibers are stabilized with oxide additives such as silica. The addition ofsilica (Figure 2b) considerably enlarges the apparent stability domain of transition aluminas, the formation ofa-alumina being shifted toalmost 1250°Cfor only 5 wt.% Si02 [34). Table II. Commercial silica-alumina fibers Fibers

Producers

Nextel 312 440 480 550 Altex

21<

Nextel 720

Precrystallized Crystallized Alumina fiber

3M 3M 3M 3M

Composition (wt.%) A!,o, SiO, B,o, 62 24 14 70 28 2 70 28 2

Majo r phases

Ref.

borate + amorphous silica

52 52 63 52

't : AJ,O,+ m + am. silica mullite

73

27

y/o-

Sumitomo

72

28

mullite

64

3M 3M

85 85 BS

15 15 15

TI/O - AJ,O, + a - AJ,O, a - AJ,O, + mullite Al-Si spinel (or y- A!,O,) + am. silica

28 28 62 43

Sumitomo

A~O,+ am . silica

As a result, sintering can be conducted athigher temperatures with limited grain growth. The resulting fiber consists of T], y, 0, or e transition alumina phases depending on the maximum temperature, with some amounts of a-alumina and mullite appearing at the highest sintering temperatures [11) [31]. Transition alumina fibers are commercially available as short Saffil staple fibers which are not covered in this book, or as continuous Sumitomo alumina and Nextel 440 fibers which can be woven. The preparation of continuous transition alumina fibers requires the use of liquid precursors with a higher viscosity, e.g. 30 - 100 Pa.s. The control of the viscosity of the spinning dope is achieved through (1) the use ofhigh molecular weight polymers and (2) proper vaporization of the solvent in a vacuum [37). After heat treatment, the fibers are composed of one orseveral transition aluminas, depending on the sintering temperature.

(e) Mullite and related fibers Mullite fibers can be prepared by dry spinning single phase ordiphasic gels. In single phase gels formed, e.g., byslow gelation of homogeneous liquid solutions of salts or alkoxides, the silicon and aluminum bearing species are mixed atthe molecular level. In trUly diphasic gels, which are formed from a mixture ofa silica sol and an alumina sol, the constituents are mixed at a much larger scale. The size of the colloidal particles ranges from 5 to 50 nm, and the resulting interdiffusion distances are longer and mullitization ismore difficult. Finally, intermediate gels are often used which derive from a mixture of a silica hydrosol with an aqueous solution ofan aluminum salt. Although the silicon and aluminum bearing species are actually mixed at small scales in all these sol-gel precursors, the formation at relatively low temperatures of homogeneous and dense mullite bodies from amorphous alumina-silica

Chapter 8

213

gels, isa complex process [15-16] [38-41]. Single phase precursors, e.g., a solution of aluminum and silicon alkoxides in an organic solvent, might appear the most suitable mullite precursors from the standpoint ofmixing scale. However, their use has several drawbacks. First, their equivalent oxide content islow and the shrinkage of related gels is high, Second, the hydrolysis/condensation rate of aluminum alkoxides is much faster than that of their silicon counterparts, which may result in independent (consecutive) formation of =AI-O-AI= and =Si-O-Si= networks instead ofa single =AI-O-Si= network. This problem can be overcome by controlling the conditions for hydrolysis and polycondensation, for example by (1) pre-hydrolysis of the silicon alkoxide, (2) avoidance of an excess ofwater, (3) use of a catalyst that accelerates the slowest reaction, i.e., that of the silicon alkoxide or(4) chelating the aluminum alkoxide toslow down the fastest reaction. Another problem is related to mullitization and sintering. Crystallization of single phase gels occurs at-980·C with the evolution of heat [15] [40-41]. It yields an alumina-rich mullite, with a stoichiometry approaching that of 2Ab03·Si0 2• This tetragonal mullite coexists with amorphous silica that is expelled from the gel network. Free silica is reincorporated into the mullite phase with increasing temperature, At 1300·C, the mullite phase is truly orthorhombic with the stoichiometry 3Ab03·2Si02 (Figure 4).

10

80

s

T~

/

60

2WC »>: ---------

I I I

~o

o 40 20

-' .:.:.:

~"

.~

:

/ :/ ..... j.'.:/ 950

...2..D

----_.. ..- .> • •

1050

--..:-

:

.:'

:'

/

DG 3 :

:

--

.... ;'./r / .. / /

-> »>.

:

:

I

I

I

/

/

/

/

/

/

I I BG

.- ._....:.: -: / I

1150 1250 Temperature, · C

1350

1450

Figure 4, Dynamic XRO data ofsingle phase or diphasic alumina-silica gels. (2WC) single phase polymeric gel formed byslow hydrolysis yielding tetragonal mullite at 980·C, (20)single phase gel with shorter gelation due to faster hydrolysisyielding spinel and tetragonal mullite at980·C. (CG) colloidal gel from single-phase gelformed by introduction ofaluminum nitrate after hydrolysis, and showing the direct fomnation of orthorhombic mullite, but not of tetragonal mullite. (OG3) & (BG) diphasic gels with dispersed boehmite and a-Ab03 powders respectively, forming orthorhombic mullite directly by a diffusion-oontrolled mechanism [25] [52]. Reproduoed with permission of the Materials Research Society, Warrendale, PA,

214

Chapter 8

Sintering, which requires temperatures higher than 980·C, is made more difficult by the early mullitization. As aresult, the material has to be treated at>1300·C to achieve a high density. Thus, the main advantage of single phase precursors, their easy mullitization, might impede their sintering atrelatively low temperatures. The interdiffusion distances in diphasic gels are much longer and hence their mullitization is more difficult. The crystallization of the gels follows different paths depending on the heterogeneity scale. When the mixing scale is large, as in truly diphasic gels, mullitization occurs at relatively high temperatures (1200-1400·C) with no intermediate phase formation and no exothermic OTA peak at980·C (Figure 4). For intermediate mixing scales, crystallization of diphasic gels involves an intermediate exothermic step at 980·C. This step has been assigned to the formation of a poorly crystallized Al-Si spinel (alone ormixed with tetragonal alumina-rich mullite) that further reacts exothermically with the amorphous phase at 1270· C to give stable orthorhombic mullite [15] [40-41]. In this case, sintering can take place before mullitization, and isenhanced by the low crystallization state ofthe material (Figure 4). Most commercially available fibers have a molar ratio of AbOJSi02that is 3/2 that of nominal mullite (Table I). They are prepared from diphasic precursors which contain additives such as boria. For example, Nextel 312 and Nextel 440 (or 480) fibers display the following molar compositions: 3 Ab03·2 Si02·B2OJ and 3 Ab03·2 Si02·O.13 B203, respectively [32] [43-45]. Furthermore, a variety of experimental mullite fibers have been prepared, from single phase ordiphasic precursor fibers [46-52]. Fibers with a molar composition consisting of 3Ab03.B203·(2 to 3)Si02have been prepared from a diphasic precursor consisting of a mixture of (1) an aqueous solution of basic aluminum acetate stabilized with boric acid, AI(OHh(00CCH3)·1/3 H3BOJ, (2) an aqueous silica sol and (3) various additives such as dimethyl formamide [57]. The mixture is concentrated in a vacuum to a viscosity of 40-70 Pa.s and dry spun. The multifilament precusor yarn is dried in warm air, prefired at 870·C and fired at 1000·C, yielding a continuous transparent, colorless, multifilament yarn. Fibers with a composition either close to 3 Ab03·2 Si02, orfibers with a small amount of boria, e.g., 98 wt.% 3 Ab03·2 Si02and 2 wt.% B203 (or 3 Ab03·2 Si~ ·0.13 B203, have recently been reported [58-59]. The diphasic precursor is an aqueous mixture of a solution of aluminum formoacetate stabilized with lactic acid, used alone orwith addition of basic aluminum acetate stabilized with H3B03. Alternatively the diphasic precursor is an aqueous silica sol. The mixture is concentrated to a viscosity of 100-150 Pa.s and dry spun. The resulting green fibers are dried inwarm air,calcined to1000·C and heat treated to 1400·C. Boria-free fibers consist of nanocrystalline TI- or y-transition alumina and are porous when fired at 1000·C. They are fully sintered and crystallized as mullite at 1400·C. B203-modified fibers are poorly crystallized as transition alumina when fired at 1000·C, display a low specific surface area and transform completely to mullite at 1400·C (density 3.00 g/cm3). Boria seems to enhance sintering at low temperatures before mullitization, and the boria-modified fibers exhibit a lower mullite grain size. Various precursors can be used to prepare mullite fibers by dry or wet spinning. An oftenused precursor is a mixture of a silica and an alumina sol. The silica sol is prepared by hydrolysis/condensation of TEOS or Si(OC2H s). that, in turn, had been prepared in an alcohol-

Chapter 8

215

water solution inthe presence ofan acidic catalyst, e.g., HCI. The alumina sol isprepared by hydrolysis/condensation of chelated aluminum sec-butoxide or aluminum diisopropoxide under similar conditions. The chelating agent most used is acetylacetone, CH3COCH 2COCH 3 (or acac) [48-49]. The aluminum-chelated alkoxide can also be used unhydrolyzed, in a mixture with prehydrolyzed TEOS [51]. This mixture is concentrated under reduced pressure and the fibers are dry [49] [51] orwet spun [48]. Mullite fibers were prepared from a single phase precursor [52], whereby the alumina source was a viscous solution of aluminum formoacetate. After concentration, the dope was dry spun, yielding fibers which contained orthorhombic mullite after firing at 1250°C. Mullite fibers can also be prepared from another single phase liquid precursor, a mixture of an alkoxysilane, such as tetramethoxysilane or ethyl silicate, and an aluminum chlorohydrate polyol complex, such as the 1,2- dihydroxypropane complex, Ab(OH)sCI·H zO·CH3CH(OH)CHzOH [46]. Mullite fibers often contain traces of a glassy, silica based phase at grain boundaries that increases creep. Creep is overcome by increasing the AbOJ SiOz ratio, i.e., with compositions in the two phase mullite-corundum region of the alumina-silica phase diagram (Figure 2a). Two fibers, i.e., Nextel 720 [18] and a Sumitomo fiber [53] having a composition of about 85 wt.% AbOJ and15 wt.% SiOz(Table I), fulfil this condition. The Nextel fiber has a mullitization temperature of 1250°C [18]. It consists of a mixture of n and 8 transition aluminas and corundum when fired below 1250°C, and oforthorhombic mullite and corundum when fired above 1250°C [18]. The other fiber is spun from a viscous solution of polyaluminoxane [53]. The precursor of a similar fiber was prepared by polycondensation of an organoaluminum compound, such as monoisopropoxydiethyl aluminum, dissolved in ethylether [60]. Some isopropoxy groups were presumably replaced by a phenoxy group, such as ethyl 0hydroxybenzoate, inorder toimprove the spinnability of the final dope. The polyaluminoxane was dissolved in benzene, the ether was distilled off and ethyl silicate was added. After concentration, the dope was dry spun and the green fibers were aged in a humid atmosphere and calcined. The fibers (Table II) had a glassy appearance and were composed of a nanocrystalline AI-Si spinel phase (or 'Illy-transition alumina) in an amorphous silica based matrix [33] [53]. After mullitization that starts at 1150·C and is complete after 2 min at 1400°C,the fibers were composed ofmullite and corundum [33]. (d) Alumina-zirconia fibers Incorporating zirconia in crystalline a-alumina bulk ceramics improves their tensile strength and their toughness by phase transformation toughening. Tetragonal zirconia is in a metastable state and can be transformed to the stable monoclinic phase in the stress field ahead of a propagating crack. The zirconia grains are larger than a critical size orotherwise the phase transformation would occur spontaneously during processing. This concept has been transferred to corundum fibers. The experimental fiber, PRD-166, is an example ofa zirconia-corundum fiber. PRD-166 ispolycrystalline and contains 80 wt.%aalumina and 20 wt.% yttria-stabilized zirconia and it is slurry spun, like Fiber FP [19] [61]. Zirconia-corundum fibers can be prepared by the slurry route, the slurry being an aqueous suspension of fine alumina particles, to which are added a basic aluminum salt (such as aluminum chlorohydroxide), zirconyl acetate and yttrium chloride hexahydrate. After

216

Chapter 8

concentration to a viscosity of about 4S Pa.s, the mixture is spun and the green fibers are dried, preheated at,.,600·C and fired atabout 1900·C ina propane/oxygen flame (62). 8.2.3 Structure and microstructure The microstructural features of transition alumina fibers which can be formed by calcining dry spun alumina-silica gel fibers are shown inTable III. The first phase that crystallizes from the amorphous gel is n-alurrma. This phase is a spinel with vacancies distributed in the octahedral sites [10]. The calcination temperatures corresponding to each transition alumina can be estimated from Figure 2b.

(a) Transition alumina fibers According to Figure 2b, the 11 (ory) phase isstable within the temperature range 740-1070·C (34). Thus, 11 (or y) transition alumina fibers, which are still poorly crystallized, display some interesting properties. They have small grains (e.g., a few nm in size), pronounced porosity (mean pore size of 2.S nm), low density (2.7 g/cm3) , a very large specific surface area (200 m2/g by BET nitrogen adsorption), and relatively high tensile strengths (12) (31). The formation of C)-alumina at-1070·C (Figure 2b) isaccompanied by a decrease in porosity and an increase in grain size (Table III), but the fiber retains high strength. Saffil fibers consistof c)-alumina nanocrystals (10-S0 nm in size) whose structure is related to spinel (10), small amounts of y (and y') phase and a-alumina platelets (63). As temperature is further increased, the amount of a-alumina increases, the porosity decreases, grain growth continues, and fiber strength decreases. With calcined alumina gels containing 4 wi.% SiCh (Figure 2b), transition aluminas transform to corundum at a sufficiently high temperature, i.e., 1240·C and, according to the nucleation/growth mechanism, there is a significant increase ingrain growth (30). The degree of conversion of transition aluminas to corundum, a, was calculated from quantitative XRD data. When plotted as a function of annealing time for 9S0 < T < 10S0·C, sigmoidal curves were obtained. The kinetics data were treated according toa modified Avrami-Erofeev kinetic equation.

a = 1- exp (- kt" )

(Sa)

which can be rewritten as.

LnLn[1 / (1- a)}

= Ln k + n Ln t

(Sb)

where k is a rate constant, n is the time exponent whose value is related to the crystal growth mechanism and t = I.- 'ti with I. the actual annealing time and r, an incubation period related to the nucleation step. The value for n was found to be equal to 1.6. Furthermore, the growth of a-alumina crystals appeared to be thermally activated with an apparent activation energy of4S0 kJ/mol for spinnable precursor gels, which isclose tothat observed for the formation of corundum from boehmite, i.e.,431 kJ/mol.

(b) Mullite and related fibers Commercially available fibers with a composition close tothat ofmullite (3AbOJ·2Si02 or71 .80 wt.% AbOJ and 28.20 wt.% Si~) consist ofpoorly crystallized mullite ora mixture oftransition

Nextel312 Nextel 440 Nextel480 Nextel 550 Altex 2K Ne xtel 720 crystallized Altex

Fibers

200 18 2000 catalytic

TJ 6 50 187 14 . .. ..

68

62 ... 17 ...

r

~r

121 8 1800

77

r/'O '0 10-50 80-97 5-20 0-73 6.5 1000-2000 low a u ..... 86 16 46 3.5 .. ...

'0/8 100 97 20-50 0 0.5 1000 h igh a

al 81m

500

o o

aim

200+ 100 100

Composition (wt.%) AlP, SiO, Bp, 14 62 24 70 28 2 70 28 2 ... 73 27 ... 72 28 15 85 15 85 Alwnina structure 9 A~OJ.2B,OJ + am. SiO, r + m + amorphous SiO, mullite r/'O + amorphous SiO, mullite mullite + corundwn 1\ly+ amorphous SiO,

Grain size <500nm <500nm <500nm <500nm 115nm 9-30nm 1O-25nm 3.40 3.20

2.7 3.05 3.05 3.03

Densitl

(g/cm)

Strength (GPa) 1.7 2-2.1 1.9 2-2.2 1.2 1.98-2.1 1.8

260-304 210

Modulus (GPa) 150-152 190 220 193-220

0.81 0.80

Elonllation ( Yo) 1.12 1.11 0.86 0.98

Table IV. Composition, structure and properties of commercially ava ilable fibers with a composition equal or close to that of mullite.

Major phase{s) Approx. Crystal size (nm) % crystallinity % a-alumina Pore volume (mm'l g) % Shrinkage@1400°C (1 hr) Tensile strength (Mpa) Grade

Table III. Property change during the processing of alwnina fibers containing ~ 4% silica.

28,53,74 74,75

53,74 53,74 64,74 65

Ref.

Ref. 22,41,73 22,41 22,41 22,41 22 41

(")

~

"'"

~ co

iSr

218

Chapter 8

aluminas or A1-Si spinel and amorphous silica. The crystal phases depend on whether the fibers have been fired below or above the mullitization temperature (Table IV). For example, the main phase in Nextel 440 is y-alumina with an average grain size of 15nm. Mullitization further increases the average grain size with increasing firing temperature from 80 nm (1200°C) to 135 nm (1400°C)[55]. Fibers with a large excess of alumina with respect to the nominal mullite composition exhibit similar features. When fired below the mullitization temperature, they consist of a mixture of transition alumina and amorphous silica with (pre-crystallized Nextel 720 fiber) or without (Altex fiber) corundum. Conversely, after mullitization in the 11 00-1200°C temperature range (Figure 2a), the crystalline phases present in the fibers are mullite and corundum [18] [33]. Finally, the Nextel 312 fiber is a specific case. Although its AbOJlSi02molar ratio is that of nominal mullite, it also contains a large amount of boria. When fired at 900-11 OO°C, the fiber consists of an aluminum borosilicate crystalline phase mixed with amorphous silica. If the firing temperature is raised above 1100°C, a quite significant weight loss occurs due to 8203 volatilization. As a result, the chemical composition of the fiber is shifted toward that of mullite with a grainsize increase [32]. The kinetics of formation of mullite have been studied qualitatively for the Altex fiber (85Ab03'15 Si~ wt.%) [33] and quantitatively for the Nextel 440 fiber (70Ab03'28Si02·28203 wt. %) [66]. The nucleation and growth of mullite from mixtures of transition alumina and a silica based amorphous phase is governed by diffusion and is therefore temperature dependent. For example, at 112rC , the conversion of the Altex fiber into mullite is only of the order of 5-10% after an annealing time of 100 h, whereas the transformation is completed after 2 min at 1400°C [33]. Similarly, conversion of Nextel 440 is complete after 167 h at 1128°C, but requires only 1.6h at 1215°C(66). The kinetic data forNextel 440 between 1128 and 1215°C[66] follow the Avrami law (Equation 5) or an exponentiallaw with a temperaturedependent induction period (Equation 6). K , ( T) ( t - t )

K; ( T)(t- r)

= [- Ln (1 - a )r /" = -Ln (1 - a )

(5c) (6)

where a is the degree of transformation (orXRD weight fraction) of newly formed mullite; tis the annealing time; t is the induction period; n is the time exponent and K., 1<., are kinetics constants. With an induction period corresponding arbitrarily to a = 0.02, the experimental data plotted as Ln [- Ln (1 - a)] vs Ln (t - t) can be fitted with straight lines yielding n values close to 1 (for n = 1, Equations 5 and 6 are identical). The Arrhenius plots of the kinetic constant K yield a lower apparent activation energy (900 kJ/mol) than that reported for diphasic Ab03-Si02gels free of 8203, (987-1070 kJ/mol). Thus, the increase in nucleation and growth of mullite might be related to the presence of 8203which is believed to lower the viscosity of the silica based glassy phase and hence tofavor atomic diffusion [66].

(c) Corundum and related fibers The morphological and structural properties of corundum based fibers are shown in Table V. All these fibers consist of a-alumina crystals, alone or mixed with crystals of tetragonal zirconia stabilized with yttria. Since they have been fired at high temperature during processing, their grain size is relatively large. With the exception of the A1max fiber, which displays a significant microporosity [67], all the other fibers are dense.

219

Chapter 8

Table V. Properties of corundum-based fibers Mechanica l prooerties Ref. 0" (GPa) 3.92 385-414 > 1.4 74,75 FP 4.2 366-380 2.1 29,74 PRD-166 3.60 344-350 1.02 74, 77 Almax 3.88 373 1.9; 2.4 53, 80 Nextel-6lO 3.9 380 3.0 78 FP, 99.9 wt % Al,O,; PRD-166, 80% A~O, and 20% yttria-stabilized z-o, Almax, > 99.5% Al,O,; Nextel61O, >99% A~O, with 0.2-0.3% SiO, and 0.4-0.7% Fe,O,. Fibers

CJ:ystalphases type grain size (um) a-AI,0, 0.50 a-AI,0, 0.34-0.50 a-Al,O, 0.5 a-Al,O, 0.1

Dens ity

E (GPa)

(g/cm")

The kinetics of grain growth in a-alumina fibers has been studied for the Almax corundum fiber [71]. As shown in Figure 5, grain growth becomes noticeable at 1300-1400°C. The grain size, which is of the order of0.3-0.5 mm in the as-processed fiber, reaches values close to the fiber diameter, or 10 IJm, after one hour annealing at 1600-1700°C. Within the temperature range 1400-1700°C,grain growth obeys the following kinetic law.

with

JL)

K = L . k . exp( T () Rr

(7) (8)

where Go is the initial grain size; G is the grain size at annealing time t; K is the kinetic constant; y, the surface energy; T, the annealing temperature; Q, the apparent activation energy; R, the gas constant; and k, is a nominal pre-exponential constant. The apparent activation energy, derived from the related Arrhenius plot. Q = 800 kJ/mol, agrees with data reported for the growth kinetics of aluminas and is consistent with an impurity drag, limited grain growth mechanism. 8.2.4 Mechanical properties Alumina based fibers display linear elastic behavior at room temperature. Their Young's modulus depends on the chemical composition, the degree of crystallinity and any residual porosity.

(aJ Atroom temperature As shown in Figure 5 of Chapter 4, the modulus increases as the alumina content of the fibers increases. For a given composition, the stiffness of alumina-silica fibers produced by the solgel route depends on the residual porosity and hence the processing conditions. For example, the modulus of alumina fibers with a 96 AbO,Si02 (wt. %) composition increases from 110 GPa for the highly porous n-transiton alumina fiber, to 280-300 GPa for the dense a-Ab03fiber prepared athigher temperature.

220

Chapter 8

10 -

1200

1400

1600

1600

Temperature, ° c

Figure 5. Variations of the grain size of a-AI2OJ crystals in Almax fibers as a function of temperature after 1 h annealing [81); reproduced with permission ofTrans Tech. Publications.

The tensile strength of alumina based fibers is governed by the population of flaws. It displays a statistical character, which isdepicted with a Weibull distribution function.

PR

=1- Ps =1- exp [-

~ (.5!..-J"'] V"

(9a)

= m Ln (J'+ Ln (~)-m Ln v , v"

(9b)

(J'"

or

Ln [Ln ( 1/ Ps))

wherein PR and P, are the probabilities of failure and survival, G is the tensile failure stress, V is the fiber volume under stress, Vo isa reference volume, mis the Weibull modulus and Go is a scale parameter. The value of mcan be experimentally measured from a Ln [Ln (1/Ps)) vs. Ln G plot drawn from tensile test data. The large diameter of Fiber FP formed at high temperature and eXhibiting a large grain size (0.50 IJm), has a relatively low tensile strength, i.e., 1.54 GPa for L = 25 mm. The addition of 20 wt.% of t-Zr02 (Y203) to a-alumina increases the tensile strength of the fiber by about 20% [75). This improvement is assigned to the stress-induced phase transformation of tetragonal zirconia. Logically, the tensile strength of fibers produced by a sol-gel route, fired atrelatively low temperatures and hence exhibiting small grain size (Tables IVand V), is high and of the order of 2 GPa for L = 25 mm. Finally, the strength of a Nextel 610 fiber, 3.2GPa [68), isthe highest value reported so farfor a polycrystalline alumina based fiber, although somewhat lower values were reported by other investigators (Table V). When performed at a high enough temperature, heat treatments degrade the mechanical properties, e,g., tensile strength, of alumina based fibers. Two phenomena are responsible for this degradation: phase transformation orland grain growth. All the fibers composed of transition aluminas and amorphous silica undergo a phase transformation to a-alumina or mullite, with some grain growth and hence a room temperature strength decrease once they have been annealed above the transformation temperature.

Chapter 8

221

3 . 0 ' - - - - - - - - - - - - - - - - - - - - - - - ' 1()4 Nextel440

ell 2.0 a.

e

~

'6>

-

-

<:

E

<:

103 ai .~

'"<:

.~

l!!

Cl

Ci5

.s! 'iii <:

CD

Cl

-102 l!! CD ~

1.0

~

I

Transformation tomullite I

0

400

800

1200

Mullite grain growth \

(a)

1600

10

Annealing temperature, °c

2.2

PRO 166

2.0 ell a. 1.8

o -6Cl <:

1.6

l!!

Ci5 1.4

.s! 'iii <:

e

FiberFP

1.2 1.0

Held2 hoursat temperature Tested at room temperature (b)

0.8 0

400

800

1200

1600

Temperature, "C

Figure 6. Effect ofanannealing treatment on the residual tensile failure strength atroom temperature ofaluminabased fibers: (a) for Nextel 440 fiber; reproduced with permission from Elsevier Science Ltd., The Boulevard, Langford Lane, Kidlington OX5 1GB, UK, and (b) for fiber FP and PRD-166, according to ref. [65] and (29); reproduced with permission oIThe American Ceramic Society, PO Box 6136, Westerville, OH 43086-6136.

In contrast, the fibers which have (in terms of phases) reached their equilibrium state during processing, e.g., mullite orcorundum fibers, simply undergo grain growth when annealed ata sufficiently high temperature and their room temperature tensile strength decreases [18-19] [45] [54·55] [71].

222

Chapter 8

Nextel 440 is fabricated bydry spinning. Its composition is close to that of mullite (Table IV) and because it had been fired below the mullitization temperature, it consists of y-transition alumina and amorphous silica. As a result, its tensile strength is high [55J. Annealing at or above 1100-1200·C yields mullite with grain growth and a corresponding loss in strength. Above 1400·C, the fiber becomes sobrittle that it can no longer be tensile tested afterwards atroom temperature.

Bolh Fiber FP and PRO-166 have been fired at very high temperatures. Their room temperature tensile strengths remain almost constant after annealing to 1000·C. The strength decrease observed for higher annealing temperatures is assigned to grain growth [19J. In a similar manner, pre-crystallized Nextel 720 fibers, fired during processing below the mullitization temperature, consist of a mixture of transition aluminas, corundum and amorphous silica (Table IV). Annealing at 1093·C, i.e. , below the mullitization temperature, lowers their room temperature strength slightly whereas their strength degrades rapidly when the fibers are treated above 1204·C, i.e, atorabove the mullitization temperature (18). (b) Athigh temperatures All the alumina based fibers display a linear elastic behavior in airupto about 1000·C. Bofh their modulus and strength decrease sharply when the test temperature is raised from 800 to 1300·C [17) [34J [54J [67J [72J [76). As shown in Figure 7, the Nextel 720 fiber with a two phase microstructure (mullite + a-alumina) retains 85% of its room temperature (RT) tensile strength at 1200·C, but the single phase a-alumina fiber (Nextel 610) and mullite fibers having 2 wt.% 8203 retain only 43% and 58%, respectively. Similarly, PRO-166, the two phase zirconia/a-alumina fiber, displays higher retention of its RT strength at 1200·C than Fiber FP, itspure a-alumina counterpart, i.e., 64 and 44% respectively. Finally, Almax, the aalumina fiber, retains only 20% ofits RT tensile strength at 1200·C [67]. The high temperature characteristics of alumina based fibers are inferior to those of non-oxide ceramic fibers. Only the a-aluminalmullite fiber (Nextel 720) has a higher strength retention at 1200·C (85% of RT strength) than that of the commonly used Si-C-O Nicalon fiber (78% of RT strength) [76J. The most advanced non-oxide fiber, Hi-Nicalon, has higher strength and higher RT strength retention than Nextel 720 and Si-C-O fibers. It has a tensile strength of 2.6 GPa at 1200·C and retains 87% of its RT strength at this temperature; has a tensile strength of 2,3 GPa at 1600·C and retains 78% of its RT strength at this temperature; and has very low, but still measurable, tensile strength at1800·C (77). Alumina based fibers are subject to creep, even at low temperatures. This phenomenon is exacerbated by the fine grain size of the fibers [20J [54J [67J [70J [73J [79-80J. At 1200·C and with an applied stress of 70 MPa, the strain rate for the fine-grained Nextel 610 a-alumina fiber is higher than that of Fiber FP with a coarser microstructure, Both strain rates are about one order of magnitude higher than that for a bulk alumina ceramic with a grain size of 1.2 um, The creep of Nextel 610 is already measurable at a temperature as low as 900·C under an applied stress of 200-500 MPa (70). The strain-time creep curves of alumina based fibers exhibit a steady state (or secondary) s-t linear domain, the corresponding strain rate, E, depending on the applied stress, o: the grain size, a; and the test temperature, T, according to the Oorn equation.

Chapter 8

223

2.0

Nextel610 Nextel720

8: o li

1.5

Nextel 480

l!! 1ii

~ 1.0

~ ~

'iii

c:

~

0.5

o

200

400

1400

600

Testtemperature, °C

Figure 7. High temperature tensile strength in air of alumina-based fibers. Figure redrawn from [27J [64]: with permission ofthe American Ceramic Society, Westerville, OH 43086.

(10)

In this equation, A is a constant, Q is the apparent activation energy, R is the gas constant, G is the shear modulus, and p and n are the inverse grain size and stress exponents, respectively. The values of nand Q, measured for the fibers considered here, are listed in Table VI. The high values of n suggest that creep deformation is rate controlled by an interface reaction-limited grain boundary diffusional process [78]. Mullite fibers are more resistant tocreep than a-alumina fibers. For example, the strain rate ofthe Nextel 480 mullite fiber at 1200°Cis almost two orders ofmagnitude lower than that of the Nextel 610 a-alumina fiber [55].

Table VI. Creep parameters for alumina-based fibers FP

Phases a-alumina

PRD-166

a-alumina

Nextel610 Nextel480 Nextel720

a-alumina mullite a-alumina

Fibers

i-z-o, (Y)

Temperature (0C) 1150 -1250 1150-1250 1000 -1300 1150-1250 1150 -1300 1000 -1300 900 -1200 1000 - 1300 1000

n

2.75 1.06 -1.40 2 1.25 - 2.15 0.96 -1 .84 2 3 5.4 3-3.6

Q (kJ/mol) 588 564 648 757 600 660

472

Ref. 30 88 77

30 88 77

80 64 83

224

Chapter 8

The creep of alumina based fibers can be reduced by adding a second phase. For example, the steady state creep rate of a PRO-166 a-aluminalt-Zr02 (Y) fiber is about one order of magnitude lower than that of the single phase a-alumina fiber (Fiber FP) [78J. This effect is thought to be related, at least at low temperatures (T<11 OO°C), to the fact that the dispersed zirconia particles atgrain boundaries limit the mobility ofintergranular dislocations [67J. The best improvement in terms of creep resistance is observed for mullite! a-alumina fibers (such as the Nextel 720 fiber). At 1200°C, and under an applied stress of 70 MPa, an early version of Nextel 720, a 85 AbCh·15 SiOl fiber, displayed a strain rate which was one to three orders of magnitude lower than that of the mullite, Nextel 480, and a-alumina Nextel 610 fibers. This improvement is related to (1) the two phase character of the microstructure, (2) reduced grain boundary sliding and (3) the presence of elongated grains having an aspect ratio of4/1 and mosaic crystals as large as 500 nm [17J. The creep resistance of a fiber can also be studied by analyzing the bend stress relaxation (BSR) test parameter m, which is defined as m = 1 - RJR. In this equation, R, is the initial radius imposed on the fiber. R is the permanent radius taken by the fiber after a one hour relaxation treatment at a temperature T. When m approaches 1, the fiber does not creep at temperature T, and its ability to exhibit creep increases as the value of m decreases. For Fiber FP and PRO-166 fibers, a value m= 0.5is achieved after a relaxation test is performed at1000 and 1100°C, respectively. In contrast. the test temperature has to be raised to about 1500°Ctoobserve the same mvalue for the best SiC based fibers known todate [77J [79J. The fime-to-failure, tt, is another parameter which can be used to compare the fibers. It is related to the steady state strain rate, E, by an empirical relationship, such as that of Monkman-Grant. I j 'E"'= C

(11)

Inthis equation, mand C are constants (with m= - 1,20 and - 1.08 for Fiber FP and PRO-166 fibers). This relation can also be modified toinclude stress and/or temperature terms,in order tocompute lifetime [78J. 8.2,5 Physical properties As shown in Table VII, alumina based fibers exhibit coefficients of thermal expansion (CTE) which are relatively high, particularly for fibers with high alumina concentrations. Alumina based fibers fabricated by a sol-gel process are colorless and often transparent when fired at relatively low temperatures, their refractive index being of the order of 1.6 for compositions close to that of mullite. The color of the fibers can be changed by adding pigments. For example, the color of Ab03-Si~ or Ab03-Si02-B203doped with NiO changes from green tobluish as the firing temperature israised [81J. Alumina based fibers are usually good electrical insulators with low dielectric constants (Table VII). Their electrical and magnetic properties can be changed by modifying their composition and processing conditions. For example, cermet fibers, i.e., fibers comprising an oxide matrix with dispersed particulates of metal, can be produced from Ab03-Si02-B2Ch gels containing a reducible oxide (such as NiO) by heat treatment at 800-900°C in a hydrogen containing atmosphere [32J [81J.

Chapter 8

225

Table VII. Physical characteristics of alumina-based fibers . Fibers Nextel610

FP

PRD166 Nextel440 Nextel550 Nexte1312 Nextel720 Altex at 9.375 x 10 Hz

7.9 6.8 9 4.4- 5.3 5.3 3-3.5 6.0 8.8

Refractive index

1.617 1.57 1.65

Die1ectic constant 9.0

'5::;6'

5.8 5_5.2rl 5.8

Ref. 53 71 29 53 54 53 53 63

8.2.6 Applications Applications of ceramic oxide fibers are discussed in Chapter 12, along with related applications ofcarbon and silicon carbide fibers.

8.3 Zirconia based fibers Zirconia (Zr02) has the stiffness of steel, a density which is only 50% higher than that of alumina, a relatively high coefficient of thermal expansion (Table I) and excellent chemical inertness. 8.3.1 General considerations One of the key features ofzirconia lies inits polymorphism. Zirconia exhibits three polymorphs. The monoclinic phase is stable up to 1170°Cwhere it transforms to the tetragonal phase, which is itself stable up to 2370°C. Above this temperature, zirconia exists as a cubic CaF2type phase. The reversible m- to t-Zr02 transformation is key to the use of zirconia in ceramics. First. it is reversible but occurs with a thermal hysteresis. Second, it is rapid and takes place by a diffusionless shear process similar to that of a martensitic transformation. Finally, it is dependent on particle size and occurs with a volume change (3to5%) [82]. As a result. zirconia is not used as a pure single oxide but is mixed with another refractory oxide (CaO, MgO, Y203, orCe02) referred toas a stabilizer. For high stabilizer concentrations (e.g., 10-20 mol.%), the mixed oxides form cubic solid solutions (CaF2-structure), which are known as fully (or cubic) stabilized zirconias (CSZ). For low stabilizer concentrations (a few mol%), tetragonal solid solutions are formed. They can be maintained in a metastable form at room temperature if their grain size is sufficiently small. They are transformable into the monoclinic phase in a stress field that lies ahead of the tip of a propagating crack. If their grain size is too small, tetragonal zirconia polycrystals (TZP) are no longer transformable. Conversely, if it is too large, they will spontaneously undergo the t ~ m transformation upon cooling. For intermediate stabilizer concentrations, the oxide mixtures, when annealed under proper conditions, consist of a mixture of tetragonal and cubic solid solutions referred to as partially stabilized zirconia (PSZ).

226

Chapter 8

The use of transformable TZP in polycrystalline ceramics, alone or mixed with a second phase, is the origin of the so-called transformation toughening mechanism, the positive volume change and the shear strain developed by the t-Zr02 to m-Zr02 martensitic transformation induced by the stress field opposing the crack opening and hence increasing the resistance tocrack propagation. Zirconia based fibers consisting of TZP, PSZ orCSZ can be prepared by sol-gel routes similar to those used for alumina based fibers, but only a few have been produced ina continuous multifilament yarn form. 8.3.2 Processing ofzirconia based fibers The oldest process known to yield zirconia fibers is the relic process [12] [83J [84]. The simplest process is a dry spinning process using a concentrated, aqueous solution of a zirconium salt [85-86]. Neither yields technically advanced fibers. The latter are best prepared by dry spinning and heat treatment of sol-gel derived or polyzirconoxane derived precursor fibers.

(c) Fibers from zirconia sols Zirconia based fibers can be produced from aqueous sols of zirconia particles. Such sols are prepared, e.g., through the hydrolysis of a zirconium alkoxide, by the simplified overall following equations [87]. Zr(O "pr)" + H 20 ~ ZrO (0 "Pry , + 2 II PrOH

(12)

2 ZrO(O"prh

(13)

~

Zr02 +Zr (O"pr)"

The reSUlting precipitate is then peptized with an acid, such as HCI, to yield a zirconia aqueous sol with particle sizes in the range of 30-60 nm. A typical dope comprises the zirconia aqueous sol, a high molecular weight polymeric additive, and a stabilizer such as yttrium acetate, Y(OAc)J·3H20 . The viscosity of the dope is adjusted to the required value via partial evaporation of the solvent. The green fiber isdried, calcined and sintered to 1500°Cin air. Dehydration and decomposition of additives takes place at temperatures up to 250°C. The zirconia gel transforms first to poorly crystallized zirconia at about 370°C and then to tetragonal zirconia at a temperature above 480°C. Volatilization of residual carbon occurs at 800°C. The total weight loss of the zirconia gel is of the order of 40% [88J. Y-TZP fibers containing a dispersion of elo"'gated alumina grains can be prepared by the same route, except that an aluminum salt is added to the initial composition. The elongated alumina grains are formed atsufficiently high firing temperatures, i.e., 1400°C[89-90].

(b) Fibers from polyzirconoxanes Stabilized zirconia fibers can also be prepared from viscous solutions of polyzirconoxanes. The starting reagent is a zirconium alkoxide, such as zirconium tetra n-propoxide (Zr(O'Pr)4) in n-propanol or 2-methoxyethanol or zirconium tetra n-butoxide (Zr (O"BU)4) in n-butanol. Polyzirconoxanes are formed through hydrolysis/condensation reactions in the presence ofan acid catalyst (HCI or HN03) and a controlled amount of water. Since the rate of these reactions is high, the functionality of the alkoxide is lowered from 4 to 2 by formation of a complex with a chelating agent, such as acetylacetone (acac), according to the following equation written for dimeric zirconium tetra n-propoxide inan excess ofpropanol [91-92].

Chapter 8

Zr2(On Pr)8 · 2HOnpr + 2H -acac ~ Zr2 (OnPr)6 (acach +4 HOnPr

227

(14)

The acid catalyzed hydrolysis/condensation reactions yield chain-like polyzirconoxanes, when performed with appropriate acac/Zr and H20/Zr ratios. The stabilizer (Y, Ce, Ca or Mg) is introduced either as a salt (chloride, acetate or nitrate), a complex such as Ce (3) 2, 4pentane dionate, or an alkoxide (yttrium triisopropoxide, Y (OPrh). Furthermore, various additives [92-93) are often used tocontrol the dehydration and the porosity of the zirconia gel, toimprove itsability toform, orto act as a plasticizing additive (e.g., poly(ethyleneglycol)). The green gel precursor fibers are dried at 100-180°C. Then, a broad exotherm is observed by DSCIDTA at intermediate temperatures (350-700°C) indicating the decomposition of the salts and burnout of the residual organic species. It is accompanied by the crystallization of zirconia. The fibers, fired at 1000-1200°C, still contain some residual porosity and carbon (they are black). Conversely, at 1400°C, the precursor fibers are fully sintered with a density close totheoretical and their color ispale yellow. A variety of experimental Zr01-Ce01(91) [93-95), Zr01-Y203[96-98) and Zr01-CaO as well as Zr01-MgO fibers (95) (99) have been prepared by the alkoxide/polyzirconoxane route. Most of them were simply hand drawn with a glass rod from the viscous precursor. Only few fibers were produced as continuous fibers (91) (93) (98). 8.3.3 Properties and applications The physical and mechanical properties of zirconia based fibers are not sufficiently known, inasmuch as high performance continuous fibers are not available from the market and experimental materials are scarce. The applications for ceramic oxide fibers are discussed in Chapter 12, along with related applications ofcarbon and silicon carbide fibers.

8.4 Yttrium aluminum garnet (YAG) fibers Three processes are known to fabricate continuous yttrium aluminum garnet (YAG) fibers. Single crystal YAG fibers are obtained by the edge defined film fed growth process and by the laser heated float zone process (Chapter 4.5). Both are slow processes. Amorphous YAG glass fibers have recently been demonstrated by a containerless laser heated melt process (Chapter 4.4). Polycrystalline YAG fibers can be obtained with sol-gel and related processes (this chapter). These are potentially fast processes. 8.4.1 General considerations Yttrium aluminum garnet (YAlsOl1 or YAG) has a cubic crystal structure with a large cell parameter and a melting point close to that of corundum (Table I). Compression creep tests performed on YAG single crystals show that (1) it is the most creep resistant oxide presently known, and (2) its creep resistance is not strongly dependent on the crystal orientation (contrary towhat isobserved for corundum) [103-104]. It has been suggested that this high creep resistance might be related to the low density and the low mobility ofdislocations present inYAG single crystals [105). Further, YAG is stable in contact with alumina up to about 1700°C. Both materials display similar CTEs; they do not react with each other and their mutual solubility in the solid state is negligible. Hence, YAG

228

Chapter 8

fibers are potential reinforcements for alumina based all-oxide composites. 8.4.2 Processing ofYAG fibers Polycrystalline YAG fibers can be obtained from precursor fibers derived from (a) diphasic gels, (b) polymer precursors and (c) YAG powders. (a) From diphasic gels

YAG fibers can be prepared from acolloidal yttria sol and a colloidal pseudo-boehmite AIOOH sol. However, these sols are not compatible with each other. Water soluble polymers, such as poly(ethyleneoxide), poly(vinylpyrrolidone), or poly(ethyleneglycol) are added to each sol before mixing to prevent particles from interacting with each other [106-107]. A dope is prepared by mixing the two sols, previously stabilized with e.g., low molecular weight PEa, and then adding a small amount of high molecular weight PEa which acts as a spinning aid to control the dope rheology. The green as-spun fiber is dried and fired at high temperature and yields a polycrystalline YAG fiber, 120 IJm in diameter. This fiber consists of a mixture of yttria and y-alumina when fired at 900°C, as expected from the diphasic character ofthe gel. Above 1400°C, only YAG is present, and the grain size grows from 0.18 to3.21Jm when the sintering temperature israised from 1400 to1700°C. (b) From polymer precursors

YAG fibers can also be produced from organometallic precursors, such as carboxylate precursors [106] oralkoxides [108], with the advantage that the yttrium and aluminum bearing species are mixed at the molecular level. For example, yttrium and aluminum isobutyrates, Y(ChCRh and AI(02CRh with R =i - CH 2(CH 2h, are mixed in the Y/AI =3/5 stoichiometry, and dissolved in tetrahydrofuran. Partial removal of the solvent yields a dope which isdry spun in a glove box, the isobutyrates being moisture sensitive. The green fiber is pyrolyzed in air, the main weight loss occurring at 200-400°C. The amorphous residue transforms directly toYAG above 900°C, yielding a dense fiber with a grain size ofthe order of1 IJm if it israpidly heated to1500°C[106]. (c) From YAG powder

Finally, YAG fibers can be fabricated from fine YAG powders mixed with a polymer binder, by a technique having common features with that used for a-SiCfibers. For example, a mixture of amorphous (with a high sintering ability) and crystalline YAG powders is compounded with a poly (ethylene-ethylacrylate) binder [109]. The paste is then extruded through a heated spinneret, the green fiber being fired at 1700°C and resulting in a ~100 IJm polycrystalline YAGfiber. 8.4.3 Properties and applications Polycrystalline YAG fibers are highly crystalline, with a grain size ranging from 1 to 3 IJm and a density close tothe theoretical value (4.55 g/cm3j. The failure strength is relatively low, i.e., ~1000 MPa for small diameter fibers (15-30 IJm) with a grain size of ~ 1 IJm [108] and 500600 MPa for fibers of larger diameters (100 -120 IJm). Young's modulus for these fibers is assumed to be that ofbulk YAG, i.e., 283 GPa.

