Ceramic matrix composites
i © Woodhead Publishing Limited, 2006
Related titles: Multi-scale modelling of composite material systems (ISBN-13: 978-1-85573-936-9; ISBN-10: 1-85573-936-4) This new book focuses on the fundamental understanding of composite materials at the microscopic scale, from designing microstructural features, to the predictive equations of the functional behaviour of the structure for a specific end-application. The papers presented discuss stress and temperature-related behavioural phenomena based on knowledge of physics of microstructure and microstructural change over time. The science and technology of materials in automotive engines (ISBN-13: 978-1-85573-742-6; ISBN-10: 1-85573-742-6) This new book provides an introductory text on the science and technology of materials used in automotive engines. It focuses on reciprocating engines, both four- and twostroke, with particular emphasis on their characteristics and the types of materials used in their construction. It considers the engine in terms of each specific part: cylinder, piston, camshaft, valves, crankshaft, connecting rod and catalytic converter. The intention is to describe the metallurgy, surface modification, wear resistance, and chemical composition of these materials. It also includes supplementary notes that support the core text. The book will be essential reading for engineers and designers, as well as lecturers and graduate students in the fields of combustion engineering, machine design and materials science, looking for a concise, expert analysis of automotive materials. Nanostructure control of materials (ISBN-13: 978-1-85573-933-8; ISBN-10: 1-85573-933-X) Nanotechnology is an area of science and technology where dimensions and tolerances in the range of 0.1–100 nm play a critical role. It encompasses precision engineering as well as electronics, electromechanical systems and mainstream biomedical applications in areas as diverse as gene therapy, drug delivery and novel drug discovery techniques. Nanostructured materials present exciting opportunities for manipulating structure and properties on the nanometre scale. The ability to engineer novel structures at the molecular level has led to unprecedented opportunities for materials design. This new book provides detailed insights into the synthesis/structure and property relationships of nanostructured materials. A valuable book for materials scientists, mechanical and electronic engineers and medical researchers. Details of these and other Woodhead Publishing materials books and journals, as well as materials books from Maney Publishing, can be obtained by: • visiting www.woodheadpublishing.com • contacting Customer Services (e-mail:
[email protected]; fax: +44 (0) 1223 893694; tel: +44 (0) 1223 891358 ext. 30; address: Woodhead Publishing Ltd, Abington Hall, Abington, Cambridge CB1 6AH, England) If you would like to receive information on forthcoming titles, please send your address details to: Francis Dodds (address, telephone and fax as above; e-mail:
[email protected]). Please confirm which subject areas you are interested in. Maney currently publishes 16 peer-reviewed materials science and engineering journals. For further information visit www.maney.co.uk/journals.
© Woodhead Publishing Limited, 2006
Ceramic matrix composites Microstructure, properties and applications Edited by I. M. Low
Woodhead Publishing and Maney Publishing on behalf of The Institute of Materials, Minerals & Mining CRC Press Boca Raton Boston New York Washington, DC
WOODHEAD
PUBLISHING LIMITED
Cambridge England
© Woodhead Publishing Limited, 2006
Woodhead Publishing Limited and Maney Publishing Limited on behalf of The Institute of Materials, Minerals & Mining Woodhead Publishing Limited, Abington Hall, Abington, Cambridge CB1 6AH, England www.woodheadpublishing.com Published in North America by CRC Press LLC, 6000 Broken Sound Parkway, NW, Suite 300, Boca Raton, FL 33487, USA First published 2006, Woodhead Publishing Limited and CRC Press LLC © Woodhead Publishing Limited, 2006 The authors have asserted their moral rights. This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. Reasonable efforts have been made to publish reliable data and information, but the authors and the publishers cannot assume responsibility for the validity of all materials. Neither the authors nor the publishers, nor anyone else associated with this publication, shall be liable for any loss, damage or liability directly or indirectly caused or alleged to be caused by this book. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming and recording, or by any information storage or retrieval system, without permission in writing from Woodhead Publishing Limited. The consent of Woodhead Publishing Limited does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific permission must be obtained in writing from Woodhead Publishing Limited for such copying. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation, without intent to infringe. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library. Library of Congress Cataloging in Publication Data A catalog record for this book is available from the Library of Congress. Woodhead Publishing Limited ISBN-13: Woodhead Publishing Limited ISBN-10: Woodhead Publishing Limited ISBN-13: Woodhead Publishing Limited ISBN-10: CRC Press ISBN-10: 0-8493-3476-4 CRC Press order number: WP3476
978-1-85573-942-0 (book) 1-85573-942-9 (book) 978-1-84569-106-6 (e-book) 1-84569-106-7 (e-book)
The publishers’ policy is to use permanent paper from mills that operate a sustainable forestry policy, and which has been manufactured from pulp which is processed using acid-free and elementary chlorine-free practices. Furthermore, the publishers ensure that the text paper and cover board used have met acceptable environmental accreditation standards. Project managed by Macfarlane Production Services, Dunstable, Bedfordshire, England (
[email protected]) Typeset by Replika Press Pvt Ltd, India Printed by TJ International, Padstow, Cornwall, England
© Woodhead Publishing Limited, 2006
Contents
Contributor contact details Introduction Part I: 1
1
Fibre-whisker- and particulate-reinforced ceramic composites
Fibrous monolithic ceramics
9
Y-H KOH, Seoul National University, Korea
1.1 1.2 1.3 1.4 1.5 1.6 1.7
Introduction History Processing Structures Mechanical properties Future trends References
9 9 11 14 15 28 29
2
Whisker-reinforced silicon nitride ceramics
33
M D PUGH, Concordia University, Canada and M BROCHU, McGill Univeristy, Canada
2.1 2.2 2.3 2.4 2.5
Introduction Fabrication Properties Applications References
33 34 37 54 55
3
Fibre-reinforced glass/glass-ceramic matrix composites
58
R BANERJEE and N R BOSE, Central Glass and Ceramic Research Institute, India
3.1 3.2
Introduction Types of fibre suitable as reinforcements in different glass/glass-ceramic matrix composites
© Woodhead Publishing Limited, 2006
58 60
vi
Contents
3.3 3.4 3.5 3.6 3.7 3.8
Methods for manufacturing different fibre-reinforced glass/glass-ceramic matrix composites Properties of glass/glass–ceramic matrix composites Microstructural observation Application areas Future trends References
72 80 92 93 95 95
4
Particulate composites
99
R I TODD, University of Oxford, UK
4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9
Introduction Powder processing and microstructural development Thermal microstresses Toughening Room-temperature strength High-temperature strength Wear Future trends References
99 100 103 105 110 116 120 124 125
Part II Graded and layered composites 5
Functionally-graded ceramic composites
131
I M LOW, R D SKALA and P MANURUNG, Curtin University of Technology, Australia
5.1 5.2 5.3 5.4 5.5 5.6 5.7
Introduction Infiltration kinetics and characteristics Infiltration processing of LGMs Characterisation and properties of alumina-matrix LGMs Concluding remarks Acknowledgements References
131 132 137 138 150 151 151
6
SiAION based functionally graded materials
154
H MANDAL, Anadolu University, Turkey and N ÇALIS ACIKBAS, MDA Advanced Ceramics, Ltd, Turkey
6.1 6.2 6.3 6.4 6.5 6.6 6.7
Introduction Functionally graded materials SiAlON ceramics Functionally graded SiAlON ceramics Production techniques of functionally graded SiAlON ceramics Concluding remarks References
© Woodhead Publishing Limited, 2006
154 154 155 160 161 174 175
Contents
7
Design of tough ceramic laminates by residual stresses control
vii
178
N ORLOVSKAYA, Drexel University, USA, M LUGOVY, Institute for Problems of Materials Science, Ukraine, J KUEBLER, EMPA, Lab for High Performance Ceramics, Switzerland, S YARMOLENKO and J SANKAR, North Carolina A&T State University, USA
7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8
Introduction Laminate design for enhanced fracture toughness Processing of Si3-T4–TiN and B4C–SiC ceramic laminates Si3N4 based laminates B4C based laminates Future trends Acknowledgements References
178 179 189 193 201 210 211 211
8
Hardness of multilayered ceramics
216
W J CLEGG, F GIULIANI, Y LONG, S J LLOYD, University of Cambridge, UK and J M MOLINA-ALDAREGUIA, Centro de Estudios e Investigaciones Tecnicas de Gipuzkoa (CEIT), Spain
8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8
Introduction Behaviour of multilayer structures Hardening mechanisms in multilayers Microstructural changes due to making a multilayer Conclusions Future trends Further reading References Part III:
9
216 217 219 230 236 237 237 237
Nanostructured ceramic composites
Nanophase ceramic composites
243
L YONGLI, Beijing University of Technology, China
9.1 9.2 9.3 9.4 9.5 9.6
Introduction Micro–nano type ceramic composites Nano–nano type ceramic composites Fabrication of nanoceramics Conclusions and future trends References
243 244 248 255 257 257
10
Nanostructured coatings on advanced carbon materials
260
Y MORISADA and Y MIYAMOTO, Osaka University, Japan
10.1 10.2
Introduction Coating method of nanostructured SiC
© Woodhead Publishing Limited, 2006
260 261
viii
Contents
10.3
Applications of nanostructured SiC coatings in advanced composites Conclusions References
273 281 281
Processing and microstructural control of metalreinforced ceramic matrix nanocomposites
285
10.4 10.5 11
W D KAPLAN, and A AVISHAI, Technion – Israel Institute of Technology, Israel
11.1 11.2 11.3 11.4 11.5 11.6 12
Introduction Processing Microstructure Properties Future trends References
285 285 290 300 304 304
Carbon nanotubes-ceramic composites
309
A PEIGNEY and CH LAURENT, CIRIMAT, Université Paul-Sabatier, France
12.1 12.2 12.3 12.4 12.5 12.6 12.7 13
Introduction Structure, synthesis and properties of carbon nanotubes Preparation of CNT-ceramic composites Properties of CNT-ceramic composites Conclusions and future trends Sources of further information References
309 309 313 320 329 330 331
Machinable nanocomposite ceramics
334
R WANG, Arizona State University, USA
13.1 13.2 13.3 13.4 13.5 13.6 13.7
Introduction Design principles of machinable ceramics Al2O3–LaPO4 Si3N4/h-BN Machinable nanocomposites Conclusions References Part IV:
14
334 334 335 344 351 352 354
Refractory and speciality ceramic composites
Magnesia–spinel (MgAl2O4) refractory ceramic composites
359
C AKSEL, Anadolu University, Turkey and F L RILEY, University of Leeds, UK
14.1
Introduction
© Woodhead Publishing Limited, 2006
359
Contents
14.2 14.3 14.4 14.5 14.6 14.7 14.8 14.9 14.10 14.11 14.12 14.13 15
ix
Crystal structures Production of MgAl2O4 Densification In-situ formed/preformed spinel based refractories Industrial applications and properties of magnesia–spinel materials Thermal shock Mechanical properties and thermal shock behaviour of magnesia–spinel composite refractory materials Conclusions Future trends Acknowledgements Sources of further information References
362 362 363 365
Thermal shock of ceramic matrix composites
400
366 372 375 389 390 392 392 393
C KASTRITSEAS, P SMITH and J YEOMANS, University of Surrey, UK
15.1 15.2 15.3 15.4 15.5 15.6 15.7 15.8 16
Introduction Thermal shock of brittle materials: the induced stress field Experimental methods Thermal shock of monolithic ceramics Thermal shock of particle- and whisker-reinforced CMCs Thermal shock of fibre-reinforced CMCs Concluding remarks References
400 401 407 410 413 416 427 428
Superplastic ceramic composites
434
A DOMÍNGUEZ-RODRÍGUEZ, D GÓMEZ-GARCÍA, Universidad de Sevilla, Spain, and F WAKAI, Tokyo Institute of Technology, Japan
16.1 16.2 16.3 16.4 16.5 16.6 16.7 16.8
Introduction Macro- and microscopic superplastic characteristics Accommodation processes controlling superplasticity Parameters improving superplasticity Applications of superplasticity Future trends Acknowledgements References Part V:
17
434 435 439 445 448 452 454 454
Non-oxide ceramic composites
Interfaces in non-oxide ceramic composites
461
S TURAN, Anadolu University, Turkey and K M KNOWLES, University of Cambridge, UK
17.1
Introduction
© Woodhead Publishing Limited, 2006
461
x
Contents
17.2
Assessment of the accuracy of TEM techniques for the detection and measurement of film thickness at interfaces Wetting, non-wetting and dewetting behaviour of interphase boundaries in non-oxide ceramic composites Equilibrium film thickness at interphase boundaries Effect of intergranular film composition on equilibrium film thickness Crystallography of interphase boundaries Future trends Further reading References
17.3 17.4 17.5 17.6 17.7 17.8 17.9 18
Sialons
463 467 469 473 475 482 484 486 491
Z B YU, Queen’s University, Canada and D P THOMPSON, University of Newcastle, UK
18.1 18.2 18.3 18.4 18.5 18.6 19
Introduction Sialons Challenges in toughening and strengthening sialons Sialon composites Future trends References
491 492 493 494 510 510
Carbon-ceramic alloys
514
C BALÁZSI, Research Institute of Technical Physics and Materials Science, Hungary
19.1 19.2 19.3 19.4 19.5 19.6 19.7 20
Introduction Carbon as fugitive additive for porous silicon nitride processing Comparison of silicon nitrides with carbon additions prepared by hot isostatic pressing and pressureless sintering In situ processing of Si3N4/SiC composites by carbon addition Silicon nitride ceramics reinforced with carbon fibres and carbon nanotubes Concluding remarks References Silicon nitride and silicon carbide-based ceramics
514 515 518 524 530 533 533 536
Y ZHANG, New York University College of Dentistry, USA
20.1 20.2 20.3 20.4 20.5
Introduction Material selection Material characterization Erosion response Microstructure and mechanical properties
© Woodhead Publishing Limited, 2006
536 537 540 543 550
Contents
20.6 20.7 20.8 21
xi
Microstructure and erosion mechanisms Conclusions References
552 556 557
Oxynitride glasses – glass ceramics composites
560
S HAMPSHIRE, University of Limerick, Ireland
21.1 21.2 21.3 21.4 21.5 21.6 22
Introduction Potential applications Oxynitride glass/glass ceramic composites Oxynitride glass–silicon carbide composites Conclusion References
560 560 561 569 572 572
Functionally graded ceramics
575
G ANNÉ, J VLEUGELS and O vAN University Leuven, Belgium
22.1 22.2 22.3 22.4 22.5 22.6 22.7 22.8
DER
BIEST, Katholieke
Introduction Functionally graded ceramics concept Classifications of FG ceramics Processing of FGMs FGM design for structural applications Future trends Further reading References
© Woodhead Publishing Limited, 2006
575 575 577 577 584 590 591 591
Contributors contact details
(* = main contact)
Chapter 1
Chapter 3
Professor Young-Hag Koh School of Materials Science and Engineering Seoul National University Seoul, 151-742 Korea
Dr Rajat Banerjee* and Dr Nripati Ranjan Bose Central Glass and Ceramic Research Institute Jadavpur Kolkata-700 032 India
Tel: 82-2-880-1397 Fax: 82-2-884-1413 E-mail:
[email protected]
E-mail:
[email protected]
Chapter 4 Chapter 2 Dr Martin Pugh* Department of Mechanical and Industrial Engineering Concordia University 1455 de Maisonneuve Blvd, West Montréal, Québec, H3G 1M8 Canada E-mail:
[email protected] Dr Mathieu Brochu Department of Mining, Metals and Materials Engineering 3610 University Street Montréal, Québec H3A 2B2 Canada E-mail:
[email protected]
Dr Richard I. Todd University of Oxford Department of Materials Parks Road Oxford OX1 3PH UK E-mail:
[email protected]
Chapter 5 Professor It-Meng (Jim) Low* Department of Applied Physics Curtin University of Technology GPO Box U1987 Perth, WA 6845 Australia E-mail:
[email protected]
© Woodhead Publishing Limited, 2006
xiv
Contributor contact details
Dr Robert D. Skala Millennium Chemicals, Inc. 6752 Baymeadow Drive Glen Burnie Maryland, MD 21060 USA Dr P. Manurung Department of Physics University of Lampung Bandar Lampung 35145 Indonesia
Chapter 6 Professor Dr Hasan Mandal* Anadolu University Department of Materials Science and Engineering 26470, Eskisehir Turkey Tel: +90 222 322 36 62 Fax: +90 222 323 95 01 E-mail:
[email protected] Nurcan Çalıs Acıkbas MDA Advanced Ceramics Ltd Organize Sanayii Bölgesi Eskisehir Turkey Tel: +90 222 2301880 Fax: +90 222 2301881 E-mail:
[email protected]
Chapter 7 Professor N. Orlovskaya* Department of Materials Science and Engineering Drexel University 3141 Chestnut Street Philadelphia, PA 19104 USA E-mail:
[email protected] Dr M. Lugovy Institute for Problems of Materials Science 3 Krzhizhanovskii Street 03142 Kiev Ukraine E-mail:
[email protected] J. Kuebler EMPA, Laboratory for High Performance Ceramics Ueberlandstrasse 129 CH-8600 Duebendorf Switzerland E-mail:
[email protected] S. Yarmolenko and J. Sankar Department of Mechanical Engineering North Carolina A&T State University 1407 E. Market St. Greensboro, NC 27411 USA E-mail:
[email protected] E-mail:
[email protected]
© Woodhead Publishing Limited, 2006
Contributor contact details
Chapter 8 Dr William John Clegg*, F Giuliani, Dr Y. Long, Dr S. J. Lloyd Gordon Laboratory Department of Materials Science and Metallurgy University of Cambridge Pembroke Street Cambridge CB2 3QZ UK Tel: +44 (0)1223 334470 E-mail:
[email protected] Dr J. M. Molina-Aldareguia Centro de Estudios e Investigaciones Tecnicas de Gipuzkoa (CEIT) Paseo de Manuel Lardizabal, 15 San Sebastian 20018 Spain
Chapter 10 Dr Yoshiaki Morisada Joining and Welding Research Institute Osaka University Ibaraki Osaka 567-0047 Japan Tel: +81-6-6879-8693 E-mail:
[email protected] Professor Yoshinari Miyamoto* Joining and Welding Research Institute Osaka University Ibaraki, Osaka 567-0047 Japan E-mail:
[email protected]
Chapter 11 Chapter 9 Dr Yongli Li Key Laboratory of Advanced Functional Materials, Ministry of Education of China Beijing University of Technology Pingleyuan 100, Chaoyang District Beijing 100022 China Tel: +86 10 67391760, Fax: +86 10 67392840 E-mail:
[email protected],
[email protected]
© Woodhead Publishing Limited, 2006
xv
Professor Wayne D. Kaplan* and Dr Amir Avishai Department of Materials Engineering Technion – Israel Institute of Technology Haifa 32000 Israel E-mail:
[email protected]
xvi
Contributor contact details
Chapter 12
Chapter 14
Professor Alain Peigney* CIRIMAT, UMR UPS-INPT-CNRS 5085 Université Paul-Sabatier, Bat 2R1 118 Route de Narbonne 31062 Toulouse Cedex 4 France
Assistant Professor Dr Cemail Aksel* Anadolu University Faculty of Engineering and Architecture Department of Materials Science and Engineering Iki Eylül Campus 26470 Eskisehir Turkey
Tel: 05 61 55 61 75 E-mail:
[email protected] Professor Christophe Laurent CIRIMAT, UMR UPS-INPT-CNRS 5085 Université Paul-Sabatier, Bat 2R1 118 Route de Narbonne 31062 Toulouse Cedex 4 France Tel: 05 61 55 61 63 E-mail:
[email protected]
Chapter 13 Dr Ruigang Wang Science and Engineering of Materials Program and Center for Solid State Science Arizona State University Tempe, AZ 85287-1704 USA E-mail:
[email protected]
Tel: 00-90-222-3350580/6362 Fax: 00-90-222-3239501 E-mail:
[email protected] Professor Frank L. Riley Department of Materials, School of Process, Environmental and Materials Engineering University of Leeds Leeds LS2 9JT UK Tel: 00-44-113-3432531 Fax:00-44-113-2422531 E-mail:
[email protected]
Chapter 15 Dr Christos Kastritseas, Dr Paul Smith and Dr Julie Yeomans* Reader in Ceramic Materials School of Engineering Postal Area H6 University of Surrey Guildford Surrey GU2 7XH UK Tel: +44 (0) 1483 689613 E-mail:
[email protected]
© Woodhead Publishing Limited, 2006
Contributor contact details
Chapter 16
Chapter 18
Arturo Domínguez-Rodríguez* and D. Gómez-García Departamento de Física de la Materia Condensada Universidad de Sevilla Apartado 1065 41080 Sevilla Spain
Dr Z. B. Yu* Centre for Manufacturing of Advanced Ceramics and Nanomaterials Queen’s University Kingston, Ontario, K7L 3N6, Canada
E-mail:
[email protected]
E-mail:
[email protected]
F. Wakai Materials and Structures Laboratory Tokyo Institite of Technology 4259 Nagatsuta Midori Yokohama 226-8503 Japan
D. P. Thompson Advanced Materials Group School of Chemical Engineering and Advanced Materials University of Newcastle Newcastle upon Tyne NE1 7RU UK E-mail:
[email protected]
Chapter 17 Professor Servet Turan* Anadolu University Department of Materials Science and Engineering Iki Eylül Campus 26555 Eskisehir Turkey E-mail:
[email protected] Dr Kevin M. Knowles University of Cambridge Department of Materials Science and Metallurgy Pembroke Street Cambridge CB2 3QZ UK E-mail:
[email protected]
© Woodhead Publishing Limited, 2006
Chapter 19 Dr Csaba Balázsi Ceramics and Composites Laboratory Research Institute of Technical Physics and Materials Science Hungarian Academy of Sciences 1121 Budapest XII Konkoly-Thege u. 29-33 Hungary Tel: +36-1-3922222/3279 Fax: +36-1-3922226 http://www.mfa.kfki.hu E-mail:
[email protected]
xvii
xviii
Contributor contact details
Chapter 20 Dr Yu Zhang Department of Biomaterials and Biomimetics New York University College of Dentistry Room 813C 345 East 24th Street New York, NY 10010 USA E-mail:
[email protected]
Chapter 21 Professor Stuart Hampshire Materials Ireland Research Centre & Materials and Surface Science Institute University of Limerick Limerick Ireland E-mail:
[email protected]
Chapter 22 Guy Anné, Jozef Vleugels and Omer Van der Biest* Department of Metallurgy and Materials Engineering (MTM) Katholieke University, Leuven, Kasteelpark Arenberg 44 B-3001 Heverlee (Leuven) Belgium E-mail: Omer.Vanderbiest@mtm. kuleuven.ac.be
© Woodhead Publishing Limited, 2006
Part I Fibre-whisker- and particulate-reinforced ceramic composites
© Woodhead Publishing Limited, 2006
1 Fibrous monolithic ceramics Y - H K O H, Seoul National University, Korea
1.1
Introduction
Fibrous monolithic ceramics (FMs) consist of a hexagonal arrangement of submillimeter ‘cells’ of strong polycrystalline ceramic and a network of crack-deflecting weak ‘cell boundaries’ [1, 2]. These composites are sintered or hot-pressed monolithic ceramics with a distinct fibrous texture. This unique architecture opened new avenues for ceramic composites, in which they fail in a nonbrittle manner because of crack interactions with weak cell boundaries, such as crack deflection or crack delamination. This approach provides a simple and versatile method for manufacturing nonbrittle ceramic composites from a variety of different material combinations that include oxide ceramics (Al2O3/Al2O3–ZrO2 [3]) and non-oxide ceramics (SiC/graphite [4, 5], SiC/ BN [6] and Si3N4/BN [1, 7–16]). This chapter presents an overview of these composites with a variety of material combinations, as well as their architectures. The major objectives of this chapter are the following: (a) to briefly discuss the history and main concepts of FMs in section 1.2; (b) to address the experimental ways to prepare these composites, including coextrusion, microfabrication by coextrusion, and hybrid extrusion and dip-coating, in section 1.3; (c) to demonstrate the various types of FMs in the form of oxide-based or nonoxide-based composites, as well as their various architectures in section 1.4; (d) to discuss the mechanical properties of Si3N4/BN FMs at room and high temperatures, as well as their fracture mechanisms, in section 1.5; and finally (e) to give a personal perspective on the future of these wonderful composites in section 1.6.
1.2
History
Ceramics have been well known to offer potential benefits over metal parts in high-temperature environments, such as higher strength, lower density, and greater resistance to oxidation. In spite of these merits, their low fracture 9 © Woodhead Publishing Limited, 2006
10
Ceramic matrix composites
resistance with poor reliability has long been a major concern. A number of potential toughening mechanisms have been proposed [17, 18]. One of the most successful techniques is the fiber-reinforced ceramic composites [19, 20]. In a tough ceramic composite, the matrix crack is deflected at the interface and ceases to propagate, allowing the composite to possess high toughness along with nonbrittle failure by extensive crack interactions, such as bridging of the primary crack, crack deflection along the fiber/matrix interface and frictional sliding with fiber pullout of the matrix [20]. However, fiber-reinforced ceramic composites have not been fully utilized yet because of their manufacturability and cost. Also, densification techniques such as (hot pressing and hot isostatic pressing) often lead to fiber damage, deteriorating its mechanical properties. Another toughening concept was introduced by Cook and Gordon in 1964 [21 They suggested crack propagation in brittle materials could be controlled by incorporating a fabric of microstructural features that change the crack path, resulting in high toughness by crack blunting at the weak interface and crack delamination. Rather than rely on bridging mechanisms to improve toughness, the advancing crack was blunted at the weak interface and forced to reinitiate in order to continue propagation. Based on this concept, in 1988 Coblenz described a method for producing a pressureless sinterable ceramic composite consisting of a strong load-bearing phase surrounded by a continuous weak crack-deflecting phase [2]. Another variation on the concept of Cook and Gordon was introduced by Clegg et al. in 1990 [22]. This approach involves alternating layers of the strong load-bearing phase and the weak crack-deflecting phase, allowing the composite to fail in a nonbrittle manner with high strength and toughness (Fig. 1.1(a)). Following failure of this layer, a series of decreasing steps were observed in the load–deflection behavior of the laminate, as additional layers failed with subsequent crack arrest via deflection along the weak interfaces. In 1991, Halloran and co-workers pioneered an exciting new class of structural ceramics, so-called ‘fibrous monolithic ceramics’ that exhibit mechanical properties similar to ‘continuous fiber ceramic composites’, including very high fracture energies, damage tolerance and graceful failure [1, 3–5]. They consist of 250-micron ‘cells’ of a strong polycrystalline ceramic, such as silicon carbide or silicon nitride, separated by ‘cell boundaries’ from materials, such as boron nitride, which promote crack deflection and delamination (Fig. 1.1(b)). Fibrous monolithic ceramics are produced most often by extrusion, followed by lay-up of filaments into laminates. The extruded filaments consist of a cell phase surrounded by a sheath that forms a continuous cell boundary. This approach creates analogs of many composite architectures, allowing these composites to fail in a nonbrittle manner with energy dissipation arising from sliding of the cells, and branching and deflection of cracks. Such fibrous monolithic ceramics constitute lower-cost alternatives
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
11
Strong layer Weak layer
(a) Laminated composite
Strong cell
Weak cell boundary
(b) Fibrous monolithic ceramic
1.1 Schematics illustrating (a) laminated composite, consisting of the strong layer and the weak layer, and (b) fibrous monolithic ceramic, consisting of the strong cell and the weak cell boundary.
to conventional continuous-fiber ceramic composites in some applications, and a wide variety of fibrous monolithic ceramics are available commercially. A number of the different cell/cell-boundary combinations have been investigated, including oxide and non-oxide ceramics.
1.3
Processing
In this section, we discuss three prevalent processing methods for producing fibrous monolithic ceramics, i.e. coextrusion [1, 23], microfabrication by coextrusion [24], and hybrid extrusion and dip-coating [25].
1.3.1
Coextrusion
A significant advancement in the processing of FMs was made by the development of coextrusion. The schematic illustrations showing processing are shown in Fig. 1.2. This process involves forming a feedrod consisting of a core of the cell material surrounded by a shell of the cell boundary material, prepared using a cylindrical mold and a half-pipe shape (Fig. 1.2(a)). Both the core and the shell are blends of thermoplastic polymer and ceramic powder. The feedrod is then used to coextrude a ceramic green fiber identical to the feedrod in core/shell proportion, but 100 times smaller (Fig. 1.2(b)). The coextruded fibers are uniaxially aligned and then warm pressed to fabricate green billet (Fig. 1.2(c)). In addition, for the multilayer FMs, the filament direction can be rotated between the layers. The resulting green billets undergo binder removal and hot-pressing to produce densified FMs (Fig. 1.2(d)).
© Woodhead Publishing Limited, 2006
12
Ceramic matrix composites
Cell boundary
Cell
(a) (b) Hot-pressing
(c)
(d)
1.2 Schematic illustrations showing coextrusion process to fabricate FMs: (a) initial feedrod formation, (b) coextrusion, (c) lamination, and (d) hot-pressing.
There are many advantages to this technique. In general, the coatings on fibers produced by this method are more uniform than those on dip-coated fibers, improving the overall uniformity of the composite. Unlike dip-coating, the coating thickness is determined by the shell thickness on the feedrod, not the rheological properties of the dip-coating slurry. Thus batch-to-batch repeatability is also improved. The coextruded fibers are also much easier to handle than the dry-spun green fibers, making processing much easier to perform.
1.3.2
Microfabrication by coextrusion (MFCX)
A variation on this approach used multifilament coextrusion, so-called ‘microfabrication by coextrusion (MFCX )’. A limitation of the single-filament process is the size of the filament. The rheological properties of the polymer/ ceramic blends make spinning fibers smaller than 250 µm very difficult. Additionally, spooling fine-diameter fibers is quite challenging. The MFCX is shown schematically in Fig. 1.3. The setup is the same as that used to spin fibers except that the spinneret is replaced with an extrusion die with a diameter between 1 mm and 6 mm. Two separate extrusion steps are used. In the first step, coarse primary filaments are extruded from the feedrod (Fig.
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
13
Primary filament (b)
Cell
Cell boundary
Second filament (c)
(a)
1.3 Schematic illustrations showing MFCX process to fabricate FMs: (a) coextrusion, (b) assembly of primary filament, and (c) filament after second coextrusion.
1.3(a)). The primary filaments are bundled together to form a second feedrod. On extrusion of the second feedrod, a filament containing many cells in its cross-section is formed (Fig. 1.3(b)). The second filament is referred to as a multifilament strand, indicating the origin of the strand (Fig. 1.3(c)). The similar procedure to coextrusion is employed to fabricate specimens, including strand alignment, lamination, and thermal treatment. The scale of the cells within the multifilament strand is controlled by two factors: (1) the number of primary filaments bundled into the second feedrod, and (2) the size of the extrusion die used to form the second filament. The number of filaments bundled into the second feedrod is determined by both the size of the primary filaments and the size of the second feedrod. The decreasing ratio of extrusion die sizes results in a dramatic increase in the number of cells within a strand. In addition to making finer cell sizes possible, it is also much easier to lay coarse multifilament strands into a die than it is to lay small fibers. The strands lay straight and flat, whereas fibers tend to curl up and become intertwined. However, despite the fine cell size, the coarseness of the strands limits their use to architectures in which the scale of a cluster of cells is no less than ~ 0.75 mm.
1.3.3
Hybrid extrusion and dip-coating
As previously mentioned, FMs were first fabricated by hybrid extrusion and dip-coating [4–6]. Firstly, highly concentrated, viscous slurries are extruded and then dried to form a green fiber. The resulting green fibers are dip-coated
© Woodhead Publishing Limited, 2006
14
Ceramic matrix composites
Si3N4 polymer
Monolithic Si3N4 Si3N4/BN FM 20 wt%
0 wt%
Monolithic Si3N4
BN-containing slurry (a)
(b)
1.4 (a) Schematic illustrating hybrid extrusion and dip-coating process to fabricate a Si3N4/BN FM composite; (b) schematic of the FM composite, consisting of monolithic Si3N4 and Si3N4/BN FM.
in a slurry containing the cell boundary material. The coated fibers (core/ shell) are cut to length and stacked in a die to form a green body. The green body undergoes the binder removal and hot-pressing. Although this process limits the uniform coating of cell boundaries, it can create FMs containing gradient cell boundary thickness in a simple way. For instance, in 2002, Koh et al. fabricated the three-layered Si3N4 /BN FMs by simply adjusting the BN-containing slurry for the cell boundary phase during the dip-coating stage (Fig. 1.4) [25]. In this process, Si3N4-polymer dough instead of thermoplastic ceramic compound is used for spinning. For the cell boundary, two kinds of dip-coating slurries with different BN concentrations were used. The Si3N4-polymer is extruded using a syringe with a piston through a 300 µm orifice, and then dip-coated by passing the BN-containing slurry with 20 wt% and 0 wt% concentration (Fig. 1.4(a)). The coated fiber was uniaxially arrayed and then dried in an oven at 80°C for 12 h to improve the shape and strength of the fiber by hardening. The green billet was pressed using a square mold. Thereafter, the green billet undergoes thermal treatment to produce FM composite, consisting of monolithic Si3N4 and Si3N4/BN FM (Fig. 1.4(b)).
1.4
Structures
1.4.1
Various material combinations
Fibrous monolithic ceramics consist of dense cells separated by a continuous cell boundary, in which the cells provide most of the strength of the FM and
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
15
the cell boundary provides the toughness by isolating the cells from each other and promoting dissipation of fracture energy by mechanisms such as pullout of the cells or deflection of a crack through the cell boundary [1]. The cell boundaries must be either weak themselves or poorly bonded to the cells to dissipate fracture energy and exhibit minimal or no reaction with the cells for long-term use at elevated temperatures. To date, many kinds of structural ceramics have been examined for a strong cell phase. They are in the forms of either oxides (Al2O3 [3], ZrO2 [26], and ZrSiO4 [27]) or nonoxides (SiC [4–6], Si3N4 [8], and borides [1, 7–16]). Oxides have the advantage of stability in oxidizing environments, while non-oxides have the advantage of substantially higher strength and superior creep resistance. In all-oxide FMs, porous cell boundaries are generally employed because they can minimize transmission of fracture energy to the cells. They have been formed from large-grained ceramic powders or a mixture of ceramic particles and platelets. In addition, the cell-boundary phase must be resistant to microstructural alteration with prolonged heating for elevated-temperature application. For non-oxide FMs, similar criteria for selecting a cell boundary should be considered. To date, hexagonal-BN has been extensively used as cell boundary because BN forms a dense, highly textured cell boundary that bonds only weakly to the cell by hot-pressing. In most cases, Si3N4 has been used for the cells. Oxide sintering aids (e.g. Al2O3 and Y2O3) in the Si3N4 migrate into the BN cell boundary during hot-pressing and impart the bonding [1, 24].
1.4.2
Various architectures
The processing methods used to manufacture FMs are very versatile, allowing any number of unique, submillimeter architectures to be designed and fabricated. For example, sheets of filament can be rotated with respect to one another to form a multiaxial architecture (Fig. 1.5(a)) [1, 25]. Another good example is to tailor the cell boundaries, in which the cell boundaries contain a thin web of Si3N 4 reinforcement (Fig. 1.5(b)) [1, 25]. Such Si3N 4 reinforcement may alter the fracture behaviors and mechanical properties of Si3N4/BN FMs at both room and high temperatures. This capability of tailoring the architecture can offer the opportunity to allow FMs to achieve better performance at both room and high temperatures.
1.5
Mechanical properties
Since fibrous monolithic ceramics are intended for use in applications where stresses are primarily generated due to bending, strength and work-of-fracture in flexure are measured to evaluate their basic mechanical properties. In addition, factors determining the manner of crack propagation should be
© Woodhead Publishing Limited, 2006
16
Ceramic matrix composites (b)
(a)
Si3N4 Cell BN Si3N4 50 µm
1.5 (a) Low-magnification SEM composite showing three sections of a fibrous monolith with a [0°/90°] architecture; (b) cross-sectional view of FM fabricated with a thin web of Si3N4 reinforcing the BN cell boundary (adapted from ref. [1]).
examined, including the fracture resistance of cells, the interfacial fracture resistance, and the interfacial sliding resistance, and residual stresses should be investigated to understand how cracks propagate in flexure. Among the many FMs that have been investigated, Si3N4/BN FMs have achieved the best overall properties and have been reliably manufactured on a commercial scale [28]. Therefore, this section deals with the mechanical properties of Si3N4/BN FMs primarily at room and high temperatures, as well as with their failure mechanisms.
1.5.1
Room-temperature properties
Mechanical properties and failure mechanisms Si3N4/BN FMs have shown promise as a structural and tough ceramic material [1, 7–15, 29]. The unique flexural response of FMs is a result of the submillimeter architecture that is engineered into the material. A typical microstructure of fibrous monolithic Si3N4/BN ceramic (FM) is shown in Fig. 1.6. Low-magnification SEM micrographs of polished sections show three-dimensional representations of the submillimeter structure of two architectures of fibrous monoliths. The polycrystalline silicon nitride cells (~250 µm) appear in dark contrast, while the continuous boron nitride cell boundaries (20 µm) appear in bright contrast. The Si3N4 cells are surrounded by the cell boundaries consisting of BN particles bonded with yttrium aluminosilicate. The typical stress versus crosshead deflection response associated with a Si3N4/BN FM is contrasted with that of a monolithic Si3N4 in Fig. 1.7. The monolithic Si3N4 exhibits greater strength but negligible apparent work-offracture (WOF) because of brittle failure (Fig. 1.7(a)). On the other hand, the Si3N4/BN FM demonstrates some load-bearing ability following the first
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
17
250 µm
1.6 Low-magnification SEM micrograph of polished sections, showing three-dimensional representations of the submillimeter structure of two architectures of Si3N4/BN FM (adapted from ref. [29]). 800
Apparent stress (MPa)
(a) Monolithic Si3N4 600
400
(b) Fibrous monolith
200
0 0.0
0.2 0.4 0.6 0.8 Crosshead displacement (mm)
1.0
1.7 Flexural response of (a) monolithic Si3N4 and (b) Si3N4/BN FM (adapted from ref. [29]).
failure event with significant WOF (Fig. 1.7(b)). To date, the flexural strength and WOF of Si3N4/BN FMs have significantly improved up to ~700 MPa and ~8000 kJ/m2, by optimizing respectively manufacturing routes and microstructures [1]. In unidirectional FMs, fracture can initiate when the tensile stress carried by the cell exceeds the strength of the cell [1]. This is generally favorable when the cell boundaries are tough in comparison to those of the cells. A
© Woodhead Publishing Limited, 2006
18
Ceramic matrix composites
flaw on the tensile surface of FMs will cause failure of only a single cell. The maximum applied load is typically achieved at the point just prior to failure in the layer of the cells closest to the tensile surface, because the loadbearing capacity is reduced due to the reduction in the effective cross-section of the sample. Thereafter, each stress drop in FMs is frequently followed, owing to the fracture of one or several layers of cells (Fig. 1.7(b)). As well as the strength of the cell, the strength of unidirectional FMs depends on their orientation with respect to the loading axis. The flexural strength of FMs decreases dramatically as the direction of stress application changes from on-axis to off-axis [1, 24]. The strength of FMs tested off-axis is much lower than that for on-axis orientation because failure is determined by the strength of the cell boundaries, rather than that of the cells. FMs with multiaxial architectures will be an important part of designing for complex stress states because many applications involve biaxial or off-axis loading conditions. The strength is determined by both cell and cell boundary fractures. If cells on the tensile surface are not aligned in the direction of the applied stress, failure of the cell boundary on the tensile surface can occur at a relatively low load, but cells with 0° orientations that are just beneath the tensile surface can continue to bear substantially more load [1, 24]. In addition to strength and WOF of FMs, the elastic behavior of these architectures should be considered. Simple brick models were proposed to accurately predict elastic properties of FMs [1, 24]. Figure 1.8 shows the elastic modulus versus orientation for uniaxially aligned Si3N4/BN FMs with experimentally measured values, indicating that there is very good agreement between experiment and prediction. This prediction can be used for FMs with multiaxial architectures.
Young’s modulus E1 (GPa)
300 Measured Predicted 250
200
150
100 0
20
40 60 Orientation, θ (degrees)
80
1.8 Young’s modulus versus orientation for uniaxially aligned Si3N4/ BN FM (adapted from ref. [1]). The line is the predicted behavior using the brick model and laminate theory.
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
19
Crack propagation and energy absorption The unique nonbrittle failures of FMs are accomplished as a consequence of crack deflection at the BN cell boundaries as well as significant crack delamination and sliding, as shown in Fig. 1.9. During crack propagation, tensile-initiated cracks are turned on weak interfaces, creating surface area and thus absorbing energy in the process. This process, known as crack deflection and crack delamination, is repeated as the crack works its way through the thickness of the sample. There are several factors that determine the degree of the crack interactions with the BN cell boundaries. In general, crack deflection is predicted when the fracture resistance of the interface is low and when the elastic mismatch between the cell and the cell boundary is high. In 1998, Kovar and coworkers. observed that the addition of a Si3N4 phase into a BN interphase in laminate composite decreased the interfacial fracture resistance, reducing the degree of crack delamination [30], as shown in Fig. 1.10(a)–(d). Extensive delamination and high energy absorption are observed only in materials that have the lowest interfacial fracture resistance (Fig. 1.10(a)). In addition to Si3N4/BN laminate composites, the fracture resistances of Si3N4/BN FMs were also examined [31, 32]. The energy-absorption capacity of FMs is primarily influenced by the crack path after the initial crack deflection occurs. Long delamination distances are favored when the interfacial fracture resistance is low, the flaw size in the layers is small, and the fracture resistance of the layers is high [30, 33]. A map of crack propagation behavior in Si3N4/BN FM is shown in Fig. 1.11. At very low values of the interfacial fracture resistance, increasing the interfacial fracture resistance causes more energy to be dissipated through the creation
1 mm
1.9 Optical photograph of crack propagation of the fibrous monolithic Si3N4/BN ceramic after flexural testing (adapted from ref. [29]).
© Woodhead Publishing Limited, 2006
20
Ceramic matrix composites
(a)
(b)
(c)
(d)
1.10 SEM micrographs of the side surface of flexural specimens containing (a) 10, (b) 25, (c) 50, and (d) 80 vol% Si3N4 in the interphase (adapted from ref. [30]).
Critical flaw size, a (µm)
200
Crack kinking
150
100
Delamination
50
0 0
0.1
0.2
0.3 Γi / ΓSi3N4
0.4
0.5
0.6
1.11 Critical flaw size predicted to cause a crack to kink out of the BN interphase, plotted against the ratio of the interfacial fracture resistance to the fracture resistance of Si3N4 (adapted from ref. [30]).
of interfacial crack area. However, if the interfacial fracture resistance is too high, crack kinking will reduce the delamination area. This observation suggests that there is an optimum interfacial fracture resistance that maximizes the energy-absorption capability, and this optimum value is determined by the transition from delamination and crack kinking.
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
1.5.2
21
High-temperature properties
Elevated temperature mechanical properties Since Si3N4/BN FMs are candidates for high-temperature applications, for instance in gas turbine engines, the failure behavior of FMs at various elevated temperatures must be understood. In 2000, Trice and co-workers measured the high-temperature flexural strength of Si3N4/BN FMs and monolithic Si3N4 from room temperature through 1400°C [11]. A larger drop in flexural strength occurs in the monolithic sample from room temperature to 1000°C as compared to the FM sample, as shown in Fig. 1.12. The difference in strength trends through 1000°C between the monolithic and FM samples may be attributed to the different amounts of glassy phase in the Si3N4 grain of two materials. They also observed three different failure modes depending on testing temperatures, as shown in Fig. 1.13. As previously mentioned, the apparent stress peak is related to tensile crack initiation in the Si3N4 cell and the subsequent drop is related to crack deflection and delamination in the BN cell boundary. In relatively low-temperature regimes (25°C to 1000°C), tensileinitiated failure occurred on the outer ply of the Si3N4 cells. However, a side view of the flexure sample at 1000°C showed more crack delamination than at 25°C, as shown in Fig. 1.14(a) and (b). This change is probably a result of the changing interfacial fracture energy [10, 11]. In middle-temperature regimes (1100°C to 1300°C), shear-initiated failure occurred at the beam midplane in the BN cell boundary. Following the peak strength and subsequent reloading of the specimen, a distinct slope was observed in the apparent stress versus crosshead deflection curve. The presence of a lengthwise horizontal crack that split part of the beam into halves, characteristic of midplane-initiated
Peak strength (MPa)
1000 6Y/2AI-M
800
600 6Y/2AI-FM 400
200
0 0
200
400
600 800 1000 Temperature (°C)
1200
1400
1.12 Comparison of flexural strengths of Si3N4/BN FM and monolithic specimens with 6 wt% Y2O3 and 2 wt% Al2O3 sintering additives (adapted from ref. [11]).
© Woodhead Publishing Limited, 2006
22
Ceramic matrix composites 600
25°C 600°C 800°C
Apparent stress (MPa)
500 400
1000°C Individual cell breakage 1100°C
300
1200°C
200
1300°C 1400°C
100 0 0
1
2 3 Crosshead deflection (mm)
4
1.13 Typical characteristic apparent stress versus crosshead deflection curves as a function of temperature for Si3N4/BN FM (adapted from ref. [11]).
shear failure, is visible in Fig. 1.14(c). At relatively high temperatures (1400°C), viscoelastic flow of the Si3N4 on the tensile surface took place. The test bar indicated an appreciable bending moment, presumably due to the easy flow of the grain boundary glassy phase present within the Si3N4 cells, as shown in Fig. 1.14(d). Oxidation behavior Another critical factor affecting mechanical properties of Si3N4/BN FMs at elevated temperatures is their oxidation resistance. In 2002, Koh and coworkers studied the oxidation behavior of Si3N4/BN FMs after exposure to air at temperatures ranging from 1000°C to 1400°C for up to 20 h [34]. They observed an overall weight loss due to the extensive vaporization of B2O3 liquid that was formed by oxidation of BN cell boundary. During oxidation, the BN cell boundary began to oxidize at temperatures above 1000oC, forming B2O3 liquid, followed by its vaporization as oxidation proceeded (not shown), and the Si3N4 cell oxidized at above 1200°C, leaving Y 2Si2O 7, SiO 2 (cristobalite) crystals surrounded by a glassy phase (Fig. 1.15(a)). At 1400°C, the Si3N4 cells were also severely damaged, revealing an oxide layer composed of large Y2Si2O7 crystals and a glassy phase, as shown in Fig. 1.15(b). The surface cracks were due to crystallization of a glassy phase on cooling. However, the BN cell boundary layers were completely covered by oxidation products, implying that the oxidation of Si3N4/BN FM is retarded with further oxidation. Koh and colleagues observed that the maximum apparent strength and WOF decreased with increasing exposure temperature (Fig. 1.16(a) and (b)).
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
23
(a)
(b)
(c)
(d)
1.14 Side view of tested flexure Si3N4/BN FM samples at (a) 25°C, (b) 1000°C, (c) 1200°C, and (d) 1400°C (adapted from ref. [11]).
After severe exposure to air at 1400°C for 20 h, the sample maintained 41% (~226 MPa) and 21% (~2.3 kJ/m2) of its initial strength and WOF, respectively. These reductions were apparently due to degradation of both the Si3N4 cell and the BN cell boundary. In other words, the reduction in strength at
© Woodhead Publishing Limited, 2006
24
Ceramic matrix composites (a)
30 µm
(b)
30 µm
1.15 SEM micrographs of the Si3N4/BN FM samples exposed to air for 20 h at (a) 1200°C and (b) 1400°C (adapted from ref. [34]).
temperatures above 1000°C was presumably due to the damage of Si3N4 cells on the surface by the tensile-initiated failure, while the reduction in WOF is primarily influenced by the damage of the BN cell boundary with increasing oxidizing temperatures. Thermal shock resistance Some kind of thermal shock loading is inevitable during service of FMs. In addition, most FMs have anisotropic thermal-expansion coefficients due to their unique architectures. In 2004, Koh and co-workers investigated the thermal shock resistance of Si3N4/BN FMs [29]. They observed their excellent thermal shock resistance by measuring the retention of the flexural strength after thermal shock test, as shown in Fig. 1.17. The monolithic Si3N4 exhibited
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics 12
800 (b) Work-of-fracture
10
600 8 (a) Flexural strength 6
400
4 200
Work-of-fracture (kJ/m2)
Flexural strength (MPa)
25
2
0 0
200
0 400 600 800 1000 1200 1400 1600 Exposure temperature (°C)
1.16 (a) Flexural strength and (b) work-of-fracture of the Si3N4/BN FM sample as a function of the exposure temperature (adapted from ref. [34]).
1000 (a) Monolithic Si3N4
Flexural strength (MPa)
800
600
(b) Fibrous monolith 400
200
0 0
200
400 600 800 1000 1200 Temperature difference (°C)
1400
1.17 Flexural response of (a) monolithic Si3N4 and (b) Si3N4/BN FM after thermal shock test. Monolithic Si3N4 showed brittle fracture, while fibrous monolith showed graceful fracture due to its unique architecture (adapted from ref. [29]).
© Woodhead Publishing Limited, 2006
26
Ceramic matrix composites
the typical thermal shock behavior of brittle ceramics, with a critical temperature (∆Tc = 1000°C) at which the strength decreases catastrophically [35–38], as shown in Fig. 1.17(a). On the other hand, the flexural strength of Si3N4/BN FM after thermal shock was not changed much without ∆Tc, as shown in Fig. 1.17(b), indicating the material’s excellent thermal shock resistance. They observed similar failure modes for FM samples (i.e., shear-initiated cracks), regardless of temperature difference (∆T), as shown in Fig. 1.18 (a)– (d). With increasing ∆T, more extensive crack interactions were observed, increasing the apparent WOF (Fig. 1.19). The increase in WOF after thermal shock suggests that thermal shock reduces the interfacial crack resistance of the cell boundary, which is a composite of boron nitride and glass. Thermal shock damage seems to be absorbed by pre-existing microcracks within the BN cell boundaries [1, 24], which would decrease the cell boundary fracture resistance, enabling easier crack deflection and higher WOF. Koh and colleagues suggested models to explain the thermal shock resistance of Si3N4/BN FMs, as shown in Fig. 1.20(a) and (b). After thermal shock, two thermal stresses are generated in the transverse and longitudinal directions of the sample, due to its unidirectional architecture (Fig. 1.20(a)). Less thermal stress was generated than in monolithic Si3N4. They also found that the calculated crack propagation parameter value of the FM sample is much larger (>16 times) than that of monolithic Si3N4, indicating that crack propagation is more restricted for the Si3N4/BN FM sample. 300
(a) ∆T = 800°C
(b) ∆T = 1000°C
Apparent stress (MPa)
200
100
0 300 (c) ∆T = 1200°C
(d) ∆T = 1400°C
200
100
0 0.0
0.5
1.0
1.5 2.0 0.0 0.5 1.0 Crosshead displacement (mm)
1.5
2.0
1.18 Flexural strength of Si3N4/BN FM ceramic after thermal shock with temperature difference ∆T (adapted from ref. [29]).
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
27
Apparent work-of-fracture (kJ/m2)
14
12
10
8
6
4 0
200
400 600 800 1000 1200 Temperature difference (°C)
1400
1.19 Work-of-fracture (WOF) of Si3N4/BN FM ceramic after thermal shock with temperature difference ∆T (adapted from ref. [29]).
σtrans
σlong
(a) Thermal stress after thermal shock
σtrans ≈ 0.55σm BN-rich cell boundary Cell cracking σlong ≈ 0.80σm
Si3N4 cell BN-rich cell boundary 100 µm
(b) Cracking due to thermal stress
1.20 (a) Thermal stress after thermal shock in transverse and longitudinal directions; (b) a schematic of single Si3N4 cell between two BN-rich cell boundaries, illustrating (i) pre-existing cracks in cell boundaries (- - -) and cell boundary cracks extended by thermal shock (—), (ii) possible transverse cracks in Si3N4 cell (adapted from ref. [29]).
© Woodhead Publishing Limited, 2006
28
Ceramic matrix composites
In addition to these factors, thermal shock-induced cracks and crack propagation during subsequent flexural testing are also important to explain the thermal shock resistance of Si3N4/BN FM. The longitudinal thermal stress may fracture occasional Si3N4 cells, and the transverse thermal stress most likely causes localized extension of the pre-existing flaws in the BNrich cell boundary and is unlikely to cause cracks within the Si3N4 cells, as shown in Fig. 1.20(b). The extension of BN-cell boundary cracks is believed to decrease the cell boundary fracture resistance, which is consistent with the observation of more extensive delamination after flexural testing of severely shocked samples. To clarify this hypothesis, Koh and colleagues measured the degree and length of crack delamination after tensile testing. The degree of delamination cracks significantly increased with increasing temperature difference (∆T). Similarly, a cumulative distribution plot of pullout lengths increased with increasing ∆T. These results support the idea that the fracture resistance decreased through the extension of pre-existing microcracks on BN-rich cell boundaries, promoting the delamination cracks.
1.6
Future trends
In the last 10 years, significant advances in fibrous monolithic ceramics have been achieved. A variety of materials in the form of either oxide or nonoxide ceramic for cell and cell boundary have been investigated [1]. As a result of these efforts, FMs are now commercially available from the ACR* company [28]. These FMs are fabricated by a coextrusion process. In addition, the green fiber composite can then be wound, woven, or braided into the shape of the desired component. The applications of these FMs involve solid hot gas containment tubes, rocket nozzles, body armor plates, and so forth. Such commercialization of FMs itself proves that these ceramic composites are the most promising structural components at elevated temperatures. Nevertheless, much work is still required to fulfill performances of FMs at room and elevated temperatures. For this goal, we address two prospective approaches, from the viewpoints of micro-scale and macro-scale engineering. In micro-scale engineering, we must clearly understand the fundamental mechanisms of crack propagation and energy absorption through FM samples and tailor microstructures of the cell and cell boundary. Although several factors have been found that determine how far the deflected crack travels down in the cell boundary, there are still many unexamined factors affecting mechanical properties of FMs. Such close observations allow for FM designs with optimized properties (i.e., high energy absorption and high strength). Once these are accomplished, we can tailor the microstructures of the cell
*Advanced Ceramics Research, Inc., Tucson, Arizona
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
29
and cell boundary. We may reinforce the cell and cell boundary by incorporating strengthening or toughening phases, in the form of particulate, whisker, platelet, and fiber, into them [39–41]. In addition, a glassy phase may be incorporated into the cell boundary to enhance the densification of a weak cell boundary [42]. These approaches may be more effective in improving mechanical properties of FMs at elevated temperatures. In addition to conventional approaches, emerging nanotechnologies may open new opportunities to design FMs in submicron-scale engineering [43]. For instance, as far as the particle size that serves as a strong phase is concerned, nanosized particles with better control of chemical and physical characteristics can offer potential advantages to decrease the sintering temperature and to lead to new properties. Another prospective adoption of nanotechnology is to utilize micron- or nano-sized fibers, giving an extremely large number of cell boundaries [44]. In macro-scale engineering, we must develop innovative techniques that are capable of producing complex FM architectures without difficulty. To date, coextrusion and hot-pressing are commonly used to manufacture the dense FM parts. This approach may increase manufacturing cost and limit the shape of components. To replace this conventional method, we may adopt solid freeform fabrication (SFF) techniques to create three-dimensional FM architectures. These SFF techniques automatically allow freeform fabrication of parts with complex geometries directly from their CAD models by accumulatively building three-dimensional parts layer by layer [45–47]. For instance, we may employ fused deposition of ceramics (FDC) using a green filament, prepared by conventional coextrusion [47]. To date, only a few manufacturing tools are accessible; hence, there are great opportunities to develop new manufacturing tools. In addition to devolvement of novel manufacturing tools, we must exploit the new way to consolidate FM components by pressureless sintering. In the case of the BN cell boundary, we may incorporate a glassy phase into the cell boundary to enhance the consolidation of BN material. From these viewpoints, there is no doubt that significant advances in FMs can be achieved in the near future and these FMs can be used as structural ceramic components over a broad range of applications.
1.7
References
1. Kovar, D., King, B.H., Trice, R.W., and Halloran, J.W. (1997), ‘Fibrous monolithic ceramics’, J. Am. Ceram. Soc., 80(10) 2471–2487. 2. Coblenz, W.S. (1988), ‘Fibrous monolithic ceramics and method for production’, US Patent 4 772 524. 3. Baskaran, S., Nunn, S.D., Popovic, D., and Halloran, J.W. (1993), ‘Fibrous monolithic ceramics: I, Fabrication, microstructure, and indentation behavior’, J. Am. Ceram. Soc., 76(9) 2209–2216.