Chapter 8

229

The main interest ofYAG fibers istheir creep resistance. The BSR behavior of polycrystalline fibers prepared from diphasic gel (107) is intermediate between those of commercially available alumina based polycrystalline fibers (except corundum/mullite Nextel 720) and that ofmonocrystalline sapphire (c-axis) fiber. Hence, the creep resistance of polycrystalline YAG fibers (m = 0.5 at 1300°Cfor a 1 h BSR test) represents a significant improvement over that of the other alumina based fibers (m = 0.5 at 950-1100°C). However, it is much lower than the creep resistance of the monocrystalline fibers. Polycrystalline YAG creeps four orders of magnitude faster than single crystal alumina and five orders of magnitude faster than single crystal YAG. The creep parameters of polycrystalline YAG were derived from creep tests performed on bulk ceramic samples prepared by conditions close to those used for the fibers. The stress exponent was close to unity and the apparent activation energy (Equation 10) was equal to 584 kJ/mol. It has been suggested that the operative mechanism is Nabarro-Herring creep, with a rate limited by the bulk diffusion ofone ofthe y3+ orAI3+ cations (105). 8.4.4 Applications Applications of ceramic oxide fibers are discussed in Chapter 12, along with related applications ofcarbon and silicon carbide fibers. REFERENCES [1) [2) [3) [4) [5) [6]

[7] [8)

[9] [10) [11) [12] [13] [14] [15) [16) [17] [18] [19]

E. M. Rabinovitch, Solilel processing. general principles, in Sol-Gel Optics. Processing and Applications, L.C. Klein, ed., chap. 1, 1-37, Kluwer Acad. Publ., Boston, Dordrecht, and London (1994). OW. Johnson Jr., Solilel processing ofceramics and glass, Ceram. Bulletin, 64[2], 1597-1602 (1985). R. Roy, Ceramics bythesoluton-sol-qel route, Science, 238,1664-1669 (1987). Colomban P., Gel technology in ceramics, glass-ceramics and ceramic-ceramic composites, Ceramics Intern., 15,23-50 (1989). G. Wilson and A. Patel, Recent advances in sol-qel processing forimproved materials synthesis, Mater. Sci. and Technology, 9, 937-944 (1993). J. Livage, Les matenaux ceramques bEmeficient des proceces sol-qel, Mater. et Techniques, 6-7, 23-27 (1994). M. Pozo de Fernandez, C. Kang and P. L. Mangonon, Process ceramic fibers by solilel, Chemical Engineering Progress, 49-53, Sept. 1993. G. D.Kim, D.A. Lee, H. I. Lee and S.J. Yoon, A study onthedevelopment of mullite fibers using thesolilel process, Mater. Sci. and Engineering, A 167, 171-178 (1993). J. Phallipou, Processing of monolithic ceramics viasolilel, in Chemical Processing of Ceramics, 265-285, B. Lee and E. Pope, eds., M. Dekker Publisher, New York (1994) W. H. Gitzen, Alumina asa ceramic material, 7-42, The Amer. Ceram. Soc., Columbus, OH (1970). J. D. Birchall, The preparation and properties of polycrystalline aluminum oxide fibers, Trans. J. Br. Ceram. Soc., 82, 143-145 (1983). J. D.Birchall, J. A. A. Bradbury and J. Dinwoodie, Alumina fibres; preparation, properties and applications, in Handbook of COmposites, Vol. 1 Strong Fibers, W. Watt and B.V. Perov, eds., 115-154, Elsevier, Amsterdam (1985). A. Douy, Stabilisation of transition aluminas byalkaline earth ions, Key Engineering Materials, Vols. 132-136, 101-104, Trans. Tech. Publications, ZOrich (1997). J. A. Pask, Critical review ofphase equilibria in theAlz03-SiOz system, Ceram. Trans., 6, 1-13 (1990). J. A. Pask and A. P. Tomsia, Formation of mullite from solilel mixtures and kaolinite, J. Amer. Ceram. Soc., 74[10], 2367-73 (1991). M. D. Sacks, Hae-Weon Lee and J. A. Pask, A review of powder preparation methods and densification procedures forfabricating high densitymullite, Ceram. Trans., 6,167-207 (1990). D. MWilson, S. L. Lieder and D. E. Lueneburg, Micro-structure and high temperature properties of Nextel720 fibers, Ceram. Eng. Sci. Proc., 16, 1005-1014 (1995). G. Das, Stability of polycrystalline Nextel720 fiber, Ceram. Eng. Sci. Proc., 17[4], 25-34 (1996). J. C. Romine, New high temperature ceramic fiber, Ceram. Eng . Sci. Proc.,8 [7-8], 755-765 (1987).

230

Chapter 8

(20)

D. J.Pysher and R. E.Tressler, Creep rupture studies oftwo alumina based ceramic fibres, J. Mater. Sci., 27, 423-428 (1992). B. Claus, Preparation and characterization of spinning dopes for dry-spinning of contin-uous alumina green fibers, Angewandte Makromoleculare Chemie, 217,139-158 (1994). A. K. Dhingra, Alumina FiberFP, Phil.Trans. R. Soc. Lond., A294, 411-417 (1980). L. E. Seufert, Alumina fiber, US Pat. 3,808,015, Apr. 30, 1974. Hyunkook Shin, Firing process foralumina yam, US Pat. 3,953,561, Apr. 27,1976. K. Koba, T. Utsunomiya, Y. Saitow, K. Iwanaga, M. Matsue and N. Nishitani, Continuous process for producing long y-alumina fibers, US Pat. 4,812,271, Mar. 14, 1989. T. Maki and S. Sakka, Preparation of alumina fibers by sol-gel method, J. Non-crystalline Solids, 100, 303308 (1988). Y. H. Chiou, M. T. Tsai and H. C. Shih, The preparation ofalumina fibre by sol-gel processing, J. Mater. Sci., 29, 2378-2388 (1994). T.Yogo and H.Iwahara, Synthesis ofpolycrystalline alumina fiber with aluminium chelate precursor, J.Mater. se. 26, 5292-5296 (1991). T. Yogo and H. Iwahara, Synthesis ofy-alumina fibre from modified aluminium alkoxide precursor, J. Mater. Sci.,27,1499-1504 (1992). T.Nishio and Y. FUjiki, Phase transformation kinetics of precursor gel to a-alumina, J. Mater. Sd ., 29, 34083414 (1994). J. Dinwoodie, Saffil alumina fiber. manufacture, properties and application in high temperature furnaces, Canadian Ceramics Quarterly, 23-29, Feb. 1996. H. G. Sowman, Alumina-boria-silica ceramic fibers from the sol-gel process, in So/-gel technology forthin films, fibers, preforms, electronics and specialty shapes, L. C. Klein, ed., 162-183, Noyes Publications, Park Ridge, NJ (1988). C. Lesniewski, C. Aubin and A. R. Bunsell, Property-structure-characterization of a continuous fine aluminasilica fibre, Composites Science &Technology, 37, 63-78 (1990). M. H. Stacey, Developments incontinuous alumina-based fibres, Br. Ceram.Trans., 87, 168-172 (1988). Babcock and Wilcox Company, Brit. Pat., 1141207, Jan. 29, 1969. J. D. Birchall, MJ.Morton, J. S. Kenworth and M. D. Taylor, British Pal., 1,425,934, Feb. 25, 1976. M. J.Morton,J. D. Birchall and J. E. Cassidy, British Pat., 1360197, July 17, 1994. J. C. Huling and G. L. Messing, Surface chemistry effects on homogeneity and crystallization of colloidal mullite sol-gels, Ceram.Trans., 6, 221-229 (1990). S. Komameni and R. Roy, Mullite derived from diphasic nanocomposite gels, Ceram. Trans., 6, 209-219 (1990). B. E. Yoldas, Mullije formation from aluminum and silicon alkoxides, Ceram.Trans.,6,255-261 (1990). I.Jaymes, A. Douy, D. Massiot and J. P. Coutures, Characterization of mono- and diphasic-mullite precursor powders prepared by aqueous routes. 27AI and 29Si MAS-NMR spectroscopy investigations, J. Mater. Sci., 31, 4581-4589 (1996). D. X. Li and W. J. Thompson, Tetragonal toorthorhombic transformation during mullite formation, J. Mater. Res., 6 [4J, 819-824 (1991). T. L. Tompkins, Ceramic oxide fibers, building blocks for new applications, Ceramic IndUStry, 144 [4], 45-50 (1995). H. G. Sowman and D. D. Johnson, Ceramic oxide fibers, Ceram. Eng. Sci. Proc., 6 [9-10], 1221-1230 (1995). D. D. Johnson, Properties and applications for 3M's Nextel 440 ceramicfiber, inLooking Ahead forMaterials and Processes, J. de Bossu, G. Briens and P. Lissac, eds., 443-454, Elsevier Science Publishers, Amsterdam (1987). J. M. BOUlton, K. Jones and H. G. Emblem, Gels, filaments and fibres from alkoxysilanes and aluminium chlorohydrate·polyol complexes, J. Mater. Sci., 24, 979-990 (1989). K. R. Venkatachari, L. T. Moeti, MD. Sacks and J.H. Simmons, Preparation ofmullite-based fibers by sol-gel processing, Ceram. Eng. Sci. Proc.,11 [9-10J 1512-1525 (1990). D. S. Tucker, J. S. Sparks and D. C. Esker, Production of continuous mullite fiber via sol-gel processing, Ceram. Bull., 69 [12], 1971-1974 (1990). G. D. Kim, D. A. Lee, H.I. Lee and S. J. Yoon, A study on the development of mulling fibers using sol-gel process, Mater. Sci. Eng.,A 167, 171-178 (1993). S. AI-Assafi, T. Cruse, J. H. Simmons, A. B. Brennan and M. D. Sacks, Sol-gel processing of continuous mullne fibers, Ceram. Eng. Sci. Proc.,14[7-8],1060-1067 (1993). S. AI-Assafi, T. Cruse, J. H. Simmons, A. B. Brennan and M. D. Sacks, Processing of fine-diameler continuous mullite fibers, Ceram. Trans.,46, 65-72 (1994). L. Chen, B. Wang, S. Liu and Y. Van, Preparation of mullite fiber, J. Amer. Ceram. Soc., 79 [6), 1494-1498 (1996). Y. Abe, S. Horikiri, K. Fujimura and E. Ichiki, High-performance alumina fiber and alumina/aluminum

[21]

[22) [23] (24) (25) [26] (27) (28) (29) [30] [31] [32) (33) (34) [35] [36] [37] [38) [39] [40) [41) [42] [43) (44) (45) (46) (47) (48) (49) [50) [51) [52) [53)

Chapter 8

[54] [55] [56] [57) [58) [59) [60) [61) [62J [63) [64] [65] [66] [67) [68) [69] [70] (71) [72] [73] [74] [75] [76] [77] (78) [79J [80) [81] [82J [83]

231

composites, in Progress inScience and Engineering ofComposites (T. Hayashi, K. Kawata and S. Umekawa, eds.), ICCM-IV, 1427-1434, Tokyo (1982). D. D. Johnson, A. R. Holtz and M. F. Grether, Properties of Nextel 480 ceramic fibers, Ceram. Eng. Sci. Proc., 8 [7-8), 744-754 (1987). M. Schmucker, F. Flucht and H. Schneider, High temperature behavior ofpolycrystalline aluminosilicate fibres with mullite bulk composition. I. Micro-structure and strength properties, J. Europ. Ceram. Soc., 16, 281-285 (1996). A. Borer and G. P. Krogseng, Method offiring dry spun refractory oxide fibers, US Pat. 3,760,049, Sept. 18, 1973. H. G. Sowman, Aluminum borate and aluminum borosilicate articles, US Pat. 3,795,524, Mar. 5, 1974. K. A. Karst and H. G. Sowman, Non-frangible alumina-silica fibers, US Pat. 4,047,965, Sept. 13, 1977. E. A. Richards, C. J. Goodbrake and H. G. Sowman, Reactions and microstructure development in mullite fibers, J.Amer. Ceram. Soc.,74 [10], 2404-2409 (1991). Kadokura Hidekimi, Harakawa Masashi, Seagusa Kumo and Yamagiwa Masao, Processing for producing inorganic fiber, Europ. Pat., 0083 839, Nov. 30, 1982. J.K. Weddel, Continuous ceramic fibers, J. Text. Inst., 81 [4J, 333-359 (1990). M. C. Wolfe, Zirconia-modified alumina fiber, US Pat. 4 753 904, June 28, 1988. S. Wang and H. J. Dudek, Structure of phases in the o-A1203 fiber studied by convergent beam electron diffraction, Metal. Mater. Trans.,27A, 3318-3329 (1996).

A. R. Bunsen and M. H. Berger, Ceramic fibre development and characterization, in Key Engineering Materials, Vols. 127-13115-26, Trans.Tech. Publ., Switzertand (1997). A. R. Bunsen, Development offine ceramic fibres for high temperature composites, Materials Forum, 11 , 7884 (1988). B. O. Hildmann, H. Schneider and M. Schmucker, High temperature behaviour of polycrystalline aluminosilicate fibres withmullite bulk composition. II.Kinetics ofmullite formation, J.Europ. Ceram. Soc., 16, 287-292 (1996). V. Lavaste, M. H. Berger, A. R. Bunsen and J. Besson, Microstructure and mechanical characteristics of alpha-alumina-based fibres, J. Mater. Sci., 30, 4215-4225 (1995). D. M. Wilson, Statistical tensile strength of Nextel 610 and Nextel 720 fibres, J. Mater. Sci., 32, 2535-2542 (1997). S. Nourbakhsh, F. L. Lian9 and H. Margolin, Characterization of a zirconia toughened alumina fibre, PRD166, J. Mater. Sci. Letters, 8, 1252-1254 (19B9). D. M. Wilson, D. C. Lueneburg and S. L. Lieder, High temperature properties of Nextel 610 and aluminabased nanocomposite fibers, Ceram. Eng. Sci. Proc., 14 [7-B], 609 (1993). J. M. Heintz, J. C. Bihr and J. F. Silvain, Grain growth in alumina polycrystalline fibres during NiAI/AI203 composite processing, in Key Engineering Materials, Vols. 127-131, 211-21B, Trans. Tech. Publications, Switzerland (1997). A. S. Kim, S. Bengtsson and R. Warren, Fracture strength testing of o-alumina fibres with variable diameters and lengths, Composites Science and Technolgy, 47, 331 -337 (1993). J. Goring and H. Schneider, Creep and subcritical crack growth of Nextel 720 alumino-silicate fibers as received and after heat treatment at1300·C, Ceram. Eng. Sci. Proc., 1B (3), 95-102 (1997). K. Jakus and V. Tunuri, Mechanical behavior of a Sumitomo alumina fiber at room and high temperature, Ceram. Eng. Sci.Proc., 10[9-10], 133B-1349 (19B9). V. Lavaste, J. Besson and A. R. Bunsen, Statistical analysisof strength distribution of alumina based single fibers accounting for diameter variations, J. Mater. Sci., 30, 2042-204B (1995). D. J. Pyscher, K. C. Goretta, R. S. Hodder and R. E. Tressler, Strengths of ceramic fibers at elevated temperatures, J. Amer. Ceram. Soc., 72 [2], 284-2BB (19B9). G. Chollon, R. Pailler, R. Naslain and P. Olry, Structure, composition and mechanical behavior at high temperature ofthe oxygen-free Hi-Nicalon fiber, Ceram. Trans., 5B, 299-304 (1995). D. J. Pysher and R. E. Tressler, Tensile creep rupture behavior ofalumina based poly-crystalline oxide fibers, Ceram. Eng. Sci. Proc.,13 [7-B], 218-226 (1992). R. E. Tressler and J. A. DiCarlo, High temperature mechanical properties ofadvanced ceramic fibers, in High Temperature Ceramic Matrix Composites, R. Naslain, J. Lamon and D. Doumeingts, eds., 33-49, Woodhead Publishing Ltd., Cambridge, UK (1993). R. E. Tressler and J. A. DiCarlo, Creep and rupture ofadvanced ceramic reinforcements, Ceram. Trans., 57, 141-155 (1995). H. G. Sowman, A new era inceramic fibers via solileltechnology, Ceram. Bull., 67[12], 1911-1916 (1988). R. Stevens, Zirconia and zirconia ceramics, 2nd edition, publication No. 113, Magnesium Elektron Ltd., Twickenham, UK (1986). P. A. Vityaz, I.L. Fyodorova , I. N. Yerrnolenko and T. M. Ulyanova, Synthesis ofalumina and zirconia fibers,

232

Chapter 8

Ceram. International, 9 [2), 46-48 (1983). B.H. Hamling. A. W. Naumann and W. H. Dresher. CeramIc fibers and textiles from organic precursors. Appl. Polymer Symp., 9.387-394 (1969). [85] D. B. Marshall, F. F. Lange and P. D. Morgan. High-strength zirconiafibers, J. Amer. Ceram. Soc., 70[8], C187-188 (1987). [86J M. E. Khavari, F. F. Lange. P. Smith and D. B. Marshall. Contnuous spinning of zirconia fibers. retatons between processing and strength, in Better Ceramics through Chemistry III. C.J. Brinker, D.E. Clark. D.R Ulrich, eds., Mater. Res. Soc. Symp. Proc.,21, 617 (1988). [87J B. Clauss, A. Grub and W. Oppermann, Continuous yttria-stabilized zirconia fibers. Adv. Mater.. 8 [2), 142146 (1996). [88] S. M. Sim and D. E.Clark, PreparaUon ot zirconia fibers bysol"gel method. Ceram. Eng. Sci. Proc.•10[9-10], 1271-1282 (1989). [89] S. M. Sim, A. Morrone and D. E. Clark, Processing and microstructure of Y-TZP/AI203 fibers, Ceram. Eng. Sci. Proc., 11 [9-10). 1712-1728 (1990). (90) M. K. Naskar, and D. Ganguli, Rare-earth doped zirconia fibres by sol"gel processing. J. Mater. Sci., 31 . 6263-6266 (1996). [91) G. Emig. E. Fitzer and R Zimmermann-Chopin, Sol"gel process forspinning of conUnuous (Zr, Ce)02fibers, Maler. Sci. Engineering, A 189, 311 -317 (1994). [92] G. De, A. Chatte~ee and D. Ganguli, Zirconia fibres from the zirconium n-propoxide-acetylacetone-waterisopropanol system, J.Mater. Sci. Letters. 9, 845-846 (1990). (93) G. Emig. R. Wirth and R Zimmermann-Chopin, Sol-pel based precursors formanufacturing refractory oxides, J. Mater. Sci..29, 4559-4566 (1994). [94J R Di Maggio, F.Farina, T. Mangialardi and P. Scardi, Preparation ofCe02-stabilized Zr02 fibers by a chemically modified alkoxide method. in Advanced Structural Fiber Composites (P. Vincenzini, ed.), 37-44, Techna Srl. , Italy (1995). [95] K. Kamiya. K. Takahashi. K. Maeda and H. Nasu. Sol"gel derived CaO- and Ce02-stabilized Zr02 fibers. Conversion process ot gel tooxide and tensile strength, J. Europ. Ceram. Soc., 7.295-305 (1991). [96] T. Yogo. Synthesis ofpolycrystalline zirconia fibre withorganozirconium precursor, J. Mater. Sci., 25. 2394-2398 (1990). (97) C. Sakurai. T. Fukui and M. Okuyama, Hydrolysis method forpreparing zirconia fibers. Ceram. Bull.. 7 [4), 673-674 (1991). [98J H. Goto, H. Tomioka, T. Gunji, Y. Nagao, T. Misono. and Y. Abe. Preparation ofcontinuous Zr02-Y203 fibers byprecursor method using polyzirconoxane, J. Ceram. Soc. Japan. 101 [3],336-341 (1993). [99J M. Chatterlee, A. Chatte~e and D. Ganguli, Preparation of Zr02-CaO and ZrCh-MgO fibres byalkoxide solgel processing. Ceram.lntem., 18,43-49 (1992). [100) Y. Abe, H. Tomioka. T. Gunji, Y. Nagao and T. Misono. A one-pot synthesis of poly-zirconoxane as a precursorforconUnuous zirconia fibres, J. Maler. Sci. Letters, 13, 960-962 (1994). [101] K. J. McClellan, H. Sayir, A. H. Heuer, A. Sayir. J. S. Haggerty and J. Sigalovsky, High-strength, creepresistantY203-stabilized cubicZr02 singJe-crystal fibers, Ceram. Eng. Sci. Proc., 14[7-8J, 651-659 (1993). [102) Birkby.1. and Stevens. R, Applications ofzirconia ceramics. inKey Engineering Materials. 122-124.527-552, Trans. Tech. Publications. Switzerland (1996). (103) G. S. Corman, Creep ofyttrium-aluminum garnet single-crystal, J. Mater. Sci. Letters, 12,379-382 (1993). (104) S. Karato, Z. Wang and K. FUjino, High temperature creep of yttrium-aluminum garnet single-crystals, J. Mater. ss, 29, 6458-6462 (1994). [105] T. A. Parthasarathy, T. I. Mah and K. Keller, Creep mechanism of polycrystalline yttrium aluminum garnet, J. Amer. Ceram. Soc.•75[7], 1756-1759 (1992). [106) B. H. King, Y. Liu, R M. Laine and J. W. Halloran, Fabrication of yttrium aluminate fibers. Ceram. Eng. Sci. Proc., 14 [7-8], 639-650 (1993). [107] B. H. King and J. W. Halloran, Polycrystalline yttrium aluminum garnet fibers from colloidal sols, J. Amer. Ceram. Soc.. 75[8], 2141-2148 (1995). [108] G. N. Morscher, K. C. Chen and K. S. Mazdiyasni, Creep resistance ofdevelopmental polycrystalline yttriumaluminum garnet fibers, Ceram. Eng. Sci. Proc., 15(4). 181 (1994). [109] D. Popovich and J. L. Lombardi, Fabrication and mechanical propenies of polymer melt spun yttrium aluminum garnet (YAG) fiber. Ceram. Eng. Sci. Proc.,18[3J. 65-72 (1997). [84]

CHAPTER 9 CARBON FIBERS FROM PAN AND PITCH R. Naslain Commercial carbon fibers are obtained from solid organic precursor fibers such as polyacrylonitrile fibers, or from solid carbonaceous precursor fibers derived from pitch or mesopitch, the semisolid residue ofoil refineries.

9.1 General considerations Carbon fibers, like elemental carbon, remain solid up to extremely high temperatures. They vaporize in an inert atmosphere without melting atabout 3600°C under 1 atm. Their stability in an oxidizing atmosphere is limited, but their surface can be protected by suitable coatings. In addition, carbon fibers have moderate to high strength. They are stiffer than glass fibers and tolerate higher in-use or service temperatures. Carbon fibers are widely used in today's market where their high stiffness and resistance to elevated temperatures can justify their relatively high price. 9.1.1

History ofcarbon fibers

The interest in carbon fibers started in the late 1950s with the development of a modem aerospace industry requiring stiff, strong and light structural materials. The first carbon fibers were prepared from rayon, but they had low strength, stiffness and thermal conductivity. Post treatment under tension at2200°Cyielded carbon fibers with amodulus exceeding 500 GPa. In the early 1960s, polyacrylonitrile (PAN) fibers afforded a total carbon yield after pyrolysis that was higher, and high strength carbon fibers were obtained by stretching PAN fibers in steam and oxidizing them under stress before carbonization. Carbon fibers from pitch precursors are a more recent development. Pitches are low value residues of the petroleum industry. Pyrolysis of pitch fibers yields low cost fibers with low strength (1 GPa) and low modulus «40 GPa). An advance in the 1970s afforded ultrahigh modulus carbon fibers (>380 GPa) from mesophase pitch. Mesophase is a liquid crystal phase formed in the transformation of organic precursors to carbon. The chemical pretreatment of the pitch is complex and the potential cost advantage ofusing a pitch precursor isstill not fully realized [1-8). 9.1.2 Elemental carbon Elemental carbon exhibits two main crystalline modifications. Diamond is cubic and graphite is hexagonal. Fullerene, a new form ofcarbon, will not be discussed; fullerene related carbon nanotubes are discussed inChapter 3.

234

Chapter 9

In graphite, the atomic layers, referred to as graphene layers, are stacked in the c-direction. Within these layers, the C-C bond is extremely strong. Conversely, the graphene layers are only weakly bonded to one another. The two-dimensional character of the graphite crystal structure is responsible for the strong anisotropy, which is observed for most of the mechanical and physical properties ofthe single crystal ofgraphite. Since the layers in graphite are only weakly bonded along the c-axis, they can easily slide over one another orland undergo in-plane rotation. These motions lead to turbostractic carbon structures, which consist of roughly parallel and equidistant graphene layers rotated randomly. The terms: graphite material or graphite carbon should be used only if the interlayer spacing is less than 0.344 nm; otherwise the terms: carbon material ornon-graphitic carbon are preferred [3}. Most fibers fabricated from organic precursors should be referred to as carbon fibers, even when they exhibit rather high Young's moduli, and not as graphite fibers, with the exception ofsome fibers derived from mesophase pitch [3}. Graphitization is a solid state transformation of thermodynamically unstable non-graphitic carbon into graphite bythermal activation. Non-graphitizable carbons cannot be transformed into graphitic carbon by heat treatment up to 3000 K under atmospheric or lower pressure. Graphitizable carbons can be converted into graphite. Generally, non-graphitizable carbons are produced from non-fusing organic precursors, whereas most graphitizable carbons usually pass through a liquid crystal intermediate (the mesophase) during the carbonization process [9].

1200

Cr(GPa)

5r(llt5GPa") 1000

. e

11.

!i 's

~

5" = ~=

5.. =

800

5'2 = 5'3 =

98 2750 25000 ,16 ·33

800


'co C

"

~

400

IIII

200

o

Figure 1.

C" - 1060 C33 = 3&.5 4 C.. = C'2 = 160 15 C'3 =

20

Lc

40 60 Orientation 9, 0

60

Young's modulus ofgraphite as a function oforientation angle e [2].

In carbons derived from carbonizing an organic precursor, the structural unit is the graphene layer. Graphene layers form small turbostratic stacks. In many poorly organized carbons, each graphene layer is made of less than 10 to 20 atom rings and the stack consists of 2 to4 layers lying roughly parallel to each other. This stack is referred to as basic structural unit (BSU) [4} [10j orbasic microstructural unit (BMU) [11}. The BSU size isofthe order ofone or a few nm. Since BSUs are thin, they are flexible and often distorted. BSUs are associated

Chapterg

235

edge-to-edge, are more or less parallel to each other, and have tilt and twist boundaries, leading tovarious microtextures. Boundaries between BSUs often involve heteroatoms (such as N) and/or C atoms with Sp3 bond configuration. The association of some BSUs edge-toedge and face-to-face, with the carbon layers being roughly parallel to each other, yields local molecular orientation (LMO). Models have been used to describe the microtextural change that occurs in a graphitizable carbon as the heat treatment temperature is increased. The transformation ofmicrostructures, built on BSUs, into graphite crystals involves four steps [10). In stage 1, single BSUs are randomly distributed. In stage 2, single BSUs coalesce, forming distorted columns as the interlayer defects are released. Stage 3 takes place above 1600°C and corresponds to the edge-to-edge association of the columns. Larger but still distorted carbon layers are formed. Up to the end of this stage, the material remains turbostratic. Finally, stage 4 begins above 2000°C after the release of all defects. In this stage, the graphene layers become flat, stiff and almost perfect. Three-dimensional ordering of hexagonal layers, characteristic of graphite, takes place only in this final stage. 9.1 .3 Classification ofcarbon fibers Most commonly, carbon fibers are classified bytheir mechanical properties [1). Accordingly, they are divided into five categories. Ultrahigh modulus (UHM) carbon fibers have a longitudinal Young's modulus of >500 GPa. These fibers are fabricated from mesophase pitch and are graphitized atvery high temperatures. High modulus (HM) carbon fibers have a modulus >300 GPa. Intermediate modulus (1M) carbon fibers have a modulus up to300 GPa, and low modulus (LM) carbon fibers have a modulus as low as 100 MPa and a low strength. These carbon fibers are fabricated from isotropic pitch and have an isotropic microtexture. Finally high tenacity (HT), also known as high strength (HS), carbon fibers have high tensile strength. 9.2 Processing of carbon fibers

Carbon fibers reviewed in this chapter are continuous multifilament yarns, fabricated from organic polymer precursors whose molecular architectures prefigure, in a more or less pronounced manner, the hexagonal structure of the graphene layers [3). The typical process consists of three stages: (1) fabrication of the solid precursor fiber, (2) orientation and stabilization ofthe precursor fiber, and (3) carbonization ofthe precursor fiber. 9.2.1

Principles offiber formation

A variety of organic polymers can yield carbon fibers after spinning and carbonization. Polyacrylonitrile (PAN) is either wet spun or dry spun from a high viscosity solution. PAN derived precursor fibers must be stretched (thereby oriented) before they can be stabilized and carbonized. Pitch compositions are melt spun. Mesophase pitch consists of large discoid polyaromatic molecules which prefigure the hexagonal two-dimensional (2-D) network of the graphene layers. When it melts, it forms either an oriented liquid crystal, the mesophase, or a mixture of liquid crystals and isotropic liquid. During melt spinning, the discoid molecules orient spontaneously along the fiber axis, and the resulting orientation is maintained during the subsequent stabilization and carbonization steps.

236

Chapter 9

The carbonization of a PAN based fiber yields a non-graphitizable carbon fiber with a relatively low modulus (250 GPa), but with a tensile strength that can be as high as 7000 MPa. The stiffness of PAN based carbon fibers can be increased to about 500 GPa by stretching at very high temperatures. The carbonization of mesophase pitch based fibers yields graphitizable carbon fibers, which can be transformed into UHM fibers by a short heat treatment at very high temperatures. These fibers exhibit a high modulus (700 to 900 GPa) that isclose to Ell, for the single crystal ofgraphite. However, most have low tensile strengths (e.g., 2200 MPa).

Pan

Pitch

Stretching

Stretching

Hot

Stretching

Figure 2.

Main steps inthe fabrication ofacarbon fiber from PAN orpitch precursors.

Pitch as a precursor material is cheaper than PAN as a precursor fiber, but the conversion of pitch into mesophase pitch and subsequent fiber formation is complex and costly. When a pitch is not transformed into a mesophase and is spun as an isotropic liquid, the resulting carbon fibers have extremely poor mechanical properties. These considerations explain why more than 90% of today's carbon fibers are fabricated from PAN based precursors. Processes for fabricating carbon fibers from PAN or pitch based precursor fibers differ in important aspects, but also share important commonalties (Figure 2). Finally, the carbon yield from PAN based precursor fibers is 50%, that from mesophase pitch is 70·80% , and that from rayon is 25%.

Chapler9

237

9.2.2 From polyacrylonitrile based precursor fibers The common precursor fibers are not usually based on PAN homopolymer, but on carefully tailored copolymers [1J [3J [6J. The acrylonitrile monomer is therefore copolymerized with one orseveral other monomers such as acrylonitrile, itaconic acid (ITA) ormethylacrylate (MA). (a) Nature ofthe precursor PAN copolymers are more readily soluble in spinning solvents than homopolymer. The resulting solutions have a higher percentage of solids and are therefore more readily dry or wet spinnable. The stabilization temperature of PAN homopolymer fibers is high and the stabilization rate is low. Copolymers have a lower glass transition temperature. Their use reduces the stabilization temperature and cuts the stabilization time bya factor of ten or more [3J. The stabilization of PAN precursor fibers is an exothermic process and requires careful control. (b) Spinning of PAN based precursor PAN homo- and copolymers are infusible, decompose below their melt temperature, and cannot be melt spun. PAN copolymers are soluble in highly ionizing solvents, such as dimethylacetamide (DMAC) or dimethyl-formamide (DMF). Two commercial processes are known. In the wet spinning process, a viscous DMAC solution ofa suitable PAN copolymer is extruded through tiny spinneret orifices into an aqueous coagulant bath which extracts the solvent from the fiber. The rate at which the solvent is extracted controls the shape of the fiber cross section and the distribution of nanoporosity. Optimally, fibers with circular cross sections and minimal nanoporosity are obtained [6J. In the dry spinning process, a viscous DMF solution of a suitable PAN copolymer is extruded through a spinneret into a hot dry spinning column that extracts most, if not all, ofthe solvent. Dry and wet spinning of PAN have the disadvantage of requiring large amounts of toxic solvents. This drawback is overcome by using melt assisted spinning, a highly experimental process. In this process, acrylonitrile copolymer is plasticized under pressure bya controlled addition of water, allowing it to form a homogeneous melt well below its decomposition temperature. The plasticized melt can be melt spun directly into a steam pressurized solidification zone, with no need for expensive solvent [6J. (c) Stretching PAN polymer chains have a random coil configuration. They have to be stretched along the fiber axis to yield improved mechanical properties after carbonization. The fiber is first stretched moderately during the spinning step itself (e.g., 2.5 times) and then extensively (e.g., up to 14 times) in steam or boiling water. After this treatment, the PAN fiber exhibits a highly oriented texture which still has tobe locked into place. (d)

Stabilization

The stretched PAN fiber is slowly heated under tension in air between 220 and 280°C to stabilize the stretched zigzag chains aligned along the fiber axis. The following phenomena occur during stabilization. (1) Cyclization of the nitrile groups and dehydrogenation of the saturated C-C bonds yield a ladder type polymer. (2) Oxidation introduces oxygen bearing

238

Chapter 9

groups into the ladder polymer. (3) The ladder polymer is rigid and thermally stable, and the oxygen atoms contribute to chain crosslinking. During the stabilization step, the PAN fiber, which isinitially white, turns brown.

PAN

spool

Stretching

Stabilization

Steam atmosphere

220·280·C

Airat

Figure 3. Stretching and stabilization of PAN-based fibers (3). reproduced with permission of Elsevier ScienceNL. Amsterdam.

The stabilization of PAN precursor fibers is highly exothermic. Reaction temperature and amount of heat generated [12] depend upon the precursor composition and on the reaction environment. Optimally, the fiber will take up 8-10 wt. %oxygen without overheating. During stabilization, scission and oxidation reactions result in an evolution of gaseous species, i.e., HCN, CO2, and H20, and these reactions cause the fiber to shrink [13]. Comonomers exert a catalytic effect. The stabilization of the homopolymer in airis very slow, and requires several hours when run isothermally at220-230·C. When PAN copolymers are used, the stabilization time can be reduced toless than one hour [1] [3] [13]. (e) Carbonization

The endothermic carbonization of PAN fiber proceeds slowly by heating the stabilized fiber in an inert atmosphere to a final temperature ranging from 1000 to 1600·C (Figure 4). The evolution ofgases isaccompanied bya further weight loss and by further shrinkage. The gas evolution occurs mostly below 1000·C and the off gasses consist of (1) nitrogen containing species, i.e., HCN, NH3, and molecular nitrogen, (2) short chain hydrocarbons, including CH4, (3) carbon oxides and (4) H20 and molecular hydrogen. The final fiber diameters range from 5 to 10 IJm, or about one half those of the green PAN fiber. For the PAN homopolymer, the carbon yield (45-55 wt.%) ismuch lower than the theoretical value (67.9wt.%). The transformation of an organic into an inorganic fiber occurs by complex mechanisms. The carbonization temperature range can be divided into three domains. The first domain (200500·C), where the evolution rates of HCN, H20 and C~ are high, corresponds mainly to crosslinking reactions [1]. The second domain (400-800·C) ischaracterized by the formation ofhydrocarbons and ammonia as well as an initial evolution of molecular hydrogen. The third domain (700-1000·C) corresponds to acontinuing evolution of HCN and hydrogen, and to the onset of the formation of molecular nitrogen. Most of the oxygen that was introduced during

239

Chapter 9

the stabilization step is subsequenty eliminated during the carbonization; its residual concentration is lower than 1 wt.% at 1000°C. At 1000°C, the PAN based fiber still contains large amounts of nitrogen «7 wt.%) and hydrogen «0.3wt.%) which can be released only at higher temperatures [13].

50 .--------------:=:-------,

.~

q,

~

~

g

s ~

~

'0

~ a:

40

PAN + 6.4% MA + 2.3% ITA Heatingrate : 5°C/min.

30

20

10

Temperature. °C

Figure 4. Evolution of gaseous species during the carbonization step for PAN with MA and ITA additives (3); reproduced with permission ofElsevier Science-NL, Amsterdam.

m Post heat treatment As mentioned, the carbons resulting from the pyrolysis of stabilized PAN precursor fibers are non-graphitizable. As a result, further heating of PAN based carbon fibers at very high temperatures (>2500°C) and without any mechanical stretching only slightly improves the ordering ofthe carbon atoms and the fiber stiffness. 9.2.3 From pitch based precursor fibers Pitch is a carbonaceous solid that consists primarily of a complex mixture of polycyclic aromatic compounds [14]. Petroleum pitches are residues of crude oil distillation or of catalytic cracking of petroleum distillates. Coal tar pitches are products of the distillation of coal, whereas synthetic pitches are residues of the treatment of other organic substances. Irrespective oforigin, all pitch products have complex chemical compositions.

(a) Nature ofpitches Pitches consist mostly of fused ring, polycyclic aromatic hydrocarbons with a broad mass spectrum and a molecular mass of 200 to 800 [3] [14]. Since the corresponding molecules are not large enough, they interact only weakly with one another. Most untreated pitches are therefore isotropic.

240

Chapter 9

The pitches ofinterest for preparing anisotropic carbon fibers, known as graphitizable pitches, are characterized bya high degree of aromaticity with an overall composition ranging from 88 to 96 wt.% carbon and from 12 to 4 wt.% hydrogen. Natural pitches also contain some heteroatoms, such as nitrogen, oxygen, and sulfur, which are undesirable, and a small fraction that is insoluble in pyridine orquinoline. The insoluble fraction consists ofcoke, carbon black, ash or water. Quinoline insolubles are determined after extraction at 75°C, and pyridine insolubles by Soxhlet extraction in boiling pyridine (115°C). The combined amount of insolubles should be <2 wt.% . (b) The carbonaceous mesophase stage Isotropic pitches with a high degree of aromaticity are capable of forming an anisotropic liquid crystal phase, referred to as the mesophase orthe carbonaceous mesophase, when heated in the 350-500°Ctemperature range inan inert atmosphere [13-16J. The mesophase first appears as small, urn size, insoluble, liquid spherules within the bulk of the isotropic pitch. Since the mesophase is optically anisotropic, the spherules can be observed inpolarized light by hot stage microscopy, oralternatively on polished samples after cooling. As heating is continued, the mesophase spherules grow in size, come into contact with one another, and progressively coalesce with each other to produce large domains of an anisotropic liquid phase. When the mesophase reaches a critical mesophase concentration (i.e., 60-80%), a phase inversion is observed, and the mesophase now becomes the continuous phase. Highly oriented pitches are known as mesophase pitches. Pitches with a mesophase content ranging from 40 to 90 wt. %are useful precursors to carbonaceous fibers [15J. The mesophase prepared by this route is mainly insoluble in solvents like pyridine or quinoline. Thus, extraction techniques have been extensively used todetermine the degree of the isotropic pitch/mesophase conversion and then the reaction kinetics. At 350°C, the formation of the mesophase is extremely slow. One week is required to achieve a 40 wt.% concentration of mesophase. At 400-450°C, one toforty hours are required to achieve a 50 wt. % concentration of mesophase. Finally, temperatures above 500°C are considered undesirable due to the risk of coke formation. The formation of pyridine and quinoline extractables, t.e., mesophase, has been reported to follow first order kinetics.

= - k(J - x) Ln (1 - x) = - kt x = l-e- kJ

dx/ dt

with

k

-£. /

=

k e I RT ()

(1 ) (2a) (2b) (3)

where x is the mass fraction of pyridine insoluble material; 1-x is that of pyridine soluble material; k is the kinetic constant at temperature T; ko is the pre-exponential term; E. is the apparent activation energy; and Ris the gas constant. For a petroleum pitch, E. :::; 22 kd.rnot' (14)[17]. Studies of aromatic model hydrocarbons have demonstrated that the main reactions, which are responsible for mesophase formation, are molecular rearrangement, cleavage of aliphatic substituents (e.g., methyl groups) attached to aromatic rings, and dehydrogenative

241

Chapter 9

polymerization by a free radical process [17). As the dehydrogenative polymerization progresses, planar polycyclic aromatic molecules are formed with relatively high molecular weights (Mw = 800-1000); interactions between the discoid molecules are sufficient to give rise toliquid crystal ordering. The formation of the mesophase by this route has the disadvantage of requiring rather long processing times. An alternative route is the solvent route [18). Isotropic aromatic pitches contain a separable fraction which, when heated at 230.400°C, develops an optically anisotropic liquid crystal phase in <10 minutes. This mesophase has been called a neomesophase since it is highly soluble in solvents such as pyridine or quinoline, while the mesophase derived by the thermal route isinsoluble. The separable fraction of isotropic pitch isinsoluble in solvents like benzene, toluene, ormixtures oftoluene and heptane. Thus, it can be separated bysolvent extraction from isotropic orheat soaked pitches (Figure 5, band b'). The use of solvated mesophases has been another key breakthrough in the processing of spinnable mesophase pitches [19). A solvated mesophase is a material with a mesophase liquid crystal structure that contains up to 40 wt.% of a solvent. The remaining consists of mesophase forming molecules referred to as mesogens or pseudomesogens, depending on whether ornot they melt upon heating. Suitable solvents are toluene, benzene, halogenated benzene, pyridine, tetrahydrofuran or quinoline. Solvated mesophases are formed from soaked pitches as intermediates during solvent extraction of mesogens (or pseudomesogens) as shown inFigure 5c.

(a) 400°C- 2 hrs,

1 250~ml

(c) 400°C- 12 hrs. 1 250~ m l

(b) 400'C - 6 hrs.

1250~ml

(d) 400°C- 20 hrs. 1250~ml

Figure 5. Polarized light micrographs of mesophase formed in petroleum pitch; reproduced withpermission from the Societe Francaise inParis.

242

Chapler9

Solvated mesophases can be obtained bydifferent procedures. One consists of fluxing the pitch in the solvent (e.g., benzene), removing the insolubles by filtration, and precipitating the mes6gens by diluting the filtrate with additional solvent (the rejection step). Further heating of the rejection mixture to230°Cgives the solvated mesophase. Solvated mesophases can also be obtained from supercritical solvent separated fractions. In this case, the solvated mesophase can be separated as a solid cake by filtration under supercritical conditions. Solvated mesophases exhibit two key properties. First, the solvent lowers the melting point without disrupting the liquid crystal state. As a result, fibers can be spun from solvated mesophase precursors at lower temperatures. Second, solvated mesophase fibers can become unmeltable on loss ofsolvent (19). The carbonaceous mesophase belongs to the family of discotic nematic liquid crystals. Generally speaking, these liquid crystals are formed from discotic molecules which are more or less planar, and are comprised of a rigid (usually aromatic) core having lateral chains. Discotic molecules can be stacked in piles ofinfinite lengths byvarious types ofarrangements (20). Mesophase spherules appear within isotropic pitches upon heating, as a result of dehydrogenative condensation of polycyclic aromatic molecules when the size of the molecules becomes sufficiently large. Their weights range from 400 to 3000 with a number average molecular weight of about 1400 [21]. Large discotic molecules consist of small aromatic regions bonded together by non-aromatic bridges and aliphatic substituents on aromatic rings (Equation 4).

CH,

(4)

Discotic molecule Mllecularweight = 1203 CJH (at) = 1.50 H~)H.....(at) = 1.10 C~)C"., (at) =5.7

Discotic molecules within a spherule lie roughly parallel to one another and perpendicular to the polar diameter (16). Further, they splay outward to lie perpendicular to the spherical interface with the isotropic pitch (16). Discotic molecules tend to form distorted and curved

Chapter 9

243

columns, the discotic molecule planes being slightly misoriented, Le., they lie within 22-25° from the mean layer orientation [21 -22]. When the elementary spherules of mesophase coalesce to form anisotropic domains, the microtexture of the carbonaceous mesophase becomes more complex. Disclinations (rotational defects) in the arrangement of the discotic molecules are often present. Disclinations in 2-D media play a role similar to that of dislocations in crystals and the evolution of gas bubbles results from condensation reactions in a medium which is still fluid. Disclinations, which occur during the mesophase formation, remain after carbonization and are key to understanding the relationships between the microtexture and the properties of carbon fibers formed from mesophase pitches.

Precursor pitch, e.g. petroleum p~ch

Heat treatment

4QO'C, 14-32h

Spinning I drawing ~

P~ch fibers w~h 95% neomesophase

a

b

b'

c

Figure 6. Fabrication of mesopiteh fibers, (a) thennal process without mesophase extraction, (b) solvent extraction process with heat treatment, (b') solvent extraction process without heat treatment, (e) solvated mesophase process.

(c) Spinning and stabilization Pitches are thermoformable and suitable, in the molten state, to be melt extruded through the orifices ofa spinneret. Upon heating, the apparent melt viscosity of a pitch initially decreases within a narrow temperature range (the softening range), then passes through a broad minimum, and increases sharply above 450-500°Cas bulk mesophase isformed [23].

244

Chapter 9

Mesophase pitch

Pumpingzone Melting zone

Die head

Filler

Extruder

Spinneret

Rota ting sc rew Que nch air

- -.-

.- - - - -..-

-------~

Figure 7. The Brooks and Taylor Model ofthe structure ofa mesophase spherule [16); reproduced with permission.

l--

--'

Figure 8. Melt extrusion of mesophase pitch fibers with a rotation screw and multi hole spinneret (6); reproduced with permission from Noyes Publications, Park Ridge, NJ.

The precursor is melled in an extruder which pumps the melt into a die head equipped with a filter and a multihole spinneret [5-6] [24]. As the precursor fibers exit the spinneret holes they cool and solidify, and are drawn before windup. The window for achieving successful and continuous fiber formations is small. The temperature dependence of the viscosity is large and the failure strength of the mesophase pitch fibers is low (30-40 MPa). Thus, the extrusion temperature must be precisely controlled. The largest diameter reduction occurs within less than 1 cm of the spinneret surface. Spinning of a mesophase pitch with 40-80% mesophase requires a melt viscosity of 10-40 MPa.s (or log 2.6 poise). This viscosity corresponds to a fiber forming temperature of 340380°C, and requires a winder speed ranging from 100 to 500 m/min [15] [24]. Extrusion or melt spinning of 100% mesophase pitch melts can be carried out athigher viscosities and with higher winder speeds [6] [25-26]. As-spun mesophase pitch filaments exhibit a high degree of preferred orientation of the discotic molecules along the fiber axis, because the mesogen molecules are oriented under low shear stresses and because of the flow pattern of the melt. The degree of preferred orientation depends on the viscosity of the mell and the final diameter of the fiber. For 100% anisotropic mesophase coal tarpitch, the best results are obtained with a low viscosity melt and small, I.e., 10 IJm diameter, fibers [25]. As-spun mesophase pitch fibers consist of large and elongated anisotropic domains aligned almost parallel tothe fiber axis [15] [251. Since a high degree oforientation isachieved during melt extrusion, they do not need to be further stretched. Moreover, the transverse microtexture of these fibers can be tailored by controlling the flow of the discotic molecules in

Chapter 9

245

the die head and spinneret capillaries. The nature ofthe transverse microtexture isaffected by the use of spinneret capillaries with non-standard shapes, an optional stirrer above the capillaries, capillaries containing porous media, and close control of melt temperature and melt viscosity 126-27]. Fibers which were spun either from a heat soaked pitch containing >85 wt.% mesophase 115], or from a solvated mesophase prepared from high melting mesogens require little or no further stabilization in an oxidative environment 119]. Most mesophase pitch fibers must be stabilized in an oxidative environment. Oxidative stabilization is achieved through a mild oxidation ata temperature slightly below the softening point ofthe mesophase precursor fiber in air, oxygen, or ozone. Oxygen uptake is about 6-7 at.% . Oxygen reacts with aliphatic chains to further aromatize and crosslink mesophase molecules [29]. However, if only the surface isoxidized, oxidation will yield a skin/core microtexture. (d) Carbonization and graphitization

As-spun and stabilized mesophase pitch fibers have no functional utility but serve as the solid precursor fibers from which mesopitch (MP) based carbon fibers are derived. The conversion of mesophase pitch fibers into carbon fibers is carried out in an inert atmosphere at temperatures above 500·C. Initially, free radicals [17] are formed by a dehydrogenation condensation mechanism and the formation and evolution ofgaseous species isaccompanied by weight loss. Most of the gaseous species are formed and evolve at 1000·C, except hydrogen which isstill formed in small amount above this temperature. Carbonization of mesophase pitch yields highly graphitizable carbon due to the structure and arrangement of the mesogens. A very short exposure of carbon to very high temperatures in an inert environment yields the three-dimensional atomic order ofgraphite. Graphitized fibers with very high axial moduli are therefore obtained in a few minutes from mesopitch based carbon fibers by acontrolled heating cycle between 2500 and 3000·C in Ar.

9.3 Structure ofcarbon fibers Structurally, all carbon fibers consist of graphene layers which have been oriented along the fiber axis during processing. Due to the structure of a given precursor and the specific processing conditions, the graphene layers are often distorted, bent or twisted, and their morphology, size, shape and arrangement can differ from fiber to fiber. As a result, pitch based carbon fibers exhibit different microtextures, particularly in cross section which, in turn, are responsible for the variety ofproperties which can be achieved. 9.3.1

Structural parameters

Carbon fibers scatter both x-rays and electrons, and the corresponding diffraction patterns can be used to assess structural parameters [4] [31). The diffraction patterns of carbon fibers contain a limited number ofreflections, i.e., usually a strong 002 reflection and (when present) weaker 004, 10 and 11 reflections, due to the turbostractic character of the carbon. The 001 and hk reflections become less diffuse as the heat treatment temperature (HIT) is increased. Finally, the separation of the 10 reflection into its 100 and 101 components and the appearance of the 112 reflection are observed only for true graphite fibers, prepared from mesopitch fibers and formed at3000·C.

246

Chapler9

The size of the scattering domains, La and le, and their orientation with respect to the fiber axis, Z (also referred to as arc opening, A,), are derived from diffraction patterns by a proper treatment of the intensity data in the equatorial, meridional and azimuthal directions, respectively (Table I). L: is calculated from the half maximum width, B, of the 002 (or 004 when present) reflection, according to the Scherrer formula, le = 0.9 IJBcose, where A is the wavelength and e the Bragg angle. It characterizes the thickness of coherently stacked carbon layers. Similarly, the apparent size, La of the carbon layers in the coherent domain is derived from the half maximum width, B(hk), ofthe diffuse hk rings.

Table 1. Structural parameters of PAN and MP based carbon fibers [31-33J. Fiber PAN-based (liT) PAN-based (HM) MP-based (HM) MP-based (UHM)

d OO2 (nm)

XRD 0.360 0.350 0.342 0.337

L, (nm) XRD SAD 1.7 1.4-2.2 6.0 3.0 12.5 27

L" , (nm)

XRD 3.9 9.8 5.2 9.2

L,.L(nm)

XRD SAD 2.7 2.0 8.0 3.5-4.5 9.8 11.9

Z(")

XRD 41 20 16 8

SAD 35-40 30-37

The meridional width is used tocalculate the apparent length parallel tothe fiber axis, La/I, and the equatorial width the length perpendicular to the fiber axis, La l , by the Ruland and Warren formulas, respectively [1] [4] [32]. The interlayer spacing doo2, is derived from the 002 reflection along the equatorial direction. Itcharacterizes the turbostratic feature ofthe carbon. A parameter characterizing the misorientation of the carbon layers with respect to the fiber axis, Z, is derived from the half maximum width of the azimuthal distribution intensity curve along the 002 arc (arc opening). The coherent domains are isometric, i.e., the values of ~I and Lal are of the same order. Their size is higher for mesophase pitch based carbon fibers, which are prepared at higher temperatures, than for PAN based carbon fibers (Table I). Furthermore, the interlayer spacing doo2, is lower for the former than for the latter. These important differences are self-evident. The carbon structure formed during the carbonization of PAN based precursor fibers is not graphitizable, but the structure formed during the carbonization of mesophase pitch based carbon fibers isgraphitizable. As a result, the densityof PAN based carbon fibers, ranging from 1.76 g/cm 3 for HT fibers to 1.87 g/cm 3 for HM fibers, is lower than that of mesopitch (MP) based carbon fibers, i.e., 2.02.20 g/cm 3• As a result, the density of MP based carbon fibers approaches that of the ideal graphite single crystal, Le., 2.278 g/cm [8]. Finally, a more preferred orientation of the coherent domains is achieved with MP based carbon fibers than with PAN based carbon fibers and it is improved for MP based fibers byraising the heat treatment temperature (HTT) and thereby reaching Z-values as low as 5°. Raman spectroscopy and magnetoresistance data provide further structural information. The Raman spectra of carbon fibers are comprised of two main peaks, the E20 Raman allowed graphitic (G) peak at 1580 crrr' and a disorder induced (D) peak at 1360 em' [34-35]. A dramatic change isobserved in the Raman patterns when moving from PAN based HT fibers, which are characterized by broad and overlapping Dand Gpeaks, tomesopitch based carbon fibers treated athigh temperatures. The latter exhibit two well separated Dand Gpeaks. The

Chapter 9

247

D peak almost vanishes for true graphite fibers. Similarly, transverse magnetoresistivity (ilp/po) is positive only for true graphite fibers, e.g., for P100 and P120 mesopitch based fibers made by Amoco, but negative for all the other carbon fibers. 9.3.2 Microtexture In polycrystalline materials, the microstructure is the arrangement of the grains or crystallites. Polycrystalline materials have texture when there is a preferred orientation of the grains. Generally, most carbon fibers have no grains or crystallites. The basic microtextural unit or SMU [11] is the graphene layer or a smallturbostratic stack of aromatic carbon layers. These SMUs tend to group together edge-by-edge, forming wrinkled and folded sheets (like crumpled sheets of paper), which lie approximately parallel to the fiber axis, and forming elongated pores. The microtexture ofthe fiber is the arrangement ofthese sheets ofcarbon in both the longitudinal and transverse directions. Its complexity depends on the nature of the precursor and the processing conditions [4] [10-11] [31] [37). (a) PAN based high tenacity carbon fibers The microtexture of PAN based high tenacity (HT) carbon fibers is very complex [4-11]. The basic microstructural units (SMUs) are associated edge-to-edge in a zigzag form with tilt and twist boundaries, forming wrinkled sheets [4] [11]. The misorientation angles are about 40°. Individual sheets are oriented roughly parallel to the fiber axis but appear highly folded in a cross sectional view. The wrinkled carbon sheets are entangled in an intricate manner, forming the walls of elongated pores. When two sheets are sufficiently close to each other, transverse bonding may take place when heteroatoms (such as nitrogen) or/and tetrahedral bonds are formed atthe SMU boundaries.

(b)

PAN based high modulus carbon fibers

The microtexture of PAN based high modulus (HM) carbon fibers is more regular than that of PAN based high tenacity (HT) carbon fibers. It was first described with fibril [1] [31) and then with lamellar models [4) [31] (35). The lamellar model is based on crumpled and entangled sheets of turbostratic carbon lying almost parallel to the fiber axis. The graphene sheets are oflarger size (25-50 nrn) but they are still bent along the fiber axis and folded. The transverse microtexture isnot homogeneous. The transverse radius ofcurvature of the graphene layers, rt, increases from the core of a fiber to its surface, and graphene sheets near the fiber surface usually lie parallel to it. This structure is the cause of the notable skin/core effect and the low reactivityof the fiber surface. The reactivity of aromatic carbons is lower in-plane than at the layer edge. The turbostratic character of carbon in PAN based HM carbon fibers and the occurrence of elongated pores explain their low density (1 .8-1 .9g/cm3) .

(c) Mesopitch (MP) based carbon fibers Mesopitch based carbon fibers exhibit a variety ofmicrotextures depending on the structure of the specific precursor fiber and the spinning conditions. Mesopitch based fibers prepared from 100% mesophase pitch have a more homogeneous microtexture than those prepared from mixtures of anisotropic and isotropic pitches. Further, disturbing the flow pattern of the

248

Chapter 9

mesogen molecules during extrusion can affect the transverse microtextures. The most frequently observed microtextures in commercial fibers are radial, oriented core and random microtextures. The microtextures of P55S, P75S and P100 fibers (Table II) are mixtures of three components: a polycrystaliline graphite phase, a microporous carbon phase and a lamellar turbostratic carbon phase similar tothat of PAN based HM fibers. The amount ofthe graphite phase increases from the P55S to the P100 fiber [4]. Furthermore, mesopitch based graphite fibers [32] from the Amoco P-series, i.e., P-100 and P-120, have an oriented core microtexture, and mesopitch based carbon fibers from the Kashima Carbonic series, l.e., HM50 to HM-80, have a radial folded microtexture. Only the P-100 and P-120 fibers contain a graphite phase. The fibers from the Carbonic series are turbostratic like PAN based HM fibers. The microtexture of fibers from the P-series have flat carbon layers lying almost perfectly along the fiber axis whereas that of fibers from the Carbonic series have folded carbon layers oriented radially with a higher misorientation with respect to the fiber axis.

I I

Long itudinal section External surface

Figure 9. Microtextures ofHM-fibers derived from PAN [4); reproduced with permission from Elsevier Science-NL, Amsterdam.

249

Chapter 9

Table II. Properties of MP-based carbon fibers [41]. Fiber code P25 P55 P75 P75S PI00 P120 P130(X)

Diameter (pm)

10 11 10 12 10 12 11

Density

p (g.cm")

2.00 2.00 2.16 2.18 2.19

Modulus (GPa) 170 379 490 524 765 827 923

Strength (GPa) 1.3 1.9 1.89 1.75 2.41 2.2 2.87

Cross section C C C C 0, C C, C, P

°

Radial microtexture WL RD WL OC OC OC OC

Direct evidence for the existence of true graphite layers in mesopitch based carbon fibers is obtained [39-40] by in-situ dark field TEM imaging on fibers which are thin and therefore are transparent to the electron beam. This approach includes the use of the 112 reflection characteristics to determine the three-dimensional atomic ordering in graphite; the microstructure of the fibers consists mainly ofgraphite layers. Each layer exhibits a mosaic of <2 nm thick and >50 nm long graphite platelets oriented along the fiber axis. Most mesopitch based carbon fibers from the P-series (P25, P55, P75S, P100, P120, and P130X), except P25, P55, and P75, exhibit an oriented core microtexture [41]. Microtextures are domains ofmicrotextural needle-like units, 0.1 -0.2 IJm in diameter and up to 100 mm long. These domains are parallel to each other and to the fiber axis, and consist of graphene layers arranged in two modes. In dense domains, the graphene layers are stacked predominantly parallel to the fiber axis with little void space. In microporous domains, they have little or no preferred orientation and many pores are present. When viewed in cross section, the microtexture consists of an equatorial zone where the orientation of layering is continuous, two polar zones of chaotic structure where layering is discontinuous, and two transitional zones. Within the P-series of carbon fibers, the volume of micropores decreases from 25% for P75 to 1% for P1 00, and <1 % for P120 and P130X. Graphite ispresent in fibers with the highest stiffness, and the composite microtexture of most of the fibers results from the precursor composition and rheology. Most of these are two phase systems (Chapter 9.2.3). The dominant fraction of the pitch, the mesophase, is the precursor of the dominant domain of the carbon fiber, the dense domain. The minor fraction of the pitch, the isotropic phase, is the precursor of the minor domain of the carbon fiber, the microporous domain. In dense domains, the graphene layers are parallel to the fiber axis because melt spinning and drawing facilitates a very effective flow alignment of the polyaromatic mesogens. In microporous domains, the absence of preferred orientation ofthe graphene layers iscaused by the structure ofthe isotropic pitch fraction. The microtexture of carbon fibers obtained from the E-series of 100% anisotropic, mesopitch based carbon fibers produced by DuPont was found to depend on disclinations [42]. Conceptually, disclinations are introduced into the liqUid crystal precursor during extrusion, remain entrapped during the subsequent high temperature steps, and control the microtexture of the resulting carbon fiber. For example, the microtexture of a fiber is radial when the density ofdisclinations in a given pitch is normal. A local loss ofmicrotexture can occur when the density of disclinations in a given pitch is very high, possibly as a result of disturbing the

250

Chapler9

liquid crystal flow pattern during extrusion. In this case the carbon layers are less densely packed and possess higher curvatures and elongated pores oflower transverse sizes. Transverse microtextures of mesopitch based carbon fibers are achieved and controlled by changing the flow pattem ofthe mesogen molecules during extrusion, and by the stabilization conditions [7J [26-281 [43J. The flow pattern is affected by melt temperature, viscosity and uniformity, and bythe shape and dimension of the spinneret hole. Radial microtextures are formed when the extrusion is performed with a circular capillary, without a stirrer, and at a relatively low temperature, i.e., at a relatively high melt viscosity. Quasi-onion or random microtextures are formed when a stirrer is used and onion microtextures are obtained with melts having a relatively low viscosity, i.e., at a relatively high temperature. Capillaries filled with a porous body alter the structure of the mell, yield fibers with random microtextures [7J and lower the degree ofgraphitization [27J. Transverse microtextures can also be achieved by using properly designed non-circular spinneret capillaries orholes and a properly adjusted mell viscosity. A line origin microtexture is obtained when 100% anisotropic, methyl naphthalene containing, mesopitch based carbon fibers are spun at a low temperature with a trilobal spinneret. The line origin microtexture of the resulting trilobal carbon fiber is shown in Figure 10. However, when the extrusion is performed at too Iowa viscosity, the cross section becomes circular, and the resulting fiber will either retain a transverse microtexture [27J or exhibit a random onion-like microtexture [26J. A skin/core microtexture can be developed not only in PAN based carbon fibers but also in mesopitch based carbon fibers by proper control of the oxygen uptake during the stabilization step, e.g., for 10 pm diameter carbon fibers derived from coal tar based mesopitch precursor fibers [43]. 9.4 Properties of carbon fibers

At room temperature, carbon fibers exhibit an approximately linear elastic behavior under tensile loading up to failure. However, some non-linearity in the stress-strain curve was recently reported. The slope increases slightly as the fiber is strained [44-46J. After mechanical properties, the most important properties are those which characterize the electrical transport behavior ofcarbon fibers. Up to now, high strength (HS) or high tenacity (HT) carbon fibers were obtained with high strength or tenacity levels (3000-7000 MPa), but low stiffness or Young's modulus (230-250 GPa). In contrast, high modulus (HM) or ultrahigh modulus (UHM) carbon fibers have been obtained with very high modulus (350-900 GPa), but only low strength (2000 MPa). Recent advances in controlling the microtexture have afforded experimental UHM carbon fibers derived from mesopitch based carbon precursor fibers with high strength (4000 MPa) and ultrahigh modulus (900 GPa). (a) Young'smodulus

The longitudinal Young's modulus of a carbon fiber is directly controlled by the orientation of the graphene layers along the fiber axis. Its upper limit, which would be achieved for a perfect orientation (Z= 0), is the Young's modulus Ell = Cl1 = 1060 GPa for a graphite single crystal in

251

Chapter 9

the basal plane. E-values as high as 923 GPa have been reported for some mesopitch based carbon fibers with Z"" 5° (Table II).

•

Flat layer

Line origin

Radial folded

•

Radial

Onion skin

•

Random

Quas i-onion

Figure 10. Transverse microstructures observed in fibers spun from mesophase pitches [7]; reproduced with permission from Noyes Publications, Park Ridge, NJ.

- -- --Lc----7" I

~~

I I I

Leg

Figure 11. Microtexture model of the oriented core of a very thin carbon fiber derived from MP; reproduced with permission from Chapman and Hall, london.

252

9.4.1

Chapter 9

Mechanical properties

For mesopitch based carbon fibers, the planar mesogen molecules are already aligned along the fiber axis (with Z :::; 15°)during the extrusion ofthe liquid crystal precursor. They are then locked into place during the stabilization step. The preferred orientation is further improved (with Z = 5-15°) during the treatment at high temperature (2500 < HTT < 3000°C). For PAN based carbon fibers, the preferred orientation results from stretching the precursor fiber during both the steam stretching and the stabilization steps (Figure 3). Its modulus is significantly improved during the high temperature treatment (with Z = 20° for PAN based HM carbon fibers), as shown inTable I. Further, stretching the fiber during the high temperature treatment (hot stretching >2000°C) is an efficient way to further improve the preferred orientation (and therefore the modulus, E). The fiber becomes plastic under such conditions and experiences both a longitudinal elongation and a reduction in cross-sectional area [47). Finally, Young's modulus regularly increases asthe heat treatment temperature (HTT) increases irrespective of the nature of the precursor and whether ornot hot stretching was employed.

5

4

(a)

UHS

D DIM

290-300 GPa

230-250 GPa

a.


a:-

e

.c

E

HS

III

500

3

400 III

a.


S

'5 <: l!!

'0

0

iii

DUH

.9! 2 'iii

E 200

<:

_f/)

Cl

<: ::;,

§?

~

100

o

\ , - J . -_ _--'-

-'--_ _--I...

l.-.-_ _- l

0

HTI, OC Figure 12. Young's modulus and tensile strength of PAN based carbon fibers as a function of heat treatment temperature [3] [45]: reproduced with permission from Elsevier Science-NL, Amsterdam.

Attempts have been made to correlate Young's modulus (Figure 13) with the various parameters characterizing the preferred orientation of the graphene layers [3). Accordingly, the modulus increases asa function of preferred orientation when moving from low modulus, isotropic pitch based carbon fibers (E :::; 50 GPa) to HM/UHM mesopitch based carbon fibers (350 < E < 900 GPa) [3). Young's modulus, corrected for zero porosity (Ecorr =

253

Chapter 9

(d~IJdmoas) 'Emoas, with d~te = 2.23 q.crrr' for mesopitch based carbon fibers), is related to preferred orientation, by:

(5) where Eg and Gg are the Young's and shear moduli of the graphite single crystal, respectively (Eg = Cl1 '" 1060 GPa, Figure 13; Gg = C« '" 5 GPa), and is an orientation parameter with = 0.722 sin'Z [44-45]. The Young's modulus of PAN based HM carbon fibers varies linearly with the reciprocal ofthe carbon layer mean size, La [35]. Values of some elastic characteristics for various mesopitch based and PAN based carbon fibers can be computed from those ofsingle fiber composites with a resin matrix [44].

1000 \\\\\\ I Pitch

'"

800

0.. (!)


A Pitch

1111111 PAN

600

"S

GY80 GY70

'C

0

E

,'"Ol

P55

400

c:

::l

~

UHM HM

.<W'.@'jM

~HT

200

0 0

0.2

0.4

0.6

0.8

1.0

Prefer red orientation

Figure 13, Young's modulus and preferred orientation parameter of PITCH and PAN based carbon fibers; reproduced with permission from Elsevier Science-NL Amsterdam and Chapman &Hall.

(b) Tensile strength In brittle materials such as carbon fibers, tensile strength (crR) is controlled by defects introduced in the materials during processing orland handling. For a defect of size a" it is given bythe Griffith relationship:

a R = J2EYa ;rae

(6)

where Eis Young's modulus and y. the apparent surface energy. Since a carbon fiber usually contains many defects of different sizes, its tensile strength exhibits a statistical character. Assuming one single population of flaws, the probability of failure, PR, of a sample of volume, V, loaded ata stress level, c, is often expressed with the Weibull distribution function as:

Chapter 9

254

(7a) Inthis equation, cr. is the stress below which PR =0 (cr. isoften taken ascru =0). o, is a scale parameter and m is the Weibull modulus that characterizes the width of the distribution. The right hand side of equation (7a) should bedimensionless which means that crodoes not have the dimension of a stress. One way to take this feature into account is to normalize the volume (e.g., to replace V by VNo where Vo is a reference volume). Additionally, if one considers a lotofcylindrical fibers of length L with a constant diameter, V can bereplaced by L (or in a similar manner by UL. where L. is a reference length). Assuming cr. = 0, equation (7a) can be rewritten in a linear form and used toderive the value ofmfrom tensile test data.

LnLn(J/ (J - PR ))

= ml.na + (LnV - mLna,,)

(7b)

First, a group of N fibers is tested. N ranges from 20 to 40, and all fibers have the same volume i.e., the same length and diameter. Then, the tensile strength values are ranked in ascending order and the failure probability is calculated utilizing an appropriate estimator, e.g., 1- PR =(i-0.5)/N or 1 - PR =i/(N+1) where i is the rank, for each o value [47]. If the o values obey a Weibull distribution, the plot of LnLn (1/{1-PR)) vs Lncr should yield a straight line whose slope is equal to m byequation (7b). The literature values of m for carbon fibers usually fall within the 5-10 range. The mean strength value can be assessed in different ways. It is equal to:

a = a"r[l+(J /m)] W

(8)

VI/III

In this equation, r is the gamma function. V is replaced by L when the fiber diameter is constant [48). The mean strength can also be characterized by the stress value cr{0.5) corresponding to PR = 50%. The actual values of o; and cr{0.5) are close to one another [4849). Introducing cr{0.5) in equation (7b) and assuming that the fibers are of constant diameter and the Weibull parameters (rn, cro) donot depend on L yields:

= - -I

LnL + constant (9) m Equation (9) shows that the variations of Ln c (0.5) as a function of LnL can be represented with a straight line, which can be used to derive m. This equation also shows that the strength of a fiber increases with decreasing gauge length. The value given by fiber producers usually refers toa gauge length ofone inch (L = 25.4 mm). Theoretically, the linear relationship can beused toderive the failure strength of a fiber atvery short lengths, e.g., the critical length, Ie, for a carbon fiber/resin composite of the order of 0.1 - 0.3 mm. Failure strength at critical lengths can also be derived from the in-situ fragmentation test (ISFT) performed on a model 1-D composite with a single fiber embedded in a transparent resin [4850). Experimental tensile strength data fora lotoffibers ofsame length, L, can not always befitted with a single straight line. Such a situation usually corresponds to the occurrence of two populations of defects, for example, a population of internal flaws and a population of surface flaws. Lna (0.5)

The nature and the origin ofthe defects which control failure in carbon fibers are discussed in detail in reference [1]. They include diameter and cross section variations, both along the length of the fiber and from fiber to fiber in a yarn bundle, internal flaws and surface flaws.

255

Chapter9

Internal flaws include voids and surface flaws include dust particles carried through from the precursor fiber made from dry orwet spinning solutions. Internal flaws may also include voids of various sizes related to the evolution of gases during stabilization and pyrolysis as well as extended structural defects. Surface flaws also include surface pits and microscratches resulting from handling the precursor and resulting carbon fiber. The size of critical defects can be derived from equation (6) and rewritten as equ.(6b}, assuming that y. remains constant (with y. "" 4.2J/m 2 for carbon). Table III gives the order of magnitude ofac for representative carbon fibers.

2Ya

E

a = _. c

(6b)

"(J'~

The most severe flaws in carbon fibers can be largely avoided byusing clean precursors and by working under clean room conditions. In an experimental process, solid particles were eliminated by filtering the liquid precursor, and bubbles of gas were removed by centrifuging. Experiments show that PAN fibers [52] formed under such conditions are stronger than control fibers fabricated in commercial conditions. The strength of the resulting carbon fibers which were derived from PAN based precursor fibers increased as the heat treatment temperature (HTT) was raised, whereas that ofthe control fibers first increased and then decreased. Table III. Critical defect size calculated according to equation (6'). Carbon fibers Ultrahigh strength (UHS) H igh strength (HS) High modulus (HM) Ultrahigh modulus (UHM)

if (GPa) 7-5 3.5 2.7 2.4

E(GPa)

275 230 400 800

a, (run)

15 - 29 50 146 370

The failure strength of carbon fibers also depends on the microtexture of the fibers and thus on the nature of the precursor and processing conditions. When a microcrack is initiated by a defect, it must propagate toinduce failure of the fiber. A crack can easily propagate between the carbon layers of graphite but crack propagation in a perpendicular direction is more difficult. This feature means that fibers consisting of well organized carbon atoms, i.e., of straight parallel graphene layers of large size such as those observed in mesopitch based fibers, provide easy crack propagation paths. Conversely, fibers characterized by highly entangled and folded graphene layers such as those formed from PAN based precursor fibers exhibit a higher resistance to crack propagation and thus a higher failure strength. In PAN based HS carbon fibers, the occurrence of crosslinking between the graphene layers (which are of very small size), such as that related to Sp3bonds or heteroatoms at the BMU's edges, is assumed to render crack propagation more difficult. The strength of carbon fibers fabricated from mesopitch based precursor fibers is usually lower than that of the PAN based counterparts. However, it can be improved without significantly lowering Young's modulus by modifying the transverse microtexture. Fibers with pure radial or oriented core transverse microtextures (Figure 10) have lower strength than those with random or random onion transverse microtextures [26) [41]. HM-80 Carbonic mesopitch based (MP) carbon fibers are stronger than P120 MP based carbon fibers (cf = 3.5 vs. 2.4 GPa) but have equal stiffness (800 GPa).

256

Chaplerg

MP based fibers with oblong cross section, show a composite transverse microtexture with a core consisting of planar and parallel graphene layers surrounded by a zone of curved turbostratic carbon layers, which might impede crack propagation. These fibers have an extremely high tensile strength, 5.85 GPa, and an ultrahigh Young's modulus, 650 GPa [11]. It isclear that a new generation ofmesopitch based carbon fibers has emerged. These fibers have a random, radially folded, ora specific, non-circular, transverse microtexture. Thus, they have ultrahigh stiffness and ultrahigh strength [7] [11] [26] [46] [51]. Temperature is one of the processing parameters which has a pronounced effect on the failure strength of carbon fibers. This effect is quite different for MP based and PAN based fibers. As the heat treatment temperature (HIT) is raised, the strength of a mesopitch based fiber, tested atroom temperature, progressively increases [3] [38] [54-55]. Conversely, that of most PAN based fibers undergo a maximum atabout 1300-1500°C(Figure 12). The causes for this decrease of crR are still a matter of speculation. New and more severe defects can be created within this temperature range from impurities [52]. PAN based fibers also contain large amounts ofheteroatoms including nitrogen. Thus, it is possible that carbon layer crosslinking [4] and nitrogen loss which occur in the 1200-1600°C temperature range are partly responsible for the observed strength loss [3] [56]. (c) Compressive strength

Carbon fibers exhibit up to 50% lower compressive than tensile strengths. The compressive strength offibers iscalculated from compressive test data recorded from composite materials. Acritical analysis ofthe compressive tests and compressive strength data for a variety of PAN based and mesopitch based carbon fibers can be found in reference [43]. The compressive strength ofcarbon fibers decreases with increasing stiffness [7] [57]. For a given stiffness, the compressive strength is higher for carbon fibers made from PAN fibers than for those made from mesophase pitch precursor fibers. The failure modes under compressive loading are different [7] [44] since compressive strength depends on microtexture. It is high for PAN based HS carbon fibers having a microtexture consisting of entangled carbon layers, and low for those where the graphene layers are larger and well oriented parallel tothe fiber axis. Finally, most carbon fibers, including mesopitch based UHM fibers, which show a pronounced graphitic character are extremely weak in compressive loading. The compressive strength of weak mesopitch based carbon fibers can be improved, as already discussed, by tailoring their transverse microtexture and the shape of their cross section. Fibers with randomly or radially folded microtextures are stronger than fibers with layered microtextures and multilobal fibers are assumed to be stronger incompressive loading than their round counterparts. Finally, ion (e.g., boron) implantation also increases the compressive strength ofcarbon fibers. (d) High temperature properties

Tensile strength and modulus of carbon fibers increase with increasing test temperature to about 1500°C[58]. Carbon fibers begin tocreep above about 2000°C [47] [51] [59]. Creep is actually used during hot stretching to increase their stiffness or Young's modulus.

257

Chapter 9

Experiments performed on PAN based fibers at a heat treatment temperature (HIT) of 1300·C) have shown that the creep is primary and that the creep rate decreases regularly with time. Furthermore, creep appears to be logarithmic and the inverse of the creep rate increases linearly as a function oftime [59]. 9.4.2 Thermal and electrical properties Thermal expansion, electrical resistivity and electrical conductivity are key properties of carbon fibers which strongly depend upon their structure.

(a)

Thennalexpans~n

The thermal expansion of anisotropic fibers is usually characterized by two coefficients of thermal expansion (CTEs), the longitudinal CTE, ai, and the radial CTE, a" The direct measurement of al is straightforward but that of a r is difficult since the gauge length is the fiber diameter. Further, the use of one single a, CTE to depict the transverse thermal expansion of carbon fibers is only appropriate for those fibers which have a transverse microtexture with axial symmetry, an assumption that isclearly not correct for all fibers. The longitudinal CTE ofcarbon fibers (Table IV) is low and negative near room temperature. This uncommon feature is closely related to the intrinsic properties of graphite (all = - 1 .5 X 10-6 ·C·' vs. a .l= 27 x 10-6 ·C' for a single crystal) and tothe preferred orientation ofthe carbon layers along the fiber axis. Consequently, the value of at is much higher for mesopitch based carbon fibers with a graphitic character than for PAN based HS carbon fibers. Above 600·C, it becomes positive and increases with increasing temperature [59-60].

2.5

333 mg/filament

2.0 c:

.~ 1.5 c:

~

W

1.0

0.5

o

20 mglfilament

2

3

4

5

Time. min

Figure 14. Creep of PAN based carbon fiber tows at selected temperatures and applied loads in an inert atmosphere [59).

258

Chapter 9

The radial eTE of selected carbon fibers has been measured upto 800 e by a hot stage TEM technique using surface defects of the fibers as markers, and between 800 to 1600 0 e bylaser diffraction [60]. The data are considerably scattered because of the small fiber diameter and the inhomogeneity of the transverse microtexture. Nevertheless, the transverse thermal expansion ofthose carbon fibers which have been studied is large and positive (Table IV). 0

Table N. Coefficients of thermal expansion of PAN-based and MP-based carbon fibers.

Fibers T3OO(PAN) T40(pAN) TSO (PAN) P55S(pAN) P75S(PAN) PlOOS(pAN) P12OS(PAN) 3K75(MP)

p (g/an')

1.79

E(GPa) 230

2.00 2.04 2.15 2.18 2.14

380 515 700 830 400

a,(10"'C1) - 0.1 to - 0.5 - 0.56(1) - 0.85"' -1.15 P ' - 1.24P ' - 1.29"' -1.33"' - 1.9")

6.8'"

NO NO 11.0P'

12.9'"

9.7'" NO

9.7'"

Graphite is an excellent conductor in the basal plane but an insulator in a perpendicular direction [8]. The electrical and thermal conductivities of carbon fibers are low in fibers which have been fabricated at low heat treatment temperatures (HIT) and exhibit a poorly ordered microtexture, e.g., in the PAN based HS carbon fibers or the fibers spun from isotropic pitch. They regularly increase and become more and more anisotropic as the size (La), the preferred orientation along the fiber axis (2) and the perfection of the stacking of the graphene layers are raised. Ultimately, they trend toward values which are relatively close to those characterizing highly oriented pyrolytic graphite (HOPG), in the UHM mesopitch based fibers treated atvery high temperatures. In PAN based fibers, the electrical conductivity is low, is dominated by the microtextural defects, and has a semiconducting character. It increases when the fibers are stretched during the high temperature treatment and when the temperature is raised, i.e., when the microtexture becomes more and more ordered [1]. Interestingly, the variations in electrical resistivity of PAN based HM fibers appear to be a function of the inverse of the carbon layer size and obey a linear relationship: --I

p; =3.91 + 1510·La

(10)

In this equation, pe is expressed in IJO.m and La in A[35]. The electrical conductivity of mesopitch based fibers remains similar to that of PAN based fibers for a low heat treatment temperature (HIT). It becomes much higher when the heat treatment temperature is raised from 2500 to 3000 o e, i.e.,where graphitization actually takes place. Finally, and as illustrated in Table V, the electrical conductivity along the fiber axis of the best mesopitch based fibers (P 120 and P 140) is still lower than those of highly oriented pyrolytic graphite and copper

[61].

The electrical conductivity of carbon fibers is improved by intercalating donors (e.g., alkali metals) or acceptors (e.g., Br2, eb, F2, chlorides or fluorides) between the carbon layers [1] [62]. These intercalation compounds are more easily formed in fibers which exhibit a highly ordered graphitic structure, te., the mesopitch based UHM carbon fibers. Some intercalation

259

Chapter 9

compounds which correspond to chloride or fluoride intercalatesare stable in air between 25 and 300°C. The formation of intercalation compounds in carbon fibers increases the electrical conductivity byabout a factor often. The thermal conductivity of carbon fibers shows the same general features as its electrical counterpart although there are much fewer data presently available. As shown in Figure 15 and Table V, the thermal conductivity ofthe best mesopitch based UHM carbon fibers (P-100, P-120 and P-140) ishigher than those ofcopper and silver [3) [38) [61).

-~-. Kcu

~:::::::: :: 2

HOPG

__

P1~~

~~ AI- - - - - - - - - - - - - - - - - - - - 100

~ Mg

P750

P55

TSO Ti

P25 T300

Electrical resistivity, u-ohmrn

Figure 15. Transport properties ofcarbon fibers; comparison between the transport properties ofcarbon fibers and those ofmetals [61); reproduced by permission from SAMPE.

9.4.3 Oxidation ofcarbon fibers Oxygen reacts with Sp2 carbons at low temperatures to form gaseous oxides such as CO orland CO2. Thus, the oxidation ofcarbons is accompanied byweight loss, usually expressed as ~mo, where m, is the initial sample mass. Reaction occurs preferentially at active sites, i.e., atthe edges of graphene layers and defects. Oxidation can be accelerated bytraces of catalytic impurities (e.q.,alkali ortransition metals) and residual porosity [8) [63). The oxidation reaction starts at a temperature that depends on composition, structure, and microtexture. For example, the oxidation of carbon fibers derived from isotropic pitch precursor fibers in flowing air starts at about 400°C whereas that of the mesopitch based UHM fibers starts at a higher temperature [64). At a given temperature, the variation of the weight loss, i1m1m o, as a function oftime is linear, to a first approximation, within a rather large domain ofweight loss [65-66). This property is used to characterize the oxidation rate with a kinetic constant, k, defined as:

260

Chapter 9

k

or

1

drn

m,

dt

(11)

= -_.-

where k isthe slope ofthe linear portion of the kinetic curve. Ata given temperature, k ishigh for PAN based HS fibers, which contain a high density of active sites. It is lower for the mesopitch based UHM fibers, which have a low porosity and a graphitic character (Figure 16). Atlow temperatures, the thermal variations of the kinetic constant obey an Arrhenius plot. A clear ranking isapparent among the fibers. Table V. Transport properties of selected materials at room temperature [8]. Electrical resistivity at 25°C (@.m) 0.016 0.017

Thermal conductivity at 25°C (W.m-'K'j

Materials Silver Copper Diamond Pyrolytic graphite, / / ab plane / / c-axis PAN-based fibers along the axis MP-based fibers along the axis

420

385 2000 390 - 4180

2.5 -5 3000 9.5 -18 2.2 - 2.5

2

8 - 70 530 -1180

Fibers prepared at low HIT have poorly ordered microtextures, high oxidation rates and low activation energies (80-112 klrnol'). They include carbon fibers derived from isotropic pitch based precursor fibers and as-received PAN based T300 fibers. Conversely, mesopitch based UHM fibers, which are fabricated at a high temperature, and PAN based T300 fibers after a post heat treatment, Le., annealing at 1600°C, have low oxidation rates and higher activation energies (140-160 kl .mol").

20

'#. 40


E .~ 60

::

80

100

'----'-_-'-_-'-_.L----'==----'-_-'-_-'--_'-----'_--'-_-J

o

30

60

90

120

Time, min

Figure 16. Oxidation rate of carbon fibers [65J; reproduced with permission from Societe Franco-Japonaise des Techniques Industrielles , Tokyo .

Chapler9

261

These results illustrate the decrease in active site concentration as the materials become more ordered. The higher values of E. reported above, (140-160 kl.mol"), lie at the low end of the activation energy range (160-200 kJ.mol·1) where the rate limiting step of the oxidation of carbon is assumed to be a surface reaction . The lower E. values reported for PAN based T300 and carbon fibers derived from isotropic pitch based precursor fibers (80-110 klrnot') might be related toa mixed in-pore diffusionl chemical reaction mode. These fibers are highly porous (p = 1.79 and 1.60 g/cm·3, respectively), and their nm-size pores are being enlarged as soon asoxidation starts [66J. The oxidation of carbon fibers can beinhibited to some extent bythe use of dopants such as boron. Three mechanisms could be involved in the inhibition process: active site blockage resulting from the formation of a boron oxide layer, chemical inhibition by electron transfer, and development of fiber structure/microtexture which is catalyzed byboron. For example, at 700°C, the oxidation rate of the T300 fiber decreases 30 fold when 2000 ppm 8 are added and the P55 fibers having 5% 8 never reach 25% burn-off[67J. 9.4.4 Coated carbon fibers Thin ceramic coatings are applied to carbon fibers to lower their oxidation rate, improve their ability to be wetted by liquid metals, or control the fiber-matrix bonding in CMCs. Silica and glass-forming SiC or 8,C coatings increase the oxidation resistance of carbon fibers. Coatings of TiC or Ti82 significantly improve the wetting behavior of carbon fibers by aluminum. Finally, coating carbon fibers with a material exhibiting a layered crystal structure, such as pyrocarbon, provides a rather simple way to weaken fiber-matrix bonding, in e.g., SiC-matrix composites. Coatings can be deposited on carbon fibers either from liquid precursors, e.g., organometallic species or sols, or from gaseous precursors, e.g., by chemical vapor deposition (CYD) or infiltration (CYI). For example, a silica coating deposited from tetraethylorthosilicate lowers the oxidation rate ofcarbon fiber at650°Cbya factor of 5 and at430°Cbya factor of 30 [68J. Likewise, a 50 nm thick 84C or SiC coating deposited byreactive CYD significantly lowers the oxidation rate ofcarbon fibers [69]. Most coatings are weaker than, and strongly bonded to, the carbon fibers. When a coated fiber is tensile loaded, the coating will fail and undergo microcracking. These microcracks form a new surface flaw population, and in turn control the failure of the fiber if their mean size is larger than that, a, of the intrinsic defects controlling the failure of the pristine fiber. If the size of coating microcracks is of the order of the coating thickness, x, two situations can be envisaged. If the coating is very thin, i.e., when Xc < a, (see equations (6) and (6b)), fiber failure remains controlled by the intrinsic defects of the pristine fiber. Conversely, if Xc > a, the new population of surface flaws controls fiber failure. The failure strength is given by equation (6), where a, is replaced bythe coating thickness, Xc; it decreases asthe square root of the inverse ofthe coating thickness [70J.

9.5 Applications The applications of carbon fibers along with those of ceramic oxide and silicon carbide fibers are discussed in Chapter 12.