© Woodhead Publishing Limited, 2006
30
Ceramic matrix composites
4. Baskaran, S., and Halloran, J.W. (1993), ‘Fibrous monolithic ceramics: II, Flexural strength and fracture behavior of the silicon carbide/graphite system’, J. Am. Ceram. Soc., 76(9) 2217–2224. 5. Baskaran, S., and Halloran, J.W. (1993), ‘SiC-based fibrous monolithic ceramics’, Ceramic Engineering and Science Proceedings, 14(9–10) 813–823. 6. Baskaran, S., and Halloran, J.W. (1994), ‘Fibrous monolithic ceramics: III, Mechanical properties and oxidation behavior of the silicon carbide/boron nitride system’, J. Am. Ceram. Soc., 77(5) 1249–1255. 7. Popovich, D., Halloran, J.W., Hilmas, G.E., Brady, G.A., Somas, S., Bard, A., and Zywicki, G. (1997), ‘Process for preparing textured ceramic composites’, US Patent 5 645 781. 8. Hilmas, G., Brady, A., and Halloran, J.W. (1995), ‘SiC and Si3N4 fibrous monoliths: non-brittle fracture from powder processed ceramics’, Ceram. Trans., 51, 609–614. 9. Hilmas, G., Brady, A., Abdali, U., Zywicki, G., and Halloran, J.W. (1995), ‘Fibrous monoliths: non-brittle fracture from powder processed ceramics’, Mater. Sci. Eng., 195A, 263–268. 10. Trice, R.W., and Halloran, J.W. (1999), ‘Influence of microstructure and temperature on the fracture energy of silicon nitride/boron nitride fibrous monolithic ceramics’, J. Am. Ceram. Soc., 82(9) 2502–2508. 11. Trice, R.W., and Halloran, J.W. (2000), ‘Elevated-temperature mechanical properties of silicon nitride/boron nitride fibrous monolithic ceramics’, J. Am. Ceram. Soc., 83(2) 311–316. 12. Lienard, S.Y., Kovar, D., Moon, R.J., Bowman, K.J., and Halloran, J.W. (2000), ‘Texture development in Si3N4/BN fibrous monolithic ceramics’, J. Mater. Sci., 53, 3365–3371. 13. Tlustochowitz, M., Singh, D., Ellingson, W.A., Goretta, K.C., Rigali N., and Sutaria, M. (2000), ‘Mechanical property characterization of multidirectional Si3N4/BN fibrous monoliths’, Ceram. Trans., 103, 245–254. 14. Singh, D., Cruse, T.A., Hermanson, D.J., Goretta, K.C., Zok, F.W., and McNulty, J.C. (2000), ‘Mechanical response of cross-ply Si3N4/BN fibrous monoliths under uniaxial and biaxial loading’, Ceram. Eng. Sci. Proc., 21(3) 597–604. 15. Koh, Y.H., Kim, H.W., and Kim, H.E. (2004), ‘Mechanical properties of fibrous monolithic Si3N4/BN ceramics with different cell boundary thicknesses’, J. Eur. Ceram. Soc., 24(4) 699–703. 16. Lienard, S.Y., Kovar, D., Moon, R.J., Bowman, K.J., and Halloran, J.W. (2000), ‘Texture development in Si3N4/BN fibrous monolithic ceramics’, J. Mater. Sci., 35, 3365–3371. 17. Harmer, M.P., Chan, H.M., and Miller, G.A. (1992), ‘Unique opportunities for microstructural engineering with duplex and laminar ceramic composites’, J. Am. Ceram. Soc., 75(7) 1715–1728. 18. Evans, A.G. (1990), ‘Perspective on the development of high-toughness ceramics’, J. Am. Ceram. Soc., 73(2) 187–206. 19. Kerans, R.J., and Parthasarathy, T.A. (1999), ‘Crack deflection in ceramic composites and fiber coating design criteria’, Composites, A30, 521–524. 20. Okamura, K. (1995), ‘Ceramic-Matrix Composites (CMC)’, Adv. Comp. Mat., 4(3) 247–259. 21. Cook, J., and Gordon, J.E. (1964), ‘A mechanism for the control of crack propagation in all-brittle systems’, Proc. Roy. Soc. Lond., 282, 508–520.
© Woodhead Publishing Limited, 2006
Fibrous monolithic ceramics
31
22. Clegg, W.J., Kendall, K., Alford, N McN, Button, T.W., and Birchall, J.D. (1990), ‘A simple way to make tough ceramics’, Nature, 357 (4 October) 455–457. 23. Trice, R.W. (1998), The Elevated Temperature Mechanical Properties of Silicon Nitride/Boron Nitride Fibrous Monoliths, PhD Thesis. University of Michigan, Ann Arbor, MI. 24. King, B. (1997), Influence of Architecture on the Mechanical Properties of Fibrous Monolithic Ceramics, PhD Thesis. University of Michigan, Ann Arbor, MI. 25. Koh, Y.H., Kim, H.W., and Kim, H.E. (2002), ‘Mechanical properties of threelayered Si3N4/ fibrous Si3N4-BN monolith’, J. Am. Ceram. Soc., 85(11) 2840–2842. 26. Brady, G.A., Hilmas, G.E., and Halloran, J.W. (1994), ‘Forming textured ceramics by multiple coextrusion’, Ceram. Trans., 51, 297–301. 27. Polzin, B.J., Cruse, T.A., Houston, R.L., Picciolo, J.J., Singh, D., and Goretta, K.C. (2000), ‘Fabrication and characterization of oxide fibrous monoliths produced by coextrusion’, Ceram. Trans., 103, 237–244. 28. Advanced Ceramics Research, 3292 East Hemisphere Loop, Tucson, AZ 857055013, USA. 29. Koh, Y.H., Kim, H.W., Kim, H.E., and Halloran, J.W. (2004), ‘Thermal shock resistance of fibrous monolithic Si3N4/BN ceramics’, J. Eur. Ceram. Soc., 24(8) 2339–2347. 30. Kovar, D., Thouless, M.D., and Halloran, J.W. (1998), ‘Crack deflection and propagation in layered silicon nitride/boron nitride ceramics’, J. Am. Ceram. Soc., 81, 1004–1012. 31. McNulty, J.C., Begley, M.R., and Zok, F.W. (2001), ‘In-plane fracture resistance of a cross-ply fibrous monolith’, J. Am. Ceram. Soc., 84, 367–375. 32. Singh, D., Goretta, K.C., Richardson, J.W. Jr., and de Arellano-López, A. (2002), ‘Interfacial sliding stress in Si3N4/BN fibrous monoliths’, Scripta Mater., 46, 747– 751. 33. He, M.Y., and Hutchinson, J.W. (1989), ‘Crack deflection at an interface between dissimilar elastic materials’, Int. J. Solids Structures, 25(9) 1053–1067. 34. Koh, Y.H., Kim, H.W., Kim, H.E., and Halloran, J.W. (2002), ‘Effect of oxidation on mechanical properties of fibrous monolith Si3N4/BN at elevated temperatures in air’, J. Am. Ceram. Soc., 85(12) 3123–3125. 35. Hasselman, D.P.H. (1969), ‘Unified theory of thermal shock fracture initiation and crack propagation in brittle ceramics’, J. Am. Ceram. Soc., 52, 600–604. 36. Hasselman, D.P.H. (1970), ‘Thermal stress resistance parameters for brittle refractory ceramics: a compendium’, Am. Ceram. Soc. Bull, 49, 1033–1037. 37. Wang, H. and Singh, R.N. (1994), ‘Thermal shock behavior of ceramics and ceramic composites’, Int. Mater. Rev., 39, 228–244. 38. Hirano, T. and Niihara, K. (1996), ‘Thermal shock resistance of Si3N 4/SiC nanocomposites fabricated from amorphous Si–C–N precursor powders’, Mater. Lett., 26, 285–289. 39. Hai, G., Yong, H., and Wang, C.A. (1999), ‘Preparation and properties of fibrous monolithic ceramics by in-situ synthesizing’, J. Mater. Sci., 34(10) 2455–2459. 40. Li, S.Q., Huang, Y., Wang, C.G., Luo, Y.M., Zou, L.H., and Li, C. W. (2001). ‘Creep behavior of SiC whisker-reinforced Si3N4/BN fibrous monolithic ceramics’, J. Eur. Ceram. Soc., 21(6) 841–845. 41. Li, S.Q., Huang, Y., Luo, Y.M., Wang, C.A., and Li, C.W. (2003), ‘Thermal shock behavior of SiC whisker reinforced Si3N4/BN fibrous monolithic ceramics’, Mater Lett., 57(11) 1670–1674. 42. Trice, R.W. and Halloran, J.W. (1999), ‘Investigation of the physical and mechanical
© Woodhead Publishing Limited, 2006
32
43. 44. 45. 46. 47.
Ceramic matrix composites properties of hot-pressed boron nitride/oxide ceramic composites’, J. Am. Ceram. Soc., 82(9) 2563–2565. Sternitzke, M. (1997), ‘Structural ceramic nanocomposites’, J. Eur. Ceram. Soc., 17(9) 1061–1082. Li, D., and Xia, Y.N. (2004), ‘Direct fabrication of composite and ceramic hollow nanofibers by electrospinning’, Nano Letter, 4(5) 933–938. Cawley, J.D. (1999), ‘Solid freeform fabrication of ceramics’, Curr. Op. Solid State Mater. Sci., 4, 483–489. Halloran, J.W. (1999), ‘Freeform fabrication of ceramics’, Br. Ceram. Proc., 59, 17– 28. Rangarajan, S., Qi, G., Venkataraman, N., Safari, A., and Danforth, S.C. (2000), ‘Powder processing, rheology, and mechanical properties of feedstock for fused deposition of Si3N4 ceramics’, J. Am. Ceram. Soc., 83(7) 1663–1669.
© Woodhead Publishing Limited, 2006
2 Whisker-reinforced silicon nitride ceramics M D P U G H, Concordia University, Canada and M B R O C H U, McGill University, Canada
2.1
Introduction
Silicon nitride as a monolithic ceramic has earned a place in the inventory of commercial high-performance ceramics with excellent performance in applications such as ball bearings, seals, cam followers, and molten metal handling. However, like most other monolithic advanced ceramics, Si3N4 has suffered from an inherent brittleness which limits its application where some toughness or impact resistance is required. In response to this, work began in the mid-1980s to use whisker reinforcement to increase the toughness of ceramics such as alumina and silicon nitride. The work was mainly centred around two main whisker types – SiC and Si3N4 – until the development of ‘in-situ whiskers’, or ‘self-reinforcing’ Si3N4 wherein β-Si3N4 seed crystals could be grown into whisker-like grains during sintering. The principal work has since been on self-reinforced Si3N4 and SiC(w)–Si3N4 composites. An appreciable increase in toughness has been accomplished through the use of whisker (or elongated grain) reinforcement, and K1C values of 10 MPa.m1/2 are commonly reported. This chapter will summarize the materials and fabrication processes involved in producing these materials and then go on to discuss the ensuing properties and applications. Due to various factors including processing difficulties encountered with these materials and socioeconomic factors, commercialization of Si3N4 composites has been somewhat limited, but SiC(w)–Si3N4 composites are finding applications in cutting tools and self-reinforced Si3N4 has become a popular replacement for conventional silicon nitride. Work has been continuing in this area and optimization of processing routes to produce a viable commercial product (physically and economically) is still continuing. These routes are required in order to bring the enhanced properties of silicon nitride composites to the relevant industries with near-net shape capabilities and acceptable costs. In addition to ongoing work into the properties and behaviour of these materials, the future outlook is for more emphasis into this aspect of ‘manufacturability’ and cost. 33 © Woodhead Publishing Limited, 2006
34
2.2
Ceramic matrix composites
Fabrication
There has been considerable progress in the processing of monolithic engineering ceramics in the last 25 years, especially in the field of sintering, microstructural control, and grain boundary refinement of silicon nitridebased ceramics. This has led to the development of gas pressure and even pressureless sintering of silicon nitride components having excellent characteristics (except in the area of toughness) from green precursors prepared by a variety of techniques (die-pressing, injection moulding, slip-casting, etc.). As with most engineering ceramics, the primary fabrication route relies on production of a green compact from Si3N4 powder (typically 40–60% dense) which is then densified by a high-temperature process such as pressure/ pressureless sintering, hot-pressing, and hot isostatic pressing (HIPing). This densification ipso facto results in shrinkage of the compact which produces the major fabrication challenge for whisker-reinforced silicon nitride composites. This section will describe the commonly used starting materials and fabrication routes for processing whisker-reinforced silicon nitride composites.
2.2.1
Materials
Silicon nitride matrix The materials used for whisker-reinforced silicon nitride ceramics depend on two major factors: firstly the chosen processing route for fabricating the composite (hot-pressing, sintering, reaction-bonding, etc. – see Section 2.2.2) and secondly the properties required of the whiskers. If the silicon nitride matrix is to be formed by a sintering mechanism (hot-pressing, gas-pressure and pressureless sintering) then the starting materials will include firstly silicon nitride powder, usually with a sub-micron size (giving a high surface area and hence driving force to promote sintering) and with an α-phase crystal structure to benefit from the α → β transformation that occurs during sintering, and secondly, sintering additives. These sintering additives are required to provide a liquid phase at high temperature through which a dissolution–transport–precipitation mechanism can occur, leading to densification and the formation of a solid structure on cooling (liquid phase sintering). Unlike ceramics such as alumina and silicon carbide, silicon nitride powders do not densify easily through solid state mechanisms. A range of materials have been tried as sintering aids for silicon nitride, including magnesia, yttria, beryllia, calcium oxide and some rare earths, but common formulations are now centred around additions of yttria and alumina totalling 10–15 wt% for sintering and lesser amounts for hot-pressing. Magnesia at levels of 4–5 wt% is used for hot-pressing as lower temperatures can be used, but this will
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
35
also limit the maximum use temperature of the material. Unless specifically tailored to crystallize, this liquid phase forms a glassy intergranular phase on cooling which will reduce the creep resistance of the material. Whiskers The whiskers that are commonly added to silicon nitride to form composites are Si3N4 and SiC (common nomenclature is to add the suffix (w) to denote whiskers and this will be used where appropriate). Although other ceramic and metallic whiskers are available (e.g. BN, TaC, TiC, B4C and Fe), Si3N4 and SiC each have properties that make them prime candidates as reinforcements, not least of which is that they can sustain the high temperatures and reactivity of the sintering process without being degraded. Silicon nitride whiskers are added to improve toughness through various mechanisms as described in Section 2.3, without introducing problems with regard to mismatch of coefficients of thermal expansion (CTE) and Young’s moduli. Since the discovery and development of self-reinforced silicon nitride in which large, elongated grains can be grown in Si3N4 during sintering to produce an in situ, whisker-like reinforcement, ‘conventional’ silicon nitride whiskers have not been utilized extensively as reinforcements per se but are replaced with rod-like, β-Si3N4 seed crystals – which are, in essence, very short whiskers.1,2 The principal whiskers used for toughening of silicon nitride are those based on SiC. Silicon carbide whiskers show an excellent combination of properties including strength, stiffness and high temperature stability, and they are also amenable to a variety of surface treatments, such as oxidation, which are used to augment toughening mechanisms including debonding and pullout. Silicon carbide whiskers are available in a variety of sizes, chemistries and surface conditions depending on their fabrication route. The dimensions of these whiskers can vary but are typically of the order of 0.1– 1 µm in diameter and 5–100 µm in length. The aspect ratio of whiskers ranges from ≈5 to ≈50, but in many cases the mixing/milling processes required to separate whisker flocs and uniformly disperse the whiskers in the powder mixture can result in a reduction of whisker lengths and hence aspect ratio. This may not necessarily be a bad thing, however, since a reduced aspect ratio has been shown to result in improved densification, especially in pressureless and gas-pressure sintering.3
2.2.2
Fabrication
Green fabrication and sintering of compacts containing second phases poses several problems and when the second phase is the form of whiskers these problems are magnified. These problems include inhomogeneous distribution
© Woodhead Publishing Limited, 2006
36
Ceramic matrix composites
of the second phase, thermal expansion mismatch on cooling from sintering, and most importantly whisker impingement during sintering. This last factor presents a major obstacle in the fabrication of Si3N4 composites using sintering (gas pressure or pressureless). In order to overcome this problem there are currently three main approaches: application of high external pressure (hotpressing / hot isostatic pressing (HIPing)), orientation of whiskers such that bridging is minimal (extrusion, tape casting, etc.) and finally reaction-bonding of silicon in which no shrinkage occurs. The first is the most commonly used in research as high uniaxial pressure can provide ~98% or more of theoretical density.4 The major drawbacks of this method are the limited geometry and the preferred alignment of whiskers perpendicular to the pressing direction. This imposes limitations on the versatility and application of material processed this way. Generally, whiskers (up to 30 wt%) and powders (Si3N4 plus sintering additives) are mixed or milled, dry or wet: if wet, the liquid is usually a solvent such as ethanol which is subsequently removed by drying. Hot-pressing conditions depend on the additives used but typically involve temperatures from 1650 to 1800°C and pressures of ≈30 MPa in graphite dies surrounded by a nitrogen or argon atmosphere.5 HIPing is considerably more complex and expensive than hot-pressing but for monolithic silicon nitride it normally produces a superior product with near-isotropic properties at low additive contents. However, with silicon nitride whisker composites, whisker bridging and the formation of whisker ‘nests’ can be more of a problem.5 In conventional (pressureless) sintering of whisker composites in an inert or nitrogen atmosphere as well as in gas-pressure sintering (≈1 MPa of nitrogen overpressure), densification can be hampered by impingement of whiskers to form a skeletal network which can inhibit further shrinkage and thus result in a lower sintered density and higher percentage porosity.6 Typically for more than 20 wt% whiskers the final density is less than 90% of theoretical when processed this way, which leads to a reduction in many mechanical properties such as strength, and changes to some physical properties such as thermal conductivity. In order to alleviate this problem several workers have attempted to orient the whiskers in the green compact by one means or another prior to sintering. These methods include extrusion of a powder/whisker/binder mixture, slipcasting and tape-casting of slips.7–9 These techniques, although giving some improvement, are generally limited to about 95% theoretical density (TD) after sintering, and 15–20 wt% whiskers. For higher whisker contents there is a significant decrease in TD and flexural strength. Once again, these processes result in an anisotropic structure. The only method currently available that avoids the whole problem of whisker impingement is reaction bonding of silicon powder to form the
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
37
silicon nitride matrix (RBSN).10 In this process, a compact containing silicon powder is reacted in a nitrogen atmosphere at temperatures around the melting point of silicon (1300–1400°C) for extended times (≈20 hours). Because this nitridation generally occurs by a controlled vapour phase reaction, the increase in volume of solid material (Si → Si3N4) is accommodated in the pore space of the compact, giving a final component that has no net shape change, i.e. no shrinkage (unlike sintering or hot-pressing), but does result in a residual porosity of the order of 23%. For the manufacture of composites, whiskers are mixed with the silicon powder and the nitridation takes place in the usual way, although the kinetics may be changed, thus allowing formation of a composite free from the problems of whisker impingement.11 Because the nitridation produces the silicon nitride matrix in situ, there is no requirement for liquid-forming sintering aids and hence no residual glassy phase unless a secondary sintering step is to be performed to reduce the residual porosity, which then reintroduces the problem of shrinkage, albeit of a smaller magnitude.12 As the RBSN composite consists of Si3N4 directly bonded to itself and the whiskers, this material is less prone to creep at elevated temperatures but may be more susceptible to oxidation due to the higher porosity levels. As the silicon powder compact can be formed by conventional powder processing routes (pressing, extrusion, casting), there is a choice in whether or not the whisker reinforcement is oriented, with no effect on the reaction bonding process. Purely from a processing viewpoint, one can see that hot-pressing is very straightforward and produces a good but anisotropic product, though from a manufacturing perspective this process is the least accommodating in terms of shaping. The most versatile would be the reaction bonding route except for its Achilles’ heel – its high residual porosity.
2.3
Properties
One of the main strategies to improve the performance of monolithic silicon nitride ceramic is the addition of whiskers, leading to crack bridging and crack deflection. As mentioned earlier, the whiskers most commonly used to reinforce monolithic Si3N4 are SiC and Si3N4, but more recently β-Si3N4 seeds are now used to promote in situ growth of whisker-like crystals in place of Si3N4 whiskers. The control of the microstructure–properties relationship leads to the design of materials with enhanced properties. In addition to the type of whisker reinforcement, the microstructure is also modified by the volume fraction of whiskers, their orientations and respective properties and by interfacial reactions. This section describes the different microstructures typically obtained for Si3N4 reinforced with Si3N4(w) or SiC(w) and the resulting changes in physical and mechanical properties.
© Woodhead Publishing Limited, 2006
38
2.3.1
Ceramic matrix composites
Microstructure
The main mechanism reported to improve the toughness of whisker-reinforced Si3N4 is based on crack-bridging which depends on the orientation of the fibres to the propagating crack. Components possessing whiskers oriented perpendicular to the crack propagation direction will exhibit an increase in fracture strength and creep resistance at high temperature. Obviously, a reduction of these properties is observed when the whisker orientation is parallel to the crack propagation direction. Therefore, the orientation of the reinforcement will dictate whether the composite has isotropic (random orientation) or anisotropic (aligned orientation) properties. SiC(w) reinforced Si3N4 The introduction of silicon carbide whiskers in a silicon nitride matrix is commonly performed by mixing and densification as described in Section 2.2.2. As mentioned, the sintering of silicon nitride requires liquid-forming additives. In SiC(w)–Si3N4 composites, the reaction and wettability of the whiskers by the sintering additives used become important parameters, as the debonding characteristics of this interface have a strong influence on mechanical properties. Fortunately, like Si3N4, SiC possesses a native SiO2 surface layer, which is a common component of the liquid-phase sintering process for silicon nitride. Studies with different sintering additives have been performed to identify the optimum composition. Dogan and Hawk13 have detected with TEM and XRD two scenarios regarding the crystalline state of the sintering additive: (1) crystalline Y2Si3N4O3 (millilite), β-Y2Si2O7 and an amorphous layer bridging the grains and the crystalline phase, or (2) a probably completely amorphous yttrium–aluminium silicate. In both cases, the sintering additives wet both the matrix and the whiskers. Unfortunately, insufficient experimental details are available yet to try to demystify the wetting mechanism. Highresolution TEM of the interface for monolithic Si3N4 and Si3N4–20%SiC(w) with Y2O3 as the sintering additive has indicated a grain boundary phase of α-Y2Si2O7 with a thin, continuous and amorphous film at the interface for the monolithic silicon nitride. However, the addition of SiC(w) modifies the chemical composition of the sintering additive, and a glassy phase at the interface containing a high concentration in SiO2, presumably coming from the surface oxide film of the whiskers, is observed. The consolidation mechanism of many of these ceramic composites involves a thin glassy phase linking the matrix grains, the reinforcement and the crystalline portion of the sintering additive.14 The thermal and mechanical properties of all these phases are quite different, which will therefore influence the behaviour of the composite. Although the coefficient of thermal expansion
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
39
(CTE) of the crystalline phase would change considerably depending on the initial sintering additive added, the Y2Si3N4O3 is a member of the gehlenite family, which possesses a CTE of around 1–2 × 10–6/°C15 as opposed to 5 × 10–6/oC for the amorphous phase.16 Knowing that the CTE of Si3N4 is around 3.2 × 10–6/oC and that of SiC is 4.5 × 10–6/oC, the volume fraction and composition of the different phases present will lead to complex interfacial residual stress patterns. For example, the presence of a crystalline phase with such a low CTE will lead to compressive residual stresses at room temperature for the crystalline boundary phase. The opposite will be observed for the amorphous boundary phase where tensile residual stresses will be present. This change in residual stress will influence the mechanical properties, as one scenario will promote better toughening and the other better strength. The morphology of the added whiskers also plays an important role during processing as does the actual mixing technique. Whiskers with high aspect ratio can have a deleterious effect on green density consolidation and sinterability, decreasing the fracture toughness of the composite.3 Another important modification of the processing parameters by the presence of SiC(w) is that of the kinetics of the α → β transformation. This transformation is well known in monolithic Si3N4 ceramics to improve the fracture toughness by the development of elongated β-grains. The presence of SiC(w) retards the α → β phase transformation, thus reducing this beneficial effect. Si3N4(w) reinforced Si3N4 As mentioned earlier, this type of microstructure can be obtained by adding silicon nitride whiskers to the Si3N4 powder as for SiC whisker composites or more commonly by the growth of large and elongated β-grains from rodlike β-seed crystals in an otherwise equiaxed β-Si3N4 matrix during sintering. These high-aspect-ratio grains act similarly to whiskers. Therefore, control of the processing parameters (mixture and composition of the raw materials, hot-pressing vs. liquid phase sintering, sintering additives, etc.) directly affects the composite microstructure and therefore its response to service stresses. The growth of the elongated β-grains is anisotropic; a higher growth rate occurs along the c-axis of the grain and is also influenced by the presence and composition of the sintering additive. The control of the orientation of the resulting elongated grain is performed by orienting the initial β-grain seeds, which grow at the expense of the small matrix grains. Therefore, their initial alignment leads to an oriented microstructure, and the most favourable processes to produce an oriented seed structure are tape casting and extrusion. The growth of the Si3N4 whiskers in an oriented green body is such that they have a lower probability of impeding each other, resulting in a microstructure with a reinforcement phase possessing a higher aspect ratio than for the randomly oriented seed mixtures.
© Woodhead Publishing Limited, 2006
40
Ceramic matrix composites
Lee et al.17 have reported that the growth of elongated β-Si3N4 whiskers is a function of the initial starting powder size and occurs by the dissolution of surrounding grains. Whiskers with a higher aspect ratio are obtained when finer seeds are used, and a critical particle size is important as 2-D nucleation and growth must prevail, since larger particles will promote the growth of a more equiaxed microstructure. Obviously, the sintering parameters modify this behaviour; whisker-containing microstructures will be obtained at higher temperatures when larger seed particle sizes are used. Control of the microstructure (in this case, primarily the volume fraction of whiskers) is usually achieved through the initial ratio of α/β. As mentioned earlier, the addition of β-seeds in an α-powder produces the elongated βgrains in a fine β-matrix (subsequent to the α → β transformation) accompanied by a residual glassy phase. In addition to growth, these β-seeds act as nucleating agents for the α → β transformation. However, the existence of a critical amount of nuclei, 10 wt%, was proposed by Emoto and Mitomo18 above which a reduction of the growth driving force occurs. Therefore, the generation of a distinct bimodal microstructure, especially one with large, elongated grains, well dispersed in a fine, submicron-grain-size matrix, is an important factor in optimizing the fracture resistance and the fracture strength of the material. However, the presence of the whisker reinforcement is not the only factor affecting the response to mechanical stresses. Silicon nitride is commonly produced through liquid phase sintering, and therefore a residual quantity of sintering additive phases (crystalline or amorphous) will remain at the grain boundaries, and their respective volume fractions will modify the toughening mechanisms. In addition, the chemical composition of the sintering additive will influence the grain growth rate during processing, and different grain morphologies and sizes can be obtained. The use of a sintering additive containing a high concentration in alumina and nitrogen has been shown to favour the development of a distinct epitaxial β′-SiAlON layer at the interface between the Si3N4 whisker and the oxynitride glass.19 Interfacial debonding was found to vary with the glass composition, where for low yttrium–aluminium ratios, the β′-SiAlON layer grew on the Si3N4 whiskers and the β′-SiAlON/oxynitride glass interface was reported to be more resistant to debonding as the chemical gradients are more gradual. Therefore, the mechanical properties of the composite are strongly influenced by the z value (concentration of substituted Al and O) of the SiAlON near the β′-SiAlON/glass interface. Experiments have shown that for sintering additives with lower z value, an improvement of toughness of 30% was obtained.19 The absence of the β′-SiAlON layer was found to improve the fracture resistance of the composite by debonding, crack deflection and whisker pullout.20 However, the concentration in Al in the β′-SiAlON layer is controlled not only by the sintering additive but also by the extent of grain growth. The
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
41
processing conditions used to establish the interface, such as temperature and time, will modify the interfacial chemistry by the control of grain growth. Therefore, the use of Al2O3 as sintering additive should not be proscribed, as alumina plays an important role in the densification and a low quantity will enable a limited growth of β′-SiAlON layer, which will still have low interfacial strength and promote the targeted debonding behaviour that leads to improvement in toughness. As opposed to composites reinforced with SiC(w), the residual stresses formed at the interface should be lower as the CTE of the matrix and the reinforcement phase are similar. Several modeling studies have shown that the composition of the grain/glass interface has a negligible effect on the magnitude of the residual stresses and therefore has no significant influence on the fracture behaviour.19
2.3.2
Improved properties
In whisker-reinforced ceramic composites, the modification of the mechanical properties is dictated by the strength at the grain/grain boundary phase interfaces as discussed above. This alteration of the mechanical properties compared to the monolithic Si3N4 is observed on a macro-scale through several mechanisms: whisker debonding, whisker pullout, crack bridging and/or crack deflection. However, the common idea in these three toughening mechanisms is to increase the energy needed to extend a crack. A brief description of these mechanisms will be presented to help understand their impact on mechanical properties. The concept of crack deflection arises from the lower-strength grain boundary present in polycrystalline materials and is governed by the whisker shape and volume fraction, although elastic moduli and thermal expansion coefficient have a role. A propagating intergranular crack will have a tendency to follow the easiest path and therefore the crack orientation may change as the crack propagates. This change in orientation reduces the average stress intensity at the crack tip, because the crack propagation plane is no longer normal to the applied tensile stress. This toughening mechanism is well accepted by the community for explaining the lower toughness of single crystal versus polycrystalline ceramics. Crack bridging implies the mechanical support by a second phase to help reduce the opening of a propagating crack. The second phase acts as a ligament behind the crack front, reducing the stress intensity factor. The stress supported by the ligament increases slowly with distance behind the crack tip, and greater crack-opening displacements are achieved in the bridging zone, enhancing the fracture toughness. Crack bridging is often enhanced by increasing the volume of the reinforcement phase, the ratio of the elastic modulus of the ceramic over that of the fibre, Ec/Ef, and the ratio of the
© Woodhead Publishing Limited, 2006
42
Ceramic matrix composites
fracture energy of the bridging ligament to that of the reinforcement/matrix interface. Crack bridging is also complemented by the contribution of pullout of the failed whisker reinforcements. The pullout operation consumes energy which would otherwise contribute to the advancement of the crack front and thus enhances toughness. However, these strengthening mechanisms are not the answer to all the problems. The improvement of the mechanical properties by the addition of whiskers greatly influences the long-crack toughness but has little or no influence on the short-crack toughness. In the design of a component from whisker-reinforced ceramics, the geometry and size of the final part influences the desired microstructure. Because of the difficulties in processing homogeneous, large-scale ceramic components, long-crack toughness is desirable in this case as it increases the tolerance to flaws (e.g. microstructural heterogeneities, such as weak grain boundaries, large grains and residual stresses) and improves the service life. The toughness of this composite increases with the crack length due to internal friction resistance. On the other hand, the design of small-sized ceramic composites requires a more homogeneous microstructure – the fracture toughness is dictated by the heterogeneity in the microstructure, which should be reduced as much as possible in small pieces. In this case, the toughness becomes more independent of the microstructure as crack bridging due to internal friction resistance is less efficient. Therefore, more optimization of the microstructure must be done to improve the resistance to short-crack propagation. The following sections summarize the improvements of physical and mechanical properties obtained for whisker-reinforced silicon nitride composites to date. Physical properties Coefficient of thermal expansion The coefficient of thermal expansion (CTE) of composite materials usually follows the simple rule of mixtures (or more complex models), based on the CTE of the respective components, their volume fraction and the volume fraction of interfacial phases. Based on these models, a Si3N4–Si3N4(w) composite should possess a similar CTE to monolithic Si3N4 ceramic (3.2 × 10–6/oC). Obviously, the chemical composition of the sintering additive will have a certain influence but should remain within the variations observed for monolithic Si3N4. However, in the case of SiC(w), the CTE of the reinforcement phase (SiC: 4.5 × 10–6/oC) is higher than that of the composite matrix, implying an increase in the composite CTE when whiskers are added to silicon nitride. Jia et al.21 reported that the CTE of 30 vol% SiC(w)–Si3N4 reaches nearly 7 × 10–6/oC,
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
43
which is more than the double the value of monolithic Si3N4. However, their measurements were performed on samples with densities varying between 86–91% and with residual α-phase levels reaching up to 10%, both of which can affect the CTE values. Thermal conductivity The thermal conductivity of a Si3N4 composite is controlled by the level of porosity, the β-phase content, the amount and conductivity of the reinforcement phase (Si3N4 vs. SiC), the amount of glassy phase and any orientation effect. In Si3N4–Si3N4(w), the thermal conductivity will increase with the density and with the amount of β-phase, as it possesses a higher thermal conductivity than α-phase. In addition, orientation of the whiskers produces an anisotropic thermal conductivity behaviour, where a higher thermal conductivity, up to 1.5 times higher in fact, is observed in the direction of orientation compared with the direction perpendicular to the grain orientation.22 Therefore, long βgrains with fewer triple junctions are favourable for conductivity. Finally, the volume fraction of the glassy phase influences conductivity, as the sintering additives typically used possess a lower thermal conductivity than the matrix and act as a thermal barrier. The prediction of the thermal conductivity is not as simple for the composite reinforced with SiC whiskers. In fact, as mentioned earlier, the thermal conductivity of the reinforcement phase plays an important role. However, for SiC whiskers, their chemical composition can vary drastically depending on the manufacturing process, which is not the case for Si3N4 whiskers. In fact, the influence of the manufacturing process is drastic: the thermal conductivity of SiC(w) produced by the vapour–solid process is around 20 W/m K as opposed to 100–250 W/m K for whiskers produced by the vapour–liquid– solid process. Using a Si3N4 matrix with a nominal thermal conductivity of 48.5 W/m K, the thermal conductivity of the Si3N4–SiC(w) composite can increase or decrease depending on the type of SiC whiskers added.23 Mechanical properties Flexural strength The targeted properties in Si3N4 composites are high strength and fracture toughness for high-temperature applications. Figure 2.1 presents four-point flexural strength values for hot-pressed Si3N4–Si3N4(w) composites with different whisker volume fractions as a function of temperature (MgO as sintering aid).24 The strength for a monolithic Si3N4 is added for comparison purposes. The results demonstrate that the addition of whiskers improves the flexural strength at low and intermediate temperature (approximately 1000oC)
© Woodhead Publishing Limited, 2006
44
Ceramic matrix composites
Flexural strength (MPa)
1000
800
600
400
200
0% Si3N4w 20% Si3N4w
5% Si3N4w 35% Si3N4w
0 0
300
600 900 Temperature (°C)
1200
1500
2.1 Flexural strengths as a function of temperature for silicon nitride composites.24
but becomes less significant at higher temperature (1400oC). The addition of some whiskers can improve strength properties but additions over a critical amount will have a deleterious effect on the microstructure. The presence of more than 15 vol% of Si3N4 whiskers reduces the elongated grain growth of the Si3N4 matrix, thus reducing the desired effect of adding whisker.25 This can be observed by the similarity in the curves for the composites possessing 0, 20 and 35 vol% whiskers. By increasing the test temperature, the strengthening mechanism of the composite changes, passing from crack bridging and whisker pullout to a softening of the grain boundary and stress relaxation at high temperature, resulting in easier whisker pullout. The change in failure mechanism is also apparent by the change in the fracture surface, where round, smoothed surfaces and grain boundary cavities are observed for samples tested at 1400oC. As mentioned in Section 2.2 the use of MgO as sintering additive is known to reduce the high-temperature strength of the composites compared to Y2O3–Al2O3 additives. Therefore, higher flexural strength at higher temperature can be expected for composites sintered with yttria–alumina mixtures. The influences of the Si3N4 matrix grain size distribution and Si3N4(w) aspect ratio on the flexural strength at room temperature have been reported by Becher et al.26 and are presented in Table 2.1. The results show that improved strength is obtained for a distinct bimodal grain size distribution and that the characteristics of the raw materials are very important. In such cases, the mechanical behaviour of the composite is getting closer to the optimum microstructure for crack bridging and whisker pullout. Another factor that appears to affect strength is the ratio of yttria to alumina in the sintering additive which changes the phases present at the interface; the presence of a β′-SiAlON layer reduces the flexural strength.19
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
45
Table 2.1 Fracture strength as a function of grain size and distribution26 Average grain size
Type of distribution
Fracture strength (MPa)
0.4 µm 1 µm 0.2 µm/0.5 µm 0.3 µm/2 µm
Monomodal Broad Bimodal Bimodal
660 850 925 1144
± ± ± ±
165 75 75 126
As opposed to the processing of Si3N4–Si3N4(w) composites, where the majority of whisker formation occurs during sintering by growth of β-Si3N4 seeds, the processing of composites reinforced with SiC(w) is harder due to the impingement of the whiskers, which can lead to lower final densities that in turn also affect the composite properties. Figure 2.2 presents a summary of the room-temperature flexural strength of some Si3N4–SiC(w) composites. For hot-pressed composites, it appears that the presence of SiC(w) reduces the strength and there is no sign of an optimum volume fraction of reinforcement, as in Si3N4–Si3N4(w). In these two cases, the sintering additives were similar (a mixture of yttria and alumina) but their ratio differs. The flexural strength results correlate with the final density and with the chemical phases present at the grain boundaries. As noted earlier, improved flexural strength is obtained with sintering additives comprising a high ratio of Y2O3 to Al2O3 combined with high density. For composites produced by a combined reaction-bonding/sintering route where there is less shrinkage, higher flexural strengths are obtained for 5% compared to 15% SiC(w).27 High-temperature four-point bending tests of Si3N4 and Si3N4/20%SiC(w) composites prepared with no external sintering aids by hot isostatic pressing 1000
Flexural strength (MPa)
900 800 700 600 500 400 300
Process
200
Ratio Y2O3/Al2O3
Reaction bonded – hot-pressed Hot-pressed Hot-pressed
100
8 1.5 2.66
Ref [27] [21] [28]
0 0
5
10 15 20 Volume fraction SiC(w) (%)
25
2.2 Flexural strength as a function of SiC whisker content.
© Woodhead Publishing Limited, 2006
30
46
Ceramic matrix composites
show no reduction in strength between room temperature and 1400oC for both the monolithic ceramic and the composite, demonstrating the importance of the additives and grain boundary softening on the high-temperature failure mechanism.29 However, the strength values were limited to ≈500 MPa, which is half the room temperature strength obtained when sintering additives are used in the processing. Young’s modulus It is well known that the Young’s modulus of a composite can be calculated by the rule of mixtures for long-fibre reinforced material. In the case of whiskers, the rule of mixture is also applied to estimate the change of modulus (conventionally, reinforcements are added to improve the stiffness of a material, though for ceramic matrix composites this is not always the primary concern). For Si3N4–Si3N4(w) composites one would not expect any significant change in elastic modulus; however, as 100% solid density is rarely achieved, there may be some decrease depending on the level of porosity present. For SiC(w) reinforced composites, due to the higher Young’s modulus of the SiC whiskers (≈420 GPa) compared to monolithic silicon nitride (320 GPa), an increase in stiffness should be observed, and for Si3N4–20vol%SiC(w) composites produced by extrusion, reaction bonding and hot isostatic pressing, the highest value observed was 350 GPa which is actually higher than that predicted by the rule of mixtures.14 The reason for this is most likely in the estimation of the elastic modulus of the SiC whiskers. Toughness There are two main techniques used to measure the fracture toughness of ceramics: fracture stress and hardness indentation. The former measures the load to fracture of a pre-cracked specimen using a single edge notched beam (SENB) or a chevron notched beam (CNB) sample. The main drawback of this technique is ensuring that the crack tip is atomically sharp. The second method uses the crack formed at the corners of the indentation produced during a Vickers indentation hardness test. This technique is rapid and relatively inexpensive. However, the toughness values measured are those of the surface, unlike the values obtained by fracture of the pre-cracked beams which are a measure of the bulk material properties. The main tool for comparing the toughness of monolithic and whiskerreinforced ceramics is through R-curve behaviour. As mentioned previously, the presence of whiskers can result in reduced crack opening and therefore the toughness becomes dependent on the propagating crack length, as opposed to the typical Griffith theory in which the initial crack or flaw is assumed to propagate instantaneously and completely. Figure 2.3 presents a schematic
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
47
c
c
a2
a1
(b)
(a)
R-curve behaviour Strength
Toughness
R-curve behaviour
No R-curve behaviour No R-curve behaviour Flaw size (c)
Flaw size (d)
2.3 Schematic representation of cracking in (a) a monolithic ceramic and (b) a ceramic composite reinforced with whiskers (a1 > a2); (c) influence of R-curve behaviour on fracture toughess, and (d) influence on strength, in ceramics and composites.
representation of a propagating crack in (a) a monolithic ceramic, and (b) a ceramic composite reinforced with whiskers, and also a representative view of the influence of R-curve behaviour on (c) fracture toughness and (d) strength. In essence, ceramics that exhibit R-curve behaviour are more tolerant to the presence of flaws than those which do not. Si3N4-Si3N4(w) composites As opposed to composites reinforced with SiC(w), the formation of the elongated microstructure in the Si3N4–Si3N4(w) system is occurring in situ, during the consolidation operation. Several researchers have reported that the main factor affecting the toughness of these composite is the volume fraction of reinforcement, but the presence of a bimodal grain distribution was also found to have a major role in the toughening mechanism. Figure 2.4 nicely summarizes the microstructure evolution vs. toughening relation occurring for seeded and non-seeded silicon nitride composites.30 Monolithic Si3N4 with a fine equiaxed matrix shows a modest R-curve response with a steady state toughness of only 3.5 MPa.m1/2. Longer hot-pressing times or higher pressing temperatures will enhance the formation of elongated grains and will be accompanied by an increase in R-curve behaviour. However,
© Woodhead Publishing Limited, 2006
48
Ceramic matrix composites 12
Fracture resistance (MPa.m1/2)
Gas pressured sintered – seeded 10 8 Hot-pressed 2 h; unseeded 6 Hot-pressed 0.33 h; unseeded 4
Hot-pressed
2 0 0
300
600 900 Crack length (µm)
1200
1500
2.4 Effect of processing conditions on R-curve behaviour for in-situ Si3N4 composites.26
materials with a broad grain diameter distribution and with a large fraction of elongated grains exhibit only a modest increase in toughness combined with a lower strength. The control of the microstructure by the addition of βseeds shows that a major improvement in the R-curve behaviour can be achieved and fracture toughness values of 10 MPa.m1/2 have been reported.26 The microstructure developed in such cases comprises large elongated grains embedded in a fine matrix. From a microstructural aspect, increasing the size of the reinforcing grains will enhance contributions from both frictional bridging and pullout. In addition, enhanced interfacial debonding, number of bridging reinforcements, and length of the debonded interfaces should all result in a more rapid rise in the R-curve, and therefore the toughness. However, the control of the initial composition is very important, as an overabundance of seeds will lead to grain impingement and will reduce their growth and consequently inhibit the formation of the bimodal microstructure. A surplus of seeds in conjunction with poor processing will favour the formation of clusters of large grains, which are known to reduce the strength and toughness. As mentioned earlier, the toughness of the material is a function of the orientation of the whiskers, where isotropic properties will be obtained for randomly oriented whiskers and anisotropic properties for oriented microstructures. The influence of whisker orientation as well as the ratio of yttria to alumina on the toughness of Si3N4–Si3N4(w) composites as measured by Sun et al.20 are presented in Table 2.2. In all cases, the fracture toughness of the composite is higher in the direction normal to the whiskers, which confirms that more energy is required for a crack to propagate through whiskers by crack deflection and crack bridging than it is along the whisker/
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
49
Table 2.2 Dependence of fracture toughness on grain orientation as a function of sintering additive20 Sintering additive
Toughness perpendicular to whisker direction (MPa)
Toughness parallel to whisker direction (MPa)
4.0Y2O3–2.8Al2O3 5.0Y2O3–2.0Al2O3 6.25Y2O3–1.0Al2O3
7.5 ± 0.2 8.7 ± 0.4 10.6 ± 0.2
6.9 ± 0.2 7.8 ± 0.3 9.2 ± 0.2
Table 2.3 Variation in toughness as a function of sintering time32 Sinterig time (min)
10
45
90
180
540
Toughness (MPa.m1/2)
6.7 ± 0.2
7.6 ± 0.2
7.9 ± 0.3
8.1 ± 0.3
6.9 ± 0.2
matrix interface. As discussed earlier in the microstructure section, the concentration and chemistry of the sintering agent can change the interfacial reaction, and for low Y2O3 to Al2O3 ratios, a β′-SiAlON/oxynitride glass layer is formed, giving a much stronger matrix/whisker interface, leading to a reduction in the efficiency of the toughening mechanisms. Table 2.3 presents the variation in toughness of gas-pressured, sintered, seeded Si3N4–Si3N4(w) as a function of sintering time. The results clearly demonstrate that the toughness evolves with the microstructure. The kinetics of formation of the elongated grain population is a function of several parameters such as time, temperature and volume of seeds. The high-toughness samples can also be obtained faster by using higher sintering temperatures, but other problems may occur such as low cost-effectiveness, high residual stresses or interfacial reactions. The toughness values reach a peak after 90– 180 minutes of sintering time, which is in accordance with Faber’s theory stating that the crack deflection mechanism becomes independent of volume fraction above 20%.31 For smaller volume fractions, the toughness of the composite is proportional to the mean grain diameter. The reduction of toughness for the composites sintered for 9 hours is associated with interfacial chemistry – possibly increased crystallization of the grain boundary phase or increased thickness of the interfacial reaction zone. Si3N4–SiC(w) composites One of the main objectives in adding SiC(w) to Si3N4 matrices is to improve the fracture toughness. Figure 2.5 shows a crack propagating through a reaction-bonded silicon nitride reinforced with SiC whiskers which has been deflected by a whisker, resulting in an increase in absorbed energy. This
© Woodhead Publishing Limited, 2006
50
Ceramic matrix composites
2.5 Crack deflection in a SiC(w)-reinforced reaction bonded silicon nitride.
toughening behaviour rests on numerous variables including characteristics of the SiC whiskers and their surfaces. There are several manufacturing processes available to produce SiC whiskers, which result in whiskers, and hence composites, possessing a wide range of properties. For example, Rossignol et al.5 have demonstrated that Si3N4–30%SiC(w) composites manufactured with the same procedure but with differing SiC whiskers may have nearly 20% difference in indentation fracture toughness (6.8 ± 0.9 vs. 8.2 ± 0.6 MPa.m1/2) for samples with similar porosity levels (98.7 vs. 99.7% TD). As mentioned earlier, fracture toughness values can vary significantly due to the method of measurement, the reinforcement properties, manufacturing defects, etc. Taking into account these possible factors, a small sample of the fracture toughness values reported in the literature for Si3N4–SiC(w) composites with different volume fractions and different sintering additives is presented in Fig. 2.6. The fracture toughness values of the samples reaction-bonded and then sintered with yttria and alumina are high even in the monolithic state and the sample do not show any major improvement of toughness with addition of SiC(w). In contrast, the composites sintered with the AlN-rare earth additives possess a lower toughness for the monolithic sample which improves with the addition of whiskers. As mentioned earlier, two major factors influence the toughening mechanisms: (1) the interfacial reaction layer, which affects the debonding and pullout, and (2) the bimodal grain size distribution for crack deflection. For the sample reaction-bonded and sintered with Y2O3–Al2O3 the microstructure revealed, in addition to the
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
Fracture toughness (MPa.m1/2)
11
51
Notch-beam technique
10 9 8 7
Vickers indentation technique
6 5 4 3
Process
2
Rx bonded – hot pressed 8wt% Y2O3–1wt% Al2O3 Hot-pressed 5wt% AlN–5wt% CRE* Hot-pressed 5wt% AlN–2.5wt% CRE
1
Sintering additives
Ref [27] [33] [33]
0 0
2
4
6 8 10 12 14 16 Volume fraction SiC(w) (%)
18
20
2.6 Fracture toughness as a function of SiC whisker content for different composites.
SiC(w), the presence of Si3N4 needles in an equiaxed Si3N4 matrix. The toughening is attributed to these phases and thus is not significantly affected by the presence of SiC(w). Regarding the samples sintered with the AlNbased sintering additive, the presence of elongated Si3N4 grains was not reported and the improvement of toughness is directly related to the presence of SiC(w). These results also reinforce the necessity of having higher Y2O3 concentrations in order to improve toughness, as better interfacial reactions occur at the interface with the Si3N4 grains. Finally, the results demonstrate that composites reinforced with 20 vol% of SiC(w) can reach similar values of toughness to microstructurally optimized Si3N4–Si3N4(w) composites. Figure 2.7 summarizes the variation of toughness (indentation method) of Si3N4 composites as a function of volume fraction of SiC whiskers possessing different aspect ratios (R). In all cases, similar sintering additives were used. The maximum fracture toughness, approximately 10 MPa.m1/2 for 20 vol% SiC(w), is reached for the largest aspect ratio (15). The moderate improvement for whiskers with lower aspect ratio is attributed to their smaller contribution to the main toughening mechanisms: crack deflection and crack bridging. The rapid decrease in toughness of the composites with the highest aspect ratio is associated with the reduced density of the composite. In fact, such long whiskers impede consolidation and densities of only 2.5 g/cm3 were obtained. Therefore, the optimum whisker aspect ratio for producing a composite with high fracture toughness should be kept at around 15, which also allows a respectable level of density.
© Woodhead Publishing Limited, 2006
52
Ceramic matrix composites 10
Fracture toughness (MPa.m1/2)
9 8
R = 15
7 6
R=5
R = 25
5 4 3 2 1
Process
Sintering additives
Ref.
Hot-pressing Pressureless sintered Pressureless sintered
Y2O3–Al2O3 8wt% Y2O3–2wt% Al2O3 8wt% Y2O3–2wt% Al2O3
[34] [3] [3]
0 0
5
10 15 20 Volume fraction SiC(w) (%)
25
30
2.7 Effect of aspect ratio R on fracture toughness of SiC(w)–Si3N4 composites.
Microhardness The microhardness values reported for silicon nitride reinforced with SiC whiskers are somewhat contradictory when compared to unreinforced Si3N4; sometimes they are higher and sometimes lower. A main factor affecting the hardness is the presence, chemical content and composition of the sintering additive, which dictates the bonding characteristics between the reinforcement and the matrix. Dogan and Hawk13 reported a higher increase of microhardness over the monolithic material for their Si3N4–20%SiC(w) composite with crystalline sintering additives (19.0 vs. 15.6 GPa) than for their composite with an amorphous phase boundary (16.5 vs. 15.0 GPa). Baldacim et al.35 have measured the variation of the microhardness as a function of the sintering additive (mixtures of AlN and Y2O3) and volume fraction of SiC(w) (10, 20 and 30 vol%). Their results show that the microhardness decreases with the volume fraction of reinforcement and, in addition, the measured values are higher for the sintering additive with a lower AlN/Y2O3 ratio. They correlate this higher hardness with the presence of crystalline Y2Si3N4O3, not detected for the higher ratio of AlN/Y2O3. Wear resistance In monolithic or ceramic matrix composites, increases in hardness or toughness are not the only factors that lead to improved wear resistance. Microstructure
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
53
is also a major factor in the equation. The addition of a second phase, whiskers for instance, creates internal stresses. The difference in CTE in multiphase components, as well as anisotropic thermal expansion in single-phase materials, form tensile and compressive residual stresses at phase boundaries. As seen previously, the addition of silicon carbide can increase the toughness and the hardness of silicon nitride composites, though the wear behaviour may be decreased. As described above, some sintering additives result in lower wettability of the whiskers, thus giving a lower interfacial strength, leading to debonding of the whiskers which, although increasing toughness, increases the wear rate by surface pullout compared to monolithic Si3N4.13 Thermal shock resistance Thermal shock is one of the main drawbacks in the utilization of ceramics for high-temperature applications, and one of the aims of making CMCs is to improve thermal shock resistance. The thermal shock of ceramic materials is influenced by many factors such as strength, Young’s modulus, fracture toughness, thermal conductivity and thermal expansion coefficient. As whiskerreinforced silicon nitrides can show an improvement in room-temperature strength and fracture toughness, one would expect that these CMCs will also possess increased thermal shock resistance. Jia et al.21 have reported that the addition of SiC whiskers improves the thermal shock resistance to catastrophic failure but decreases the resistance to fracture initiation, demonstrating that the situation is not straightforward. Creep resistance Creep resistance for both types of whisker reinforcement in sintered composites is governed by grain boundary sliding via a viscous flow mechanism occurring along the amorphous phase. Depending on the temperature, the volume fraction and the viscosity of the glass system, if the rigidity of the residual sintering additive phase can be maintained no significant creep is observed. Nixon et al.36 have reported that creep of hot-pressed Si3N4–20%SiC(w) with Y2O3 and Al2O3 as additives is negligible below 1300°C in a nitrogen environment as the grain boundaries are sufficiently rigid and that relatively little displacement along the grain boundaries is observed. Similar results have reported similar creep results for Si3N4 reinforced with 0–20 vol% SiC(w) with MgO additive.37 In both cases, the researchers reported that the volume fraction of the reinforcement had no effect, showing that the viscosity of the glassy phase is the main active mechanism for creep. Above 1300oC, cavitation at Si3N4–Si3N4 grain boundaries becomes more important as the viscous flow of the sintering additives is increased;38 however, no such cavities were observed at the SiC–Si3N4 boundaries. In some cases, 20 vol% SiC(w)
© Woodhead Publishing Limited, 2006
54
Ceramic matrix composites
composites show higher creep rates than the unreinforced Si3N4 and this is attributed to a higher resistance to devitrification of the grain boundary phase due to the SiO2 present at the surface of the SiC whiskers. For a 30 vol% SiC(w)–Si3N4 composite (MgO additive), the creep resistance in air was improved by the whiskers, which may suggest that the higher volume fraction of reinforcement may play a more important role in improving creep resistance as more mechanical impingement should occur, reducing deformation.39 This is contrary to the behaviour of monolithic ceramics sintered with MgO additives: they exhibit lower creep resistance as the amorphous Mg2SiO4 glass composition is known to have a lower viscosity than Y2O3–Al2O3 base glass for the same test temperature, since Y2O3 is considered as a devitrifying agent in this glass composition. Fatigue resistance The crack growth rate during cyclic fatigue of materials is commonly recognized to be lower than for a sustained load test. Whisker-reinforced silicon nitride composites are no exception to this rule. However, the presence of the second phase (whiskers) and sintering additive, depending on the processing route, will also influence the fatigue resistance. Zhu et al.40 have reported that for gas-pressure sintered Si3N4–20vol%SiC(w), the static and cyclic fatigue life are equal for temperatures up to 1000oC but a decrease of the static strength was observed at higher temperature. This change in properties is related to the change of fracture mode which is related to the microstructure. At temperatures up to 1000oC, the fracture propagation occurs in a mixture of intergranular and transgranular modes, whereas nucleation, growth and interlinkage of cavities at the front of the propagating crack is observed at higher temperatures.
2.4
Applications
As discussed previously, ceramic matrix composites were originally developed to overcome the brittleness of monolithic ceramics. Thermal shock, impact and creep resistance can also be improved, making CMCs premium replacement choices for some technical ceramics. Industrial applications such as in automotive gas turbines or advanced cutting tools are already taking advantage of such characteristics.
2.4.1
High-temperature gas turbines
Between 1990 and 1997, the Japanese successfully developed a 100 kW automotive ceramic gas turbine in which the severe operating conditions required high-performance materials. 41 Among the different CMCs developed,
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
55
the turbine rotor (rotational component) and the back plate (stationary component) were made of SiC(w)–SiAlON or Si3N4(w)–Si3N4 ceramics. The turbine rotor, made of in situ reinforced Si3N4 ceramic, exhibits a Weibull modulus of over 20, indicating the reliability of the component. This part has confirmed its potential through hot spin tests at 1200oC (70,000 rpm), a flexural strength at 1200oC of 960 MPa, a fracture toughness of 7 MPa.m1/2 and an oxidation resistance at 1200oC for 200 hours (890 MPa). Regarding the back plate, both SiC(w)–SiAlON and Si 3N 4(w)–Si 3N 4 compositions sustained the stationary test (26 and 31 hours respectively at 1350oC, 5 atm), proving their suitability for such components.
2.4.2
Cutting tools
The trend in machining is to raise throughput, i.e. increase machining speed and material removal rate. The application of CMCs becomes eminently favourable under such conditions. Silicon nitride ceramic (sintered with additive) is conventionally used to mill or rough-turn cast irons. Self-reinforced Si3N4(w)–Si3N4 ceramics, produced by sintering, are showing improved hightemperature properties and excellent fracture toughness compared to the conventional Si3N4 as no softening of the glass phase occurs at high temperature.42 A number of suppliers are using this fabrication method for cutting tool inserts. SiC(w)–Si3N4 ceramics are also used for turning hard parts (hard steels or iron-based parts with hardnesses of 48–64 HRC). Turning has been shown to give higher material removal rates compared to conventional grinding operations for these materials, but this process generates higher temperatures and cutting forces, justifying the utilization of CMCs. Unfortunately, the silicon nitride matrix possesses a lower resistance to chemical wear arising from reaction between the materials being machined and the tool tip, limiting their uses.43
2.5
References
1. Tani, E., Umebayashi, S., Kishi, K. and Kobayashi, K. ‘Gas-pressure sintering of Si3N4 with concurrent addition of Al2O3 and 5 wt% rare earth oxide: high fracture toughness Si3N4 with fiber-like structure’, Am. Ceram. Soc. Bull., 65[9] (1986) 1311–1315. 2. Li, C.W. and Yamanis, J. ‘Super-tough silicon nitride with R-curve behaviour’, Ceram. Eng. Sci. Proc., 10[7–8] (1989) 632–645. 3. Sneary, P.R., Yeh, Z. and Crimp, M.J. ‘Effect of whisker aspect ratio on the density and fracture toughness of SiC whisker reinforced Si3N4’, J. Mat. Sci., 36 (2001) 2529–2534. 4. Shalek, P.D., Petrovic, J.J., Hurley, G.F. and Gac, F.D. ‘Hot-pressed SiC whisker/ Si3N4 matrix composites’, Am. Ceram. Soc. Bull., 65[2] (1986) 351–356.
© Woodhead Publishing Limited, 2006
56
Ceramic matrix composites
5. Rossignol, F., Goursat, P. and Besson, J.L. ‘Microstructure and mechanical behaviour of self-reinforced Si3N4 and Si3N4–SiC whisker composites’, J. Eur. Ceram. Soc., 13 (1994) 299–312. 6. Tiegs, T.N. and Becher, P.F. ‘Sintered Al2O3/SiC–whisker composites’, Am. Ceram. Soc. Bull., 66[2] (1987) 339–342. 7. Muscat, D., Pugh, M.D., Drew, R.A.L., Pickup, H. and Steele, D. ‘Microstructure of an extruded β-silicon nitride whisker-reinforced silicon nitride composite’, J. Am. Ceram. Soc., 75[10] (1992) 2713–2718. 8. Park, D.-S., Roh, T.-W., Han, B.-D., Kim, H.-D. and Park, C. ‘Microstructural development of silicon nitride with aligned β-silicon nitride whiskers’, J. Euro. Ceram. Soc., 20 (2000) 2673–2677. 9. Yonezawa, T., Saitoh, S.-I., Minamizawa, M. and Matsuda, T. ‘Pressureless sintering of silicon nitride composites’, Composites Sci. Tech., 51 (1994) 265–269. 10. Moulson, A.J. ‘Reaction bonded silicon nitride’, J. Mat. Sci., 14 (1979) 1017. 11. Pugh, M.D. and Gavoret, L. ‘Nitridation of whisker reinforced reaction bonded silicon nitride ceramics’, J. Mat. Sci., 35 (2000) 3257–3262. 12. Lundberg, R.L., Kahlman, L., Pompe, R., Carlsonn, R. and Warren, R. ‘SiC-whisker reinforced Si3N4 composites’, Am. Ceram. Soc. Bull., 66[2] (1987) 330–333. 13. Dogan, C.P. and Hawk, J.A. ‘Influence of whisker reinforcement on the abrasive wear behavior of silicon nitride- and alumina-based composites’, Wear, 203–204 (1997) 267–277. 14. Campbell, G.H., Rühle, M., Dagleish, B.J. and Evans, A.G. ‘Whiskers toughening: a comparison between aluminium oxide and silicon nitride toughened with silicon carbide’, J. Am. Ceram. Soc., 73[3] (1990) 521–530. 15. Orsini, P.G., Buri, A. and Marotta, A. J. Am. Ceram. Soc., 58 (1975) 306. 16. Turner, W.E.S. J. Am. Ceram. Soc., 12 (1929) 760. 17. Lee, C.J., Chae, J.I. and Kim, D.J. ‘Effect of β-Si3N4 starting powder size on elongated grain growth in β-Si3N4 ceramics’, J. Eur. Ceram. Soc., 20 (2000) 2667–2671. 18. Emoto, H. and Mitomo, M. ‘Control and characterization of abnormal growth grains in silicon nitride ceramics’, J. Eur. Ceram. Soc., 17 (1997) 797–804. 19. Sun, E.Y., Alexander, K.B., Becher, P.F. and Hwang, S.L. ‘Beta-Si3N4 whiskers embedded in oxynitride glasses: interfacial microstructure’, J. Am. Ceram. Soc., 79[10] (1996) 2626–2632. 20. Sun, E.Y., Becher, P.F., Hsueh, C.H., Alexander, K.B., Waters, S.B., Plucknett, K.P., Hirao, K. and Brito, M.E. ‘Microstructural design of silicon nitride with improved fracture toughness: II, Effect of additives’, J. Am. Ceram. Soc., 81 [11] (1998) 2831– 2840. 21. Jia, D.C., Zhou, Y. and Lei, T.C. ‘Thermal shock resistance of SiC whiskers reinforced Si3N4 ceramic composites’, Ceramics International, 22 (1996) 107–112. 22. Lee, S.W., Chae, H.B., Park, D.S., Choa, Y.H., Niihara, K. and Hockey, B.J. ‘Thermal conductivity of unidirectionally oriented Si3N4/Si3N4 composites’, J. Mat. Sci., 35 (2000) 4487–4493. 23. Russell, L.M., Donaldson, K.Y., Hasselman, D.P.H., Corbin, N.D. Petrovic, J.J. and Rhodes, J.F. ‘Effect of vapor–liquid–solid and vapor–solid silicon carbide whiskers on the effective thermal diffusivity/conductivity of silicon nitride matrix composites’, J. Am. Ceram. Soc., 74[4] (1991) 874–877. 24. Chu, C.Y., Singh, J.P. and Routbort, J.L. ‘High-temperature failure mechanisms of hot-pressed Si3N4 and Si3N4/Si3N4-whisker-reinforced composites’, J. Am. Ceram. Soc., 76[5] (1993) 1349–1353.
© Woodhead Publishing Limited, 2006
Whisker-reinforced silicon nitride ceramics
57
25. Chu, C.Y. and Singh, J.P. ‘Mechanical properties and microstructure of Si3N4-whiskerreinforced Si3N4 matrix composites’, Ceram. Eng. Sci. Proc., 11[7–8] (1990) 709– 720. 26. Becher, P.F., Sun, E.Y., Plucknett, K.P., Alexander, K.B., Hsueh, C.H., Lin, H.T., Waters, S.B., Westmoreland, C.G., Kang, E.S., Hirao, K. and Brito, M.E. ‘Microstructural design of silicon nitride with improved fracture toughness: I, Effects of grain shape and size’, J. Am. Ceram. Soc., 81[11] (1998) 2821–2830. 27. Shih, C.J., Yang, J.-M. and Ezis, A. ‘Microstructure and properties of reactionbonded/hot-pressed SiCw/Si3N4 composites’, Composites Sci. Techn., 43 (1992) 13–23. 28. Bellosi, A. and De Portu, G. ‘Hot-pressed Si3N4-SiC whisker composites’, Mat. Sci. Eng., A109 (1989) 357–362. 29. Pezzotti, G., Tanaka, I. and Okamoto, T. ‘Si3N4/SiC-whisker composites without sintering aids: III, High temperature behaviour’, J. Am. Ceram. Soc., 74[2] (1991) 326–332. 30. Becher, P.F. ‘Microstructural design of toughened ceramics’, J. Am. Ceram. Soc., 74 (1991) 255. 31. Faber, K.T. and Evans, G.A. ‘Crack deflection processes: theory and experiment’, Acta Metall., 31[4] (1983) 565–584. 32. Peillon, F.C. and Thevenot, F. ‘Microstructural designing of silicon nitride related to toughness’, J. Eur. Ceram. Soc., 22 (2002) 271–278. 33. Baldacim, S.A., Santos, C. Silva, O.M.M. and Silva, C.R.M. ‘Ceramics composites Si3N4-SiC(w) containing rare earth concentrate (CRE) as sintering aids’, Mat. Sci. and Eng., A367 (2004) 312–316. 34. Rajan, K. and Šajgalík, P. ‘Microstructurally induced internal stresses in β-Si3N4 whiskers-reinforced Si3N4 ceramics’, J. Eur. Ceram. Soc., 17 (1997) 1093–1097. 35. Baldacim, S.A., Santos, C., Silva, O.M.M. and Silva, C.R.M. ‘Mechanical properties evaluation of hot-pressed Si3N4-SiC(w) composites’, Int. J. Refrac. Met. and Hard Mat., 21 (2003) 233–239. 36. Nixon, R.D., Chevacharoenkul, S., Davis, R.F., Huckabee, M.L. and Buljan, S.T. ‘Deformation behaviour of SiC whisker reinforced Si3N4’, in Materials Research Society Symposium Proceedings, vol. 78, ed. Becher, P.F., Swain, M.V. and Somia, S. Materials Research Society, Pittsburg, PA (1987) 295–302. 37. Backhaus-Ricoult, M., Castaing, J. and Roubort, J.L. ‘Creep of SiC-whiskers reinforced Si3N4’, Revue Phys. Appl., 23[3] (1988) 239–249. 38. Nixon, R.D., Koester, D.A., Chevacharoenkul, S. and Davis, R.F. ‘Steady-state creep of hot-pressed SiC whisker-reinforced silicon nitride’, Composites Sci. Tech., 37 (1990) 313–328. 39. Desmarres, J.M., Goursat, P., Besson, J.L. Lespade, P. and Capdepuy, B. ‘SiC whiskers reinforced Si3N4 matrix composites: Oxidation behavior and mechanical properties’, J. Eur. Ceram. Soc., 7 (1991) 101–108. 40. Zhu, S., Mizuno, M., Kagawa, Y., Nagano, Y. and Kaya, H. ‘Static and cyclic fatigue in SiC whisker-reinforced silicon nitride composite’, Mat. Sci. and Eng., A251 (1998) 113–120. 41. Kaya, H. ‘The application of ceramic-matrix composites to the automotive ceramic gas turbine”, Composites Sci. and Tech., 59 (1999) 861–872. 42. Bhola, R., Das Gupta, S. and Jacobs, J.K. ‘Ceramic cutting tool inserts’, Key Engineering Materials, 122–124 (1996) 235–246. 43. Brandt, G. ‘Advanced tool materials’, Euro PM 2001 Proceeding, 1 (2001) 90–95.
© Woodhead Publishing Limited, 2006
3 Fibre-reinforced glass/glass-ceramic matrix composites R B A N E R J E E and N R B O S E, Central Glass and Ceramic Research Institute, India
3.1
Introduction
The traditional or conventional ceramics are generally in monolithic form. These include bricks, pottery, tiles and a variety of art objects. The advanced or high-performance monolithic ceramic materials represent a new and improved class of ceramic materials where, frequently, some sophisticated chemical processing route is used to obtain them. Generally, their characteristics are based on the high quality and purity of the raw materials used. Examples of these high-performance ceramics include oxides, nitrides, carbides of silicon, aluminium, titanium and zirconium, alumina, etc. Monolithic high-performance ceramics combine some very desirable characteristics, e.g. high strength and hardness, excellent high-temperature capability, chemical inertness, wear resistance and low density. They are, however, not very good under tensile and impact loading, and, unlike metals, they do not show any plasticity and are prone to catastrophic failure under mechanical or thermal loading (thermal shock). The difference in the behaviour of metals and ceramics can be categorized by saying that metals are forgiving while ceramics are not forgiving. The forgiving nature of metals has its source in the high mobility of dislocations of their atoms in them, which allows them to deform plastically before fracture. Plastic deformation being an energy-absorbing process, the fracture process in metals involves extensive energy dissipation. Ceramic materials are lacking such energy-dissipating phenomenon, which causes them to fail in a catastrophic fashion, i.e. makes them unforgiving. Therefore, a critical need exists for increasing the toughness of ceramic materials. With a view to achieving high fracture toughness of ceramic materials, the major effort of the materials community in the field of structural materials has been directed towards incorporating a variety of energy-dissipating phenomena in the fracture process of ceramics, i.e. imparting damage-tolerant behaviour. Improving the toughness and in-service reliability of ceramic materials is thus the major objective. One of the important approaches to attaining these goals is through composite and glass/glass58 © Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
59
ceramics matrix composites, and these are among the materials attracting major attention today. During the last quarter of the twentieth century, materials scientists have been able to revolutionize a major change in the development of materials that are considered suitable for applications at temperatures of more than 300oC with characteristics of high fracture toughness. In the past, designers worked with data on the properties of homogeneous, isotropic materials and designed their components to fit the ranges of ‘design allowances’ by using fibre-reinforced polymer matrix and metal–matrix composites for structural applications. The use of such composites increased to the point where boronand graphite-fibre reinforced epoxy resin and boron-reinforced aluminium were restricted to applications up to 300°C. Recently, however, the concept of composite materials has permitted almost limitless tailoring of composites to create entirely new designs never previously possible. By selecting judiciously the types of material constituents, their relative percentages, their orientation and fabrication methods, the designer can now work closely with the materials scientist to optimize system performance. The widespread acceptance of this philosophy, combined with new challenges to create engineering advances in aerospace and commercial areas, has created tremendous opportunities for new composites possessing greater environmental stability, higher temperature capabilities and high fracture toughness characteristics. The quest for such excellent environmental aspects and improved characteristics of materials prompted the development of fibrereinforced glass/glass-ceramic matrix composite (CMC). The combination of starting materials with suitable properties and appropriate fabrication procedure ultimately determines the properties of the resultant composite. Although this simple and obvious statement encompasses too many facets to be considered to achieve the high-performance characteristics of the resultant composite, two facets can be considered for comment. First, the fibres should not be greatly degraded during processing either by handling or by chemical reaction, and second, the resultant fibre–matrix interface must have the characteristics necessary to prevent excess fibre–matrix bonding. As should be clear from the discussion above, high-performance ceramics must have superior structural and/or mechanical characteristics because they find application in some very demanding environments, e.g. rocket nozzles, heat exchangers, automobile engines and cutting tools. Yet another important factor is the cost of ceramics. The great challenge is to produce consistent and reliable ceramic components having superior properties but without excessive cost, i.e. they should be competitive on a cost/performance basis with the materials they seek to replace. In this regard fibre-reinforced glassceramic composite has taken a leading role. Incorporation of fibres, whiskers or particles in a ceramic/glass-ceramic matrix can result in a tough ceramic material. This happens because
© Woodhead Publishing Limited, 2006
60
Ceramic matrix composites
incorporation of reinforcements introduces energy-dissipating phenomena such as debonding at the fibre/matrix interface, crack deflection, fibre bridging, fibre pullout, etc. In this regard, proper control of the characteristics of the interface region is of obvious importance. Yet another point to note is that while the ratio of the modulus of the reinforcement and the polymer or metal matrix is generally between 10 and 100, this ratio for a CMC is rather low, and can frequently be equal to unity or even less. The fibre may be continuous or discontinuous with a high aspect ratio. The high modulus ratio in polymer– matrix composites (PMC) and metal–matrix composites (MMC) allows an efficient load transfer from the matrix to the fibre via a strong interface. However, in a CMC, unlike PMC and MMC, the low modulus ratio means that the reinforcement and the matrix are not very different in their loadbearing capacity, i.e. a simple increase in strength of a ceramic material is rarely the objective. It is therefore necessary to consider low modulus based matrix material (e.g. glass/glass-ceramic matrix) in CMC for developing high strength in the materials.
3.2
Types of fibre suitable as reinforcements in different glass/glass-ceramic matrix composites
Reinforcements in the form of continuous fibres, short fibres, whiskers or particles are available commercially. Continuous ceramic fibres are very attractive as reinforcements in high-temperature structural materials. They provide high strength and elastic modulus with high temperature-resistant capability and are free from environmental attack. Ceramic reinforcement materials are divided into oxide and non-oxide categories, listed in Table 3.1. The chemical compositions of some commercially available oxide and non-oxide reinforcements are given in Table 3.2 and Table 3.3.
Table 3.1 Ceramic reinforcement materials56 Category Continuous fibres Oxide Non-oxide Discontinuous fibres Whiskers Short fibres Particles
© Woodhead Publishing Limited, 2006
Reinforcement materials
Al2O3, (Al2O3 + SiO2), ZrO2 Silica based glasses, etc. B, C, SiC, Si3N4, BN SiC, TiB2, Al2O3 Glass, Al2O3, SiC, (Al2O3 + SiO2) Vapour-grown carbon fibre SiC, TiC, Al2O3
Fibre-reinforced glass/glass-ceramic matrix composites
61
Table 3.2 Composition of oxide ceramic reinforcements56 Trade name
Manufacturer
Composition (%) Al2O3
Continuous Nextel 312 Nextel 440 Nextel 480 Nextel 550 Nextel 610 Astroquartz Saphikon
3M 3M 3M 3M 3M J.P. Stevens Saphikon
62 70 70 73 >99 —
Discontinuous (various silica-based glasses) Saffil ICI 96 Fibermax Carborundum 72 Fiberfrax Carborundum 52
ZrO2
SiO2
B2 O 3
— 24 14 — 28 2 — 28 2 — 27 — — 0.2–0.3 — — 99.95 — Al2O3 single crystal — — —
4 28 48
— — —
Fe2O3
— — — — 0.4–0.7 —
— — —
Table 3.3 Composition of non-oxide ceramic reinforcements56 Trade name Continuous fibres Nicalon SCS, Sigma Tyranno Tyranno Discontinuous fibres Silar Tokawhisker
Manufacturer
Composition (%)
Nippon Carbon Co. Textron Specialty Materials, BP Ube Ind. Ube Ind.
SiC + O + C SiC on tungsten or carbon substrate SiC + Ti + C SiC + Zr + C
Adv. Composite Material Corp. Tokai Carbon
SiC whiskers SiC whiskers
Many researchers have used glass and glass-ceramic matrices for reinforcing with high-modulus graphite fibres [1, 2], silicon carbide fibres and silicon carbide mono-filaments [3–7]. Very strong, tough and refractory composites were obtained from these efforts. Incorporation of discontinuous graphite fibres in glass and glass-ceramic matrices has also been reported [8, 9]. The basic mechanism of strengthening in fibre composite is that of load transfer by the matrix to the fibres through shear. This load transfer takes place at the fibre ends within a few fibre diameters. Depending on the length of the fibre, the amount of load transferred by the matrix to the fibre changes. Therefore, the fibre may not be loaded to the breaking stress and full advantage cannot be taken of its reinforcing ability. The stress concentration at the fibre ends also results in lower strength than would be possible. The load transfer depends on the properties of the
© Woodhead Publishing Limited, 2006
62
Ceramic matrix composites
matrix, the reinforcing fibre and the interface between them. These properties will determine the extent of reinforcement that can be put into the composite with the given aspect ratio of the fibre [10]. The salient features of oxide fibres and non-oxide fibres and their fabrication process are discussed below.
3.2.1
Oxide fibres
Ceramic oxide fibres, both continuous and discontinuous, have been commercially available since the 1970s, and processing and microstructure control are very important in obtaining the desired properties. Among the desirable characteristics in any ceramic fibre for structural applications are: • • • •
High theoretical density, i.e. low porosity Small grain size for low-temperature applications Large grain size for high-temperature applications High purity.
Alumina fibres have γ, δ, η and α allotropic forms. α-Alumina is the thermodynamically stable form. In practice, it is very difficult to control the time and temperature conditions to proceed from γ to α. At low firing, the fibres will give a smaller grain size and therefore an unacceptable level of porosity. At higher processing temperatures, porosity can be eliminated but excessive grain growth will result. This dilemma can be avoided by introducing a second phase that restricts grain boundary mobility while the porosity is removed at high temperature. It is possible to select the type and amount of the second phase that inhibits the grain growth at the service temperature. There are various ways to select this second phase by trial and error. It has been made possible to lower the working temperature by introducing oxides of silicon, phosphorus, boron or zirconium as the second phase, thereby inhibiting the formation of a grain boundary. Some results obtained after experimental studies are α-alumina + 15–20% ZrO2, δ-alumina + 4% SiO2 and α-alumina + 0.4% Fe2O3 + 0.25% SiO2 for lowering the working temperature during the formation of fibres. Various types of oxide fibres available commercially can be considered as suitable reinforcements other than the types of fibres listed in Table 3.2: • Fibre FP: a polycrystalline continuous α-alumina fibre yarn produced by DuPont in the 1980s. • Fibre PRD-166: another polycrystalline continuous alumina fibre yarn produced by DuPont in the 1980s. PRD-166 fibre yarn is a modified form of FP fibre yarn. The diameter of this fibre filament is 20 µm. The modification of FP fibre is made by incorporating 15–20 wt% yttriastabilized zirconium particles. The idea of incorporating Y2O3-stabilized zirconia particles was to take care of problems such as unstable mechanical
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
63
properties caused by grain growth and creep at high temperature. PRD166 fibre had a rough surface and an average grain size of about 0.5 µm. The zirconia particles were about 0.1 µm across and located mostly at grain boundary triple points. Their function was to inhibit grain growth in alumina fibre. Although DuPont no longer produces these fibres, their fabrication represented an important step in the processing of aluminatype fibres [11–13]. • Minnesota Mining and Manufacturing Co., also known as 3M Co., has developed an α-alumina fibre, trade name Nextel 610, via the sol-gel route. Figure 3.1 shows the 3M process schematically. • Sumitomo Chemical Co. produces a fibre that is a mixture of alumina (85%) and silica (15%). The fibre structure consists of fine crystallites of spinel. SiO2 serves to stabilize the spinel structure and prevents it from transforming to α-alumina [14]. The flow diagram of this process is shown in Fig. 3.2. • Series of various Nextel fibres produced by 3M Co. are mainly Al2O3 + SiO2 and some B2O3. The properties of some oxide fibres and monoxide fibres are given in Table 3.4. ICI Co. uses a sol-gel method to produce silica-stabilized alumina (Saffil) and calcia-stabilized zirconia fibre [15]. The saffil fibre is a δ-alumina short staple fibre that has about 4% SiO2 and a very fine diameter (3 µm).
Filter
Reservoir of organic basic Al salt solution
Pump Spinneret
Drawing wheels
Pyrolysis furnace zone (1400°C)
Winder Low temperature furnace zone for straightening fibre
3.1 3M’s process for making Al2O3 fibre (reproduced by permission of Chapman & Hall)56.
© Woodhead Publishing Limited, 2006
64
Ceramic matrix composites
Polyaluminoxanes–Si compound
Alkoxy alumino gel compound AIR3
Polymerization reactor Al AIR3 + H2O
O
R
Alkyl silicate and organic solvent
Dry spinning zone
Winder for drawing in organic fibre Al2O3: 70–100% SiO2: 30–0%
3.2 Flow diagram of the Sumitomo process for making a mixture of alumina and silica fibre (reproduced by permission of Chapman & Hall)56.
Sowman has provided details of the process used by 3M Co. for making the Nextel oxide fibre [16]. The starting material for this fibre is aluminium acetate, Al(OH)2OOCCH2 . 1 H3BO3. It is a product of Niacet Corporation 3 under the trade name Niaproof. A continuous polycrystalline α-alumina, trade name Almax, has been prepared by researchers at the Mitsui Mining Co. [17] by dry spinning a viscous slurry consisting of an aluminium salt, a fine powder of intermediate alumina, and an organic binder to produce the precursor fibre; this is followed by prefiring (calcining) and firing (sintering) the precursor fibre to produce an alumina fibre. Figure 3.3 shows the flow diagram of Almax alumina fibre. A continuous monocrystalline sapphire (Al2O3) fibre has been prepared as single-crystal fibres by LaBelle and Mlavsky using a modified czochralski puller and radio frequency heating. The technique adopted in this method is called edge-defined film-fed growth (EFG) [18–22]. Figure 3.4 shows a schematic of the EFG method.
© Woodhead Publishing Limited, 2006
Table 3.4 Composition and properties of some oxide fibres56 Fibre type
Composition (wt%) ——————————————————————— Al2O3 SiO2 B2 O 3 Fe2O3
Diameter (µm)
Density (gm/cm3)
Tensile strength (MPa)
Tensile modulus (GPa)
— — — — 0.7
10–12 10–12 10–12 10–12 10–12
2.7 3.05 3.05 3.03 3.75
1700 2000 2070 2240 1900
152 186 220 220 370
—
3 70–250 9
2.3 3.8 3.2
1000 3100 2600
100 380 250
3M 3M 3M 3M 3M
Nextel Nextel Nextel Nextel Nextel
312 440 480 550 610
62 70 70 73 79
24 28 28 27 0.2–0.3
ICI
Saffil Saphikon Sumitomo
96
4 — Al2O3 single crystal 15
© Woodhead Publishing Limited, 2006
85
14 2 2 — —
66
Ceramic matrix composites Al2(OH)5Cl
Al2(OH)5Cl powder
Dispersion Spinning and sintering aid AlCl36H2O Mixing
Filtration
Forming gel
Spinnable mixture
Dry spinning
1500–3000 poise at 25°C
50–100 m/min
Precursor fibre
Prefiring
Firing
250–500°C
1400–1600°C
α-Alumina fibre
3.3 Flow diagram for Almax Alumina fibre (reproduced by permission of Chapman & Hall)56.
Several investigators using a laser-heated floating zone method have prepared a variety of ceramic fibres. Gasson and Cockayne [22] used laser heating for the crystal growth of Al2O3, Y2O3, MgAl2O4 and Na2O3. Haggerty [23] used a four-beam heated float zone method to grow single-crystal fibres of Al2O3, Y2O3, TiC and TiB2. The laser-heated float zone technique is shown in Figure 3.5.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
67
Sapphire seed
Molybdenum capillary
Molten alumina
3.4 Edge-defined, film-fed, growth process for making a single-crystal alumina fibre (reproduced by permission of Chapman & Hall)56. Pull Seed crystal
Laser
Molten
Feed source rod
3.5 Laser-heated float zone technique (reproduced by permission of Chapman & Hall)56.
Another novel technique of making oxide fibres is called the inviscid melt technique [24]. In principle, any material in a molten state can be drawn into a fibre shape. Organic polymeric fibres such as nylon, aramid, etc., as well as a variety of glasses can be routinely converted into fibrous form by passing a molten material, having an appropriate viscosity, through an orifice. The inviscid (meaning low viscosity) melt technique uses this principle to make oxide fibres.
© Woodhead Publishing Limited, 2006
68
3.2.2
Ceramic matrix composites
Non-oxide fibres
Commercially available non-oxide ceramic reinforcements are in three categories: continuous, discontinuous, and whiskers. The great breakthrough in the ceramic fibre area has been the concept of pyrolysing polymers under controlled conditions, containing the desired species to produce hightemperature ceramic fibres. Silicon carbide fibre is a major development in the field of ceramic reinforcements. Non-oxide fibres via polymers The SiC fibre obtained via CVD is very thick and not very flexible. By an alternative route, very fine, continuous and flexible fibre was obtained by Yajima and his colleagues [25, 26] in Japan using a process of controlled pyrolysis of polymeric precursor. The ceramic fibres produced by this process possess good mechanical properties; good thermal stability and oxidation resistance have enormous potential for the development of ceramic matrix composites. Figure 3.6 shows the various steps involved in processing nonoxide fibres through the polymeric route, and Fig. 3.7 shows schematically the Yajima process of making SiC fibre from a polycarbosilane. Non-oxide fibres via CVD Silicon carbide (SiC) deposited on a substrate of tungsten or carbon heated to about 1300oC [27] is called sigma fibre (BP Sigma). A detailed schematic of the process used by BP to make its Sigma fibre is shown in Fig. 3.8. Textron Specialty Material Co. has developed a series of surface-modified Polymer precursor
Melt or solution spinning
Precursor fibre Curing
Cured or stabilized fibre
Controlled pyrolysis
Ceramic or glass fibre
3.6 Various steps involved in processing non-oxide fibres through the polymeric route (reproduced by permission of Chapman & Hall)56.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
CH3
69
Cl Si
CH3
Cl
Dichlorodimethylsilane
Dechlorination with Na (to NaCl) CH3 Si CH3 n Polydimethylsilane
Polymerization at 470°C in autoclave CH3
H
Si
C
CH3
H
n
Polycarbosilane
Polycarbosilane fibre Polycarbosilane fibres with molecular crosslinking by oxygen to avoid subsequent melting SiC fibre Amorphous or microcrystalline β-SiC
Melt spinning at 350°C (N2) Curing at 190°C in air or RT in ozone
Pyrolysis to 1300°C in vacuum (1000°C h–1)
3.7 Yajima process for making SiC from a polycarbosilane (reproduced by permission of Chapman & Hall)56.
silicon carbide fibres, called SCS fibres. These fibres have a complex throughthickness gradient structure. SCS fibre is a thick fibre (142 µm) and is produced by CVD of silicon- or carbon-containing compounds onto a pyrolytic graphite-coated carbon core. The pyrolitic graphite coating is applied to a carbon monofilament to give a substrate of 37 µm. SiC is then coated by CVD to give a final monofilament of 142 µm diameter. Figure 3.9 shows schematically the cross-sections of the two Textron SCS-type SiC fibres.
© Woodhead Publishing Limited, 2006
70
Ceramic matrix composites Hydrogen supply
Exhaust W. filament +
Flow meters Scrubber Silane vaporizer Refrigeration Silane supply Distillation
Waste byproduct
+
Photo optical diameter sensor
SiC/W Gases for recuperation
3.8 Schematic process for making SiC monofilament fibre by the CVD method (reproduced by permission of Chapman & Hall)56.
SCS-6 (~140 µm)
Pyralytic graphitecoated carbon core
SCS-0 (~75 µm)
Inner zone carbon-rich β-SiC Outer zone: stoichiometric β-SiC Carbon-rich surface coating (0–4 µm)
3.9 Schematic of two Textron SCS-type silicon carbide fibres (reproduced by permission of Chapman & Hall)56.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
71
Nicalon-type SiC fibres The Nicalon fibre (10–20 µm) available commercially consists of a mixture of β-SiC free carbon and SiO2 [28]. The properties of Nicalon start to degrade above about 600oC because of the thermodynamic instability of the composition and microstructure. Ceramic-grade Nicalon fibres, designated the NL series, having low oxygen content are also available. Other SiC/Si3N4-type fibres Various SiC-type fibres with elemental compositions of Si–C, Si–N–C–O, Si–B-N, Si–C–O and Si–Ti–C–O are commercially available. These fibres are made from polymeric precursors. A multifilament SiC fibre, called Tyranno, is produced by Ube Industries, Japan [29]. This fibre is made by the pyrolysis of poly(titano carbosilanes) and contains 1.5–4% titanium by weight. Another multifilament fibre is called silicon carbonitride, trade name HPZ, produced by Dow Corning Corporation, USA. Several researchers also produce silicon nitride fibre by using various organometallic compounds. SiC-based silicon nitride fibre is produced by the pyrolysis of organosilazane polymers with methyl groups on Si and N [30]. Researchers at Tonen Corporation of Japan also produce pure silicon nitride fibre. This fibre is made by using perhydropolysilazane polymer (PHS). Another silicon nitride fibre, called Tonen SiNB, is based on boron. Silicon nitride (Si3N4) fibres can also be prepared by reactive chemical vapour deposition (CVD) using volatile silicon compounds. The reactants are generally SiCl4 and NH3. Si3N4 is deposited on carbon or tungsten substrate. All these fibres can be used for the fabrication of glass/glass-ceramic matrix composite with proper interface modification. The interface modification is outside the scope of this discussion. Whiskers Whiskers are normally obtained by vapour phase growth. They are monocrystalline, short fibres with extremely high strength because of their high aspect ratio (50 to 10 000). They have a diameter of a few microns, but they do not have uniform dimensions and properties. Other non-oxide fibres There are other promising ceramic fibres, e.g. boron carbide and boron nitride. Boron nitride fibre has the same density (2.2 g cm–3) as carbon fibre, but has a greater oxidation resistance and excellent dielectric properties. Boron carbide fibre is a very light and strong material.
© Woodhead Publishing Limited, 2006
72
Ceramic matrix composites
Rice hull
Carbon tube reactor
Dispersion of shredded rice hull
Whisker and carbon separation
Whisker and hull relict separation
Drying
Carbon oxidation
SiC whisker
3.10 Manufacturing process for silicon carbide whiskers by VLS from rice husk (reproduced by permission of Chapman & Hall)56.
Production SiC particles and whiskers are produced from rice hulls. The rice hulls are heated at about 700°C in the absence of oxygen. This system is called coking. The coked rice hulls consist of equal amounts of C and SiO2. When the coked rice hulls are heated at a temperature of 1500–1600°C for about an hour in an inert atmosphere (N2 or NH3 gas) they form SiC whiskers. Figure 3.10 shows a schematic of the manufacturing process for SiC whiskers by the Vapour–Liquid–Solid (VLS) processing method. In this process silicon and carbon are obtained from SiO and CH4 gases respectively. Figure 3.11 shows the chemical transformation of the VLS whisker process.
3.3
Methods for manufacturing different fibrereinforced glass/glass-ceramic matrix composites
The techniques for fabricating glass/glass-ceramic matrix composites are based on the final properties and performance of the resultant composite.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
73
SiO(g) + C = Si + CO(g) CO(g)
SiO2 and C
Catalyst
SiO(g) Co(g)
Fe–Si–C
Generator Si + C = SiC CH4(g)
2H2(g)
SiC CH4(g) = C + 2H2(g) Whisker
3.11 Chemical transformation of the vapour-liquid-solid (VLS) phases for growing whiskers (reproduced by permission of Chapman & Hall)56.
Some of the techniques are unconventional and few are conventional powder processing techniques. The techniques are described below.
3.3.1
Conventional techniques
Cold pressing and sintering The first stage in producing a component is to press the powder and the binder into a desired shape using a sufficiently high pressure so that a relatively dense and strong green body is formed. In certain circumstances, a variety of fast production methods can be used such as extrusion, blow moulding, injection moulding, etc. It is necessary to remove the organic binder before a fully sintered body with a near-theoretical density can be obtained. The greatest uniformity of density is obtained by the application of pressure from all directions, which is known as isostatic pressing. To improve the strength of the resultant composite, compacted ingredients have to be heated to elevated temperatures in order to burn off the binder and to consolidate the powder further by sintering. Generally, in the sintering step, the matrix shrinks considerably and the resultant composite develops many cracks. There are other limitations on sintering of glass, ceramic or glass-ceramic matrix materials containing high aspect ratio reinforcements. Because of the difference in
© Woodhead Publishing Limited, 2006
74
Ceramic matrix composites
thermal expansion coefficients of the reinforcement and matrix, a hydrostatic tensile stress may develop in the matrix on cooling. This effect retards the sintering process [31, 32]. In glass matrix composites sintering is retarded if the reinforcement is greater than about 15 vol% [33, 34]. Yet another limitation to consider is the bridging phenomenon, which is caused by the formation of networks resulting from whiskers or fibres. So it is important to optimize the volume fraction as well as the aspect ratio of the whiskers or reinforcement for getting the densified product [35]. Experimental evidence was established by the fact that the sintered density of silicon carbide reinforced alumina decreased as the aspect ratio of the whisker increased [36]. Hot-pressing Simultaneous application of pressure and high temperature can accelerate the rate of densification and production of a pore-fee and fine-grained compact of the ceramic materials. The most important technique used to produce continuous fibre-reinforced glass and glass-ceramic composite is the slurry infiltration process [37–40]. This was developed about 20 years ago in the United Kingdom for the production of glass-matrix composites but it is now also widely used for glass–ceramic matrix composites. The intimate mixing of continuous fibres and the glass or glass–ceramic matrix is achieved by drawing bundles of fibres, called tows, through a slurry of powdered glass in water and a water-soluble resin binder. Figure 3.12 shows a schematic of this
Glass-impregnated fibre tape Glass slurry tank
Fibres
Fibre/glass composite
Stack of glassimpregnated fibre tapes
Pressure
Binder burnout 500°C
Hot pressing 800–925°C Graphite die
3.12 Schematic of the slurry infiltration process for making a fibrereinforced glass and glass–ceramic composite (reproduced by permission of Woodhead Publishing Limited)74.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
75
process. The tows, impregnated with the slurry, are wound onto a mandrel to form a monolayer tape. The tape is cut into plies, which are stacked into the required stacking sequence before burnout of the binder. Wetting agents may be added to ease the infiltration of the fibre tows. The impregnated tape is then hot-pressed for consolidating the matrix. The CMCs produced by hot pressing are of a superior quality. The flaws encountered in CMCs produced by hot-pressing may be due to the presence of a combination of matrix-rich regions and fibre-rich regions. This kind of inhomogeneity weakens the composite. Some CMCs produced by various investigators using hot-pressing techniques are silicon carbides, alumina and carbon fibres in a variety of glass, glass–ceramic and oxide–ceramic matrices [39–41]. The slurry infiltration process is ideal for making glass or glass–ceramic matrix composites because the processing temperatures for these materials are lower than those used for crystalline matrix materials. Figure 3.13(a) shows an optical micrograph of a transverse section of a unidirectional alumina fibre/glass matrix composite, while Fig. 3.13(b) shows the pressure and temperature schedule used during hot-pressing of this composite. Some porosity can be seen in this picture. Lanxide process The Lanxide process was developed by Lanxide Corporation [42]. It involves the formation of a ceramic matrix by the reaction between a molten metal and a gas. For example, when molten aluminium is reacted with oxygen, alumina is formed. The ceramic matrix occurs outwards from the original metal surface. In the case of fibrous composites, filament winding or a fabric lay-up may be used to make a preform. A fabric made of a continuous fibre can also be used. The fabric is coated with a proprietary chemical solution to protect the fibre from highly reducing aluminium and to provide a weak interface. A barrier is placed on the preform surfaces to control the rate of growth of the matrix material. In this method, the transport of the ceramic matrix occurs continuously at the oxidation reaction front. The desired reaction product forms on the surface of the molten metal and grows outward. This process is considered as a low-cost process and there is a possibility to get near-net shape products. It has been reported [42] that products with good mechanical properties (strength, toughness, etc.) can be obtained by this process, which can make glass/glass–ceramic matrix composites. A schematic of the Lanxide process is shown in Fig. 3.14. Various researchers have used different techniques for the fabrication of ceramic matrix composite (CMC) by using (a) chemical vapour deposition (CVD), (b) chemical vapour infiltration (CVI) and (c) modelling of CVI [43–54], but it is difficult to fabricate glass/glass–ceramic composite using these techniques.
© Woodhead Publishing Limited, 2006
76
Ceramic matrix composites
(a) Temperature Pressure
1000
6
5
4 600 3 400
Pressure (MPa)
Temperature (°C)
800
2 200
0
1
0
50
100 Time (min)
150
200
0
(b)
3.13 (a) Optical micrograph of a cross-section of a unidirectional alumina fibre/glass matrix composite made by slurry infiltration; (b) pressure and temperature schedule used during hot-pressing of this composite (reproduced by permission of Chapman & Hall)56.
Sol-gel technique One of the most innovative approaches to ceramic and glass processing is the sol-gel technique. A brief description of the process is given below. The sol-gel route of making any glass or ceramic involves the formation of the
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
77
Reactive gaseous environment Preform Infiltrated preform Liquid metal
(a) γSV
Air
γSL
Molten metal (b)
3.14 Lanxide process: (a) infiltration of preform; (b) wicking of liquid metal along grain boundaries (reproduced by permission of Woodhead Publishing Limited)74.
appropriate glass or ceramic structure by chemical polymerization of suitable compounds in the liquid state (sol) at low temperatures, followed by chemical reactions such as hydrolysis or condensation at temperatures much lower than those used in powder processing or direct melting. The particle size in sol generally varies between 1 and 100 nm. It can also be obtained by mixing a metal-containing precursor (e.g. an acid) and water. Hydrolysis and precondensation reactions make the sol viscosity increase until a gelled state is obtained. This gel is like a wet solid and is termed a precursor. The wet gel is dried to remove any unwanted residue (water, organic compounds, etc.). The gel is then converted into glass or ceramic by heating at temperatures much lower than those used in direct melting processes. The glass or ceramic thus obtained may be in the form of powder, film, fibre, etc. The slurry infiltration process results in a fairly uniform fibre distribution inside the composite and low porosity, and enhanced high-strength values of the composites. Figure 3.15 shows the sol-gel process flow diagram. For the fabrication of ceramic matrix composite conventional polymer handling and processing techniques are used. Fibrous preforms are made by impregnation of sol in vacuum or filament winding techniques. In filament winding, fibre tows or ravings are passed through a tank containing sol and the impregnated tows are wound on a mandrel to a desired shape and thickness. The sol is converted to gel and the structure is removed from the mandrel. A final heat treatment is done at 1400oC to get ceramic or glass–ceramic
© Woodhead Publishing Limited, 2006
78
Ceramic matrix composites Pour Pour sol over preform
Mix Mix sol with reinforcement or gel with reinforcement
Dry
Dry
Repeat Repeat infiltration and drying until required density
Fire
Calcine Heat to produce required ceramic
Hot press
Film
Fibre
Powder (a)
(b)
3.15 Sol-gel processing: (a) infiltration of a preform; (b) mixing reinforcement in a sol or a gel (reproduced by permission of Woodhead Publishing Limited)74.
composite. A fibrous preform can be vacuum impregnated with a polymer precursor in liquid form and an appropriate heat treatment then given to convert the polymer into ceramic. Particulate or whisker reinforcement can be disposed in the sol or gel state. Sol-gel filament winding of a fibre preform and vacuum impregnation of woven preforms are shown in Fig. 3.16 and Fig. 3.17 respectively. Researchers at GEC in England have emphasized the fabrication of continuous fibre-reinforced ceramics in various complex shapes. They have used liquid precursors to produce a ceramic matrix in a fibrous preform [55, 56]. Sol-gel processing is a viable means of preparing glass, ceramics and thin films through hydrolysis and condensation of metal alkoxides in organic solvents [57–61]. Compared with conventional techniques, the sol-gel method has several advantages because many multi-component oxides can be prepared with a higher degree of chemical purity and easier control of stoichiometry. The sol-gel technique offers excellent composition control, low-temperature processing and short fabrication times at comparatively low cost. One of its disadvantages consists of the fact that only a small thickness (up to ~ 200 nm) of high-quality film can be achieved in one coating cycle, so several
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
Dryer Spool
Filament winding Sol
Gelled body
Heating coils Heat to convert the gel into glass or ceramic
3.16 Sol-gel filament winding of a woven preform (reproduced by permission of Chapman & Hall)56.
Vacuum Sol
Preform
Vacuum impregnation by sol
Gelation
Heating coils
Heat to convert gel into glass or ceramic
3.17 Vacuum impregnation of a woven preform (reproduced by permission of Chapman & Hall)56.
© Woodhead Publishing Limited, 2006
79
80
Ceramic matrix composites
coating cycles are necessary when thick films are required. One of the ways to make a thick layer is to form a composite material. Several researchers have reported the preparation of cermets [62] or sol-gel polymeric composite [63, 64]; lead zirconate titanate (LZT) and ZrO2 thick films have been synthesized [57] using the dispersion of ceramic powders in sol containing zirconium alkoxides.
3.4
Properties of glass/glass–ceramic matrix composites
The important ceramic matrix materials are glass, silicon carbide, silicon nitride, alumina, glass–ceramics, sialons, intermetallics and some elemental materials. A list of some ceramic matrix materials is given in Table 3.5. The characteristic high strength and brittleness of ceramic matrix materials can be judged by the types of bonding in their structure [65, 66]. In ceramic matrix materials with ionic bonding, there occurs a transfer of electrons between the atoms, and in case of covalent bonding, the electrons are shared between atoms. The properties of some ceramic matrix materials are given in Table 3.6. Table 3.5 List of some ceramic matrix materials56 Nitrides
Silicon nitride (Si3N4), boron nitride (BN)
Carbides
Silicon carbide (SiC), boron carbide (B4C), titanium carbide (TiC)
Mixed oxides
Mullite (3Al2O3.2SiO2), spinel (MgO.Al2O3)
Single oxides
Alumina (Al2O3), zirconia (ZrO2), titania (TiO2), magnesium oxide (MgO), Silica (SiO2)
Intermetallics
Nickel aluminide (NiAl, Ni3Al), titanium aluminide (TiAl, Ti3Al), molybdenum disilicide (MoSi2)
Elements
Carbon (C), boron (B)
Table 3.6 Properties of some ceramic matrix materials56 Ceramic matrix materials
Physical and mechanical properties —————————————————————————————— Density Melting Young’s Coefficient of Fracture ρ (g.cm–3) point modulus thermal toughness (°C) E (GPa) expansion KIC α (10–6K–1) (MPa. m1/2)
Al2O3 SiC Si3N4 MgO Mullite Borosilicate glass Soda-lime glass
3.9 3.2 3.1 3.6 3.2 2.3 2.5
© Woodhead Publishing Limited, 2006
2050 — — 2850 1850 — —
380 420 310 210 140 60–70 60–70
7–8 4.5 3.1 3.6 5.3 3.5 8.9
1–3 2.2–3.4 2.5–3.5 — 3.0–4.0 0.5–2 0.5–1
Fibre-reinforced glass/glass-ceramic matrix composites
81
Table 3.7 Characteristics of some important varieties of glass56 Glass
Softening point (°C)
Density ρ (g.cm–3)
Toughness KIC (MPa.m1/2)
Soda-lime glass Borosilicate glass 96% Silica glass Fused quartz
700 825 1500 1580
2.4 2.3 2.5 2.6
0.7 0.8 — —
Glass-matrix materials can be considered as a non-crystalline solid with the frozen-in structure of a liquid. Characteristics of some important varieties of glass are given in Table 3.7. Glass–matrix materials are polycrystalline materials having fine ceramic crystallites in a glass matrix. Important glass– ceramic matrix materials are as follows. • Li2O–SiO2 (LAS). The trade names of such glass–ceramic matrix materials are Corningware, Zerodur and Ceran. This type of glass–ceramic matrix material has nearly zero thermal expansion and high thermal shock resistance. It is used for the production of optical and telescopic mirrors. • MgO.Al2O3.5SiO2. This system is utilized for the production of various stable crystalline phases. These are (i) 2MgO.2Al2O3.5SiO2 (known as cristobalite, tridymite and cordierite), (ii) enstatite (MgO.SiO2), and (iii) mullite (3SiO2.2Al2O3). These are used for making radar antennae and radomes for aircraft. • SiO2–Al2O3–MgO–K2O–F. The presence of the mica phase in this system helps easy machinability of the product. • SiO2–Al2O3–CaO. In this system wollastonite (CaO–SiO2) or anorthite (CaO–Al2O3–2SiO2) is present as main crystalline phase.
3.4.1
Properties of fibre-reinforced glass/glass–ceramic matrix composites
During the past few years, the interest in composite materials that could extend this temperature capability has been stimulated by the results of several research programmes dealing with glass-matrix composites reinforced with either graphite [67–70] or alumina fibres [71]. In all of these programmes it was found that the use of fibres exhibiting high strength and stiffness was successful in reinforcing lower-modulus glass matrices. The graphite fibrereinforced glass system demonstrated exceptionally high levels of strength, fatigue resistance and fracture toughness but was susceptible to fibre oxidation during elevated temperature oxidation. In contrast, the alumina fibre-reinforced silica matrix composite [71] was unaffected by exposure to temperatures above 1000oC in air; however, the overall levels of strength and toughness
© Woodhead Publishing Limited, 2006
82
Ceramic matrix composites
attained were far lower than those of the graphite fibre-reinforced glass system. The silicon carbide fibre-reinforced glass-matrix composites possess a unique combination of both high levels of mechanical performance along with excellent oxidation resistance. The AVCO System Division (Lowell, MA) is now producing a SiC monofilament of 140 µm diameter that is fabricated by CVD onto a carbon filament core in a pilot plant. This fibre exhibits an average tensile strength of up to 3450 MPa, has a temperature capability of over 1300°C, and is stable in oxidizing environments. It has a density of 3.2 g cm–3 and an elastic modulus of 415 GPa. This fibre is now available on the commercial market. The second type of SiC fibre, recently synthesized by Yajima et al. [72] in Japan, consists of continuous length SiC yarn that is produced from an organometallic polymer. The tows of yarn contain 2000 fibres per tow with an average fibre diameter of 10 µm. The SiC fibre is highly flexible with an extremely smooth surface. The manufacturer, with a use temperature of up to 1300°C, has reported tensile strengths of up to 3450 MPa for this fibre. The SiC yarn density is approximately 2.7 g cm-3 and the elastic modulus is 221 GPa. Composite fabrication The steps in the fabrication of a SiC monofilament-reinforced glass composite are as follows. The SiC monofilament is wound with the desired spacing on a drum, bonded together at periodic intervals with polystyrene, cut into individual tapes and then stacked up in the hot-pressing die to form the composite by alternating SiC tapes with 7740 glass powder (7740 is the Corning Glass Works designation for a borosilicate glass). The amount of powder is varied to make two types of composite samples, one with 65 vol% SiC and the other with 35 vol% SiC fibre. The composite lay-up is then hot-pressed for 20 min at a maximum temperature of 1150°C and a pressure of 6.9 MPa in an argon atmosphere. It is found that this hot-pressing schedule results in complete densification of the 7740 glass with very little bubble or void formation. A cross-section of a typical 65 vol% SiC composite is shown in Fig. 3.18. The SiC yarn-reinforced 7740 glass specimens are fabricated using the identical procedure developed for the fabrication of graphite fibre-reinforced glass-matrix composites [69, 70]. The SiC yarn is passed through a slurry of glass powder and isopropyl alcohol, dried and then cut into tapes of the appropriate length to fit the hot-pressing die. Sufficient tape is then stacked in the die to obtain the desired thickness composite and hot-pressed in vacuum for 1 hour at 1200oC and at a pressure of 14 MPa. This hot-pressing schedule is found to result in void-free composites with complete densification of the glass matrix. The coefficients of thermal expansion of 7740 glass-matrix
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
83
50µm
3.18 Cross-section of 65 vol% SiC monofilament-reinforced 7740 borosilicate glass composite (reproduced by permission of Matrix Composite Mat. Sci.)66.
composites are given in Table 3.8. These values were computed using the change in dimension measured between 22 and 500oC and indicate a relatively minor anisotropy in thermal expansion as compared with other composite systems such as graphite-reinforced glass [70]. The properties of SiC fibrereinforced 7740 glass matrix composites are given in Table 3.9. The mechanical properties of freeze-gelled, unidirectional carbon-fibre-reinforced CMC are given in Table 3.10 and the mechanical properties of some fibre-reinforced sol-gel glass matrix composites are given in Table 3.11. From the fracture characteristics of the silicon carbide yarn glass-matrix system, it appears that the crack blunting ability of this system at its present state of development is somewhat less than that for the silicon carbide monofilament glass system (Fig. 3.19).
Table 3.8 Code 7740 glass-matrix composite coefficients of thermal expansion, CTE (average value between 22°C and 500°C)66 Filament
Orientation
CTE (10–6 oC)
35 vol% SiC monofilament
0o 90o
4.20 4.60
40 vol% SiC yarn
0o 90o
3.25 2.70
© Woodhead Publishing Limited, 2006
84
Ceramic matrix composites
Table 3.9 Properties of SiC fibre-reinforced 7740 glass-matrix composites66 Monofilament Fibre content (vol%)
35
Density (g.cm–3) Axial at at at
65
2.6
flexural strength (MPa) 22°C 350°C 600°C
22oC axial elastic modulus (GPa)
Yarn
2.9
40 2.4
650 — 825
830 930 1240
290 360 520
185
290
120
Axial fracture toughness (MN.m–3/2) at 22°C at 600°C
18.8 14.3
— —
11.5 7.0
Table 3.10 Mechanical properties of freeze-gelled, unidirectional carbon-fibrereinforced CMC made by hand lay-up73 Sample
Density (×103) (kg m–3)
Flexural strength (MPa)
Strain at peak load (%)
Dynamic modulus (GPa)
Without added CDM 105* glass-ceramic
1.6 (0.02)†
118 (17)
0.4 (0.05)
40 (4)
With added CDM 105* glass–ceramic
1.7 (0.01)
212 (10)
0.7 (0.15)
43 (5)
*CDM 105 is a proprietary glass–ceramic, manufactured by the freeze gelation method. † Figures in parentheses are standard deviations.
Table 3.11 Mechanical properties of some fibre-reinforced sol-gel glass-matrix composites73 Fabrication route
Fibre
Final density (×103 kg m–3)
Dynamic modulus (GPa)
Flexural strength (MPa)
Work of fracture (KJ m–2)
Cast Hand lay-up Filament Wound
Saffil Carbon Mat Nextel F P Alumina Carbon
2.06 1.72 2.26 2.19 1.60
53.1 42.6 — — 36.0
25 212 179 210 270
0.16 — 3.10 1.70 13.30
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
85
20 µm
3.19 Fracture surface of SiC yarn-reinforced 7740 borosilicate glass (reproduced by permission of Matrix Composite Mat. Sci.)66.
3.4.2
Glass-ceramic matrix composites
A range of glass-ceramics, each having different crystalline phases, has been used as the matrix material. The most widely used matrices are based on the lithium aluminosilicate (LAS) system. The constituents of typical glass– ceramic matrices are given in Table 3.12. The strain to failure of SiC fibres is greater than that of the glass–ceramic matrix. During loading on silicon carbide fibre-reinforced glass–ceramic composite, matrix cracking occurs before failure of the fibres. The load is being transferred from the matrix to the fibres. The room-temperature modulus value of LAS is less than 100 GPa compared to over 200 GPa for the SiC fibres. It is therefore considered that there is potential for increasing the stiffness of the glass–ceramic matrix by incorporating SiC fibres. The tensile modulus data of LAS–SiC composites made with both unidirectional and cross-plied SiC fibres are presented in Table 3.13. Compared with other structural materials there is a relative dearth of fatigue data for ceramics, and it is therefore not surprising that little is known of the fatigue behaviour of glass–ceramic matrix composites. It appears Table 3.12 Typical compositions of lithium aluminosilicate (LAS) glass–ceramics74 LAS system
Constituents
LAS I
Li2O–Al2O3–MgO–SiO2 with addition of ZnO–ZrO2 and BaO
LAS II
Li2O–Al2O3–MgO–SiO2 with addition of Nb2O5, ZnO, ZrO2 and BaO
LAS III
Li2O–Al2O3–MgO–SiO2 with addition of Nb3O5 and ZrO2
© Woodhead Publishing Limited, 2006
86
Ceramic matrix composites Table 3.13 Modulus data for LAS–SiC composites74 Matrix system
Vol. % SiC
Tensile modulus (GPa)
LAS LAS LAS LAS LAS LAS
0 46 (unidirectional) 46 (unidirectional) 44 (unidirectional) ~50 (cross-plied) ~50 (3-D braid)
46 133 130 136 118 79–111
I II II I III
Table 3.14 Fatigue behaviour of SiC fibre-reinforced LAS74 Material
Tensile strength, σ (MPa)
Max. fatigue stress, σmax (MPa)
LAS I (unidirectional)
261 261 261
207 207 172–138
0.79 0.79 0.66–0.53
LAS II (unidirectional)
550 550 550 550
355 310 275 225
0.65 0.56 0.50 0.41
>105 >105 >105 >105
485 525 485 620
LAS III (unidirectional)
575 575 575 575 575 575
456 421 357 315 280 223
0.79 0.73 0.62 0.55 0.49 0.39
5 × 101 5 × 102 >105 >105 >105 >105
— — 462 480 602 538
LAS II (cross-plied)
325 325 325 325 325 325 325
361 285 280 280 262 256 245–210
1.10 0.88 0.86 0.86 0.81 0.79 0.75–0.65
LAS III (cross-plied)
269 269 269 269
181 179 179 186–163
0.67 0.67 0.67 0.69–0.61
σ max σ
Fatigue cycles
Residual strength (MPa)
1.9 × 102 2.2 × 103 >105
— — 286–226
× 101 × 102 × 102 × 102 × 102 × 104 >105
— — — — — — 385–315
104 3 × 103 5 × 103 >105
— — — 291–227
6 1.2 1.5 3 3 3
that at room temperature the fatigue performance is quite good; in tension– tension fatigue the maximum fatigue stress has to exceed about 0.7 of the tensile fracture stress to cause failure within 105 cycles in any of the unidirectional LAS matrix composites (Table 3.14). Residual strength measurements on specimens which survive 105 cycles suggest that significant damage, and hence a reduction in residual strength, occurs mainly when the maximum fatigue stress is greater than the stress required for matrix micro-
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
87
cracking. This is evident from the data of Table 3.14 bearing in mind that LAS-1 exhibits a linear stress–strain curve to failure, i.e. no micro-cracking, whereas the matrix cracking stress for LAS II and LAS III is about 270 MPa and 320 MPa respectively [74].
3.4.3
Silicon carbide whisker-reinforced glass and glass–ceramic composites
Several-fold improvements in strength and fracture toughness of glass and glass–ceramics have been obtained by reinforcing these matrices with silicon carbide whiskers. Incorporation of whiskers in the matrices results in a tremendous increase in viscosity, thus necessitating much higher composite processing temperatures compared to that of the matrix alone. Glass and glass–ceramic composites eliminate the residual glassy phase problem (which is creating the thermal expansion mismatch) and the increasing refractoriness of the matrix, thus enhancing the high-temperature properties. Several glass– ceramic compositions were studied [10] as matrices for whisker reinforcement: see Table 3.15. The first composite system is a barium osumilite composition reinforced with 30 wt% (25 vol%) No. 1 whiskers. The second composite system evaluated is a barium-stuffed cordierite matrix 30 wt% No. 1 whisker composite. The room-temperature properties of these two composites are given in Table 3.16. Figure 3.20 shows the change in modulus of rupture (MOR) and fracture toughness of a Ba–osumilite composite system with temperature [10]. The room-temperature properties are maintained to 900oC. Beyond 900oC the properties begin to deteriorate. The flexural strength drops from 406 MPa at room temperature to 55.1 MPa at 1200oC. There is an increase in KIC beyond 900oC. The fracture toughness value peaks at 1000oC and decreases again. The small amount of residual glass affects the composite performance significantly beyond 900oC. Figure 3.21 shows the change in MOR and fracture toughness of a Ba-stuffed cordierite composite system with temperature. The room-temperature properties are maintained to 900oC. Beyond Table 3.15 Glass–ceramic composition range10 Ba–stuffed cordierite
Ba–osumilite
4MgO.4Al2O3.10 SiO2 BaO2+ + 2Al3+ = 2Si4+ x BaO.4 MgO.(4 + x)Al2O3 × (10 – 2x) SiO2 Where x = 0 to 0.5
2BaO.4 MgO.6Al2O3.18SiO2 x [Ba2+ + 2Al3+] = 2x Si4+ n [Mg2+ + Si4+] = Al3+
© Woodhead Publishing Limited, 2006
(2 – x)BaO.(4 + n) MgO.(6 – x – n) Al2O3 × (18 + 2x + n) SiO2 where x = 0 to 0.5, n = 0 to 1.0
88
Ceramic matrix composites Table 3.16 Properties of Ba–osumilite and Ba-stuffed cordierite matrix composites containing 30 wt% No. 1 whiskers10 Property
Ba–osumilite matrix
Ba–stuffed cordierite matrix
MOR (MPa) KIC (MPa.M1/2) Young’s modulus (GPa) Shear modulus (GPa) Poisson’s ratio Thermal expansion at 25–1000°C (×10–7/°C Density (kg/m3)
400 4.5 156.4 62.0 0.262 35.5 2808
358 4.5 186 73 0.274 36.2 2770
7 MOR
350
5
280
4
210
3
140
2
70
1
25
m)
6
KIC
900 1000 1100 Temperature (°C)
KIC (MPa.