Chapter 9

262

REFERENCES [1) (2)

[3)

[4) [5) [6) [7)

[8) [9) [10) [11) [12) [13) [14) [15] [16] [17) [18) [19) [20) [21) [22) [23) [24) [25) [26) [27] [28] [29) [30] [31) [32)

J. B. Donnet and R. C. Bansal, Carbon Fibers, Vol. 10 in Int. Fiber Sci. and Technology Series, 2nd Edit. Revised and Expanded, Marcel Dekker, New York and Basel (1990). K. K. Chawla, Ceramic reinforcements, in ceramic Matrix Composites, chap. 3, 45-125, Chapman and Hall, London (1993). E. Fitzer and M. Heine, Carbon fibre manufacture and surface treatment, in Fibre reinforcements forcomposite materials, A. R. Bunsell, ed., chap. 3,73-148, Elsevier, Amsterdam and New York (1988). A. Oberlin and M. Guigon, The structure ofcarbon fibres, in Fibre reinforcements forcomposite materials, A.R. Bunsell, ed., chap. 4,149-210, Elsevier, Amsterdam and New York (1988). G. Savage, Carbon fibres, in Garbon-carbon composites, chap. 2,37-83, Chapman and Hall, London (1993). D. D. Edie and R. J. Diefendorf, Carbon fiber manufacture, in Garbon-carbon materials and composites, J. D. Buckley and D. D. Edie, eds, ch. 2,1940, Noyes Publ., Park Ridge, NJ (1993). D. D. Edie and E. G. Stoner, Effect of microstructure and shape oncarbon fiber properties, in Garbon-carbon materials and composites, J. D. Buckley and D. D. Edie, eds., chap. 3, 41-70, Noyes Publications, Park Ridge, NJ (1993). H. O. Pierson, Handbook of carbon, graphite, diamond and fullerenes, Noyes Publications, Park Ridge, NJ (1993). I. A. S. Edwards, Structure in carbons and carbon forms, in Introduction to Garbon Science, H. Marsh, ed., chap. 1,1-18, BUllerworths, London (1989). A. Oberlin, High-resolution TEM studies of carbonization and graphitization, in Chemistry and Physics of Carbon, P. A. Thrower, ed.,vol. 22, 1-143, Marcel Dekker, NY and Basel (1989). M. Guigon M., Microtexture and mechanical properties of carbon fibers: relationship with the fiber-matrix adhesion ina carbon-epoxy composite, Polym. Eng. Sci., 31 [17), 1264-1270 (1991). W. Wall, Chemistry and physics ofthe conversion ofpolyacrylonitrile fibers into high-modulus carbon fibers, in Handbook ofComposites, Vol I - Strong Fibers, W. Wall and B. V. Perov, eds., Elsevier Science Publ., 327387 (1982). E.Fitzer, W. Frohs and M. Heine, Optimization ofstabilization and carbonization treatment of PAN fibres and structural characterization ofthe resulting carbon fibres, Carbon, 24, 387-395 (1986). I. C. Lewis, Chemistry and development of mesophase inpitch, J. Chim. Phys. Physico-chim. BioI., 81 [11/12), 751-758 (1984). L. S. Singer, High modulus, high strength carbon fibers produced from mesophase pitch, U.S. Pat. 4,005,183, Jan. 25, 1977. J. D. Brooks and G. H. Taylor, The formation of some graphitizatizing carbons, in Chemistry and Physics of Carbon, P. L. Walkered., Vol. 4,243-286, Marcel Dekker, NY (1968). L. S. Singer and J. C. Lewis, ESR study ofthe kinetics ofcarbonization, Carbon, 16,417423 (1978). R. J. Diefendorf and D. M. Riggs, Forming optically anisotropic pitches, US Pat. 4,208,267, June 17, 1980. W. M. Kalback, H. E. Romine and X. M. Bourrat, Solvated mesophase pitches, US Pat. 5,259,947, Nov. 9, 1993. H. Gasparoux, Liquid crystals and carbonaceous mesophase (in french), J. Chimie Physique, 81 [11/12), 759767 (1984). J. E. Zimmer and J. L. WMe, Disclination structures inthe carbonaceous mesophase, Adv. in LiquidCryst., 5, 157-213 (1982). D. Auguie, M. Oberlin, A. Oberlin and P. Hyvemat, Microtexture of mesophase spheres as studied by high resolution conventional transmission electron microscopy (CTEM), Carbon, 18,337-346 (1980). F. F. Nazem, Flow ofmolten mesophase pitch, Carbon, 20[4], 345-354, (1982). D. D. Edie and M. G. Dunham, Melt spinning pitch-based carbon fibers, Carbon, 27(5), 647-655 (1989). T. Hamada, M. Furuyama, Y.sajiki, T.Tomioka and M. Endo M., Preferred orientation ofpitch precursor fibers, J. Mater. Res., 5[6),1271-1280 (1990). I. Mochida, Seong-Ho Yoon and Y. Korai, Control of transversal texture in circular mesophase pitch-based carbon fibre using non-circular spinning nozzles, J. Mater. Sci., 28, 2331-2336 (1993). T. Hamada, T. Nishida, Y. Sajiki , M. Matsumoto and M. Endo , Structures and physical properties of carbon fibers from coal tarmesophase pitch, J.Mater. Res., 2(6),850-857 (1987). D. D. Edie, N. K. Fox, B. C. Barnell and C. C. Fain, Melt-spun non-circular carbon fibers, Carbon, 24[4), 477482 (1986). Sin-Shang Lin, Oxidative stabilization in production of pitch based carbon fiber, SAMPE Journal, 27 (1), 9-14 (1991). X.Bourrat, E. J. Roche and J. G. Lavin, Structure ofmesophase pitch fibers, Carbon, 28(213), 435446 (1990). D. J. Johnson, Structure and properties of carbon fibres, in Garbon Fibers Filaments and Composites, J. L. Figueiredo etaI., eds., 119-146, Kluwer Acad. Publ., Dordrecht NL (1990). M. Endo, Structure ofmesophase pitch-based carbon fibres, J. Mater. Sci., 23, 598-605 (1988).

Chapter 9

263

[33] A. M. Rao, A. W. P.Fung, M. S. Dresselhaus and M. Endo, Structural characterization ofheat-treated activated carbon fibers, J. Mat. Res., 7[7], 1788-1794 (1992). [34] E. Fitzer, PAN-based carbon fibers: present state and trend ofthe technology from the viewpointofpossibilities and limits to influence and to control the fiber properties by the process parameters, Carbon, 27[5), 621-645 (1989). [35) M. Guigon, A. Oberlin and G. Desarmot, Microtexture and structure of some high-modulus Pan-based carbon fibres, Fibre Science and Technology, 20, 177-198 (1984). [36) M. Guigon and A. Oberlin, Heat treatment of high tensile strength PAN-based carbon fibres: microtexture and mechanical properties, Composites Sci. Tech., 27, 1-23 (1986). [37) A.Oberlin, Carbonization and graphitization, Carbon, 22[6], 521 -541 (1984). [38) D. D. Edie, Pitch and mesophase fibers, in Carbon Fibers Fifaments and Composites, J. L. Figueiredo, C. A., Bernardo, R. T. K. Baker and K. J. HUllinger, eds., 43-72, Kluwer Academic Publishers, Dordrecht (1989). [39) E. J. Roche, J. G. Lavin and R. G. Parrish, The mosaic nature of the graphne sheet in pitch-based carbon fibers, Carbon, 26, 911 -913 (1988). [40) E. J. Roche, Electron microscopy study of mesophase pitch-based graphitefibres, J. Mater. Sci., 25, 21492158 (1990). [41J J. D. FitzGerald, G. M. Pennock and G. H. Taylor, Domain structure in MP (mesophase pitch)-based fibres, Carbon, 29[2], 139-164 (1991). [42] X. M. Bourrat, E. J. Roche and J. G. Lavin, The role ofdisclinations inthe structure of mesophase pitch-based carbon fibers, Proc. Int. Symp. Carbon, Tsukuba, Tanso ed.,320-323, Nov. 4-8, 1990. [43] I. Mochida, Shu-Ming Zheng, Y. Korai and I. Toshima, The introduction of a skn-core structure in mesophase p~ch fibers by oxidative stabilization, Carbon, 28[1], 193198 (1990). [44) L. H. Peebles, Jr., Carbon fibres: structure and mechanical properties, Intern. Mater. Rev., 39(2),75-92 (1994). [45) Yanling Huang and R. J. Young, Micro-structure and mechanical properties of pitch-based carbon fibres, J. Mater. Sci.,2!.\, 4027-4036 (1994). [46) J. D. H. Hughes, Strength and modulus ofcurrent carbon fibres, Carbon, 24(5), 551 -556 (1986). [47) S. Ozbeck, G. M. Jenkins and D. H. Isaac, Mechanical properties ofhot stretched carbon fibres, Ext. Abstracts 20"'. Biennal Conf. on Carbon. Santa Barbara, The Amer. Carbon Soc., 308-309 (1991). [48] E. M. Asloun, J. B. Donnet, G. Guilpain, M. Nardin and J. Schultz, On the estimation ofthe tensilestrength of carbon fibres atshort lengths, J. Mater. Sci., 24, 3504-3510 (1989). [49) J. W. Hitchon and D. C. Phillips, The dependence of the strength of carbon fibres on length, Fibre SCi. and Technology, 12, 217-233 (1979). [50) M. C. Waterbury and L. T. Drzal, On the determination of fiber strengths by in-situ fiber strength testing, J. Composites Technology and Res., 13[1],22-28 (1991). [51] K. Kogure, G. Sines and J. G. Lavin, Microstructure and texture of pitch-based carbon fibers after creep deformation, Carbon, 32[8], 1469-1484 (1994). [52] R. Moreton and W. Walt, The spinning of polyacrylonitrile fibres inclean room condfions forthe production of carbon fibres, Carbon, 12, 543-554 (1974). [53] J. E. Zimmer, Disclinations and fracture, Proc. Int. Symp. Carbon: New Processing and New Applications, Toyohashi, 131-134, Kagaku Gijutsu-Sha, Tokyo (1982). [54] I. Mochida, L. Ling and Y. Korai, Some factors for the high performances of mesophase pitch based carbon fibre, J. Mater. Sci., 29,3050-3056 (1994). [55) T. Matsumoto, Mesophase pitch and its carbon fibers, Pure and Applied Chern., 57(11),1553-1562 (1985). [56] J. A. H. De Pruneda and R. J. Morgan. The effects of thermal exposure on the structural and mechanical integrityofcarbon fibres, J. Mater. Sci., 25,4776-4781 (1990). [57] S. Kumar, Structure and properties of high performance polymeric and carbon fibers: an overview, SAMPE Quarterly, 20[2], 3-8 (1989). [58] Y. Tanabe, E. Yasuda, A. R. Bunsell, Y. Favry, M. Inagaki and M. Sakai, The strength of pitch-based carbon fibre athigh temperature, J. Mater. Sci., 26, 1601-1604 (1991). [59] S. Ozbek, G. M. Jenkins and D. H. Isaac, Thermal expansion and creep of carbon fibers, Ext. Abstr. 20'" Biennal Conf. Carbon, Santa Barbara, 270-271 , Amer. Carbon Soc. Ed., (1991). [60] J. F. Villeneuve, R. Naslain, R. Fourrneaux and J. Sevely, LongitudinaUradial thermal expansion and Poisson ratio ofsome ceramicfibers asmeasured byTEM, Composites Sci. Technology, 49, 89-103 (1993). [61] J. M. Kowalski, New high performance domestically produced carbon fibers, Proc. 32"" Int. SAMPE Symp., 953-963, April 6-9, 1987. [62) M. S. Dresselhaus and M. Endo, Intercalation of graphite fibers, in Graphite Intercalation Compounds II, H. Zabel and S. A. Solin,eds., Springer Series inMater. Sci.,vol. 18, chap. 8, 347-411 , Springer (1992). [63] Fundamental Issues in Control of Carbon Gasification Reactivity, J. Lahaye and P. Ehrburger, eds., NATO ASI-series, Series E - Applied Sciences, Vol. 192, Kluwer Academic Publishers, DordrechllBoston/London (1990).

264

Chapter 9

[64) T. L.Dhami, L. M. Manocha and O. P. Bahl, Oxidation behaviour of pitch based carbon fibers. Carbon, 29(1). 51-60 (1991). [65) Sciences etTechnologies au Japon: les Materiaux Carbones, 129-139. SF-JTI, French Foreign Ministry. Paris (1993). [66) F. Lamouroux. X. Bourrat, R. Naslain and J. Sevely, Structure-oxidation behavior relationship in the carbonaceous constituents of2D-C/PyC/SiC composites, Carbon, 31 , 1273-1288 (1993). [67] L. E. Jones and P. A. Thrower. Influence of boron on carbon fiber microstructure. physical properties and oxidation behavior, Carbon. 29(2). 251-269 (1991). [68] Y. Deslandes and F. N. Sabir, Inhibition ofoxidation ofcarbon fibres by501-gel coatings. J. Mater. Sci. Letters, 9,200-202 (1990). [69] J. Bouix. R. Favre. C. Vincent. H. Vincent, S. Cardinal. P. Fleischmann, P. F. Gobin and P. Merle, Carbon fibre coating and its influence onthe mechanical behavior of the composites, Proc. 101h Tsukuba General Symp., 259-272, Oct. 2-3, 1990. [70] T. Helmer, H. Pete~ik and K. Kromp, Coating of carbon fibers: the strength of the fibers, J. Amer. Ceram. Soc.• 78(1),133-136 (1995).

CHAPTER 10 SILICON CARBIDE AND OXYCARBIDE FIBERS R. Naslain The major deficiency of carbon fibers is their sensitivity to oxidation even at relatively low temperatures. Although silicon carbide (SiC) fibers are also sensitive to oxidation, their oxidation starts at higher temperatures and yields a protective silica coating. In an oxidative environment, SiC and Si-C-O fibers are generally more useful than carbon fibers [1-3]. Large diameter silicon carbide fibers are obtained by chemical vapor deposition (Chapter 4). Small diameter silicon carbide and oxycarbide fibers are derived from solid polydimethylsilazane precursor fibers (this chapter).

10.1 General considerations SiC isa covalent solid in which silicon istetrahedrally coordinated tocarbon (Sp3 bonds). The SiC4 tetrahedra are interconnected only by corner sharing. SiC exhibits two simple crystal structures, a cubic blende type modification or the p (or 3C) form, and a hexagonal wurtzite modification orthe ex (or 2H) form. SiC also displays numerous polymorphs orpolytypes with hexagonal or rhombohedral unit cells which correspond to complex stacking sequences [4]. SiCisrigid and brittle (Table I), and has high thermal and electrical conductitivities. SiC melts at2500°Cwith peritectic decomposition. Finally, the diffusion coefficients in SiC are low even at relatively high temperatures, a feature which correlates with its resistance to sintering and creep. Table I. Silicon carbide fiber properties. Materials Single crystal SiC whiskers SiC/C fibers made by CVD PC5-based Hi-Nicalon SiC fibers

g/~' 3.21

3.10

E GPa 480-580 360-390 420

cr' GPa 6-20 3.6-4.6 2.6

P.

lO-6oC ' 5.0

IX.

n.an

3.1

0.1

Silicon carbide fibers are derived [1-3] [5-6] from a polymer precursor fiber bya process that issimilar to the carbon fiber process. It starts offwith polydimethlysilane (PDMS) rather than polyacrylonitrile. First generation Si-C fibers are oxygen containing fibers [7-13]. Second generation Si-C fibers are oxygen-free. They can be produced in a first generation process wherein the green fiber is either cured by electron beam irradiation [27-29], or derived from a high molecular weight polycarbosilane precursor fiber [30]. These SiC fibers contain free

266

Chapter 10

carbon and possess a fine microstructure that is stable at elevated temperatures. Their modulus islow «270GPa) [27] [29-30] and they creep, but the creep rate islower than that of first generation Si-C(O) fibers [31-32]. Third generation, oxygen-free Si-C fibers can also be produced bypolymer precursor routes. They are quasi-stoichiometric, Le., they have an atomic C/Si ratio close to unity. Free carbon that is present in E-beam cured PCS fibers can be removed by performing the pyrolysis in a decarburizing atmosphere [33]. A precursor fiber can be used with an atomic C/Si ratio close to unity [34-36]. Boron or aluminum can be introduced into the green fiber to remove excess carbon and oxygen as CO, while the pyrolytic residue undergoes sintering with boron or aluminum acting as sintering aids [37-40]. Finally, quasi-stoichiometric SiC fibers can be produced by extrusion of a viscous slurry of a submicrometer SiC powder in suspension in a liquid [41-42]. Quasi-stoichiometric fibers have aYoung's modulus ofabout 420 GPa. The criteria for designing fibers for use in ceramic matrix composites (CMCs) are different from those for designing fibers for use in polymer or metal matrix composites. The key properties are thermal stability and mechanical properties at high temperatures [43]. As a consequence, relatively coarse microstructures are obtained at elevated temperatures, corresponding to somewhat lower failure strengths (-2 GPa), but high thermal stability and creep resistance are preferable toultrafine microstructures.

Synthesis of polydimethylsilane Autoclave (470°C)

or

Catalyst (PBDPSO)

Si-C-O fiber

Figure 1.Fabrication ofSi-C-O fibers from PCS precursors

10.2 Preparation of Si·C·O fibers The generic polymer based process that yields oxygen containing silicon carbide fibers consists of five steps. (1) Polydimethylsilane, or PDMS, is synthesized. (2) PDMS is rearranged into polycarbosilane, or PCS. (3) PCS is melt spun and yields a solid, green, or

267

Chapter 10

precursor fiber. (4) The solid PCS precursor fiber is rendered infusible byoxygen crosslinking and (5) the stabilized fiber is pyrolyzed and yields the final functional Si-C-O fiber. Polycarbosilane (PCS) is the most commonly used organosilicon polymer precursor for producing SiC based fibers. Polycarbosilane is a generic name forpolymers containing Si-C bonds intheir backbones [5) [6) [44). 10.2.1

The Yajima process

In the generic process, dimethyldichlorosilane is polymerized and yields polydimethylsilane (POMS), asshown in Equation 1, through alkali metal promoted dehalocoupling of chlorosilanes. The polymerization of dimethyldichlorosilane is carried out in xylene in a nitrogen atmosphere. POMS has a polymeric chain structure with a complex empirical formula (SiCl940oo2Hs46). It cannot beused directly as a SiC precursor. Its backbone has a SiSi chain and itspyrolytic residue at800·C is almost nil [7]. n(CH 3)] SiCI] +2nM ~[(CH3h Si]" +2nMCI

(1)

with M = Li, Na (CH 3JJ Si-Si(CH 3JJ

400· C

) (CH 3) 3 Si-CH]-Si-(CH3hH

(2)

POMS must therefore be converted into polycarbosiliane (PCS) by the Kumada rearrangement as shown in Equations 2-3 [6). The rearrangement inserts methylene (-CH2-) groups from pendent methyl groups into the Si-Si chain [6] [44). This produces a polymer with a -Si-C-Si- backbone, which prefigures that of silicon carbide, and lateral Si-H bonds, which help to stabilize the green fibers. Insertion of methylene groups into the POMS backbone yields polysilapropylene (PSP), te., SiC 2H6 , with C/Si at. = 2, but there is no gas evolution. The PDMS/PCS conversion proceeds by a thermal, or catalytic thermal, rearrangement. The actual mechanism is more complex than assumed in Equation 2; the PCSs are partly crosslinked and their formulas are different from that of PSP. CH]

CH]

CH]

- l i - - l i - - - +li--CH l

I)

CH]

1

CH]

1

H

~

(3)

The thermal rearrangement of POMS (Equation 4) was initially performed at 470·C under atmospheric pressure in nitrogen (PC-N-470 in Table II). However, under such conditions, the reaction is slow. The rearrangement step can be shortened byone order of magnitude if the reaction is performed in an autoclave, the final internal pressure being about 10 MPa owing to the formation of gaseous species (PC-A-470 in Table II). The resulting product is dissolved in n-hexane and the solution is filtered. After removal of the solvent, the residue is vacuum distilled yielding a yellowish brown PCS. The chemical composition of PC-A-470, i.e.•49.99 wt. % Si; 37.88 wt. % C; 0.99 wt. % 0 and 6.60 wt.% H, corresponds to SiC 17700 03H370 [7). Such PCSs are solids atroom temperature with high softening points (Ts ~300°C) .

268

Chapter 10

CH J I

ci-si-ct I CH J

I

Dim ethyldichlorosilane Na

1

CH J I

[-Si-] +NaCI

~6Hs

J

Polydimethysilane / I ~ N N 2 gas fl~:as f 10w Autoclave

,0

[-Si-O-B I

C 6H s

'0 <,

(PBDPSO)

~n

I 1I

I / I

-, I

[-Si-C-]

[ -Si-C-]

I

(4)

CH n

)n

I

n

(PC-B)

I

(PC-N)

I

[-Si-C-] n

I

I

n

(PC-A)

An improved version of the PCS synthesis relies on the use of a catalyst, e.g., polyborodiphenylsiloxane (PBOPSO) that favors the POMS/PCS conversion (Equation 4) [1] [6] [44]. Such PCSs (referred toas PC-B followed by the percentage ofcatalyst) are obtained at a lower temperature (Le., 350°C) with a higher yield under an inert gas at atmospheric pressure (Table II). Table II. Synthesis, yield and structure of PCS precursor fibers [1, 13) PCS PC-N-470 PC-A-470 PC-B-3.2 PC-B-5.5 PC-TMS N, flowing

Synthesis T, °C T, h 470 120 470 14 350 10 350 10 760 27 N,; A, autoclave; B, B-based

Structural Units Da SiC, SiC,H 50-55 1500 58.5 1680 0.53 0.47 50.0 1740 0.40 0.16 64.8 1310 0.44 0.15 6.5 620 0.82 0.18 catalyst, flowing N,; TMS, tetramethyisilane Yield %

Mn

SiC,Si... 0.00 0.44 0.41 0.00

The insertion of CH2 groups inthe -Si-Si- backbone ofPOMS and the formation ofSi-H bonds can be observed by the change occurring in the IR spectrum [7] [9] [10] [13] [45]. The POMS/PCS conversion via the Kumada rearrangement was once assumed to occur by a radical mechanism starting with the pyrolytic cleavage of the relatively weak Si-Si bond in POMS [7]. PBOPSO was believed to act as a trapping agent for the low molecular weight species resulting from the decomposition of POMS, and to enhance the abstraction of hydrogen from the methyl groups by boron radicals. More recently however, the radical mechanism and the role played by PBOPSO in the rearrangement of the POMS backbone has been questioned [44]. The difference between the formula for PSP, namely SiC2H6, and that of the Yajima type PCSs from the Kumada rearrangement of POMS, e.g., SiC1770003H370, suggests that the latter

Chapter 10

269

are not true linear PSPs (Equation 2). A model linear PSP was synthesized [44) by a ringopening polymerization of 1,1 ,3,3-tetramethyl-1 ,3-disilacyclobutane (Equation 5). n

(5)

The resulting linear PSP is a liquid oligomer at room temperature, with Mn = 2300 Da, in contrast to the high softening point of PCSs with similar Mn- Furthermore, the pyrolytic residue yield of linear PSP at 1000·C is less than 20%. However, this yield increases if PSP is further heat treated, i.e., thermally crosslinked. For example, at400-500·C, PSP yields a soluble solid precursor ( M n = 8600 Da) with a yield of65% and, at900·C, a ceramic material with a yield of 45%. Analysis of the preceramic polymer shows a loss of both carbon and hydrogen compared to SiC2H6, as did the Yajima type PCSs. A more severe heat treatment (e,g., at 450·C in an autoclave) results in a material which becomes progressively insoluble with a still higher, up to85%, ceramic yield (Equation 4)[44). All these features show that the PCSs produced in the Kumada rearrangement of PDMS by the Yajima route are not linear PSPs but branched polymers. Viscosity measurements and spectroscopic data (NMR, IR, UV) have shown that the structure of PCS polymers is roughly planar with three different atomic bonding schemes for the silicon atoms [11 [6) [9). The respective fraction of each atomic bonding scheme depends upon the PDMS/PCS conversion conditions (Table II). The structure of PCSs resulting from the catalytic thermal rearrangement ofPDMS (PC-B type polymers) contains more Si-Si bonds. It could thus be belter described as polysilane chains connected through highly branched carbosilane nodes [44). 10.2.2 Melt spinning of PCS PCSs are usually spun in the molten state at about 300·C. When necessary, PCSs are pretreated in order to adjust their molecular weight distribution [29) [46). Spinning is performed under nitrogen since PCSs are sensitive to oxygen and moisture. The molten filaments are drawn down by mechanical pulling in order toachieve a diameter of15-20 IJm. 10.2.3 Stabilization and curing Since PCSs are melt spun, they have to be cured before being pyrolyzed. In the generic process, infusible fibers are achieved by oxidation ofthe green PCS fibers. Oxidation of PCS in air begins atabout 150·C. The reaction is exothermic and the weight gain, which depends on the nature of the precursor, ranges from 8 wt.% (PC-A-470) to 17 wt.% (PC-B-3.2). The empirical formulas for the oxidized PCSs are SiC163H3340020 (PC-A-470) and SiC 176H mOo 6B (PC-B-3.2), respectively [8) [13).

270

Chapter 10

Wave number v, crn-t Figure 2.Infrared spectra of PDMS, PC·A470 and its pyrolytic residues at increasing pyrolysis temperatures according toreference (8); reproduced with permission.

IR spectroscopy shows that curing proceeds mainly by oxidation of Si-H bonds and, to a lesser extent, of Si-CH 3 bonds to form OH, C = 0, Si-O-Si or Si-O-C groups [6] [8] [13]. Oxidation of Si-H and Si-CH 3 bonds increases the degree of crosslinking ofthe PCS units, via the formation of Si-O-Si orland Si-o-C bridges, rendering the polymeric precursor fiber progressively infusible. Radical mechanisms have been proposed to account for the formation ofsuch bridges via the preferred oxidation ofthe Si-H bonds [6]. 10.2.4 Pyrolysis of PCS fibers PCS precursor fibers are converted to SiC based ceramic fibers by pyrolysis in an inert atmosphere at 1200-1300·C [6] [8] [13] [18] [28] [46-47]. This conversion is accompanied by evolution of gas (Figure 3) and weight loss. A large fraction of the organic bonds break at 800-900·C, and the pyrolytic residue is an amorphous, still hydrogenated, Si-C (or Si-C-O)

Chapter 10

271

containing material. With increasing temperature, heteroelements (H and 0) are released as a result of the scission of the last C-H bonds and decomposition of SiC,Ct. At 1200°C, oxygen cured PC-A-470 loses 20 wt.%. Above 1600-1800°C, the pyrolytic residue reaches its thermodynamic equilibrium state. However, since the tensile strength of the fibers produced from oxygen cured PCS undergoes a maximum at 1200°C, the pyrolysis of the commercial fibers ends atabout this temperature. They are below their equilibrium state and are metastable when heated above their processing temperature.

6

400

800

1200

1600

2000

Temperature, K

Figure 3.Gas evolution during the vacuum pyrolysis under of PCS fibers cured in helium by E-beam irradiation (28); reproduced with permission from the American Ceramic Society, PO Box 6136, Westerville, Ohio 43086-6136.

The pyrolysis of PCS can be discussed by considering four temperature domains overlapping one another to some extent: 200-600°C (domain 1); 500-900°C (domain 2); 700-1200°C (domain 3) and above 1200°C(domain 4). The majority of the weight loss occurs in domain 1 as a result of volatilization of low molecular weight PCSs and dehydrogenation/dehydrocarbonation, condensation and radical polymerization reactions. At the end ofdomain 1, PCSs are almost fully crosslinked. The organic/inorganic transition occurs within both the second and third temperature domains [28] [47] with a fresh but minor weight loss and a density increase. The gaseous species which are formed are mainly CH4 and Hzfrom scission of the relatively weak Si-CH 3 and Si-H lateral bonds indomain 2, and mainly Hz from scission ofthe stronger C-H bonds from the SiCH 2-Si backbone in domain 3. At the end of domain 2, i.e., at 800-900°C, the pyrolytic residue isan amorphous Si-C (or Si-C-O) material with a composition close to SiC 16Ho 65 [6]. It isstill hydrogenated and contains an excess of carbon with respect tothe stoichiometric PSP having a C/Si at. = 1. It also displays numerous structural defects and itsdensity (2.21 g/cm 3) isstill low. Atentative structural model has been proposed [6].

272

Chapter 10

As the temperature is raised to domain 3, hydrogen corresponding tothe residual C-H bonds from the Si-CH2-Si backbone is progressively released with a continuous but slow density increase. Above about 1000°C, clusters of free carbon and nanocrystals of P-SiC are formed . At the end of domain 3, ceramic grade Nicalon Si-C-O fibers consist of a dispersion of free carbon clusters and SiC nanocrystals in an amorphous Si-C-O matrix [14). As the temperature is increased to domain 4, growth of P-SiC nanocrystals and decomposition of ternary silicon oxycarbide occurs in the pyrolytic residue, and evolution of CO and SiO is accompanied by further weight loss. 10.2.5 Related Si-C-O(Ti) fibers Titanium can be introduced in PCS precursors as titanium tetrabutoxide, Ti{OC4H9)4, to yield Si-C-O{Ti) ceramic fibers such as Tyranno (Equation 6) [48). Fibers produced from such PCS{Ti) precursors have a slightly higher pyrolysis temperature (Tp) than that previously mentioned for their pure PCS counterparts [49]. As a result, they also retain their amorphous state and hence their tensile strength toa slightly higher temperature. PCS{Ti) precursors are produced in a one (or two) step process by crosslinking PCS chains with titanium tetrabutoxide [48) orisopropoxide [49). The PCS chains are formed with the use of PBDPSO catalyst and no autoclave. As a result, PCS{Ti) already contains a significant amount ofoxygen in the as-prepared state.

OC H I 4 9

~ ~

~ ~

HC-Si--O-- Ti--O--Si-CH 3

I

I

CH ~

I

OCH 2

4

9

CH ~

3

(6)

2

PCS{Ti) precursors with M n '" 1600 Da are melt spun at 270°C and the green continuous fibers are cured in air at 170°C. During the curing step, part of the Si-H bonds still present in the precursor isused to further crosslink the polymer through the formation of Si-O-Si bridges, resulting in an increase of the oxygen content. The cured fiber is pyrolyzed in an inert atmosphere. Ata pyrolysis temperature of 1300°C, the chemical composition (34.14 at.%Si, 43.67% C, 20.79% 0, 1.22% Ti, 0.13% Nand 0.04% B) corresponds to a formula of SiC1260061Tio OJ. Thus, the Si-C-O-Ti fiber contains an excess of carbon and the pyrolytic residue is amorphous. Crystallization occurs above 1300°C with an evolution of gaseous oxides, formation ofSiC and TiC nanocrystals, and a decrease in tensile strength.

10.3 Preparation ofoxygen-free Si·C fibers

Nearly-oxygen-free Si-C fibers can be made by two routes. In one process, the precursor fibers are dry spun from a Yajima type PCS precursor, but the green fibers are cured by yrays or E-beam irradiation [25-29). In the other process, curing is not needed. The green precursor fibers are dry spun from a solution ofsoluble but infusible PCSs [30). 10.3.1

From radiation cured PCS precursor fibers

The first approach to nearly-oxygen-free SiC fibers is to mell spin green fibers from Yajima type PCSs ( Mn = 1600-2000 Da) and to rendered them infusible by anaerobic curing with y-

273

Chapter 10

ray (e.g., GOCo y-rays) orelectron beam (e.g., 2 MeV) irradiation atroom temperature. E-beam curing is usually preferred since the electron beam can be easily deflected to scan large fiber samples and provide a higher irradiation dose in a much shorter time. Radiation curing of PCS requires a minimum dose of radiation, which is expressed in gray or in rad (1 Gy =100 rad =1 J/kg). The degree of crosslinking can be estimated by measuring the gel fraction, i.e., the percentage of PCS insoluble in tetrahydrofuran (THF) after irradiation (Figure 4) [26]. In order to keep the fiber sample from melting, the sample containers are maintained atroom temperature by water cooling.

75 ~

° ~

c0

'€3

£1

50

Qi

<.!l

25

o Vacuum • He

0

5

10

15

20

Dose, MGy

Figure 4. Irradiation curing ofgreen PCS fibers: gelfraction in irradiated pes [26J; reproduoed with permission from the American Ceramic Society, Westerville, Ohio.

Free radicals are formed during the room temperature irradiation of PCS. Their amount (Figure 5) can be estimated by electron spin resonance (ESR) [47]. The free radicals, which do not contribute to the crosslinking reaction, remain entrapped in the green fibers. Most of them are extremely sensitive to oxygen and moisture. Irradiated fibers can therefore not be handled directly in air, but the relative amount of radicals issignificantly reduced (Figure 4) by annealing in an inert gas atmosphere orvacuum [25]. Cured fibers are subsequently pyrolyzed. Residual oxygen in ceramic SiC fibers depends on the nature ofthe atmosphere during the post-irradiation annealing and pyrolysis treatments. It is about 2.5- 4 wt.% after pyrolysis at1400°Cin argon, when no post-irradiation annealing is performed [25] [29], but is only 1.9 wt%when the cured fibers are annealed at 250°C before pyrolysis. Finally, residual oxygen drops to <0.5 wt.% for fibers annealed in vacuum at 1000°C and pyrolyzed in argon at1200°C [29]. This procedure yields Si-C fibers with a

274

Chapter 10

composition close to SiC141 which consist of P-SiC nanocrystals and free carbon and are virtually free ofoxygen [25J [50J.

6.------------------., Electron beam crossllnking (1.1 wt"k 0) ·· · ··Oxldatlve curing (13wt% 0)

.c--;

5

/

..../ /

\ .\.

\

.... 400

800

1200

1600

2000

Heat treatment temperature, K

Figure 5.Variations of the entrapped radical concentration at room temperature as a function of the pyrolysis temperature in E-beam cured PCS fibers (1 .1 wt.% 0), from (47); reproduced with permission fnom the American Ceramic Society, Westerville, Ohio 43086.

Evolution ofH2 and CH4 (Figure 3) and weight loss can be observed during the pyrolysis of Ebeam cured PCS [28]. Evolution ofCH4 occurs as one single peak whereas that of hydrogen occurs as two overlapping peaks, suggesting a two step pyrolysis mechanism. The second peak at 1025°C is almost the same for E-beam and oxygen cured fibers. Conversely, the intensity of the first peak (at 700°C) strongly depends upon the percentage of Si-H bonds already involved in the curing step. The change in free radical concentration as a function of the pyrolysis temperature (Figure 5) shows that free radicals are formed and remain entrapped inthe fibers [47]. The curve for the E-beam cured fibers can be decomposed into two peaks at 625°C and 1025°C, the former being more intense than the latter (whereas it is the reverse for the oxygen cured fibers). There is a correlation between the gas evolution (Figure 3) and the entrapped radical concentration (Figure 5), suggesting a two step pyrolysis process. A radical reaction mechanism has been proposed. The evolution of H2 and CH4 at low temperat ures is related tothe formation of=Si· based free radicals corresponding tothe scission ofthe weaker Si-H and Si-CH 3 bonds; evolution of H2 at higher temperatures is related to the formation of=C· based free radicals, which result from the scission ofthe stronger C-H bonds inthe Si-CH 2-Si groups.

Chapter 10

275

10.3,2 From infusible PCS precursor fibers The second approach to the production of Si-C fibers with a low oxygen content is based on the use of PCS precursors which are infusible (Mw = 5000-10000) but completely soluble in organic solvents such as toluene [30] [51-52]. Such PCSs can be prepared via the thermal rearrangement of POMS in an autoclave at 435-480°C, as in the Yajima route, but the duration of the heat treatment islonger. Green fibers are dry spun from the PCS solution with spinning aids such as high molecular weight polyvinylsilazane (PVSZ) and a small amount of a hydrosilylation catalyst, such as dicumylperoxide [52], to generate free radicals. The vinyl groups contribute to the crosslinking of PCS chains through reaction with Si-H bonds, The solution of PCS and PVSZ isfiltered and is concentrated todevelop appropriate rheological properties. The solution is dry spun at room temperature and the green fibers are wound onto a drum. The solvent is partly evaporated between the spinnerets and the drum, producing infusible green fibers. These fibers are slowly heated to 200°C in an inert atmosphere to release all residual solvent and tofully crosslink the polymer. They are finally pyrolyzed in argon below 1000°C to yield a black Si-C ceramic fiber. This fiber still contains a small amount of hydrogen and oxygen, and a significant amount of excess carbon. It is amorphous and has an empirical formula ofSiC178HolSOoosNoo4, but tends to crystallize above 1200°Cinto a mixture of P-SiC and free carbon. 10.4 Preparation of quasi-stoichiometric SiC fibers When the pyrolysis of PCS precursor fibers is carried out in the presence of hydrogen, or when boron or aluminum doped PCS precursor fibers are pyrolyzed, it is possible to obtain quasi-stoichiometric silicon carbide fibers, Alternatively, quasi-stoichiometric fibers are also obtained from precursor fibers consisting of SiC powder reinforced polymers. 10.4.1

Pyrolysis of pes precursor fibers under hydrogen

The excess ofcarbon in Si-C fibers can be lowered bypyrolyzing Yajima type PCS precursor fibers in an atmosphere of hydrogen [33]. Hydrogen is assumed to favor release of the pendent methyl groups and formation of CH4 during the organic/inorganic transition. Complete pyrolytic release of pendent methyl groups from the ideal polysilapropylene chain polymer (Equation 1) would yield stoichiometric SiC, assuming that the carbon from the SiCH 2-Si backbone remains in the solid state. The residue after pyrolysis at 1000°Cis thought tobe more hydrogenated than the residue prepared in argon. Quasi-stoichiometric SiC fibers have been recently produced from fusible [53] and infusible [34-35] PCSs. The processing details have not been disclosed, but it is likely that a hydrogen or hydrogen containing pyrolysis atmosphere was used. 10.4.2 Pyrolysis of boron doped PCS precursor fibers Yajima type fibers, produced from oxygen cured PCS, decompose above 11 00-1200°C in an inert atmosphere with evolution of CO, This process can be used to simultaneously remove oxygen and carbon from the fiber. If the temperature is sufficiently high (e.g., 1800°C), removal of oxygen is almost complete leaving behind a pyrolytic residue that is either a mixture ofSiC + Corstoichiometric SiC having less than 0,5wt.% residual oxygen.

276

Chapter 10

Unfortunately, the Yajima type fibers, when treated at 1800°C,become porous and therefore extremely weak unless they are doped with SiC sintering aids. For example, doping Si-C-O Yajima type fibers with 0.2 - 0.6wl.% B yields dense fibers after pyrolysis at 1800°C[37-39J. Ifenough oxygen is introduced when the green fibers are cured, stoichiometric SiC fibers may result having 97% of the theoretical density and less than 0.5 wt. %O. The boron dopant can be introduced in the green fiber by curing it in an atmosphere containing a gaseous boron bearing species [38-54J. Boron is a well-known sintering aid for silicon carbide, but temperatures of at least 2000°C are required for sintering polycrystalline SiC powders with a grain size of the order of 1 IJm. Its effectiveness in sintering SiC based fibers at lower temperatures (1600-1800°C) is probably related tothe extremely small SiC grain size ofPCS based materials. Similarly, an experimental, quasi-stoichiometric, oxygen-free SiC fiber has been produced from a Si-AI-C-O precursor. In this process, aluminum is introduced into the polymer as aluminum (III) acetylacetonate, and acts like boron as a SiC sintering aid during heat treatment at 1800°C[40J. 10.4.3 From extruded SiC pOWder/polymer mixtures Continuous quasi-stoichiometric SiC fibers can also be produced from SiC powder/polymer mixtures [41J [57-58J. These mixtures consist of a fine SiC powder (usually a-SiC) with a grain size less than about 1 IJm and a boron bearing species, i.e., boron carbide, B,C. The source ofcarbon is typically a phenolic resin and various polymers to render the mixture melt spinnable. After melt spinning, the green fibers can be directly fired orthey can be woven and then fired . During firing in nitrogen, residual oxygen is released as CO, and B,C acts as a sintering aid. Since the grain size of the SiC powder is relatively large, firing has to be performed at 200Q-2300°C to achieve a high degree of densification. The draw ratio during melt spinning islow and the ultimate fiber diameter remains high «100 IJm). 10.5 Structure of silicon carbide fibers

Fibers resulting from the pyrolysis of PCS and related precursor fibers consist of pure SiC, SiC + Cor SiC + SiO,Cv + C, depending on their processing conditions. Silica is not present except eventually as a discontinuous thin film atthe fiber surface. 10.5.1 Silicon oxycarbide fibers Fibers resulting from the pyrolysis of oxygen cured PCS precursor fibers at 850-1000·C remain amorphous and still contain significant amounts of hydrogen and C-C bonds. With a further increase in pyrolysis temperature to 1200-1300°C, hydrogen is progressively released and SiC nanocrystals and free carbon clusters are formed. Specifically, the resulting fibers consistofa mixture of P-SiC nanocrystals and partly hydrogenated carbon clusters which are dispersed inan amorphous silicon oxycarbide phase, usually formulated as SiO,Cv• The formation of both P-SiC and carbon is evidenced by transmission electron microscopy (TEM) [1J [16J. The P-SiC phase appears as tiny crystals with a mean size of 2 nm. The carbon phase appears as randomly oriented 0.7 to 0.8 nm BSUs, which consist of polyaromatic species associated face to face and which could be related to coronene, C2,H,2 [16). PCS pyrolytic residues contain 20 at. % residual hydrogen after pyrolysis at850·C and 4

277

Chapter 10

at.% after pyrolysis at 1000·C (46). Since the C-H bond is a strong bond, it is likely that residual hydrogen is present as C-H bonds. Finally, free carbon can also be observed by Raman spectroscopy. The Raman spectrum of Si-C-O fibers exhibits an unresolved band after pyrolysis at 850·C. This Raman band splits into two lines at 1580 and 1350 crrr' after pyrolysis at1000·C. (1). The formation of a ternary phase in Si-C-O fibers is evidenced in x-ray photoelectron spectra orXPS (14) [59-63]. The C1s and Si2p photopeaks recorded from the surface of a typical SiCoO fiber are asymmetrical (Figure 6) and can be decomposed into several components. Component I of the C1s and Si2p photopeaks (binding energy, BE, of 283.3 and 100.5 eV, respectively) corresponds to carbon and silicon from SiC (BE = 282.9 and 100.5 eV, respectively). Similarly, component III of the Si2p photopeak (BE = 103.1 eV) is assigned to silicon in silica (BE = 103.4 eV). The fact that its intensity decreases dramatically after argon ion etching shows that silica is only present at the fiber surface [16] [63]. Component II of the Si2p photopeak which corresponds to a binding energy, i.e., 101 .5eV, intermediate between those for silicon in SiC (100.5 eV) and silicon in silica (103.4 eV), is assigned to a ternary SiO,C, phase, where Si is tetrahedrally bonded to both C and 0 atoms. Component II of the C1s photopeak (BE = 284.6 eV) corresponds either to C atoms from the hydrogenated free carbon phase or to C atoms from the ternary SiO.C, species.

SiC

(b)

290

SiC

C1s

106

98

Binding energy, eV

Figure 6.C1s and Si2p core level spectra recorded from: (a) unetched and (b) Ar-etched Si-C-O Nicalon fibers derived from PCS, - - , experimental; - • - • - background substraction; • - - •• calculated (63); reproduced with permission.

Determining the molar composition of the Si-C-O fibers is complicated [15-16] [59-60) (63). There is already some dispersion in the elemental overall analysis data even for the major elements (Table III). Neglecting residual hydrogen, the mean composition ofthe Si-C-O fiber

278

(Nicalon NL 200) is close to 38 a1.% Si, 48 a1.% C and 14 a1.% overall formula SiCl260037.

Chapter 10

°

and corresponds to the

Table ill. Elemental composition of Si-C-C fibers derived from PCS precursor fibers Silicon oxySilicon carbide fibers at% wt% 37.70 57.90 Nicalon NL-200'I/ PeN-based 470 34.40 51.30 39.00 58.00 Nicalon NL-200(21 Nicalon NL-22~' 36.49 55.64 37.00 54.20 Nicalon NL-20i" 39.50 58.90 Nicalon NL-202(21 41 37.80 57.10 Nicalon NL-20i ,II [16J; (21 EPMA-data, [59J[15J[50J;

Carbon Oxygen Hydrogen at% wt% at% wt% at% wt% 43.10 28.40 15.00 13.20 3.7 0.20 53.40 34.10 11.20 9.55 0.9 0.05 47.00 30.00 14.00 12.00 49.95 32.57 13.56 11.78 47.40 29.80 15.60 13.00 48.50 30.90 12.00 10.20 49.80 32.20 12.40 10.70 W chemical analysis, [61J; (41 XPS-data, [60J

Nitrogen at% wt% 0.5 0.3

The calculation ofthe molar composition supposes a structural model for the SiO,Ct phase. It is usually assumed that the oxycarbide consists of tetrahedral units, SiCaO"" in which the silicon atom is bonded to carbon and oxygen. The compositional domain of the tetrahedral unit ranges continuously from Si04 (a =0) to SiC 4 (a =4). Additionally, it isassumed that the carbon atom remains tetrahedrally bonded to 4Si (as in SiC), and the oxygen atom is bonded to2 Si (as in silica), with no other crossbonding. Under such assumptions, the formula for the silicon oxycarbide derived from that ofthe tetrahedral unit is: SiCa/ .PI / 2( -I-a)

If Y =a/4 and x = 1/2{4-a), the general overall formula of the oxycarbide SiO,Ct can be written as [59]: a=4y

x =2(1- y)

and

x 2

or

(7)

y=J--

The variations of x and y as a function of a, and reciprocally those of a and 4 - a as a function of y, which give the correlation between the nature of the tetrahedral unit, SiCaO",. and the composition ofthe silicon oxycarbide, SiO,Ct, are shown in Figure 7.

Si02

4

Si03l2C1 /4

SiOC,12

Si0112C3I4

SiC

0.50

0.75

1

0.25

0

~

3

3

2~

~2

tf

---

i ......... ......_. . . . -. .--t----~~ ~.~

00 Si04

4

-

~:a:~

2 Si03C

Si02C2

a

1

3 SiOC3

SiC4

Figure 7.Correlation between the nature ofthe tetrahedral SiC.04-a un~ and the chemical overall formula inthe silicon oxycarbide.

279

Chapter10

The molar composition of Nicalon NL 200 can be computed as 46.06% SiC, 23.03% SiOxC,. xI2[X= 1.1053] and 30.64% free carbon, assuming: (1) the Nicalon NL 200 Si-O-C fiber consists of SiC, SiOxC, .x12 and free carbon; (2) its atomic composition is 38% Si, 48% C and 14% 0; and (3) the intensity ratio of the two photopeaks assigned to SiC and SiOxCv (and reported to be 2for Nicalon NL 200) gives the molar ratio between the two species. Thus, the fibers resulting from the pyrolysis at 1200-1300°C of oxygen cured Yajima type PCS precursor fibers in an inert atmosphere are far from consisting of pure SiC since the molar fraction of SiC is about 50%. They also contain significant amounts of partly hydrogenated free carbon and silicon oxycarbide. It is noteworthy that the value, or perhaps mean value [14] [64], of the term 1-x/2 = 0.447 is close to 0.5 in Nicalon NL-200 (Figure 7) and that the main tetrahedral units are therefore close to Si02C2 [15]. The atomic distribution of silicon, carbon and oxygen in the three phases (SiC, SiOxC,.x12 and free carbon) can be calculated as a function of x for a given overall atomic composition [59]. Other approaches have been used to calculate the molar composition of PCS based Si-C-O fibers from the overall atomicpercentages [16] [60]. 10.5.2 Silicon carbide fibers Silicon carbide fibers with low oxygen content fall intwo groups. One group of fibers displays a C/Si ratio higher than one and contains free carbon, e.g., 30 mol.% for Hi-Nicalon. The other group offibers displays a C/Si ratio close toone, e.g., 1.05-1.07, for quasi-stoichiometric SiC fibers having a free carbon content of <5 mol.% (Table IV). Fibers with a C/Si ratio lower than one, e.g., 0.85, presumably containing free silicon, have also been reported [53] [65]. Finally, SiC fibers contain very little residual hydrogen, a feature consistent with their higher processing temperatures [50]. Table IV. Elemental composition of silicon carb ide fibers with low oxygen contents Silicon Silicon carbide at % wt% (SiC) fibers 42.02 PeS-based EB-cured Hi-Nicalon(l ) 61.80 41.00 Hi-Nicalon'" 61.80 41.60 62.40 Hi-Nicalon'" Si-C UF-fiber(4) 55.00 Hi-Nicalon type 5(3) 48.70 68.90 SiC UF-HM fiber(2x,) 48.20 68.50 50 .00 70.05 Stoichiometric SiC (I) chemical analysis, [50]; \l1 EPMA, [50];

Carbon Oxygen Hydrogen C/Si at% wt%. at%. wt% at % wt % at% 55.04 0.84 2.10 1.31 37.20 0.6 <0.3 1.40 58.00 37.40 1.00 0.8 1.41 57.80 37.10 0.60 0.5 1.39 42.00 1.1-2.6 <0.5 1.78 51.10 30.90 0.20 0.2 1.05 51.80 31.50 < 0.1 1.07 50.00 29.95 1.00 (3) [65]; ~) from high'M~ PC:S;'i30]

[35'] "

The microstructure of SiC + C fibers depends on (1) the heat treatment temperature (HTT) used during fiber processing, (2) the C/Si atom ratio, whereby free carbon impedes SiC grain growth and (3) the use of sintering aids such as boron bearing species. Raising the HTT increases the SiC grain size and improves the crystalline state of the carbon phase. For a pyrolysis temperature of 1300-1400°C and a C/Si ratio of 1.40 corresponding to Hi-Nicalon fibers, the mean SiC grain size is about 5 nm, [31] [50] [65]. SiC is mainly present as the ~ (or 3C) form [46] [50]. Here, free carbon isbetter organized than inSi-C-O fibers and consists of thicker stacks ofcarbon layers having a larger extension (La = 2-3 instead of 0.8nm). The carbon layer stacks form incomplete cages between and around the SiCgrains.

280

Chapter 10

For a pyrolysis temperature of 1600-1800·C and C/Si ratio of 1.05, the fibers made from PCS precursor fibers have an improved crystallization state [34-36] [53-54] [65]. Their SiC grain size is 5-20 nm. Although the dominant phase is still P-SiC (or 3C), the fibers also contain small amounts of hexagonal a-polytypes. Numerous crystal defects such as twins and stacking faults are observed, features which are common in SiC ceramics. The amount of free carbon islower and "clean" SiC/SiC boundaries are locally present. The carbon phase is better organized, and the carbon layer stacks are thicker and more extended. Finally, the fiber density (3.1 g/cm 3) isclose tothat ofdense SiC (d lheor. = 3.21 g/cm 3) .

Figure 8.TEM-micrograph of a nearly-stoichiometric SiC UF-HM fiber showing the SiC grain size. The electron diffraction pattern isshown inthe inset; reproduced with permission from the American Ceramic Society, Westerville, Ohio.

For a pyrolysis temperature of 2000·2300·C and a C/Si ratio assumed to be slightly higher than one, fibers produced from sintered powders display a state of crystallization resembling that of sintered bulk SiC ceramics [41] [57-58]. The microstructure of such fibers is much coarser, with SiC grains of about 1 urn in size and impurities (e.g., B4C) at the SiC grain boundaries. Finally, the SiC fibers prepared at high temperatures usually exhibit a surface consisting ofalmost pure carbon over a depth of50-200 nm [54] [65]. 10.6 Thermal stability of silicon fibers

The Si-C-O fibers produced from oxygen cured PCS precursor fibers at 1200-1300·C are unstable at high temperatures. Si-C fibers, i.e., almost pure SiC or SiC + C mixtures, are thermodynamically stable to 2500·C. The only cause of microstructural instability is the extremely small grain size of the phases. When such fibers are maintained at high

Chapter 10

281

temperatures, grain size tends to increase. The driving force is the reduction in surface energy. Chemical reactions occurring in Si-C-O fibers and SiC grain growth in Si-C-O, Si-C and SiC fibers cause a dramatic reduction in tensile strength. 10.6.1 Silicon oxycarbide fibers When a Si-C-O fiber is heated in vacuum or an inert atmosphere above its processing temperature, an added weight loss occurs that is accompanied by evolution of gaseous species and by a profound change in its microstructure [16-17] [21] [23-24] [29] [66-71]. The weight 1055 rate remains slow up to about 1250°C(Figure 9). The overall asymptotic weight 1055 (~W<Xl) at 1500-1700°Cis 24-26% for the Si-C-O Nicalon fiber NL 200) [16] [23-24] [69]. The main gaseous species formed during decomposition are CO, SiO and, toa lesser extent, hydrogen. Residual hydrogen is released between 1000-1250°C[16] [71-72]. The formation ofboth CO and SiO has been documented by multiple Knudsen cell mass spectrometry. The relative contribution of CO (Weo =x) and SiO (WSiO= y) to the total weight loss can be calculated from the initial sample mass and weight 1055, the weight fraction of elemental Si, and the molar ratio MsJM SiO[69].

2Sr-------------, 14OO°C o

~


~

135O°C

20

15

iii


~ 10 s:

f

Time, ks

40

Figure 9.Weight loss occurring during the decomposition of Nicalon NL 200 Si-C-Ofibers asa function of time [69); reproduced with permission from the CeramicSociety ofJapan, Tokyo.

Variations of x and y as a function of decomposition ratio X = ~WJ~W<Xl (~W<Xl is the weight time t ~ <Xl attemperature T) are shown in Figure 9. CO accounts for the majority of the weight loss when Xtends to 1(e.g., 18 ofthe 24% weight 1055 for Nicalon NL 200).

1055 for

As gas evolution proceeds, Si-C-O fibers undergo a two step microstructural transformation [16]. In the first step that corresponds to the formation of hydrogen, the free carbon phase is reorganized and the BSUs form distorted sheets. Distorted layers form a network of open cages around each SiC crystal, helping to prevent its growth. In the second step, which occurs at higher temperatures and corresponds to the formation of both CO and SiO, free carbon isconsumed and the SiC nanocrystals grow. They are no longer separated from each

Chapter 10

282

other by C cages and amorphous SiO,C,.
- Ln ( I - X)

= kt"

(8a)

Ln {- Ln ( I - X)1 = Ln k + m Ln t

or (8b)

where k is a rate constant and ma constant whose value is in the range 0.5 - 4 and depends upon the mechanism of the transformation. It appears from Figure 10 that the fiber decomposition kinetics obey Equation 8b with a value of m of the order of 1.5for sufficiently high temperatures. Thus, decomposition might be kinetically governed bythe rate of growth of the SiC nanocrystals by diffusion, the nanocrystals growing 3-dimensionally. Finally, the noted decomposition is thermally activated and the apparent activation energy, 633-791 kJ/mol, isvery similar to that reported for the diffusion ofCinSiC (560 - 840 kJ/mol).

·4 '--"'-_ _' - - _ - - l ._ _---'-_ _...J 6.0 6.2 6.4 6.6 6.8 Reciprocal temperature T-1 , 10-4K-1

Figure 10. Decomposition kinetics of Si-C-O (Nicalon NL 200) fibers: Avrami-Erofeev plot of the data[69): reproduced w~h permission from the CeramicSociety ofJapan, Tokyo.

From the experimental data reported above, the decomposition might occur by a mechanism involving the two following equations: 4 SiOxCI-x I2~ (4 -3x) SiC+x CO+3 x SiO

(9)

y SiO + 2yC ~ y SiC + Y CO

(10)

4 SiOxCI-x l2+2yC ~ (4-3x+ y ) SiC+(x+ y) CO+(3x- y ) SiO

(11)

Chapter 10

283

This mechanism accounts for the evolution of a gaseous SiO + CO mixture, the growth of the SiC crystals and the decrease in the amounts of free carbon and silicon oxycarbide. The relative amounts of CO and SiO which are formed and the composition of the final residue (i.e., SiC or SiC + C) depend upon the relative amounts of free carbon and silicon oxycarbide in the fiber. For the Nicalon fiber (NL 200), the main species in the gas phase is CO and the solid residue is SiC. There is therefore enough SiO formed by decomposition of the silicon oxycarbide phase (Equation 9) to consume all the free carbon by Equation 10 [731. This mechanism also accounts for one of the processes used to produce oxygen-free nearlystoichiometric SiC fibers [38] [54]. Since CO and SiO diffuse and escape from the fiber, the decomposition starts near itssurface and the decomposition front moves radially towards the fiber axis, yielding a skin/core microstructure [16]. The decomposition of Si-C-O can be impeded, or at least shifted to higher temperatures, by subjecting the fibers to a pressurized gas or a gas tight coating. The decomposition of Nicalon fibers is shifted to about 1800°C when an isostatic argon pressure of 138 MPa is applied to the fibers [68] [74-76]. Similarly, the stability of the fibers is enhanced at 1300°C when they are treated inAr-CO gas mixtures with Peo = 40 kPa. In addition, a carbon buildup on the fiber surface results from a decomposition ofCO: (12) Slightly pre-oxidized Si-C-O fibers have a uniform glassy silica coating. Their decomposition occurs at higher temperatures. This suggests that the silica coating remains gas tight and prevents the evolution ofCO and SiO up to atleast 1400°C[69]. 10.6.2 Silicon carbide fibers Oxygen-free Si-C fibers, consisting of SiC and free carbon, are expected to be stable up to 2500°C, but experimental data show that the fibers experience surface decomposition and grain growth when aged at high temperatures. Surface decomposition, which is related to a surface vaporization of Si, is driven by the formation of a surface layer of carbon. Grain growth of the nanometer size SiCcrystals isdriven bya reduction ofsurface energy. A Si-C fiber experiences no noticeable weight loss when aged up to about 1800°C in an atmosphere of argon and itschemical composition does not change much. However, some evolution ofhydrogen continues tooccur [50] and a carbon layer builds up on the fiber surface whose thickness increases with time ata given aging temperature, e.g., to30 nm for one hour ofpyrolysis at 1600°Cand to70 nm for ten hours ofpyrolysis. Some growth of the SiC grains occur when the pyrolysis temperature is raised, even in HiNicalon fibers containing a large amount of free carbon. However, the maximum grain size remains relatively small, e.g., 20 nm at 1400°Cand 50 nm at 1600°C, compared with those of heat treated Si-C-O fibers. Free carbon is in equilibrium with SiC. However, it undergoes some reorganization with an increase of both the carbon layer size (La) and the stack thickness (N). This might occur by de-wrinkling of the layer stacks and their edge-to-edge association into larger layers tending to lieflat on the faces of the growing SiC crystals. Near stoichiometric, dense SiC fibers are produced at higher temperatures (1800 to 2300°C) with a boron bearing sintering aid; they do not contain significant amounts of free carbon [3435] [41] [53] [72]. As a result, the SiC grain size is relatively large inthese fibers, e.g., 50-200

284

Chapter 10

nm, and a few crystals are as large as 1.0 IJm (35). Some of the secondary intergranular phase, e.g., carbon, may impede the growth of the SiC crystals and the SiC fiber microstructure may have a higher thermal stability between 1200 and 1600·C when they have been processed at higher temperatures. Therefore, the microstructure is to some extent stabilized during processing. 10.7 Mechanical properties of SiC fibers SiC based fibers display a linear elastic behavior at room temperature. Their Young's modulus, E, and tensile strength, OR, depend on the processing conditions, particularly the pyrolysis temperature, Tp. 10.7.1 Atroom temperature At first, both the modulus and tensile strength of Si-C-O fibers increase as the pyrolysis temperature is raised to >850·C, i.e., as hydrogen is released and densification occurs. Then, they decrease sharply above 1200·C when the fibers undergo decomposition. There is thus an optimum Tp value. The modulus and tensile strength of the oxygen-free SiC + C fibers produced from E-beam cured PCS precursor fibers also show maxima as the pyrolysis temperature is raised but the maxima are shifted to higher temperatures and the decrease in both E and OR is gradual. The fibers still have high stiffness and strength at 2000·C (Figure 11 ).

4r-----------------, ..,

~

-oK

---

SiC+Cfiber

---)(--)(--.,.; -

....

Si-C-O fiber

1200

1600

2000

Temperature. °C

Figure 11 . Variations of the room temperature tensile properties of SiC-based fibers as a function of the temperature. The pyrolysis temperature forSi-C-O fibers derived from PCS and Si-G fibers derived from radiationcured PCS with0.4 wi. %ofoxygen (77]: reproduced with permission from the Woodhead Publishing Ltd.

The modulus of SiC based fibers strongly depends upon the occurrence of intergranular phases. It is of the order of 580 GPa for SiC whiskers and 400-450 GPa for the nearly stoichiometric, oxygen-free polycrystalline SiC fibers [54) (65). Conversely, it is much lower

Chapter 10

285

when the fibers contain free carbon and/or amorphous silicon oxycarbide. The modulus of oxygen-free SiC + Cfibers decreases when the C/Si ratio increases from 1 to 1.6. Finally, for three reasons, the Si-C-O fibers consisting ofa SiC + C+ SiO,C 1.J
P = l_exP[_~(.!!.-)m] V a R

or

o

PR

(14a)

o

= l_exp[_~(.!!.-)IIl]

(14b)

o;

L;

where Vo and L. are reference volume and reference length. Equations 14a or 14b can also be rewritten in linear form as:

Ln Ln _1_ = m

i-PR

Lna+(Ln~-m Lna o ) Vo

Ln Ln _i_ = m Ln a + (Ln

(15a)

~ - m Ln a , )

(15b) L; corresponding to the Weibull plots. Finally, if two populations of flaws are present (characterized by two different pairs of Weibull parameters, mdcrol and m2/cro2), a bimodal Weibull distribution isused which can be expressed for fibers ofconstant volume, as: or

i-PR

P = i-exp - ~ - ~ )1Il}] mJ

R

[

(

aol

)

(

a

0

(16a)

2

Equation 16has been used indifferent forms, to show the influence of the fiber length L [21]:

286

Chapter 10

(16b) orthe occurrence ofspecific flaw populations, e.g., a population of internal flaws (ml/crol) and a population ofsurface flaws (mJcr02) [79-82] :

P=l-exP[-~(~J",J _~(~)",1] R

Vo

O"oJ

So

(16c)

0"02

The data for recent Nicalon NL-200 fibers of a given gauge length (L=25 mm) are spread over a wide range from 1500 to 4000 MPa. The average is about 3000 MPa (Figure 12). In a first approximation, these experimental data can be fitted to a unimodal two parameter Weibull distribution. There are often a few data points, usually at low stress levels, which do not actually fall on the straight line predicted by equations of type 15a or 15b, suggesting that more than one flaw ispresent in the fiber.

••

0.90

•

0.60

l!! ::l

S (;

I

0.30

.~

0.10

•

D..

•

• • 1500

Nicalon NL200

2500

3500

Tensile strength. MPa

Figure 12. Weibull plot of tensile strength data for Si-C-O (Nicalon NL 202) fibers. with a unimodal Weibull distribution (80); reproduced with permission from VSP, Zeist, NL.

About 96% ofthe fracture origins in Nicalon NL200 fibers are surface flaws and only 4% are internal flaws. Since the internal flaws are minimal, the corresponding data can be treated as censored data and the tensile strength of the fiber can be statistically analyzed (Figure 12) with a unimodal Weibull distribution, where m = 4.5 and cro = 2670 MPa [80]. Data for an older Nicalon fiber. however, gave a better fit with a bimodal Weibull distribution corresponding to a family ofsurface flaws (m, = 3.64; crOl = 4.64 GPa) and a familyof internal flaws (rn, = 9.41 and cr02 = 5.08) [79]. More recently, the observation of two partly concurrent populations offlaws in Nicalon NL200 fibers has been attributed [82] to extrinsic and intrinsic flaws. Accordingly, a family ofextrinsic flaws (severe flaws at the fiber surface) is responsible for failure at the lowest stress levels

287

Chapter 10

(with m2 = 1,92), and a family of intrinsicflaws (located both atthe surface and in the volume) control failure athigher stress levels (with m, = 4,5), The identification of families of flaws, which are responsible for the failure of a batch of fiber specimens, is not straightforward. Before testing, the specimens must receive a soluble damping coating ofa wax ora polymer such as 1-3 polypropanediol, orthey must be tested in a liquid medium such as glycerol to absorb the shock wave energy due to specimen bursting atfailure. Under such precautionary conditions, the primary failure surfaces can be recovered and the origin ofthe fracture determined [79-80]. The surface flaws are mainly microvoids and microcracks induced by spinning and mechanical abrasion whereas the internal flaws are microvoids or inclusions originating from the PCS precursor fiber or formed during pyrolysis. The failure surface often shows the fracture origin-mirror-mist-hackle-crack branching morphology characteristic of brittle materials. The depth a and the width c of surface flaws in Nicalon NL200 fell in the ranges 0.09 < a < 0.6 IJm and 0.25 < c < 1.7IJm, respectively. The surface flaws were modeled as semi-elliptical cracks, allowing calculation ofthe fiber toughness, KIC, by the Griffith equation: I

O"R

K/c

= y' (J(Q c/

12

(17)

where Y isa geometric parameter depending on the shape ofthe defect (here, Y = 0.8331 , for alc = 0.3). K,C, calculated from failure stress (crR) and flaw depth (a) data, is of the order of 1.55 -1.9MPa m". The tensile strength of Si-C-O fibers decreases after exposure to elevated temperatures. When Nicalon NL 200 fibers are exposed for 1 hour to 1300°Cin argon (P = 100 kPa), their mean tensile strength and scale parameter, cr., decrease by 45% while their Weibull modulus remains unchanged [80-83]. Fibers exposed to more severe conditions (e.g.,for 5 hours in a vacuum at1500°C)are so weak that they cannot be tested. Finally, the fact that oxygen-free fibers maintain their tensile strength under similar conditions relates to the absence of silicon oxycarbide and its decomposition process. 10.7.2 Athigh temperatures The high temperature (HT) properties of SiC based fibers depend upon their pyrolysis temperature, their thermal stability, the time and the test atmosphere. When mechanical tests are performed at a test temperature that is higher than the pyrolysis temperature, some change occurs in the composition orland microstructure ofthe fibers which may strongly affect their mechanical behavior. (a) Tensile tests

Young's modulus and tensile strength of Si-C-O and Si-C fibers decrease when the test temperature is increased (Figure 13). The room temperature tensile strength of the Si-C fibers (Hi-Nicalon) is higher than that of Si-C-O fibers (Nicalon NLM 200), but the two fibers have almost the same strength, (1200 MPa), when tested at 1400°C. At any test temperature, the modulus ofthe Si-C fiber is higher than that ofthe Si-C-O fiber [31] [84]

288

Chapter 10

4 (a)

3 ca

Q.

e

~

15>

<:: ~

2

U;

'"<::

~

~

o HiNicalon • Nicalon NL 202 0

500

0

1000

400

1500 (b)

300/9'"---_ _

t:-

el
.2

~200 100

o HiNicalon • Nicalon NL 202

500

1000

1500

Testtemperature, °C

Figure 13. High-temperature tensile strength (a) and Young's modulus (b) of SiC-based fibers, as assessed through short-term tensile tests performed inaironsingle filaments [31): reproduced with permission.

Si-C-O fibers remain linear elastic and brittle up to ",1300°C [85]. Si-C fibers (Hi-Nicalon) display a non-linear behavior above 1400°C that almost totally disappears after heat treating the fibers at 1600°C, prior to HT tensile testing. This behavior suggests that the viscoplastic character ofthe fibers is related tomicrostructural changes occurring inthe fibers [32]. (b) Creep tests

SiC based fibers derived from PCS precursor fibers are prone to creep at relatively low temperatures due to the presence of a secondary phase (silicon oxycarbide) and extremely small SiC grain sizes. The most creep resistant quasi-stoichiometric fibers are produced at high temperatures (1800°C) and exhibit a relatively large SiC grain size (0.1 to 0.5 IJm). Finally, a highly creep resistant fiber having a large grain size (1-2 IJm) can be produced from SiC powder at2000-2300°C[21] [31-32] [59] [86-89].

Chapter10

289

1.2 tl 1350~

o~ 0.8 ff .= o 0.4

200 -

o

O

I 400

I

I 1200

Pure Argon I

2000

Time, min

Figure 14. Creep of Si-C-O (Nicalon NL 200) fibers under an applied stress of 0.45 GPa as measured in argon [59]; reproduced with permission from the American Ceramic Society, PO Box 6136, Westerville, Ohio 43086-6136.

Si-C-O fibers are prone to creep above 1100~ (Figure 14). The creep rate does not reach steady state prior to failure during tensile tests performed in argon. It decreases continuously with time, indicating that primary (or logarithmic) creep dominates the entire creep life of the fibers. The strain-time curves obey the following classical law: 1

c = - - L n ( 1 + f l ~o t )

P

(18)

where p is a constant and /; o is the strain rate at zero strain. Creep is rate controlled by the viscous flow of the glassy silicon oxycarbide containing free carbon and SiC nanocrystals in suspension [21] [31]. Above 1100~ silicon oxycarbide decomposes with formation of more SiC, resulting in a continuous increase of the viscosity and thus a decrease of the creep rate. Conversely, the creep curves for fibers tested in CO-rich atmospheres also display a large steady state domain (secondary creep) [59]. This linear domain is related to the fact that, under such conditions, the silicon oxycarbide and the fibers are stable. For higher test temperatures (Tt ___1400"C),the fibers are no longer stable for Pco _<100 kPa and the linear domain can no longer be observed. The strain rate & ,/time data in the steady state domain obey the Dorn relationship: ~s~. = A or" exp ( - Q / R T )

(19)

where A is a dimensionless constant, n the stress exponent, Q the apparent activation energy and R the gas constant. The observed value of n (n = 1) corresponds to a Newtonian type viscous flow and that of the activation energy (Q = 435 kJ/mol)is consistent with the activation energies reported for a thermally activated, viscous flow of glasses at high temperatures [31].

290

Chapter 10

SiC + C fibers (e.g., Hi-Nicalon) creep at temperatures as low as 1000°C although they are almost totally free of silicon oxycarbide. This suggests a different creep mechanism (32). A steady state domain is always observed ata test temperature of::;;;1400°C, i.e., when the test temperatures is lower than the assumed processing temperature (Figure 15). Conversely, at a test temperature Tt ~1500°C, the strain rate decreases continuously with time up to failure and primary creep dominates the entire creep life ofthe fiber. The apparent activation energy is 0, = 220 kJ/mol for 1000 < Tt < 1250°C and ~ = 700 kJ/mol for 1250 < Tt < 1400°C, suggesting that an important change occurs in the creep mechanism. The creep rate of Hi-Nicalon type fibers is significantly lowered when the fibers are heat treated at 1400·1600°Cin argon prior to the creep tests (also performed in argon). Further, fibers which are heat treated at 1600°C exhibit a steady state creep domain at Tt = 1500°C, whereas the untreated fibers do not creep when tested at Tt = 1500°C (32). The creep resistance of heat treated fibers at 1400°C in air is higher than that of the untreated fibers (87). In the family ofTyranno Si-M-C-O fibers (M = Ti, ZrorAI) byUbe, a quasi-stoichiometric fiber with a composition close to 68 wt.% Si, 31% C, 0.3% 0, and 0.6% AI (C/Si = t07) offers the highest creep resistance at1300°Cunder a tensile load of 1GPa. The creep mechanism has been reported to be transgranular diffusion (89). Finally, a sintered a-SiC fiber produced by Carborundum has a measurable creep rate ranging from of 5.1 x 10-8 to 1.4 X 10·7S·1 between 1300 and1500°C atstress levels ranging from 70 to 300 MPa (88). From preliminary data, n =1.86 ± 0.50 and 0 =273 ± 12 kJ/mol.

2.0 . . . - - - - - - - - - - - - - - - - - - , 1 GPa

Ar

3.0

2.0

E

E. 1.0
1.0

l-

....I.....

....L...

5

10

---l

0.0

15

Time, ks

Figure 15. Creep of Si-C(Hi-Nicalon) fibers. asassessed from tests performed in argon (P= 100 kPa) on single filaments (32): reproduoed bypermission.

291

Chapter 10

(c) Bend stress relaxation test A simple bend stress relaxation test can be used to compare the creep resistance of individual filaments [43]. A bending stress is applied with a graphite jig to the fiber which acquires a radius of curvature, Ro. The assembly is submitted to a heat treatment. After cooling, the graphite jigpieces are separated revealing a fiber with a creep induced curvature ofradius R (with R 2: Ro). The stress relaxation isquantified by a parameter mdefined as m = 1 - (RJR). For given test conditions, creep resistant fibers are characterized by m-values close to 1(i.e., R » Ro) whereas creeping fibers display low mvalues (R "" Ro). Si-C-O fibers (Nicalon NLM-200) creep at low temperatures (1000-1200°C), whereas the sintered a-SiC fibers (Carborundum) creep at much higher temperatures (1300-1600°C). The other fibers, which are derived from PCS precursor fibers, i.e., oxygen-free Si-C fibers (Hi-Nicalon) and the quasi-stoichiometric fibers (Hi-Nicalon S or Dow Corning fiber) fall between these two limits. The fibers which have been heat treated at 1400-1600°C and display a larger SiC grain size (e.g., 30-50 nm) exhibit a creep resistance similar tothat ofthe a-SiC sintered fiber. In summary, there isaclose relationship between the creep resistance and the SiC grain size. The larger the grain size, the higher the creep resistance. However, since tensile strength usually decreases as the grain size increases, it appears that optimizing the fiber microstructure in order to achieve both a high tensile strength and a good creep resistance may be contradictory [90]. 10.8 Oxidation ofsilicon carbide fibers Depending on the oxidation conditions, the oxidation of silicon-rich non-oxide ceramics proceeds by passive or active oxidation. At low temperatures and high oxygen partial pressures, a protective layer ofsilica isformed by passive oxidation (Equations 20a and 20b). SiC(s)+202( g) ~ Si02(s) + CO2( g)

(20a)

SiC(s)+3 /2 02

(20b)

(g)~Si02(s,l)+CO(g)

At high temperatures and low oxygen partial pressures, silica is no longer formed; it is converted to gaseous SiO by active oxidation (Equation 21). SiC(S)+02 (g)

~SiO(g)+CO(g)

(21)

In the passive oxidation regime, the ability of Si based ceramics to form a continuous protective layer of silica depends upon their silicon content and the volume change occurring during oxidation. It isoften characterized with a parameter, Ll, defined as:

~=A._d_

(22a)

d Si0 1

with

A = CSi

M sio

. __ 1

M.w

(22b)

where CSi is the weight fraction of silicon in the material, MSi0 2 and MSi, the molar weights of silica and silicon, and d and dSi02, the densities of the ceramic material and silica. The oxidation yields a silica scale covering the material when ~ 2: 1. For dense SiC (CSi = 0.70;

292

Chapter 10

dsiC =3.2 g/cm3) , t!. '" 2.2, assuming that glassy silica is actually formed (with dSio2 g/cm 3) .

=2.20

SiC based fibers have t!. values ranging from 1.35 for the Si-C-O fibers (Nicalon NLM 200) and 1.66 for oxygen-free SiC + C fibers (Hi-Nicalon) to almost 2.2 for quasi-stoichiometric fibers (Hi-Nicalon S). Thus, oxidation yields a protective silica layer, which remains glassy up toabout 1200°C, but becomes cristobalite at 1400°C. The glassy sheath contains nanopores and probably some hydrogen as hydroxyl ions arising from the residual hydrogen in the fibers. Conversely, the cristobalite sheath may crack after cooling. When the silica sheath is continuous, it impedes evolution of CO/SiO and fiber decomposition as long as the internal CO/SiO pressure islower than the external pressure [22-23) [29) [32) [91-92]. The oxidation of SiC based fibers becomes noticeable above about BOO°C. It occurs with an overall fractional weight gain t!.rrlmo. The oxidation kinetics can be determined by TGA under flowing oxygen (or air) or/and by SEM ofthe thickness of the silica sheath (Figure 16). When the weight gain remains <10% , the thickness ofthe silica sheath can be calculated from t!.rrlmo [50): I 2

t!. s; A-I mo

e"'- ·r . - _ . 0

(23)

where r, isthe initial radius ofthe fiber and t!., Aare the parameters defined by Equations 22a and 22b. 1600

1400

-6

8

-8

-

In

N'

[ ~ c: ..J

\ \

6°Q

-10

\

'i

1200

1000

T,OC

°

Hi - Nicalon (TGA)

•

Hi - Nicalon (SEM)

t:.

Nicalon (NL202)

600

"

-12

-14 0 0.5

0.6

0.7

0.8

0.9

1.0

1000ff. K-l

Figure 16. OxidaUon kinefics forSi-C fibers under flowing oxygen (100 kPa): Arrhenius plots of the kinetic constant [50): reproduced bypermission.

The variation of t!.rrImo (or of the sheath thickness of silica, e) as a function of time [50) is parabolic up to about 1400°C for the Si-C (Hi-Nicalon) fibers (Figure 16) but only up to 1200°C for the less thermally stable Si-C-O (Nicalon) fibers [22]. This parabolic feature

Chapter 10

293

suggests the oxidation process is rate controlled by diffusion, the rate limiting step being presumably the diffusion ofoxygen across the silica layer. In cylindrical geometry, the surface through which diffusion occurs is not constant. The external surface of the silica layer is larger than the inner surface and both change with time. As long as &1m. (thus, e) remains small «10%), this difference and the gas evolution can be neglected to a first approximation, and the kinetic constant can be calculated byapplying the classical parabolic law for diffusion across aplanar interface: (24a) (24b) kl and k2 are kinetic constants, &Jmo and ei are the weight gain and silica thickness att = O. kl is expressed in reciprocal time unit whereas k2 has the dimension of a diffusion constant (L2t 1) . k1 or k2 is calculated byfitting the experimental data toEquations 24a or 24b. Thermal variations of the kinetic constants obey Arrhenius law since it is a diffusion process (Figure 16): K r = Kooexp - (E; / RT)

(25a)

or

(25b)

Ln K r = Ln Koo - Ea / RT

The corresponding apparent activation energy, Ea, is 107 kJ/mol for Si-C (Hi-Nicalon) fibers [50] and 69-77 kJ/mol for Si-C-O (Nicalon) fibers [22]. These values are lower than those reported for pure dense SiC (e.g., 128 kJ/mol for SiC) deposited by CVD/CVI at1000·C. A different treatment ofTGA based &1m. data accounts for the decrease of the surface area between the growing silica layer and the unreacted fiber, assuming the oxidation process is also rate controlled by diffusion (90). The kinetic constant, I<J, is derived from the contracting disc rate equation: (26) where Xisthe degree ofoxidation, defined for Si-C-O fibers as : 1 !1.m

x=----M SiO ]

-

M SiCO

mo

(27a)

M SiCO MSi02 is the molarweight of Si~ and MSiGO the molecular weight ofthe overall formula depicting the fiber composition. For a fiber having a low oxygen content and an overall formula SiCl~ 0 005, MSiCO = 45.70 g/mol, Equation 27 can be rewritten as: X=_I_ .!1.m 0.315 mo

(27b)

The kinetic constant I<J iscalculated by fitting the experimental Xdata. 10.9 Transport properties of SiC fibers SiC based fibers display semiconductor behavior (Figure 17). The electrical conductivity of the Si-C (Hi-Nicalon) fibers isfour orders of magnitude higher than for the Si-C-O (Nicalon NL

294

Chapter 10

202) fibers [50) atroom temperature. For a given precursor, the electrical properties depend strongly on the pyrolysis temperature. Selected Si-C-O [93) and Si-C-o-Ti [49) fibers are available which exhibit either a low (10-6 to 10·7n-1cm·') or a high (0.5 to 101'11 crrr') electrical conductivity atroom temperature. The electrical conductivity of a PCS residue increases dramatically when the pyrolysis temperature is increased from 700-800°C to 1400-1600°C. This increase is as large as 10 orders ofmagnitude for uncured bulk PCS [62) or6 orders of magnitude for oxygen cured and electron beam cured PCS fibers [18) [29). Further, the activation energy, Ea, decreases with increasing pyrolysis temperature. It is0.4eV for Si-C-Ofibers produced at1000-1200°Cand approaches zero for 1400-1600°C. The dramatic change observed in the electrical behavior of SiC based fibers with increasing pyrolysis temperature isrelated tothe formation orland the organization offree carbon around the SiC crystals. In Si-C-O fibers it occurs above 1200°C. Below 1200°C, the microstructure ofthe fibers iseither amorphous ornanocrystalline, and carbon exists as isolated small BSUs. The material has semiconducting properties (Ea = 0.4 eV). Above 1200°C, decomposition occurs with formation of ~-SiC crystals and free carbon. The increase in electrical conductivity might be related to the removal ofthe glassy silicon oxycarbide and the formation ofacontinuous network ofcarbon around the SiC crystals [18).

o

E o

2:

t:i

Ol

.3

Si-Cfiber

(Hi-Nicalon)

-2 -4

Si-C-O fiber

(NicalonNL202)

-6

2

4

6

8

10

12

14

1oo0rr, K_l

Figure 17. Electrical conductivity of SiC-based fibers asa function of test temperature: Si-C-O (Nicalon, NL 202) and Si-C (Hi-Nicalon) fibers [50): reproduced with permission.

The electrical conductivity of Si-C fibers with low oxygen content increases sharply at first when the pyrolysis temperature is increased from 1200 to 1400°C, and then more slowly. Initially, most of the residual hydrogen is released and the carbon is formed. SUbsequently the SiC crystals grow and an intervening carbon network isformed . The thin carbon layer (or sheath) on the surface of Si-C fibers is not alone responsible for the observed gain in electrical conductivity since its removal by a brief oxidation treatment at 600°C does not markedly affect its electrical behavior [291. Very little is known about the thermal properties of SiC fibers. A thermal conductivity of 12W/m.K and acoefficient ofthermal expansion of3.1 x 10-6 IK have been reported for Si-C-O fibers (Nicalon NL200) [93).

Chapter 10

295

10.10 Applications

Applications for silicon carbide fibers are discussed in Chapter 12, along with applications for carbon fibersand ceramic oxide fibers. REFERENCES [1) [2] [3] [4] [5] [6]

[7]

[8) [9) [10) [11] (12) (13) (14) [15) (16) (17) (18) [19) (20) [21] [22] (23)

K. Okamura, Ceramic fibers from polymer precursors, Composites, 18[2], 107-120 (1987). J. Lipowitz, J..A. Rabe and R. M. Salinger, Ceramic fibers derived from organosilicon polymers, in Handbook of Fiber Science and Technology: Vol. //I. High Technology Fibers, Part C, M. Lewinand J. Preston, eds., Marcel Dekker, New York, 207-273 (1993). W. H. Atwell, Polymeric routes tosilicon carbide and silicon nitride fibers, in Advances inChemisfry Series 224: Silicon-Based Polymer Science, AComprehensive Resource, J.M. Zeigler, F. W. Gordon and M. Fearon, eds., The Amer. Chern. Soc., Wash.,593-606 (1990). A. A. Verma and P. Krishna, Polymorphism and polytypism incrystals, Wiley, NY (1966). D. Seyferth, Synthesis ofsome organosilicon polymers and their pyrolytic conversion toceramics, in Advances in Chemistry Series 224: Silicon-Based Polymer Science, A Comprehensive Resource, J. M. Zeigler, F. W. Gordon and M. Fearon, eds., The American Chemical Soc., Washington, 565-59 (1990). R. M. Laine and F. Babonneau, Preceramic polymer routes tosilicon carbide, Chern. Mater., 5, 260-279 (1993). S. Yajima, Y. Hasegawa, J. Hayashi and M. limura, Synthesis of continuous silicon carbide fibre with high tensile strength and high Young's modulus. Part 1: Synthesis of polycarbosilane as precursor, J. Mater. Sci., 13,2569-76 (1978). Y. Hasegawa, M. Iimura and S. Yajima, Synthesis of continuous silicon carbide fibre. Pt 2: conversion of polycarbosilane fibre into silicon carbide fibres, J.Mater. Sci., 15,720(1980). Y. Hasegawa and K. Okamura, Synthesis of continuous silicon carbide fibre. Part 3: Pyrolysis process of polycarbosilane and structure ofthe products, J. Mater. Sci., 18,3633-48 (1983). Y. Hasegawa and K. Okamura, Synthesis of continuous silicon carbide fibre. Part 4: The structure of polycarbosilane asthe precursor, J. Mater. Sci., 21 , 321 -328 (1986). H. Ichikawa, F. Machino, S. Mitsuno, T. Ishikawa, K. Okamura and Y. Hasegawa, Synthesis of continuous silicon carbide fibre. Part 5:Factors affecting stability of polycarbosilane tooxidation, J. Mater. Sci., 21, 435258(1986). Y. Hasegawa, Synthesis ofcontinuous silicon carbide fibre. Part 6: Pyrolysis process ofcured polycarbosilane fibre and structure ofSiCfibre, J. Mater. Sci., 24, 1177-90 (1989). K. Okamura, Continuous silicon carbide fibers, in Silicon Carbide Ceramics. 1. Fundamental and Solid Reaction, S. Somiya and Y. Inomata, eds., Elsevier Applied Science, London and New York, 99-118 (1991). C. Laffon, A. M. Flank, P. Lagarde, M. Laridjani, R. Hagege, P. Olry, J. Cotteret, J. Dixmier, J. L. Miquel, H. Hommel and A. P. Legrand, Study of Nicalon-based ceramic fibres and powders by EXAFS spectrometry, Xray diffractometry and some additional methods, J. Mater. Sci., 24, 1503-12 (1989). S. M. Bleay, A. R. Chapman, G. Love and V. D. Scott, Electron probe microanalysis ofNicalon fibres, J. Mater. Sci., 27, 5389-96 (1992). P. LeCoustumer, M. Monthioux and A. Oberlin, Understanding Nicalon Fibre, J. Europ. Ceram. Soc., 11,95103 (1993). T. Mah, N. L. Hecht, D. E. McCullum, J. R. Hoenigman, H. M. Kim, A. P. Katz and H. A. Lipsitt, Thermal stability ofSiC fibres (Nicalon),J. Mater. Sci., 19, 1191-1201 (1984). E. Bouillon, D. Mccaer, J. F. Villeneuve, R. Pailler, R. Naslain, M. Monthioux, A. Oberlin, C. Guimon and G. Pfister, Composition-microstructure-property relationships in ceramic monofilaments resulting from the pyrolysis ofa polycarbosilane precursor at800 to 1400·C, J. Mater. Sci., 26, 1517-30 (1991). D. J. Pysher, K. C. Goretta, R. S. Hodder Jr. and R. E. Tressler, Strengths of ceramic fibers at elevated temperatures, J. Amer. Ceram. Soc.,72[2], 284-288 (1989). G. Simon and A. R. Bunsell, Creep behaviour and structural characterization at high temperatures of Nicalon SiCfibres, J.Mater. Sci., 19,3658-70 (1984). N. Jia, R. Bodet and R. E. Tressler, Effects of micro-structural instability on the creep behavior of Si-C-O (Nicalon) fibers inargon, J.Amer. Ceram. Soc., 76[12], 3051 -60 (1993). L. Filipuzzi and R. Naslain, Oxidation kinetics ofSiC-based ceramic fibers, inMater. Sci. Monographs 68. Adv. Structural Inorganic Composites, P. Vincenzini, ed., Elsevier, Amsterdam, 35-46 (1991). T. Shimoo, H. Chen and K. Okamura, High temperature stability of Nicalon under Ar or 02 atmosphere, J. Mater. Sci., 29,456-463 (1994).