MOR (MPa)
420
1200
5
280
KIC
4
MOR (MPa)
210
3
140
2
70 25
900 1000 Temperature (°C)
1100
KIC (MPa.
MOR
●
350
m)
3.20 Variation of properties of Ba–osumilite matrix composite with temperature (reproduced by permission of Am. Ceram. Soc. Bull.)10.
1 1200
3.21 Variation of properties of Ba-stuffed cordierite matrix composite with temperature (reproduced by permission of Am. Ceram. Soc. Bull.)10.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
89
(b)
(a)
3.22 Fracture surface of Ba–osumilite matrix composite tested at (a) 25oC and (b) 1100oC (bar = 1 µm) (reproduced by permission of Am. Ceram. Soc. Bull.)10.
900oC the properties begin to deteriorate. The composite strength decreases from 358 MPa to 213.6 MPa at 1200oC. A much larger fraction of roomtemperature strength (60%) is thus retained to 1200oC compared to the Ba– osumilite system (13%). Figure 3.22 shows SEM micrographs of the fracture surfaces of Ba–osumilite composite tested at room temperature and at 1100oC. The flow of the glass is quite evident on the 1100oC fracture surface. SiC whiskers are available from various sources. Table 3.17 gives the composition of the glass matrices. Code 0080 (a soda lime glass), code 1723 (an aluminosilicate glass), code 7052 (a borosilicate glass) and code 7740 (another borosilicate glass) were used as matrices. Figure 3.23 shows the effect of the volume fraction of SiC whiskers in composite with code 1723 matrix on the critical aspect ratio, according to Dow’s analysis and Rosen’s analysis. Figure 3.24 shows the effect of porosity on flexural strength of composites with code 1723 glass matrix [75, 76]. Figure 3.25 shows a SEM micrograph of a well-consolidated 30 wt% SiC whisker-reinforced composite [10]. Table 3.18 gives the mechanical properties Table 3.17 Composition of glass matrices10 Oxide
SiO2 Al2O3 B 2O 3 Li2O Na2O K 2O MgO CaO BaO
© Woodhead Publishing Limited, 2006
Amount present (wt%) Code 0080
Code 1723
Code 7052
Code 7740
73 1 — — 17 — 4 5 —
57 16 4 — — — 7 10 6
64 8 19 1 2 3 — — 3
81 2 13 — 4 — — — —
90
Ceramic matrix composites 140 120
(l/d)C
100 80 60 40 20 0 0
10
20 30 Volume fraction
40
3.23 Effect of volume fraction of SiC whiskers in composite with code 1723 matrix on critical aspect ratio, according to (䊊) Dow’s analysis and(䊉) Rosen’s analysis (reproduced by permission of Am. Ceram. Soc. Bull.)10.
Flexural strength (MPa)
350 280
40 wt%
210 50 wt%
30 wt% 140 70 0 10
30
40
% porosity
3.24 Effect of porosity on flexural strength of composites with code 1723 matrix (reproduced by permission of Am. Ceram. Soc. Bull.)10.
3.25 Fracture surface of composite containing 30 wt% No. 1 whiskers in code 1723 matrix (bar = 10 µm) (reproduced by permission of Am. Ceram. Soc. Bull.)10.
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
91
Table 3.18 Properties of silicon carbide whiskers10 No. Type of whisker
Dimensions (µm) Diameter
1 2 3 4 5
α + β mixture α + β mixture α + β mixture α + β mixture β–SiC
Mechanical properties
Length
0.6 10–80 0.6 10–80 1–10 20–400 0.05–0.2 10–40 Tangled woolly whiskers
Strength (MPa)
Modulus (GPa)
689 689 551 482 No data
6.89 6.89 18 20 No data
and aspect ratio data on all these whiskers as available from the manufacturers [79]. SiC whisker-reinforced glass matrix composites were fabricated at the same process viscosity of the matrices and were well consolidated. All the composites were 30 wt% whiskers (No. 1) composites. The properties of these composites are given in Table 3.19. A comparison of the 1723 matrix composite and the 7052 composite shows that the latter is much weaker and has a lower modulus. Comparing the 7052 and 7740 systems, the 7740 composites are weaker still. A comparison of the 0080 and 1723 systems again shows a lower performance for the 0080 composite. The experiments were done [10] with the 1723 glass matrix and No. 1, No. 3 and No. 4 whiskers. Table 3.20 summarizes the results of these Table 3.19 Properties of SiC whiskers-reinforced glass matrix composites10 Glass matrix
Matrix expansion (10–7/°C)
MOR (MPa)
Code Code Code Code
52 53 34 95
337.6 241.1 194.3 179.1
1723 7052 7740 0080
(± (± (± (±
8%) 8%) 10%) 9%)
Composite modulus (GPa)
Matrix modulus (GPa)
141.6 107.5 92.3 96.4
86 (± 5%) 57.8 (± 8%) 62.6 (± 8%) 70.3 (± 7%)
(± 7%) (± 6%) (± 8%) (±10%)
Table 3.20 Comparison of whisker performance in code 1723 glass matrix composites10 Composite no.
1 2 3 4 5
Whisker no.
1 4 3 4 4
© Woodhead Publishing Limited, 2006
Whisker MOR content (MPa) of composite (wt.%) 30 30 30 30 30
338 193 263 190 250
(± (± (± (± (±
Modulus (GPa)
Fracture toughness (MPa. m1/2)
8%) 10%) 8%) 11%) 8%)
141.6 (± 5%) — 133.7 (± 6%) — 128.8 (± 5%)
3.4 (± 7%) — 2.1 (± 10%) — 2.8 (± 8%)
92
Ceramic matrix composites
experiments. All the experiments were done at 30 wt% (25 vol%) whisker loading, except for composite 4, which contained 20 wt% whiskers. The composite consolidation was conducted under the same conditions for composites 1 through 4, while composite 5 was processed at a higher temperature. The flexural strength measurements show that the No. 1 whiskers give the best strength values. The composite containing No. 1 whiskers is much stronger (338 MPa) than the No. 3 composite (263 MPa). The modulus of the No. 3 whisker composite is also lower (133.7 GPa) compared to the No. 1 composite (141.6 GPa).
3.5
Microstructural observation
Composite samples are sectioned with a diamond saw and mounted in cold curing epoxy resin. Because of their porous nature, the composites are infiltrated under vacuum and subsequently cured under pressure in order to force the mounting resin into the pores. Mounted samples are ground flat on 240 grit silicon carbide paper, finely ground with a 9 µm oil-based diamond slurry and finally polished with a 1 µm diamond slurry and a 50 nm silica suspension.
3.5.1
Scanning electron microscopy (SEM)
The polished samples are sputtered with a thin layer of gold for analysis in a scanning electron microscope (SEM), a Jeol JSM 35c fitted with a link AN 10000 energy-dispersive X-ray spectrometer (EDS). The fractured surfaces and polished sections through fractured specimens can also be prepared and analysed in this manner. SEM analysis may reveal a non-uniform fibre distribution in the composite. In composites sintered at different temperatures, cracking in the matrix phase and residual porosity can be identified and the filler particles are discernible. The EDS indicates the higher particles and the matrix constituents.
3.5.2
Transmission electron microscopy (TEM)
TEM analysis is performed in a Jeol 2000 FX equipped with an EDS system. Thin sections (approximately a few hundred nanometres thick) suitable for TEM are prepared by cutting 3 mm slices, grinding them to a thickness of ~300 µm and dimpling them to leave a central region ~10 µm thick. Thinning with argon ion bombardment in a Gatan Duomill may be carried out until specimen perforation occurs. TEM studies of the composite sintered at different temperatures may reveal the possibility of forming a fused network with the spherical particles originally present in the samples. The filler particles may
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
93
exhibit strain bands and the matrix may contain several twinned regions, both suggestive of residual strains within the material.
3.5.3
X-ray diffraction (XRD) analysis
XRD spectra of composites sintered at different temperatures are obtained by using a defractometer. Samples sintered at a particular temperature indicate the structural behaviour as either amorphous or crystalline in nature. Such indication is of immense help to researchers for improving the properties of the resultant materials by optimizing the rate of sintering temperature.
3.5.4
Optical microscopy analysis
An extensive network of porosity of the sintered samples is obtained by using an optical microscope. Dark-field illumination in the optical microscope reveals the pores to have a crystalline surface texture and readily distinguishes between the denser (darker) and more porous (lighter) regions of the specimens. Also, a degree of subsurface detail is revealed. Optical microscopy is carried out in reflected light with Nomarski differential interference contrast (DIC) and dark-field modes on a suitable microscope. The fibre volume fraction can be estimated for all the samples. Matrix cracks around fibres arising from residual stresses can be observed in all samples and tend to reach the specimen surface via dense matrix regions. Circumferential crack patterns may indicate residual stresses arising from fibre/matrix thermal expansion mismatches.
3.6
Application areas
Very important application areas of glass/glass–ceramic matrix composites are supersonic planes, US high-speed civil transport planes, turbine blades, heat shields, rocket nozzles and propulsion components, re-entry thermal protection for spacecraft, rocket cone frustra, braking materials, cutting tool inserts, high-wear parts such as wire drawing or extruding dies, valve seats, high-precision balls, bearings for corrosive environments, and plungers for chemical pumps. Other application areas for glass/glass–ceramic matrix composites are disc brakes for racing cars and aircraft, gas turbine components (e.g., exhaust nozzle flaps and seals), nose cones and leading edges for missiles, and biomedical implants such as bone plates. The applications of glass/glass–ceramic matrix composites (CMC) can be divided into two specific categories: aerospace applications and non-aerospace applications. In aerospace applications, performance is the prime consideration, while in non-aerospace applications cost-effectiveness is paramount. The characteristic properties of materials for aerospace applications should be
© Woodhead Publishing Limited, 2006
94
Ceramic matrix composites
high strength-to-density ratio, high stiffness-to-density ratio, improved damage tolerance at significantly higher temperatures and/or faster cruising speeds, and improved flight performance at higher altitudes. Continuous fibrereinforced ceramic matrix composites potentially offer higher specific strength properties, which can be utilized in various high-temperature aerospace applications. Silicon carbide coated carbon/carbon composites and carbon/ silicon carbide composites are the right candidate materials for such hightemperature aerospace applications [65, 66]. The applicable temperature range of CMCs is 800–1650°C. The tensile and compressive strengths of CMCs are in the range of 175–350 MPa and their moduli in the range of 100–175 GPa throughout the temperature range of 800–1650oC. In non-aerospace applications of CMCs, cutting tool inserts, wear-resistant parts, energy-related applications such as heat exchanger tubes, nozzles, exhaust ducts, etc., are the emerging areas. Particle and whisker-reinforced ceramics are commonly used for cutting tools, wear-resistant parts and heat engine applications. TiC particle-reinforced Si3N4 and Al2O3 and SiC(w)– Al2O3 composites are used for cutting tool inserts. Toughened zirconia and SiC whisker or continuous fibre-reinforced composites are used for making wear-resistant parts. The other non-aerospace application areas of carbon– carbon composites are biomedical implants and internal fixation of bone fractures because of their excellent biocompatibility. They are also used for making moulds for hot-pressing. Carbon–carbon moulds can withstand higher pressures and offer a longer life than polycrystalline graphite and tool steel material. In general, their high cost limits applications to aerospace and other special applications. Whisker-reinforced glass–ceramic matrices are expected to find several applications in automotive components, metal forming, cutting tools, etc., due to their low thermal expansion, high thermal shock resistance, high reliability and low material and processing costs. Some industrial applications for continuous fibre-reinforced ceramic matrix composites (CMCs) are listed below. • Heat engine liners, combustors, high-wear parts, etc. CMCs are used in high-temperature gas turbines. • CMCs are used in the manufacture of preheaters and recuperators in heat recovery equipment. They are used for indirect heating and energy-intensive industrial internal processes such as glass melters, steel reheaters and aluminium remelters. • Radiant tube burners are made by using CMC. These are used for indirectfired, high-temperature zones, controlled atmosphere heating and melting applications. • Reformers and reactors of chemical process equipment. • Handling equipment, internals and cleanup of waste incineration systems. • Filters, substrates and centrifuges of separation and filtration systems. Such filtration and separation systems can be used for gas turbines,
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
95
particulate traps for diesel exhausts, molten metal filters and sewage treatment equipment.
3.7
Future trends
The gel synthesis of ceramic powders offers many advantages over more conventional methods based on solid-state reactions: lower processing temperature, better control of the morphology and microstructure, powderless processing of ceramics. However, a better knowledge of sol-gel chemistry has to be developed before a real mastery of the process can be reached. This requires the careful characterization of all the chemical species formed during the course of the sol-gel process. The chemical modification of alkoxide precursors opens future possibilities for the molecular design of advanced ceramic matrix composite (CMC) materials. There is a possibility of developing potential high-strength viable CMC-based engineering materials by increasing toughness, decreasing sensitivity to flaws and increasing the reliability of the materials. The incorporation of carbon fibre into glass and glass–ceramic matrices may produce composite materials having toughness several times greater than that of monolithic matrix materials, which can withstand large tensile strains prior to final failure and can exhibit a significant degree of damage tolerance material. Conventional polymer-matrix composite processing techniques such as sheet moulding compound (SMC), filament winding and resin transfer moulding (RTM) may be utilized for the fabrication of advanced ceramic matrix composite prepregs before putting them into the firing schedule. Improvement of the freeze-gelation method may eradicate many problems associated with the CMC fabrication process and may develop high-fracture– toughness materials by reducing porosity and internal shrinkage.
3.8
References
1. Phillips, D.C., Sambell, R.A.J. and Brown, D.H., ‘The mechanical properties of carbon fibre reinforced pyrex glass’, J. Mat. Sci., 7, 1454–1464 (1972). 2. Phillips, D.C., ‘Interfacial bonding and the toughness of carbon fibre reinforced glass and glass-ceramics’, J. Mat. Sci., 9, 1847–1854 (1974). 3. Prewo, K.M. and Brennan, J.J., ‘High strength silicon carbide fibre reinforced glassmatrix composites’, J. Mat. Sci., 15, 463–468 (1980). 4. Brennan, J.J. and Prewo, K.M., ‘Silicon carbide reinforced glass-ceramic matrix composites exhibiting high strength and toughness,’ J. Mat. Sci., 17, 2371–2783 (1982). 5. Chyung, K., et al., ‘Nicalon fibre reinforced LAS glass-ceramic composites’, presented at the 9th Conf. on Composite Materials, January 1985, Cocoa Beach, FL. 6. Brennan, J.J., Chyung, K. and Taylor, M.P., ‘Glass-ceramic compositions of high refractoriness’, US. Patent 4,415,672, 15 Nov. 1983. 7. Brennan, J.J., Chyung, K. and Taylor, M.P., ‘Reaction inhibited-silicon carbide fibre
© Woodhead Publishing Limited, 2006
96
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. 33. 34. 35. 36.
Ceramic matrix composites reinforced high temperature glass-ceramic composites’, US Patent 4,485,179 27 Nov. 1984. Sambell, R.A.J., Bowen, D.H. and Phillips, D.C., ‘Carbon fibre composites with ceramic and glass materices’, J. Mat. Sci., 7, 663–675 (1972). Prewo, K.M., ‘A complaint high failure strain, fibre reinforced glass matrix composite’, J. Mat. Sci., 17, 3549–3563 (1982). GadKaree, K.P. and Chyung, K., ‘Silicon-carbide-whisper-reinforced glass and glassceramic composites’, Am. Ceram. Soc. Bull., 65 (2), 370–376 (1986). Dhingra, A.K., Phil. Trans. R. Soc. London, A294, 411 (1980). Romine, J.C., Ceram. Eng. Sci. Proc., 8, 755 (1987). Nourbakhsh, S., Liang, F.L. and Margolin, H., ‘Characterization of a zirconia toughened alumina fibre, PRD-166’, J. Mat. Sci., Letters, 8, 1252 (1989). Chowla, K.K., J. Metals, March, 35 (1983). Birchall, J.D., Bradbury, J.A.A. and Dinwoodie, J., in Strong Fibres, Handbook of Composites, Vol. 1, North-Holland, Amsterdam, p.115 (1985). Sowman, H.G., in Sol-Gel Technology, Noyes Publishing, Park Ridge, NJ, p. 162 (1988). Saitow, Y., Iwanaga, K. and Itou, S., et al., Proc. SAMPE Annual Meeting. LaBelle, H.E. and Mlavsky, A.I., Nature, 216, 574 (1967). LaBelle, H.E., Mat. Res. Bull., 6, 581 (1971). Pollack, J.T.A., ‘Filamentary sapphire. Part 3. The growth of void-free sapphire filament at rates up to 3 cm min’, J. Mat. Sci., 7, 787 (1972). Hurley, G.F. and Pollack, J.T.A., Met. Trans., 7, 397 (1972). Gasson, D.G. and Cockayne, B., J. Mat. Sci., 5, 100 (1970). Haggerty, J.S., NASA–CR–120948, May 1972. Wallenberger, F.T., Weston, N.E., Motzfeldt, K. and Swartzfager, D.G., ‘Inviscid melt spinning of alumina fibres: chemical jet stabilization’, J. Am. Ceram. Soc., 75, 629 (1992). Yajima, S., Okamura, K., Hayashi, J. and Omori, M., ‘Synthesis of continuous SiC fibres with high tensile strength’, J. Am. Ceram. Soc., 59, 324 (1976). Yajima, S., Phil. Trans. R. Soc. London, A294, 419 (1980). De Bolt, H.E., Krukonis, V.J. and Wawner, F.E., in Silicon Carbide 1973, University of South Carolina Press, Columbia, SC, p. 168 (1974). Laffon, C., Flank, A.M. and Lagarde, P., et al., ‘Study of Nicalon-based ceramic fibres and powders by EXAFS spectrometry, X-ray diffractometry and some additional methods’, J. Mat. Sci., 24, 1503 (1989). Yamamura, T., Ishirkawa, T. and Shibuya, M., et al., ‘Development of a new continous Si–Ti–C–O fibre using an organometallic polymer precursor’, J. Mat. Sci., 23, 2589 (1988). Milewski, J.V., Sandstrom, J.L. and Brown, W.S., in Silicon Carbide 1973, University of South Carolina Press, Columbia, SC, p. 634 (1974). Ray, R. and Bordia, R.K., Acta Met., 1003 (1989). Kellett, B. and Lange, F.F., ‘Stresses induced by differential sintering in powder compacts’, J. Am. Ceram. Soc., 67, 369 (1989). Rahaman, M.N. and De Jonghe, L.C., J. Am. Ceram. Soc., 70, C-348 (1987). Prewo, K.M., in Tailoring Multiphase and Composite Ceramics, Materials Science Research, Vol. 20, Plenum Press, New York, p. 529 (1986). Holm, E.A. and Cima, M.J., J. Am. Ceram. Soc., 72, 303 (1989). Tiegs, T.N. and Becher, P.F., Am. Ceram. Soc. Bull., 66, 339 (1987).
© Woodhead Publishing Limited, 2006
Fibre-reinforced glass/glass-ceramic matrix composites
97
37. Phillips, D.C., in Fabrication of Composites, North-Holland, Amsterdam, p. 373 (1983). 38. Cornie, J.A., Chiang, Y.-M. and Uhlmann, D.R. et al., Am. Ceram. Soc. Bull., 65, 293 (1986). 39. Prewo, K.M. and Brennan, J.J., J. Mat. Sci., 17, 2371 (1980). 40. Sambell, R.A.J., Phillips, D.C. and Bowen, D.H., in Carbon Fibres: Their Place in Modern Technology, The Plastics Institute, London (1974). 41. Briggs, A. and Davidge, R.W., in Whisker-and Fibre-toughened Ceramics, ASM International, Materials Park, OH, p. 153 (1988). 42. Urquhart, A.W., Mat. Sci. Eng., A144, 75 (1991). 43. Fitzer, E. and Hegen, D., Angew. Chem., 91, 316 (1979). 44. Fitzer, E. and Schlichtin, J., Z. Werkstofftechnik, 11, 330 (1980). 45. Fitzer, E. and Gadow, R., Am. Ceram. Soc. Bull., 65, 326 (1986). 46. Stinton, D.P., Caputo, A.J. and Lowden, R.A., Am. Ceram. Soc. Bull., 65, 347 (1986). 47. Burkland, C.V., Bustamante, W.E., Klacka, R. and Yong, J.M., in Whisker- and Fibre-toughened Ceramics, ASM International, Materials Park, OH, p. 225 (1988). 48. Middleman, S., J. Mat. Res., 4, 1515 (1989). 49. Currier, R.P., ‘Overlap model for chemical vapor infiltration of fibre yarns’, J. Am. Ceram. Soc., 73, 2274 (1990). 50. Tai, N.H. and Chou, T.W., ‘Analytic modeling of chemical vapor infiltration in fabrication of ceramic composites’, J. Am. Ceram. Soc., 72, 414 (1989). 51. Tai, N.H. and Chou, T.W., ‘Modeling of an improved chemical vapor infiltration process for ceramic composite fabrication’, J. Am. Ceram. Soc., 73, 1498 (1991). 52. Chung, G.Y. and Benjamin, J.M., ‘Modeling of chemical vapor infiltration for ceramic composites reinforced with layered, woven fabrics’, J. Am. Ceram. Soc., 74, 746 (1991). 53. Stinton, D.P., in Proc. 10th Int. Conf. on Chemical Vapour Deposition, The Electrochemical Society, Pennington, NJ, 1147.36 (1987). 54. Starr, T.L., ibid. 55. Hyde, A.R., GEC J. Res., 7, 65 (1989). 56. Chawla, K.K., Ceramic Matrix Composites, Chapman & Hall, New York (1993). 57. Barrow, D., PhD thesis, Queens University (1995). 58. Brinker, C.J. and Scherrer, G.W., in Sol-Gel Science, Chapter 13, Academic Press, New York (1990). 59. Nicolaon, G.A. and Teichner, S.J., Bull. Soc. Chim. France, 5 (1900). 60. Dislich, H. and Husmann, E., Thin Solid Films, 77 (1981). 61. Jabra, R., PhD Thesis, University of Montpellier, France (1979). 62. Colomer, M.T. and Jurado, J.T., ‘Thick film cermet of ZrO2–Y2O3–TiO2/Ni: polarization study’, J. Eur. Ceram. Soc., 19 (1999). 63. Dias, C., Wenger, M., Das-Gupta, D.K., Blanas, P. and Ahuford, R.J., ‘Intelligent piezoelectric composite materials for sensors’, NDT&E International, 30 (1997). 64. Igreja, R., Dias, C.J. and Marat-Mendes, J.N., ‘Processing and characterization of sol-gel derived modified PbTiO2 for ferroelectric composite applications’, Integrated Ferroelectrics, 8, 721–723 (1977). 65. Naslain, R., Lamon, J. and Donmeingts, D., High Temperature Ceramic Matrix Composites, 6th Eur. Conf. on Composite Materials, 20–24 September 1993, Bordeaux, Woodhead Publishing, Cambridge, UK, CEACM (1993). 66. Prewo, K.M. and Brennan, J.J., ‘High strength silicon carbide fibre-reinforced glassmatrix composite, J. Mat. Sci., 15, 463 (1980).
© Woodhead Publishing Limited, 2006
98
Ceramic matrix composites
67. Phillips, D.C., Sambell, R.A.J. and Bowen, D.H., ‘The mechanical properties of Carbon fibre reinforced Pyrex glass’, J. Mat. Sci., 7, 1454 (1972). 68. Levitt, S.R., ibid., 8, 793 (1973). 69. Prewo, K.M. and Bacon, J.F., Proc. 2nd Int. Conf. on Composite Materials, Toronto, Canada (AIME, New York, p. 64 (1978). 70. Prewo, K.M., Bacon, J.F. and Dicus, D.L., SAMPE Q., 42 (1979). 71. Bacon, J.F. and Prewo, K.M., Proc. 2nd Int. Conf. on Composite Materials, Toronto, Canada (AIME, New York), p. 753 (1978). 72. Yajima, S., Okamura, K., Hayashi, J. and Omori, M., ‘Glass-ceramic matrix composites’, J. Amer. Ceram. Soc., 59, 324 (1976). 73. Russel–Floyd, R.S., Harris, B., Cooke, R.G., Laurie, J. and Hammett, F.W., J. Am. Ceram. Soc., 76(10), 2635 (1993). 74. Matthews, F.L., and Rawlings, R.D., Composite Materials: Engineering and Science, Woodhead Publishing, Cambridge, UK (1999). 75. Dow, N.F., GEC Missile and Space Division, Report No. R635D61. 76. Rosen, B., Fibre Composite Materials, ASM, pp. 37–75 (1964). Process – I
© Woodhead Publishing Limited, 2006
4 Particulate composites R I T O D D, University of Oxford, UK
4.1
Introduction
Ceramic materials have been produced by man for at least 9000 years. The microstructures of most traditional ceramics resemble particulate ceramic composites in that at least one of the phases present consists of approximately equiaxed, discontinuously distributed particles. Although particulate phases may be present naturally in the clay used for shaping, for much of the history of ceramic technology particulates have also been added deliberately as a ‘temper’ of quartz, limestone, sand, shell, ‘grog’ (recycled pulverised pottery) or other easily available substances. The function of these particulates in traditional ceramics is usually to give high-temperature strength so that the shape is retained during firing or to act as a cheap filler, and thus has little relevance to this publication. There is evidence, however, that variations in the choice of temper occurring over periods of many years in particular communities resulted in improvements in mechanical properties such as strength, toughness or thermal shock resistance [1] and point to at least an accidental application more than 1000 years ago of some of the principles described in this chapter. The development of ‘engineering ceramics’ with sufficient strength for application in load-bearing situations did not begin in earnest until the 1960s. Research concentrated initially on simplifying and refining the coarse and flaw-ridden microstructures found in traditional ceramics. This was very successful, and the outcome was the production of almost pore-free, finegrained, single-phase ceramics with strengths of several hundred MPa (compared with several tens of MPa for traditional ceramics). Investigations then switched to methods of increasing the toughness. The range of innate ceramic toughness values is relatively limited, although the use of ceramics such as carbides and nitrides with strong covalent bonds and therefore inherently high surface energy succeeded in pushing strengths towards 1 GPa. More substantial increases in fracture toughness have proved possible only by making the microstructure more complex again, though this time in a systematic 99 © Woodhead Publishing Limited, 2006
100
Ceramic matrix composites
and targeted fashion, particularly by making many of the types of composite and related structure described in this book. The particulate composites described in this chapter constitute perhaps the simplest departure from a fine-grained single-phase ceramic. The particulates do not provide the highest strengths or the greatest degree of toughening to be found in ceramic composites, but against this they are relatively cheap and easy to process compared with other shapes of reinforcement. Particulate reinforcements also provide inherently isotropic properties (cf. long-fibre composites) and are less toxic and easier to handle than whiskers. Although particulate composites already see commercial application, their simplicity makes them useful as model materials as well, easier to understand and simulate than more complicated microstructures. This chapter aims to describe the principles behind the processing, microstructural development and properties of particulate ceramic composites and to illustrate these using experimental results. The main emphasis is on examples where the addition of particulates to a ceramic matrix causes new mechanisms to operate that give an improvement in properties greater than would be expected from a ‘rule of mixtures’. The chapter concentrates almost exclusively on structural composites, since this is where most work has been done to date. Particulate ‘nanocomposites’ are included in the chapter, since the important examples described are currently at the coarse end of the ‘nanoscale’, and the principles underpinning their properties seem to be a simple extension of those relevant to the ‘microcomposites’ with which the rest of the chapter is concerned. The next section describes the processing and microstructural development of particulate composites, and is followed by a section on thermal residual stresses. These stresses are often the most obvious consequence of adding second-phase particles to a matrix and can have a profound effect on properties. Factors determining the toughness, strength and wear resistance of particulate composites are then considered in turn, and the chapter concludes with an assessment of possible future developments in this area.
4.2
Powder processing and microstructural development
Unlike fibre- or whisker-reinforced composites, particulate composites have the advantage of being compatible with conventional powder processing, and in many cases can be pressurelessly sintered. As with other ceramic microstructures, a myriad of other ingenious fabrication routes have also been reported, but these are too numerous and system-specific to describe here. This section merely outlines the main points of powder processing where the production of composites in chemically compatible systems (i.e. those in which the components do not react chemically with one another) differs from that of monolithic ceramics.
© Woodhead Publishing Limited, 2006
Particulate composites
101
We begin with milling and dispersion of the powders in a liquid. In addition to the role of breaking down hard agglomerates, as for monolithic ceramics, this step must also thoroughly mix the component powders of the composite. For composites in which the particulates need to be relatively large, however, it is important not to reduce the mean size of the particulates by using a milling treatment that is too aggressive or very long in duration. More careful control of the milling procedure is often required than for monolithic ceramics. The particulate and matrix powders usually have differing surface responses in the development of electrical double layers or adorption of steric dispersants during milling. This does not usually degrade their ability to mix. It might even help it if, for example, the components develop opposite surface charges, though this might affect the viscosity and achievable solid loading of the slurry. Problems arise mainly when one of the components shows no response to these deagglomeration mechanisms. In these cases the remedy is the same as for monolithic processing, namely to try different liquids or dispersants, or to coat the particles with a substance that does offer a response. The other principal differences between monolithic ceramics and powder composites occur during sintering. When a particulate second phase that is considerably larger than the matrix powder is incorporated into the green body, it represents a region that will not shrink with the matrix as sintering takes place. The resulting mismatch in shrinkage inhibits sintering of the matrix and can also lead to stresses sufficient to cause cracking [2]. The diffusional fluxes during sintering can also relax the stresses in the matrix [3], however, essentially through simultaneous diffusion creep. This can be sufficient to enable sintering to proceed to completion and for cracking to be avoided. There are many examples of pressurelessly sintered composites containing relatively large particles [4–7]. When the particulate phase is smaller than, or of comparable size to the matrix powder, this source of inhibition does not arise. If the particulate phase has similar diffusional properties to the matrix at the sintering temperature, sintering can actually be improved because the particles oppose grain growth by pinning the grain boundaries [8]. Examples of this type of composite include Al2O3–ZrO2 and Al2O3–Cr3C2. The ability of the particles to participate in diffusion usually means that they are mobile. Grain growth is therefore not entirely prevented and the particles are dragged around by the migrating grain boundaries, coalescing in the process, so that typical final microstructures are characterised by rounded particles of equilibrium shape, often coarser than the original particles added as powder, and predominantly situated at grain boundaries and (especially) triple lines. The addition of fine particles that are much more refractory than the matrix has a different effect. The outstanding example of this behaviour is in alumina/SiC ‘nanocomposites’ [9]. The SiC/Al2O3 interface may have a low
© Woodhead Publishing Limited, 2006
102
Ceramic matrix composites
diffusion coefficient, and it is widely accepted [10, 11] that it may be difficult to remove matrix material from the particle matrix interfaces, so that the essential sintering process of removal of material from the grain boundaries and deposition in the pores is severely impeded. Figure 4.1 shows that the addition of only 2.5 wt% SiC severely impedes sintering, and further additions continue to reduce the sintered density. Another consequence of the inability of the SiC/Al2O3 interface to participate in diffusional processes is that the particles are immobile. They therefore retain their original size, are uniformly distributed and are situated both within the grains and on the grain boundaries. They pin grain growth more strongly than mobile particles, approximately in accord with Zener’s theory [12, 13]. This striking property of the Al2O3/SiC interface can be understood in terms of the observation of Ashby and Centamore [14] that the more refractory of two phases at an interface (the covalently bonded SiC in this case) controls the interface reaction because in general atoms in both phases must be involved in the reaction. The majority of the Al2O3/SiC interfaces in the nanocomposites have been observed to be free of any glassy phase, the presence of which would presumably allow alumina to be removed or deposited at the interface without the involvement of the SiC, and consequently much more rapidly. The introduction of an interfacial layer may be the source of the ability of sintering aids such as Y2O3 to enable these materials to be pressurelessly sintered [15, 16] (Fig. 4.2). In conclusion, particulate composites are more difficult to process using powders than monolithic ceramics, but are easier than other kinds of composite nevertheless. They can often be sintered to full density without pressure. When this is not possible, sintering aids or the superimposition of pressure (hot pressing, hipping) can be used to alleviate the problems, and there are many examples of this in the literature and in commercial practice.
Density (%)
100 95 90 85 80 0
2
4
6 wt% SiC
8
10
4.1 Sintered density (1700°C, 2 hours) against SiC content for alumina/SiC nanocomposites, demonstrating the inhibition of sintering caused by the SiC particles [10].
© Woodhead Publishing Limited, 2006
Particulate composites
103
100
Density (%)
95
90
1650°C 1600°C 1550°C 1500°C 1450°C
85
80 0.0001
0.001
0.01 0.1 % Yttria
1
10
4.2 Effect of yttria additions on sintered density of alumina–2% SiC nanocomposites for various sintering temperatures (data from [16]). The ‘0.0001%’ points indicate no added yttria.
4.3
Thermal microstresses
If the matrix and particles in a composite have different thermal expansion coefficients then thermal microstresses develop during cooling from processing temperatures. These stresses can be very large in particulate ceramic composites, firstly because the processing temperatures are high so that the temperature change on cooling is large, secondly because ceramics are typically very stiff so that a large stress develops for a given thermal expansion mismatch, and thirdly because, unlike metals, most ceramic phases have little scope for plastic relaxation of the stresses during cooling, at least below 1000°C. The change of these stresses during cycling of a MgO–SiC ‘nanocomposite’ from room temperature to 1550ºC and back again is demonstrated in Fig. 4.3 [17] from which it can be seen that the stress level in the SiC particles is almost 4000 MPa at room temperature. For most particulate composites the mismatch between the particles and the matrix is more important than the anisotropy of either component (though alumina/aluminium titanate composites provide a notable exception and are described below). The main features of the stresses can therefore be understood in terms of a simple elastic model assuming thermoelastic isotropy and consisting of a spherical particle in a concentric spherical shell of matrix with dimensions chosen to give the appropriate volume fractions. The particles are predicted to be under a uniform hydrostatic stress, σp, after cooling. If the particles have a larger thermal expansion coefficient than the matrix, this stress is tensile, and vice versa. For small particle volume fractions the stress
© Woodhead Publishing Limited, 2006
Ceramic matrix composites Average hydrostatic stress (GPa)
104
0 Error = 0.52 GPa –1
–2 Heating –3
Cooling Predicted
–4 0
500 1000 Temperature (°C)
1500
4.3 Hydrostatic stress in the SiC particles in a magnesia–10%SiC nanocomposite during a thermal cycle to 1550°C. Note the small amount of relaxation during the cycle and the good agreement with the prediction of a simple elastic model for the stresses [17].
state in the matrix immediately outside the particle is purely deviatoric with a radial stress of σp, and hoop stresses of –σp /2. These matrix stresses fall away rapidly as 1/r3 where is r is the radial distance from the particle centre. For higher volume fractions of particles, a uniform hydrostatic image stress is superimposed on these stresses, of opposite sign to the particle stress. Thermal microstresses influence the fracture of composites. Davidge and Green [18] showed the ability of the stresses to deflect cracks using glass matrices with a range of expansion coefficients containing large thoria spheres. The model stress field described above shows that when the particle thermal expansion coefficient is larger than that of the matrix, the radial stress close to the particles is tensile so that cracks are deflected away from the particles. With the reverse expansion mismatch, cracks are attracted towards the particles. This deflection of crack path is of relevance to the apparent toughness of the composite and to the formation of microstructural elements bridging the crack interfaces. Even on a planar crack path, the fluctuations between tensile and compressive thermal stresses influence the propagation of the crack, which is impeded when passing through the compressive regions. A further effect of thermal stresses is that when combined with the crack tip stress field they encourage the formation of a microcracked zone around the crack, and the accompanying dilatation can shield the crack tip from the externally applied stress. These toughening mechanisms are described in more detail in the next section. As well as producing these broadly beneficial effects, thermal microstresses can also degrade the strength of composites. The tensile components of stress can help in crack initiation. In a composite with a uniform distribution of particles, the tensile components act only over distances comparable with
© Woodhead Publishing Limited, 2006
Particulate composites
105
the particle spacing, but non-uniform distributions of particles can lead to mean tensile stresses over considerably greater distances, comparable with the scale of the local volume fraction variations. The most obviously deleterious effect of the thermal stresses, however, is the possibility that they are sufficiently large to cause spontaneous microcracking during cooling from the processing temperature. Davidge and Green [18] first pointed out that since the stored elastic strain energy scales with the particle volume, and the energy required to create new crack surfaces scales with its area, there is a critical particle size for spontaneous cracking below which there is not enough energy available to allow a crack to form. This is why the MgO/SiC nanocomposite containing thermal stresses approaching 4 GPa described above did not crack spontaneously, despite the fact that the microstresses were considerably larger than the macroscopic strength of the material. A detailed study of Al2O3– 20%SiC composites [19] showed that there was an abrupt transition from negligible microcracking with small SiC particles (Fig. 4.4(a)) to general cracking of the alumina matrix when a mean particle size of 10 ± 3 µm was exceeded (Fig. 4.4(b)). Figure 4.4(b) shows that the microcracks run radially from the particles as is consistent with the tensile hoop stresses in this system (αSiC < αalumina). This behaviour was explained by a fracture mechanics model which correctly predicted the critical particle size, and showed that once nucleated, microcracks can propagate from one particle to the next so that microcracking can spread throughout the material from a small number of nucleation sites when the median particle size exceeds this critical value.
4.4
Toughening
One of the primary motivations for the deliberate addition of second-phase particles to a ceramic matrix is to increase its toughness. If the particles are tougher than the matrix then the crack resistance energy, R, will be increased, approximately according to the rule of mixtures if the crack simply passes through the particles and the difference in toughness between the particles and the matrix is relatively small. This is obviously of limited value, since the composite cannot exceed the toughness of the particles. The composite approach is much more powerful if it causes new mechanisms to operate that either do not occur or are weak in single phase materials. The following toughening mechanisms have been investigated for non-transforming particulate composites.
4.4.1
Crack deflection
All else being equal, a toughening effect occurs if the crack tilts or twists away from a planar geometry because this reduces the net crack driving force. In homogeneous materials such as glass, cracks tend to propagate in
© Woodhead Publishing Limited, 2006
106
Ceramic matrix composites
(a)
(b)
4.4 (a) Backscattered SEM micrograph of unetched alumina–3 µm SiC composite. The straight lines on the specimen surface are scratches from metallographic preparation. (b) Unetched alumina–13 µm SiC composite showing radial microcrack network in the matrix [19].
a planar fashion for the same reason, but non-uniform features such as weak interfaces and residual stresses can lead to such a deflection in other materials. These may occur in single-phase polycrystals, but there is scope for augmentation of the effect in composites, and particles with higher stiffness than the matrix can also lead to deflection. According to the purely geometrical analysis of Faber and Evans [20], spherical particles are less effective than plate- or rod-shaped particles and lead to a maximum increase in apparent toughness, Kc, of around 30%. Even this modest increase represents an overestimate of the effect, however, since it does not take into account the
© Woodhead Publishing Limited, 2006
Particulate composites
107
reason for the deflection, which occurs because it is easier for the crack to deflect than to propagate in a planar fashion. This may be because the nonplanar crack path has a lower toughness (weak interfaces), because the residual stress provides extra driving force if the crack is deflected (thermal expansion mismatch), or because the strain energy release rate is greater in the direction of deflection (stiff particles), the implication being that the measured toughness would be higher if the crack remained planar. A complete argument should consider the driving force required to cause crack propagation at every point on the crack path, but it is clear that although crack deflection is important in understanding the net toughness of a composite exhibiting this effect, it is not itself a potent toughening mechanism.
4.4.2
Crack bridging
If intact or interlocking ligaments remain behind the advancing crack front, the restraining force they exert reduces the stress intensity at the crack tip, causing an increase in the macroscopically measured toughness. Because the bridges accumulate behind the crack front, the toughening effect increases as the crack propagates, a phenomenon known as R-curve behaviour. Crack bridging is a very potent toughening mechanism in long-fibre composites and operates in a similar manner with whisker reinforcements [21]. These reinforcement geometries are particularly conducive to crack bridging, but the mechanism can also operate in less favourable situations. Crack deflection along weak interfaces can lead to bridging through geometrical interlocking and causes toughening in monolithic alumina exhibiting intergranular fracture [22, 23]. The presence of particulate reinforcements can enhance this effect. If a particle is to act as a bridge, the key requirement is that the crack path must be deflected around its periphery and in doing so tilt or twist through 90º or more to form an interlocking section. The main factors determining whether or not this is possible are (i) the relative toughnesses of the matrix, the particle and the interface, (ii) the residual stress state around the particle, and (iii) the size of the particles. It is important that the interface is relatively weak. If, for example, the interface and matrix are as tough as the particle, the crack will go through the particle instead of around it. If the particle is tough but the interface is only marginally weaker than the matrix, the crack will tend to detach from the particle instead of undergoing the severe deflection required for interlocking to occur [24]. Particles with thermal expansion coefficients greater than that of the matrix will have tensile stresses across the interface, effectively weakening it, thus favouring bridge formation. The influence of residual stresses on a crack path increases with the distance over which the stresses act, so this effect is more important with bigger particles. The balance between these different considerations can vary considerably between different composite systems. Figure 4.5, for instance, shows
© Woodhead Publishing Limited, 2006
108
Ceramic matrix composites
(a)
(b)
4.5 Indentation crack paths in alumina matrix composites containing 20 vol% of particles of (a) TiN and (b) Cr3C2. The alumina matrix is the darker phase in each case.
indentation crack paths in alumina matrix composites containing 20 vol% of (a) TiN and (b) Cr3C2 particles [25]. The particles are of similar size in the two composites, and neutron diffraction measurements of the residual stresses showed that both types of particle were in tension with a mean stress close to 400 MPa. Despite these similarities, Fig. 4.5 shows a profound difference in the crack paths in the two composites. The crack tends to cut through the particles in the alumina/TiN composite (Fig. 4.5(a)), indicating a relatively strong particle–matrix interface. In contrast, the crack goes around the periphery of all the particles encountered in the alumina/Cr3C2 composite (Fig. 4.5(b)) even when large tilts are required to do so, and this has resulted in extensive
© Woodhead Publishing Limited, 2006
Particulate composites
109
crack bridging. Fracture of the surrounding matrix has occurred as the interlocking areas have been pulled apart. This behaviour must be attributed to a particle/matrix interface that is considerably weaker than that in the alumina/TiN composite. The differing crack behaviour in these materials is reflected in the toughening increment relative to pure alumina that the particles produced. The addition of Cr3C2 increased the toughness considerably, ∆Kc ~ 6 MPa m1/2, whilst TiN particles give only a small toughening effect compared with alumina, ∆Kc < 2 MPa m1/2. This demonstrates the ability of crack bridging to cause significant toughening, even in particulate composites.
4.4.3
Microcrack toughening
Section 4.3 showed how thermal microstresses in particulate ceramic composites can cause spontaneous microcracking when the particles exceed a critical size. For composites in which the particles are below the critical size for spontaneous fracture, the imposition of additional stress can lead to stress-induced microcracking. A potential consequence of this is the development of a process zone of microcracked material ahead of the crack tip. The consequent reduction in modulus ahead of the crack tip reduces the stress intensity [26], though this small effect is countered by the reduction in toughness as a result of the microcracking. The energy dissipated in the wake of a propagating crack as the newly microcracked material is unloaded provides a stronger effect. This originates both in the irreversible dilatation of the material as the microcracks form in a manner analogous to the transformation toughening of zirconia, and in the accompanying reduction in stiffness. The existence of a microcracked zone around cracks in SiC–TiB2 composites, in which the thermal expansion mismatch puts the TiB2 particles in a state of tension after sintering, has been confirmed using small-angle X-ray scattering and by direct observation in the TEM [27]. The toughness was observed to increase with crack propagation (R-curve behaviour), as would be expected by this mechanism which, like crack bridging, relies on the development of features behind the crack tip. The observations of microcracking were used to estimate the extent of microcrack toughening expected, and the results were of similar magnitude to the measured toughening increments, defined as the difference between the toughness on initial crack propagation and the plateau value at large extensions. Since the observed toughness increases were all less than 2 MPa m1/2, however, it is difficult to separate unequivocally the contribution of microcracking from those of other mechanisms capable of causing R-curve behaviour such as crack bridging, which would also be favoured by the tendency of the residual stress to aid circumferential crack formation.
© Woodhead Publishing Limited, 2006
110
4.4.4
Ceramic matrix composites
Direct influence of thermal microstresses
As a loaded crack propagates through a particulate composite, any thermal microstresses of the type described in Section 4.3 will contribute to the total stress intensity at the crack tip. Stresses applied immediately behind the crack tip make the biggest contribution to the stress intensity, so the crack can be expected to propagate easily through regions of residual tension, and conversely will be impeded in regions of residual compression. In order for significant crack extension to take place, the externally applied stress intensity must be sufficient to force the crack front through the compressive regions, and the apparent toughness will therefore be higher than it would be without the microstresses. This principle applies regardless of the sign of the thermal expansion mismatch because the average microstress on a planar surface through the composite must be zero; any region of tension must therefore be balanced by a region of compression, though the details of the crack geometry at the critical position of maximum resistance can be expected to differ according to whether the particles are in compression or tension. A simple estimate for the toughening increment, ∆Kc, attainable by these means can be obtained by assuming that the regions of compression comprise a uniform microstress, σ, acting over a distance d behind the crack tip. Ignoring microstresses acting further behind the crack, which is assumed to be long, gives [28]:
∆K c = 2σ
2d π
(4.1)
The mean compressive stress in the matrix of the composites shown in Fig. 4.5 is ~100 MPa, and the distance between particles ~5 µm. Inserting these values into eq. (4.1) gives a toughening increment of only 0.4 MPa m1/2, demonstrating that this mechanism makes only a weak contribution to the toughness of most particulate composites. In summary, crack interface bridging provides the most effective toughening mechanism in non-transforming particulate ceramic composites and can produce significant toughening increments compared with the unreinforced matrix, particularly when combined with several weaker toughening mechanisms which are also known to operate. The aim of increasing the toughness is to influence the more directly applicable properties of ceramics, and the next section examines the effect of particulates on their strength.
4.5
Room-temperature strength
4.5.1
Microcomposites
The aim of toughening a matrix by adding second-phase particles is essentially to increase its strength. For cracks in homogeneous materials under uniform
© Woodhead Publishing Limited, 2006
Particulate composites
111
loading the toughness, Kc, and strength, σf, are related by the following expression: K c = Y σf c
(4.2)
where c is the critical crack length and Y is a factor depending on the geometry of the crack. Equation (4.2) predicts that the strength should be proportional to the toughness assuming that the particulate additions do not affect the flaw size. In some cases, this prediction is borne out in practice, as is demonstrated in Fig. 4.6 for glass–alumina particle composites [29]. When combined with the value of Y appropriate for semicircular edge cracks, the gradient of the line in Fig. 4.6 indicates a crack radius of about 100 µm. In most cases, however, the strength change on the addition of particles that increase the measured toughness is less than this naive prediction would suggest, and often the strength is actually reduced. There are several related reasons for this. One which is obvious from eq. (4.2) is that the addition of the particulate may increase the critical flaw size as well as increasing the toughness. This may occur because of some of the processing problems mentioned in Section 4.2, most simply the presence of pores arising from the inhibition of sintering. Non-uniform particle dispersions can also increase the effective flaw size dramatically. Clusters of particles can fail to sinter properly if they are refractory and act as critical flaws. Less extreme local volume fraction variations can also lead to large strength reductions because of the possibility of differential shrinkage cracking during sintering. In systems with a pronounced thermal expansion mismatch between the particles and the matrix, regions that are particularly rich or deficient in particles lead to thermal stresses that can aid crack initiation. The preceding reasons for strength reductions are a consequence of microstructural deficiencies, but even ‘perfect’ microstructures, in which the critical flaw size is the same as for the unreinforced matrix material, can yield smaller strength increases from particle-induced toughening than would
Strength (MPa)
150
100
50
0 0
0.5 1 1.5 Toughness (MPa m1/2)
2
4.6 Relationship between strength and toughness for glass/alumina composites with best fit straight line to eq. (4.2). Data from [29].
© Woodhead Publishing Limited, 2006
112
Ceramic matrix composites
be predicted on the basis of eq. (4.2). This is because the most potent toughening mechanisms, crack interface bridging and microcrack toughening, rely on processes that occur behind the crack tip, and consequently exhibit R-curve behaviour, as described in Section 4.4.2. Taking the example of a particulate composite toughened by bridging of the crack by the particles, as an initially small crack grows between particles the toughness is simply that of the matrix. If tensile thermal stresses are present in the matrix, the effective toughness at first gets smaller as the crack grows longer because these internal stresses aid crack propagation. After further propagation, however, the crack tip passes a particle or other bridging element which subsequently exerts a closure force on the crack, causing an increase in the macroscopically measured toughness. As crack propagation continues, more bridges are formed and the toughness continues to rise until the bridges that were formed first are broken, beyond which point a steady state is established and the toughness becomes constant. Detailed consideration of the strength under these conditions (see, e.g., [30]) shows that for small critical flaws, the strength is determined solely by the toughness versus crack length behaviour in the very early stages of crack growth, before any bridges have been formed. Conversely, the strength for very large flaws is determined essentially by the plateau toughness. The strength for intermediate flaws depends on the form of the toughness–crack length relationship as it rises steeply after formation of the first bridges. In this region, some stable crack growth can occur prior to final failure as the toughness rises more quickly with crack growth than the applied stress intensity, and the strength becomes insensitive to flaw size. Combining the information in the previous two paragraphs enables the following common observations regarding strength–toughness relationships in particulate composites to be explained straightforwardly: • Most standard methods of toughness measurement rely on the use of large flaws and therefore measure values close to the plateau toughness. The lack of correlation between measured toughness and strength that is frequently observed can partly be attributed to the fact that flaws in particulate composites are usually sufficiently small for the strength to be determined by the toughness during initial crack propagation rather than by the plateau toughness. • If there are no residual stresses in the composite, the strength of samples containing only very small flaws can be expected to be similar to that of the matrix alone. If tensile stresses are present because of the thermal expansion mismatch between the matrix and the particles, the initial reduction in apparent toughness during crack growth means that the strength of materials containing small to moderate critical flaws may be lower than that of the pure matrix, even for a composite with a homogeneous
© Woodhead Publishing Limited, 2006
Particulate composites
113
particle distribution, and notwithstanding the fact that the toughness for much longer cracks can exceed that of the matrix by a considerable margin. • For a given composite system, the reduction in strength compared with the unreinforced matrix material is often greatest for the composites showing the greatest amount of steady-state toughening because bridging is most effective with large particles owing to the retention of high bridging forces for greater crack face separations. Larger particles also imply an increased particle spacing for a given volume fraction, however, so cracks must grow further before encountering the first bridging element that is responsible for the increasing toughness. • Finally, materials with very large flaws, such as may result from heavy surface damage, benefit most from the toughening effect. The weakest specimens in a batch of particle-toughened composites may, therefore, be considerably stronger than those in a batch of pure matrix material, and the range of strengths present is reduced. This is most important from the point of view of applications for these materials because it is the weakest specimens that determine the usable strength, and the smaller range of strengths present and resistance to damage are attractive to designers. The Al2O3–Al2TiO5 particulate composites investigated by Bennison and co-workers [7, 31] provide an excellent illustration of several of these effects. The volumetric thermal expansion of aluminium titanate, which has an orthorhombic crystal structure, is similar to that of alumina, but unlike alumina it exhibits extreme anisotropy of linear expansion. This leads to sizeable thermal stresses in the composites which are at least partly responsible for the significant crack interface bridging observed [7]. Figure 4.7 shows a log–log plot of strength against indentation load for biaxial bend tests with Vickers hardness indentations at the point of maximum tensile stress. The indentations are a method of introducing controlled flaws to the specimens, and simulate contact damage. The experimental points on the plot are from an Al2O3–20vol%Al2TiO5 composite with a grain size of 6 µm. The mean strength of this material when unindented was ~250 MPa. This is considerably below the strength of pure alumina with the same grain size, which can be estimated as ~510 MPa by interpolation of strengths obtained for other grain sizes in the same laboratory [32]. The strength of the composite is remarkably insensitive to indentation, however, and remains substantially unaltered even with indentation loads as large as 300 N. The experimental curve (A) for alumina with a grain size of 2.5 µm and the prediction for an alumina with the same grain size as the composite, 6 µm (B), are substantially more sensitive to indentation load, and for indentation loads of around 30 N or more become weaker than the composite, the strength falling to around half that of the composite for the highest indentation loads used. Alumina with a grain size of 80 µm (C) is also flaw tolerant, but is significantly weaker than
© Woodhead Publishing Limited, 2006
114
Ceramic matrix composites Indentation dimension (µm) 10 30 100
800
300
A
Strength (MPa)
600
B
400
200 C
100 10–1
100 101 102 Indentation load (N)
103
4.7 Plot of strength versus indentation load for alumina with various grain sizes and alumina/aluminium-titanate composite. The data points and solid line are from the composite. The other lines are for alumina with various grain sizes: A, 2.5 µm, experimental; B, 6 µm, interpolated; C, 80 µm, experimental (reproduced from Bennison et al. [7] by kind permission of Taylor and Francis Ltd (http:// www.tandf.co.uk/journals).
the composite. Another attraction of this composite system is that it can be pressurelessly sintered in air, either by mixing Al2O3 and Al2TiO5 powders, or by reaction sintering [33].
4.5.2
Ceramic nanocomposites
Figure 4.7 shows that particulates of several micrometres in dimension can yield desirable properties such as flaw tolerance, but there is a limit to how far this approach can be extended owing to the inverse relationship frequently observed between strength and toughness. An alternative approach has been pioneered by Niihara and his co-workers, who in 1991 published a review of results from a variety of ‘ceramic nanocomposites’ [34], i.e. composites in which at least one of the phases is ‘nanoscale’. The most striking results are from microstructures consisting of alumina grains of conventional size (a few microns) containing SiC particles with a mean diameter of ~0.25 µm. The addition of 5 vol% submicron SiC to alumina was reported to increase its strength by a factor of three, from 350 MPa to 1050 MPa. Annealing of the composites increased the strength further to 1520 MPa. These results have proved to be controversial because they have been difficult to reproduce, although it should be stated at the outset that the high strength of Niihara’s materials has been verified independently. The main area of uncertainty is over the extent of the strength increase on adding SiC
© Woodhead Publishing Limited, 2006
Particulate composites
115
(in the absence of any annealing treatment). Subsequent investigations by other workers in which the strengths of alumina/SiC nanocomposites have been compared with those of alumina processed in the same way and with the same grain size and grain size distribution (i.e. no abnormal grain growth in either material) also show the nanocomposites to be the stronger material, but typically by only 10–50%, e.g. [13, 35, 36]. Several factors appear to contribute to this inconsistency. One is that the extent of the strength increase originally reported by Niihara [34] is partly attributable to the unusually low strength of the alumina used in the comparison, as well as to the high strength of the composite. Such low strengths in alumina are normally indicative of processing defects such as porosity or abnormally grown grains, neither of which was present in the nanocomposites. At the other end of the scale, strengths of ~1000 MPa have previously been achieved in monolithic alumina by using a special processing technique to break down powder agglomerates [37]. Though this processing method was not used in producing the nanocomposites in Niihara’s work [34], it is clear that a full assessment of the effect of the SiC nanoparticles on strength can only usefully be made using materials that differ only in whether they contain SiC or not. Even then, composite results from different sources may differ in the size and distribution of the SiC particles. Agglomerations of SiC particles may be capable of acting as critical flaws. The distribution of particles, and consequently the strength of the composite, will thus depend on fine details of the processing that are difficult to reproduce precisely from one laboratory to another [10]. The variable nature of the strength increases in these particulate nanocomposites has also made it difficult to identify the explanation for them with certainty, though there is wide agreement that the SiC produces no significant toughening [13, 35, 36, 38] and that the explanation therefore lies elsewhere. There is some evidence that crack initiation may be inhibited in the nanocomposites [38], and several explanations for this have been put forward based on interactions between cracks and the large thermal stresses (~ −2 GPa in the particles [39]) which are one of the most obvious consequences of the SiC particle additions [40, 41]. Another important factor is undoubtedly that surface damage and machining stress development during grinding and polishing differ dramatically between alumina and the nanocomposites, with the nanocomposites exhibiting significantly less surface cracking and pullout. If the critical defects are machining-induced, this can be expected to explain at least partially the increased strength. The response to surface abrasion is described more fully in Section 4.7.2. The further increase in strength on annealing reported by Niihara has been reproduced in several investigations [35, 42, 43]. The extent of the strength increase depends on the surface finish [42], and there is convincing evidence that this is caused by crack healing, aided by a glassy phase resulting
© Woodhead Publishing Limited, 2006
116
Ceramic matrix composites
from oxidation of SiC particles near the surface that fills the cracks and bonds the faces together [42]. This has been shown to occur even in atmospheres such as standard laboratory argon that are nominally inert. The few reports in which annealing did not increase the strength (e.g. [38]) may be a consequence of subtle differences in the annealing atmosphere (e.g. lower oxygen partial pressure), or failure from subsurface cracks where oxidation cannot take place. This self-healing effect brings to oxide ceramics a potential advantage otherwise available only to silicon-containing non-oxides such as Si3N4. In conclusion, the strengthening of alumina/SiC ‘nanocomposites’ looks set to remain controversial owing to its capricious nature. Commercially, the room-temperature strength increase is not sufficiently large for it to repay the extra cost of processing compared with alumina, since the nanocomposites require an inert atmosphere and slightly higher temperatures, even with sintering aids such as yttria, which have been shown to alleviate the inhibition of sintering caused by the SiC additions. The high-temperature properties and the polishing and wear behaviour of the nanocomposites offer much more significant improvements, however, which may well be cost effective (see Sections 4.6 and 4.7.2).
4.6
High-temperature strength
One of the main drivers for the application of ceramics is their ability to maintain their strength at high temperature. In monolithic ceramics without potent toughening, there are several stages of high-temperature behaviour. At moderate temperatures, below the level at which solid-state diffusion or other high temperature mechanisms become significant, standard measurements of toughness and strength show little temperature dependence, although slow crack growth may be accelerated considerably, particularly in oxide ceramics when crack growth is caused by the interaction of water vapour with the material at the crack tip. Similarly, composite systems with a small thermal expansion mismatch between the phases such as Al2O3–TiC exhibit neither thermal residual stresses, nor in this case strong toughening mechanisms, so the toughness is moderate and independent of temperature until new mechanisms operate at high temperature. In particulate composites exhibiting strong toughening mechanisms such as crack bridging and stress-induced microcracking, a more marked change in toughness and strength might be expected at moderate temperatures owing to the reduction in the thermal residual stresses locked into the microstructure. This would clearly inhibit stress-induced microcracking and may also reduce both the number of bridging elements formed and the closure force they exert if the thermal stress clamps them in place. The more minor direct toughening effect of the fluctuating residual stress field would also be reduced.
© Woodhead Publishing Limited, 2006
Particulate composites
117
SiC–TiB2 particle composites provide a good illustration of this, since microcrack toughening, crack bridging and residual stress toughening are all expected to operate at room temperature. Jenkins, Salem and Seshadri [44] found that the long-crack (chevron notch) fracture toughness of such composites fell gradually as the temperature was increased from room temperature to 1400°C, at which temperature the toughness was close to that of monolithic SiC (Fig. 4.8), indicating that the toughening mechanisms operative at room temperature had ceased to be effective. Interestingly, toughness values for the SiC–TiB2 particle composites in [44] derived from SENB tests using blunt notches made with a 300 µm diamond saw did not exhibit such a marked reduction with temperature. Similarly, McMurtry et al. [6] found that the flexural strengths of similar composites was independent of temperature between room temperature and 1200ºC. This suggests that the initial portion of the R-curve, which determines the strength when failure is from small flaws such as those at the tip of a sawn notch or the surface of a flexural strength specimen, is not greatly influenced by the toughening mechanisms mentioned. At very high temperatures, typically in excess of 1000ºC, the deformation and fracture behaviour of monolithic ceramics becomes complicated by the operation of new mechanisms such as solid-state diffusion, grain boundary sliding, the activation of dislocation slip systems, the melting of thin grain boundary films, and oxidation. All of these can also occur in particulate ceramic composites. One example of such effects is the observation of a sharp toughness increase, which is well known to be caused by crack blunting or healing associated with softening of grain boundary phases, followed by a rapid loss of strength with further temperature increases as the grain boundary phase loses its strength completely [45]. This has been observed in Si3N4– TiC and Al2O3–TiC composites by Baldoni et al. [46], who point out that it is unlikely that the particulate reinforcement plays an important role. At
Toughness (MPa m1/2)
4.5 4 3.5 3 2.5 2 0
500 1000 Temperature (°C)
1500
4.8 Fracture toughness against testing temperature for a commercial SiC–16vol%TiB2 particle composite tested using chevron notched beams in three-point bending. Data from [44].
© Woodhead Publishing Limited, 2006
118
Ceramic matrix composites
higher temperatures, creep and the associated cavitation or cracking associated with grain boundary sliding can lead to composite failure in much the same way as for monolithic ceramics. Although the processes occurring in particulate composites at high temperatures qualitatively resemble those in monolithic ceramics, there are nevertheless several examples of particulate additions leading to significant property improvements. Modifications to the grain boundary structure, associated phases or segregants are often involved. French et al. [47], for instance, have reported that duplex microstructures consisting of equal proportions by volume of Al2O3 and yttrium aluminium garnet (YAG) exhibited creep rates at 1250ºC that were slower than those of ‘pure’ alumina and YAG by factors of 20 and 4 respectively. This was explained by the ability of yttria additions to reduce the creep rate of alumina by approximately two orders of magnitude. When the alumina in the composite was considered as being yttria-doped, the composite obeyed the rule of mixtures. Although the lowest creep rate was obtained from single-phase yttria-doped alumina, the composite might be preferable in some situations because of the increased microstructural stability conferred by the duplex structure. The composite suffered negligible grain growth during the creep tests, for instance, but the grain size of the yttria-doped alumina increased noticeably. This can be attributed to the greater diffusion distance required for grain growth in multiphase structures. Another notable example of a reduction in creep rate through the addition of second-phase particles concerns ‘nanocomposites’. In alumina–SiCn systems, several investigations have reported significant reductions in creep rate compared with monolithic alumina [48, 49]. Figure 4.9 shows the results of Ohji and co-workers [48]. At 1200ºC the creep rate of an Al2O3–17vol%SiC nanocomposite was less than that of alumina for a given stress by a factor of 250, and the time to rupture at 50 MPa was increased from 120 h to 1120 h. The SiC inhibits creep primarily because it is difficult to remove or deposit
Log strain rate (s–1)
–5
alumina
nanocomposite
–6 –7 –8 –9 – 10 1.4
1.6
1.8 Log stress (MPa)
2
2.2
4.9 Log–log plot of tensile creep strain rate against applied stress for alumina and an alumina–17vol%SiC nanocomposite tested in tension at 1200°C.
© Woodhead Publishing Limited, 2006
Particulate composites
119
material at the interface with the alumina matrix to allow diffusional transport to occur. Intergranular particles therefore inhibit diffusion creep and grain boundary sliding in the same way that they inhibit sintering (Section 4.2) and, through their consequent immobility, prevent grain growth. The smaller improvement in time to rupture in these observations shows that the strain to failure was reduced by the SiC additions. This is attributable to the nucleation of cavities at the intergranular SiC particles. The suppression of creep has also been reported in Si 3 N 4 –SiC nanocomposites, and similar explanations have been given [34, 50, 51], although others have found no improvement [52]. The reasons for these discrepancies have yet to be resolved, but it is likely that they originate in the different processing methods and sintering aids used in producing these materials and hence the differences in grain boundary phases, as well as in the wide variety of other additive-induced microstructural variations possible in Si3N4 materials (e.g. the presence of elongated, whisker-like grains). As well as being used to inhibit creep, second-phase particle additions can be used under different conditions to achieve the opposite, in fabricating ceramic microstructures that enable superplastic deformation. This term refers to the ability to achieve large, uniform tensile elongations (100%) at moderate strain rates (10–5–10–4 s–1) without failure. The underlying mechanism of this type of deformation involves diffusion, and the main requirements are that a fine grain size (of the order of microns or finer) can be maintained at the high temperatures necessary to give rapid deformation at sufficiently low stresses to avoid failure. A common strategy for producing and maintaining a fine grain size is to use microstructures comprising two or more mutually insoluble phases, often in roughly equal volume fractions. This severely limits grain growth as described in connection with duplex Al2O3–YAG composites above, and many superplastic particulate composites have now been reported. An alternative method of maintaining a fine grain size is to use a lower volume fraction of fine second-phase particles which can restrict grain growth by Zener pinning. These include ZrO2–Al2O3 [52], ZrO2–mullite [54] and Si3N4–SiC [34, 55]. Perhaps the most impressive results to date are those of Hiraga and co-workers [56], who have fabricated three-phase microstructures consisting of zirconia, magnesium aluminate spinel and αalumina with a mean grain size of around 200 nm which are capable of tensile elongations of >1000% with a strain rate of 0.4 s–1 at a temperature of 1650°C. These exceptional results stem from the combination of the high testing temperature and very fine and stable grain size that the three-phase structure allows. Superplastic ceramics have several obvious potential advantages for commercial application. These include net size and shape forming and the possibility of forming complex components from initially flat sheets. Whilst the practical problems of forming at temperatures in excess of 1200°C obviously
© Woodhead Publishing Limited, 2006
120
Ceramic matrix composites
add cost to the process, the diamond machining which is the only practical competitor for the production of many complex shapes to high dimensional tolerance is also expensive. Despite these attractions, the phenomenon remains a scientific curiosity at the time of writing. This is much the same as the situation for superplastic metals until the late 1960s, when a few practical demonstrations of their commercial benefits led to their widespread application. It remains to be seen whether the industrial superplastic forming of ceramics will take off in the same way.
4.7
Wear
4.7.1
Microcomposites
Another primary motivation for the use of ceramics in engineering applications is their high wear resistance. At its simplest, wear involves plastic deformationcontrolled mechanisms such as cutting or ploughing and, in ceramics, the removal of pieces of material by brittle fracture (‘pullout’). This is the origin of figures of merit for wear of the form K cm H n , where Kc is the toughness and H the hardness, and m and n are positive exponents. In reality, however, these wear mechanisms are much more complex than this suggests, with the formation of modified surface microstructures and compacted layers being common, and additional mechanisms such as chemical interaction between ceramic and substrate, or atmosphere and ceramic, are frequently important. The high temperatures generated locally during the wear process add to this complexity. Even in cases where it can be argued that the simple plasticity or brittle fracture mechanisms are dominant, the appropriate values of Kc and H to use in models are not clear, as the scale of the plastic deformation or fracture is much smaller than that in tests used for the measurement of these properties, and the temperature at which these properties should be measured is ill defined. Furthermore, the dominant mechanism and the rate at which it operates depend not only on the ceramic itself, but on the wear conditions and substrates involved. Many of the reports of wear tests on particulate ceramic composites are abrasive tests (e.g. grinding on different grades of SiC paper [57]) or measurements associated with specific applications, the outstanding example being cutting tools, in which this class of composite finds widespread application. The agreement in raw results from different studies is sometimes contradictory. Sarin et al. [58, 59], for instance, tested a range of composites by abrasion with dry 45 µm diamond in argon, and found that the wear resistance scaled approximately with K c3/4 H 1/2 and the order of ranking of the materials tested (lowest wear resistance first) was Al2O3, Al2O3–ZrO2, Al2O3–TiC, Si3N4+Y2O3, SiAlON and Si3N4–TiC composite. Holz et al., however, performed abrasive pin-on-disc tests on a wide range of composites
© Woodhead Publishing Limited, 2006
Particulate composites
121
using two grades of SiC paper with 15 µm and 70 µm grit particles, and found no clear correlation with a relationship of the form K cm H n [57]. The addition of ZrO2 to alumina had little effect its wear resistance in this study, but the further addition of TiC/TiN reduced the wear rate by a factor of ~3 to produce one of the most wear-resistant materials tested, which, along with a hot-pressed monolithic β-SiAlON, was far superior to either of the two Si3N4–TiC/TiN composites tested. The sensitivity of wear to so many experimental factors is undoubtedly a major part of the reason for some of these apparently contradictory conclusions. Another is that important details of the microstructures of the materials being compared, such as the matrix grain size, particle size and amount of porosity, differ between the two studies. Such features can have a profound effect on the wear rate. Indeed, although the original motivation for adding TiC particulate to Al2O3 cutting tools was that TiC was harder, stiffer and more thermally conductive than alumina, though difficult to process as a monolith, it is now thought that the main reason why the particulate improves the hardness, strength and wear resistance is its grain refining effect [4]. The success of Al2O3–TiC cutting tools for machining steels and cast iron is interesting in the context of the good bonding between the particles and the matrix, the small thermal expansion mismatch, and consequently the limited amount of toughening in this composite system [60]. In tougher composites, the microstructural features such as thermal stresses and weak interfaces which are instrumental in the operation of toughening mechanisms such as crack bridging, microcracking, crack deflection and the direct toughening effect of residual stresses are also a potential aid to the initiation and propagation of the short, near-surface cracks that are responsible for severe wear by surface fracture and pullout, and so are potentially damaging to the wear resistance. Holz et al. [57], for instance, observed that SiC platelets in a reaction-bonded silicon nitride matrix were only weakly bonded to the matrix and pulled out during abrasive wear, increasing the wear rate both directly, and indirectly by acting as abrasive particles themselves. In cases where the mechanical properties of the composites play a dominant role, the extent to which the ease of crack initiation in toughened composites is overcome by the potential advantages of the particles (higher hardness, global toughness, etc.) depends on the relative scales of the microstructure and the cracks caused by abrasion. If the microstructural scale is greater than the crack size, the wear resistance is likely to be degraded, and vice versa. Figure 4.10 shows the effect of microstructural scale on the abrasive wear resistance of Si3N4/TiCp composites from the work of Wayne and Buljan [61]. Fine-scale composites, with a mean TiC particle size of 0.4 µm, improve the wear resistance relative to monolithic Si3N4, but coarser particles (≥ 1.5 µm) degrade it. The same work also demonstrates the influence of the type of test on the ranking of materials: composites with all three particle sizes
© Woodhead Publishing Limited, 2006
122
Ceramic matrix composites
1/ V (105 cm–2)
6 5 Si3N4
4 3 2 0
1
2 Particle size (µm)
3
4
4.10 Wear resistance, expressed as the reciprocal of the abraded volume, V, against microstructural scale, represented by the TiC mean particle diameter, for Si3N3–20vol%TiCp composites. Data from the work of Wayne and Buljan [61].
showed significant wear rate reductions relative to monolithic Si3N4 when subjected to a gas-jet particle erosive wear test. In other cases, the effect of particulate additions is dominated by chemical rather than mechanical effects, usually in less aggressive tests or in reactive environments. There is now a substantial body of literature describing the beneficial effects on wear rates of adding titanium-containing particles to ceramic matrices. These form a soft adherent tribochemical film of Ticontaining oxide that reduces the wear rate of the underlying surface from damage. Examples include unlubricated ball (Si3N4 or steel)-on-disc wear of Si3N4–TiB2 in laboratory air [62] and SiC–TiC and SiC–TiC–TiB2 in oscillating sliding against SiC and α-Al2O3 in water [63].
4.7.2
Nanocomposites
In view of the above comments on the importance to wear resistance of a fine microstructural scale, it is perhaps not surprising that the most remarkable effect of submicron particle additions in alumina/SiC ‘nanocomposites’ is to increase the resistance to severe wear dramatically compared with monolithic alumina. Zhao et al. [35] first commented that the nanocomposites were easier to polish to a mirror finish than alumina, and significant improvements in surface finish are also found for more aggressive surface treatments as Fig. 4.11 demonstrates [9]. Figure 4.11(a) shows the surface of a piece of alumina ground with 45 µm diamond paste. Grain facets resembling a conventional fracture surface can be seen behind the wear debris, showing that much of the material has been removed as large pieces by intergranular fracture around their periphery. Figure 4.11(b) shows an alumina–11vol%SiC nanocomposite, which has the same alumina grain size, after being subjected to the same treatment. The appearance of the surface contrasts sharply with
© Woodhead Publishing Limited, 2006
Particulate composites
(a)
123
(b)
4.11 Comparison of the response of pure alumina and an alumina/ SiC nanocomposite to abrasion with 45 µm diamond paste [9]: (a) alumina, grain size 2.6 µm; (b) alumina–11 vol% SiC, alumina grain size 2.6 µm.
that of the alumina, there being little brittle fracture and pullout, wear having become dominated by plasticity-controlled mechanisms as shown by the scratches covering most of the surface. The addition of only 2 vol% SiC reduces the area fraction of pullout by more than a factor of 2, and the corresponding reduction with 10% SiC is a factor of more than 50 [64]. Walker et al. [11] were the first to report improved wear resistance, finding that the addition of SiC reduced the wet erosive wear rate by a factor of 3 compared with alumina with the same grain size. It has now become clear that this is a quite general observation in conditions of severe wear, with reductions in wear rate of a similar order having been found for abrasive wear [64] and dry sliding wear [65]. In addition, Chen et al. [66] have shown that the mild to severe wear transition seen in sliding wear of alumina is either delayed or completely suppressed in the nanocomposites. They also found that there is no improvement in wear resistance in mild wear, where surface fracture is absent. Recent work correlating microstructure, abrasive wear rate and the appearance of the worn surfaces [64] has concluded that the most important reason for the improved surface finish and wear resistance on adding SiC is that the dimensions (diameter, depth) of the individual pullouts are reduced. This was explained in terms of the well-established change in fracture mode from intergranular in alumina to transgranular in the nanocomposites, which allows pullouts smaller than the grain size to be formed when fracture initiates in the nanocomposites, whereas in alumina, near-surface cracks tend to follow grain boundaries, giving a minimum pullout dimension of the order of the grain size. SiC additions of 10 vol% were also shown to inhibit the initiation of the cracking responsible for pullouts directly, and there is evidence that this is because the SiC particles inhibit the subsurface twinning to which
© Woodhead Publishing Limited, 2006
124
Ceramic matrix composites
alumina is prone [67]. It is also worth noting that the suppression of fracture initiation and surface damage in the nanocomposites provides a natural explanation for at least some of the strength improvement in these composites described in Section 4.5.2.
4.8
Future trends
The understanding of the mechanical behaviour of particulate composites is well advanced, and it seems unlikely that step changes in properties will be forthcoming through the discovery of new strengthening mechanisms, at least for materials with microstructural scales of 100 nm upwards. The most obvious area for further investigation is in particulate composites based on length scales smaller than this. Whilst the ‘nanocomposites’ described in this chapter have proved controversial, the striking results shown in Fig. 4.11 provide convincing evidence that the continuing refinement of microstructure can lead to novel and striking effects. The first task to be undertaken is to find processing methods capable of producing such microstructures, and here it is likely that research on particulate composites will proceed in close relationship with the efforts to produce single-phase ceramics with truly nanoscale grain sizes that are beginning to bear fruit [68]. Another area for novel work may be in functional particulate composites. Most functional ceramics in use at present are essentially single phase. The insertion of particles into the structure may interact in novel ways with the functional elements of microstructure. A simple example might be the pinning of ferroelectric domain wall motion during formation or operation, which may improve dielectric and electromechanical losses. Strength improvements might also be obtained. A further possible area of interest is the interaction between thermal stresses around particles with the strains occurring during cooling through the Curie temperature of BaTiO3 and other perovskites, which offers the possibility of producing new domain structures. At present the only commercial use of particulate ceramics of any note is alumina–TiC cutting tools, which as noted above rely more on the grainrefining effect of the particulate rather than on any of the true composite effects described. Thus a further area for development is in the application of these ceramics. The lack of application to date should not come as a surprise. The history of new materials has often shown a lag between development in the laboratory and commercial use. Monolithic ceramics are no exception to this. Despite the remarkable developments in structural ceramics that have taken place since the 1960s, truly structural applications remain rare. This is beginning to change, although it is important to realise that the applications for which structural ceramics are well suited are limited in scope by their nature. Monolithic nitride ceramics are beginning to be used in semi-structural applications in jet engines, and it is reasonable to suppose that as confidence
© Woodhead Publishing Limited, 2006
Particulate composites
125
and experience are gained with these, the use of ceramics will spread. The evidence in this chapter suggests that the potential role of most particulate ceramics is to give modest property improvements over monolithic ceramics with little cost penalty. The last point is important, and realistically requires the use of particles that allow pressureless sintering and that are chemically compatible with the matrix in terms of processing environment and oxidation. For improving the flaw tolerance of alumina, for instance, Cr3C2 particles are unlikely to be cost effective since they require the use of an inert or reducing environment. Al2TiO5 particles, however, allow pressureless sintering in air and are therefore an excellent candidate for exploitation in alumina components that require tolerance to surface damage. The alumina/SiC ‘nanocomposites’ may be an exception to this rule, in that the improvements in wear resistance may be sufficiently spectacular for it to be worth paying the premium associated with sintering in inert gas. Instead of regarding these materials as a superior (but expensive) alternative to alumina, it may be beneficial to market them as a cheap alternative to SiC or Si3N4.
4.9
References
1. Tite, M.S., Kilikoglou, V. and Vekinis, G. ‘Strength, toughness and thermal shock resistance of ancient ceramics, and their influence on technological choice’, Archaeometry 43 (2001) 301–324. 2. Lange, F.F. and Metcalf, M. ‘Processing-related fracture origins: II, Agglomerate motion and cracklike internal surfaces caused by differential sintering’, J. Am. Ceram. Soc. 66 (1983) 398–406. 3. Raj, R. and Bordia, R.K. ‘Sintering behavior of bi-modal powder compacts’, Acta Metall. 32 (1984) 1003–1019. 4. Cutler, R.A., Hurford, A.C. and Virkar, A.V. ‘Pressureless-sintered Al2O3–TiC composites’, Mat. Sci. Eng. A105/106 (1988) 183–192. 5. Taya, M., Hayashi, S., Kobayashi, A.S. and Yoon, H.S. ‘Toughening of a particulatereinforced ceramic-matrix composite by thermal residual stress’, J. Am. Ceram. Soc. 73 (1990) 1382–1391. 6. McMurtry, C.H., Boecker, W.D.G., Seshadri, S.G., Zanghi, J.S. and Garnier, J.E. ‘Microstructure and material properties of SiC–TiB2 particulate composites’, Am. Ceram. Soc. Bull. 66 (1987) 325–329. 7. Bennison, S.J., Padture, N.P., Runyan, J.L. and Lawn, B.R. ‘Flaw-insensitive ceramics’, Phil. Mag. Lett. 64 (1991) 191–195. 8. Huang, J.L., Huang, J.J., Jeng, C.A. and Li, A.K. ‘Investigation of Al2O3/Cr3C2 composites prepared by pressureless sintering: 3’, Ceramics International 25 (1999) 141–144. 9. Winn, A.J. and Todd, R.I. ‘Microstructural requirements for alumina–SiC nanocomposites’, Brit. Ceram. Trans. 98 (1999) 219–224. 10. Stearns, L.C., Zhao, J. and Harmer, M.P. ‘Processing and microstructural development in Al2O3-SiC “nanocomposites” ’, J. Eur. Ceram. Soc. 10 (1992) 473–477. 11. Walker, C.N., Borsa, C.E., Todd, R.I., Davidge, R.W. and Brook, R.J. ‘Fabrication, characterisation and properties of alumina matrix nanocomposites’, British Ceramic Proc. 53 (1994) 249–264.
© Woodhead Publishing Limited, 2006
126
Ceramic matrix composites
12. Zener, C.S. as described in Smith, C.S. ‘Grains, phases and interfaces: an interpretation of microstructure’, Trans. Met. Soc. AIME 175 (1948) 15–51. 13. Borsa, C.E., Jones, N.M.R., Brook, R.J. and Todd, R.I. ‘Influence of processing on the microstructural development and flexure strength of Al2O3/SiC nanocomposites’, J. Eur. Ceram. Soc. 17 (1997) 865–872. 14. Ashby, M.F. and Centamore, M.A. ‘The dragging of small oxide particles by migrating grain boundaries in copper’, Acta Metall. 16 (1968) 1081–1092. 15. Jeong, Y.K. and Niihara, K. ‘Microstructure and mechanical properties of pressureless sintered Al2O3/SiC nanocomposites’, Nanostructured Materials 9 (1997) 193–196. 16. Cock, A.M., Shapiro, I.P., Todd, R.I. and Roberts, S.G. ‘Effects of yttrium on the sintering and microstructure of alumina–silicon carbide “nanocomposites” ’, accepted for publication in J. Am. Ceram. Soc., 88 (2005) 2354–2361. 17. Wain, N. and Todd, R.I. paper in preparation. 18. Davidge, R.W. and Green, T.J. ‘The strength of two-phase ceramic/glass materials’, J. Mat. Sci. 3 (1968) 629–634. 19. Todd, R.I. and Derby, B. ‘Thermal stress induced microcracking in alumina-20% SiC composites’, Acta Mater. 52 (2004) 1621–1629. 20. Faber, K.T. and Evans, A.G. ‘Crack deflection processes – I. Theory’, Acta Metall. 31 (1983) 565–576. 21. Becher, P.F. and Wei, G.C. ‘Toughening behavior in SiC whisker-reinforced alumina’, J. Am. Ceram. Soc. 67 (1984) C267–C269. 22. Hübner, H. and Jillek, W. ‘Sub-critical crack extension and crack resistance in polycrystalline alumina’, J. Mat. Sci. 12 (1977) 117–125. 23. Knehans, R. and Steinbrech, R.W. ‘Memory effect of crack resistance during slow crack growth in notched Al2O3 bend specimens’, J. Mat. Sci. Lett. 1 (1982) 327–329. 24. Merchant, I.J., Macphee, D.E., Chandler, H.W. and Henderson, R.J. ‘Toughening cement based materials through the control of interfacial bonding’, Cement Concrete Research 31 (2001) 1873–1880. 25. Todd, R.I., Morsi, K. and Derby, B. ‘Neutron diffraction measurements of thermal residual microstresses in ceramic particle reinforced alumina’, Brit. Ceram. Proc. 57 (1997) 87–101. 26. Evans, A.G. and Faber, K.T., ‘Crack growth resistance of microcracking brittle materials’, J. Am. Ceram. Soc. 67 (1984) 255–260. 27. Gu, W.-H., Faber, K.T. and Steinbrech, R.W. ‘Microcracking and R-curve behaviour in SiC-TiB2 composites,’ Acta Metall. Mater. 40 (1992) 3121–3128. 28. Cutler, R.A. and Virkar, V. ‘The effect of binder thickness and residual stresses on the fracture toughness of cemented carbides’, J. Mat. Sci. 20 (1985) 3557–3573. 29. Boccaccini, A.R. and Trusty, P.A. ‘Toughening and strengthening of glass by Al2O3 platelets’, J. Mat. Sci. Lett. 15 (1996) 60–63. 30. Heuer, A.H. ‘Transformation toughening in ZrO2-containing ceramics’, J. Am. Ceram. Soc. 70 (1987) 689–698. 31. Padture, N.P., Bennison, S.J. and Chan, H.M. ‘Flaw-tolerance and crack-resistance properties of alumina–aluminum titanate composites with tailored microstructures’, J. Am. Ceram. Soc. 76 (1993) 2312–2320. 32. Lawn, B. Fracture of Brittle Solids (2nd edn), Cambridge University Press, Cambridge, UK, 1993. 33. Taruta, S., Itou, Y., Takusagawa, N., Okada K. and Otsuka, N. ‘Influence of aluminum titanate formation on sintering of bimodal size-distributed alumina powder mixtures’, J. Am. Ceram. Soc. 80 (1987) 551–556.
© Woodhead Publishing Limited, 2006
Particulate composites
127
34. Niihara, K. ‘New design concept of structural ceramics – ceramic nanocomposites’, J. Ceram. Soc. Japan 99 (1991) 974–982. 35. Zhao, J., Stearns, L.C., Harmer, M.P., Chan, H.M., Miller, G.A. and Cook, R.E. ‘Mechanical behaviour of alumina–silicon carbide “nanocomposites” ’, J. Am. Ceram. Soc. 76 (1993) 503–510. 36. Pérez-Rigueiro, J., Pastor, J.Y., Llorca, J., Elices, M., Miranzo, P. and Moya, J.S. ‘Revisiting the mechanical behaviour of alumina/silicon carbide nanocomposites’, Acta Mater. 46 (1998) 5399–5411. 37. Alford, N.M., Birchall, J.D. and Kendall, K. ‘High-strength ceramics through colloidal control to remove defects’, Nature 330 (6143) (1987) 51–53. 38. Meschke, F., Alves-Riccardo, P., Schneider, G.A. and Claussen, N. ‘Failure behavior of alumina and alumina/silicon carbide composites with natural and artificial flaws’, J. Mat. Res. 12 (1997) 3307–3315. 39. Todd, R.I., Bourke, M.A.M., Borsa, C.E. and Brook, R.J. ‘Neutron diffraction measurements of residual stresses in alumina/SiC nanocomposites’, Acta Mater. 45 (1997) 1791–1800. 40. Jiao, S. and Jenkins, M.L. ‘A quantitative analysis of crack-interface interactions on alumina-based nanocomposites’, Phil. Mag. A78 (1998) 507–522. 41. Hoffman, M. and Rödel, J. ‘Suggestion for mechanism of strengthening of “nanotoughened” ceramics’, J. Ceram. Soc. Japan 105 (1997) 1086–1090. 42. Wu, H.Z., Lawrence, C.W., Roberts, S.G. and Derby, B. ‘The strength of Al2O3/SiC nanocomposites after grinding and annealing’, Acta Mater. 46 (1998) 3839–3848. 43. Ando, K., Chu, M.C., Tsuji, K., Hirasawa, T., Kobayashi, Y. and Sato, S. ‘Crack healing behviour and high-temperature strength of mullite/SiC composite ceramics’, J. Eur. Ceram. Soc. 22 (2002) 1313–1319. 44. Jenkins, M.G., Salem, J.A. and Seshadri, S.G. ‘Fracture of a Tib2 particle/SiC matrix composite at elevated temperature’, J. Comp. Mat. 23 (1989) 77–90. 45. Davidge, R.W. ‘Mechanical Behaviour of Ceramics’, Cambridge University Press, Cambridge, UK, 1979. 46. Baldoni, J.G., Buljan, S.T. and Sarin, V.K. ‘Particulate titanium carbide–ceramic matrix composites’, Proc. 2nd Int. Conf. Science of Hard Materials, Inst. Phys. Conf. Ser. 75, IoP/Adam Hilger, Bristol, UK, 1986. 47. French, J.D., Zhao, J. Harmer, M.P. Chan, H.M. and Miller, G.A. ‘Creep of duplex microstructures’, J. Am. Ceram. Soc. 77 (1994) 2857–2865. 48. Ohji, T., Nakahira, A., Hirano, T. and Niihara, K. ‘Tensile creep behavior of alumina/ SiC nanocomposite’, J. Am. Ceram. Soc. 77 (1994) 3259–3262. 49. Thompson, A.M., Chan, H.M. and Harmer, M.P. ‘Tensile creep of alumina–silicon carbide “nanocomposites” ’, J. Am. Ceram. Soc. 80 (1997) 2221–2228. 50. Niihara, K., Suganuma, K., Nakahira, A. and Izaki, K. ‘Interfaces in Si3N4–SiC nanocomposite’, J. Mat. Sci. Lett. 9 (1990) 598–599. 51. Park, H., Kim, H.E. and Niihara, K. ‘Microstructure and high-temperature strength of Si3N4–SiC nanocomposite’, J. Eur. Ceram. Soc. 18 (1998) 907–914. 52. Pezzotti, G. and Sakai, M. ‘Effect of silicon carbide “nanodispersion” on the mechanical properties of silicon nitride’, J. Am. Ceram. Soc. 77 (1994) 3039–3041. 53. Wakai, F. ‘A review of superplasticity in ZrO2-toughened ceramics’, Brit. Ceram. Trans. 88 (1989) 205–208. 54. Yoon, C.K. and Chen, I.W. ‘Superplasticity of two-phase ceramics containing inclusions – zirconia mullite composites’, J. Am. Ceram. Soc. 73 (1990) 1555–1565. 55. Wakai, F., Kodama, Y., Sakaguchi, S., Murayama, N., Izaki, K. and Niihara, K. ‘A superplastic covalent crystal composite’, Nature 344 (1990) 421–423.
© Woodhead Publishing Limited, 2006
128
Ceramic matrix composites
56. Kim, B.N., Hiraga, K., Morita, K. and Sakka, Y. ‘A high strain-rate superplastic ceramic’, Nature 413 (2001) 288–291. 57. Holz, D., Janssen, R., Friedrich, K. and Claussen, N. ‘Abrasive wear of ceramicmatrix composites’, J. Eur. Ceram. Soc. 5 (1989) 229–232. 58. Sarin, V.K., Buljan, S.T. and Smith, J.T. in Science and Technology, ed. S.P. Parker, pp. 441–449, McGraw Hill, New York (1985). 59. Warren, R. and Sarin, V.K. ‘Particulate ceramic-matrix composites’, in CeramicMatrix Composites, ed. R. Warren, pp. 146–166, Blackie, Glasgow, UK (1992). 60. Wahi, R.P. and Ilschner, B. ‘Fracture behaviour of composites based on Al2O3–TiC’, J. Mat. Sci. 15 (1980) 875–885. 61. Wayne, S.F. and Buljan, S.T. ‘Microstructure and wear resistance of silicon nitride composites’, in Friction and Wear of Ceramics, ed. S. Jahanmir, pp. 261–285, Marcel Dekker, New York (1994). 62. Jones, A.H. Dobedoe, R.S. and Lewis, M.H., ‘Mechanical properties and tribology of Si3N4–TiB2 ceramic composites produced by hot pressing and hot isostatic pressing’, J. Eur. Ceram. Soc. 21 (2001) 969–980. 63. Wäsche, R. and Klaffke, D. ‘Ceramic particulate composites in the system SiC– TiC–TiB2 sliding against SiC and Al2O3 under water’, Tribology International 32 (1999) 197–206. 64. Ortiz, Merino, J.L. and Todd, R.I. ‘Relationship between wear rate, surface pullout and microstructure during abrasive wear of alumina and alumina/SiC nanocomposites’, Acta Mater. 53 (2005) 3345–3357. 65. Rodríguez, J., Martín, A., Pastor, J.Y., Llorca, J., Bartolomé, J. and J. Moya, ‘Sliding wear of alumina/silicon carbide nanocomposites’, J. Am. Ceram. Soc. 82 (1999) 2252–2254. 66. Chen, H.J., Rainforth, M. and Lee, W.E. ‘The wear behaviour of Al2O3–SiC ceramic nanocomposites’, Scripta Mat. 42 (2000) 555–560. 67. Wu, H.Z., Roberts, S.G. and Derby, B. ‘Residual stress and subsurface damage in machined alumina and alumina/silicon carbide nanocomposite ceramics’, Acta Mater. 49 (2001) 507–517. 68. Binner, J.G.P. ‘The art of the possible – processing nanostructured ceramics’, Mater. World 12 (2004) 30–32.
© Woodhead Publishing Limited, 2006
Part II Graded and layered composites
© Woodhead Publishing Limited, 2006
5 Functionally-graded ceramic composites I M L O W , R D S K A L A and P M A N U R U N G, Curtin University of Technology, Australia
5.1
Introduction
Layered-graded materials (LGMs) exhibit a stepwise or progressive change in composition, structure, and properties as a function of position within the material [Koizumi, 1993; Hirai, 1996]. This innovative design eliminates ever-present sharp boundaries in conventional composites which may impart undesirable physical and mechanical properties. An example is debonding or separation at the boundary due to thermal or residual stress induced by mismatch in thermal expansion. LGMs have been processed by a variety of methods. A list of the common synthesis methods & examples of these materials have been described elsewhere [Sakai & Hirai, 1991]. Recently, liquid infiltration of preforms [Marple & Green, 1990; Low, 1998a] has emerged as an innovative technique for the processing of graded composite materials. Using this infiltration process, it is possible to design new materials with unique microstructures (e.g. graded, multiphase, microporous, etc.) and unique thermomechanical properties (e.g. graded functions, designed residual strains, thermal shock, etc.). Recent developments in layered ceramics have provided a strategy for laminating the ceramic structure with an outermost homogeneous layer to provide wear resistance and an underlying heterogeneous layer to provide toughness [An et al., 1996; Liu et al., 1996; Padture et al., 1995]. These layered structures promote toughness by interlayer crack deflection through weak interfacial bonding, or strength by incorporating macroscopic compressive residual stresses through strong interlayer bonding. Layered ceramics produced in this second way have shown uncommonly high damage resistance under Hertzian loading, with retention of strength and wear resistance. However, these layered structures are disadvantaged by either the counterproductive effects of weak interlayers or the excessively large residual stresses that can cause enhancement of delamination failures. Here we consider a new approach, in which microstructural elements are tailored to provide graded compositions and generate different modes of 131 © Woodhead Publishing Limited, 2006
132
Ceramic matrix composites
strengthening and toughening. The basic idea is to produce a graded dispersion of particles within the alumina matrix through an infiltration process to yield a layer of homogeneous alumina for hardness and wear resistance, and a heterogeneous layer of tough graded alumina for damage dispersion. The design of this layered-graded material can provide a unique mechanical performance of both flaw tolerance and wear resistance. The proposed strategy of designing layer structures with graded interfaces for crack arrest is fundamentally different from the conventional laminar approach, where there is a sudden change in composition at the interface between layers. This abrupt interface can cause cracking or delamination due to thermal expansion and elastic modulus mismatches. The toughening processes envisaged, in which the heterogeneous graded layers act to inhibit crack penetration by interlayer ‘stress shielding’ and by ‘crack bridging’, offer major advantages over conventional layer composites, where toughness is introduced via deflection of transverse cracks along weak interfaces. In this chapter, we describe the synthesis and characterisation of the microstructure and properties of layered-graded alumina-matrix composites through liquid infiltration. This approach is relatively simple and offers excellent control over the depth of the graded layer. The presence of a graded dispersion of reinforced particles in the alumina matrix has a profound influence on the physical and mechanical properties of the composites. An overview of the infiltration kinetics and the use of the infiltration process as a new philosophy for tailoring novel graded ceramic systems are also presented.
5.2
Infiltration kinetics and characteristics
Liquid-phase infiltration of preforms has emerged as an extremely useful method for the processing of composite materials. This process involves the use of low-viscosity liquids such as sols, metal- or polymer-melts. Using this infiltration process, it is possible to design new materials with unique microstructures (e.g. graded, multiphase, microporous) and unique thermomechanical properties (graded functions, designed residual strains and thermal shock). Liquid infiltration into dry porous materials occurs due to capillary action. The mechanism of infiltrating liquids into porous bodies has been studied by many researches in the fields of soil physics, chemistry, powder technology and powder metallurgy [Carman, 1956; Semlak & Rhines, 1958]. However, the processes and kinetics of liquid infiltration into a powdered preform are rather complex and have not been completely understood. Based on Darcy’s fundamental principle and the Kozeny–Carman equation, Semlak & Rhines (1958) and Yokota et al. (1980) have developed infiltration rate equations for porous glass and metal bodies. These rate equations can be used to describe the kinetics of liquid infiltration in porous ceramics preforms, but
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
133
a single capillary model will result in the pore size measured from the experimental infiltration rates being an order of magnitude smaller than the pore sizes seen from SEM and porosimetry measurements [Dullien et al., 1977; Einset, 1996]. Low and co-workers (2000) have shown that a combination of the Washburn model [Washburn, 1921] and Dullien’s analysis [Dullien et al., 1977] is able to reconcile the infiltration rates and the pore sizes determined from SEM and porosimetry measurements for the infiltration of water and TiCl4 into alumina preforms.
5.2.1
Modelling infiltration kinetics
There are a number of formulae which are relevant for modelling the infiltration kinetics of a liquid into preforms. The first equation to calculate the height of infiltration against time was formulated by Washburn (1921):
h=
γ r cos θ 2η
1/2
t 1/2
(5.1)
where h and t are the height of liquid and the time, respectively, and γ, θ, r and η are the surface tension, contact angle, pore radius and viscosity, respectively. Hence, the rate of infiltration (i.e. h/t) is a function of surface tension, contact angle, pore radius and viscosity. Another model, proposed by Yokota et al. (1980), involves tortuosity, T, and shape factors, Cs:
Cs pγ r cos θ h= 2 T η
1/2
t 1/2
(5.2)
The value for T is normally 1.4142 and Cs is 0.4. When water wets the preform completely and liquid is spreading, the contact angle θ is 0° [Yokota et al., 1980]. A third formula was proposed by Travitzky & Shlayen (1998):
P h = r net 2η
1/2
t 1/2
(5.3)
where the net pressure, Pnet is given by: Pnet =
2 γ cos θ – ρ gh – Patm r
(5.4)
Here ρ is the density of the infiltrant, g is the acceleration due to gravity (9.8 m/s2) and Patm is the pressure of the surrounding space above the liquid infiltrant. This model is useful for predicting the influence of pressure on the rate of infiltration. Another formula is the Kozeny–Carman equation [Carman, 1956]:
© Woodhead Publishing Limited, 2006
134
Ceramic matrix composites
h dh = k T dt
(5.5)
where T is the tortuosity of the capillary tube and k is a constant.
5.2.2
Studies on infiltration kinetics
Mercury porosimetry measurements for a partially sintered alumina preform showed a bimodal pore size distribution with neck diameter Dn = 0.15 µm [Manurung, 2001]. As a comparison with the pore sizes and distribution of the preform measured by porosimetry, SEM micrographs (Fig. 5.1) were taken before and after infiltration. Based on SEM examination, the pores in the preform before infiltration ranged in size from r ~ 0.1–0.5 µm. Assuming an average pore radius of 0.3 µm, this radius is approximately four times larger than the pore-neck radius (Dn = 0.15 µm, so pore radius = 0.075 µm) determined by mercury porosimetry. It is interesting to note that the pore sizes appear to remain virtually unchanged following the infiltration process. This suggests that (i) the capillary forces involved did not appear to cause any shrinkage or size reduction of the pores, and (ii) the infiltrant had not filled up the pores but only formed a thin layer deposit on the walls of pores. The small particles entering the grains are believed to arise from the infiltrant which had entered the pore channels and adhered to the walls of the pores following drying [Manurung, 2001]. In order to successfully model the infiltration kinetics in terms of the effects of presintering temperature, type of infiltrant, infiltration environment, and multiple infiltrations, the pore radius of alumina preform (presintered at 1000°C) was measured using water as infiltrant, since the viscosity and
(a)
(b)
5.1 SEM micrograph of the partially sintered alumina preform (1000°C) showing the pore microstructure (a) before and (b) after infiltration with TiCl4 [Manurung, 2001].
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
135
surface tension of water are well known. The pore radius of the alumina preform was calculated from the Washburn model. By assuming a contact angle θ = 0° [Einset, 1996; Ligenza & Bernstein, 1951] the pore radius of the preform can be calculated if the height and time of infiltration are known. The rate of infiltration is determined from the slope in Fig. 5.2(a) and then from this slope the pore radius can be found. From the measurements, it was found that the pore radius of the alumina preform is 0.015 ± 0.001 µm. Similarly the pore radius found from alumina preform infiltrated with TiCl4 is 0.018 ± 0.002 µm [Manurung, 2001]. The errors in the radii only reflect the experimental uncertainty in the measured values for surface tension and viscosity. However, the measured pore radius is an order of magnitude smaller than the pore radius determined from porosimetry and SEM (Fig. 5.1). The pore radius determined from infiltration kinetics can be reconciled to the experimentally determined radii from SEM and porosimetry, by assuming a two-pore-size model (pore neck and pore bulge), instead of a single capillary pore-size [Dullien et al., 1977; Einset, 1996]. The schematic diagram for this two-pore-size model is shown in Fig. 5.2(b). Dullien (1979) considered the rate of capillary rise of a fluid in a model three-dimensional network pore structure consisting of a repeating pore element with step changes in diameter. The effective diameter, Deff, model is given by: Deff
3 2 D = 1 Σ Dk Σ Dk Σ k 3 k j Dj k
–1
(5.6)
0.07
Height (m)
0.06
Db
0.05 0.04
Dn
0.03 0.02 0.01 0 0
20 40 60 Square root of time (s1/2) (a)
Infiltration direction
80 (b)
5.2 (a) Height of infiltration of water into alumina preforms (sintered at 1000ºC) as a function of square root of time; (b) a schematic of the two-size single-capillary. Db is the pore-bulge diameter, Dn is the pore-neck diameter.
© Woodhead Publishing Limited, 2006
136
Ceramic matrix composites
where the summations are over the number of segments of the repeating pore unit. The effective pore diameters are therefore smaller than the individual diameters of the pore segments. Consider a two-pore-size repeating unit model in equation (5.6), i.e. j = 1, 2 and k = 1, 2 for pore necks and pore bulges as shown in Fig. 5.2(b). Thus Db can be calculated by substituting Deff and Dn into equation (5.6). The results are 0.27(3) and 0.18(4) µm for water and TiCl4 respectively. The calculated value of Db for both infiltrants is in reasonable agreement with the average pore diameter estimated from SEM (Fig. 5.1). The kinetics of liquid infiltration in porous alumina preforms have been found to depend on several parameters such as surface tension and viscosity of infiltrant, porosity and pore size of preforms, and pressure [Manurung, 2001]. An enhanced infiltration rate is most favourable when (a) a preform has a high porosity (>45%), (b) an infiltrant has a low viscosity, (c) the infiltrate is in vacuum due to a greater driving force, and/or (d) there are multiple infiltrations by virtue of a self-lubrication effect [Manurung, 2001]. A more likely reason for the rate increase during subsequent cycles can be attributed to a decrease in the tortuosity. The path that the infiltrant takes during each cycle is smoothed by the preceding infiltration cycle. Therefore the tortuosity is decreased. Inspection of equation (5.2) shows that the infiltration rate will increase if the tortuosity is decreased. From cycle 1 to cycle 3, the rate increases by a factor of approximately 2 (Fig. 5.3). The tortuosity would have to decrease by a factor of ∼2 if the rate increase was attributed entirely to a decrease in tortuosity. This is unlikely, therefore it is suggested that there may be a slight increase in the pore radius and shape factor during subsequent cycling, as well as a decrease in tortuosity. It is suggested that the major cause in the rate increase in going from cycle 1 to cycle 3 can be attributed to a decrease in the tortuosity. The experimental results show that the kinetics of infiltrating water and titanium tetrachloride into an alumina preform are parabolic with time (Fig. 5.3). It has been shown that the viscosity of infiltrants influences the rate of infiltration and that the rate of infiltration is pressure dependent. The experimental result of faster kinetics in a vacuum as opposed to that in 1 atm agrees with Travitzky & Shlaken (1998) model. The presintering temperature has a strong influence on the kinetics of TiCl4 in alumina preforms and multiple infiltrations increase the rate of infiltration. According to Yokota et al. (1980) the increase in the infiltration rate due to multiple cycling is predominantly attributable to a decrease in the tortuosity of the preform during subsequent cycles.
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
137
0.06
Height (m)
0.05
Water
0.04
TiCl4
0.03 0.02 0.01 0 0
20
40 60 80 100 Square root of time (s1/2)
120
(a) 0.06 Cycle 3 Cycle 2 Cycle 1
Height (m)
0.05 0.04 0.03 0.02 0.01 0 0
10
20 30 40 50 60 70 Square root of time (s1/2)
80
90
(b)
5.3 (a) Effect of infiltrants using water and titanium tetrachloride (TiCl4); (b) effect of multiple infiltrations with TiCl4 under vacuum.
5.3
Infiltration processing of LGMs
Two steps are involved in the design of LGMs using the infiltration process. The first involves the fabrication of a partially sintered porous preform which is then followed by its infiltration with an appropriate infiltrant such as TiCl4, Si(OC2H5)4 or calcium acetate solution. The infiltrated preform is subsequently heat-treated at elevated temperatures to form the desired phases in situ. The process of infiltration can involve either partial or complete immersion of the preform in the infiltration. The latter gives rise to samples with an outer graded layer and an inner core of the host material [Marple & Green, 1993]. In contrast, the former produces an outer layer of host material and an inner layer with a graded composition [Low, 1998a]. The depth or thickness of the graded layer can be further increased by multiple cycles of
© Woodhead Publishing Limited, 2006
138
Ceramic matrix composites Pellet/bar
Presintering 1000–1200°C [2 h]
Preform (~ 40–45%)
Infiltration in TiCl4 solution
Partial infiltration
Full infiltration
Sample Sample
Heat 1400°C [12h] 1650°C [2h] TiCl4 solution
TiCl4 solution Functionally-A/AT
5.4 Liquid infiltration processing of layered-graded alumina-matrix composites.
infiltration. A typical infiltration process for the alumina/AT and other systems is shown in Fig. 5.4.
5.4
Characterisation and properties of aluminamatrix LGMs
5.4.1
Alumina/mullite and mullite/ZTA/mullite systems
Electron microprobe analysis of concentration profiles across sections of the sintered samples revealed the existence of concentration gradients, the mullite content decreasing with increasing distance from the surface of the bodies [Marple & Green, 1993]. SEM examination also revealed a microstructural effect: the alumina grain size tended to increase from the surface of the
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
139
samples. This suggests that the presence of mullite limits grain growth in alumina. From evaluating the mechanical properties of the resulting ceramic composite material, large increases, of up to 60%, for both the strength (biaxial flexure) and indentation fracture toughness were achieved (see Table 5.1). These increases were attributed to the presence of the mullite case and the resulting residual compressive surface stresses due to the thermal expansion mismatch between the mullite and alumina [Marple & Green, 1991]. The strength of the indented samples deviated from the ideal behaviour, indicating the possibility of R-curve behaviour in these materials. The abundance of mullite in the mullite/ZTA system increased with increasing infiltration time [Low et al., 1993]. The density (ρ) and mullite content of the sintered sample as a function of infiltration time are shown in Table 5.2. The results suggest that the infiltration process was time (t) dependent and diffusion-controlled with the infiltration front travelled as a function of t1/2. The content of mullite was greatest near the surface and decreased sharply towards the core of the sample. The presence of mullite and hence compressive surface stresses appears to improve the hardness and fracture toughness (see Table 5.2). These values are at least two to three times higher than those reported for the mullite/ alumina system described above. Clearly, the presence of mullite is desirable for inducing compressive stresses in the vicinity of the surface region by virtue of the mismatch in thermal expansion between ZTA and mullite. This significant improvement in the observed fracture toughness was attributed to Table 5.1 Mechanical properties of graded mullite/alumina Mullite content (vol%)
Strength (MPa)
Modulus (GPa)
Toughness (MPa.m1/2)
0 5 10 13 19
305 400 425 440 510
400 375 370 355 335
3.9 5.5 5.5 6.5 7.0
Table 5.2 Physical and mechanical properties of graded mullite-ZTA Infiltration time (h)
Density (g/cm3)
Mullite content (vol%)
Hardness (GPa)
Toughness (MPa.m1/2)
0 1 4 6
4.15 4.10 4.00 4.06
0 4.7 5.9 6.2
1679 1692 1693 1733
8.0 10.5 11.5 13.0
© Woodhead Publishing Limited, 2006
140
Ceramic matrix composites
the presence of the compressive stresses near the surface in addition to other well-established energy-dissipative processes such as transformation toughening, microcracking and/or crack deflection [Low et al., 1994]. This explanation concurs with that proposed by Marple & Green [1992] for their mullite/alumina composites where the presence of residual compressive stresses was observed to be the major contributor to increases in strength and fracture toughness [Marple & Green, 1992].
5.4.2
Alumina/aluminium titanate and alumina–zirconia/ aluminium titanate systems
Layered-graded Al2O3/AT and Al2O3–ZrO2/AT systems were synthesised using TiCl4 or Ti(OC2H5)4 as an infiltrant [Skala, 2000; Low et al., 1996a]. Figure 5.5 shows a typical graded microstructure of this material where the content of AT is most abundant near the surface and decreases with increasing depth. SEM micrographs of the cross-section of the graded Al2O3/AT at various depths are presented in Fig. 5.6. The alumina grains are a dark colour while the AT grains are a lighter shade of grey. Figure 5.6(a) shows the microstructure in the near-surface regions of the LGM. The presence of AT can be clearly seen to have a beneficial effect on the growth of the alumina grains. The alumina grains within the surface regions of the LGM are relatively small and equiaxed when compared to those of the core (Fig. 5.6(c)). The alumina grains within the core region are extremely large and elongated, with some of the grains containing intragranular porosity due to exaggerated
5.5 Back-scattered scanning electron micrograph showing the gradation of AT distribution within the sample. Direction of infiltration is from right to left.
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites (a)
141
(b)
10 µm
10 µm (c)
10 µm
5.6 Typical microstructures of the graded alumina/AT system at (a) surface, (b) 1.0 mm, and (c) 2.0 mm below surface.
grain growth and grain boundary migration. The grain growth behaviour of the alumina phase is clearly modified by the presence of AT, probably due to a solid-solution effect. Very similar observations in compositional and microstructural gradations were obtained for the layered-graded ZTA/AT system [Pratapa, 1997]. LGMs of AT/Al2O3 have also displayed some very unique but interesting properties which include excellent machinability, low thermal expansion coefficient, improved thermal shock resistance, low hardness, low Young’s modulus and enhanced tolerance to damage [Low, 1998a, 1998b; Skala, 2000; Low et al., 1996a]. The XRD depth profile shows that the top surface region is very rich in AT (~88 wt%) with the concentration decreasing slowly as the depth increases towards the middle region of the sample. The amount of AT formed here is very large when compared to that found by other workers (Fig. 5.7). For instance, Pratapa (1997) found 46 wt% AT phase on the surface of a graded AT/ZrO2–alumina system, with the value decreasing substantially to 7 wt% at a depth of 0.8 mm. Low (1998b) found 50 wt% AT on the surface of graded AAT composites, decreasing to 10 wt% at a depth of 0.5 mm. Similarly Skala (2000) obtained 52 wt% on the surface which decreased to only 3 wt% at a depth of 0.5 mm in AAT composite. This suggests that unidirectional and double infiltrations with the aid of a plastic shield provide an effective method for increasing the content of AT near the surface and for preventing
© Woodhead Publishing Limited, 2006
142
Ceramic matrix composites
Weight percentage of AT (wt%)
100 This work Skala (2000) Pratapa (1997) Low (1998a)
90 80 70 60 50 40 30 20 10 0 0
0.5
1 1.5 Depth (mm)
2
2.5
5.7 Depth profiles for the alumina–AT LGM as obtained by various researchers, including this study. Error bars indicate two error estimated deviations (σ).
Intensity ratio (AT/alumina)
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0
1
2 3 4 Grazing-incidence angle (°)
5
6
5.8 Variation of peak-intensity ratio between AT (110) and the alumina (104) as a function of grazing-incidence angle.
a very rapid decrease in the graded AT content. The existence of composition gradation at both the nanometre and micrometre scale has been verified by grazing-incidence synchrotron radiation diffraction (Fig. 5.8) [Singh et al., 2002]. The development of induced residual strains within the alumina layer of Al2O3−Al2TiO5 bilayers with and without graded interfaces in the temperature range 20–1500°C has been observed from the time-of-flight neutron diffraction in terms of line-shift of the (113) reflection. As would be expected, the presence of a sharp interface in the non-graded sample resulted in the formation of much higher residual strains when compared to the bilayer sample with 300 µm thin graded interfaces, due to mismatch in thermal expansion
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
143
coefficients and elastic moduli between Al2O3 and Al2TiO5 in the former. The presence of graded interfaces serves as a ‘buffer region’ to modulate the differences in material properties within the bilayer and thus the attenuation of the thermally induced strains. It is also interesting to note that the residual strains induced in the non-graded sample were most profound along the planes parallel with the sharp interfaces, resulting in the concomitant cracking and sample disintegration. In contrast, the graded sample showed little or no cracking. This observation verifies the importance of designing bilayer ceramics with graded interfaces to reduce the formation of undesirable residual strains or stresses which may cause cracking and delamination at the interface. As would be expected, the microhardness (Hv) of the graded material increases with depth from 12.1 to 14.3 GPa (Fig. 5.9), since the concentration of relatively soft AT decreases with depth [Pratapa et al., 1998; Skala, 2000; Manurung, 2001]. The stress–strain curves obtained from spherical indentations on both graded and control samples are illustrated in Fig. 5.10. Solid curves are empirical fits for the data. The data deviate from the Hertzian elastic limit at stresses above Po ≈ 2 GPa for the control and Po ≈ 0.2 GPa for the graded sample, marking the onset of ‘yield’. This result serves to verify the ‘quasi-plastic’ nature of the graded layer, a phenomenon which has also been observed in other ceramics with a heterogeneous microstructure [An et al., 1996; Padture et al., 1995; Liu et al., 1996]. Figure 5.11 is an optical micrograph showing the Vickers indentation damage around the indent at 200 N load. There is a distinct upheaval in the vicinity of the indent as a result of pronounced grain uplift. The presence of a profuse damage zone surrounding the indent is vividly highlighted under the Nomarski illumination. However, no radial cracks are observed at 200 N. Unlike the brittle control sample which exhibits cracks emanating from all 25 surface
end
AT rich
AT poor
Hardness (GPa)
20
15
10
5
0 0
0.5
1
1.5 2 Depth (mm)
2.5
3
5.9 Vickers hardness of alumina–AT LGM as a function of depth with load of 3 kg. Error bars indicate two mean deviations (±).