296

Chapter 10

(24) C. Valhas, P. Rocabois and C. Bemard, Thermal degradation mechanisms of Nicalon fibre : a thermodynamic simulation, J. Mater. Sci., 29, 5839-46 (1994). [25) K. Okamura, M. Sato, T. Seguchi and S. Kawanishi, Preparation of high-temperature strength SiC fiber, in Controlled Interphases inComposite Materials, H. Ishida, ed., Elsevier Science Publishing, 209-218 (1990). [26) M. Takeda, Y. Imai, H. Ichikawa, T. Ishikawa, T. Seguchi and K. Okamura, Properties of the low oxygen content SiC fiber on high temperature heat treatment, Ceram. Eng. Sci. Proc., 12[7-8),1007-18 (1991). (27) M. Takeda, Y.Ymai, H. Ichikawa, T. Ishikawa, N. Kasai, T. Seguchi and K. Okamura, Thermal stability of the low oxygen silicon carbide fibers derived from polycarbosilane, Ceram. Eng. Sci. Proc., 13 [7-8], 209-217 (1992). [28] M. Sugimoto, T. Shimoo, K. Okamura and T. Seguchi, Reaction mechanisms ofsilicon carbide fiber synthesis byheat treatment ot polycarbosilane fibers cured by radiation: I Evolved gas analysis, J. Amer. Ceram. Soc., 78[4), 1013-17 (1995). [29J G. Chollon, M. Czerniak, R. Pailler, X. Bourrat, R. Naslain, J. P. Pillot and R. Cannet, A model SiC-based fiber with a low oxygen content prepared from a polycarbosilane precursor, J. Mater. Sci., 32, 893-911 (1997).

[301 W. Toreki, C. D. Batich, M. D. Sacks, M. Saleem, G. J. Choi and A. A. Morrone, Polymer-derived silicon carbide fibers with low oxygen content and improved thermo-mechanical stability, Compostes Sci. and Technology, 51 , 145-159 (1994). [31) R. Bodet, X. Bourrat, J. Lamon and R. Naslain, Tensile creep behavior of a silicon carbide-based fiber with a low oxygen content, J. Mat. Sci., 30, 661-677 (1995). (32) G. Chollon, R. Pailler, R. Naslain and P. Olry, Correlation between microstructure and mechanical behavior at high temperatures ofaSiCfiber with a low oxygen content (Hi-Nicalon), J.Mater. Sci., 32, 1133-47 (1997). [33] A. Tazi-Hemida, R. Pailler and R. Naslain, Continuous SiC-based model -mono-filaments with a low free carbon content. Part I: From the pyrolysis ofa polycarbosilane precursor under an atmosphere of hydrogen, J. Mater. Sci., 32, 2359-2366 (1997). [34] M. D. Sacks, A. A. Morrone, G. W. Scheiffele and M. Saleem, Characterization of polymer-derived silicon carbide fibers with low oxygen content, near-stoichiometric composition, and improved thermo-mechanical stability, Ceram. Eng. Sci. Proc., 16[4], 25-35 (1995). [35] M. D. Sacks, G. W. Scheiffele, M. Saleem, G. A. Staab, A. A. Morrone and T. J. Williams, Polymer-derived silicon carbide fibers with near-stoichiometric composition and low oxygen content, Mater. Res. Soc. Symp. Proc.,Vol. 365, Mater. Res. Soc., Pgh., 3-10 (1995). [36) Z. F. Zhang, C. S. Scotto and R. M. Laine, Pure silicon carbide fibers from polymethylsilane, Ceram. Eng. Sci. Proc., 15[4J, 152-161 (1994). [37] J. Lipowitz, J. A. Rabe, G. A. lank, Y. Xu and A. Zangvil, Nanocrystalline silicon carbide fibers derived from organosilicon polymers, inChemical Processing ofAdvanced Materials, L. L. Hench and J. K. West, eds., John Wiley, New York, 767-776 (1992). [38) Y. Xu, A. Zangvil, J. Lipowitz, J. A. Rabe and G. A. lank, Microstructure and micro-{;hemistry of polymerderived crystalline SiC fibers, J. Amer. Ceram. Soc., 76(12), 3034-3040 (1993) [39] J. Lipowitz, J. A. Rabe, L. D. Orr and R. R. Androl, Polymer derived, stoichiometric SiC fibers, Spring MRS meeting, April 1994. [40) T. Ishikawa, S. Kajii, T. Hisayuki and Y. Kohtoku, New type of SiC-sintered fiber and its composite material, Proc. 22Annual Cocoa Beach ConI.,January 20-24, 1998. [41) F. Frechette, B. Dover, V. Venkateswaran and J. Kim, High temperature continuous sintered SiC fiber for composite applications, Ceram. Eng. Sci. Proc., 12[7-8), 992 (1991). [42] G. V. Srinivasan and V. Venkateswaran, Tensilestrength evaluation of polycrystalline SiC fibers, Ceram. Eng. Sci. Proc., 14[7-8), 563-572 (1993). [43] J. A. DiCarlo, Creep limitations of current polycrystalline ceramic fibers, Composites Sci. and Technology, 51 ,

213-222 (1994). [44) M. Birot, J. P. Pillot and J. Dunogues, Comprehensive chemistry of polycarbosilanes, polysilazanes and polycarbosilazanes, asprecursors ofceramics, Chem. Reviews, 95, 1443-77 (1995). [45) T.lshikawa, M. Shibuya and T. Yamamura, The conversion process from polydimethylsilane topolycarbosilane inthe presence ofpolyborodiphenylsiloxane, J. Mater. Sci., 25, 2809-14 (1990). [46] G. D. Soraru, F.Babonneau and J. D. Mackenzie, Structural evolutions from polycarbosilane toSiC ceramic, J. Mater. Sci., 25, 3886-93 (1990). [47] M. Sugimoto, T. Shimoo, K. Okamura and T. Seguchi, Reaction mechanisms ofsilicon carbide fiber synthesis byheat treatment ofpolycarbosilane fibers cured byradiation: II, radical reaction, J. Amer. Ceram. Soc., 78 (7), 1849-52 (1995). [48) S.Yajima, T. Iwai, T. Yamamura, K. Okamura and Y. Hasegawa, Synthesis of polytnano-{;8rbosilane and its conversion into inorganic compounds, J. Mater. Sci., 16, 1349-55 (1981). [49] T. Yamamura, T. Ishikawa, M. Shibuya and K. Okamura, Development of a new continuous Si-Ti-C-O fibre using an organo-metallic polymer precursor, J. Mater. Sci., 23, 2589-94 (1988).

Chapter 10

297

[50) G.Chollon, C. Laporte, R. Pailler, R. Naslain, F.Laanani, M. Monthioux and P. Olry, Thermal stability of PCSderived SiC fiber with a low oxygen content (Hi-Nicalon), J. Mater. Sci.,32, 327-347 (1997). [51) W. Toreki, G.J. Choi, C. D. Batich, M. D. Sacks and M. Saleem, Polymer-derived silicon carbide fibers with low oxygen content. Ceram. Sci. Eng. Proc., 13, 198-208, (1992). [52J W. Toreki and C. D. Batich, SiC fibers having low oxygen content and methods of preparation, US Patent 5,171,722, Dec. 15, 1992. [53] M. Takeda, J. Sakamoto, Y. lmai, H. Ichikawa and T. Ishikawa, Properties ofstoichiometric silicon fiber derived from polycarbosilane, Ceram. Sci. Eng. Proc., 15(4),133-141 (1994). [54) J. Lipowitz, T. Barnard, D. Bujalski, J. Rabe, G. Zank, A. zangvil and Y. Xu, Fine-diameler polycrystalline SiC fibers, Composites Sci. Technology, 51,167-171 (1994). [55] Z. F. Zhang, Y. Mu, F. Babonneau, R. M. Laine, J. F. Harrod and J. A. Rahn, Polymethylsilane asa precursor tohigh purity silicon carbide, inInorganic and Organo-metallic Oligomers and Polymers, J.F. Harrod and R. M. Laine, eds., Kluwer Academic Pub!. NL, 127-146 (1991). [56) A. Tazi-Hemida, R. Pailler, R. Naslain, J. P. Pillot, M. Birot and J. Dunogues, Continuous SiC-based model monofilaments with a low free carbon content. Part 2 : from the pyrolysis of a novel copolymer precursor, J. Mater. Sci., 32, 2367-72 (1997). [57) F. J. Frechette, W. D. G. Boecker, C. H. McMurtry and M. R. Kasprzyk, Non-oxide sintering ceramic fibers, US Patent 4,908,340; March 13, 1990. [58) H. Tenailleau, X. Bourrat, R. Naslain, R. E. Tressler and L. A. Giannuzzi, TEMlEELS characterization of a sintered polycrystalline silicon carbide fiber, J. Amer. Ceram. Soc., 81 , [8], 2037-44 (1998). [59) R. Bodet, J. Lamon, N. Jia and R. E. Tressler, Microstructural stability and creep behavior of Si-C-O (Nicalon) fibers incarbon monoxide and argon environments, J. Amer. Ceram. Soc., 79[10), 2673-86 (1996). (60) P. Schreck, C. Vix-Guterl, P. Ehrburger and J. Lahaye, Reactivity and molecular structure of silicon carbide fibres derived from polycarbosilanes. Part II: XPS-analysis, J. Mater. Sci., 27, 4243-46 (1992). [61) C. Vix-Guterl, Comportement en atmosphere oxydante de composites thermostructuraux SiC/C/SiC, PhD Thesis, no91 MULHO 198, Univ. Haute Alsace, Mulhouse, Oct. 18,1991. [62) E. Bouillon, F. Langlais, R. Pailler, R. Naslain, J. C. Sarthou, A. Delpuech, C. Laffon, P. Lagarde, F. Cruege, P. V. Huong, M. Monthioux and A. Oberlin, On the conversion mechanisms ofa polycarbosilane precursor into a SiC-based ceramic material, J. Mater. Sci.,26,1333-45 (1991). [63) L. Porte and A. Sartre, Evidence fora silicon oxycarbide phase in the Nicalon silicon carbide fibre, J. Mater. Sci., 24, 271-275 (1989). [64) J. Lipowitz, H. A. Freeman, R. T. Chen and E.R. Prack, Composition and structure ofceramic fibers prepared from polymer precursors, Adv. Ceram. Mater., 2 (2), 121 (1987). (65) H. Ichikawa, K. Okamura and T. Seguchi, Oxygen-free ceramic fibers from organosilicon precursors and Ebeam curing, in High-Temperature Ceramic-matrix Composites II: Manufacturing and Materials Development, A. G. Evans, R. Naslain, eds., Ceram. Trans.,58, 65-74 (1995). (66) S. M. Johnson, R. D. Brittain, R. H. Lamoreaux and D. J. Rowcliffe, Degradation mechanisms ofsilicon carbide fibers, J. Amer. Ceram. Soc., C132-C 135, March 1988. (67) Y. Sasaki, Y. Nishina, M. Sato and K. Okamura, Raman study of SiC fibres made from polycarbosilane, J. Mater. Sci., 22, 443-448 (1987). (68) G. S. Bibbo, P. M. Benson and C. G. Pantano, Effect of carbon monoxide partial pressure on the high temperature decornposjton ofNicalon fibre, J. Mater. Sci., 26, 5075-80, (1991). [69] T. Shimoo, H. Chen and K. Okamura, Pyrolysis of Si-C-O fibers (Nicalon) at temperature from 1473 K to 1673 K, J. Ceram. Soc. Japan, Int. Edition, 100,48-52 (1992). [70) O. Delverdier, M. Monthioux, D. Mocaer and R. Pailler, Thermal behavior ofpolymer-derived ceramics. I. Si-C and SI-C-O systems from both commercial and new polycarbosilane (PCS) precursors, J. Europ. Ceram. Soc., 12,27-41 (1993). [71] P. Rocabois, C. Chatillon and C. Bernard, Mass spectrometry experimental investigation and thermodynamic calculation of the Si-C-O system and Si,CA fibre stabil~ , inHigh Temperature ceramic Matrix Composites, R. Naslain, J.Lamon, D. Doumeingts, eds., Woodhead Pub!. Ltd., Abington-Cambridge, UK, 93-100 (1993). [72) P. Schreck, C. Vix-Guterl, P. Ehrburger and J. Lahaye, Reactiv~ and molecular structure of silicon carbide fibres derived from polycarbosilanes. Part I: Thermal behavior and reactivity, J. Mater. Sci., 27, 4237-42 (1992). [73) C. Labrugere, A. Guette and R. Naslain, Mechanical behaviorand microstructural evolution of thermally aged 2D-SiC (ex-PCS)/C/SiC composltes, Revue Composites et Materiaux Avances, Vol. 3, Hors Serie, J.L. Chermantand G. Fantozzi, eds., Hermes, 1993, Paris, 91-111 (1993). [74) M. H. Jaskowiak and J. A. DiCarlo, Pressure effects onthe thermal stability ofsilicon carbide fibers, J. Amer. Ceram. Soc., 72(2), 192-197 (1989). [75] R. Bodet, N. Jia and R. E. Tressler, Therrnomechanical stability of Nicalon fibres in a carbon monoxide environment, J. Europ. Ceram. Soc., 15, 997-1006 (1995).

298

Chapter 10

[76) J. lipowitz, J. A. Rabe, K.T. Nguyen, L. D. Orr and R. R. Androl, Structure and properties of polymer-
CHAPTER 11 SILICON NITRIDE AND BORIDE BASED FIBERS R. Naslain The present chapter is devoted to (1) Si-C-N-O and Si-C-N fibers, which are derived from SiCoO and Si-C fibers byaddition of nitrogen atoms to the organosilicon precursors, (2) Si-N-O and Si-N fibers and (3) boron containing fibers, including the Si-S-O-N; Si-S-N and Si-S-N-C fibers.

11.1

General considerations

Silicon nitride (ShN4) whiskers have excellent mechanical properties, good oxidation resistance, and relatively low density (Chapter 2). They have high modulus (380 GPa) and strength (6 GPa), but lower thermal stability and thermal conductivity than SiC and SiOC fibers. The advantages of continuous ShN4 [1-6] and related fibers [2] [7-12] appear limited relative to continuous SiC and Si-C-O fibers (Chapter 10) but, for applications where insulating fibers are needed, they might be preferred. The addition of heteroatoms (nitrogen, oxygen, and boron) to polymeric organosilicon precursors stabilizes the amorphous state during pyrolysis. For example, in SiSON and SiSCN fibers, this amorphous state has been reported to be stabilized up to surprisingly high temperatures [14-17]. 11.2 Si·C·N·O and Si·C·N fibers Si-C-N-O and Si-C-N fibers are made by spinning, curing and pyrolyzing polysilazane (PSZ) or polycarbosilazane (PCSZ) precursors. PSZs have a Si-N backbone, carbon being present inpendent groups, whereas, in PCSZs, carbon iscontained ina Si-C-N backbone.

11.2.1 Processing Oxygen in Si-C-N-O fibers is introduced in significant amounts when the melt spun precursor fibers are cured by oxidation, but oxygen is present only in trace amounts in Si-C-N fibers where curing ofgreen fibers is achieved under anaerobic conditions.

(a) From polysilazane (PSl) and related fibers Only a few examples ofprecursors are known which have been actually converted into fibers. A review of the processing routes for PSZ precursors leading to Si-C-N(O) ceramics should be consulted for details [18].

Chapler 11

300

The first PSZ precursors were obtained from the thermal condensation! deamination of tris (Nmethylamino) methylsilane (shown in Equation 2) or tris (N-methyl-amino) phenylsilane [19] [20]. Experimental precursor fibers were hand drawn from the melt, cured and pyrolyzed in nitrogen at 1200°C. The overall pyrolysis reaction is shown in Equation 1. The weight loss is about 32%. The resulting fibers were amorphous and black. They had an empirical composition corresponding toSiNl667Co 667 .

(1 ) CHJ

HC

3 n CHfi{NHCH)J 5?!!S 1-4h

/C N--+--

I

J -, /N .... / -+---SI si: CH

I

I

J

N -, /N -CH

/1" NH CH

HC/ J

Hp

I

(2)

J

J

+4n CH~H2

n

Hydridopolysilazane (HPZ) can also beused as a precursor to produce Si-C-N(O) fibers. It is obtained by the exothermic reaction of trichlorosilane HSiCb with hexamethyldisilazane [21] [18]. Condensation yields a HPZ polymer with a (Si-N)n backbone; carbon is in the pendent methyl groups. HPZ polymers have a formula close to (SiH)39J(Me3Sih42(NH)373N226 [21]. Fibers must be drawn from the HPZ melt in inert atmosphere since HPZ is sensitive to moisture and oxygen. Chemical curing is achieved by exposing the green fibers to an argon stream containing trichlorosilane vapor ata temperature below the softening point: (3) In this step there is a decrease in carbon content. The cured fibers are pyrolyzed under flowing nitrogen at 1200°C. Their chemical composition after pyrolysis is about 58 wt. % Si, 10% C, 29% Nand 3% 0 [21-22]. The carbon content of the fibers depends on processing conditions. The empirical formula assigned tothe fibers is close to SiC0403Nl00300091 [22]. The ammonolysis of dimethyldichlorosilane offers another route to PSZ precursor fibers and ultimately silicon nitride fibers. Ina first step, Me2SiCh and MeHSiCh are mixed in nearly a 1:1 molar ratio inbenzene. The mixture is treated with NH3, NH4CI precipitates, and the solvent is removed. The ammonolysis product, a viscous polymer, is converted in a second step at 400°Cinto melt spinnable polymer, and is melt spun. The resulting PSZ precursor fiber has a (Si-N), backbone. Carbon is present in pendent methyl groups, and the empirical formula of the fiber is SiC 17No90H s7 [23]. (b) From polycarbosilazane (PCSl) fibers InPCSZ precursors, carbon atoms are in the polymer backbone. The C-Si-N bonds, yielding tetrahedral Si(C,N)4 units after pyrolysis, are already present. Most oligomers with carbon in the backbone [18] pyrolyze to yield ceramics with an undesirably high carbon content. In contrast, a copolymer route that combines a given silazane monomer with dimethyldichlorosilane (the starting material in the Yajima route) can yield melt spinnable

Chapter 11

301

PCSZ precursor melts and precursor fibers with a -Si-CH 2-NH-Si- backbone (via the Kumada rearrangement) having tailored Si/C/N ratios [8-11] [18] [24]. In the copolymer route, a polysilasilazane (PSSZ) oligomer is prepared by sodium copolycondensation of dimethyldichlorosilane (Me2SiCb) and 1,3-dichloro-1 ,3dimethyldisilazane (CIMeHSi-NH-SiHMeCI) ina 1:1 molar ratio inboiling toluene (Equation 4). Insolubles are removed by filtration, the product is concentrated and NH3is bubbled at O·C into the solution to break all residual Si-Cl bonds. Ammonium chloride is removed by filtration, solvent and low boiling oligomers are distilled offand a low molecular weight PSSZ is obtained ( M w = 2500). It is converted into soluble, fusible, high molecular weight PCSZ (M w =15 000; soft. point =170·C) by heating at 300-370·C. The PSSZlPCSZ conversion occurs with a viscosity increase and the evolution ofhydrogen. CH 3 x

CH3

I

I

+ (I-x) (CH) ] SiCI]

Cl-Si-N-Si-Cl I

H

I

H

I

H

Naj Refluxing toluene

f

C~CHj-

I 3 I CH si- N -Si

I

H

I

H

3

I

Si

I

H

(4)

+ 2 NaCI

I

x

CH 3

l -x

Continuous, 25 IJm diameter monofilament PCSZ precursor fibers, referred to as PCSZ-II, were melt spun [8-9]. Their empirical formula was close to SiC122No45H oOoOJ. They were either chemically cured by oxidation [9] or physically cured by irradiation with y-rays. Two successive weight losses occur during the pyrolysis in argon or nitrogen. The first, at 25450·C, is related to the volatilization of light oligomers and the second, at 450-950·C, was assigned to an evolution of hydrogen and CH.. The pyrolysis yields amorphous fibers whose chemical compositions remain constant within the 950-1400·C temperature range [9] [11]. The empirical formulas for fibers pyrolyzed at 1400·C, are SiCol8 N039 0061 for the fibers cured with oxygen, and SiCo.93No4600 05 for the fibers cured byy-radiation. The evolution of CH. during the pyrolysis step and the empirical formulas of pyrolyzed fibers suggest that carbon isonly partly present in the backbone of the polymer (assuming that CH4 isformed from pendent methyl groups). Finally, the NISi ratio in PCSZ based fibers, Le. 0.460.49, islower than that, 0.90-1 .00, inthe fibers derived from PSZ. 11.2.2 Structure and properties Si-C-N-O and Si-C-N(O) fibers are amorphous (Figure 1) but the molecular composition is still a matter of speculation. Generally speaking, they can be depicted as consisting of a continuum of tetrahedral units wherein silicon is surrounded by carbon, nitrogen and oxygen atoms; the latter are connected by their summits. Furthermore, some free carbon, which results from the pyrolysis of the pendent methyl groups, is usually present asBSUs.

302

Chapter 11

16000 C,Ar

jJ\

1500 C, Ar

-.

0

A

1400 C,Ar 0

16000 C, N2 I

20

I

~ I

I

I

60

40

I

I

80

29, degrees

Figure 1. XRD-paltems of Si-C-N-O fibers produced from PCSZ precursor cured byoxidation and pyrolyzed at various temperatures inargon ornitrogen [10]; reproduced with permission.

(a) Fiber Structure

Si-C-N(O) fibers derived from HPZ precursor fibers are nanoporous and heterogeneous with a skin/core structure. The composition changes from SiOxCy in the external porous surface to SiNxCr in the core. The molecular formula of this fiber is close to 4 mol.% Sieh, 81 mol.% SiNxCr (x = 1.02, Y= 0.23) and 15mol.% free C [22). The presence of complex tetrahedral units is supported by the 29Si NMR spectrum which shows a broad signal covering the chemical shift region expected for silicon oxycarbide, silicon oxynitride and silicon carbonitride units [21). The occurrence of free carbon, expected from the nature of the precursor, is supported by the C 1s XPS pattern [22). The structure of Si-C-N-O and Si-C-N(O) fibers derived from PCSZ is quite similar to that depicted for fibers derived from HPZ with, however, some noticeable differences. In Si-C-N-O fibers derived from PCSZ, the oxygen concentration is relatively high (e.g., 12-20%) and oxygen is distributed homogeneously over the entire cross section whereas it is concentrated in the skin offibers derived from HPZ [9-10). Thus, the tetrahedral units in the continuum are probably SiOxNyC, units rich in oxygen. In Si-C-N(O) fibers derived from PCSZ and cured by irradiation, the concentration of oxygen is very low and the continuum may consist mainly of SiNxCyunits. (b) Thermal stability

Si-C-N-O and Si-C-N(O) fibers remain amorphous up to 1300-1400°C (Figure 1). The N heteroatom in the fiber has a beneficial effect on the thermal stability of the amorphous

Chapter 11

303

tetrahedral continuum. The glassy Si-C-O phase in Nicalon fibers undergoes decomposition/crystallization at1100-1200·C. The core of the fiber derived from HPZ retains its amorphous state and elemental composition, even after long annealing in argon at 1300 or 1400·C. However, noticeable structural change occurs near the fiber surface where oxygen is concentrated, with formation ofcrystalline silica (cristobalite), SbN 20, I)-SiC and channels ofporosity [22). If it is assumed that silica is formed first by decomposition of the outer silicon oxycarbide skin, then SbN20 and SiC might result from slow interactions in the solid state between silica and the SiNxCy core. SiC could also result from decomposition ofthe core near the skin/core interface

0,---__ Organiclinorganic transition

•~

{

20


en<Jl

:::i ai

.Q

80...-----------, Vacuum

c:: o

E

Cl

'Qj

:5:

Argon

40

Decomposition

~40

~ <Jl o

as O'--L..~,,=:x......-.L_-'---'

1000

1200

1400

1600

Tp' · C

60 '-_-L.._--'_ _-'-_---l._ _-'-_--L_ _.L.-_....J..-l o 400 800 1200 1600 Pyrolysis temperature, Tp' ·C

Figure 2. Weight loss during the pyrolysis ofoxygen cured fibers derived from PCSZ, in a flow ofargon (P = 100 kPa ; Q = 1/hourt ). The inset shows the gas evolution by high resolution mass spectrometry from a fiber prepyrolyzed at1200'C under argon (10); reproduced with permission.

Decomposition of Si-C-N-O fibers derived from PCSZ occurs above 1400·C when performed in an argon flow, with evolution of CO (and presumably SiO), N2(and to a lesser extent H2) and a corresponding 44% weight loss [10]. At 1600·C, when the decomposition is complete (Figure 2), the fibers consist of a mixture of SiC + C which is almost free of oxygen and nitrogen. Decomposition/crystallization starts at the fiber surface and progresses radially toward the fiber axis. The skin is highly porous and consists of a I)-SiC + C mixture almost free of oxygen and nitrogen. The loss of the heteroatoms (N, 0) facilitates the formation of SiC. tetrahedral sites and therefore the nucleation and growth ofcrystalline SiC.

304

Chapler 11

The decomposition of PCSZ-based Si-C-N-O fibers in nitrogen proceeds by a mixed decomposition/nitriding process, rather than pure decomposition. It proceeds by a radial diffusion process and yields a skin/core microstructure in partially transformed fibers, but there are two diffusion fluxes in opposite directions. The skin loses part of its oxygen and gains nitrogen. It remains amorphous since the overall concentration ofthe two heteroatoms (N, 0) remains high. This prevents the formation of SiC4 tetrahedral sites and impedes nucleation and crystal growth [10J. When the process is complete, the chemical composition ofthe fiber isSiCoaNo/Oo2. PCSZ based Si-C-N(O) fibers, which have been cured by irradiation, are almost free of oxygen, display high thermal stability in nitrogen, and retain a composition that remains constant up to 1600°C[11 J. The only noticeable phenomena are the formation of a very thin carbon skin and of tiny P-SiCcrystals in an amorphous Si-C-N matrix. Thus, the tetrahedral SiC,Ny units are stable at 1600°Cunder an external nitrogen pressure of 100 kPa. Further, the high nitrogen concentration of the fiber (19 at.%) prevents a rapid formation of SiC4 sites and thus impedes to some extent the nucleation and growth of SiC crystals. Conversely, the fibers undergo some decomposition above 1400°C in argon, with formation of a skin/core microstructure. However, the process kinetics remain relatively slow. (c) Mechanical properties

Si-C-N-O and Si-C-N(O) fibers exhibit linear elastic tensile behavior up to failure. At room temperature HPZ based fibers have tensile strengths ranging from 1.9 GPa [22J to 3.1 GPa [21J, and elastic moduli ranging from 200 GPa [22J to260 GPa [21J. PCSZ based fibers have similar mechanical properties. Oxygen cured Si-C-N-O ceramic fibers and radiation cured SiC-N(O) fibers have moduli of 175 GPa and 215 GPa, respectively, and tensile strengths of 1.8 GPa and 2.4GPa, respectively [10-11]. The room temperature properties of Si-C-N-O as well as Si-C-N(O) fibers depend upon the thermal history of the fibers. For example, the strength of HPZ based fibers shows a decrease of 28,40 and 68% after a 15 hour exposure, in argon, to 1200, 1300 and 1400°C, respectively (22). The strength degradation is probably related to surface damage since the composition and the amorphous nature of the fiber remains constant. The effect of thermal history on mechanical properties of PCSZ based fibers tested at room temperature is shown in Figure 3 [10-11]. After pyrolysis at 1600°C, Si-C-N(O) fibers still exhibit a room temperature strength of 2 GPa, whereas the Si-C-O Nicalon fibers can no longer be tensile tested. As expected from their amorphous structure, Si-C-N-O and Si-C-N(O) fibers are prone to creep. For HPZ based fibers, the apparent energy for creep is -200 kJ/mol in the temperature range between 1150 and 1350·C, a value that is consistent with activation energies for thermally activated viscous flow ofglasses athigh temperatures [25]. (d) Oxidation resistance

Oxidation of PCSZ based Si-C-N-O fibers proceeds with an overall weight increase and evolution of carbon oxides, molecular nitrogen and presumably some H20 [12J. Under conditions ofpassive oxidation, the silica sheath is protective. The ~-parameter (see Chapter 10.10.8), Le., ~ = 1.26, is lower than that reported for Si-C-O Nicalon fibers but still higher than unity.

305

Chapter 11

3000 . - - - - - - - - - - - - - - - - - - - - - - ,

Si-C-N

O'-----'------L-----'------'---.l...---' 600

800

1000 1200 1400 Pyrolysis temperature. °C

1600

250 . - - - - - - - - - - - - - - - - - - - - - - ,

eo ui 1Il

:l

Si-C-N

150

"5 "C 0

E 1Il

-0>

c:

:l

~

50

600

Si-C-N-O

800

1000 1200 1400 Pyrolysis temperature, °C

1600

Figure 3. Room temperature strength and modulus of y-ray cured Si-C-N fibers and oxygen cured Si-C-N-O fibers [11]: reproduced with permission.

Growth kinetics for the silica sheath obey parabolic laws (Equations 24a and 24b, Chapter 10). Thermal variations of the parabolic constants follow an Arrhenius law (Equation 25, Chapter 10). The apparent activation energy is 170 kJ/mol and the pre-exponential factor is Ko= 3 X 106 nm2 S·l for ~ [12]. Thus, the oxidation of Si-C-N-O fibers is rate controlled by diffusion. Compared with the oxidation kinetics of the PCS based Si-C-O fibers, those of PCSZ based Si-C-N-O fibers differ in important points. (1) The oxidation parabolic constant, Figure 4, is lower. (2) The activation energy for Si-C-N-O fibers, E.= 170 kJ/mole, is higher than that for comparable Si-C-O fibers, E. = 70 kJ/mol [26]. (3) The parabolic constant for SiC-N-O fibers ishigher atlow temperatures than those for bulk SbN4 and CVD-SbN4. However, the differences strongly decrease as temperature israised.

306

Chapter 11

HP-SiC

3

Bulk

Si3N4

in

'"Ec

-1

Si-C-N-O fiber

C\i ~

c

...J

-3

-5

-7

' - - _ - ' - - _ . . . L - _ - ' - _ - - ' - _ - . l . _ - '_ _"----I

0.5

0.6

0.7

0.8

0.9

Temperature,103K-l

Figure 4. Arrhenius plots ofthethermal variations ofthe kinetic parabolic constant ks, fortheoxidation of Si-C-Nfibers derived from PCSZ and related Sibased ceramics, according toref. (12); reproduced bypermission.

o

(e) Other properties

Si-C-N-O and Si-C-N(O) fibers have a low density (2.3-2.4 g/cm 3 for HPZ based fibers and 2.5 g/cm 3 for PCSZ based fibers). They also exhibit a low coefficient of thermal expansion. Its value is 3.0-3.9 x 10-6 " for the HPZ based fibers in the 350-1000·C temperature range. PCSZ based Si-C-N-O fibers exhibit low electrical conductivity with semiconducting behavior and alow apparent activation energy [10J. 0C

11.3 Si·N·O and Si·N fibers Continuous Si-N-O and Si-N fibers, almost free of carbon, can be prepared by pyrolysis of polycarbosilane precursor fibers in ammonia. The precursors have been discussed for fabricating Si-C-O and Si-C fibers (Chapter 10). 11.3.1

Processing

The pyrolysis of oxygen cured PCS fibers yields continuous Si-N-O fibers, whereas that of PCS based fibers, cured by irradiation under anaerobic conditions, affords continuous Si-N fibers [7J [27J. Their main advantage istheir extremely low electrical conductivity. (a) From Yajima type polycarbosilane (PCS) fibers

Chapter 11

307

When a Yajima type PCS is heated in an ammonia gas flow, it undergoes a large weight loss that is partly related to phenomena which have been already described for PCS pyrolysis in an inert atmosphere (Chapter 10). It is also related to nitridation of the Si-CH3, Si-CH2-Si and Si-H bonds which takes place between 400 and 800°C with a sharp decrease in the carbon content of the fiber. When the nitridation is complete, the fiber contains almost no carbon. 80th the pendent methyl groups and the methylene groups in the backbone have reacted with ammonia [28). The nitridation of PCS can be followed from the change occurring in the 29Si high resolution NMR signal [29]. Starting from a PCS fiber with a composition close to SiC193H 47IOool, the material is first cured by E-beam irradiation under anaerobic conditions. In a second step, the cured fiber is pyrolyzed in a gas flow of ammonia, yielding at1200°Can amorphous Si·N ceramic fiber with a composition close to SiN1.26CO.olHo.ls. When the pyrolysis temperature is raised to 1400°C, the fiber crystallizes. It is essentially free of carbon and oxygen and its empirical formula, SiN w orSbN m , isclose to that ofSbN 4 [7]. The composition of the Si-N-O fibers from the pyrolysis of oxygen cured PCS fibers in an ammonia gas flow depends on the curing conditions. Starting with the Yajima type PCS precursor, SiC1.93H 47IOool , Si-N-O fibers were obtained with compositions such as SiNll sOo.47 having an O/Si atomic ratio close to 0.4. These fibers remained amorphous up to 1500°C[7]

[27].

(b) From perhydropolysiJazane (PHPSZ) fibers PSZs without pendent carbon bearing groups, which are precursors to silicon nitride ceramics, can be obtained by ammonolysis of dichlorosilane, H2SiCb, or a stable complex, with a tertiary amine [18]. A stable H2SiCb/2-pyridine complex has been used toproduce Si-N fibers (Equations 5a, 5b). Pyridine acts as a catalyst in the condensation/crosslinking reaction, whereas other solvents, such as xylene, have the reverse effect [4-5]. The end product isa clear solution ofperhydropolysilazane (PHPSZ). A dry spinnable PHPSZ polymer (Mn =2000; Mw =12000) has an empirical composition close to SiN,.ooHo.950o.o3 and a structure consisting tentatively of rings linked by linear segments (Equation 5c). Green fibers are dry spun in an inert atmosphere from a PHPSZlxyiene solution. Vaporization of the solvent yields a material that does not melt or soften when the temperature is further raised and thus does not require a curing treatment.

308

Chapter 11

(5a) (5b)

+2Hz

-SiH-NH-

I

NH

I

-SiH-NH-

-N H, H/

.su:

H

':sr: H

S, k. .

Si

H H H .:. sf I

H

~ _~i_H 'N I

(5 c)

Si- H

I I N N, ./N -Si-NI:J........... / ' Si - N ' \ H/ Sl I -, Si I \ \ H ~ H H H/ \ H H H H

d

The green fibers are pyrolyzed in a flow of ammonia gas at 1000°C to complete the condensation/nitridation/crosslinking process and to increase the NISi ratio. The nitridation reaction occurs between 400 and 600°C. At600°C, the NISi ratio isclose tothat of ShN4 [30J. At 1000°C,the fibers consist of an amorphous hydrogenated silicon nitride. Finally, the fibers are densified by a heat treatment at 1400°C in nitrogen. The fibers have elliptical cross sections, and they are colorless, transparent, amorphous and almost free of carbon and oxygen. Their composition, SiN'2..0o.079Co.o'6, isclose tothat of ShN4 (or SiN1333).

(e) From other polysilazane fibers The curing of HPZ fibers byHSiCh could theoretically remove all pendent methyl groups if the reaction according to Equation 3, goes to completion. The fibers, produced bypyrolysis in a nitrogen gas flow, are almost carbon -free. Their actual composition is close to SiNuI03C040300091, typical of a silicon carbonitride fiber [22J. However, early fibers which had a low NISi ratio but much lower carbon content were found to correspond to a composition close to SiN1069C0 06900 064 [21 J. 11 .3.2 Structure and Properties Si-N-O and Si-N fibers are amorphous. The thermal stability of the amorphous continuum depends on the fiber composition, which is itself related tothe nature of the precursor and the pyrolysis temperature (Tp).

(a) Thermal stability PCS based Si-N-O fibers remain amorphous up to at least Tp = 1400°C whereas their Si-N counterparts undergo crystallization at 1300 < Tp < 1400°C [7J. For both Si-N-O and Si-N fibers, the crystalline phase which is formed is a-ShN4. PHPSZ based fibers with a composition close to SiN12l16 00.108 C0028 are amorphous when prepared at Tp = 1300°C. Their

Chapter 11

309

crystallization to a-SbN4 in nitrogen, obeys a first order kinetic law within the 1400-1500'C temperature range [6]:

da ldt=k(J -a)

(6)

where a is the weight fraction of crystalline a-SbN4in the fiber at time t; k is a kinetic constant. Temperature variations of k obey an Arrhenius law with an apparent activation energy of E. = 602 kJ/mol.

(b) Mechanical properties The mechanical properties of Si-N-O and Si-N fibers, measured in tensile loading at room temperature, depend on their processing conditions. For example, the modulus of PCS based Si-N-O fibers increases from 80 to 180 GPa as pyrolysis temperature is raised from 1000 to 1500'C. Their tensile strength shows a maximum close to 1.8 GPa when pyrolyzed between 1300 and 1400'C, and then decreases. The decrease isrelated to the onset offiber decomposition/crystallization, but, even after pyrolysis at 1500'C, the fibers still display a tensile strength of about 1 GPa [7]. 10 IJm diameter PHPSZ based fibers, which are prepared at 1300'C, have a composition close to SiN12200.'0C002, a Young's modulus of about 220 GPa and a tensile strength of 2.5 GPa (L = 25 mm) at room temperature. However, tensile strengths as high as 4 GPa are achieved when the fiber diameter is reduced to 6 IJm [5]. Annealing in nitrogen does not significantly decrease the RT strength of the fibers as long as the annealing temperature is maintained below the onset of the formation ofcrystalline a-SbN4.

(c) Other properties Si-N-O and Si-N fibers display a density ranging from 2.3 to 2.4g/cm3 and a low coefficient of thermal expansion (1.5 x 10-6 'C·l for fibers derived from PHPSZ). Furthermore, they display low electrical conductivity. The resistivity of the fibers is 5.0x 10'4Oem. Finally, Si·N·O and Si-N fibers exhibit good oxidation resistance, which is related to the formation of a silica surface in the passive oxidation regime. 11 .4 Si·B·Q·N, Si·B·N and Si·B·N·C fibers The insertion of boron into Si-N and Si-N-O shifts the onset of SbN4 crystallization to higher temperatures. This recognition has been used to produce experimental Si-B-O-N [15-16], SiS-N [17] and Si-S-N-C [17] [31] fibers. Pyrolysis in ammonia yields colorless and nearly carbon-free fibers, while pyrolysis in nitrogen or argon produces black, carbon containing fibers. The fibers are amorphous after pyrolysis at 1000'C; the thermal stability of the amorphous state is very high. For example, Si-S-N-C fibers with a composition close to SiSN 3C crystallize only at 1800-1900'C (17). 11.4.1

Processing

Two process routes are known. One starts with perhydropolysilazanes (PHPSZs), the other with trichlorosilylaminodichloroborane (TADB).

Chapter 11

310

(a) From perhydropolysilazane (PHPSZs) fibers Polyborosilazanes (PBSZs) are produced by reaction of PHPSZ with trimethyl borate (SIIB = 3) in pyridine, The solution is heated for 3 hours at 120·C. After cooling, o-xylene is added and the solvent is removed by vacuum distillation at50·C, yielding a white PBSZ powder (Mn =2100-2200) that is soluble in toluene and chloroform but does not melt < 500·C [16]. The PBSZ composition isclose to SiBo ,380o ,3iN1.o5Co ,31Hu~ . The chemical bonding between PHPSZ and trlmethyl borate is mainly via B-N bonds formed according to the following reaction:

t ~-NHt- + H

/I

(CHP)JB -

t~-N+I

H

+CHpH

(7)

/I

B(OCH J)2

SI-O bonds can also be formed as the result of a reaction of Si-H bonds in PHPSZ with CH30H, orwith the trimethyl borate itself:

tf-NHto I

+CH4

(8)

/I

B(OCH)2

PBSZ is a polymer consisting mainly of the structural units HSiN3, H2SIN2, H3SIN, HSION2, NB(OCH3h and NB(OCH3)N. PBSZ fibers are dry spun at 60·C from a filtered solution. Curing is not necessary since the material isinfusible after vaporization ofthe solvent. The green fibers can be pyrolyzed so as to either retain or to remove carbon. Nearly-carbon-free Si-B-O-N fibers are obtained by a two step pyrolysis. The green fiber is first pyrolyzedat 800·C in ammonia, yielding an amorphous fiber with a composition close to SiBo,.aOo,~NwCo ,olHo,ae that is nearly hydrogenfree. This fiber is then annealed in nitrogen at1500·C, yielding a composition that isclose to SIBo,.aOO,304Nl ,~CO ,OlHo ,~. If the pyrolysis is directly performed in nitrogen at 1500·C, very little carbon is removed, and the composition ofthe fiber isclose to SIBo,380o,38N1.1ICo,23Ho," [16]. Nearly oxygen-free PBSZs can be prepared by the reaction of PHPSZ with tris (dlmethylamino) borane in pyridine, in the presence of ammonia, by a procedure similar to that used for the synthesis of PBSZs from PHPSZ and trimethyl borate. The product is a clear solution of PBSZ in o-xylene or a white powder insoluble in organic solvents and infusible up to 500·C when the solvent is removed by distillation [31]. This PBSZ isa polymer consisting of HSIN3, H2SIN2, H3SiN and substituted borazine. Its empirical formula is close to SiBo,aeN2,38C1.380ooaH5.25. The pyrolysis of this PBSZ in a gas flow yields a residue which remains amorphous up to Tp i'= 1700·C and whose composition Is close to SiBo,aaNl,IaCo,330o,05HMI.

Chapter 11

311

(b) From trichlorosilylamino-dichloroborane (TADB)fibers

PBSZs, precursors to Si·B·N or SI-B-N-C oxygen-free fibers, can also be prepared by reacting trichlorosllylamino-dichloroborane (TAOB) with CI3Si-NH-BCb and methylamine, H3C-NH2 [17] [32]. The advantage ofTAOB lies in the fact that the Si-N·B bonds are already present in the main starting material. The low temperature reaction yields hexane and yields a soluble, thermosetting oligomer that can be further polymerized by heating at 160·C in a vacuum. Polymerization gives a yellow, glass·like PBSZ polymer (Mn =2500) that is fusible and soluble in organic solvents, but is sensitive to moisture and oxygen. Its composition, SiBN432C3le0007Hl1.11, corresponds to the formula SbB2(NHMNHCH3MNCH3) and its structure consists ofslx·membered Si3(NCH3)3 rings connected via HN-B- and N(CH3)·B- groups. Fibers are obtained from the PBSZ precursor either by melt spinning at 140·180·C or by dry spinning. No specific curing step is necessary. Pyrolysis at 1500·C yields amorphous fibers free ofoxygen. When the pyrolysis is performed in nitrogen or argon, the fibers are black and smooth, with a composition close to SIBN3C, and contain a significant amount of carbon. Conversely, when the pyrolysis is performed up to 1000·C in ammonia, the majority of the organic functionality Is removed at 800·C and the fibers are colorless; their composition is close to SbB3N7 [17] [32]. 11.4.2 Structures and properties Si-B·O-N, Si·B-N and Si-B-C-N fibers are amorphous; the thermal stability oftheir amorphous state is much greater than that observed for Si·N, Si·N·O and Si·C·N·O fibers. Thus, boron, alone or associated with oxygen, Impedes the crystallization ofthe pyrolytic residue In SbN4.

(a) Structure and thermal stability Amorphous SI-B-O-N ceramics are known to consist oftetrahedral SiN4.•O. units (with x =0, 1, 2, or 3) and trigonal BN3-yOy units (with y =0 or 1), where each unit is crosslinked through a three-dimensional network of Si·N·Si, Si-N·B, SI-O-Si, Si-O-B, B-N-B and B-O-B bonds [30]. SI·B-O-N fibers are mainly amorphous when pyrolyzed at 1700·C, but they crystallize at 1800·C In nitrogen (Figure 5). Crystallization, which is associated with a major (23%) weight loss and a decrease in the oxygen content, yields a mixture of c- and P-SbN4 but no crystalline h·BN. Its composition is close to Si3BN4. Oxygen-free Si-B-C·N ceramics resulting from the pyrolysis of polyborosilazanes behave similarly when annealed at high temperatures in nitrogen. When pyrolyzed at 1700·C, the pyrolytic residue ofPBSZ that resulted from the reaction ofPHPSZ with tris (dimethylamino) borane is amorphous, with each boron atom surrounded by three nitrogen atoms. Its composition is close to SbB3N3C. Crystallization, which occurs at about 1800·C is accompanied by weight loss; ityields amixture ofc- and p·SbN4 and P-SiC, but no crystalline h·BN [31]. The pyrolytlc residue ofPBSZ resulting from the reaction ofTAOB with methylamine, having a composition close to SiBN3C, is thermally more stable when annealed in nitrogen. Crystallization above 1900·C yields amixture of c- and p·SbN4 but no crystalline h-BN or B4C phase. At 1800·C, when annealing is performed in argon, crystallization yields Si3N4 and SIC, and is accompanied by a7wI. %loss ofnitrogen [17].

312

Chapter 11

1800° C

10

20

30

40

50

60

29, degrees

Figure 5. XRD patterns of Si-B-O-N fibers annealed in an atmosphere of nitrogen at increasing temperatures [16): reproduced with permission.

(b) Mechanical properties

The room temperature tensile strength of annealed Si-B-O-N fibers, measured at room temperature, is much higher than that of the Si-N fibers prepared from the same PHPSZ precursor (Figure 6). The tensile strength of SoN fibers decreases above an annealing temperature of 1300°C, the onset of crystallization, and the fibers are too brittle to be tested after pyrolysis at 1600°C. The tensile strength of the Si-B-O-N fibers is still >2 GPa after annealing at 1600°C, and the fibers are too brittle tobe tested after heat annealing at1800°C. The difference of about 300°C where the fibers exhibit the same residual tensile strength correlates with the increase in thermal stability of the amorphous state related to the addition ofboron [16].

11.5 Applications The fibers described in this chapter are still experimental materials. Si-N or Si-N-O fibers might find use in applications where good mechanical properties, good oxidation resistance and low electrical conductivity are required. Si-B-O-N or Si-B-C-N fibers, displaying high

Chapter 11

313

thermal stability, good mechanical properties at high temperatures and oxidation resistance, are potentially useful for reinforcing ceramic matrices. Chapter 12 deals with applications for carbon, ceramic oxide and silicon carbide fibers.

3

-o----'

oL--.l_ _....1..._ _..L-_--'I..lir_--'-_ _ 1700 1800 1600 1400 1500 1300

Temperature, · C

Figure 6. Tensile strength measured at room temperature after annealing at high temperatures for Si-N fibers derived from PHPSZ (ti) and Si-B-Q-N fibers derived from PBSZ (0). The fibers were first pyrolyzed under flowing ammonia at 1200·C and then annealed under flowing nitrogen at 1300-1800·C, except (e) Si-B-O-N fiber, derived from PBSZ, which was simply pyrolyzed to 1500·C under flowing nitrogen according to ref. (16); reproduced with permission.

REFERENCES (1) (2) [3] (4) (5) [6J

[7J

K. Okamura, Ceramic fibers from polymer precursors, Composites, 18[2], 107-120 (1987). J. lipowitz,J.A. Rabe and R. M. Salinger, Ceramic fibers derived from organosilicon polymers, in Handbook of Fiber Science and Technology: Vol. 11/. High Technology Fibers, Part C, M. Lewinand J. Preston, eds., Marcel Dekker, New York, 207-270 (1993). W. H. Atwell, Polymeric routes tosilicon carbide and silicon nitride fibers, in Advances inChemistry Series 224: Silicon-Based Polymer Science, A Comprehensive Resource, J. M. Zeigler and F. W. Gordon, M. Fearon, eds., The American Chern. Soc.,Wash., 593 (1990). T. Isoda, Surface of high purity silicon nitride fiber made from perhydropolysilazane, in Controlled Interphases ;n CompositeMaterials, H.lshida, ed., Elsevier Science Publishing, Amsterdam, 255-265 (1990). O. Funayama, M. Arai, Y. Tashiro, H. Aoki, T. Suzuki, K. Tamura, H. Kaya, H. Nishii and T. Isoda, Tensile strength of silicon nitride fibers produced from perhydropolysilazane, Nippon Seramikkusu Kyokai Gakujutsu Ronbushi, 98(1), 104-107 (1990). H. Matsuo, O. Funayama, T. Kato, H. Kaya and T. Isoda, Crystallization behavior of high purity amorphous silicon nitride fiber, J. Ceram. Soc. Japan, 102 [5], 409-413 (1994). K. Okamura, M. Sato and Y. Hasegawa, Silicon nitride fibers and silicon oxynitride fibers obtained by the nitridation ofpolycarbosilane, Ceramics Intern., 13,55-61 (1987).

314 (8) (9)

[10]

[llJ [12J [13J [14] (15) (16) [17J [18] [19] [20] [21] [22] [23] [24] [25] [26] (27) [28J [29J [30J [31] [32]

Chapter 11 D. Mocaer, R. Palller, R. Naslaln, C. Richard, J. P. Plllot, J. Dunogues, C. Gerardln and F. Taulelle, SI·C·N ceramics with a high microstructural stability elaborated from the pyrolysis of new polycerbosllazane precursors. Part I.The organic-Inorganic transition, J. Mater. ScI., 28, 2615·31 (1993). D. Mocaer, R. Palller, R. Naslaln, C. Richard, J.P. Plllot, J.Dunogues, O. Delverdler and M. Monthloux, SI-C·N ceramics with a high micro-structural stability elaborated from the pyrolysis of new polycarbosllazane precursors. Part II. Effect ofoxygen curing on properties ofex-PCSZ monofilaments, J. Mater. Sci., 28, 2632· 38 (1993). D. Mocaer, R. Palller, R. Naslaln, C. Richard, J. P. Plllot, J. Dunogues, O. Delverdler and M. Monthloux, SI·C·N ceramics with a high micro-structural stability elaborated from the pyrolysis of new polycarbosllazane precursors. Part III. Effect ofpyrolysis conditions on the natura and properties ofoxygen-curad derived monofilaments, J. Mater. Sci., 28, 2639-2653 (1993). D. Moceer, R. Palller, R. Naslaln, C. Richard, J. P. Plllot, J. Dunogues, C. Damez, M. Chambon and M. Lahaye, SI·C·N ceramics with a high microstructural stability elaborated from the pyrolysis of new polycerbosllazane precursors. Part IV. Oxygen·frae model monofilaments, J.Mater. ScI., 28, 3049-58 (1993). D. Moceer, G. Chollon, R. Palller, L. Flllpuzzi and R. Neslaln, SI-C-N ceramics with a high mlcrostructurel stability elaborated from the pyrolysis of new polycarbosllazane pracursors. Part V. Oxidation kinetics of model filaments, J.Mater. ScI., 28, 3059-68 (1993). T. Ishikawa, Recent developments ofthe SIC fiber Nlcalon and Its composites, Including properties ofthe SIC fiber HI·Nlcelon for Ultra-high temperature, Composites ScI. and Technology, 51,135·144 (1994). S. R. Rlceltlello, M. S. Hsu and T. S. Chen, Ceramic fibers from SI-B·C polymer pracursors, Sampe Quarterly, April Issue, 9·14 (1993). O. Funayama, T. Kato, Y. Tashiro and T. Isoda, Synthesis of a polyborosllazane and Its conversion Into Inorganic compounds, J. Amer. Ceram. Soc.,76 (3), 717·723 (1993). O. Funayama, H. Nakahara, A. Tezuka, T. IshII and T. Isoda, Development of SI-B-O·N fibras from polyborosllazane, J. Mater. Sci., 29, 2238-44 (1994). H. P. Baldus, G. Passing, D. Spom and A. Therauf, SI·B(N, C) a new ceramic material for high performance applications, Proc. In!. Conf. High Temperature Ceramic Matrix Composites, HT-CMC-2, A. G. Evans and R. Naslaln, eds.,Ceram.Trans., 58, 75·84, The Amer. Ceram. Soc" Westerville, USA (1995). M. Blrot, J. P. Plllot and J. Dunogues, Comprehensive chemistry of polycerbosllanes, polysllazanes and polycerbosllazanes, as precursors ofceramics, Chem. Reviews, 95, 1443-77 (1995). W. Verbeck, German Patent 2,218,960 (1973). B. G. Penn, J.G. Daniels, F. E. Ledbetter and J.M. Clemons, Preparation ofsilicon cerbide-slllcon nitride fibers by the pyrolysis ofpolycerbosllazane precursors: areview, Polymer Eng. ScI., 26 [17], 1191·94 (1986). G. E. Legrow, T. F. Lim, J. Llpowltz and R. S. Reaoch, Ceramics from hydrodopolysllazane, Amer. Ceram. Soc. Bull., 66 12J, 363-367 (1987). R. Bodet, N.Jla and R. E.Tressler, Microstructural Instability and the resultant strength ofSI-C-O (Nlcelon) and SI·N·C·O (HPZ) fibres, J. Europ.Ceram. Soc., 16,653-664 (1996). Y. C. Song, Y. Zhao, X. Feng and Y. Lu, Synthesis and pyrolysis ofpolysllazane as the precursor ofShNJSIC ceramic, J.Mater. ScI., 29, 5745·56 (1994). O. Delverdler, M. Monthloux, D. Moceer and R. Palller, Thermal behavior ofpo1¥mer-derlved ceramics. IV: SIC·N·O fibers from oxygen-curad polycerbosllazane, J. Europ. Ceram. Soc.,14, 313-325 (1994). R. E. Tressler and D. J.Pysher, Mechanlcel behavior ofhigh strangth ceramic fibers athigh temperatures, In Advanced Structurellnorganlc Composltas, P. Vlcenzlnl, ed., Elsevier, Amsterdam, 3·18 (1991). G. Chollon, Oxidation behavior ofceramic fibers from the SI·C·N·O system and SUbsystems, Key Engineering Materials, 164/165, 395-398 (1989). K. Okamura, M. Sato and Y. Hasegawa, SI-N-O fiber and SI·TI·C fiber obtained from polycarbosllane, In Proc. 5th In!. Conf. Composite Matarlals, W. C. Harrigan, J. Strife and A. K. Dhlngra, eds., The Metallurgical Society, Warrendale, PA, 535-544 (1985). S. Kamlmura, K. Watanabe, N. Kasal, T. Seguchl and K. Okamura, Silicon nitride fiber synthesis from polycerbosllane fiber by radiation curing and pyrolysis under ammonia, Ceram. Transactions, 58, 281·286 (1995). T. Takl, M. Inul, K. Okamura and M. Sato, A stUdy ofthe nitridation process ofpolycarbosllane by solld·state high resolution NMR, J.Mater. ScI. Letters, 8, 1119·24 (1989). O. Funayama, Y. Tashiro, A. Kamo, K. Okamura and T. Isoda, Conversion mechanism ofperhydropolysllazane Into silicon nltride·based ceramics, J. Mater. Sci., 29, 4883·88 (1994). O. Funayama, T. Aokl and T. Isoda, Synthesis and pyrolysis ofpolyborosllazane with low oxygen content, J. Ceram. Soc. Japan, 104 [5J, 355·360 (1996). H. P. Baldus, O. Wagner, and M. Jansen, Synthesis ofadvanced ceramics In the systems SI·B·N and SI·B·N·C employing novel precursor compounds, Mater. Res. Soc. Proc., Vol. 271, The Materials Research Society, Pittsburgh, PA, 821·826 (1992).

CHAPTER 12 APPLICATIONS OF CARBON AND CERAMIC FIBERS R. Naslain This chapter is an overview of the applications for carbon and ceramic fibers. These applications mainly involve composite materials where the fibers are embedded in a polymer or a metallic or ceramic matrix, depending on the service temperature or the specific properties which are required. 12.1 Fiber applications Carbon and ceramic fibers are usually embedded in a matrix. There are applications in the fields ofthermallnllulatlon, catalyst support, filters, porous electrodes, and battery separators where they are utilized as mats or felts, or with asmall amount ofabinder. Thermal insulation isby far the most important present application ofoxide fibers. Transition alumina fibers, e.g., eta-alumina fibers, are produced atIntermediate firing temperatures and are used as supports for catalysts and Insulation tiles such as those used for the space shuttle orbiter [1-2]. Carbon fiber felts are used as internal thermal insulation for vacuum furnaces at extremely high temperatures. Activated carbon fibers, which are obtained by partial oxidation ofselected carbon fibers, have extremely small pores and very high specific surface areas, ranging from 500 to 3000 m2/g. They are of great Interest In ultrafiltration as membranes for the treatment ofused waters and liquids [3-5]. Rechargeable lithium cells are being developed to overcome difficulties observed in charging the lithium electrode orlithium-Ion battery. These problems are caused by reaction between Li and electrolytic solutions, and by cell shorting due to Li dendrite growth. One way to solve these problems is to use a negative graphite-lithium electrode, and to intercalate Li into and de-intercalate LI from the carbon host material: with 0 s x S J

(1)

Carbon fibers derived from PAN or mesophase pitch appear to be promising electrode materials [6-9]. Lithium intercalation Is highly reversible and both the current densities and the cycle lives are reported to be compatible with power batteries and electric vehicle applications [8]. Zirconia fiber cloths, such as those made by the relic process (see Chapter 8), are used as separators in aerospace nickel-hydrogen and nickel-cadmium batteries [9]. These fibers display a high resistance to many corrosive media, Including hot potassium hydroxide. Zirconia fiber felts are also present in aerospace solid oxide fuel cells.

316

Chapter 12

12.2 Composite applications Polymer matrix composites (PMCs), also referred to as fiber reinforced plastics (FRPs), consist of strong and stiff fibers reinforcing, unidirectionally (10) or multidirectionally (nO), thermosetting orthermoplastic polymers. They have high specific strength (crR/p) or/and high specific stiffness (E /p), as well as high fatigue resistance; crR, E and p are tensile strength, Young's modulus and density. Thermoset polymers, including polyesters, vinylesters and epoxies, are resins which irreversibly crosslink during curing. Thermoplastic polymers are fusible, high molecular weight, long chain molecules, which melt upon heating and can give, in a reversible manner, either an amorphous (polycarbonates) or semicrystalline (nylon) solid when cooled toroom temperature. 12.2.1

Polymer matrixcomposites

Three reinforcements are mainly used in PMCs: glass fibers, aramid fibers and carbon fibers [11 -13J. Carbon fibers have the advantage of being simultaneously strong and stiff (3 ~ crR s 7 GPa; 250 ~ E s 950 GPa and 1.8 s p s 2.1 g/cm3) (as discussed in Chapter 9). Glass and aramid fibers are strong but lack stiffness (E = 75 and 130 GPa, respectively). Strong carbon fibers (E = 250 GPa; crR= 3 GPa and p = 1.8 g/cm 3) yield 1-0 epoxy composites with the stiffness ofsteel (200 GPa), ahigher strength and alower density. PMCs have a much higher fatigue resistance than metals. However, the cost ofcarbon fibers ishigh, and carbon fibers are therefore often used as a hybrid reinforcement with glass fibers to increase composite stiffness and take advantage of hybrid economics. PMCs are practically used from cryogenic temperatures to a maximum temperature of about 300°C. Silicon carbide (or nitride) fibers as well as oxide fibers with low electrical conductivities are used in PMCs for specific applications where good dielectric properties are required, e.g., in electronic devices oraerospace structures with alow radar signature. The fabrication of PMCs iswell-documented [14J. A prepreg route is often preferred in order to facilitate the handling of the plies and to avoid the evolution of large amounts of organic solvents. A prepreg is an intermediate product consisting of a layer of impregnated fibers in which the polymer matrix is only partially crosslinked and hence is still suitably soft, the solvent being partially removed. The full crosslinking of the matrix is achieved, after plies stacking, in an autoclave. Another frequently used technique isfilament (or tape) winding, in which an impregnated fiber tow (or tape) is wound on a mandrel, and the mandrel is removed after autoclave curing. This technique is used to fabricate axisymmetrical parts and complex shapes. In the pultrusion technique, impregnated fibers are pulled continuously through a heated die, yielding a 1-0 composite with a constant cross section and a high fiber volume fraction. More specific techniques, such as hot pressing, structural reaction injection molding (S-RIM), resin transfer molding (RTM) and others, have been designed mainly toreduce processing time [141. Space and aerospace were the first important fields for applications of composite materials, particularly for carbon fiber reinforced polymer matrix composites [12-13J [15J. Carbon/epoxies are main constituents of the rudder, elevators, stabilizers, outboard and inboard ailerons, flaps, landing gear doors, engine cowlings and even wing skins of advanced civilian aircraft. High performance PMCs reduce weight and improve performance and

317

Chapter 12

maneuverability of fighter aircraft. Composites make up 1% and 2% by weight of the F-15 and the F-16, while carbon/epoxies will make up about 50% of the outer surface of the F/A18, where the composites represent 10% ofthe total weight ofthe fighter plane. In the next fighter plane. carbon/epoxy components will be about 26% of the total weight, resulting in a 20-30% weight reduction for the aircraft. The use of PMCs in the structures of the stealth bomber contributes to lower its radar signature. More and more PMCs are also used in helicopters in the airframe/fuselage, and even in the most critical parts, such as the rotor blades themselves. Composite rotor blades exhibit much better fatigue resistance and much higher service life than their metal counterparts [15-17].

_ _

=

Figure 1.

Carboncomposite Aluminum Aramidcomposite SPFDBtitanium

Advanced composite materials inaircraft structures: Rafale fighter aircraft; courtesy Dassaut Aviation.

Carbon/epoxies are also extensively used for space applications where stiffness, dimensional stability and weight are key considerations. The shuttle orbiter 18 meter cargo bay doors are among the largest carbon/ epoxy structures ever built. The 15 meter long mechanical arm of the space shuttle orbiter can handle space payloads weighing 20 tons on earth. Most satellites and spacecraft have large antennas made of carbon/epoxy composites to achieve dimensional stability. Carbon fibers have a negative coefficient of thermal expansion (CTE) and epoxies yield composites which have no thermal expansion or contraction within the temperature range corresponding to the space environment. As a result, both the shape ofthe antenna surface and its precise alignment are properly maintained. Further, the space stations of the future will be built from structural beams fabricated from carbon fiber/polymer matrix composites, to obtain the required stiffness, and assembled in

318

Chapter 12

space. Finally, PMCs are used in strategic missiles (rocket motor cases), cruise missiles (tail fins and ailerons), and insatellite launchers [15] [18] [20]. The weight of automobiles could be significantly reduced and their fuel economy improved if most of the steel components could be replaced with much lighter PMCs. However, cost considerations so far have limited the use of high performance PMCs in automobiles. Metal replacement inautomobiles continues tobe dominated by glass fiber reinforced PMCs. Carbon fiber/polymer matrix composites are extensively used in sporting goods and recreational applications. This is the second most important field of applications for carbon fibers. It has been largely responsible for the rapid growth of carbon fiber production during the last two decades and hence for the decrease in fiber prices [12-13]. The properties justifying the use of PMCs in this field are the high specific strength and stiffness of the composites, coupled with rapid vibrational damping and fatigue resistance. Golf club shafts, tennis rackets, fishing rods, cross country skis and poles, alpine skis, and bows and arrows are well known examples of sporting goods fabricated with carbon fibers [21]. PMCs are also finding more and more use in marine applications because of their strength and stiffness as well as their corrosion resistance. Typical examples ofhighly loaded components are the masts, booms, and rigging of racing sail boats, the cross beams of oceanic catamarans and trimarans, as well as in recently designed hydrofoil sail boats (Figure 2).

Figure 2.Aquitaine Innovations, a60-foot oceanicracing sloop designed bythe Fino-Conq architecfs group, islargely buiK with carbon fiber reinforced composites, including hull, deck, mast, centerboard, rudder rods, and blades. The hull contains by itsen 2.5metric tons ofcarbon fibers and was fabricated according totechnology previously used for aerospace structures by Composites Aquitaine an Aerospatial

Sailplanes and modern racing bicycles [21-22] represent other applications for PMCs. Here, the high specific stiffness ofcarbon fiber/epoxy composites isa key factor in structural design and performance improvement. Carbon fiber PMCs have found a few niches in industry on

Chapter 12

319

the basis of their specific stiffness, low moment of inertia, dimension stability or fatigue and corrosion resistance. The high cost of materials is compensated by the improvement of component performance. The use of carbon fiber PMCs has enabled robotics engineers to design manipulators which display low inertia and small end effect vibrations, whereas high acceleration and precision were previously considered to be mutually exclusive [23-24]. Pressure vessels offer key opportunities for properly designed MMCs [12). They are filament wound over a permanent stainless steel or ductile aluminum alloy liner that acts as a mandrel and also prevents potential gas or fluid leakage under pressure. Such light tanks are used in emergency situations by firemen ordoctors (Figure 3). e Group subsidiary; courtesy Henri TibaultlDPPI.

Figure 3. Pressure vessels with a capacityof 3 to 9 liters and an internal pressure of 30 MPa, fabricated with carbon fiber/epoxy composites, are used byfiremen in extreme situations. The pressure vessels provide firemen an autonomy upto one hour and correspond to 50% weight saving; courtesy Composites Aquitaine. an Aerospatiale Group subsidiary.

High performance carbon fiber PMCs are also of interest in energy storage or energy conversion involVing components oflarge size (windmill generators) orspinning athigh speed (flywheels). Stored kinetic energy is proportional to the moment of inertia and tothe square of the spinning speed. For this application, lightweight carbon fiber reinforced composites rely on high rotational speed in preference to large mass [12). Carbon fiber PMCs are also used to fabricate large blades for wind turbine generators. Carbon fiber PMCs, consisting of light elements (H, C, 0, N), display high X-ray transparency. This property combined with high specific stiffness has been exploited in X-ray equipment for non-destructive testing and medical analysis. Rigid and lightweight tables for computerized

320

Chapter 12

X-ray tomography are fabricated by applying skins ofcarbon fiber polymer laminates toacore made offoam orhoneycomb (27). The use of SiC or alumina based ceramic fibers in organic polymer matrix composites remains limited. Their production cost ishigher and their performance isoften lower than that for carbon fibers. The exception has to do with their dielectric properties. The dielectric constant of epoxy composites reinforced with Nicalon fiber can be widely varied by merely changing the nature of the fiber itself. When high resistivity NL-400 grade fiber is used (p = 106 - W n cm), the dielectric constant of the composite is high (20-30 at 10 GHz), and with low resistivity N500 grade fiber (p = 0.5 - 5.0 n cm), the dielectric constant is low or zero. Composites with low orhigh microwave absorption are used in advanced military aircraft (28). 12.2.2 Metal matrix composites With a few exceptions, metal matrix composites (MMCs) are still in a development stage although it is well known that the presence of inorganic fibers in a metal matrix has a major effect on the properties of the metal. The reinforcement and manufacturing costs are inevitably too high for the potential markets, such as that ofautomotive engines [29-33). In an aluminum alloy, for example, the presence of inorganic fibers: (1) increases its stiffness, failure strength and fatigue resistance at room and high temperatures; (2) improves its wear resistance if the reinforcement is a hard material, such as alumina or silicon carbide; and (3) increases its dimensional stability in thermal cycling, the CTE ofthe fibers usually being lower than that of the matrix. In UHM carbon fiber/AI (or Mg) composites, the CTE of the composites can even be nil for a specific fiber volume fraction, the contraction of the fibers and the dilatation ofthe matrix compensating one another when the temperature israised. Most MMCs are systems out of thermodynamic equilibrium. Diffusion phenomena and chemical reactions occur at the fiber/matrix interfaces which degrade the mechanical properties when such systems are heated for short periods of time at high process temperatures or maintained for prolonged periods of time at intermediate service temperatures. This is the case with C/ftJ, SiC/AI and SiCITi composites [34-38]. Hence, the main advantage of MMCs, their potential for high temperature applications, is altered unless specific alloys and/or fiber coatings are used. For example, Al-Si alloys are less reactive toward SiC fibers than pure AI, and SiC/C filaments are usually coated with multilayered carbon coatings when used to reinforce titanium alloys. Under such conditions, reactions occur between coating and matrix and the fibers remain undamaged as long as there issome carbon in between acting as amechanical fuse [34-35)[39). Compared to PMCs, the fabrication of MMC components is more difficult owing to the high melting points of metal matrices (660°C for aluminum; 1083°C for copper, and 1668°C for titanium) and their high chemical reactivity for most carbon and ceramic fibers. MMCs can be fabricated bydifferent techniques (12)[40-45). In the field of MMCs, continuous carbon and ceramic fibers compete with other reinforcements on a cost/performance basis. Continuous, and hence costly, fibers can be used in applications where high specific performance, e.g., in space applications, is required. In other potential applications, e.g., in automotive engines, metal matrix composites are not cost effective.

Chapter 12

321

One of the oldest applications of MMCs in space is the use of S/AI tubes, serving as a frame for the mid-fuselage section ofspace shuttle orbiters. The key requirements are high specific stiffness, high specific strength, and high fatigue resistance. The reinforcement is a continuous, 140 IJm diameter boron/tungsten bicomponent fiber (Chapter 4), which is unidirectionally embedded in an aluminum alloy. Each structure, consisting of more than two hundred tubes and ranging from 60 to 180 cm in length, results in a 44% weight saving over the original aluminum design, or 140 kg per vehicle. Carbon fiber/AI (or Mg) composites have been fabricated into beams for large antennas and benches for optical instruments in satellites. The main requirements are specific stiffness, thermal conductivity and dimensional stability. UHM carbon fibers derived from mesophase pitch (Chapter 9) are well suited for these applications, owing totheir extremely high stiffness, high thermal conductivity and negative CTE [45]. MMCs are considered promising for the design of specific parts for the cars of the future, which will be more fuel efficient, safer and have lower emissions. Although reducing the vehicle weight is also an important target, this industry is extremely competitive and price sensitive. MMCs have been envisioned for drive shafts, brake rotors, cylinder liners, connecting rods and pistons [46-47]. These parts experience severe service conditions and require the use of materials with good stiffness and strength at high temperatures, excellent fatigue and wear resistance, and compatibility with hot hydrocarbons, oil or combustion gas. A 250-300·C peak temperature is observed at the top of a piston. MMCs meet these requirements and alumina as well as silicon carbide fiber reinforced piston rods have performed well in racing cars. Although the feasibility offibrous MMCs istechnically demonstrated, the cost ofthe fibers and of the manufacturing processes, e.g., squeeze casting, is still too high. Thus, particulate reinforcements are presently preferred. Potential applications exist for metallic (MMCs) and intermetallic matrix composites (IMCs) in next generation aerospace propulsion systems for advanced commercial subsonic aircraft and high speed civil transport. These applications require stiff, strong and light materials that operate athigher temperatures with long lifetimes [48]. One candidate is a titanium alloy that is reinforced with large diameter SiC/C filaments (see Chapter 4) and isfabricated by superplasticforming/ diffusion bonding. This MMC is suited to the fabrication of bladed compressor rings, shafts, ducks, fan components or structural rods for jet engines. Their use for parts submitted to still higher temperatures is limited by fiber/matrix reaction and environmental considerations. Titanium aluminide ThAI (or y-TiAI) matrices could permit an increase in the service temperature ofthe composites. Substantial improvements, e.g., better fiber coatings, matrix alloying and oxidation protection, are needed before IMC technology can be seriously considered for jet engine applications. An important property of MMCs, compared to their PMC counterparts, is their electrical and thermal behavior; they have been used for some time in innovative applications such as in electrical and electronic devices.

322

Chapter 12

Both carbon fiberllead and alumina/lead composites have been utilized for the fabrication of electrodes in lead acid batteries with a significant weight reduction. For example, reinforcing lead unidirectionally with 25 vol.% of a-alumina fibers increases the matrix stiffness by a factor of20 and reduces the electrical conductivity only slightly. Due tothe stiffening effect of the fibers, pure lead can be used instead of a complex alloy, thus reducing the cost of the matrix and improving its corrosion resistance [49). Due to their tailorable combinations of properties, MMCs may play a key role in advanced electronic packaging, e.g., in power semiconductor diodes. In this application, the silicon wafer cannot be directly soldered to the copper electrode, due to a CTE mismatch (a (Si) = 4.1 x 10-6; a (Cu) = 17 X 10-6 °C·'). Under service conditions, interface delamination would occur as the result of thermal stresses arising from temperature changes. In conventional power diodes, the problem is partly solved byusing an intermediate molybdenum ortungsten layer with a CTE between those ofsilicon and copper (a (Mo) = 5.0x 10-6 and a (W) = 4.6 X 10-6 °C·1) . Another concept is the use a carbon fiber reinforced copper electrode. Carbon fibers have negative CTEs with a CTE matching that ofSi [50). Packaging materials support and protect integrated circuits in electronic devices. They also playa key role in heat removal. The amount of heat to be removed increases rapidly as the density of the integrated circuit packaging is raised. A key requirement is a high thermal conductivity coupled with a low CTE (semiconductor chips and ceramic carriers display low CTEs compared to metals). For some applications, such as aircraft avionics, weight is an important parameter. Among conventional packaging materials, such as copper, aluminum, molybdenum, tungsten, Invar, Kovar, W/Cu, or Mo/Cu blends and laminates, none exhibits simultaneously a low CTE and a high specific thermal conductivity. Two different MMC families have been studied for integrated circuit packaging, namely SiC/AI and UHM carbon fiber/Cu composites. P120 UHM carbon fibers are unsurpassed in terms of thermal conductivity. They have a 50% higher thermal conductivity than copper and a negative CTE), but they are expensive, whereas poorer performers are inexpensive [51 -52). 12.2.3 Carbon and ceramic matrix composites Ceramic matrix composites (CMCs), in which carbon or ceramic fibers are embedded in a ceramic matrix, have been designed to overcome the intrinsic brittleness of monolithic ceramics with a view toward structural uses at extremely high service temperatures. The most commonly used are carbon (C/C) and SiC matrix composites (C/SiC and SiC/SiC). Ceramic matrix composites with a silica based glass or glass-ceramic matrices have also been studied [12) [53-56). Carbon fiber reinforced carbon (C/C) composites, which are fabricated by the polymer impregnation/pyrolysis process (PIP) or by chemical vapor infiltration (CVI), are commonly used in rocket engine nozzles, where the temperatures can exceed 3000°C, and divergents, such as the very large booster nozzles of the Ariane V satellite launcher (Figure 4). C/C composites are also utilized in the heat shields of space vehicles and in the nose cap and leading edge of space shuttle orbiters. These are the structural components which experience the highest temperature (1300 to 1500°C) during reentry into the earth's atmosphere. However, C/C composites must be protected against oxidation since many are reusable.

Chapter 12

323

In the US space shuttle orbiters, the C/C outer surface is first converted to SiC; then the material is impregnated with tetraethylorthosilicate which yields, after curing, a silica residue in the porosity; finally, a seal coating is applied to the surface [571. In the European project of space shuttle orbiter Hermes, several parts, including the very large wing box components, spar bars, panels and leading edge segments, were directly fabricated with C/SiC composites by isothermal/isobaric chemical vapor infiltration (I-CVI). Tests were successfully performed on a leading edge segment under conditions simulating reentry up to1550-1700·C [58]. Nickel based superalloys which have been successful in the development of gas turbine engines over the last 50 years have reached their limit. To operate the engines at higher temperatures, new materials are much needed. CMCs are thought to have enough potential to satisfy such conditions (SiC has a melting point of 2500·C and a density of 3.2 g/cml However, they are still very new and their behavior in long term exposures (the lifetime of an engine is of the order of 10,000 hours) to corrosive atmospheres under load is still partly unknown [59-60]. Feasibility and test rig evaluation have been conducted on a flame holder and an exhaust cone (operating temperature: 800-1100·C and 800-950·C, respectively) fabricated from Nicalon/SiC composites by I-CVI. Weight saving was 50% and 30%, respectively. Nicalon/SiC composite inner flaps (operating temperature: 850·C, weight saving: 60%) were mounted on a SNECMA M53-2 engine and flight tested on a Dassault Mirage 2000 fighter (Figure 5). C/SiC outer flaps (operating temperature less than 650·C, weight saving: 50%) have been flight tested in SNECMA M88-2 engines of the Dassault Rafale fighter and are now produced in series [59].

Figure 4. Nozzle of the booster rockets of the European satellite launcher Ariane V having anouter diameter of 2.8 m, an inner diameter of 90em, and anoverall weight of 6 metric tons. The nozzle is fabricated with C/C and carbon fiber/phenolic matrix composites; courtesy SEP division deSNECMA.

324

Chapter 12

Figure 5. C/SiC outer flaps, in-flight tested on a SNECMA-M53.2jetengine and a Dassault Mirage 2000 flQhter aircraft. The flaps were fabricated byCVI and represent a 50% weight savings; courtesy SEP division deSNECMA.

Recently, the lifetime of SiCISiC composites has been increased through the use of SiC based fibers, with much higher thermal stability such as the Hi-Nicalon fibers, and self-healing multilayered interphases and matrices. Hence, it is anticipated that most of the hot static components of gas turbine engines, including the combustor liner (operating temperature: 1400°Cand above) would be fabricated with SiC matrix composites [61]. Over one-half of the CIC composites in the world are used in aircraft brake systems (Figure 6). The use of CIC braking in racecars is cost effective, but in other land vehicles such as high speed trains it isnot [54].

Figure 6. C/C discs for the braking systems of civilian Airbus aircraft (left) and military Mirage 2000 fighter aircraft (right). The discs were fabricated according tothe I·CVI process; courtesy SEP division deSNECMA.

Chapter 12

325

Aircraft braking systems have a multiple disc rotorlstator configuration, the rotor discs being sandwiched between the stator discs, and the rotor driven by the wheel. The braking action is performed by pressing the discs together. The brake discs have two main functions: they provide the frictional torque which stops the vehicle and serve as heat sinks to absorb the large amount of heat generated during the braking action. Replacing steel with C/C composite in an aircraft braking system results in a 40% weight saving and a doubling of the service life [54] [62-63].

C/C braking systems are already extensively used in racing cars (such as Formula 1) with, however, a different configuration. Here, the braking action is performed by pressing pads against a ventilated disc. C/C composites fabricated by either the PIP process (impregnation of the fibers with e.g. a phenolic resin and pyrolysis) or the CVI process are actually used. Finally, as already mentioned, tests have been conducted in order to explore the potential of C/C braking systems in high speed trains [54]. C/C composites may be used in place of aluminum alloys in spark ignition (or diesel) engine applications, e.g. as pistons and valves, on the basis of their low CTE, high thermal conductivity, good tribological properties, and low density. The benefits ofusing C/C parts are reduced reciprocating weight (and hence reduced vibration) and higher operating temperatures. Further, the combined use of C/C pistons and C/C cylinder liners may eventually eliminate the need for piston rings, owing to the tribological properties of C/C composites. However, such applications suppose low raw material and manufacturing costs [64]. CMCs are also being evaluated for applications in a 100 kW ceramic gas turbine for automobiles, with a turbine inlet temperature of1350·C [65]. Heat management systems such as combustion systems, burners, heat exchangers, waste incinerators and steam generators are potential applications for CMCs. These systems produce or recover heat under severe temperatures and exhibit a higher resistance to corrosion . For example, metal heat exchangers that extract heat from combustion gases cannot be operated above 750·C and their durability is limited by corrosion. SiC/SiC composites could be operated at1000·C, increasing the efficiency of, e.g., a coal fired power generation plant [56].

C/C composites find increasing use in industry both at medium and high temperatures in place of asbestos or monolithic graphite. Although much more expensive than the materials they replace, C/C composites display higher mechanical properties, shock resistance and durability. C/C composites are also used as heating elements for furnaces operating at high temperatures with non-oxidizing atmospheres, and as charging stages or trays for carbonizing, carbiding, orheat treatment furnaces, inplace ofmonolithic graphite orrefractory metals (Figure 7). Their use increases part lifetime and facilitates the handling of the parts [53-54]. C/C composites are used for hot-pressing dies, molds for superplastic forming of Ti alloys at 900·C, and for components of oil-less compressors, pumps and devices used in materials processing.

326

Chapter 12

Figure 7.

C/C tray for heat treabnent furnaces; courtesy SEP division deSNECMA.

Figure 8.

C/C hipprosthetic device; courtesy SEP division deSNECMA.

Finally, carbon is known to have superior biocompatibility with bones, blood and soft tissues. As a result, carbon fiber tows or GIG composites are used as tendon replacements or as tooth implants. GIG composites with some residual porosity are used to manufacture bone prosthetic devices. Their structure can be tailored to match the mechanical properties of bones and their biocompatibility allows the in-growth ofnew tissue [13) [53-54).

Chapter 12

327

REFERENCES [1) [2] [3] [4] [5) [6]

[7] [8] [9) [10] [11] [12] [13] [14] [15] [16] (17) [18]

[19) [20] [21) [22) [23) [24)

J. D. Birchall, J. A. A. Bradbury and J. Dinwoodie, Alumina fibres: preparation, properties and applications in Handbook of Composites Vol. 1 Strong Fibers, W. Watt and B. V. Perov, eds., Elsevier, Amsterdam, 115-154 (1985). J. Dinwoodie, Saffil alumina fiber: manufacture, properties and application in high temperature fumaces, Canadian Ceramics Quarterly, Feb. Issue, 23-29 (1996). A. M. Rao, A. W. P. Fung, M. S. Dresselhaus, and M. Endo, Structural characterization of heat-treated activated carbon fibers, J. Material Res., 7 [7], 1788-94 (1992). K. Oshida, K. Kogiso, K. Matsubayashi, K. Takeuchi, S. Kobayashi, M. Endo, M. S. Dresselhaus, and G. Dresselhaus, Analysis of pore structure of activated carbon fibers using high resolution electron spectroscopy and image processing, J. Material Res., 10[10J, 2507-17 (1995). P. H. Wang, K. L.Hong, and Q. R. Zhu, Surface analysis of polyacrylonitrile-based activated carbon fibers by X-ray photo-electron spectroscopy, J.Applied Polym. Sci., 62,1987-91 (1996). R. Kanno, Y. Kawamoto, Y. Takeda, S. Ohashi, N. Imanishi, and O. Yamamoto, Carbon fiber as a negative electrode inlithium secondary cells, J. Electrochem. Soc., 139 [12), 3397-3404 (1992). R. Yazami, K. Zaghib, and M. Deschamps, Carbon fibres and natural graph~e asnegative electrodes forlithium ion-type batteries, J. Power Sources, 52, 52-59 (1994). M. N. Verbrugge and B. J. Koch, Lithium intercalation of carbon fiber microelectrodes, J. Electrochem. Soc., 143 [1], 24-31 (1996). N. Imanishi, H. Kashiwagi, T. Ichikawa, Y. Takeda, O. Yamamoto, and M. Inagaki, Charge-discharge characteristics of mesophase-pitch-based carbon fibers for I~hium cells, J. Electrochem. Soc., 140 [2], 315-320 (1993). D. Hamling, Using ceramic-fiber materials in corrosive environments, Amer. Ceram. Soc. Bull., 76 [9], 79-82 (1997). J.Mack, Carbon fibre: above and beyond, Materials Edge, 13, 19-30, (1989) J. Delmonte, Technology of Carbon and Graphite Fiber Composites, Van Nostrand Reinhold Co, New York (1981). J. B. Donnet, and R. C. Bansal, Carbon Fibers, 2nd ed~ ., Marcel Dekker, NY and Basel, (1990) F. R. Jones, Handbook ofpolymer-fibre composites, Longman Scien. &Tech., Harlow, UK (1994). K. Upadhya, Recent developments in composite materials: a critical review, in High Performance Metal and Ceramic Matrix Composites, K. Upadhya, ed., The Mineral, Metals and Materials Society, Warrendale, PA, 237-260 (1994). M. Iden, and D. Pham, Boeing 777 empennage manufacturing, in Proc. 10th International Conference on Composite Materials, A. Poursartip and K. Street, eds., Woodhead Publishing Ltd., Abington-Cambridge, UK, Vol. III, 647-652 (1995). T. E.Palm. Design. fabrication and test ofcomposite sandwich panel component forairframe primary structural applications, inProc. 10th Intemational Conference on Composite Materials, A. Poursartip and K. Street, eds., Woodhead Publishing Ltd., Abington-Cambridge, UK, Vol. III, 677-684 (1995). S. Beland, B. Shaver, C. Chapman, and S. Desjardins, Non destructive inspection and mechanical testing of composite arm booms forthe space station remote manlpulator system, inProc. 10th International Conference on Composite Materials, A. Poursartipand K. Street, eds., Woodhead Publ. Ltd.,Abington-Cambridge, UK, Vol. 111,661-668 (1995). C. I. Grastataro, T. A. Butler, B. G. Smith, and T. C. Thompson, Development ofa composite satellite structure forFORTE, in Proc. 10th International Conference on Composite Materials, A. Poursartip and K. Street, eds., Woodhead Publishing Ltd., Abington-Cambridge, UK, Vol. III, 653-660 (1995). D. Marzi, Usage of composite materials for commercial and scientific satellites: experience at Alenia Aerospazio, inProc. 8th European Conf. on Compos~e Materials, I. Crivelli-Visconti, ed., Woodhead Pub. Ltd., Abington-Cambridge, UK, Vol. 1,261-266, (1998). Composites, Jec, Paris, Vol. 28, 14-37 (1998). P. L.Lizotte, L.B. Lessard, and J. A. Nemes, Stress and failure analysis ofcomposite bicycle frames, in Proc. 10th Intemational Conference on Composite Materials, A. Poursartip and K. Street, eds.,Woodhead Publ. Ltd., Abington-Cambridge, UK, Vol. 111,701 -708 (1995). L. B. Lessard, V. Hayward, and M. M. Roy, Design and fabrication of a graphite/epoxy link fora light weight robotic manipulator, in Proc. 8th European Conference on Composite Materials, I. Crivelli-Visconti, ed., Woodhead Publishing Ltd.,Abington-Cambridge, UK, Vol. 1, 535-542 (1998). H. Hald, F. Kocian, P. Shanz, and R. Steinheber, Composites forlight-weight robot structures, in Developments in fhe Science and Technology ofComposite Materials: Proc. ECCM-4, J. Fiiller, G. GrOninger, K. Schulte, A.R. Bunsell and A. Massiah, eds., Elsevier Applied Science, London and New York, 811-818 (1990).