© Woodhead Publishing Limited, 2006
144
Ceramic matrix composites 8 7
Stress (GPa)
6 5 4 3 2 1 0 0.00
0.03
0.06
0.09
0.12
0.15
Strain
5.10 Indentation stress–strain curves for both graded (䊐) and control (∆) samples.
four indentation corners, the graded surface of LGM exhibits either no cracks (up to 200 N) or short cracks in only one or two corners of the indent, at 300 N. Grain pushout is routinely observed around the indent. It is believed that during the process of indentation, the weakly bonded grains are initially debonded, lifted up and eventually pushed out from their original positions. It appears that most of the indentation energy is used for debonding, lifting and pushing out the grains from the surface, thus rendering the material damage resistant. The high propensity for grain debonding is believed to arise from the presence of residual tensile stresses produced by the thermal expansion mismatch between AT and alumina. It should be pointed out that the display of profuse grain uplift and microdamage around the indent is unusual for ceramic materials. When an indentation is applied, ceramics usually exhibit a ‘sink in impression’ which is caused by the densification below the tip of the indent [Zeng et al., 1996]. This is usually accompanied by formation of radial cracks at the four cornertips of the indents, indicating the brittleness of the material. By contrast, metals usually exhibit a ‘rising of material’ or surface uplift above the unindented surface level as a result of shear or plastic deformation. It follows that the graded material exhibits indentation damage patterns that are indicative of pseudo-plastic deformation during loading. The presence of AT ‘softens’ the alumina matrix, thus rendering it effective in energy-absorption and crack attenuation [Pratapa & Low, 1998; Skala, 2000]. Optical microscopy confirms that the ‘yielding’ behaviour during Hertzian loading arises from the onset of indentation damage in the graded region of the LGM. The nature of this damage can be discerned from the micrographs in Fig. 5.12, obtained using the bonded-interface section technique previously described. The micrographs show both half-surface (upper) and section views
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
145
(a)
20 µm
(b)
5.11 Indentation damage on the surface of LGM at a load of 200 N as revealed by (a) Nomarski contrast and (b) SEM.
(lower) of indentations. The results clearly show the evolution of subsurface damage development as one progresses up the stress–strain curve in Fig. 5.10 that corresponds to increasing indentation pressure. The initiation of the deformation subsurface damage zone and subsequent expansion of this zone are immediately apparent from the grain deformations or displacements revealed by the Nomarski contrast. At Po ≈ 0.5 GPa, i.e. just above the elastic limit, only a few grains have deformed through shear-driven debonding, sliding and pushout. At increasing pressures, the number of deformed grains increases, and the damage zone expands further below the surface. At Po ≈ 1.5 GPa, the damage is more profuse and begins to take on the appearance of the well-developed, near-hemispherical deformation zone expected from
© Woodhead Publishing Limited, 2006
146
Ceramic matrix composites
(a)
(b)
5.12 Optical micrographs in Nomarski illumination showing halfsurface (top) and section (bottom) views of Hertzian damage in both graded and control samples: (a) control sample at P = 1.5 kN; (b) graded sample at P = 1.5 kN.
continuum plasticity models [Johnson, 1985]. The complete absence of any ring cracks or cone cracks on the surfaces even at the maximum pressure suggests that the graded material is damage tolerant. Cone fracture is inhibited by the ability of the material to contain the extent of microdamage to a small area around the indent via multiple energy-absorbing mechanisms which include diffuse microcracking, grain debonding and sliding, crack deflection, grain pushout, and grain bridging. In contrast, a ring crack on the surface of the control sample has initiated and attempted to run around the contact circle. The classical cone crack has also formed in the subsurface region, highlighting the brittleness of this material. LGMs of the AT/alumina and AT/ZTA displayed some very interesting properties which include excellent machinability, low thermal expansion coefficient, improved thermal shock resistance, low hardness (about 5 GPa), low Young’s modulus (E) (250 GPa) and excellent flaw tolerance [Pratapa, 1997; Pratapa & Low, 1998; Skala, 2000; Manurung, 2001]. These materials appeared to display a large degree of near-surface ‘quasi-plasticity’ under the Hertzian or the Vickers indenter which effectively inhibits the formation and propagation of cracks. The ‘ductile’ behaviour of these materials was
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
147
further verified by the dependence of Vickers microhardness (H) on indentation load and a low ratio of H to E. These materials also exhibited a superior thermal shock resistance when compared to pure alumina. No cracks were observed on the surface of these materials during the Vickers or Hertzian indentation [Low et al., 1996b; Pratapa, 1997; Skala, 2000]. Instead, a heavily microdamaged region was observed both within and in the vicinity of the indenter. The formation of these damaged zones is believed to act as an effective ‘energy sink’ for the indenter, thus shielding the material from crack formation. Energy dissipative processes such as debonding between AT and alumina grains, sliding and pushout of AT grains have been observed in these damage zones. This display of flaw tolerance is attributed to high thermal expansion anisotropy of AT grains which induce very large residual stresses. These stresses can cause alumina grains to form ‘crack bridges’ and thus apply closure forces for shielding the crack tip from the applied stress intensity field, not unlike microcrack toughening [Runyan & Bennison, 1991]. Very similar observations in Vickers and Hertzian contact damages were obtained for the layered-graded ZTA/AT system [Pratapa, 1997; Pratapa & Low, 1998].
5.4.3
Alumina/mullite/AT hybrid
The Al2O3/mullite/AT hybrid has been fabricated by infiltrating an Al2O3 preform with both TiCl4 and Si(OC2H5)4, followed by heat-treatment at 1600°C for 2 h. The phase composition on the surface of the hybrid sample as revealed by XRD showed the presence of alumina, mullite and AT. No peaks associated with that of rutile or silica were visible, which suggests that complete reactions between titania and alumina were achieved to form AT (Al2TiO5) at ∼1300°C, and between silica and alumina to form mullite (3Al2O3 · 2SiO2) at ∼1100°C as follows: TiO2 + Al2O3 → Al2TiO5
(5.7)
SiO2 + Al2O3 → 3Al2O3 · 2SiO2
(5.8)
Optical microscopy of the cross-section of an infiltrated hybrid sample revealed a distinct gradation of AT and mullite content in the infiltrated zone at the micrometre scale. However, whether this gradation occurs at the nanometre scale cannot be discerned from the microstructure. It is interesting to note that the interface between graded and non-graded regions shows exaggerated growth of alumina grains, which is not observed in the graded hybrid region. Scanning electron microscopy of the hybrid region showed the presence of needle-like mullite and equiaxed AT grains embedded within the fine microstructure. A scanning electron micrograph of an as-fired surface of the hybrid sample showed the presence of needle-like mullite grains with a dense interlocked structure. These elongated grains help to achieve flaw
© Woodhead Publishing Limited, 2006
148
Ceramic matrix composites
Percentage of phases
tolerance in the hybrid through crack bridging and crack deflection as energy dissipative processes. Small equiaxed or acicular AT grains are observed to distribute along or at the triple-point junctions of larger alumina grains. The presence of both mullite and AT has also resulted in much finer Al2O3 grains within the hybrid region, indicating their effectiveness as grain-growth inhibitors. Such grain refinement of Al2O3 has also been observed in Al2O3– mullite, Al2O3–AT and Al2O3–CaAl12O19 systems [Marple & Green, 1993; Skala, 2000; Asmi et al, 1999] and may account for the relatively high hardness observed. The combined effect of mullite and AT has allowed a much smaller reduction in hardness when compared to the much softer Al2O3– AT system. Clearly, the self-reinforcement due to the presence of mullite has compensated for the much softer AT phase. XRD and grazing-incidence synchrotron radiation diffraction (GISRD) plots of a hybrid sample at different depths from the surface showed the abundance of α-Al2O3, AT and mullite to vary with depth (Fig. 5.13). As the depth increased, the abundance of both mullite and AT decreased, but that of αAl2O3 increased. The composition depth profiles as determined from the Rietveld analysis are shown in Fig. 5.13(a). From the results it can be seen 100 90 80 70 60 50 40 30 20 10 0
Mullite AT Al2O3
Counts ratio (AT/A, M/A)
0
0.5
1 1.5 Depth (mm) (a)
2
2.5
1.6 1.4 1.2 1
AT
0.8 0.6 0.4
M
0.2 0 10
100 1000 Penetration depth (Å)
10000
5.13 Depth profiling of phase compositions in a hybrid LGM as revealed by (a) X-ray diffraction and (b) grazing-incidence synchrotron radiation diffraction.
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
149
10
20
8
Hardness
16
6 4
Toughness
14 12
2 0
18
0
0.5 1 1.5 2 Distance from the surface (mm)
Hardness (GPa)
Fracture toughness (MPa.m1/2)
that the abundances of AT and mullite in the hybrid sample decrease rapidly within the first 0.5 mm. The existence of graded compositions at the nanoand micro-scale has been established by GISRD. Figure 5.13(b) shows the GISRD plots of peak intensity ratios for AT and mullite relative to alumina as a function of penetration depth (l) [Singh & Low, 2002a,b; Low et al., 2002. As the grazing-angle (α) or depth of penetration (l) increases, the AT and mullite peaks become less intense as compared to alumina. The AT intensity ratio curve shows a rapid decrease within the depths of <50 Å, above which there is a gradual fall. It is interesting to note that the mullite intensity ratio curve shows a more gradual fall throughout the depths. This is indicative of a nanometre-scale gradation in phase composition within the hybrid material as would be expected from the time-dependent kinetics of the infiltration process. In essence, the near-surface hybrid layer is richer in both AT and mullite, which decrease in abundance rapidly with depth. Figure 5.14 shows the depth-profile variation of hardness (Hv) and fracture toughness (KIC) of the hybrid sample. Fracture toughness is evidently higher near the surface due to the abundant presence of both AT and mullite. As the depth increases, the abundance of both phases and thus the fracture toughness decreases. Hardness is lower near the surface due to the abundant presence of much softer AT, but increases with depth towards the alumina region. It should be noted that since measurements were made on a polished crosssectioned sample, the indentation data could not be collected at distances less than 0.5 mm from the surface. Hence, hardness and fracture toughness values on the surface were measured separately. When compared to the control, the presence of AT and mullite as dispersed phases has imparted a two-fold increase in the fracture toughness without causing a large reduction in hardness as in the alumina-AT system (Table 5.3) [Singh & Low, 2002a; Low, 1998a, 1998b]. This implies the display of both self-strengthening and self-toughening processes in the hybrid sample by virtue of the presence of both low thermal expansion AT and needle-like
10 2.5
5.14 Variation of hardness and fracture toughness as a function of distance for the hybrid LGM.
© Woodhead Publishing Limited, 2006
150
Ceramic matrix composites
Table 5.3 Physical and mechanical properties of hybrid and control Al2O3 samples Sample
Shrinkage (%)
Apparent porosity (%)
Bulk density (g.cm–3)
Hv (GPa)
KIC (MPa.m1/2)
Al2O3 Hybrid
19.5 6.7
0.25 1.08
3.93 3.63
17.9 (0.4)* 16.9 (0.2)
3.5 (0.4) 8.3 (0.6)
*Figures in parentheses are the estimated standard deviation of the values to the left. 10
KlC (MPa.m1/2)
9 8 7 6 5 4 0
5
10
15 20 Load (kg)
25
30
35
5.15 Display of crack-growth resistance in the hybrid LGM.
mullite and its interlocked texture within the microstructure. The fracture toughness of over 8.0 MPa.m1/2 is similar to that reported for the alumina– AT system but considerably higher than that of the alumina–mullite system [Padture et al., 1993; Chan et al., 1996; Low, 1998a, 1998b; Marple & Green, 1991]. The presence of elongated mullite grains acting as bridging sites augmented by large residual stresses due to thermal expansion mismatch between alumina and AT is believed to be responsible for the much improved fracture toughness. The characteristic elongated grain morphology of mullite and an interlocked texture would have also resulted in desirable crack-tip deflections and tilts to produce tortuous crack paths. Figure 5.15 shows the KIC versus load plot for the graded region of the hybrid sample. A rising R-curve is clearly evident which suggests the presence of flaw tolerance behaviour in the graded hybrid region. This display of flaw tolerance may be attributable to crack-tip energy dissipative processes such as crack bridging, crack deflection and crack branching by virtue of the simultaneous presence of elongated mullite and soft but low thermal expansion AT grains.
5.5
Concluding remarks
The concept and synthesis of layered and graded materials (LGMs) from the infiltration technique have been introduced. The mechanical and fracture
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
151
properties of various alumina-based LGMs have been discussed in relation to phase composition, microstructure and residual stress. These materials displayed superior mechanical performance in terms of strength, fracture and thermal shock resistance. This new approach may be further developed for designing new generation layer composites with an in-situ top wear resistant coating and a flaw-tolerant graded substrate or bulk. For instance, through unidirectional infiltration, a layer composite of a wear-resistant alumina surface and a damage-tolerant graded AT/alumina bulk can be fabricated. A similar design can also be tailored for alumina– or ZTA–mullite systems where the in-situ top layer imparts wear and fatigue resistance. The infiltration process also offers a potential for designing in-situ thin films of pristine HTSC with graded HTSC/epoxy substrates via careful control of infiltrant kinetics from a specified direction. These structures will be ideally suited for fabrication of superconducting electronic devices such as microbridges, microwave cavities and resonators. A similar avenue can be used to design graded ZrP/epoxy protonic conductors for use in a number of electrochemical devices such as fuel cells, gas sensors, and electrolysers. A new philosophy and design process for tailoring layer graded composites for both strength and damage tolerance has been highlighted for alumina-based ceramics. This approach allows the design of multifunctional new generation layer composites with a top hard layer for wear resistance and a graded bulk or underlayer for damage tolerance. A similar design strategy can also be tailored for other novel systems based on ceramic/polymer, ceramic/metal or polymer/polymer hybridisation. This approach also has generic appeal and can be extended to other more complex geometries or architectures (e.g. tubes or crucibles).
5.6
Acknowledgements
Financial support from the Australian Research Council (LX0242352 and A00001131) and the Australian Institute of Nuclear Science and Engineering (AINSE Grants 96/143 and 97/141, 98/010, 99/029 and 00/091P), the Australian Synchrotron Research Program (ANBF 99/2000-AB-26 and 00/01-AB-31) and ISIS (RB12443 and RB13709) of the Rutherford-Appleton Laboratory are acknowledged. IML is especially indebted to his past research students (R. Skala, D. Asmi, P. Manurung, S. Pratapa and D. Lawrence) and also the research colleagues (B. Lawn, D. Li, B. O’Connor, C. Buckley and M. Singh) for their invaluable contributions to the research program on layered and graded materials over the last eight years.
5.7
References
An, L., Chan, H.M., Padture, N.P. & Lawn, B.R. (1996): Damage-resistant aluminabased layer composites. J. Mater. Res. 11, 204–210.
© Woodhead Publishing Limited, 2006
152
Ceramic matrix composites
Asmi, D., Low, I.M., Kennedy, S.J. & Day, R.A. (1999): Characteristics of layered and graded calcium hexaluminate/alumina composites. Mater. Lett. 40, 96–102. Carman, P.C. (1956): Flow of Gaseous through Porous Media, Butterworths Scientific Publications, London. Chan, H., An, L., Padture, N. & Lawn, B.R. (1996): Damage resistant alumina-based layer composites. J. Mater. Res. 11, 204–210. Dullien, F.A.L. (1979): Porous Media Fluid Transport and Pore Structure. Academic Press, New York. Dullien, F.A., El-Sayed, M.S. & Batra, V.K. (1977): Rate of Capillary Rise in Porous Media with Non-Uniform Porous, J. Colloids and Interface Science, 60, 497–506. Einset, E.O. (1996): Capillary Infiltration Rate into Porous Media with Application of Silicon Composite Processing. J. Am. Ceram. Soc., 79(2), 333–338. Hirai, T. (1996): Chapter 20 in Processing of Ceramics, Part 2, edited. by R.J. Brook. VCH, Weinheim pp. 293–341. Johnson, K.L. (1985): Contact Mechanics. Cambridge University Press, London. Koizumi, M. (1993): in Ceramic Transactions, Vol. 34 – FGMs (Proc. 2nd Int. Symp. on FGMs), edited by J.B. Holt, M. Koizumi, T. Hirai and Z.A. Munir. Am. Ceram. Soc., Westerville, OH, pp. 3–10. Ligenza, J.R. & Bernstein, R.B. (1951): The Rate of Rice of Liquids in Fine Vertical Capillaries, J. Am. Chem. Soc. 73, 4636–4638. Liu, H., Lawn, B.R. & Hsu, S.M. (1996): Hertzian contact response of tailored silicon nitride multilayers. J. Am. Ceram. Soc. 79, 1009–1014. Low, I.M. (1998a): Processing of an in-situ layered and graded alumina-aluminium titanate composites. Mater. Res. Bull. 33, 1475–1482. Low, I.M. (1998b): Synthesis and properties of in-situ layered and graded aluminium titanate/alumina composites. J. Aust. Ceram. Soc. 34, 250–255. Low, I.M., Skala, R.D., Richards, R. & Perera, D.S. (1993): Synthesis and properties of novel mullite/ZTA composites. J. Mater. Sci. Lett. 12, 1985–1987. Low, I.M., Skala, R. & Perera, D.S. (1994): Fracture properties of layered mullite/ZTA composites. J. Mater. Sci. Lett. 13, 1334–1336. Low, I.M., Skala, R.D. & Zhou, D. (1996a): Synthesis of functionally-graded aluminium titanate/alumina composites. J. Mater. Sci. Lett. 15, 345–347. Low, I.M., Skala, R.D. & Zhou, D. (1996b) Sol-gel processing of functionally-graded ceramics. Proc. Int. Workshop on Sol-Gel Processing of Advanced Ceramics, 8–9 January 1996, Madras, India, Oxford & IBH Publisher, p. 143. Low, I.M., Skala, R.D., Asmi, D., Manurung, P. & Singh, M. (2000): ‘Infiltration Processing of novel functionally-graded ceramic materials’. Pp 1464–1472 in Proc 2000 Powder Metallurgy World Congress (Eds. K. Kosuge & H. Nagai), 12–16 Nov, 2000, Kyoto, Japan Low, I.M., Singh, M., Manurung, P., Wren, E., Sheppard, D.P. & Barsoum, M.W. (2002): Depth profiling of phase composition and texture in layered-graded Al2O3- and Ti3SiC2based systems using x-ray and synchrotron radiation diffraction. Key Engineering Materials 224–226, 505–510. Manurung, P. (2001): Microstructural design and characterisation of alumina/aluminium titanate composites. Ph.D Thesis, Curtin University of Technology, Perth, Australia. Marple, B.R. & Green, D.J. (1990): J. Am. Ceram. Soc. 73, 3611. Marple, B.R. & Green, D.J. (1991): Mullite/alumina particulate composites by infiltration processing: III, Mechanical properties. J. Am. Ceram. Soc. 74, 2453–2459. Marple, B.R. & Green, D.J. (1992): J. Am. Ceram. Soc. 75, 44.
© Woodhead Publishing Limited, 2006
Functionally-graded ceramic composites
153
Marple, B.R. & Green, D.J. (1993): Graded compositions and microstructures by infiltration processing. J. Mater. Sci. 28, 4637–4643. Padture, N.P., Bennison, S.J. & Chan, H.M. (1993): Flaw-tolerance and crack-resistance properties of alumina and aluminium titanate composites with tailored microstructures. J. Am. Ceram. Soc. 76, 2312–2320. Padture, N.P., Pender, D.C., Wuttiphan, S. & Lawn, B.R. (1995): In-situ processing of silicon carbide layer structures. J. Am. Ceram. Soc. 78, 3160–3162. Pratapa, S. (1997): Synthesis and character of functionally-graded aluminium titanate/ ZTA ceramics. M.Sc Thesis, Curtin University of Technology, Perth, Australia. Pratapa, S. & Low, I.M. (1998): Infiltration-processed functionally-graded AT/alumina– zirconia composites: II, Mechanical properties. J. Mater. Sci. 33, 3047–3053. Pratapa, S., Low, I.M. & O’Connor, B.H. (1998): Infiltration-processed functionallygraded AT/alumina-zirconia composites: I, Microstructure and physical properties. J. Mater. Sci. 33, 3037–3046. Runyan, J.L. & Bennison, S.J. (1991): Fabrication of flaw-tolerant aluminium-titanatereinforced alumina. J. Eur. Ceram. Soc. 7, 93–99. Sakai, M. & Hirai, T. (1991): Fabrication and properties of FGMs. J. Ceram. Soc. Japan. 99, 1002. Semlak, K.A. & Rhines, F.N. (1958): Trans. Met. Soc. AIME, 212, 325. Singh, M. & Low, I.M. (2002a): Depth-profiling of phase composition and preferred orientation in a graded alumina/mullite/aluminium-titanate hybrid using x-ray and synchrotron radiation diffraction. Mater. Res. Bull. 37, 1279–1291. Singh, M. & Low, I.M. (2002b): Layered and graded alumina-based hybrid composites by infiltration processing. Key Engineering Materials 224–226, 493–498. Singh, M., Manurung, P. & Low, I.M. (2002): Depth profiling of near-surface information in a functionally-graded Al2O3/Al2TiO5 composite using grazing-incidence synchrotron radiation diffraction. Mater. Lett. 55, 344–349. Skala, R.D. (2000): Development of a functionally-graded alumina/aluminium-titanate composite. PhD Thesis, Curtin University of Technology, Perth, Australia. Travitzky, N.A. & Shlayen, A. (1998): Microstructure and Mechanical Properties of Alumina/Cu-O. Material Science and Engineering, A224, 154–160. Washburn, E.W. (1921): Principles of Physical Chemistry, McGraw-Hill, New York. Yokota, M., Hara, A., Ohata, M. & Mitani, H. (1980): Trans. Jap. Inst. Metals, 21, 652. Zeng, K., Soederlund, E., Giannakopoulos, A.E. & Rowcliffe, D.J. (1996): Controlled indentation: a general approach to determine mechanical properties of brittle materials. Acta Mater. 44, 1127–1132.
© Woodhead Publishing Limited, 2006
6 SiAlON based functionally graded materials H M A N D A L, Anadolu University, Turkey and N Ç A L I S A C I K B A S, MDA Advanced Ceramics Ltd, Turkey
6.1
Introduction
In functionally graded materials (FGMs), the properties change gradually with position/direction and are affected by composition, microstructure, porosity and so on. The usefulness of functionally graded composites with a graded structure concept was recognized by Shen and Bever in 1972.1 On the other hand, systematic researches on production methods for functionally graded materials were carried out around four years later by Niino and colleagues of the National Aerospace Laboratory at Sendai in Japan.2–4 However, increased performance of FGMs is hampered by increased manufacturing cost and difficulties in production.5 Therefore, the usage of FGMs in the marketplace requires an extensive search for cheaper, reproducible and reliable manufacturing methods for FGMs.
6.2
Functionally graded materials
In the development of functionally graded materials, there are two approaches. One is to eliminate the boundary of laminated-type composites, thereby eliminating discontinuities in the properties at the boundary. The other option is to make non-uniform distributions of dispersoids in a homogeneous composite, thus creating multiple functions within the material.2 Therefore, continuous variation in composition, microstructure and so on, results in change in properties as a function of position in the component.6 FGMs offer increased efficiency with reduced weight and volume for motors and generators, giving a useful operating temperature to thermal barrier materials by eliminating thermal stresses between the layers. Furthermore, FGMs have the potential for use in a wide range of engineering applications, for instance aerospace, high-efficiency engine components (e.g. piston heads, engine blocks), brake discs, hot gas valves and tubes, turbine engine components, fuel cells, certain piezoelectric devices and direct metal tools (e.g. injection molding tools and cutting tools) for industrial use, 154 © Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
155
biomaterials used in artificial human implants, drug delivery devices with release rate control, armor and armament components for defense and many more.7,8 For these applications, control of material composition is necessary both for improving a variety of properties such as mechanical performance, toughness and strength, and for reducing interfacial stresses between dissimilar materials. A variety of methods to produce graded materials have been reported in the relevant literature. The most popular method is powder processing techniques, where different mixtures of powders are stacked, slip cast, tape cast, infiltrated, electrochemically processed, spark plasma sintered (SPS) and so on.9,10 Fabrication of FGMs offers a technological challenge in largescale graduation and requires elaborate processing facilities. Thus, increased performance of FGMs is mostly limited by increased manufacturing costs. Widespread transfer into the marketplace, therefore, requires an extensive search for cheap, reproducible and reliable manufacturing methods for FGMs.11,12
6.3
SiAlON ceramics
SiAlONs is a general name for a large family of the ceramic alloys based on silicon nitride.13 They were first discovered independently at about the same time (1971–72) in the United Kingdom at the University of Newcastle-uponTyne (by Jack and Wilson14) and in Japan (by Oyama15 ). It is well known that silicon nitride is a covalently bonded material and has a hexagonal structure. It can exist in two crystallographic modifications designated α and β.16,17 Both structures are built up of corner-sharing SiN4 tetrahedra. The β-Si3N4 structure is obtained by ABAB… stacking of silicon and nitrogen atoms. This structure leads to continuous channels parallel to the c-direction (Fig. 6.1). However, the α-Si3N4 structure is the result of an ABCDABCD… stacking. The channels are thus closed and as a result, there are two large interstitial sites in each unit cell, where cations can be accommodated (Fig. 6.2). The unit cell parameters of the α phase are a = 7.818 Å and c = 4.591 Å. The corresponding data for the β-modification are a = 7.595 Å and c = 2.9023 Å, based on single crystal refinement. There are two SiAlON phases that are most important for related engineering ceramics, α-SiAlON and β-SiAlON, which are solid solutions based on α and β-Si3N4 structural modifications, respectively. In single phase form, βSiAlON has higher toughness (7–8 MPa m1/2), strength and thermal conductivity. On the other hand, α-SiAlON has excellent hardness (~20 GPa) but slightly worse strength and toughness (3–4 MPa m1/2) compared to β-SiAlON ceramics. To combine the advantages of these phases, α/β-SiAlON composites have been developed.18 The other crystalline phase found in SiAlON systems is O-SiAlON, which has also drawn interest in the engineering
© Woodhead Publishing Limited, 2006
156
Ceramic matrix composites Silicon (IV) nitride
x
y
z
6.1 The β-Si3N4 structure is obtained by ABAB… stacking of silicon and nitrogen atoms.
Silicon Nitrogen
6.2 The α-Si3N4 structure is the result of an ABCDABCD… stacking of silicon and nitrogen atoms.
ceramics field. The ‘behaviour diagram’ of the SiAlON system at 1700– 1730°C is shown in Fig. 6.319 where it can be clearly seen that x-axis and yaxis scales are expressed in equivalent percentages of aluminum and oxygen, respectively. The five ordered SiAlON polytypoid phases (27R, 21R, 12H, 15R and 8H) occur close to the AlN corner in this system and these phases are structurally very similar to each other. The high-temperature reactions of Si3N4 and Al2O3 usually give β-SiAlON and SiO2-rich X-phase.20 Additions of suitable metal oxide (or nitride) will expand the SiAlON phase into a fivecomponent system, Me-Si-Al-O-N, called a Jänecke prism.21 Silicon nitride is a highly covalent bonded compound with self-diffusion coefficient of the nitrogen atoms of 6.3 × 10–20 cm2/s at 1400°C.22 Therefore, densification without any sintering additives is nearly impossible. In 1961, Deeley and Herbert.23 was the first to report that Si3N4 ceramics could be
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials 3(SiO2)
2(Al2O3)
x-p
α-
sia
157
lon
ha
se
8H β -s
ia
lon
1
5R
Si3N4
12H 21H 27H
4(AIN) Equivalent Al (%)
6.3 The ‘behaviour diagram’ of the SiAlON system at 1700–1730°C.
densified by hot pressing with the incorporation of oxides as sintering additives. On the other hand, sintering additives limit the usage of Si3N4 at hightemperature applications. Thus, Oyama.15 and Jack and Wilson14 tried to overcome the problem of high-temperature strength degradation by investigating SiAlON ceramics. The sintering behavior has a significant influence on the mechanical and thermal properties of sintered materials. Sintering of SiAlON ceramics is carried out by a liquid-phase sintering process, which contains a residual grain boundary phase inherited from the sintering medium. This phase can be glassy (amorphous) or crystalline depending on factors such as overall composition and the applied cooling conditions. Liquid-phase sintering can be performed in several ways, namely pressureless sintering, hot pressing (HP), hot isostatic pressing (HIP) and gas pressure sintering (GPS). Pressureless sintering can be employed to fabricate complex shapes, and significant stabilization can be achieved by using a protective powder bed while the products generally indicate low density and the process requires large amounts of additives for densification.24 In hot pressing, powder mixtures of Si3N4 with additives are heated to high temperatures under an applied uniaxial pressure. Traditional hot pressing uses 20–30 MPa pressure which enhances both rearrangement of particles and grain boundary diffusion. The hot pressing offers the ability to fabricate dense products, but also limits the products to simple shapes.25 Another method is to apply isostatic pressure and hot-isostatic pressing (HIP), now being another established technique. This technique is a very attractive because it offers possibilities of making dense SiAlON ceramics with a negligible residual glassy grain boundary phase and hence better high-temperature properties. However, HP and HIP techniques are very costly.
© Woodhead Publishing Limited, 2006
158
Ceramic matrix composites
They become economic only when a large number of samples is involved.26 Gas pressure sintering (GPS) is another technique to densify SiAlON ceramics. GPS has proven to be an effective approach to minimize the content of oxide additives at no sacrifice of density, because high nitrogen pressure is of advantage for suppressing decomposition of Si3N4 at temperatures above 1800°C, and provides extra driving force for sintering.27,28
6.3.1
β-SiAlON ceramics
β-SiAlONs, which are the first developed group in the SiAlON materials, are formed by substituting up to two-thirds of the Si in the β-Si3N4 by Al provided that valency compensation is maintained by the replacement of an equivalent concentration of N by O to give a range of β-SiAlONs, Si6–zAlzOzN8–z with 0 < z < 4.2.29 Thus, z (Si–N) bonds are replaced by z (Al–O) bonds, since the difference between the respective bond lengths (1.74 Å for Si–N and 1.75 Å for Al–O) is small and the extent of replacement is wide. With increasing z value the density of β-SiAlON decreases linearly, similarly lowering Young’s modulus, strength, thermal conductivity, hardness and fracture toughness.30 On the other hand, for low z values (z < 1) the hardness and fracture toughness increase and when z ≥ 1 they decrease. It has been also depicted that glass-free microstructures could be obtained in β-SiAlON polycrystals with substitutional level z ≥ 2.31 Monolithic β-SiAlON materials can be easily densified by pressureless sintering, since the presence of alumina in the starting mixture lowers the eutectic temperature of the densifying liquid phase by some 200–300°C. In single-phase form, β-SiAlON materials possess a high degree of toughness (8 MPa m1/2) by in-situ reinforcement with the elongated β-SiAlON grains. They also possess good strength up to 1000 MPa and excellent thermal shock resistance.32
6.3.2
α-SiAlON ceramics
α-SiAlONs were first observed very shortly after β-SiAlONs with the general formula of MxSi12–m–nAlm+nOnN16–n, where M is one of the cations Li, Mg, Ca or Y or most rare earths (excluding La, Ce, Pr and Eu); x is equal to m divided by the valency of the M cation; there is a minimum value for the x parameter (0.3–0.5; i.e. there is an immiscibility gap between the α-Si3N4 and α-SiAlON phases); and x cannot exceed 2 since there are only two interstitial sites in each unit cell. In α-SiAlONs, n (Si–N) bonds (1.74 Å) are replaced by similar-sized Al–O (1.75 Å). The larger lattice strain resulting from the replacement of Si–N by Al–N restricts the range of solid solution (m-value) in α-SiAlON compared with β-SiAlON. The n-values are expected to favor β-SiAlON formation.33
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
159
The α-SiAlON has twice the unit-cell parameter in the c-direction compared to β-SiAlON, and the doubling of the Burgers vector for c[0001] dislocations means that dislocation movement is more difficult and the hardness is enhanced. Studies by Ekström and co-workers34 on the sintering of α-SiAlON ceramics showed that the largest rare earth cation to enter the α-SiAlON structure alone is Nd3+, with an ionic radius of 0.99 Å. The same authors also claimed that the slightly larger cation Ce3+, with a radius around 1.03 Å, could not enter the α-SiAlON structure alone, but may enter when mixed with a smaller stabilizing cation like Y3+ with a radius of 0.89 Å. α-SiAlONs were known to have a microstructure of fine and equiaxed grains. But, more recently, it has been proved that α-SiAlON ceramics can also be produced with elongated morphologies. The first example for the elongated nature of α-SiAlON grains was observed by Hwang et al.35 using a mixture of CaO–SrO–Y2O3 as sintering additives. Therefore, fracture toughness and hardness increase at the same time in the material.36
6.3.3
O-SiAlON ceramics
O-SiAlON is another crystalline phase of interest. There is a limited solubility of alumina in silicon oxynitride structure to give O-SiAlONs, represented by the formula Si2–xAlxO1+xN2–x, where x varies from zero to ~0.2. Formation of O-SiAlON occurs in the same mechanism as β-SiAlON; i.e. Si + N is replaced by Al + O. The lattice parameters of O-SiAlON, Si2–xAlxO1+xN2–x, increase in a very typical way with the x value.37 The formation of Si2N2O from mixtures of high-purity Si3N4 and SiO2 heated at high temperatures is very sluggish due to kinetic hindrance. For Si2N2O materials intended for high temperatures, addition of metal oxides such as Y2O3 alone, which forms a refractory glass, is the most attractive alternative. The mechanical properties of monophase Si2N2O are difficult to determine because of preparation difficulties in obtaining a monophasic Si2N2O material without the use of a sintering aid. A good approximation of Vickers hardness (HV 10) of 1600 kg/mm2 and a fracture toughness of ~3.3 MPa m1/2 were obtained from measurements on the materials HIPed at 1900°C, consisting of almost pure, crystalline Si2N2O.38 A slight improvement of the toughness has been found in O-SiAlON (x = 0.1) material with Y2O3 addition. The modulus of rupture of similar O-SiAlON materials measured by Trigg and Jack is ~420 MPa.39 Besides, O-SiAlON materials have poor toughness due to strong bonding Si2N2O crystals and the SiO2-rich glassy phase. O-SiAlON is characterized by high oxidation resistance due to its high oxygen content.
© Woodhead Publishing Limited, 2006
160
6.3.4
Ceramic matrix composites
α-β-SiAlON ceramics
α- and β-SiAlON phases are completely compatible and α-SiAlON/β-SiAlON composites are readily prepared by a single-stage sintering of the appropriate mixture of nitrides and oxides.40 Therefore, in recent years, mixed α-βSiAlON materials have received increasing attention owing to their easier fabrication compared to Si3N4 materials. More importantly, good mechanical properties can be obtained due to the high hardness of α-SiAlON and the good strength and toughness of β-SiAlON.41 However, the hardness and toughness of α-β-SiAlON composites are not as high as those of their monolithic counterparts. Moreover, it has recently been found that the phase composition and microstructure of α- and α-β-SiAlON ceramics are greatly affected by post-sintering heat treatment procedures at lower temperatures (1300–1600°C) when rare earth oxides are used as sintering additives. The α-SiAlON phase is only stable at high temperatures and transforms to βSiAlON plus other crystalline or vitreous phases.42–44 The ease with which this transformation proceeds decreases with increasing atomic number of the rare earth cation. Indeed, in the absence of β-SiAlON nuclei, certain ytterbium α-SiAlON compositions do not transform to β even when a high level of the liquid phase is present.45,46 This transformation provides a convenient mechanism for controlling the mechanical properties of the final material. However, it can only be used beneficially in applications where the maximum service temperature is below the transformation temperature (1000°C) of the grain boundary glassy phase. High-temperature properties, especially oxidation and creep resistance, deteriorate significantly above this temperature.47
6.4
Functionally graded SiAlON ceramics
Microstructure, mechanical properties and high-temperature properties of SiAlON ceramics can be optimized by α-β-SiAlON composites. Although α-β-SiAlON composites possess better mechanical properties with respect to monolithic phases of either α- or β-SiAlON, α → β SiAlON phase transformation restricts the compositional design of α-β-SiAlON ceramics.45 For this reason, the concept of producing functionally graded SiAlON ceramics has been developed to improve especially the mechanical properties of the material. Depending on the application areas, the desired properties can be obtained with appropriately designed SiAlON composition and production methods. Development of functionally graded SiAlON ceramics provides better properties than either monolithic α-SiAlON or β-SiAlON, or α-βSiAlON composites with continuous change in composition, microstructure and mechanical properties. Besides, a functionally graded structure is ideal for smooth reduction of thermal stresses in joints between the layers with different composition.
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
161
Functionally graded SiAlON ceramics can be produced by lamination (powder and tape), powder bed, controlling the sintering conditions, infiltration, slip casting and so on. Some of these production techniques will be discussed in the following sections.
6.5
Production techniques of functionally graded SiAlON ceramics
The aim of the preparation of a FGM is to achieve a well-controlled distribution of composition, microstructure and so on. Although many production techniques for functionally graded materials were already developed in the early 1990s, there are still limitations concerning material combinations, specimen geometry and cost. Thus, new processing techniques for producing functionally graded materials are necessary.
6.5.1
Lamination technique
Lamination of green compacts or sheets, which are produced through the tape casting method, is a very important technique for graded material production. For this purpose, layers with different α:β SiAlON ratio are used to obtain graded structures. Further details regarding these production techniques are given below. Powder lamination Prelamination of green compacts is a simple and well-established technique for graded material production.48 In the powder lamination method, FGM fabrication stages include compositional design, compaction of different powders, and sintering at suitable conditions. In compositional design of functionally graded SiAlON ceramics, selection of cations and α:β phase ratios are very important parameters. Type of cations affects the densification and diffusion between the layers. They also control the thermomechanical properties of SiAlON ceramics. As mentioned before, α-SiAlON possesses high hardness, wear resistance and oxidation resistance while β-SiAlON has high fracture toughness and good flexural strength. Thus, the α:β ratio of a designed FGM is very important according to its application area. In the compaction stage, two or more different powder compositions are stacked by sequential filling of the die and uniaxially pressed in an automatic press. Packing of powders and grain size are effective parameters in controlling the composition and thus the properties. After compaction, the laminates are sintered to obtain a gradual structure. Shen and Nygren reported on the production of functionally graded SiAlON
© Woodhead Publishing Limited, 2006
162
Ceramic matrix composites
ceramics using the SPS method.49 However, no detailed information was given in this reference or elsewhere. Besides, our research group has produced laminar-type functionally graded SiAlON ceramics with a transition zone thickness of ~1.3 mm.50,51,52 In these studies,50 the effects of type of cations, amount of liquid phase, sintering temperature and time on the diffusion between the layers were investigated. For this purpose, two-layered functionally graded SiAlON ceramics were produced. One layer consisted of the composition with about equal proportions of α and β (B1, B2 and B3 rich in the β-phase) and another layer was rich in the α-phase (A1 composition). The difference between B1 and B2 is in the type of cations: B1 contains only a single cation while B2 has three types. The difference between B2 and B3 is in the amount of the liquid phase: B3 contains more liquid phase. Laminates were prepared by sequential filling of the die and pressing, as schematically shown in Fig. 6.4. The laminates B1–A1 (named FGM-A), B2–A1 (named FGM-B) and B3–A1 (named FGM-C) were sintered under 22 bar nitrogen gas pressure at 1800°C for 1 hour. Also, to observe the effects of sintering time and temperature, B3–A1 laminate was sintered at 22 bar nitrogen gas pressure at 1700°C for 1 hour (named FGM-D) and for 2 hours (named FGM-E). The phase composition from the surface to the interior of the samples was determined by X-ray diffractometry (XRD) through successive grinding of the surface at 100 µm intervals. The microstructural characterization of the sintered specimens was achieved by scanning electron microscopy (SEM) in backscattered mode. The hardness change from the surface to the interior of the sample was measured by the Vickers indentation method at 19.6 N load. Visual examination showed that laminates other than B1–A1 were intermixed, showing that counter-diffusion occurred. SEM analyses also confirmed the visual examinations of B1–A1 laminate (Fig. 6.5). It is obvious that a sharp transition zone exists. This can be explained by the fact that the type of cations affects the diffusion between the layers. For B2–A1 laminate, there is a gradual transition from an α-rich region to β-rich one in XRD analyses. The transition zone is about 400 µm in thickness, and the β-SiAlON proportion in this zone changed from 20% to 70% as expected (Fig. 6.6). The gradual change in the amount of phases was also
A1
B1
6.4 Basic demonstration of powder lamination method.
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
163
6.5 BE-SEM analyses of B1–A1 laminate. 90 80
A1
% α-SiAlON
70 60
Transition zone
50 40 30 20
B2
10 0 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.6 Change in α-SiAlON ratio of sintered B2–A1 laminate ground up to 100 µm through the sample thickness. 50 µm
50 µm
Transitional zone
B2
50 µm
A1
6.7 BE-SEM analyses of B2–A1 laminate.
confirmed by SEM (Fig. 6.7) in that counter-diffusion of species had taken place between B2 and A1. XRD measurements and microstructural observations coincided with the hardness measurements, as shown in Fig. 6.8. Hardness of the intermediate zone increased from 15 GPa to 19 GPa in moving from the β-SiAlON-rich side to the α-SiAlON-rich side. Another combination was prepared by using powders of B3 and A1. This laminate also showed a gradual transition zone similar to the B2–A1 pair
© Woodhead Publishing Limited, 2006
164
Ceramic matrix composites 20
A1
Hardness (GPa)
19 Transition zone 18 17 16
B2
15 14 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.8 Change in the hardness measurements of B2–A1 laminates along the sample thickness. 90 80
A1
% α-SiAlON
70 Transition zone
60 50 40 30 20
B3
10 0 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.9 Change in α-SiAlON ratio of sintered B3–A1 laminate ground up to 100 µm through the sample thickness.
(Fig. 6.9). Hardness values increased gradually from 16 GPa to about 19 GPa, shifting from the β-SiAlON-rich region to the α-SiAlON-rich one (Fig. 6.10). However, the gradient layer was longer in this sample, which may be due to easier diffusion of species in a larger amount of liquid phase in B3. In addition, the effect of sintering time and temperature on the diffusion between the layers was investigated for B3–A1 laminate. Changes in the amount of α-SiAlON from the surface of the samples sintered at different temperatures for different times (designated FGM-C, D, E) are shown in Fig. 6.11. A gradual transition from α- to β-rich regions for each FGM was observed. Sintering time and temperature did not appear to affect the diffusion between the layers. Thus, sintering at 1700°C for one hour (FGM-D) was sufficient. The gradual change in the amount of the phases was also confirmed by the SEM observations for each functionally graded SiAlON ceramic. The microstructure of each of the FGMs produced looked similar. A representative
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
165
19
Hardness (GPa)
18.5 Transition zone
A1
18 17.5 17 16.5
B3
16 15.5 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.10 Change in the hardness measurements of B3-A1 laminates along the sample thickness. 90
A1
FGMC FGMD FGME
80
%α-SiAlON
70 60 Transition zone
50 40 30
B3
20 10 0 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.11 Change in the amount of α-SiAlON from the surface of the samples sintered at different temperatures for different times (named FGM-C, D, E).
SEM image is given in Fig. 6.12. Furthermore, adding the same dopants in both layers eliminated the sharp transition zone. Along the transition zone, needle-like α-SiAlON grains were observed. This confirms that the amount of glassy grain boundary phase is high enough for needle-like α-SiAlON formation and easy diffusion between the layers. Both XRD and SEM studies support the results of hardness measurement, as illustrated in Fig. 6.13. Hardness values decreased gradually from ~19 to ~15 GPa, over a distance from α-SiAlON to β-SiAlON rich sides for each functionally graded SiAlON ceramic. The results obtained revealed that the gradient layer was longer for the B3–A1 laminate than for the B2–A1 laminate.50 This can be explained by the fact that the larger amount of liquid phase in the B3 layer facilitates the diffusion between the layers. The resulting functionally graded SiAlON ceramics may be potential candidates for wear applications.
© Woodhead Publishing Limited, 2006
166
Ceramic matrix composites
~ 50 µm
80α:20β
~ 50 µm
20α:80β
~ 50 µm
(a) ~ 50 µm
~ 50 µm
(b)
(d)
(c) ~ 50 µm
(e)
(f)
6.12 BE-SEM analyses of FGM-C laminate. 20 FGM C FGM D FGM E
Hardness (GPa)
19 A1 18 17 16
B3 15 14 0
0.2
0.4
0.6 0.8 1 1.2 1.4 Distance from surface (mm)
1.6
1.8
2
6.13 Change in the hardness measurements of FGM-C, D, E laminates along the sample thickness.
Tapes lamination Tape casting is a process for producing thin single-layer or stacked and laminated multilayer structures for which adequate thickness control is required.53 Although there are some publications on the production of Si3N4 ceramics by the tape casting method,54,55 there is only one report available on the preparation of SiAlON sheets by tape casting.56 According to this study, Xu and co-workers prepared α-SiAlON sheets by tape casting and pressureless sintering at 1750°C for 2 hours. Preparation of ceramic bodies by tape casting has been considered to be
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
167
one of the most favorable and visible routes in functionally graded materials production.57–61 According to these studies, tape casting is a candidate to produce multifunctional materials due to an adjustable layer thickness of each component. When the variations in compositions between layers are relatively small, the material approaches a continuous composition, and stress concentrations at the interfaces can be decreased to practically insignificant values. In the literature, there has been no report about production of functionally graded SiAlON ceramics by tape casting. The main advantage of this method with respect to others is that continuous change in composition, microstructure and mechanical properties can be obtained by stacking controlled layer thicknesses of different tape compositions. For the production of functionally graded SiAlON ceramics, five different SiAlON compositions were designated for tape casting. These are 85α:15β, 70α:30β, 55α:45β, 40α:60β and 25α:75β from surface to bottom.62 The tape casting was carried out with a laboratory tape caster on a glass substrate, as schematically indicated in Fig. 6.14. The blade gap was adjusted to 400 µm. After drying at room temperature, sheets were cut 1 × 1 cm in dimensions. Lamination was performed by compression of the green tapes with different compositions, where the surface was rich in α-SiAlON and the bottom rich in β-SiAlON (5 sheets × 5 compositions = 25 layers) under ~7 MPa pressure at room temperature (Fig. 6.15). The laminates were further cold isostatically Slurry Wet tape
Direction of movement
Substrate
6.14 Schematic illustration of tape production. High α-SiAlON content
A B C
High β-SiAlON content
6.15 Stacking of tapes with different compositions.
© Woodhead Publishing Limited, 2006
168
Ceramic matrix composites
pressed under 300 MPa followed by a binder burnout stage. After drying, crack-free tapes could be easily removed from the substrate; the green tapes exhibit smooth surfaces, good flexibility and adequate strength (Fig. 6.16). The thickness of the dried sheet was measured to be about 150 µm. The tapes were sintered in a gas pressure furnace at 1800°C for 60 min under 22 bar nitrogen gas pressure. Change in the amount of α-SiAlON for FGM-F from the top surface to the bottom is shown in Fig. 6.17. This clearly illustrates that α-SiAlON content changes gradually through the thickness of the sample. Hardness measurements showed high hardness at the surface (top) and low hardness at the bottom (Fig. 6.18). By comparison of the hardness measurements and scanning electron microscopy–secondary electron mode (SE–SEM) observations (Fig. 6.19), gradual change in composition and good connectivity between the layers were observed. This indicates that compaction pressure of uniaxial pressing is high enough to provide connection between the layers, providing easy removal of binders during the burnout stage without disrupting
6.16 Photograph of green tape. 90 80
% α-SiAlON
70 60 50 40 30 20 10 0 0
0.1
0.2
0.3
0.4 0.5 0.6 0.7 0.8 Distance from surface (mm)
0.9
1
1.1
6.17 Change in the amount of α-SiAlON from the surface of the samples FGM-F.
© Woodhead Publishing Limited, 2006
1.2
SiAlON based functionally graded materials
169
18
Hardness (GPa)
17.5 17 16.5 16 15.5 15 14.5 0
0.2
0.4 0.6 0.8 1 Distance from surface (mm)
1.2
1.4
6.18 Hardness measurements of designed FGM-F through the thickness of the sample.
200 µm
6.19 SE-SEM analyses of FGM-F.
the physical integrity of the compact. Microstructural observations verify the effect of further cold isostatic pressing to achieve homogeneity in the sample. As can be seen from the representative SEM image of FGM-F (see Fig. 6.20), tape casting is a good candidate for obtaining continuous changes in composition, microstructure and thus mechanical properties through the material thickness. The internal stresses can be minimized relative to multifunctional materials produced by other methods. This commonly leads to improved material performance by the tape casting method. Thus, producing SiAlON ceramics through the tape casting method may extend the application areas of SiAlON ceramics.
© Woodhead Publishing Limited, 2006
170
Ceramic matrix composites
85α:15β (a)
(b)
25α:75β (c)
6.20 BE-SEM analyses of FGM-F.
6.5.2
Powder bed technique
Previously, Chen and co-workers developed graded in-situ SiAlON ceramics by embedding β-SiAlON green compacts in an α-SiAlON powder bed.63 Their results showed that compositions, microstructures and properties of the graded SiAlON ceramic change gradually from the hard α-SiAlON with spherical morphology on the surface to the tough and strong β-SiAlON with elongated grains in the core, but the graded layer was limited to ~300 µm. Recently, Jiang and Kang developed a technique for in-situ formation of an α-SiAlON layer on a β-SiAlON surface.64 This technique consists of packing a compact of β-SiAlON composition in an α-SiAlON powder bed. They found that it was possible to control the thickness of the α-SiAlON-rich layer by changing the presintering conditions during heating to sintering temperature. Thickness of the α-SiAlON layer increased with increase at the presintering temperature. However, this can lead to grain growth and resultant decrease in strength. Thus, alternative methods must be developed.
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
171
In another study, Mandal and co-workers produced functionally graded SiAlON ceramics using the powder bed method.50 In their study, β-SiAlON compacts were embedded in two different homogeneously mixed powder bed compositions, α-SiAlON (100 wt%) and AlN:BN (50:50 wt%). The effects of powder bed composition and pressure on the formation of αSiAlON on the compact surface were investigated. For the powder bed method, a composition rich in β-SiAlON (named B2) was selected as a compact composition to observe compact–powder bed interaction. Two different powder bed compositions, α-SiAlON (100 wt%) and AlN:BN mixture (50:50 wt%), were prepared. β-SiAlON-rich pellets were embedded into the powder bed compositions, as schematically shown in Fig. 6.21. Both green and sintered pellets were embedded into the same powder bed composition in order to compare the effect of presintering on the interaction. Sintering of the pellets was carried out under 22 bar nitrogen gas pressure at 1800°C for 1 hour. To understand the effect of pressure on the interaction zone, pressureless sintering was also carried out for comparison. Changes in the amount of β-SiAlON from the surface of the samples for green and presintered compacts are shown in Figs 6.22 and 6.23, respectively, after sintering in α-SiAlON and AlN–BN powder beds. Both figures clearly illustrate that the AlN–BN powder bed is more effective for the formation of α-SiAlON at the surface than the α-SiAlON powder bed. Formation of an αSiAlON-rich layer on the component surface is due to the transfer of αSiAlON-forming ions from the powder bed. This transfer could be in various
BN crucible Pellet Powder bed
% β-SiAlON
6.21 Basic demonstration of powder bed method. 90 80 70 60 50 40 30 20 10 0
AlN:BN α-SiAlON 0
0.1
0.2
0.3 0.4 0.5 0.6 0.7 Distance from surface (mm)
0.8
0.9
1
6.22 Change in β-SiAlON ratio of green samples sintered in different powder beds.
© Woodhead Publishing Limited, 2006
Ceramic matrix composites
% β-SiAlON
172
90 80 70 60 50 40 30 20 10 0
AlN:BN α-SiAlON 0
0.1
0.2
0.3 0.4 0.5 0.6 0.7 Distance from surface (mm)
0.8
0.9
1
6.23 Change in β-SiAlON ratio of presintered samples sintered in different powder beds.
% β-SiAlON
ways: (i) solid-state reaction of the powder bed with the compact surface; (ii) formation of a liquid in the powder bed and its reaction with the surface; (iii) evaporation from the powder bed and condensation on the compact. The first mechanism appears to be unlikely as the contact area of fine powder of the loose powder bed should be rather low for effective solid-state reaction. The second mechanism is not applicable to an AlN:BN powder bed as no liquid formation is expected. For α-SiAlON powder bed, liquid formation should occur but not as much as one can expect due to the loose powder bed. Therefore, it is likely that the material transfer occurs via vaporization of reactants from the powder bed and condensation on the surface of the compact for reaction. It is possible that AlN evaporates from the AlN:BN powder bed which then reacts with the compact surface. Consequently, the surface composition becomes richer in AlN, causing α-SiAlON formation. Since less AlN is available from the α-SiAlON powder bed, its effect on α-SiAlON formation at the compact surface is naturally less. Figure 6.24 gives a comparison of the amount of β-SiAlON change of the green and presintered compacts in AlN:BN powder. The green compacts had 90 80 70 60 50 40 30 Green Presintered
20 10 0 0
0.1
0.2
0.3 0.4 0.5 0.6 0.7 Distance from surface (mm)
0.8
0.9
1
6.24 Comparison of the amount of β-SiAlON on the surfaces of the green and presintered compacts in AlN-BN powder bed.
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
173
90
% β-SiAlON
80 70 60 50 40 30 20
Sintered under 22 bar Sintered under 1 bar
10 0 0
0.1
0.2
0.3 0.4 0.5 0.6 0.7 Distance from surface (mm)
0.8
0.9
1
6.25 Comparison of the amount of β-SiAlON on the surfaces of green compacts sintered under 22 bar and 1 bar in AlN:BN powder bed.
more interaction with the powder bed than the presintered compacts. This is presumably due to easy diffusion of condensed phases through the less dense structure of β-SiAlON just before reaching full densification. The maximum thickness of the gradient zone is limited to 400 µm. The effect of nitrogen gas pressure during sintering on the thickness of the diffusion zone is determined in the AlN:BN powder bed by using 1 bar and 22 bar gas pressure. The results indicated that change in gas pressure during sintering has an influence on the thickness of the gradient zone (Fig. 6.25). The smaller gradient zone thickness at 1 bar can be attributed to sweeping of evaporating species away by flowing nitrogen gas used during sintering. As a result, it is obvious that the powder bed technique is not as effective for the production of functionally graded SiAlON ceramics as the lamination method.
6.5.3
Controlling the sintering conditions
Previously, Mandal and co-workers obtained a gradual change of α-SiAlON content from the surface through the core by rapid cooling.42 However, the aim was different from the functionally graded material concept. Recently, they have used this technique in production of functionally graded SiAlON ceramics. The achievement of gradual changes is explained as a function of α → β SiAlON transformation. In the fast cooling method, two different samples were heated and cooled very rapidly. The first was B2 (same composition in powder lamination method); the second was B4, obtained from B2 by sintering in an AlN:BN powder bed. The surface of B2 was ground until 70% β-SiAlON was obtained before fast cooling treatment. Then, the specimens were inserted in a furnace with high-speed cooling. They were heated to 1600°C at a rate of 15°C/min and cooled rapidly by immediate removal from the furnace.
© Woodhead Publishing Limited, 2006
Ceramic matrix composites
% β-SiAlON
174
90 80 70 60 50 40 30 20 10 0 0
0.05
0.1 0.15 0.2 0.25 0.3 Distance from surface (mm)
0.35
0.4
6.26 Change in β-SiAlON ratio in B1 sample ground up to 70% βSiAlON content as a function of fast cooling.
The B2 sample was treated to investigate the effect of the fast cooling method on the surface modification. There was no gradient layer after fast cooling on the surface of B2. This can be explained by the stability of βSiAlON, not transforming to α-SiAlON on heating. When the B4 sample was used in fast cooling experiments, an ~ 300 µm gradient zone was obtained at the surface (Fig. 6.26). This is probably due to its modified surface composition during sintering in the AlN:BN powder bed becoming more transformable. As a result, this method is not practical, since gradient zone thickness is limited to about 300 µm and the specimen may crack during rapid cooling. It is evident that the gradient zone thickness on the surface varies between 200 and 400 µm, but is limited to about 400 µm maximum in the powder bed and fast cooling methods. Considering that ceramic products usually need a machining operation to achieve near-net shape, and the amount of material removal during machining is about a few hundred microns depending on the application, neither method appears to be applicable. Furthermore, the methods are not practical in that the powder bed needs to be renewed at each sintering cycle and the specimen may crack during rapid cooling. Therefore, the lamination method gives more promise for functionally graded SiAlONs. This is because one can adjust physically the layer thickness as required, e.g. depending on machining allowance.
6.6
Concluding remarks
Since the macroscopic properties of a material are strongly dependent on its structure, studies of structure/property relationships are among the most important issues in material science. The properties of SiAlON ceramics are strongly influenced by their microstructure and chemical composition. Although the functionally graded materials concept has been known since 1972, the production of functionally graded SiAlON ceramics is still a subject to explore. Especially, the formation mechanisms of the functional gradient structure have not yet been fully understood. If diffusion mechanisms are explored, compositional design would be easy and the thickness of the graded
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
175
structure could be adjusted. In addition, by using different SiAlON compositions, new materials with superior properties over either monolithic β-SiAlON or α-SiAlON and α-β-SiAlON composites can be developed for industrial applications.
6.7
References
1. Shen, M., Bever, M.B. (1972), ‘Gradients in polymeric materials’, J. Mat. Sci., 7, 741–746. 2. Hirai, T. (1996), Materials Science and Technology, A Comprehensive Treatment, Vol. 17B, ed. R.J. Brook, VCH Verlags, Weinheim, Germany. 3. Holt, J.B., Koizumi, M., Hirai, T., Munir, Z.A. (1993), Ceramic Transactions, Am. Ceram. Soc., Westerville, OH. 4. Damzik, R.J., Neubrand, A., Rödel, J. (2000), ‘Functionally graded materials by electrochemical processing and infiltration: application to tungsten/copper composites’, J. Mat. Sci., 35, 477–486. 5. Suresh, S., Mortensen, A. (1997), ‘Functionally graded metals and metal ceramic composites: thermomechanical behaviour’, Int. Mater. Rev., 42(3), 85–116. 6. Koizumi, M. (1997), ‘FGM activities in Japan’, Composites, Part B, 28B, 1–4. 7. Shin, H.K., Natu, H., Dutta, D., Mazumder, J. (2003), ‘A method for the design and fabrication of heterogeneous objects’, Materials and Design, 24, 339–353. 8. Ichikawa, K. (2001), Functionally Graded Materials in the 21st Century, Kluwer Academic, Boston, MA. 9. Marple, B.R., Boulanger, J. (1994), ‘Graded casting of materials with continuous gradients’, J. Am. Ceram. Soc., 77, 2747–2750. 10. Kieback, B., Neubrand, A., Riedel, H. (2003), ‘Processing techniques for functionally graded materials’, Mater. Sci. Eng., A362, 81–105. 11. Parameswaran, V., Shukla, A. (2000), ‘Processing and characterization of model functionally graded materials’, J. Mat. Sci., 35, 21–29. 12. Carrillo-Heian, E.M., Carpenter, R.D., Paulino, G., Gibeling, J.C., Munir, Z. (2001), ‘Dense layered molybdenum disilicide–silicon carbide functionally graded composites formed by field-activated synthesis’, J. Am. Ceram. Soc., 84, 962–968. 13. Jack, K.H. (1976), ‘Review: sialons and related nitrogen ceramics’, J. Mater. Sci., 11, 1135. 14. Jack, K.H., Wilson, W.I. (1972), ‘Ceramics based on the Si–Al–O–N and related systems’, Nature (London) Phys. Sci., 238, 28–29. 15. Oyama, Y. (1971), ‘Solid solution in the ternary system Si3N4–AlN–Al2O3’, Jpn. J. App. Phys., 10, 1687. 16. Grun, R. (1979), ‘The crystal structure of β-Si3N4’, Acta Crystallogr, B35, 800–804. 17. Kohatsu, I., McCauley, J.W. (1974), ‘Re-examination of the crystal structure of αSi3N4’, Mater. Res. Bull., 9, 917–920. 18. Ekström, T., Ingelström, I. (1986), ‘Characterization and properties of SiAlON materials’, in Proc. Int. Conf. Non-oxide Technical and Engineering Ceramics, ed S. Hampshire. Elsevier Applied Science, London, 231–253. 19. Ekström, T., Nygren, M. (1992), ‘SiAlON ceramic’, J. Am. Ceram. Soc., 75, 259– 276. 20. Thompson, D.P. (1989), Preparation and Properties of Silicon Nitride Based Materials, ed. D.A. Bonnell and T.Y. Tien, Materials Science Forum, Vol. 47, Trans Tech Publications, Aedermannsdorf, Switzerland.
© Woodhead Publishing Limited, 2006
176
Ceramic matrix composites
21. Jänecke, E.Z. (1907), ‘Über eine neue Darstellungsform der van’t Hoffschen Untersuchung über ozeanische Salzablagerungen’, Z Anorg. Chem., 53, 319–326. 22. Kijama, K., Shirasaki, S. (1976), ‘Nitrogen self diffusion in silicon nitride’, J. Chem. Phys., 65, 2668. 23. Deeley, G., Herbert, J. (1961), ‘Dense Silicon Nitride’, Powder Metall., 8, 145. 24. Munakata, H., Hayashi, T., Suzuki, H., Saito, H. (1986), ‘Presureless sintering of Si3N4 with Y2O3 and Al2O3’, J. Mat. Sci., 21, 3501–3508. 25. Hwang, S.L., Chen, I.W. (1994), ‘Reaction hot pressing of α and β-SiAlON ceramics’, J. Am. Ceram. Soc., 77, 165–171. 26. Olsson, P.O., Ekström, T. (1990), ‘HIP sintered β and mixed α-β SiAlONs densified with Y2O3 and La2O3 additions’, J. Mat. Sci., 25, 1824–1832. 27. Li, H.X., Sun, W.Y., Wang, P.L., Yan, D.S., Tien, T.Y. (1997), ‘The effect of GPS parameters on mechanical properties of Y-α-SiAlON ceramics’, Ceramics International, 23, 449–456. 28. Biswas, S.K., Riley, F.L. (2001), ‘Gas pressure sintering of silicon nitride – current status’, Mat. Chem. Phys., 67, 175–179. 29. Kuwabara, M., Benn, M., Riley, F.L. (1980), ‘The reaction hot-pressing of compositions in the system Al–Si–O–N corresponding to (β-SiAlON)’, J. Mat. Sci., 15, 1407– 1416. 30. Haviar, M., Johannesen, O. (1988), ‘Unit cell dimensions of β-SiAlON’, Adv. Ceram. Mater., 3, 405–407. 31. Pezzotti, G., Kleebe, H., Okamoto, K., Ota, K. (2000), ‘Structure and viscosity of grain boundary in high purity SiAlON ceramics’, J. Am. Ceram. Soc., 83, 2549– 2555. 32. Kishi, K., Umebayashi, S., Tani, E. (1990), ‘Influence of microstructure on the strength and fracture toughness of β-SiAlON’, J. Mat. Sci., 25, 2780–2784. 33. Hampshire, S., Park, H.K., Thompson, D.P., Jack, K.H. (1978), ‘α-SiAlON ceramics’, Nature, 274, 880. 34. Ekström, T. (1993), ‘SiAlON ceramics sintered with yttria and rare earth oxides’, in Materials Research Society Symposium Proceedings, 121. 35. Hwang, C.J., Susintzky, D.W., Beaman, D.W. (1995), ‘Preparation of multication αSiAlON containing strontium’, J. Am. Ceram. Soc., 78, 588. 36. Mandal, H. (1999), ‘New developments in α-SiAlON ceramics’, J. Eur. Ceram. Soc., 19, 2349–2357. 37. Jack, K.H. (1973), ‘Nitrogen ceramics’, Trans J. Br. Ceram. Soc., 72, 376. 38. Ekström, T., Holmström, M., Olsson, P.O. (1991), ‘Yttria doped Si2N2O’, in Proc. 4th Int. Symp. on Ceramic Materials and Components for Engines, Amsterdam. 39. Trigg, M.B., Jack, K.H. (1988), ‘The fabrication of O-SiAlON ceramics by pressureless sintering’, J. Mat. Sci., 23, 481–487. 40. Ekström, T. (1989), ‘Effect of composition, phase content and microstructure on the performance of yttrium SiAlON ceramics’, Mat. Sci. Eng., A109, 341–349. 41. Cao, G.Z., Metselar, R., Ziegler, G. (1992), ‘Microstructure and properties of mixed α-β SiAlONs’, in Proc. 4th Int. Symp. on Ceramic Materials and Components for Engines, Amsterdam, p. 188. 42. Mandal, H., Thompson, D.P., Ekström, T. (1993), ‘Reversible α↔β phase transformation in heat treated SiAlON ceramics’, J. Eur. Ceram. Soc, 12, 421. 43. Ekström, T., Shen, Z.J. (1995), ‘Temperature stability of rare earth doped α-SiAlON ceramics’, in Proc. 5th Int. Symp. on Ceramic Materials and Components for Engines, p. 206.
© Woodhead Publishing Limited, 2006
SiAlON based functionally graded materials
177
44. Zhao, R., Cheng, Y.B. (1996), ‘Decomposition of Sm α-SiAlON phases during post sintering heat treatment’, J. Eur. Ceram. Soc., 16, 1001. 45. Camuscu, N., Thompson, D.P., Mandal, H. (1997), ‘Effect of starting composition, type of rare earth sintering additive and amount of liquid phase on α ↔ β phase transformation’, J. Eur. Ceram. Soc., 17, 599. 46. Mandal, H., Thompson, D.P., Liu, Q., Gao, L. (1997), ‘High temperature stability of α-SiAlON ceramics containing glass additions’, Eur. J. Solid. State. Inorg. Chem., 34, 179. 47. Klemm, H., Hermann, M., Reich, T., Schubert, C. (1998), ‘High temperature properties of mixed α-β SiAlON materials’, J. Am. Ceram. Soc., 81, 1141–1148. 48. Rabin, B.H., Heaps, R.J. (1993), in Holt, J.B., Koizumi, M., Hirai, T., Munir, Z.A. (eds), Ceramic Transactions, Vol. 34, Functionally Gradient Materials, Am. Ceram. Soc., Westerville, OH, pp. 173–180. 49. Shen, Z., Nygren, M. (2002), ‘Laminated and functionally graded materials prepared by spark plasma sintering’, Key Eng. Mater., 206–213, 2155–2158. 50. Çalıfl, N., Kuflhan, fi.R., Kara, F., Mandal, H. (2004), ‘Functionally graded SiAlON ceramics’, J. Eur. Ceram. Soc., 24, 3387–3393. 51. Çalıfl, N., Kara, A., Kara, F., Mandal, H. (2004), ‘Development of laminar type functionally graded SiAlON ceramics’, Key Eng. Mater., 264–268, 1095–1098. 52. Çalıfl, N. (2002), ‘Functionally graded SiAlON ceramics’, BSc Thesis, Anadolu University, Turkey. 53. Mistler, R.E., Twiname, E.R. (2000), ‘Tape casting theory and practice’, Am. Ceram. Soc., Westerville, OH, pp. 62–82. 54. Bitterlich, B., Heinrich, J.G. (2002), ‘Aqueous tape casting of silicon nitride’, J Eur. Ceram. Soc., 22, 2427–2434. 55. Gutierrez, C.A., Moreno, R. (2000), ‘Tape casting of non-aqueous silicon nitride slips’, J. Eur. Ceram. Soc., 20, 1527–1537. 56. Xu, X., Mei, S., Ferreira, J.M.F. (2003), ‘Fabrication of α-SiAlON sheets by tape casting and pressureless sintering’, Department of Ceramics and Glass Engineering, CICECO, University of Aveiro, Portugal, www.mrs.org/publications/jmr/jmra/2003/ jun/012.html. 57. Jung, Y.G., Ha, C.G., Shin, J.H., Hur, S.K., Paik U. (2002), ‘Fabrication of functionally graded ZrO2/NiCrAlY composites by plasma activated sintering using tape casting and its thermal barrier property’, Mat. Sci. Eng., A323, 110–118. 58. Yeo, J.G., Jung, Y.G., Choi, S.C. (1998), ‘Design and microstructure of ZrO2/SUS316 functionally graded materials by tape casting’, Materials Letters, 37, 304–311. 59. Dumont, A.L., Bonnet, J.P., Chartier, T., Ferreira, J.M.F. (2001), ‘MoSi2/Al2O3 FGM: elaboration by tape casting and SHS’, J. Eur. Ceram. Soc., 21, 2353–2360. 60. Chartier, T., Merle, D., Besson, J.L. (1995), ‘Laminar ceramic composites’, J. Eur. Ceram. Soc., 15, 101–107. 61. Carlström, E., Kristoffersson, A. (2002), ‘Water based tape casting and manufacturing of laminated structures’, Key Eng. Mat., 206–213, 205–210. 62. Çalıfl, N. (2004), ‘Functionally graded SiAlON ceramics’, MSc Thesis, Anadolu University, Turkey. 63. Chen, L., Kny, E., Groboth, G. (1998), ‘SiAlON ceramics with gradient microstructures’, Surface and Coating Technology, 100–101, 320–323. 64. Jiang, X., Kang, L. (1998), ‘Formation of α-sialon layer on β-sialon and its effect on mechanical properties’, J. Am. Ceram. Soc., 81, 1907–1919.
© Woodhead Publishing Limited, 2006
7 Design of tough ceramic laminates by residual stresses control N O R L O V S K A Y A, Drexel University, USA, M L U G O V Y, Institute for Problems of Materials Science, Ukraine, J K U E B L E R, EMPA, Lab for High Performance Ceramics, Switzerland, S Y A R M O L E N K O and J S A N K A R, North Carolina A&T State University, USA
7.1
Introduction
Ceramics have found their use in numerous crosscutting industrial applications because of excellent hardness, wear, corrosion resistance, and ability to withstand high temperatures. However, ceramics’ reliability and ductility compared to metals are not very high. The best approach to increasing the fracture toughness which enables the structural application of ceramics is through the development of ceramic composites. Fiber-reinforced composites demonstrate the highest fracture toughness and damage tolerance. However, since these materials have a very high density of weak interfaces, they are not very strong. In addition, their high cost limits their commercial applications. Particulate composites are less expensive to manufacture, but compared to monolithic ceramics, their fracture toughness increases are insignificant. Several publications on ceramics show that the use of layered materials is the most promising method for controlling cracks by deflection, bifurcation, microcracking, or internal stresses [1–4]. Layered structures clearly offer a key to greater reliability at a moderate cost and new applications may result as more complex structures are tailored to specific applications [5]. The way to achieve the highest possible mechanical properties is to control the level of residual stresses in individual layers. One can increase the strength and apparent fracture toughness of ceramics by creating a layer with compressive stresses on the surface. In such a way, surface cracks will be arrested and, therefore, higher failure stresses are achieved [6]. The variable layer composition, as well as the system’s geometry, allows the designer to control the magnitude of the residual stresses in such a way that compressive stresses in the outer layers near the surface increase strength, flaw tolerance, fatigue strength, resistance to oxidation, and stress corrosion cracking. In the case of symmetrical laminates, this can be done by choosing layer compositions such that the coefficient of thermal expansion (CTE) of the odd layers is smaller than the CTE of the even ones. The changes in compressive and tensile stresses depend on the mismatch of CTEs, Young’s moduli, as well as 178 © Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
179
on the thickness ratio of layers (even/odd) [7, 8]. The sign and value of residual stresses can be established by theoretical prediction [9–12]. There have also been a number of experimental studies of laminated ceramics that were conducted using these models, attempting to maximize the mechanical properties [13, 14]. This chapter consists of the following sections: 7.1 Introduction, 7.2 Laminate design for enhanced fracture toughness, 7.3 Processing of Si3N4– TiN and B4C–SiC ceramic laminates, 7.4 Si3N4 based laminates, 7.5 B4C based laminates, and 7.6 Future trends. Section 7.2 outlines the main approach to optimizing layered structures leading to an increase of mechanical properties. The technique describes the design of a layered composite with enhanced fracture toughness. The mechanisms responsible for the increase in mechanical performance and reliability, acting in the layered ceramics, are considered here, too. The main technological steps in manufacturing of ceramic laminates are considered in Section 7.3. Section 7.4 describes the mechanical performance of Si3N4 based laminates, and Section 7.5 presents new results on the mechanical behavior of B4C based armor laminate materials. A new design with incremental increase of the apparent fracture toughness of laminates is mentioned in Section 7.6, and a list of references is provided for further information and reading.
7.2
Laminate design for enhanced fracture toughness
A number of articles on the design of ceramic laminates leading to a significant increase of their mechanical properties were published in the past [15–19]. Our work is based on the control of thermal residual stresses by optimization of the layered structure [20, 21]. The proposed approach targets the fracture toughness increase of laminate ceramic composites and is based on the preliminary results both from our work [22, 23] and from the work of others [24–26].
7.2.1
Calculation of the residual stresses
A schematic presentation of a general case of two-component multilayered sample is shown in Fig. 7.1, where ti is the thickness of the ith layer, w is the total thickness of the specimen, b is the width, and N is the total number of layers. Often the two-component brittle layered composites with symmetric macrostructure are only considered. In this case the layers consisting of different components alternate one after another and the external layers consist of the same component. The total number of layers N in such a composite sample is odd. Often the layer of each component has some constant thickness, and the layers of same component have identical thickness. In this case all
© Woodhead Publishing Limited, 2006
180
Ceramic matrix composites x N
w
ti
i
xi 2
y
1 Surface under tension
7.1 Scheme of a two-component multilayer specimen: numbers of layers and layer boundary coordinates.
layers of the first component including two external (top) layers can be designated by index 1 (j = 1), and all layers of the second component (internal) can be designated by index 2 (j = 2). The number of layers designated by index 1 is (N + 1)/2, and the number of layers designated by index 2 is (N – 1)/2. There are effective residual stresses in the layers of each component in the layered ceramic composite. During cooling, the difference in deformation, due to the different thermal expansion coefficients of the components, is accommodated by creep as long as the temperature is high enough. Below a certain temperature, which is called the ‘joining’ temperature, the different components become bonded together and internal stresses appear. The ‘joining’ temperature is difficult to measure experimentally, and in general, it is adopted to be somewhere below the sintering temperature. In each layer, the total strain after sintering is the sum of an elastic component and a thermal component [27, 28]. The residual stresses in the case of a perfectly rigid bonding between the layers of a two-component material are [28]:
σ r1 =
E1′ E 2′ f 2 (α T 2 – α T 1 ) ∆T E1′ f1 + E 2′ f 2
(7.1)
σ r2 =
E 2′ E1′ f1 (α T 1 – α T 2 ) ∆T E1′ f1 + E 2′ f 2
(7.2)
and
where E ′j = Ej/(1 – νj), f1 =
© Woodhead Publishing Limited, 2006
( N + 1) l1 ( N – 1)l 2 , f2 = , Ej and νj are the 2w 2w
Design of tough ceramic laminates by residual stresses control
181
elastic modulus and Poisson’s ratio of the jth component respectively, l1 and l2 are the thickness of layers of the first and second components, αT1 and αT2 are the thermal expansion coefficients (CTE) of the first and second components respectively, ∆T is the difference in temperature of the ‘joining’ temperature and current temperature, and w is the total thickness of the specimen. The mismatch of thermal expansion coefficients between different layers inevitably generates thermal residual stresses during subsequent cooling of layered ceramics with strong interfaces [28]. The relative thickness of different layers determines the relative magnitudes of compressive and tensile stresses, while the strain mismatch between the layers dictates the absolute values of the residual stresses. The important trend is a decreasing of tensile residual stress and an increasing of compressive residual stress with an increase of the thickness of layers under tension and a decrease of the thickness of layers under compression. In this way a change of a layer thickness ratio allows control of the residual stress level in laminates.
7.2.2
Design principles and algorithm
While a number of mechanical properties, such as strength and reliability, can be increased by design of laminates, we specifically target an increase of the fracture toughness in our study. In the case of non-homogeneous (particularly layered) materials, the so-called apparent fracture toughness should be considered. The apparent fracture toughness is the fracture toughness calculated from test data of the layered sample considering this specimen as ‘homogeneous’. Such an approach does not meet the fracture mechanics requirement of taking into account all features of stress distribution near the crack tip in layered media, but it is still a useful characteristic allowing an effective contribution of such factors as residual stresses and material inhomogeneity to be accounted for. In fracture mechanics, both residual and applied stresses are usually included in the crack driving force. It can be useful to consider residual stresses as a part of the crack resistance, and therefore in laminates with residual compressive stress, the higher resistance to failure results from a reduction of crack driving force rather than from an increase in intrinsic material resistance to crack extension [29]. An apparent fracture toughness of laminate can be successfully modeled taking into account a layered structure. In this case the modeling is a power tool of laminate design because it allows prediction of the mechanical behavior of layered material with a crack. The compressive residual stress σr1 in the outside layers of a laminate shields natural and artificial cracks in the layer. Therefore, the effective (apparent) fracture toughness of such a structure increases. The more compressive residual stress induced, the more shielding occurs. Another important factor that contributes to the apparent fracture toughness increase
© Woodhead Publishing Limited, 2006
182
Ceramic matrix composites
is the crack length a. A longer crack promotes more shielding. The maximum length of a transverse crack in an outside compressive layer is limited by the layer thickness l1. These two factors determine the apparent fracture toughness of the material. In general, the condition for crack growth onset is Ka + Kr = Kc, (Fig. 7.2), where Ka = Ka(σa, a) is the applied stress intensity factor that can be measured, σa is the distribution of applied stress resulted from bending, Kr = Kr(σr, a) is the stress intensity factor due to a residual stress, and Kc is the intrinsic fracture toughness of a material in the layer. If a condition of a crack growth onset is fulfilled then Ka = Kc – Kr is the apparent fracture toughness. If σr1 is compressive, then Kr < 0 and Ka increases. The more | σr1 |, the greater Ka. Similarly, the greater a, the higher Ka. The largest value of a crack length in compressed layer is l1. The maximum apparent fracture toughness can be obtained for such a crack. Unfortunately, small cracks have Ka close to Kc. A schematic presentation of factors that affect apparent fracture toughness is shown in Fig. 7.3. The contribution of residual stress to the maximum apparent fracture toughness is Kr = Y(l1/w)σr1 l11/2 , where Y(l1/w) is the geometrical factor [13, 14, 30], and w is the total thickness of a layered specimen. The factor Y ( l1 / w )l11/2 increases as l1 increases (Fig. 7.3(a)). The compressive residual stress decreases as l1 increases. It can be calculated using equation (7.1). In addition, the residual stress depends on the number of layers in the sample (Fig. 7.3(b)). The final dependences of Ka on l1 for various numbers of layers are shown in Fig. 7.3(c). These dependences are non-monotonic curves with the maximum that depends on a number of layers in the laminate. The labels w/5, w/4, w/3 and w/2 designate the maximum thickness of the top layer for symmetrical layered structures with nine, seven, five and three layers, respectively. It can be seen that the highest apparent fracture toughness can be obtained for the three-layer specimen. Thus, the study of the layers’ relative thickness and their numbers reveals that the maximum crack shielding will be achieved for three-layer composites with an edge crack extending to the first interface (l1 = w/4) [13, 14]. However,
σa
σr
a σr
σa
7.2 Factors affecting the apparent fracture toughness in a laminate: σa is the distribution of applied stress resulted from bending; σr is the residual compressive stress; a is the crack length.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
183
1/2
Y(l1/w) l1
(a)
σr 9 layers 7 layers 5 layers (b) 3 layers
Ka 3 layers (c)
Kc w/5 w/4
w/3
w/2 l1
7.3 Factors affecting laminate design for maximum apparent fracture toughness.
the multilayered design is also very important to meet specific requirements, such as ballistic impact, for example. It is essential during impact loading to have more barriers to arrest cracks. In our case it is the number of compressive layers. For a three-layer design there is only one such barrier that is a top compressive layer. The top layer plays a key role, however, multilayered design is of further importance to stop cracks more effectively [3, 4]. The design technique to obtain the enhanced fracture toughness of a layered composite is as follows. First, the compositions of the layers are selected depending on the intended application of the composite. Then, the relevant material constants entering the design are determined. The constants for design are the coefficient of thermal expansion, Young’s modulus, Poisson’s ratio, and the density of the corresponding constituents. A very important but at this point of design experimentally unknown parameter is the ‘joining’ temperature. Further, effective coefficients of thermal expansion, effective Young’s modulus, average density and the thickness ratio of layers are determined using the rule of mixtures. The next step in the design is the selection of the number of layers. It can be any appropriate number depending on the required total thickness of the tile. To obtain the enhanced fracture resistance of the layered composite, the factors affecting the apparent fracture toughness should be taken into account. Usually, the thickness of the thinnest possible layer is limited by the
© Woodhead Publishing Limited, 2006
184
Ceramic matrix composites
manufacturing technology. Note that a compressive layer should be thin enough to reach a high level of residual stress. Another important requirement is the thickness ratio of layers with high CTE (tensile stress) and layers with lower CTE (compressive stress). Any appropriate thickness ratio can be used as a first approximation. Then the tensile layer thickness is found. After this, the residual stresses are calculated using equation (7.1) and (7.2). The total thickness of the sample is also determined at this step for a given layer’s thickness ratio taking into account the selected number of layers. The thickness ratio is changed after the analysis of residual stress and the total thickness of the specimen. Note that increasing the ratio of tensile layer thickness to compressive layer thickness decreases tensile residual stress. However, it can result in increasing the total thickness of the sample. After changing the thickness ratio, the calculation is repeated. Such iterations are continued to find a unique optimal layer thickness ratio that produces the maximum possible compressive residual stress, low tensile residual stress, and required total thickness of the sample. A design algorithm is presented in Fig. 7.4. The maximum possible apparent fracture toughness of the corresponding layered structure is also determined in all iterations as an indicative parameter of the design. The determination of the apparent fracture toughness uses the compressive residual stress and the thickness of an outside layer as a crack length at any given iteration. These two parameters (the compressive residual stress and the thickness of Input parameters: moduli, CTEs, etc.
Selection of composition
Layer effective properties calculation
Selection of layer number and thickness of layer under compression
Thickness ratio and tensile layer thickness determination
Residual stress calculation
Analysis of the residual stress and thickness ratio changing
Output parameters: layer thicknesses
7.4 An algorithm of laminate design.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
185
the top layer) have trends acting in opposite directions. A decrease in the top layer thickness can increase the residual stress in the layer, but it decreases the length of the maximum crack. Therefore, the maximum apparent fracture toughness was always used to analyze the optimal thickness ratio. It was shown also that in order to obtain the higher resistance to failure, the tensile layer should be made as stiff as possible (i.e., high elastic modulus), whereas the compressive layers should be as compliant as feasible (i.e., low elastic modulus) [11].
7.2.3
Calculations of the apparent fracture toughness
A weight function analysis has been used to estimate the apparent fracture toughness in laminates with residual stresses [10, 21, 31, 32]. A schematic presentation of the analyzed crack location in the layered specimen is presented in Fig. 7.5, where a is the crack length and n is the number of layers crossed by the crack. The choice of coordinate system is of great importance to the apparent fracture toughness calculations because of a significant simplification of the procedure. The most appropriate coordinate origin is on the tensile surface of the sample under bending. The geometry of the multilayered material analyzed here is such that the problem can be reduced to one dimension and that analytically tractable solutions can be used [21]. An experimental value of the apparent fracture toughness can be found using the expression [10, 21]: Ka = Y(α)σma1/2 where Y (α ) =
(7.3)
1.5 P( s1 – s 2 ) 1.99 – α (1 – α ) (2.15 – 3.93α + 2.7α 2 ) , σm = and bw 2 (1 + 2 α )(1 – α ) 3/2 x N
w
n+1 n
2
a
y
1 Surface under tension
7.5 An analyzed crack location in a layered sample.
© Woodhead Publishing Limited, 2006
186
Ceramic matrix composites
α = a/w, P is the critical load (the applied bending load corresponding to specimen failure) and s1 and s2 are outer and inner support spans of the fourpoint bending fixture. The apparent fracture toughness of a layered composite can be calculated analytically by [10, 21]: Ka =
6Y (α ) a 1/2 ( I L21 – I L 0 I L 2 )( K 1(ci ) – K r ) w 2 E n′+1
∫
a
xn
n h x , α [ I L 0 x – I L 1 ]d x + Σ E i′ i =1 a
∫
xi
x i –1
h x ,α [ I L 0 x – I L 1 ] d x a
(7.4) where K 1(ci ) is the intrinsic fracture toughness of the ith layer material, Kr is the stress intensity due to the residual stresses, h(x/a, α) is the weight function for an edge-cracked sample [9, 19, 30], xi is the coordinate of an upper boundary of the ith layer (Fig. 7.1), Ei′ = Ei/(1 – νi), and Ei and νi are the elastic modulus and Poisson’s ratio of the ith layer, respectively. The expressions for ILj (j = 0, 1, 2) and JLj (j = 0, 1) were obtained from Ref. [21] as follows: N 1 Σ E ′ ( x j +1 – x ij–1+1 ) j + 1 i=1 i i
I Lj =
N 1 Σ ε˜ E ′( x j +1 – x ij–1+1 ) i j + 1 =1 i i i
JLj =
(7.5)
(7.6)
where ε˜ i is the strain in the ith layer, which is not associated with any stress. The thermal expansion and/or volume change due to a crystallographic phase transformation might be the source of this strain. However, the case of a phase transformation is beyond the scope of this chapter. In the case of thermal expansion:
ε˜ i =
∫
Tj
β i ( T ) dT
T0
where βi(T) is the thermal expansion coefficient of the ith layer at the temperature T. T0 and Tj are the actual and ‘joining’ temperatures, respectively. If β i (T) is a linear function, ε˜ i ≤ β i > ∆ T , where ∆T = T j – T 0 ,
β i ( T0 ) + β i ( T j ) is the average value of the thermal expansion 2 coefficient in the temperature range from T0 to Tj. The stress intensity due to residual stresses is given by equation (7.7) [10, 21]: < βi ≥
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
187
1 I L21 – I L 0 I L 2
Kr =
× E n′+1 n
+ Σ Ei′ i=1
∫
a
xn
∫
xi
x i –1
h x , α [ I L 1 J L 1 – I L 2 J L 0 + ( I L 1 J L 0 – I L 0 J L 1 ) x ] dx + a h x , α [ I L 1 J L 1 – I L 2 J L 0 + ( I L 1 J L 0 – I L 0 J L 1 ) x ]dx a (7.7)
The apparent fracture toughness Ka in layered specimens can be analyzed ˜ where a˜ = Y (α ) a 1/2 . The as a function of the crack length parameter a, crack length parameter a˜ is the most appropriate to demonstrate critical conditions of crack growth. One of the advantages of this parameter is that the stress intensity factor of an edge crack for a fixed value of the applied stress σm is a straight line from the coordinate origin in the coordinate system Ka – a˜ . Since K1 = σ m a˜ , the slope of the straight line is the applied stress σm. The conditions for unstable crack growth in the internal stress field are as follows [10, 21]: K1(σm, a) = Ka(a); dK1(σm, a)/da ≥ dKa(a)/da. Using parameter a˜ , these conditions become σ m a˜ = K a ( a˜ ) and σ m ≥ dK a ( a˜ )/ da˜ , which can be reduced to: K a ( a˜ )/ a˜ ≥ dK a ( a˜ )/ da˜
(7.8)
It follows from equation (7.8) that unstable crack growth occurs if the slope of the straight line corresponding to the stress intensity factor at constant applied stress is greater than or equal to the slope of the tangent line to the fracture resistance curve at the same point (Fig. 7.6). Also the applied stress intensity factor becomes higher than the fracture resistance of the material.
7.2.4
Other toughening mechanisms in laminates
In addition to a crack shielding phenomenon that exists due to residual stress there are two other crack deflection mechanisms leading to a laminate toughening. Cracks that form in one layer can be deflected either along weak interfaces with adjacent layers [33, 34] or into layers with compressive residual stresses [35, 36]. Since we investigated laminates with strong interfaces, in our design the cracks were not deflected along interfaces; however, crack kinking and bifurcation in layers under compression has often been observed during mechanical testing (Fig. 7.7). It was shown that the crack tends to deviate as it approaches the centerline of layers with compressive stresses [7]. When the crack enters the layer with compressive stresses, tensile stresses appear near the edges of the crack. These tensile stresses are parallel to the
© Woodhead Publishing Limited, 2006
188
Ceramic matrix composites Kapp Stress intensity factor at constant applied stress
Tangent line
dKa /dã Fracture resistance
Ka /ã = σm ã
0
7.6 General criterion of stable/unstable crack growth in a brittle material.
9 mm
7.7 Crack bifurcation in the B4C–SiC woven fabric based laminate. The B4C and SiC woven fabric layers cannot be distinguished under the optical microscope.
free surfaces (crack moving direction). They appear because compressive stresses, which are perpendicular to the free surfaces, have to become zero on the free surface. These tensile stresses are maximal at the centerline of layers. The condition of kinking (and bifurcation) of a crack along the centerline is
σ r21 l1 > AK c2 where A is a constant and Kc is the fracture toughness of layers with compressive stresses [7]. The constant A takes different values for kinking and bifurcation. When compressive stresses are very high or the fracture toughness of layers is sufficiently low, the crack not only deviates at the centerline, but also bifurcates, preventing the catastrophic failure of the sample during the bending test. Compressive stresses, as well as thickness of layers, influence both crack deviation and bifurcation behavior [8, 37–39]. It was shown that a range of layer thicknesses exists where crack bifurcation can occur [40].
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
189
This range depends on the elastic constants of the layers, the layer number, the mismatch in CTEs, and ∆T. The important case of specimens with a fixed total thickness was considered in Ref. [40]. There are certain features of crack bifurcation under these conditions, such as that if the sample with a fixed total thickness has too large a number of layers there will be no bifurcation. Layer thickness and composition are important and efficient parameters to control the bifurcation in laminates. The effect is comparable with a crack bridging phenomenon [21]. The bifurcation mechanism increases the laminate fracture toughness by approximately 1.5–2 times.
7.3
Processing of Si3 N4–TiN and B4C–SiC ceramic laminates
Two systems of ceramic laminates have been chosen for manufacturing and mechanical testing of laminates – Si3N4 and B4C based ceramics. They are well-known materials for crosscutting industrial applications that can be used as cutting tools, igniters, wear parts, armor, etc. Silicon nitride is one of the most promising ceramics for structural application, with good corrosion resistance and outstanding mechanical properties. Residual stresses can be created in dense Si3N4 material by incorporating a dispersion of metal-like refractory compounds, such as nitrides or carbides. Such TiN particles create compressive tangential stresses in the surrounding matrix while remaining under tensile stress. From the point of view of chemical compatibility TiN is the most promising dispersion because of its stability in contact with Si3N4 under sintering conditions [41]. It has been demonstrated that the fracture toughness and strength of silicon nitride can be increased in composites [42]. The addition of TiN leads to increases in Young’s modulus, CTE, electrical conductivity, etc. [43]. At the same time, some properties of silicon nitride are reduced by these inclusions. Usually the metal-like refractory compounds have a much lower oxidation resistance than that of silicon nitride. Their additions result in a decrease in the oxidation resistance of the initial matrix material [44]. Also, these additions accelerate the high-temperature creep rate, which increases drastically when the TiN content is higher than 30 wt%. Boron carbide is another important ceramic material with many useful physical and chemical properties. After cubic boron nitride, it is the hardest boron-containing compound [45]. Its high melting point, high elastic modulus, large neutron capture section, low density, and chemical inertness make boron carbide a strong candidate for several high-technology applications. Due to its low density and superior hardness, boron carbide is a very promising material for lightweight ballistic protection. Boron carbide exists as a stable single phase in a large homogeneity range from B4C to B10.4C [46]. The most
© Woodhead Publishing Limited, 2006
190
Ceramic matrix composites
stable boron carbide structure is rhombohedral with a stoichiometry of B13C2, B12C3, and some other phases close to B12C3 [47, 48]. The Vickers hardness of B4C is in the range of 32–35 GPa [49]. There is an indication that the hardness of stoichiometric B4C is highest in comparison with that of boronrich or carbon-rich boron carbide compositions [50–52]. However, B4Cbased composites have a relatively low fracture toughness of 2.8–3.3 MPa m1/2 [53]. While high hardness is one of the very important requisite indicators for a material’s ballistic potential, toughness might play an equally important role in realizing that potential. Thus, materials with both high hardness and high fracture toughness are expected to yield the best ballistic performance [54]. Therefore, a significant increase in fracture toughness of boron carbidebased laminates has the potential for realization of improved armor material systems. Typically the manufacturing steps of ceramic laminates include (a) ball milling of powders in certain proportions; (b) rolling of thin tapes; (c) stacking of rolled tapes; and (d) hot pressing of the stack. A schematic presentation of the manufacturing steps is shown in Fig. 7.8. This process was used to fabricate layered specimens of the Si3N4–TiN and the B4C–SiC systems. For Si3N4 based ceramic laminates, powder mixtures of the following compositions were used: • Silicon nitride (95% of α-modification) with 2 wt% of alumina and 5 wt% of yttria as sintering aids (layers with index 1, bulk compressive stress) • The above mixture with the addition of 20–50 wt% TiN or pure TiN (layers with index 2, bulk tensile stress). α-SiC and TiN powders with grain size 1 µm and 3 µm respectively were used for layered sample fabrication. For B4C based ceramic laminates, powder mixtures of the following compositions were used: Crude rubber + petrol
Grinding
Sieving
Plasticization
Rolling
Stacking
Hot pressing
7.8 Schematic presentation of manufacturing steps of ceramic laminates.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
191
• Mixture of boron carbide with 30 wt% SiC (layers with index 1, bulk compressive stress) • Pure boron carbide (layers with index 2, bulk tensile stress). B4C and α-SiC powders with a grain size of 2–5 µm were used for laminate manufacturing. A detailed description of the manufacturing process is provided below. The mixtures of various compositions were milled in the ball mill for 48 h. The average grain size of the milled powders was about 1 µm. Crude rubber (4 wt%) was added to the mixture of powders as a plasticizer through a 3% solution in petrol. The powders were then dried up to 2 wt% residual amount of petrol in the mixture. After sieving powders with a 500 µm sieve, granulated powders were dried up to 0.5 wt% residual petrol. A roll mill with 40 mm rolls was used for rolling. The velocity of rolling was 1.5 m/min. The working pressure was about 10 MPa to obtain a relative tape density of 64%. The thickness of green tapes was 0.4–0.5 mm and the width was up to 100 mm. The formation of a thin ceramic layer is of specific importance, as the sizes of residual stress zones (tensile and compressive) are directly connected to the thickness of layers. The advantage of rolling, as a method of green layer production, is that it allows easy thickness control, achieves high green density of the tapes, and requires a rather low amount of solvent and organic additives compared to other methods such as tape casting [55]. Additional powder refinement, giving a higher sintering reactivity, might occur due to the large forces applied in the pressing zone during rolling. The modeling of rolling, as recently performed, potentially allows optimizing the process of roll compaction [56]. There is a challenging problem to produce thin tapes with a small amount of plasticizer and sufficient strength and elasticity to handle green layers after rolling. A schematic presentation of rolling is shown in Fig. 7.9. Powders are continuously supplied in the bunker and further into the deformation zone in between rolls. Powders are supplied to the deformation zone due to both the gravitational force and friction between rolls and powders. The relative density of the tape (ρr) can be calculated from
ρr =
ρp α2R 1+ hs λ
(7.9)
where ρp is the relative powder density, λ is a drawing coefficient, α is the intake angle, and R is the roll diameter. Green tapes were stacked together to form the desired layered structures and ceramic samples were prepared by hot pressing the stacked tapes. The hot pressing was performed at 1820oC and 30 MPa for 45 min for Si3N4 based laminates [22], and at 2150–2200oC and 30 MPa for 50–60 minutes
© Woodhead Publishing Limited, 2006
192
Ceramic matrix composites 1
2
3
4
5 V1
ω2
ω2
ω1
V2 7
6
1 Bunker, 2 Powders, 3 Rolls, 4 Transmission, 5 Motor, 6 Bottom support, 7 Tape (a) B4C tapes
B4C–30 wt% SiC tapes
100 mm (b)
7.9 (a) Schematic presentation of rolling; (b) a photograph of B4C and B4C–30wt%SiC rolled tapes. The thickness of an individual tape after rolling is between 0.4 and 0.5 mm.
for B4C based laminates [57]. Graphite dies were used for the hot pressing of laminates with graphite surfaces coated by a BN layer in order to prevent direct contact between graphite and ceramic material. During hot pressing of laminates, shrinkage of the individual layers occurred, and their thickness reduced to 0.15 mm after hot pressing. The interfaces between individual layers of the same composition completely disappeared and only the interface between layers of different compositions could be distinguished. Dense (95– 100% density) laminate samples were obtained. The specimens for mechanical tests were prepared by machining them from the hot pressed tiles. Standard MOR bars of dimension 50 mm × 4 mm × 3 mm were surface ground to the specification stated in EN843-1. The bars were also chamfered along the long edges with a chamfer angle at 45o to a dimension of 0.12 ± 0.03 mm. The fracture toughness was measured by the SEVNB technique [58, 59] using equation (7.3). V-notches with tip radii of the order of 10–15 µm were made in the specimens by a diamond saw followed by notching with a razor
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
193
blade with diamond abrasive to obtain a sharp notch tip. The elastic modulus was measured by a standard four-point bending technique.
7.4
Si3N4 based laminates
7.4.1
Mechanical properties
Four different laminate composites, Si3N4/Si3N4–20wt%TiN, Si3N4/2(Si3N4– 20wt%TiN), Si3N4/Si3N4–50wt%TiN and Si3N4/TiN, were chosen to study their mechanical performance. The parameters of their components, such as CTE and Young’s modulus (compiled from literature data), composition and layer thickness, are given in Tables 7.1 and 7.2. The expression 2(Si3N4– 20wt%TiN) means that the (Si3N4–20wt%TiN) layer is twice as thick as the Si3N4 layer (see Table 7.2). Besides these four designs, one more design of Si3N4/Si3N4 laminate was used as a base for comparison. The laminates of this design were prepared in the same way as the others, though all layers were of the same composition. Therefore, no residual stresses can appear during cooling. It is worth noting that both the Young’s modulus and fracture toughness of these Si3N4/Si3N4 laminates were measured to be on the same level as those of standard Si3N4 ceramics prepared by the standard powder route, which includes no rolling. The strength of the Si3N4/Si3N4 laminate was less than that of the standard Si3N4 ceramics with values of 508 ± 3.2 and 750 ± 20.7 MPa, respectively. Mechanical properties such as the strength, Young’s modulus, and fracture toughness of the laminates are presented in Table 7.3. As one can see from Table 7.3, while the strength of Table 7.1 Young’s moduli and CTE of the components Composition
E (GPa)
CTE (10–6 K–1)
Si3N4–5wt%Y2O3–2wt%Al2O3 Si3N4(5wt%Y2O3–2wt%Al2O3)–20wt%TiN Si3N4(5wt%Y2O3–2wt%Al2O3)–50wt%TiN TiN
320 335.62 364.93 440
3 3.826 5.378 9.35
Table 7.2 Calculated residual stresses in Si3N4 based laminates Composition
Si3N4/Si3N4–20wt%TiN Si3N4/2(Si3N4–20wt%TiN) Si3N4/Si3N4–50wt%TiN Si3N4/TiN
© Woodhead Publishing Limited, 2006
Thickness of layers (µm) Si3N4
Si3N4 with TiN
250 245 200 200
210 530 330 400
σcomp (MPa)
σtens (MPa)
188 280 765 2467
247 151 516 1078
194
Ceramic matrix composites Table 7.3 Mechanical properties of Si3N4 based laminates with different layer compositions Composition
σf (MPa)
Si3N4/Si3N4 Si3N4/Si3N4–20wt%TiN Si3N4/2(Si3N4–20wt%TiN) Si3N4/Si3N4–50wt%TiN Si3N4/TiN
508 356 450 158 141
± ± ± ± ±
3 76 83 15 11
E (GPa)
KIC (MPa.m1/2)
307 313 — 298 157
5.54 7.41 8.50 — 3.97
± 0.01 ± 1.79 ± 0.01 ± 0.52
Si3N4/Si3N4–20wt%TiN laminates is approximately on the same level as that of the Si3N4/Si3N4 laminates, further increases of the TiN content to 50% and 100% resulted in a significant decrease of both strength and Young’s modulus. The measured fracture toughness of the Si3N4/TiN laminates also showed a decrease similar to strength and Young’s modulus values. An explanation is sought for this reduction in mechanical properties. The Si3N4/Si3N4–20wt%TiN laminates showed an increase in apparent fracture toughness. This increase can be explained by the introduction of the residual bulk compressive stresses in the Si3N4 layers. In the case where the thicknesses of the Si3N4 and the Si3N4–20wt%TiN layers were similar, the calculated residual compressive stress was about 188 MPa and the residual tensile stress about 246.5 MPa. The measured value of the apparent fracture toughness was 7.41 ± 1.79 MPa m1/2. There was a further increase in KIC (8.5 ± 0.01 MPa m1/2) for the laminates with 20 wt% TiN when the relative thickness of the Si3N4–20wt%TiN layers was increased compared to the thickness of the pure Si3N4 layers. The reason for this is a significant increase of the residual compressive stress, and at the same time a decrease of the residual stress in the Si3N4–20wt%TiN layers (Table 7.2). However, an increase of TiN content to 50 wt% resulted in a significant increase of the residual tensile stress in the laminates. The calculated tensile stress values are higher than the tensile strength of the material, and there is therefore much cracking and a decrease in all mechanical properties (Table 7.3).
7.4.2
Apparent fracture toughness of the layered composite with residual compressive or tensile stresses in the top layer
A detailed study of the effect of the residual compressive or tensile stresses in the top laminate layers on the apparent fracture toughness values has been done for Si3N4/Si3N4–30wt%TiN. Young’s moduli of the Si3N4 and the Si3N4– 30wt%TiN monolithic samples, used as reference materials, were measured to be 308 GPa and 323 GPa, respectively. Mean values of intrinsic fracture toughness of monolith materials, measured by SEVNB, are approximately
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
195
the same both for the Si3N4 and for the Si3N4–30wt%TiN compositions and (i) were used are equal to 4 ± 1 MPa m1/2. These measured values of Ei and K IC in all following calculations. The calculated values of the apparent fracture toughness as a function of the crack length parameter a˜ in the Si3N4/Si3N4– 30wt%TiN laminate with compressive outer layers are shown in Fig. 7.10(a). The toughness increases in the layers with compressive stress with increasing crack length, and it decreases in the layers with tensile stress as the crack continues to grow. The layers with compressive and tensile stresses are shown in Fig. 7.10 in white and gray colors, respectively. As one can see, Ka reaches its maximum or minimum values as the crack approaches the interface with a new layer of an opposite stress sign. For the first Si3N4 top layer with compressive stress, the calculated apparent fracture toughness increases from 3.9 to 17 MPa m1/2 as a function of the crack length parameter. The experimentally measured Ka values, presented as solid circles in Fig. 7.10(a), show an excellent fit with the calculated values. The crack length parameters for the experimentally measured Ka were calculated from the initial notch lengths. All experimentally measured points are located on close to a straight line between the coordinate origin and the maximum Ka point at the interface between the first and second layers. Failure of all samples occurred at 351 ± 13 MPa. The calculated Ka decreases in the second Si3N4–30wt%TiN layer with a residual tensile stress from 17 to 5 MPa m1/2, followed by the next increase from 5 to 14 MPa m1/2 in the third Si3N4 layer with a residual compressive stress. The insert in Fig. 7.10(a) shows an optical micrograph of
Kapp (MPa.m1/2) Si3 N4-30% TiN Si3 N4 Si3N4-30% TiN
Si3N4
Kapp (MPa·m1/2) 8
16 6
Si3N4 Si3N4 + 30% TiN
12 4
8 4
Si3N4 + 30% TiN Si3N4
0
2
Si3N4-30% TiN
0 0
0.02
0.04 0.06 0.08 0.10 ã (m1/2) (a)
0
Si3N4 Si3 N4-30% TiN Si3N4
0.02 0.04 0.06 (b)
0.08 0.10 ã (m1/2)
7.10 Apparent fracture toughness as a function of crack length parameter a˜ in the laminate with compressive (a) and tensile (b) outer layers. Filled circles correspond to the experimental data. Inserts are optical micrographs of the two parts of Si3N4/Si3N4– 30wt%TiN laminate samples with (a) Si3N4 surface layers with a residual compressive stress and (b) Si3N4–30wt%TiN surface layers with a residual tensile stress after single-edge V-notch beam test.
© Woodhead Publishing Limited, 2006
196
Ceramic matrix composites
two parts of the Si3N4/Si3N4–30wt%TiN laminate sample with a V-notch in the top layer with residual compressive stress after the SEVNB test. As one can see, there is a relatively straight crack path with no sharp crack deviation, deflection, or bifurcation during the crack propagation. Figure 7.10(b) shows the calculated apparent fracture toughness as a function of the crack length parameter a˜ in the Si3N4/Si3N4–30wt%TiN laminate with a residual tensile stress in the outer layers. The toughness decreases from 3.9 to 0.8 MPa m1/2 within the first Si3N4–30wt%TiN layer as the crack reaches the first interface. Toughness increases from 0.8 to 6.4 MPa m1/2 in the second Si3N4 layer with a residual compressive stress, and it decreases again from 6.4 to 1 MPa m1/2 within the third Si3N4–30wt%TiN layer with a residual tensile stress. There is no continuous growth of the crack in this case. The crack starts to propagate and then becomes arrested; after this it continues to grow again. The crack arrest results in a ‘pop-in’ event on the load–displacement diagram (Fig. 7.11). A stress of such ‘pop-in’ event is the onset stress of crack propagation. This stress, as well as an initial notch length, was used to calculate the measured apparent fracture toughness. Experimentally measured values of Ka fit well with the calculated numbers. The experimental data can be considered in two different sets. The first set includes Ka measured with notch tips within the first Si3N4–30wt%TiN and the second Si3N4 layers. Failure of all samples from the first set occurred at 116 ± 2 MPa. The second set includes two Ka values measured with notch tips within the third Si3N4–30wt%TiN layer. Failure of these two samples occurred at 71 ± 1 MPa. The insert in Fig. 7.10(b) shows an optical micrograph of two parts of the Si3N4/Si3N4–30wt%TiN laminate sample with the Vnotch placed in the Si3N4–30wt%TiN top layer with a residual tensile stress 200 180 160
Final load
Load (N)
140 120 100 80 Pop-in-load
60 40 20 0 0
10
20
30
40 50 Deflection (µm)
60
70
80
7.11 Load–displacement diagram of single-edge V-notch beam sample with pop-in.
© Woodhead Publishing Limited, 2006
90
Design of tough ceramic laminates by residual stresses control
197
after the SEVNB test. As one can see from the optical image, the crack path deviates strongly from a straight line with 90o crack deflection occurring in the center of each Si3N4 layer with a residual compressive stress. While traveling only a short distance of about one Si3N4–30wt%TiN layer thickness along a centerline, the crack kinks out into the Si3N4–30wt%TiN layer with a residual tensile stress. The calculations indicate an unambiguous trend for the apparent fracture toughness behavior. The value of Ka increases in the layers with residual compressive stress and decreases in the layers with residual tensile stress as a function of the crack length (or the crack length parameter). The calculated increase of Ka is confirmed by the experimental data in the laminates with the compressive outer layer (Fig. 7.10(a)). As one can see from Fig. 7.12(a), cracks with crack length parameter from 0 to point A1 will demonstrate unstable crack growth. In this case, once the crack starts to propagate at a certain stress, it cannot be arrested; this results in complete failure of the sample, since the applied stress intensity factor is always higher than the fracture resistance of the laminate. Cracks with crack length parameter between A1 and A3 propagate in two stages. For example, a crack with crack length parameter at A2 will have an unstable growth from point B2 to point C on the Kapp– a˜ plot (Fig. 7.12(a)). Stable growth of this crack will occur from point C to point D. For all cracks with crack length parameter a˜ from A1 to A4, failure occurs at a stress equal to the slope of the straight line OD, which is a threshold stress. The threshold stress σthr is determined by the maximum value of Ka at the interface between the first (compressive) and the second Kapp D
Km
Stress intensity factor at constant applied stress
Threshold stress σthr
Kapp
σ0B1
D
Km Stress intensity factor at constant applied stress
B′3
Kc C B1
C
B′
B2
B1
B3
B2
σ0B B
Kc Compressive Tensile layer layer
B3
Tensile Compressive layer layer
Compressive layer
Tensile layer
0 Threshold stress
A1A2 A3
A4
(a)
ã
0
A1 A2 A
ã
A3
(b)
7.12 Conditions for stable/unstable crack growth in a layered structure: (a) a range of crack length parameters for stable crack growth in a laminate with a residual compressive stress in a top layer; (b) stable/unstable crack growth in a laminate with a residual tensile stress in a top layer.
© Woodhead Publishing Limited, 2006
198
Ceramic matrix composites
(tensile) layers, and no failure can occur below σthr if the sample contains the surface cracks located only in the first layer. The curvature of the Ka plot is a function of a value of the residual stress. The higher the residual stress, the more concave the curvature of Ka. At a certain small value of residual compressive stress, the line OD can have only one intersection point with the Ka plot, and therefore no stable crack growth stage can occur. The conditions for stable/unstable crack growth in the laminate with residual tensile stress in the top layer are shown in Fig. 7.12(b). Cracks with crack length parameter A1 for such laminates will propagate only unstably at stress levels above σOB1. Cracks with crack length parameter A grow unstably at stress σOB. This unstable growth occurs between points B and C (Fig. 7.12(b)), because the points belonging to the BC segment lie above the Ka plot. At point C, the condition of equation (7.8) is violated and the crack growth becomes stable between points C and D, which means that any crack advancement requires an increase of the applied stress. Point D is a maximum value of Ka at the interface between the second (compressive) and the third (tensile) layers. This point determines a stress σ0D = σthr. Above σ0D, the crack propagates unstably until complete failure. In such a way all initial cracks in the first (tensile) and the second (compressive) layers with a crack length parameter greater than A2 (Fig. 7.12(b)) will initiate specimen failure at the same σ0D = σthr stress value. The initial cracks with tips in the third and fourth layers will initiate specimen failure at the different stress value that is determined by the maximum value of Ka at the interface between the fourth and fifth layers. This stress is σthr for cracks with tips located in the third and fourth layers. It should be noted that points B and B3 in Fig. 7.12(b) correspond to the measured Ka values (using ‘pop-in’ stress), while points B′ and B ′3 or belonging to the 0D straight line are determined by the initial notch length and the failure stress of the sample. As implied by the above analysis, the surface cracks which have sufficient length to fall into the region of stable crack growth will all cause failure at the same σthr stress. At the same time, if a residual compressive stress in the top layer is not high enough, the small cracks can cause catastrophic failure once they start to grow. Therefore, it might be that different mechanisms such as crack bridging or transformation toughening can be more effective in preventing small cracks from growing unstably. As a result of this part of the work, the apparent fracture toughness as a function of the crack length parameter a˜ = Y (α ) a 1/2 has been calculated for the Si3N4/Si3N4–30wt%TiN laminates with residual compressive or tensile stresses in the top layers. The toughness increases in the layers with compressive stress as the crack length increases, and it decreases in the layers with tensile stress as the crack continues to grow. The experimentally measured Ka values for the laminates show an excellent fit with the calculated values. It was found that a threshold stress exists for cracks of a certain length. Stable crack
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
199
growth occurs for the majority of cracks with a threshold stress as indicated by the K a – a˜ graph. Short cracks will propagate unstably because the applied stress intensity factor is always higher than the fracture resistance of the laminate. If the residual compressive stress is small enough, a situation can occur where no stable crack growth exists for cracks with tips located within the first compressive layer. Therefore, it is important to introduce high residual compressive stresses that will provide a steep slope of the apparent fracture toughness curve to include short cracks in the region of stable crack growth. Obtaining a high residual compressive stress in the first layer is an effective way of providing high toughness at small crack lengths, thereby ensuring improved flaw tolerance and surface damage resistance.
7.4.3
Fracture surfaces after fracture toughness tests
The typical fracture surfaces of the pure Si3N4 layer and the Si3N4–20wt%TiN layer are shown in Fig. 7.13. The bimodal grain size distribution exists with a number of elongated grains being surrounded by small rounded grains of Si3N4. The average grain size in the Si3N4 layer was 0.4–0.5 µm. The micrograph of the Si3N4–20wt%TiN fracture surface revealed that a majority of the grain sizes were in the range of 1–2 µm, with some grains of size less than 1 µm. It was shown that the TiN has a homogeneous distribution in the Si3N4 matrix and no solid solution was detected between Si3N4 and TiN particles [60]. Fracture surfaces of Si3N4/Si3N4, Si3N4/Si3N4–20wt%TiN, Si3N4/2(Si3N4– 20wt%TiN) and Si3N4/TiN laminates are shown in Fig. 7.14. The fracture surface of the Si3N4/Si3N4 laminate, where no residual stresses were generated during cooling, is flat and smooth (Fig. 7.14(a)). As layers of a different
(a)
(b)
7.13 Micrographs of fracture surfaces of (a) Si3N4 layer, and (b) Si3N4–20wt%TiN layer, in Si3N4/Si3N4–20wt%TiN laminate.
© Woodhead Publishing Limited, 2006
200
Ceramic matrix composites
(a)
(b)
(c)
(d)
7.14 Fracture surface of laminate composite: (a) Si3N4/Si3N4 laminates; (b) Si3N4/Si3N4–20wt%TiN laminates; (c) Si3N4/2(Si3N4– 20wt% TiN) laminates and (d) Si3N4/TiN laminates.
composition are used, the fracture surface becomes rougher. For the Si3N4/ Si3N4–20wt%TiN laminate, there are two zones on the fracture surface. The first zone near the notch tip has a rough surface and corresponds to a slow crack growth. The second zone has a rather smooth surface with distinct steps only at the interfaces between layers. This zone corresponds to a fast crack growth (Fig. 7.14(b)). No crack bifurcation occurred and two equal parts of the sample could be found after failure. The Si3N4/2(Si3N4–20wt%TiN) laminates failed after crack bifurcation. The part of the fracture surface near the notch tip was the same as those shown in Fig. 7.13. At the moment when the crack bifurcated, an unusually smooth fracture surface was observed (Fig. 7.14(c)). When the value of residual tensile stresses approaches the value of the tensile strength of the layer, cracks in the layers are generated, as was the case in the Si3N4/Si3N4–50wt%TiN and Si3N4/TiN laminates. The
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
201
cracks originated during the cooling stage after the hot pressing of the laminates and appeared due to the large mismatch of CTEs and elastic moduli. Channel cracks were observed in the laminates with a difference in composition between layers, starting with 50 wt% TiN content and higher. Si3N4/TiN laminates demonstrate channel cracking (Fig. 7.14(d)) similar to the cracks described in Ref. [24]. These cracks are responsible for the dramatic decrease in the mechanical properties of Si3N4 based laminates. To reduce or eliminate cracking, it is necessary to make composites with more similar characteristics between the layers, especially the CTE and elastic moduli. The extent of channel cracking was decreased in laminates with Si3N4–50wt%TiN layers in comparison to composites where one of the layers was pure TiN. Channel cracking was fully eliminated for composites with a Si3N4–20wt%TiN layer composition. An absence of pre-existing cracks resulted in an increase of the strength and fracture toughness. The fracture surface of Si3N4/Si3N4–50wt% TiN is shown in Fig. 7.15. As one can see, there is a high roughness of the surface, and bifurcation of the moving crack occurred when it was inside the Si3N4 layer with residual compressive stresses. There are fracture steps and channel cracks at the Si3N4–50wt%TiN layers which are perpendicular to the interfaces of the composite. The fracture steps appeared only at layers with tensile stresses. Such fracture steps and other defects are responsible for a decrease in mechanical properties. Multiple bifurcations occur for pre-existing cracks inside the layers with residual compressive stresses, and in addition the moving crack bifurcates during sample loading.