328

Chapter 12

[25] J. Luyckx, R Lacroix, and J. P. Fuzier, Bridge strengthening by carbon fibers, in Proc. 8th European Conference on Composite Materials, I. Crivelli-Visconti, ed., Woodhead Publishing Ltd., Abington-Cambridge, UK, Vol. 1, 15-20 (1998). [26] A. Nanni, and W. Gold, Srengthening of PC slabs with carbon FRP composites, in Proc. 8th European Conference on Composite Materials, I. Crivelli-Visconti, ed., Woodhead Publishing Ltd., Abington-Cambridge, UK, Vol. 1,83-88 (1998). [27] T. Simonelli, Carbon fiber forX-ray applications, inProc. 8th European Conference on Composite Materials, I. Crivelli-Visconti, ed.,Woodhead PUblishing Ltd., Abington-Cambridge, UK, Vol. 1,651-654 (1998). [28) T. Ishikawa, Recent developments ofthe SiC fiber Nicalon and its composites, including properties ofthe SiC fiber Hi-Nicalon forultra-high temperature, Composites Science and Technology, 51 , 135-144 (1994). [29] A. K. Dhingra, What are fibers doing inmetal castings? CHEMTECH,11, 600-608, (1981). [30] T. W. Chou, A. Kelly, and A. Okura, Fibre-reinforced metal-matrix composites, Composites, 16 [3), 187-206 (1985). [31] J. A. Cornie, Advanced continuously reinforced inorganic composites forstructural applications: critical issues for future growth, in Advanced Structural Inorganic Composites, P. Vincenzini, ed., Elsevier Science Publishers, Amsterdam, 801-821 (1991). [32] V. K. Lindroos, and M. J. Talvitie, Recent advances in metal matrix composites, J. Mater. Processing Technology, 53, 273-284, (1995). [33] A. Vassel, M. H. Vidal-Setif, M. Rabinovitch, and R Mevrel, La Recherche Aerospatiale, 5-6, 421-440 (1996). [34) G. Das, Interlacial reactions in SiC-reinforced titanium alloy and titanium aluminide composites, in Mefal and Ceramic Matrix Composites, RB. Bhagat, A.H. Clauer, P. Kumar and A.M. Ritter, eds., The Minerals, Metals and Materials Soc.,Warrendale, PA, 413-442 (1990). [35) T. Sritharam, K. Xia, J. Heathcock, and J. Mihelich, Matrix/reinforcement development for aluminum-based composites, in Mefal and Ceramic Matrix Composites, RB. Bhagat, A.H. Clauer, P. Kumar and A.M. Ritter, eds., The Minerals, Metals and Materials Society, Warrendale, PA, 13-22 (1990). [36] J. K. Tien, and M. W. Kopp, Interlacial reactions in high temperature metallic and interrnetallic composites: a status review, in Metal and Ceramic Matrix, RB. Bhagat, A.H. Clauer, P. Kumar and A.M. Ritter, eds., The Minerals, Metalsand Materials Society, Warrendale, PA, 443-456 (1990). [37] H. J. Dudek, A. Kleine, R Borath, and G. Neite, Interfaces inalumina fibre reinforced aluminium piston alloys, Mater. Sci. Engineering, A 167, 129-137 (1993). [38) E. A. Fees!, Interfacial phenomena inmetal-matrix composites, Composites, 25 [2), 75-86 (1994). [39) C. G. Rhodes, and R. A. Spurling, Fiber/matrix interlace reactions in SiC reintorced titanium alloys, in Developments in ceramic and Metal-Matrix Composites, K. Upadhya, ed., The Minerals, Metals and Materials Society, Warendale, PA, 99-113 (1991). [40) J. A. Cornie, Y. M. Chiang, D. R Uhlmann, A. Mortensen, and J. M. Collins, Processing of metal and ceramic matrix composrtes, Ceram. Bull., 65[2), 293-304 (1986). [41) D. Huda, M. A. EI Baradie, and M. S. J. Hashmi, Metal matrix composites: Manufacturing aspects. Part I, J. Mater. Processing Technology, 37, 513-528 (1993). [42) B. Ralph, H. C. Yuen, and W. B. Lee, The processing of metal matrix composites: an overview, J. Mater. Processing Technology, 63, 339-353 (1997). [43] R Asthana, Cast metal-matrix composites. I: Fabrication techniques, J. Mater. Processing Technology, 5 [4], 251-278 (1997). [44] R Asthana, Cast metal-matrix composites. II: Process fundamentals, J. Mater. Processing Technology, 5 (5), 339-361 (1997). [45) W. R Mohn, and Vukobratovich, Recent applications of metal matrix composites in precision instruments and optical systems, J. Mater. Eng., 10, 225-235 (1988). [46] J. E. Allison, and G. S. Cole, Metal-matrix composites inthe automotive industry: opportunities and challenges, J.ofMetals, January Issue, 19-24 (1993). [47] C. G. E. Mangin, J. A.lsaacs, and J. P.Clark, MMCs forautomotive engine applications, J. ofMetals, February Issue, 49-51 (1996). [48) J. Doychak, Metal- and interrnetallic- matrix composites for aerospace propulsion and power systems, J. of Metals, June Issue, 46-51 (1992). [49) H. S. Hartmann, and R A. Sutula, "Fiber FP" reinforced pure lead composites for battery electrodes, J. Electrochem. Soc., 129 [8J, 1749-52 (1982). [50) K. Kuniya, H. Arakawa, T. Kanai, and T. Yasuda, Development ofcopper-carbon fiber composite forelectrodes of power semiconductor devices, in IEEE Transactions on Components, Hybrids and Manufacturing Technology, 6 (4), 467-472 (1983). [51] C. Thaw, R Mine!, J. Zemany, and C. Zweben, Metal matrix composite microwave packaging components, SAMPE Journal, NovlDec.,40-43 (1987). [52) C. Zweben, Metal matrix composites forelectronic packaging, J. ofMet., July, 15-23 (1992).

Chapter 12

329

[53] J.D. Buckley, and D. D. Edie, Garbon-Garbon Materials and CCmposites, Noyes Publications, Park Ridge, NJ (1993). [54] G. Savage, Carbon-Garbon CCmposites, Chapman and Hall, London, UK, (1993). [55] K. K. Chawla, ceramic Matrix CCmposites, Chapman and Hall, London, UK, (1993). [56] S. Richlen, Applications of fiber-reinforced ceramic matrix composites, in Handbook on CCnlinuous FiberReinforced ceramic Matrix Composites, R. L.Lehman, S. K. EI-Rahaiby and J. B. Wachtman, Jr., eds., JapanChapter SAMPE, Yokohama, Japan, 496-536 (1997). [57] P. R. Becker, Leading edge structural material system of the space shuttle, Amer. Ceram. Soc. Bull., 60[11], 1210-14 (1981). [58] P. Lamicq, Ceramic matrix composites: a new concept for new challenges, in High Temperature Materials for Power Engineering 1990, E. Bachelet etaI., eds., Kluwer Academic Publishers, Dordrecht. NL, 1559-74 (1990). [59) P. Spriet. and G. Habarou, Applications of CMCs toturbojet engines: overview ofthe SEP experience, in Key Engineering Materials : ceramic and Metal Matrix Composites, M. Fuentes, J. M. Martinez-Esnaola and A. M. Daniel, eds., Trans Tech Publications, Uetikon-Zuerich, Switzenand, Vol. 127-131 , 1267-76 (1997). [60] C. P. Beesley, The Applications of CMCs in high integrity gas tUrbine engines, in Key Engineering Materials: Ceramic and Metal Matrix Composites, M. Fuentes, J. M. Martinez-Esnaola and A. M. Daniel, eds.,Trans Tech Publications, Uetikon-Zuerich, Switzenand, Vol. 127-131 , 165-174 (1997). [61] K. Nishio, K. I. Igashira, and S. Okazaki, The development ofceramic matrix composite (CMC) forcombustors of gas turbine engine, in Proc. 5th Japan Intem. SAMPE Symp.,Y. Miyano and M. Yamabe, eds., SAMPE Japan Chapter, Yokohama, Japan, 1159-64 (1997). [62] C. Blanco, J. Bermejo, H. Marsh, and R. Menendez, Chemical and physical properties of carbon asrelated to brake performance, Wear, 213, 1-12 (1997). [63] J. D. Chen, J. H. Chem Lin, and C. P. Ju, Effect of humidity on the tribological behavior of carbon-carbon composites, Wear, 213, 3847(1996). [64] W. Kowbel, V. Chellappa, J. C. Withers, and P. O. Ransome, Applications ofnet-Shape molded carbon-carbon composites inICengines, J. Adv. Materials, July Issue, 2-7 (1996). [65] T. Izumi, and H. Kaya, Ceramic matrix composites applications in automotive gas turbines, J. Eng. Gas Turbines and Power: Trans. ASME, 119,790-798 (1997).

ACRONYMS 1D 3D ADFA BEl BMU BSU CAD CMC CR CSZ CTE CVD CVI DCCA DMAC DMF DRIFT DSC DTA ECR EDX EFG EMI EOPA EPMA

One-dimensional; unidirectional Three-dimensional Erbium doped fiber amplifier Backscattered electron image Basic microstructural unit Basic structural unit Computer aided design Ceramic matrix composite Corrosion resistant Cubic stabilized zirconia Coefficient of thermal expansion Chemical vapor deposition Chemical vapor infiltration Drying control chemical additive Dimethylacetamide Dimethylformamide Diffuse reflectance infrared spectroscopy Differential scanning calorimetry Differential thermal analysis High acid corrosion resistance Energy dispersive x-ray spectroscopy Edge-defined film fed growth Electromagnetic interference Ethyl 3-oxobutanodiisopropoxyaluminum Electron probe microanalysis ESCA Electron spectroscopy for chemical analysis FOG-M Military fiber optical glass HM High modulus HP High pressure HPZ Hydridopolysilazane HRTEM High resolution transmission electron microscopy HS High strength HT High temperature HTT Heat treatment temperature IC Integrated circuit 1M Intermediate modulus IMC Intermetallic matrix composite IMS Inviscid melt spinning ITA Itaconic acid LA Laurylamine LCVD Laser-assisted chemical vapor deposition LHFZ Laser heated float zone LHPG Laser heated pedestal growth L1GA X-ray lithography

332

LM Low modulus LMO Local molecular orientation LP Low pressure MA Methylacrylate MAS 2,4,S-tris(methylamino)borazine MCVD Modified chemical vapor deposition MMC Metal matrix composite MP Mesopitch MTS Methyltrichlorosilane MWNT Multi-walled carbon nanotube NMR Nuclear magnetic resonance OVD Outside vapor deposition PAN Polyacrylonitrile PSDPSO Polyborodiphenylsiloxane PSG Photonic band gap PSSZ Polyborosilazane PCS Polycarbosilane PCSZ Polycarbosilazane PCVD Plasma chemical vapor deposition PDMS Polydimethylsilane PEEK Polyetheretherketone PHPSZ Perhydropolysilazane PIP Polymer impregnation/pyrolysis PMS Polymethylsilane PSP Polysilapropylene PSSZ Polysilasilazane PSZ Partially stabilized zirconia PSZ Polysilazane RS Rapid solidification RT Room temperature SIC Sheath/core SSA Tri-sec-butoxyaluminum s-b-s Side-by-side SEI Secondary electron image SEM Scanning electron microscopy SLS SOlution-liquid-solid SM Standard modulus SNMS Sputtered neutral mass spectrometry STM Scanning tunneling microscopy TADS Trichlolosilylamino-dichloroborane TEM Transmission electron microscopy TEOS Tetraethoxysilane Si(OC2Hs)4 TGA Thermogravimetric analysis TSZM Traveling solvent zone melting TZP Tetragonal zirconiapolycrystals UHM Ultrahigh modulus VAD Vertical axial deposition VLS Vapor-liquid-solid

Appendix

333

Appendix

VS XPS

XRD YAG

Vapor-solid X-ray photoelectron spectroscopy X-ray diffraction Yttrium aluminum gam

GLOSSARY Acoustical wave velocity: The distance traversed bya periodic, orcyclic; acoustic signal per unit time. Ammonolysis: A process that involves heating a chloro compound with aqueous ammonia toform an amine. Amphibole fibers: Fibers of a class of variously colored hydrous silicates, consisting chiefly ofcalcium, magnesium, iron, aluminum and sodium. Annealing: Treatment of a material byheating to a predetermined temperature, holding for a certain time, and cooling to room temperature to improve physical properties, such as strength, orchemical resistance. Bandwidth: The number of discrete channels within a finite range of frequency, which defines the data transfer rate ofan electronic communications system. Basalt fibers: Fibers from igneous (volcenic] rock that is low in silica content, dark in color, and comparatively rich iniron and magnesium. Basic structural unit: A stack of2-4 graphene layers lying roughly parallel toeach other. Bicomponent fibers: Fibers consisting of two components. Commercially significant inorganic bicomponent fibers have either a concentric sheath/core ora side-by-side fiber structure. Bioactive glass: A glass material which bonds to, and facilitates growth ofliving tissue. Birefringence: An optical property in which a single ray of unpolarized light entering an anisotropic medium splits into two rays, each traveling in a different direction. One ray (called the extraordinary ray) is bent, or refracted, at an angle as it travels through the medium; the other ray (called the ordinary ray) passes through the medium unchanged. Calcining: Heating a material, such as an inorganic material, to a high temperature but without fusion in order to drive offvolatile matter or to effect changes, such as oxidation orpulverization. Carbonization: The pyrolyzation of organic material in an inert atmosphere at high temperatures, typically above 1000°C, all non-carbon elements being driven off in the process. Carbothermal reduction: Chemical reduction of a material by carbon at elevated temperature. Ceramic matrix composite: A composite consisting of a filler material, frequently fibers, embedded ina ceramic matrix. Chalcogenide: A binary compound of a chalcogen (Le., any of the elements oxygen, sulfur, selenium, and tellurium) with a more electropositive element orradical. Chemical vapor deposition: Deposition ofsolid material when gaseous reactants encounter ahot surface. Laser assisted CVD employs a hot laser-focus rather than a hot surface. Chemical vapor infiltration: A process whereby a reactive chemical species is generated in the vapor phase and allowed toreact with a solid substrate, thus modifying its chemistry. Coefficient of thermal expansion: The change in volume-per-unit volume produced by a one-degree rise intemperature. Compressive strength: The ability of a material to resist a force that tends to crush; the crushing load at specimen failure divided by the original cross sectional area of the specimen. Creep: The change in dimension of a plastic material under load over a period of time, not including the initial instantaneous elastic deformation.

336

Glossary

Crimp: The waviness of a fiber. Crimp can be either natural or can be mechanically or thermally induced. Debye-Scherrer pattern: A method of x-ray diffraction analysis widely used to identify polycrystalline materials. Dielectric constant: The ratio of the capacity of a condenser having a dielectric material between the plates to that of the same condenser when the dielectric is replaced by a vacuum; a measure ofthe electrical charge stored per unit volume atunit potential. Dielectric loss: A loss of energy eventually showing up as a rise in the temperature of a dielectric material placed inan alternating electric field. Diphasic gel: Agel consisting oftwo discrete, mutually incompatible phases. Discotic nematic liquid crystal: A liquid crystal formed from discotic (disk-shaped) molecules, which are more orless planar and which are stacked. Disinclinations: Rotation defects in the arrangement ofdiscotic (disk-shaped) molecules in a liquid crystal. Dispersion shifted fiber: An optical fiber engineered tocontrol the optical dispersion profile. Dry spinning: A process whereby a viscous solution is spun through spinneret orifices and the excess solvent isremoved in a drying column. E-beam curing: A process whereby green fibers are rendered infusible through irradiation with an electron beam. E-beam evaporation: Evaporation ofmaterial by electron beam irradiation. Eutectic mixture: The one mixture ofa set of substances able to dissolve in one another as liquids that, of all such mixtures, liquefies at the lowest temperature. On cooling, the components ofthis mixture separate simultaneously as an intimate mixture ofsolids. Evanescent field: That portion of the light being transmitted by an optical fiber which escapes into the sheath, outside the core. Fiber grating: A structure generated by periodically varying the refractive index of the core of an optical fiber. The period of the optical grating can be adjusted to reflect specific wavelengths. These are used in wavelength selective mirrors, resonant cavities and bandpass filters. Fragile melt: Molten material which isdifficult toprocess into fibers. Free radical: An especially reactive atom or group of atoms that has one or more unpaired electrons. Gauge length: The length over which deformation ofa sample ismeasured. Gel: A coherent mass consisting of a liquid in which particles too small to be seen in an ordinary optical microscope are either dispersed orarranged in a fine network throughout the mass. A gel may be elastic and jelly-like, as gelatin orfruit jelly, or solid and rigid, as silica gel. Glass transition temperature (T g): The temperature below which an amorphous material behaves like a solid rather than a viscous liquid. The glass transition occurs over a temperature range, whose location depends somewhat on the cooling rate and the method ofmeasurement. Graded index fiber: An optical fiber having a fixed refractive index in the sheath and a continuously variable refractive index in the core, Grain size: The size ofadiscrete crystallite orgrain within a solid material. Graphene layer: The structural unit in carbons resulting from the carbonization of an organic precursor; each layer consists ofa few fused aromatic rings. Graphitization: The solid state transformation of thermodynamically unstable non-graphitic carbon into graphite by thermal activation, usually attemperatures above 1800°C.

Glossary

337

Green fiber: A fiber precursor which must be heat-treated to obtain the final desired form. Group delay: The time delay between components ofa signal having different wavelengths. Incubation period: The time required for the onset ofnucleation in a crystallization process. Intercalation of carbon fibers: The crystal structure of graphite is of a kind that permits the formation of many compounds, called lamellar or intercalation compounds, by insertion ofmolecules orions between the graphitic layers. Internal flaws: Flaws within the volume ofamaterial, such as voids. Inviscid melt spinning: A process which allows low viscosity mollen material to be spun into fibers. The low viscosity jetischemically stabilized rather than rapidly solidified. Isotropic: Exhibiting properties, such as velocity oflight transmission orrefractive index, with the same values when measured along any direction in a material, i.e., along any of the principal axes ofacrystal. Laser heated floatzone process: In this process, a circumferential laser is placed around a preform rod to zone refine a segment of the material while simultaneously updrawing a single crystal fiber. Liquid crystal: An organic liquid whose physical properties resemble those of a crystal in the formation ofloosely ordered molecular arrays similar to a regular crystalline lattice and in the anisotropic refraction of light. Liquidus temperature: The lowest temperature atwhich a material will be completely liquid. Lithography: The process of producing patterns on semiconductor crystals for use as integrated circuits. Mat: A fibrous material for reinforced plastic use consisting of randomly oriented chopped filaments orswirled filaments with a binder. Melt spinning: Extrusion ofa molten polymer through a spinneret toform fibers. Mesophase pitch: See mesopitch. Mesophases: Liquid crystals, sometimes called mesophases, occupy the middle ground between crystalline solids and ordinary liquids with regard to symmetry, energy, and properties. Mesopitch: A carbonaceous solid, consisting primarily of a complex mixture of polycyclic aromatic compounds, which can form an anisotropic liquid crystal mesophase. Metal matrix composite: A composite consisting of a filler material, frequently fibers, embedded in a metal matrix. Microcoil: Acoiled structure of microscopic scale. Microfibers: Fibers generally 1-25 um diameter. Microhardness: The indentation hardness ofmicroscopic areas ofamaterial. Micropillar: A short inorganic fiber generally 1-25 urn diameter. Microsprings: Coiled microfibers. Microstructure: The structure ofa material on a microscopic scale. Mode coupling: Interaction of modes resulting in the transfer of light intensity between the guided (core) modes and the radiation (cladding) modes. Mode: Apath along which light may travel, e.g. in an optical fiber. Modulus: The ratio of stress to strain in a material over the range for which this value is constant. The type of modulus, which is measured, depends on the method of measurement, e.g., dynamic modulus, compressive modulus, elastic or tensile (Young's) modulus, shear modulus, torsion modulus, sonic modulus. Morphology: The size, shape and structure ofa material and its constituents. Mullitization: Crystallization ofmullite from stoichiometric alumina-silica mixtures.

338

Glossary

Nanocrystalline: Composed of crystals orcrystalline structural units which have nanometer dimensions. Nanofibers: Fibers generally 1-25 nm diameter. Nanopillar: A short inorganic fiber generally 1-25 nm diameter Nanotubes: Hollow nanofibers. Carbon nanotubes consist of one or more cylindrical structures orshells, each composed ofa single tubular graphitic layer. Nanowire: An electrically conductive nanofiber. Nucleation: The initial process that occurs in the formation of a crystal from a solution, a liquid, or a vapor, in which a small number of ions, atoms, or molecules become arranged in a pattern characteristic of a crystalline solid. This creates a site upon which additional particles are deposited as the crystal grows. Numerical aperture: Ameasure ofthe light gathering power ofan optical system. Optical amplifier: A photonic device which boosts a light signal in optical communications to remedy transmission loss. Optical dispersion: The change in refractive index for a transparent material as optical wavelength ischanged. Optical loss: Attenuation ofoptical light intensity during propagation through a fiber. Optoelectronic: Pertaining to electronic devices for emitting, modulating, transmitting, and sensing light. Photonic band gap: The energy gap between the energy minimum of the electron conduction band and the energy maximum of hole valence bands which occur at the same location in momentum space, allowing electrons and holes to recombine and radiate photons efficiently. Favorable band gaps occur with III-V compound semiconductors. Polycrystalline: Consisting ofcrystals variously oriented. Polycyclic aromatic hydrocarbons: Organic compounds consisting of fused aromatic ring moieties. Polymer matrix composite: A composite consisting of a filler material, frequently fibers, embedded in a polymer matrix. Polymorphism: The quality or state of being able to assume different forms such as the property ofcrystallizing in two ormore forms with distinct structure. Precursor fiber: A fiber from which another fiber will be formed . Preferred orientation: The anisotropic arrangement of crystals or polymer molecules; in fibers, thisis usually caused by process stresses during spinning ordrawing. Preform: A shaped form from which fibers are subsequently obtained. The use of a preform facilitates handling and control ofuniformity. Pyrolysis: Chemical change brought about by the action ofheat. Radiation modes: Modes which propagate in an optical fiber through the cladding (sheath). Rayleigh scattering: Scattering of light in a material because of microscopic fluctuations in refractive index, e.g., fluctuations in composition ordensity. Roving: A term used to designate a collection of bundles of continuous filaments either as untwisted strands or as twisted yarns. Glass rovings are predominantly used in filament winding. Sigmoidal curve: Acurve shaped like the letter S. Sintering: The welding together ofsmall particles by applying heat below the melting point. Sliver: An untwisted strand orrope oftextile fiber produced by a carding orcombing machine and ready for drawing, roving, orspinning.

Glossary

339

Softening point: The temperature above which a fiber will rapidly deform under its own weight. Sol-gel processing: A generic term covering process routes which differ from one another mainly by the nature of the starting chemicals, the most commonly used being sols or solutions of organometallic species. Gelation of a sol gives a viscous product which can be shaped as fibers. Specific modulus: Modulus divided by material density. Specific strength: Tensile strength divided bymaterial density. Spin orientation: Molecular orstructural orientation created inthe spinning process. Spinneret: A type of extrusion die, l.e., a metal plate with many tiny holes, through which a plastic melt isforced to make fine fibers and filaments which are hardened bycooling in air orwater orby chemical action. Staple fibers: Fibers of spinnable length manufactured directly or by cutting continuous filaments toshort lengths, usually one half totwo inches long. Step index fiber: An optical fiber having fixed refractive indices for the sheath and core components. Stoichiometric: Pertaining tothe quantitative relationship between constituents in a chemical substance. Strain rate: The rate ofdeformation as in a tensile test. Superconductor: A material which exhibits complete disappearance of electrical resistance especially atvery low temperatures. Supercool: To cool below the freezing point without solidification orcrystallization. Surface flaws: Flaws on the surface ofa material, such as surface scratches. Tensile strength: The maximum tensile load per unit area oforiginal cross section, within the gauge boundaries, sustained by the specimen during a tension test. Thermal conductivity: The ability ofa material to conduct heat; the physical constant for the quantity of heat that passes through a unit cube of a material in a unit time when the difference in temperature oftwo opposite faces isone degree. Thermophoresis: Transverse motion of particles in a gas stream caused by a thermal gradient perpendicular tothe direction ofthe gas stream. Thermoplastic polymer: A polymer capable of softening or fusing when heated and of hardening again when cooled. Thermosetting polymer: A polymer capable of becoming permanently rigid when heated or cured. Toughness: The energy required to break a material. This energy is equal to the energy under the stress-strain curve. High toughness leads tohigh damage resistance. Transmission loss: See optical loss. Turbostratic: A type of crystalline structure where the basal planes, such as graphene layers, have slipped sideways relative to each other, causing the spacing between planes to be greater than ideal. This structure is found in incompletely heat-treated carbon and boron nitride. Updrawing: A process whereby a rod is dipped into a pool of a viscous melt and then is raised vertically. The cooled melt adheres to the rod and a fiber is formed which cools and solidifies. Vapor-liquid-solid growth: A vapor toliquid tosolid phase transformation, the most common way to grow whiskers. Vapor-solid growth: A direct vapor to solid phase transformation, responsible for the catalyzed growth ofsome fibers.

340

Glossary

Volume resistivity: The ratio of the direct voltage applied to electrodes on opposite faces of a unit volume with unit separation to that portion of the current between them passing through a unit cross sectional area. Waveguide: Adevice (aduct, coaxial cable, orglass fiber) designed toconfine and direct the propagation ofelectromagnetic waves (Le., light). Weibull distribution function: A probability distribution function which can be used to describe the scatter in tensile strength data for a given lot of fibers having the same cross section area and test gauge length. Weibull modulus: A parameter which characterizes the width of a monomodal Weibull distribution. Wet spinning: A process whereby a viscous polymer solution isextruded through a spinneret into a spin bath. The spinneret is generally, but not always, placed in the spin bath, a coagulation bath in which solvent diffuses out of the extruded material and a nonsolvent diffuses into the extrudate. Whiskers: Filamentary single crystals X-ray amorphous: Non-crystalline, as determined by x-ray diffraction. Young's modulus: See modulus.

INDEX A Absorption, 173-174 Acid leaching, 128, 148, 158, 165--166 Acid-resistant glass fibers, 148-149 Acrylic polymers, 124 Advanced liquid phase processes, 26-29 Aerospace applications, 70-71 ,140,158,166, 315,316-318,322,323,324,325 A-glass, 149 Alkali-resistant (AR) glass fibers, 102, 146-148 Alkoxides, 206 Almax,211,219,222 Altex, 218 Alumina fibers, 51,53,64,70,88-89,207-225 properties, 90, 219-224 a-Alumina fibers, 211 ,220,224 Aluminate fibers, 4,70, 88-89 Aluminate glass fibers, 3, 98-99 Alumina-zirconium fibers, 215--216 Aluminum oxide, 53 Aluminum silicate fibers, 71, 85 Ammonia, 22, 5~0 Amoco P-series, 248 Amorphous fibers, 3,4,20, 62, 64,65, 87, 88, 92 from inviscid liquids, 103-113 Anisotropy, 88 Arc discharge, 13, 24-25 Architectural applications, 148 Asbestos, 3, 11, 39 Astroquartz (AO), 90 Automobiles, 318, 321, 324, 325

B Basal growth, 12 Basicstructural unit (BSU), 234-235 Balleries, 315 Bend stress relaxation, 291 Bicomponent fibers, 6, 41 , 85, 92, 97, 156-162 Bioactive materials, 153,326 Birefringence, 91 , 179-180 Bone biocompatibility, 153,326 Borazine fibers, 60, 69 Boron/aluminum materials, 71 Boron doping, 275--276 Boron/epoxy tapes, 70

Boron fibers, 3, 51 , 57, 61 , 64, 65-67 Boron-free fibers, 95,131-132,133-134,148, 214 Boron nitride fibers, 25, 37, 5~0 , 69 Boron oxide fibers, 5~0 , 69 Boron/tungsten (BIW) fibers, 21,56,57,61,63, 66-67,70-71 Borosilicate fibers, 130-131, 133, 165--166 Bragg gratings, 197, 198 Braking systems, 324-325 Buckyballs, 36, 37 Bushing process, 84, 85, 86, 88, 93,129,157 Butyl indium, 28

c CAD (computer-aided design), 73, 74 Calcination, 207 Calcium aluminate fibers,99,101-102,110-111 Carbon black, 23 Carbon fiber-reinforced carbon (C/C) composites, 320-326 Carbon fibers, 3, 11, 17-18, 20, 67-B8, 71 , 233-235 commercial applications, 37,41-43, 315 composaes, 41, 316-320, 322-326 infiltration, 59 processing of, 64, 235--245 structures and properties, 34-36, 245-261 Carbon ion bombardment, 25 Carbonization, 238-239, 245 Carbon nanotubes, 3, 11, 18, 20, 24, 25, 42-43, 60 endless, 60 modulus, 37, 38, 70, 72, 88-90 structures and properties, 36-39 Carbothermal reduction, 13,23-24,34 CC-glass, 149 CemFIL, 146, 147 Centrifuge processes, 11,28-29,129,160 Ceramic matrixcomposites (CMCs), 40,67-B8, 70,322-326 C-glass, 149 Chalcogenide glass fibers, 100-101 Chemical mixing processes, 23-24 Chemical resistance, 145--146 Chemical vapor deposition (CVD), 15,28,55

342 hot filament processes, 21-22, 55-56, 58, 65-68,69 metal catalyzed, 15-20,34,48,63,64,65 modified (MCVD), 185-190 plasma (PCVD), 22, 58, 190-191 See also Laser-assisted chemical vapor deposition Chemical vapor infiltration (CVI), 22-23, 59-60, 68-69,322 Coated carbon fibers, 261 Coefficient ofthermal expansion (CTE), 257-258,322 Commercial applications carbon fibers, 37,41-43,315 composites, 316-326 continuous fibers, 55-56, 65--Q8, 70-75, 129-132 glass fibers, 134-136, 147-148, 164-165 quartemary fibers, 102 silicon fibers, 39-40, 68 Compressive strength, 256 Computers, 149 Conical tips, 32, 33 Containerless laser melt process, 72, 85, 86, 107-108 Continuous fibers, 3, 4,6 dry spinning processes, 123-128 structures and properties, 60-70 vapor phase processes, 47--£0 See also Melt spinning processes Corundum fibers, 218-219 Creep,222-224,256-257,288-290 Crimped staple fibers, 161, 162 Cryogenic fibers, 3, 113 Cubic stabilized zirconia (CSZ), 225, 226 D

Damage resistance. See Mechanical toughness D-glass, 149-150 Diameter,4,15 Diamond, 21, 22,41, 58, 75 Dielectric constant, 149-152 Diodes, 39,42 Disclinations, 249 Discontinuous fibers, 3, 4 Dispersion, 174-179, 195 Double crucible technique, 97, 98, 157, 181-182 Downdrawing, 85, 86, 92 Draw towers, 191-192, 194 Dry spinning, 3,123-128,164-165

Index

E Edge-defined film-fed growth (EFG), 113-115, 118 E-glass, 84, 94-95,129 boron- and fluorine-free, 95,131-132, 133-134,148 borosilicate, 130-131, 133 corrosion-resistant (E-CR), 95, 132 hollow, 157-158 structures and properties, 83, 90, 93, 132-136, 138 Electrical and electronics applications, 140, 149, 152,315,321-322 Electrical conductivity, 258-259, 294 Electromagnetic shielding, 41, 150, 152 Endless fibers. See Continuous fibers Engines, 40,67--£8,71 ,322,323,324,325 Erbium doped fiber amplifier (EDFA), 195, 196 Etching, 29, 63 Evanescent field, 172, 173 Extrusion, 85, 244, 276 F Feedback control, 54-55 Fiber formation, 3, 103, 106, 220 Fiber FP, 70, 90, 211, 222 Fiberglass, 85,129,134-136 Fiber gratings, 196-197 Fiber optics. See Optical fibers Field-emitting structures, 30, 31, 40 Fluoride, 85, 97,101 Fluorophosphate glass fibers, 101 Four wave mixing (FWM), 196 Fracture toughness, 70. See also Mechanical toughness Fragile melts, 82, 84, 85, 87, 95-102 Fullerenes, 38, 60 Fumaces, 325

G Gelation, 206-207 General-purpose glass fibers, 94-95, 129-136 Germanium, 3,20, 21 , 34, 63, 64 Glass fibers, 3,124 chemical resistance, 145-149 commercial applications, 134-136, 147-148, 164-165 general-purpose, 94-95, 129-136 See also E-glass; Fiberglass; Inviscid melt spinning processes; Melt spinning

343

Index processes; Silica glass fibers; Silicate glass fibers Gold, 18,30 Graded index fibers, 172 Graphene layers, 234-235 Graphitization, 34-36, 234, 245 Green fibers. See Precursor fibers

H Hafnium carbide whiskers, 17 Heat management systems, 325 High-modulus (HM) fibers, 95,136,140-145, 148,235,247,250 High silica fibers, 165-166 High-strength (HS) fibers, 136-140, 148,235, 247,250 Hollow fibers, 58, 156-160 Hot filament chemical vapor deposition, 21-22, 55-56,58,65-68,69 Hybrid fiber forming processes, 100-101 Hydridopolysilazane (HPZ), 300, 302,303, 304

IMS-54,90 Indium,28 Industrial applications, 70, 149-150 Infrared (IR) absorption, 173-174 Infusible PCS fibers, 275 Insulation, 315 Inviscid melt spinning (IMS) processes, 84-86, 87-88,113-119 aluminate glass fibers, 90 amorphous fibers, 103, 105-107 metal fibers, 108-109 oxide glass fibers, 110-111 Ion bombardment, 25 Iron, 23 J

Jets. See liquid jet K Kashima Carbonic series, 248 Kumada rearrangment, 267, 269-270

L

Lanathana, 140 Laser ablation, 13,20-21 ,34 Laser-assisted chemical vapor deposhion (LCVD), 47-55,64, 65 automatic process control, 54-55 high pressure process (HP-LCVD), 53-54, 61-65,72,73 low pressure process (LP-LCVD), 49-53, 61, 62,65 rapidevaluation ofnew fibers, 71-73 rapid prototyping, 50, 53, 73-75 structure and property relationships, 61-65, 69 Laser-heated float zone (LHFZ) growth, 115-118, 159 Laser vaporization, 25, 60 Leaching, 165-166 Lead glass, 151 liquefied gases, 113 liquid jet, 103, 106 lifetime, 106, 107, 108 solidification, 111-113 liquid phase, 3-4, 26-29 liquid pitch melts, 11 liquidus temperature, 81 ,82-83, 93 lithography, 29, 30, 74 M Magnesium aluminosilicate, 95-97 Magnetron sputtering system, 58 Mats, 136 Mechanical prepolymer deposition, 74 Mechanical toughness, 70 Medical applications, 29,153,319-320 Melt assisted spinning, 237 Melt spinning processes, 81-86 generic, 85-86 and PCS, 269-272 structure and property relationship,87-92 viscous, 81-84 See also Fragile melts; Inviscid melt spinning processes; Strong melts Melt viscosity, 82 Mesophase pitch (Mesopitch, MP), 28-29, 233, 235,240-245,247-250 Metal catalyzed chemical vapor deposfion, 15-20,34,48,63,64,65 Metal matrixcomposites (MMCs),40, 320-322 Metal particle catalysts, 11 , 12, 13 Metglas, 105 Microcoils, 52,65 Microfilters,74

344 Micromechanical processess, 51, 53, 75 Micropillars, 11,29,30 Microporosity, 69 Microsolenoids, 52, 53 Microsprings, 17,51,58,69 Microstructures, 51-53, 54, 65, 74, 75 Microtexture, 60, 247-250 Microtubes, 58, 65 Microwaves, 152 Military applications, 140 Modified chemical vapor deposition (MCVO), 185-190 Modulus, 62, 70, 72, 88-90, 94 carbon nanotubes, 37, 38, 70, 72, 88-90 glass fibers, 136-137 single crystal fibers, 20 Molding applications, 41 Mullite, 208-209, 212-215, 216-218 N Nanopillars, 11 Nanotubes, 13, 24, 25. See also Carbon nanotubes Nanowires, 11, 13,20--21,26-27 Nextel fibers, 90, 218, 222, 224 Nicalon fibers, 68, 70, 71 , 72, 222, 279, 283, 286-287,290,291,292,320,323 Nickel-based alloys, 323 Niobium monocarbide (NbC),23 Nitridation, 22 Non-round cross sections, 154-155 Nylon, 154, 155

o Optical amplifiers, 179, 195-196 Optical fibers, 4,85, 92,169,198-199 devices, 194-198 drawing processes, 191-194 fabrication, 180--191 transmission principles, 169-180 Optical loss, 172-174 Outside vapor deposition (OVO), 183 Oxidation, 259-261, 291-293, 304-305 Oxide fibers, 110--111, 118-119 Oxygen-free SiC fibers, 272-275 Oxynitridefibers, 13,24,142-145 p

Packaging applications, 41, 71 , 322

Index Partially stabilized zirconia (PSZ), 225, 226 Perhydropolysilazane (PHPSZ), 307-308, 310 Phosphorescence, 116 Photonic band-gap (PBG) microstructures, 53, 75 Pitch, 235, 236, 239-245. See also Mesophase pitch Planar flow casting process, 103, 104 Plasma chemical vapor deposition (PCVO), 22, 58,190--191 Plasma processes. See Arc discharge; Carbon ion bombardment; Laser vaporization Polyacrylonitrile (PAN) fibers, 35, 36, 233,235, 236,237-239,247 Polyborodiphenylsiloxane (PBOPSO), 268 Polycarbonate substrates, 51 Polycarbosilane (PCS), 68, 266-268, 269-276, 294,307 Polycarbosilazane (PCSZ) precursors, 300--301, 302,303-304 Polycrystalline fibers, 62, 64, 65, 70, 84-85 alumina, 210--211 Polycyclic aromatichydrocarbons, 36 Polydimethylsilane (POMS), 267,268 Polymer impregnationpyrolysis (PIP), 322 Polymer matrix composhes (PMCs), 316-320 Polysilapropylene (PSP), 268-269 Polysilasilazane (PSSZ), 301 Polysilazane (PSZ) precursors, 299-300, 308 Polyzirconoxanes, 226-227 Porous hollow glass fibers, 156, 158 Potassium lithium niobate, 115 Powder-in-tube process, 118, 159 PRO-l66, 215-216, 222 Precursor (green) fibers, 3, 4,124 inchemical vapor infiltration, 59-60 PAN and pitch, 235-245 polycarbosilazane (PCSZ), 300--301, 302, 303-304 polysilazane (PSZ), 299-300, 308 Preforms,3,84,85,92,128, 163-164, 181, 191-193 Pressure vessels, 319 Prosthetics, 153, 326 Protective coatings, 192-193, 261 Pulse mode dispersion (PMD), 179 Pyrolytic processes, 23, 275-276, 322 Q

Ouaternary fibers, 84, 85, 90, 101-102 Quasi-stoichiometric SiC fibers, 266, 275-276 Quench rate, 103, 104

Index

R Radiation cured PCS fibers, 272-274 Radiation resistance, 150-151 Rapidevaluation processes, 71 -73 Rapid prototyping, 50, 53, 73-75 Rapid solidification (RS) process, 103-105 Rayleigh scattering, 173, 181 Rayleigh waves, 112 Reinforcing fibers, 66, 134 R-glass, 136-137, 138-139 Ribbons, 85, 103, 104-105, 155 Rice hull process, 13, 23, 34 Rockets, 73, 322 Rovings, 134

s Saphikon, 90 Sapphire fibers, 70, 71 , 72,114-115,118-119 Scanning devices, 32, 40 Scattering, 173 Selective laser sintering (SLS), 74 Sen-assembly processes, 11, 26-27, 60 Sen-catalyzation , 13 Sen-propagating high-temperature synthesis (SHS),24 Semiconductors, 34, 39, 40, 152 Sensor fibers, 73,74,108, 197-198 Service temperature, 67--Q8, 71,136 S-glass,83, 84, 90, 93,95, 137,149,157 Sharpened whiskers, 30-32, 33, 39-40 Sheath/core fibers, 21 , 22, 34, 58, 63,66-68, 69,97,156-160 Short fibers, 3, 6, 11 , 65. See also Amorphous fibers; Polycrystalline fibers SiAION fibers, 13,90,142-143,145 Side-by-side bicomponent fibers, 156, 160-162 Side growth, 12, 18,35,37,63,67 Silica glass fibers, 92-93 dry spinning, 123-128, 164-165 high-temperature, 162-166 Silicate glass fibers, 3,85, 88 fragile melts, 95-97 general-purpose, 129-136 non-roundcross section, 154-155 special-purpose, 136-153 structural,93-95, 129-162, 156-162 Silicon carbide/aluminum nitride composites, 71 Silicon carbide/carbon (SiC/C) fibers, 21 , 56, 63, 67--Q8,68, 70, 71, 72 Silicon carbide (SiC) fibers, 59, 64, 67--Q9, 70, 71 ,72,265-266 matrix composiles, 322, 323, 324

345 oxygen-free, 272-275 quasi-stoichiometric, 266, 275-276 structure and properties, 276, 279-280, 283-294 whiskers and nanowhiskers, 17,20,23-24, 34,40,70 Silicon carbide/titanium composites, 71 Silicon dioxide, 22 Silicon fibers, 11, 62, 63 Silicon nitride, 23, 63, 64, 69, 299 boride based fibers, 309-313 Si-C-N-O and Si-C-N fibers, 299-306, 312 Si-N-O and Si-N fibers, 306-309 whiskers, 17,20,24,63 Silicon oxide, 59 Silicon oxycarbide (Si-C-O) fibers, 70, 266-272, 276-279,281-294 passim Silicon whiskers and nanowhiskers, 3, 15-17, 18-19,20-21,39-40 structures and properties, 30-34 Silver nanowires, 11, 26-27 Single crystal fibers, 3,51,62,63,64,65,70,72 from inviscid melts, 86,113-119 melt processes, 86 See also Whiskers Sintering, 74,189-190,207 Sliver yarn, 124-125 Slurry processes, 70 Snell's law, 169-170 Soda-lime-silica, 181 Sodium borosilicate, 181 Sol-gel processing, 70,126-128 ceramicoxide fibers, 205-207 optical fibers, 193-194 Solid phase, 11, 12 advanced processes, 29-30 Solution-liquid-solid phase, 11 Solution-liquid-solid (SLS) phase, 28, 29 Solvent spinning. See Dry spinning Spacecraft, 317-318,321 ,322,323 Specialty fibers,6, 136-153 Spinorientation, 91 Sporting goods,70, 71,140,318,321 Steel,108 Step index fibers, 171-172 Stereolilhography, 74 Stiffness. See Modulus Strain sensors, 197 Strength, 66, 70, 91-92. See also Tensile strength Strong melts, 82-83, 87, 92-95 Structural fibers, 4, 93-95,129-162,156-162 Superconducting fibers, 42,116-118,152, 158-160 Supercooled melts, 86,97-99

346

Index

T Tellurite glass fibers, 98 Tensile strength, 91, 253-256, 284-285,

287-288

silicon nitride fibers, 312, 313 Ternary fibers, 88 Tetraethylorthosilicate (TEOS), 164, 193,323 Tetragonal zirconia polycrystals (TZP), 225-226 Thermal expansion, 257-258 Thermal stability silicon carbide fibers, 280-284 silicon nitride fibers, 302-304, 308-309, 311 Thermophoretic deposition, 189 Time-ta-failure, 224 Tip groW1h, 12,35,63,67 Titanium tetrabutoxide, 272 Tokawhiskers, 23 Transistors, 42 Transition alumina fibers, 211-212, 216, 222,

315

Transport properties, 258-259, 293-294 Traveling solvent zone melting (TSZM), 118 Tnchlorosilylamino-dichloroborane (TAOS), 311 Trimethylamine alane, 51 Tyranno, 272, 290

u Ultrahigh-modulus (UHM) fibers, 141-145,235 Ultrahigh temperature (UHT) fibers, 71-72, 162 Ultrapure silica fibers, 3,126-128, 162, 163-165 Uniformity. See Strength Updrawing, 85,86,97-100,102

v

Vapor-liquid-solid growth; Vapor-solid groW1h Vapor-solid (VS) growth, 3,13-14,24,34,63,64 Vertical axialdeposition (VAO), 183-185 Vibrational absorption, 173-174 Viscous melt spinning processes, 3,4,81-84

w Water glass solutions, 124-126 Waveguide physics, 169-172 Weibull distribution function, 253-254, 285 Weight-sensnive applications, 67,132,136,158,

321,323

Whiskers, 3,4, 11, 12 metal catalyzed, 15-20, 21 , 64 from organicsolvents, 27-29 structures and properties, 20, 30-34 See also Silicon carbide whiskers and nanowhiskers Wire drawing, 108, 109 Work-ta-break. See Mechanical toughness

x Xerogels, 207 X-ray lithography (L1GA), 74 X-ray transparency, 319-320 y

Yajima process, 267-269, 307 Young'smodulus, 37, 38, 250-251, 252-253 Ytlria-modified fibers, 84, 89, 90,141 Yttrium aluminum garnet (YAG) fibers, 3, 4, 72,

108,227-229

Vacuum microelectronics, 30, 40 Vapor-liquid-solid(VLS) groW1h, 3,12-13,15,

20.1 n , 23,24,25,29,34,63,64 Vapor phase, 3, 11-26,70,75. See also

Chemical vapor deposition; Chemical vapor infiltration; Laser vaporization;

z Zinc oxide modified fibers, 90, 140 Zirconia, 24,115,225-227,31