7.5
B4C based laminates
7.5.1
Design and mechanical behavior
The material systems selected for the proposed study were B4C and B4C– 30wt%SiC because of their promise for ballistic applications [61–63]. A symmetric three-layered composite was considered for the design and manufacture of armor tiles as shown in Fig. 7.16 [64]. Table 7.4 shows the relevant material constants entering the design (compiled from the literature), and Table 7.5 shows the corresponding calculated residual stresses in the three layered B4C/B4C–30wt%SiC laminate. The maximum possible apparent fracture toughness for corresponding layered structures is also presented. The layers under tensile stress have higher CTE, and in this case they are B4C layers. The layers under compressive stress have lower CTE; here they are B4C–30wt%SiC layers. A temperature T = 2150°C was used for the majority of the calculations, when residual stresses appeared in the layers upon cooling from the hot pressing temperature. There is no liquid phase present during the sintering of B4C/B4C–SiC ceramics [65], therefore the hot
© Woodhead Publishing Limited, 2006
202
Ceramic matrix composites
Step of fracture
Crack bifurcation
Si3N4–50wt% TiN layer Si3N4 layer 1 mm
1 mm (b)
(a)
100 µm
100 µm (c)
(d)
7.15 Fracture surface of Si3N4/Si3N4–50%wt%TiN composite: (a) and (c) SEI image; (b) and (d) backscattered image.
pressing temperature was used as the ‘joining’ temperature ∆T for calculations. All laminates were designed in such a way that the tensile stresses had been maintained at low values. It should be noted that the value of KIC given in Table 7.2 is a theoretical maximal apparent fracture toughness that was used to estimate the maximum possible toughening of the three-layered laminate. The experimental values measured using the single-edge V–notch beam (SEVNB) method yielded 7.42 ± 0.82 MPa m1/2 [66], which is still a very high value for brittle boron carbide based composites. Thus, the proposed approach allows a significant increase of the apparent fracture toughness values.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
203
Compressive layer B4C + 30wt% SiC
Tensile layer B4 C B4C + 30wt% SiC
7.16 Schematic presentation of symmetric three-layered design of B4C/B4C30wt%SiC laminate. Table 7.4 Properties of ceramics used in the stress calculation Composition
E (GPa)
Poisson’s ratio
CTE (10–6 K–1)
B 4C SiC
483 411
0.17 0.16
5.5 3
Table 7.5 Three-layered composite design; total thickness of a tile 10.5 mm
7.5.2
Thickness of layers (µm) B4C–30wt%SiC B 4C
σcomp(MPa)
σtens(MPa)
Apparent KIC (MPa.m1/2)
900
632
131
44
8700
Preliminary ballistic test
The manufactured 90 mm × 90 mm × 10 mm three-layered B4C/B4C– 30wt%SiC tiles were tested as armor [67]. The photographs of the experiment set-up of the ballistic test as well as a residual impression in the clay box that was used as one of the criteria in the ballistic performance of laminates are shown in Fig. 7.17. The ballistic penetration tests were performed to evaluate the ballistic performance of the laminates. Depth of penetration tests were used to evaluate the ballistic performance of the composite laminates. In addition, pure B4C monolithic ceramics were used as a standard for the test. Test panels were made using the three-layered B4C/B4C–SiC laminate and B4C monolithic ceramic material as the hard face. While the B4C monolithic tile had 100% of its theoretical density, the three-layered B4C/B4C–30wt%SiC laminates had about 3–4% of porosity. A commonly used Spectra fiberreinforced polymer composite was used as backing plates. The targets were mounted on clay and the projectile was shot at the target at a specific velocity.
© Woodhead Publishing Limited, 2006
204
Ceramic matrix composites
(a)
(b)
7.17 (a) A ballistic test setup; (b) diameter of the residual impression in the clay box used for evaluation of the performance of the B4C based tiles.
The design of the test panels was selected to ensure defeat of the threat. The depth of penetration of the projectile into the backing was measured by peeling of the unpenetrated layers of the backing plate, and the diameter of the impression on the clay after the projectile had been shot was used to evaluate the ballistic performance of the laminate composites. The results of the ballistic performance evaluation are shown in Fig. 7.18. As one can see there was no significant difference in penetration of the projectile into pure B4C monolith ceramics and three-layered composite.
7.5.3
Undesirable influence of tensile residual stresses on a laminate
During the assembly of one 100 mm × 100 mm × 12 mm multilayered tile, the inner thin B4C–30wt%SiC layers were mistakenly replaced with pure B4C thin layers [57]. As a result, instead of a multilayered tile, a threelayered laminate was produced. The parameters of this three-layered tile,
© Woodhead Publishing Limited, 2006
Fraction backing penetrated
Design of tough ceramic laminates by residual stresses control
205
0.8 0.7 0.6 0.5 0.4 0.3 0.2 B4C monolith 3-layered composite
0.1 0 1.02
1.04 1.06 1.08 1.1 1.12 Normalized projectile velocity (a)
1.14
Clay deformation (mm)
48 47 46 45 44 43 42 41 40 1.02
1.04 1.06 1.08 1.1 1.12 Normalized projectile velocity
1.14
(b)
7.18 Ballistic performance results: (a) fraction backing penetrated; (b) clay deformation.
including thickness of layers and calculated stresses, are presented in Table 7.6. The outer B4C–30wt%SiC layers had a thickness of 1650 µm, and the thick B4C layer had a thickness of 9000 µm. For such a design the level of residual tensile stress was raised to 210 MPa after cooling from THP = 2200oC. Such a high residual tensile stress leads to complete fracture of the tile during decompression of the graphite die to separate the tile after hot pressing (Fig. 7.19). The failure apparently started from the tile edges with cracks propagated further into the tile body. Table 7.6 Three-layered composite design; a total thickness of a tile 12.3 mm Composition
B4C–30wt%SiC/B4C
© Woodhead Publishing Limited, 2006
Thickness of layers (µm) B4C–30wt%SiC
B 4C
1650
9000
σcomp (MPa)
σtens (MPa)
573
210
206
Ceramic matrix composites
7.19 Photograph of fractured three-layered B4C–30wt%SiC/B4C tile hot pressed at 2200oC.
This example shows how important it is to determine a critical value of the tensile stress in a layer. Certain difficulties exist in finding this critical value. One of the problems is that the mechanical properties of an individual layer can significantly deviate from those of a corresponding bulk material. We can easily calculate the critical tensile stress if the intrinsic fracture toughness of a layer and the size of the critical flaw inside the layer are known, but the critical defect in the layer usually cannot be identified. It is possible to determine the stress for crack tunneling in the tensile layer [68]. Such stress depends only on the intrinsic fracture toughness and the layer thickness. Such transverse cracking of a tensile layer is not possible if the tensile residual stress has a lower value than the stress for crack tunneling. Therefore, an empirical value is used as a critical tensile stress. Such an approach, in fact, is also rather successful in eliminating cracking in laminates.
7.5.4
Microstructures of three-layered B4C/B4C–SiC laminates
The microstructure of a pure B4C layer of three-layered B4C/B4C–30wt%SiC laminate with 4% porosity is presented in Fig. 7.20. The three-layered B4C/ B4C–30wt%SiC tiles tested as armor material had the same microstructure and porosity level as the material shown in Fig. 7.20. As one can see, the porosity at the grain boundary of the ceramics might be a reason why threelayered laminates have not outperformed the dense monolithic boron carbide tiles. A different set of ballistic experiments are required in which fully
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
207
7.20 Microstructure of pure B4C layer in three-layered B4C/B4C– 30wt%B4C composite.
dense boron carbide based laminates will be used for comparison. Such experiments will be performed in the future. A fracture surface of a three-layer tile hot pressed at 2200oC for 1 hour and subsequently broken after hot pressing (shown in Fig. 7.19) is shown in Fig. 7.21. The layered composite demonstrates typical brittle fracture. The interface between the B4C–30wt%SiC outer layer and the pure B4C inner layer is shown in Fig. 7.21(a). The fracture surface of the B4C layer is presented in Fig. 7.21(b). Figure 7.21(c) shows the fracture surface of the B4C–30wt%SiC layer. The cleavage steps on the B4C fracture surface are presented in Fig. 7.21(d). As one can see from Fig. 7.21, the B4C grain size in B4C–30wt%SiC layers was in the range of 4–6 µm and the SiC grain size was in the range of 2–5 µm. The B4C grain size in pure B4C layers could not be calculated because of a pure transgranular fracture mode with no grains or grain boundaries revealed after fracture. Significant grain growth of boron carbide is expected during hot pressing at 2200oC. However, in B4C–30wt%SiC layers, the existence of the SiC phase prevented the exaggerated grain growth and the grain size distribution was homogeneous. Tiles hot pressed at 2200oC for 1 hour were fully dense. Tiles hot pressed at 2150oC for 30 or 45 minutes contained some amount of porosity (2–5%) that was concentrated along the interfaces and mostly in pure B4C layers. Such porosity could be detrimental to material hardness, affecting Young’s modulus and density, thus significantly lowering the ballistic performance of the laminates. As a result of the hardness and Young’s modulus decrease, material with a residual porosity of more than 2% cannot be considered as a candidate for ballistic protection. While no three-layered composite material was recoverable after the
© Woodhead Publishing Limited, 2006
208
Ceramic matrix composites
(a)
(b)
(c)
(d)
7.21 Fracture surface of a three-layered tile: (a) interface between B4C–30wt%SiC outer layer and pure B4C inner layer; (b) a fracture surface of B4C layer; (c) fracture surface of B4C–30wt%SiC; (d) cleavage steps on the B4C fracture surface.
penetration tests shown in Fig. 7.18, in a separate ballistic test designed specifically at low projectile velocity the debris of three-layered comminuted composite was collected to study the microstructural changes in material after ballistic impact. The size of the comminuted particles collected after impact varied from 5–10 µm to 1–2 mm. The density of this tile was very close to the theoretical density of the material, therefore we could consider that the tile was almost fully dense. The microstructure of the pure B4C layer comminuted by ballistic impact is shown in Fig. 7.22(a). The smooth, flat surface with transgranular fracture was typically observed for B4C layers with some amount of cleavage steps present in the material. Such cleavage mode plays an important role both in fracture and in the fragmentation event during ballistic impact [69]. There was always some amount of closed porosity which could not be eliminated by any special treatment such as increase of hot pressing temperature, pressure or dwell time. The microstructure of the B4C–30wt%SiC layer after impact is shown in Fig. 7.22(b). The B4C grains are still fractured almost always transgranularly, and a small amount of closed porosity was present in boron carbide grains. The fracture surfaces of all the SiC grains were heavily cleaved, with almost no grains observed without cleavage. Such ability of the material to form shear or cleavage steps
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
209
(a)
(b)
7.22 Micrograph of a B4C grain in (a) the pure boron carbide layer, and (b) the B4C–30wt%SiC layer, of the three-layered laminate. Closed porosity was present. Almost all SiC grains have been heavily cleaved.
should significantly increase the resistance to penetration of SiC ceramic composites. This is a topic for further intensive research; however, what is clear at the moment is that both B4C and SiC have distinctively different deformation modes under ballistic impact. This research [57, 64, 67] represents a first step in laminate ceramics development that should provide superior ballistic protection. Boron carbide– silicon carbide ceramics have been used in the design and manufacturing of three-layered composites with strong interfaces for enhanced fracture toughness. The model of a heterogeneous layered system was used to develop optimal design parameters. As a result, laminates with calculated high compressive residual stresses (up to 650 MPa) and low tensile residual stresses (below 150 MPa) were developed. The feasibility of manufacturing laminate composite systems with enhanced toughness by incorporation of thin layers with high compressive stresses in the ceramics was demonstrated. The results of this study are likely to find practical applications in the field of ballistic protection and mechanical behavior of advanced ceramic composites.
© Woodhead Publishing Limited, 2006
210
Ceramic matrix composites
7.6
Future trends
The most promising approach is the use of layered materials to control cracks by deflection, microcracking, or internal stresses. In order for these materials to become even more useful, their toughness must be increased in such a way that they could tolerate large flaws during loading. The material must be protected against the effect of the largest flaws. The most promising recent laminate designs with increased fracture toughness and high residual compressive stress have been developed with co-workers in the FP5 project ‘LAMINATES’ [70]. One of the designs for a B4C/B4C–20wt%SiC laminate based on these developments is shown in Fig. 7.23. The calculated apparent fracture toughness vs. crack length is also shown in Fig. 7.23. In the proposed design the layers create three effective barriers for crack propagation. The incremental increase of the apparent K1C is specifically targeted in the proposed design. With an increasing load the material resistance will grow further and further as the crack propagates and, therefore, more energy is required for laminate fracture. The apparent fracture toughness for this design has a
Crack
B4 C B4C + 20wt% SiC
Apparent fracture toughness (MPa m1/2)
B4 C
B4 C
14 12 10 8 6 4 2 0 0
1
2 Crack length (mm)
3
4
7.23 Design of B4C/B4C–20wt%SiC for increased fracture toughness.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
211
maximum calculated value of ~14 MPa m1/2 while the intrinsic fracture toughness of B4C was adopted to be only 2 MPa m1/2. The thick inner B4C layer serves to obtain a low level of tensile residual stress. The following directions of research in mechanical behavior improvement of ceramic laminates are proposed: • • • • • • • • • •
Improvement of the material’s structure in separate layers Obtaining macrocrack shielding Obtaining macrocrack deflection Development of gradient laminar structures. Opportunities for structure optimization of layered composites are: Optimization of statistical parameters of the strength distribution of composites Optimization of microcracking process in individual layers Design of layered structures with maximum apparent fracture toughness induced by crack shielding Design of layered composites with crack bifurcation in compressed layers Control of residual stress distribution in laminates Design of asymmetric laminar structures tailored for specific use.
7.7
Acknowledgements
This work was supported by the European Commission, project 1CA2-CT2000-10020 Copernicus-2 ‘Silicon nitride based laminar and functionally gradient ceramics for engineering application’, and by AFOSR, project F4962002-0340. EMPA was funded by BBW, the Swiss Federal Office for Education and Science, under contract 99.0785. This work was also partly performed at the Army Center for Nanoscience and Nanomaterials, North Carolina A&T State University.
7.8
References
1. Clegg, W.J., Kendall, K., Alford, N. McN., Button, T.W., Birchall, J.D. A simple way to make tough ceramics, Nature, 347, 455–457, 1990. 2. Clegg, W.J., Controlling cracks in ceramics, Science, 286, 1097–1099, 1999. 3. Rao, M.P., Sanchez-Herencia, A.J., Beltz, G.E., McMeeking, R.M., Lange, F.F., Laminar ceramics that exhibit a threshold strength, Science, 286, 102–105, 1999. 4. Moon, H., Pontin, M.G., Lange, F.F., Crack interactions in laminar ceramics that exhibit a threshold strength, J. Am. Ceram. Soc., 87(9), 1694–1700, 2004. 5. Chan, H., Layer ceramics: processing and mechanical behavior, Ann. Rev. Mater. Sci., 27, 249–282, 1997. 6. Honeyman-Colvin, P., Lange, F.F., Infiltration of porous alumina bodies with solution precursors: strengthening via compositional grading, grain size control, and transformation toughening, J. Am. Ceram. Soc., 79(7), 1810–1814, 1996. 7. Oechsner, M., Hillman, C., Lange, F., Crack bifurcation in laminar ceramic composites, J. Am. Ceram. Soc., 79 (7), 1834–1838, 1996.
© Woodhead Publishing Limited, 2006
212
Ceramic matrix composites
8. Sanchez-Herencia, A., Pascual, C., He, J., Lange, F., ZrO2/ZrO2 layered composites for crack bifurcation, J. Am. Ceram. Soc., 82(6), 1512–1518, 1999. 9. Folsom, C.A., Zok, F.W., Lange, F., Flexural properties of brittle multiplayer materials: I. Modeling, J. Am. Ceram. Soc., 77(3), 689–696, 1994. 10. Lugovy, M., Slyunyayev, V., Orlovskaya, N., Blugan, G., Kuebler, J., Lewis, M., Apparent fracture toughness in Si3N4-based laminates with residual compressive or tensile stresses in surface layers, Acta Materialia, 53, 289–296, 2005. 11. Hbaieb, K., McMeeking, R.M., Threshold strength predictions for laminar ceramics with cracks that grow straight, Mechanics of Materials, 34, 755–772, 2002. 12. He, M.Y., Evans, A.G., Yehle, A., Criterion for the avoidance of edge cracking in layered systems, J. Am. Ceram. Soc., 87(8), 1418–1423, 2004. 13. Lakshminarayanan, R., Shetty, D.K., Cutler, R.A., Toughening of layered ceramic composites with residual surface compression, J. Am. Ceram. Soc., 79(1), 79–87, 1996. 14. Blattner, A., Lakshminarayanan, R., Shetty, D.K., Toughening of layered ceramic composites with residual surface compression: effect of layer thickness, Eng. Fract. Mech. 68, 1–7, 2001. 15. Marshall, D.B., Ratto, J.J., Lange, F.F., Enhanced fracture toughness in layered microcomposites of Ce–ZrO2 and Al2O3, J. Am. Ceram. Soc., 74(12), 2979, 1991. 16. Yoo, J., Cho, K., Bae, W.S., Cima, M., Suresh, S. Transformation-toughened ceramic multilayers with compositional gradients, J. Am. Ceram. Soc., 81(1), 21, 1998. 17. Thompson, S.C., Pandit, A., Padture, N.P., Suresh, S. Stepwise-graded Si3N4–SiC ceramics with improved wear properties, J. Am. Ceram. Soc., 85(8) 2059, 2002. 18. Bai, N., Pollard, D.D., Gao, H., Explanation for fracture spacing in layered materials, Nature, 403, 753–756 (2000). 19. Yaroshenko, V., Orlovskaya, N., Einarsrud, M.-A., Berroth, K., Kovylayev, V., Laminar Si3N4–TiN hot pressed ceramic composites, Key Eng. Mater., 132–136(3) 2017– 2020, 1997. 20. Lugovy, M., Orlovskaya, N., Berroth, K., Kuebler, J., Macrostructural engineering of ceramic matrix layered composites, Composite Sci. Tech., 59, 1429–1437, 1999. 21. Lugovy, M., Slyunyayev, V., Subbotin, V., Orlovskaya, N., Gogotsi, G., Crack arrest in Si3N4-based layered composites with residual stress, Composites Sci. Tech. 64(13–14), 1947–1957, 2004. 22. Yaroshenko, V., Orlovskaya, N., Einarsrud, M.-A., Kovylayev, V., Processing of multilayered Si3N4–TiN hot-pressed ceramic composites, in Proceedings of NATO ARW ‘Multilayered and Fibre-reinforced Composites: Problems and Prospects’, Kluwer, Dordrecht, 1998. 23. Lugovy, M., Orlovskaya, N., Berroth, K., Kuebler, J., Analysis of layered composite with crack deflection controlled by layer thickness, in Proceedings of NATO ASI ‘Functional Gradient Materials and Surface Layers Prepared by Fine Particle Technology’, Kiev, Ukraine, 18–28 June 2000, ed. M.-I. Baraton and I. Uvarova, NATO Science Series, II, Mathematics, Physics and Chemistry, 16, Kluwer Academic Publishers, 273–280, 2001. 24. Green, D., Cai, P., Messing, G., Residual stresses in alumina–zirconia laminates, J. Eur. Ceram. Soc., 1999, 19, 2511–2517. 25. Lences, Z., Sajgalik, P., Toriyama, M., Brito, M.E., Kanzaki, S., Multifunctional Si3N4/(β-SiAlON + TiN) layered composites, J. Eur. Ceram. Soc., 20, 347–355, 2000.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
213
26. Sergo, V., Lipkin, D.M., De Portu, G., Clarke, D.R., Edge stresses in alumina/ zirconia laminate, J. Am. Ceram. Soc., 80(7), 1633–1638, 1997. 27. Ho, S., Hillman, C., Lange, F.F., Suo, Z., Surface cracking in layers under biaxial compressive stress, J. Am. Ceram. Soc., 78(9), 2353–2359, 1995. 28. Chartier, T., Merle, D., Besson, J.L., Laminar ceramic composites, J. Eur. Ceram. Soc., 16, 101–107, 1995. 29. Sglavo, V.M., Larentis, L., Green, D.J. Flaw-insensitive ion-exchanged glass: I, Theoretical aspects, J. Am. Ceram. Soc., 84(8), 1827–1831, 2001. 30. Orlovskaya, N., Lugovy, M., Subbotin, V., Radchenko, A., Adams, J., Chheda, M., Shih, J., Sankar, J., Yarmolenko, S., Robust design and manufacturing of ceramic laminates with controlled thermal residual stresses for enhanced toughness, J. Mater. Sci. (in press). 31. Moon, R.J., Hoffman, M., Hilden, J., Bowman, K., Trumble, K., Roedel, J. Weight function analysis on the R-curve behavior of multilayered alumina–zirconia composites, J. Am. Ceram. Soc. 85(6), 1505–1510, 2002. 32. Fett, T., Munz, D., Influence of crack-surface interactions on stress intensity factor in ceramics, J. Mater. Sci. Lett., 9, 1403–1406, 1990. 33. Kovar, D., Thouless, M.D., Halloran, J.W. Crack deflection and propagation in layered silicon nitride/boron nitride ceramics, J. Am. Ceram. Soc., 81(4), 1004– 1012, 1998. 34. Clegg, W.J., The fabrication and failure of laminar ceramic composites, Acta Metall. Mater., 40(11) 3085–3093, 1992. 35. Tomaszewski, H., Strzeszewski, J., Gebicki, W., The role of residual stresses in layered composites of Y–ZrO2 and Al2O3, J. Eur. Ceram. Soc., 19, 255–262, 1999. 36. Huang, J.-L., Chang, Y.-L., Lu, H.-H., Fabrication of multilaminated Si3N4–Si3N4/ TiN composites and its anisotropic fracture behavior, J. Mat. Res., 12(9) 2337–2344, 1997. 37. Blugan, G., Orlovskaya, N., Lewis, M., Kuebler, J., Multilayer silicon nitride laminates exhibiting high fracture toughness and crack deflection, Key Eng. Mater., 280–283, 1863–1868, 2004. 38. Blugan, G., Dobedoe, R., Gee, I., Orlovskaya, N., Kuebler, J., Failure behavior of high toughness multilayer Si3N4 and Si3N4–TiN based laminates, Key Eng. Mater., submitted, 2005. 39. Sanchez-Herencia, A.J., James, L., Lange, F., Bifurcation in alumina plates produced by a phase transformation in central, alumina/zirconia thin layers, J. Eur. Ceram. Soc., 20, 1297–1300, 2000. 40. Lugovy, M., Orlovskaya, N., Slyunayev, V., Gogotsi, G., Kuebler, J., Sanchez-Herencia, A., Crack bifurcation features in laminar specimens with fixed total thickness, Composite Sci. Tech., 62, 819–830, 2002. 41. Herrmann, M., Balzer, B., Schubert, C., Hermel, W., Densification, microstructure and properties of Si3N4–Ti(C,N) composites, J. Eur. Ceram. Soc., 12, 287–296, 1993. 42. Yaroshenko, V., Katashinsky, V., Processing and properties of silicon nitride based laminated composites, J. Can. Ceram. Soc., 66(4), 292–296, 1997. 43. Yaroshenko, V., Gogotsi, Y., Osipova, I., Effect of TiN addition on properties of silicon nitride based ceramics, in Ceramic Today – Tomorrow’s Ceramics, ed. P. Vincenzini, Elsevier, Amsterdam, 2821–2830, 1991. 44. Gogotsi, Y., Gratwhol, G., Tuemler, F., Yaroshenko, V., Oxidation of yttria- and alumina-containing dense silicon nitride ceramics, J. Eur. Ceram. Soc., 11, 375–386, 1993.
© Woodhead Publishing Limited, 2006
214
Ceramic matrix composites
45. McColm, I.J., Ceramic Hardness, Plenum Press, New York, 1990. 46. Schwetz, K.A., Lipp, A., Ulmann’s Encyclopedia of Industrial Chemistry, A4, ed. E.E. Gerhartz, VCH, Weinheim, 295, 1981. 47. Amberger, E., Stumpf, W., Buschbeck, K.-C., Handbook of Inorganic Chemistry, 8th edn, Springer-Verlag, Berlin, 1981. 48. Bylander, D.M., Kleiman, L., Structure of B13C2, Phys. Rev. B, 43, 1487, 1991. 49. Thevenot, F., Boron carbide – a comprehensive review, J. Eur. Ceram. Soc., 6(4), 205–225, 1990. 50. Champagne, B., Angers, R., Mechanical properties of hot pressed B–B4C materials, J. Am. Ceram. Soc., 62(3–4), 149–153, 1979. 51. Niihara, K., Nakahira, A., Hirai, T., The effect of stoichiometry on mechanical properties of boron carbide, J. Am. Ceram. Soc., 67, C13, 1984. 52. Thevenot, F., in Properties of Ceramics, ed. G. de With, Terpstra, R.A., Metselaar, R., Elsevier Applied Science, London and New York, 1989. 53. Lee, H., Speyer, R., Hardness and fracture toughness of pressureless sintered boron carbide (B4C), J. Am. Ceram. Soc., 85(5), 1291–1293, 2002. 54. Wilkens, M.L., Mechanics of penetration and perforation, Int. J. Eng. Sci., 16, 793– 807, 1978. 55. Hyatt, T.P., Electronics: tape casting and roll compaction, Am. Ceram. Soc. Bull., 74(10), 56–59, 1995. 56. Dec, R.T., Zavaliangos, A., Cunningham, J.C., Comparison of various modelling methods for analysis of powder compaction in roller press, Powder Technology, 130, 265–271, 2003. 57. Orlovskaya, N., Kalidindi, S., Lugovy, M., Subbotin, V., Radchenko, O., Adams, J., Chheda, M., Shih, J., Sankar, J., Yarmolenko, S., Robust design and manufacturing of ceramic laminates with controlled thermal residual stresses for enhanced toughness, J. Mat. Sci. (in press). 58. Kuebler, J., Fracture toughness using the SEVNB method: preliminary results, Ceram. Eng. Sci. Proc., 18, 155–162, 1997. 59. Kuebler, J., Fracture toughness of ceramics using the SEVNB method: from a preliminary study to a standard test method, in Fracture Resistance Testing of Monolithic and Composite Brittle Materials, ASTM STP 1409, ed. J.A. Salem, M.G. Jenkins and G.D. Quinn, ASTM, West Conshohocken, PA, pp. 93–106, January 2002. 60. Hermann, M., Shubert, C., Pabst, J., Richter, H.J., Obenau, P., Yaroshenko, V. in Symposium: Verstaekung Keramisher Werkstoffe, Hamburg, October 1991, Tangungsband, DGM. 61. Orphal, D.L., Franzen, R.R., Charters, A.C., Menna, T.L., Piekutowski, A.J., Penetration of confined boron carbide targets by tungsten long rods at impact velocities from 1.5 to 5.0 km/s, Int. J. Impact Eng., 19, 15–29, 1997. 62. Shih, C.J., Meyers, M.A., Nesterenko, V.F., Chen, S.J., Damage evolution in dynamic deformation of silicon carbide, Acta. Mater., 48, 2399–2420, 2000. 63. Orphal, D.L., Franzen, R.R., Penetration of confined silicon carbide targets by tungsten long rods at impact velocities from 1.5 to 4.6 km/s, Int. J. Impact Eng., 19, 1–13, 1997. 64. Orlovskaya, N., Lugovy, M., Subbotin, V., Radchenko, O., Adams, J., Chheda, M., Shih, J., Sankar, J., Yarmolenko, S., Design and manufacturing B4C–SiC layered ceramics for armor applications, in Ceramic Armor and Armor Systems (Ceramic Transactions, Volume 151), 2003, pp. 59–70.
© Woodhead Publishing Limited, 2006
Design of tough ceramic laminates by residual stresses control
215
65. Kislyi, P.S., Kuzenkova, M.A., Bondaruk, N.I., Grabchuk, B.L., Boron Carbide, Naukova Dumka, Kiev, 1988 (in Russian). 66. Orlovskaya, N., Lugovy, M., Kuebler, J., Mechanical performance of 3 layered B4C–SiC ceramic composites, unpublished results. 67. Orlovskaya, N., Adams, J., Chheda, M., Shih, J., Yarmolenko, S., Sankar, J., Lugovy, M., Subbotin, V. Boron carbide–silicon carbide laminate ceramics for ballistic protection, Proc. of 2003 ASME Int. Mech. Eng. Cong., Vol. 3, Paper IMECE200343323, 15–21 November 2003, Washington, DC. 68. Ho, S., Suo, Z., Tunnelling cracks in constrained layers, J. Appl. Mech., 60, 890– 894, 1993. 69. Chen, M., McCauley, J.W., Hemker, K.J., Shock-induced localized amorphization in boron carbide, Science, 299, 1563–1566, 2003. 70. Final Report 2004, EC/BBW Contract No. ICA-CT-2000-10020, FP5 INCO-Copernicus project ‘LAMINATES’ (Silicon Nitride Based Laminar and Functionally Graded Ceramic Composites for Engineering Applications), project partners: University of Warwick (UK), FCT Technologie (Germany), Institute for Problems of Materials Science (Ukraine), Materials Research Center Ltd (Ukraine), Institute for Problems of Strength (Ukraine), Institute of Chemical Physics (Armenia), Drexel University (USA), EMPA (Switzerland).
© Woodhead Publishing Limited, 2006
8 Hardness of multilayered ceramics W J C L E G G, F G I U L I A N I, Y L O N G, S J L L O Y D, University of Cambridge UK and J M M O L I N A - A L D A R E G U I A, Centro de Estudios e Investigaciones Tecnicas de Gipuzkoa (CEIT), Spain
8.1
Introduction
There is a growing demand for ultra-hard materials in applications such as high-speed cutting and forming, hard disc drives and various optical and biomechanical components. Normally properties other than hardness are also needed, such as chemical inertness, a low friction coefficient and a thermal expansivity with respect to that of the substrate so the coating does not spontaneously peel off. The resulting compromise is such that the transition metal nitrides, with hardnesses of approximately 20 GPa are often used, rather than the very hardest materials such as diamond (100 GPa) or cubic boron nitride (50 GPa). There is therefore considerable interest in developing ways of increasing the hardness of an intrinsically hard material by modifying its microstructure. The problem is not a trivial one. Most of the strengthening mechanisms that have been developed in metals give rise to an increment of strength that is independent of the strength of the base alloy. The total strength is simply the sum of the initial strength and the increment, which has a typical magnitude of a few hundred megapascals. However, in materials that are intrinsically strong, such as most ceramics, the resistance of the lattice alone to dislocation motion gives materials with strengths of tens of gigapascals. The strengthening mechanisms used in metals are therefore normally quite useless. In the late 1980s it was shown that a multilayer made of alternating layers of TiN and VN, each layer a few nanometres in thickness, could show hardnesses of over 50 GPa, twice that of the monolithic material (Helmersson et al., 1987). Since then a number of ideas have been developed to explain the strengthening that is observed, such as coherency strains or changes in dislocation line energy between the layers. However, there are such structures that on the basis of these analyses might be expected to show hardening but do not do so, suggesting either that some other effect is leading to weakening or that the hardening effect does not arise as has been suggested. The aim of this chapter, therefore, is to review what is known about how multilayer 216 © Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
217
ceramic coatings deform, why they might be hard and why this hardening is not reliably observed.
8.2
Behaviour of multilayer structures
The brittleness of the ceramic coatings has meant that their flow behaviour has been inferred from hardness tests rather than directly from tensile tests. Results from a large number of different systems are shown in Figs 8.1(a)– (c), where it can be seen that a very wide range of properties are found. In some, large increases in hardness are found (Helmersson et al., 1987; Mirkarimi et al., 1990; Shinn et al., 1992), well above those that might be expected from the properties of the individual components. In others, there is no effect (Ljungcrantz et al., 1998; Yashar and Barnett, 1999; Högberg et al., 2001; Molina-Aldareguia et al., 2002; Barnett et al., 2003). Even reductions in hardness have been observed (Jayaweera et al., 2003). Rather surprisingly the greatest increases in hardness have been observed in systems where the phases are isostructural, such as TiN and NbN, both of which have the B1 rocksalt structure: see Fig. 8.1.
Hardness (GPa)
60
TiN-VN (Helmersson et al., 1987)
50
TiN-NbN (Shinn et al., 1992)
40 NbN, L
TiN-VNbN (Mirkarimi et al., 1990)
NbN,B
NbN-VNbN (Shinn & Barnett, 1994)
30 TiC
TiN-NbN B-series (Ljungcrantz et al., 1998) TiN-NbN L-series (Ljungcrantz et al., 1998)
TiN
20 NbN VC
10
TiN-NbN (MolinaAldaregnia et al., 2002) 0 0
10
20 30 Wavelength (nm)
40
TiC-VC (Högberg et al., 2001)
(a)
8.1 The variation of hardness with multilayer wavelength in a range of different types of structures. These include multilayers of (a) isostructural transition metal nitrides and carbides, which show the greatest hardening; (b) nonisostructural multilayer materials, where slip cannot occur by the movement of dislocations across the planes of the composition modulation, because the slip systems are different in the two materials; and (c) materials where different crystal structures are stabilized at small layer thicknesses, such as AlN deposited onto TiN.
© Woodhead Publishing Limited, 2006
218
Ceramic matrix composites 60 Mo-NbN (Madan et al., 1998) Y2O3-ZrO2 (Yashar & Barnett 1999) TiN-TiB2 (Barnett et al., 2003)
Hardness (GPa)
50
TiB2
40
30 TiN 20
NbN YSZ
10
Y2 O3 Mo
0
0
10
20 Wavelength (nm)
30
40
(b) 40 Setoyama et al., 1996 Wong et al., 2000 Li et al., 2004
Hardness (GPa)
30
20
10
0
0
Cubic AlN fully stabilized by TiN
10
20 30 Wavelength (nm)
40
50
(c)
8.1 Continued
The use of indentation complicates the interpretation of the measurements. However, relationships between hardness and flow stress exist for monolithic materials and these have been used to obtain information in the multilayers. When an indenter is pressed into the surface of a material, the material that is displaced must be accommodated either by material being pushed out of
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
219
the surface (pile-up) (Tabor, 1951) or by moving radially outward from the indentation so that it is accommodated elastically within the body (Marsh, 1963). The former tends to occur in metals whilst the latter occurs in materials with a high ratio of the uniaxial flow stress, σf, to the Young modulus, E. These include most ceramics, which are the subject of this review. Numerical and analytical solutions indicate that for a soft metal the hardness, H, should be about three times the uniaxial flow stress, σf, falling gradually to about 1 as the ratio of H/E increases (Marsh, 1963; Johnson, 1970; Cheng and Cheng, 2000). This radial flow requires the movement of atoms across the planes of the composition modulation. Where the two materials have very similar crystal structures, as for instance in strained layer superlattices, where the interfaces between the layers are coherent, dislocations may be able to simply glide across the layers. In this case the dislocation must move under conditions where the coherency or elastic misfit strains between the layers alternate, giving rise to stresses that act on the dislocation. However, the changes in elastic modulus and Burgers’ vector cause changes in the dislocation line energy as the dislocation moves and hence to a force on the dislocation arising from the composition modulation, giving a possible source of hardening. However, in the more general case where the layers are not isostructural and so do not have common slip systems, slip will have to be renucleated in each layer as envisaged by Hall and Petch, even though the layers may be coherent. Alternatively, where flow across the layers becomes difficult, then the deformation required to accommodate the indentation might occur by the lateral movement of material within the individual layers. To investigate the behaviour of ceramic multilayers we therefore need to examine the various mechanisms and understand under what conditions of structure and applied indenter pressure such a mechanism might provide the dominant obstacle to deformation.
8.3
Hardening mechanisms in multilayers
8.3.1
Hardening due to coherency stresses
The initial observations of hardening in ceramic multilayers were made in strained layer superlattices of TiN/VN (Helmersson et al., 1987). At multilayer wavelengths of approximately 10 nm, the hardness was greater than 50 GPa, much greater than the 20–25 GPa normally reported for thin films of monolithic TiN. Explanations focused on the interaction of the elastic stress field of the dislocation with the alternating stress field of the superlattice, which gives rise to shear stresses that are a maximum on planes running through the thickness of the layers in both in-plane directions at an angle of 45° (Cahn, 1963). These shear stresses alternate so that one layer will aid the passage of
© Woodhead Publishing Limited, 2006
220
Ceramic matrix composites
a given dislocation, whilst the next will impede it. The effect of such alternating stress fields on the movement of a single dislocation has been analysed by Cahn (1963) and later modified by Kato et al. (1980). They analysed the situation of three orthogonal waves of a sinusoidal composition modulation, as might occur in a face centred cubic alloy undergoing spinodal decomposition. They found that the increment of the shear flow stress, ∆τc, due to the varying internal stress field could be given approximately by ∆τc = 0.14AEη
(8.1)
where A is the magnitude of the composition amplitude, E is the Young modulus and η is the misfit strain, assuming that the lattice parameter varies linearly with changing composition. Setting A to 1, that is the composition changes from pure TiN to pure VN, taking values of the lattice parameter of TiN and VN to be respectively 0.424 nm and 0.414 nm (see Table 8.1) and using the Young modulus of TiN as 450 GPa, gives an increment in the shear flow stress of approximately 1.5 GPa, suggesting an increase in the hardness of about 9 GPa, much smaller than the increment observed in Fig. 8.1(a). However, the analysis is for a composition modulation in three orthogonal directions rather than the one direction here. The simple geometry in the multilayers gives rise to a further complication. For a multilayer made of two materials with different lattice parameters, the atom spacings parallel to the planes of composition modulation will be equal if the layers are coherent. However, normal to the planes, the atom spacings will be unchanged, apart from a Poisson contraction that will act to increase the difference, so that the Burgers’ vectors in the two materials will most likely be different. Kelly (1991) has pointed out that when a dislocation moves from one phase, B, to another, A, a dislocation with a Burgers’ vector given by the difference at the Burgers’ vectors between the two phases will be left behind in the interface. This will give rise to a repulsive force on the glide dislocation when it is moving in the layer with the smaller Burgers’ vector, as the two dislocations will have the same sign. However, when the glide dislocation is moving through the layer with the larger Burgers’ vector and the interface dislocation has an opposite sign, the force between the two dislocations will be attractive, inhibiting the movement of the glide dislocation. The force between the two dislocations can be estimated using the standard Table 8.1 Single crystal elastic constants and lattice parameters of TiN, NbN and VN. The data are taken from Kim et al. (1992) Material
c11 (GPa)
c12 (GPa)
c44 (GPa)
a (nm)
TiN NbN VN
625 556 533
165 152 135
163 125 133
0.424 0.439 0.414
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
221
expressions from dislocation theory; thus it is predicted to increase as the distance between the dislocations, r, is diminished, and to rise without limit as r tends to zero. To avoid this difficulty, Kelly follows Koehler (1970) and assumes that the force will be at a maximum when r = 2b, where b is the Burgers’ vector. Substituting for r gives the maximum force on the dislocation, and as F = τb, the increase in the shear flow stress, ∆τm, required to move the dislocation a long distance through the superlattice against the forces due to these misfit dislocations is given by ∆τ m =
GA ∆b ⋅ 2π b A
(8.2)
where GA is the shear modulus of layer A, bA its Burgers’ vector and ∆b the difference between the Burgers’ vectors. Using the values of the lattice parameters of TiN and VN and the shear modulus of TiN, as before, gives an increase in the shear flow stress of 0.9 GPa, and a corresponding increment in the hardness of approximately 4.5 GPa, which again is much less than the hardening observed: see Fig. 8.1(a). One difficulty in interpreting the results in terms of coherency strains is that as the layer thickness increases, coherency is lost and an array of misfit dislocations is formed. This phenomenon has been studied in great detail in semiconductors, where retaining coherency is of considerable importance. Dunstan and co-workers have studied the effects of coherency stresses using strained layer superlattices of InGaAs where the differences in lattice parameter between the layers are introduced by doping layers with varying amounts of In (Jayaweera et al., 2003). Looking at layers with thicknesses ranging from 17 to 125 nm and strain modulations varying from 0 to 1.46%, they found that the introduction of coherency stresses weakened rather than strengthened the material. Furthermore it appeared that the magnitude of the strength reduction was not dependent on either the strain modulation or the extra strain energy associated with the strained layers. Rather the hardness was given by an expression of the form H Y = H YO – kMF ′
(8.3) H YO
where HY is the measured hardness of the superlattice, is the hardness of the monolithic material, k is a numerical constant, M is a biaxial elastic modulus and F′ is a thickness-averaged strain modulation given by the expression F′ =
| ε c hc + ε t h t | hc + h t
(8.4)
where εc and εt are the strains in the compressive and tensile layers respectively of thicknesses hc and ht. By considering the rate of change of elastic energy with respect to an unspecified variable, it is shown how an expression of this
© Woodhead Publishing Limited, 2006
222
Ceramic matrix composites
form might be obtained. The formulation in terms of a strain energy gradient involving both layers is taken to imply that the flow process must require some minimum volume to operate, although it is not clear why such effects do not also appear in the other experiments, or even in spinodal structures where the microstructural features are even smaller. It is tentatively suggested that this unknown variable may be the rate of plastic work, although it is not clear how a yield criterion can be developed by considering the rate of change of a variable with respect to the rate of plastic work. Despite this, as the authors point out, the formal derivation is correct regardless of the physical meaning of the unknown variable. However, the derivation does predict that the sign of F ′ is important, despite equation (8.4). Unfortunately there are no experimental results that might resolve this. In summary it is clear that the effects of coherency stresses are very far from understood, although experiments suggest that in intrinsically strong materials their effect is relatively small. This has led to the consideration of other possibilities, in particular the effects of elastic inhomogeneity.
8.3.2
Hardening due to changes in dislocation line energy
The importance of such effects was deduced from the data shown in Fig. 8.2 from the elegant experiments of Barnett and co-workers (Chu and Barnett, 1995). In the TiN/VN multilayers the two layers are both elastically strained with respect to one another and have different elastic moduli, that of TiN being greater than that of VN, which is similar to that of NbN. Shinn and Barnett (1994) have used this to study the effects of elastic modulus mismatch. As shown in Fig. 8.2, systems where there was a difference in elastic modulus showed a substantial increase in hardness. Where there was no difference in elastic modulus little or no hardening was observed, whilst hardening was obtained in a TiN/V0.6Nb0.4N system where there was a modulus mismatch but no lattice mismatch (Mirkarimi et al., 1990; Hubbard et al., 1992). These observations have a theoretical foundation in Cahn’s original analysis of hardening in spinodal structures (Cahn, 1963). A dislocation has a line energy, U, associated with its elastic misfit in the lattice whose magnitude is approximately equal to the product of the shear modulus, G, and the square of the Burgers’ vector. If the body containing the dislocation is not uniform but is instead made up of layers of different materials whose values of G and b give different values of U, there will be a force on the dislocation acting to move it from the layer with the higher value of U to that with the lower value. The situation of a screw dislocation moving towards an atomically sharp interface separating two isotropic layers, A and B, with shear moduli GA and GB respectively but the same Burgers’ vector, was analysed by Koehler
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics 60
TiN-VN (Helmersson et al., 1987)
50 Source activation
Hardness (GPa)
223
TiN-NbN (Shinn et al., 1992) TiN-VNbN (Mirkarimi et al., 1990) NbN-VNbN (Shin and Barnett, 1994)
40
30
Loop motion
20
10
0 0
10
20 30 Wavelength (nm)
40
8.2 The variation in hardness theoretically predicted for the TiN/NbN multilayers compared with data for the isostructural nitrides shown in Fig. 8.1(a). The predictions are based on the ideas of an increment of hardness arising from the elastic mismatch across the layers, equation (8.6), shown as the horizontal dashed line, and for the lateral flow of material within the interlayers, equations (8.10–8.12), shown as the dashed line increasing as the wavelength decreases. In the isostructural multilayers, the upper limit to the increase in hardness should occur when the stress is high enough to drive dislocation motion across the layers. In the nonisostructural, this condition does not apply, or is substantially modified. The lines shown assume that the hardness of monolithic TiN and NbN is 25 GPa and that the layer thicknesses are equal. Other data are given in Table 8.1.
(1970) using the image force approach (Benlahsen et al., 1993), thereby implicitly assuming that the Burgers’ vectors are the same in each material, or at least that the effects of differences in Burgers’ vector are unimportant. Assuming linear elasticity, Koehler showed that the magnitude of this force depended on the distance, r, between the dislocation and the interface between the two layers and the elastic mismatch across the interface, Q, given by
Q=
GA – GB GA + GB
(8.5)
As might be expected, the magnitude of the force on the dislocation increases as the dislocation gets closer to the interface. The force is repulsive when the dislocation is moving towards the stiffer layer, and attractive when it is moving towards the more compliant layer. The major difficulty with the Koehler solution is that the force on the dislocation is predicted to rise without limit as the dislocation reaches the
© Woodhead Publishing Limited, 2006
224
Ceramic matrix composites
interface and r = 0. To obtain a solution, Koehler assumed that the magnitude of the force on the dislocation due to the interface would reach a maximum when r = 2b. However, the magnitude of the force has a maximum value when, for a given increment of movement towards an interface, there is the greatest change in the elastic strain energy due to the dislocation misfit, that is when the dislocation is crossing the interface and r = 0. A solution to this problem requires a knowledge of the atom positions in the dislocation core. Using the atom positions in the Peierls dislocation, Pacheco and Mura (1969) estimated the force on a dislocation due to just a single interface. They obtained the increase in shear flow stress, ∆τE, due to an elastic modulus change across a sharp interface as
∆τ E = 22 Q sin θ GB π
(8.6)
where θ is the angle of the slip plane with the interface and the dislocation is in material B and is greater than that predicted by Koehler by a factor of 16π. An estimate of the magnitude of this effect was obtained by Shinn and Barnett, setting the shear modulus equal to the single crystal elastic constant c44 (Shinn and Barnett, 1994; Chu and Barnett, 1995). These are 163 and 125 GPa for TiN and NbN respectively: see Table 8.1. This gives a value of Q of 0.14 and a value of ∆τE of 5.3 GPa. Taking the hardness, H, to be three times the uniaxial flow stress and this to be twice the shear flow stress, and following Chu and Barnett (1995), the increment in the hardness, ∆H, due to the elastic mismatch of the layers is predicted to be 48 GPa. This assumes that the composition varies from pure TiN to pure NbN. However, X-ray diffraction showed that the variation was only one-half of this (Shinn and Barnett, 1994), so that the predicted value of ∆H was approximately 24 GPa and in very reasonable agreement with the data. However, in a cubic structure the value of G will be equal to c44 only when slip is on the {110}<001> slip system (Kelly et al., 2000). In rocksaltstructured nitrides and carbides, slip in indentation at room temperature occurs on the {110} <1 1 0> slip system (Williams and Schaal, 1962; MolinaAldareguia et al., 2002). The appropriate value of G is related to the different single crystal elastic constants, cij, by G = 1 ( c11 – c12 ) 2
(8.7)
Substituting the relevant values for TiN and NbN from Table 8.1 gives a value for Q of only 0.06, one-half of that obtained above. Accounting for the observed concentration modulation, a total increment in hardness of only 12 GPa is predicted. If the uniaxial flow stress is taken as twice the shear flow stress, the predicted value of ∆H falls to only 9 GPa, much less than that observed here.
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
225
At higher temperatures flow occurs on the {111} <1 1 0> slip system (Williams and Schaal, 1962), where G is given by (Kelly et al., 2000) G=
3c 44 ( c11 – c12 ) 4c 44 + c11 – c12
(8.8)
However, even here Q is only 0.09, giving a ∆H of approximately 14 GPa, rather than the 24 GPa given by Shinn and Barnett (1994). Only one interface has been considered. In the multilayer, there will be forces due to each of the interfaces, each decreasing as one moves further away from the dislocation. When the layers are thick this effect is negligible, so the increase in stress required to overcome the elastic mismatch across the interface will be given by equation (8.6). However, when the layers are very thin, the force due to the elastic mismatch will be reduced. As an approximation, Lehoczky considered the force on the dislocation due to just the two interfaces lying on either side of the dislocation (Lehoczky, 1978b). Adding more interfaces decreases this but, as they are further away than the two nearest interfaces to the dislocation, their effect is small and a solution including many interfaces is within 5% of that for just two interfaces. However, the effect becomes important, at least for the systems here, only when the multilayer wavelength falls below about 5 nm. There are therefore two effects that cause a reduction in the hardness of superlattices with very fine layers. The first is due to the inevitable intermixing that occurs at the interface, an effect that becomes more marked as the layers become thinner (Shinn and Barnett, 1994). The second is due to the presence of other interfaces. Both would act to reduce the hardness. These analyses describe the situation where flow occurs by the motion of single dislocations passing across the interfaces between the two materials. However, if the layers are very thick, dislocation sources within the individual layers will be able to operate. The shear stress required to operate a source in the more compliant layer, B, is given by τS ≈
GB bB lB
(8.9)
where lB is the dimension of the source. If this stress is less than the repulsive force acting on it due to the presence of the high modulus layer, then the dislocations will pile up at the interface, giving rise to a stress concentration at the head of the pile-up, τpu, which will help force the dislocation through the interface (Lehoczky, 1978a). The flow stress of the layer will therefore increase as the layer decreases in thickness until the flow stress has increased by ∆τE, at which point pile-ups will be unable to form and the increase in flow stress will be that given by equation (8.6). This is shown in Fig. 8.2 for a TiN/NbN multilayer, where the hardnesses of monolithic TiN and NbN are both taken as 25 GPa and the layer thicknesses of each are assumed to be equal.
© Woodhead Publishing Limited, 2006
226
Ceramic matrix composites
None of this explains why hardening is observed in one set of TiN/NbN data (Shinn et al., 1992) but not in three others (Ljungcrantz et al., 1998; Molina-Aldareguia et al., 2002): see Fig. 8.2. Nor does it explain the lack of hardening in the TiC/VC system (Högberg et al., 2001). All of these have the same crystal structure and were single crystal, with the exception of the data of Ljungcrantz et al., where the layers were polycrystalline. Ljungcrantz’s data is interesting because it contains two materials, denoted B and L, made in different laboratories. Apart from a slight difference in hardness between the two, neither showed any hardening. One possibility is differences in the sharpness of the interfaces. The force on the dislocation due to the elastic inhomogeneity is dependent upon the rate of change of the elastic constants and hence on the rate of composition change. Ideally this should be abrupt, but it rarely is. The magnitude of this effect has been estimated by Krzanowski (1991, 1992). By determining how the composition, and hence the modulus, changed for a small increment of movement, he was able to show that the effect of increasing the interfacial width from 1 nm to just 3 nm would decrease the increment of flow stress by approximately an order of magnitude. Using X-ray diffraction and assuming that the composition would change linearly across the interface between layers of pure TiN and NbN, Chu and Barnett (1995) measured the interface widths to be approximately 2 nm. Using Krzanowski’s analysis they found that the hardness increment predicted was no greater than that obtained previously, where it was assumed that the compositional changes were due entirely to the composition modulation. In any case it does not explain why sometimes no increase in the hardness is observed as the interface widths in some experiments were measured to be approximately 1 nm (MolinaAldareguia et al., 2002), less than those measured by Chu and Barnett (1995) in their experiments. Sharp X-ray satellite peaks indicative of sharp interfaces were also measured in the other work (Högberg et al., 2001). Furthermore it is not clear why no hardening is observed in the NbN/VN multilayers. Whilst it is true that there is little difference in G for NbN and VN, the difference in dislocation line energy is associated with a change in the dislocation line energy, U, which is proportional to Gb2, taken as |( GA bA2 – GB bB2 | /( GA BA2 + GB bB2 ). Using Table 8.1 and remembering that TiN, NbN and VN have the same crystal structures and slip systems, it can be seen that although the elastic mismatch, Q, is indeed much smaller in the case of the NbN/VN (0.07%) multilayer than it is for the TiN/NbN (6.5%) multilayer, the difference in the value of Gb2 is reversed, the difference being only 6% for TiN/NbN but 13% for NbN/VN. It is therefore not clear that the absence of any hardening in the NbN/VN system is consistent with these ideas as claimed.
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
8.3.3
227
Hardening due to lateral flow of material
It has also been suggested that flow might occur at lower stresses than those predicted above by movement of material within the individual layers (Chu and Barnett, 1995). This has been observed in pearlitic structures made up of alternating layers of ferrite and cementite, and observations in other multilayer systems suggest that that deformation might occur in this way (Gil-Sevillano, 1979). Two cases have been identified: the first where only the movement of a pre-existing dislocation loop is required, the second where the activation of a dislocation source within the layer is needed. Gil-Sevillano (1979) showed that the extra stress, ∆τM, required to move a dislocation half-loop in a layer of width l is ∆τ M =
2α Gb cos θ l ⋅ ln l b cos θ
(8.10)
where α = 1/4π and θ is the angle between the slip-plane and the normal to the interface. The extra stress required to activate a dislocation, ∆τA, is twice that for motion. The overall shear flow stress of a given layer is therefore equal to ∆τM plus any contribution from the lattice resistance, τL, which in these materials in bulk form is the dominant contribution to the flow stress. Chu and Barnett (1995) then assume that the multilayer of materials A and B is strained as it were in compression, so that the overall uniaxial flow stress is given by the volume-averaged flow stress, that is
σ = σA
lA l + σB B Λ Λ
(8.11)
where
σ A,B = m ′ ( τ L + ∆τ M )
(8.12)
where m′, the constant relating the uniaxial to the shear stress, is taken as 2. The overall hardness of the multilayer can then be obtained by taking the hardness as three times the uniaxial flow stress, taking the Burgers’ vector to be a/2√2, where a is the lattice parameter, as slip occurs on the {110}<1 1 0> slip system, and taking an average value of cosθ to be 0.5. The predicted hardnesses given by both dislocation motion and source activation are shown in Fig. 8.2. Comparing this with the data on TiN/NbN multilayers, it can be seen that the rising portion of the curve is a fair fit to the data of Shinn et al. (1992). However, it is not clear why such a process might operate at a considerably higher stress than that required to drive dislocations across the interfaces, as discussed in the previous section and shown as the horizontal line in Fig. 8.2. Indeed the predictions seem more consistent with the data where little or no hardening was observed. The possibility that hardening might arise simply because flow is restricted
© Woodhead Publishing Limited, 2006
228
Ceramic matrix composites
to the individual layers greatly increases the range of systems that might be developed. The materials considered so far have generally been isostructural (generally all having the B1 rocksalt structure). However flow can be restricted to the individual layers simply by using materials which have dissimilar structures so that flow cannot take place simply by the movement of a dislocation from one phase to the next. A range of systems have been investigated, including TiN/TiB2, ZrN/ZrB2 (Barnett et al., 2003) and systems containing metal layers, such as Mo/NbN, W/NbN (Madan et al., 1998), TiN/Cu (Ljungcrantz, 1995) and Y2O3/ZrO2 (Yashar and Barnett, 1999). Some hardening is seen in the Mo/NbN and W/NbN systems, but the effect is much less pronounced than in the TiN/NbN described earlier and no hardening at all is observed in the other systems: see Fig. 8.1(c). The reasons for this are not clear, as in all cases the interfaces are extremely sharp (Barnett et al., 2003). It can be seen from equations (8.10) and (8.11) that the contribution to the hardening comes mainly from the layers with a higher elastic modulus. However, differences in the elastic properties between the layers will cause the loops in the stiffer layers to be pulled across the interfaces, for the same reason that loops in the less stiff layers are repelled by the interfaces, greatly diminishing their contribution to the overall flow stress. Furthermore the assumption that the overall hardness is given by a thicknessaveraged hardness of the two materials applies if the multilayer were to be pulled in tension with the layers parallel to the tensile axis, but it is not clear whether this is true when the sample is being indented. For instance there are observations that, under a nanoindentation, the deformation is concentrated in the weaker of the two phases, as shown in Fig. 8.3, which shows a crosssection through an indentation in a TiB2/Al multilayer, where the compressive strain in the Al layers is greater than that in the TiB2 layers. The hardness is observed to fall very rapidly as the metal volume fraction is increased, although there is still deformation in the TiB2 layers, which may occur due to porosity in the layers. The wavelength in this material is 200 nm, which is relatively thick. In a TiN/Cu multilayer where Λ = 4.5 nm (Ljungcrantz, 1995), the hardness varies between the values for monolithic films of TiN and Cu, suggesting that the metal layer is indeed constrained when it is very thin, (see Fig. 8.4) but that there is no need to invoke atomistic processes. Only in the Mo/NbN multilayers (Λ = 5 nm) does the hardness rise above that even where the metal fraction is high, although the effect is not large (Barnett et al., 2003). The reasons for the difference between the TiN/Cu and the Mo/NbN multilayers are not clear, but are possibly associated with the difficulty of plastic flow in both molybdenum and tungsten. However, the effect is still not large. And as radial flow is preferred in monolithic materials, one might expect that the lateral flow occurring in these multilayers might be more difficult, giving
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
229
500 nm
8.3 Deformation under an indentation in an Al/TiB2 multilayer, with a wavelength of 200 nm, showing that deformation is concentrated in the Al.
30
Hardness (GPa)
TiB2 film
20 TiN film NbN film
10 = Mo/NbN = TiN/Cu = Al/ TiB2 0
0
Cu film Mo film 0.5 Metal fraction
1
8.4 The effect of varying the thickness fraction of the phases in multilayers of Al/TiB2, TiN/Cu and Mo/NbN, with wavelengths of 200, 4.5 and 5 nm respectively. In the coarser multilayer, deformation appears to be concentrated in the softer layer. In the TiN/Cu multilayer, the behaviour is close to a rule of mixtures, whereas for the Mo/NbN multilayer, some further hardening appears to be occurring.
© Woodhead Publishing Limited, 2006
230
Ceramic matrix composites
some hardening consistent with that observed. Unfortunately the magnitude of such an effect is unknown due to the lack of any theoretical studies.
8.3.4
Summary
It can be seen that the idea of flow occurring within the individual layers might explain why some materials are harder than either of the materials from which they are made, giving a fair fit to the analyses. However, most multilayers show hardnesses in between those of the individual components or with only minimal hardening, including some TiN/NbN multilayers, even though the interfacial widths are measured to be the same as in those where hardening is observed, as well as in carbide and in oxide superlattices. Whilst atomistic mechanisms may be required to explain the magnitude of the hardness in some cases, it appears that in many they do not. Despite this, the observed lateral flow is expected to give rise to some hardening, even where flow is treated in a continuum manner.
8.4
Microstructural changes due to making a multilayer
So far we have considered the properties of a multilayer only in terms of the effect of the composition modulation on the movement of dislocations either across the layers of different composition or, separately, within them. This gives a certain measure of agreement with particular sets of data, but when the experiments are considered in their entirety, these ideas cannot account by themselves for the observed behaviour. This is noticeable in Fig. 8.1(a), which shows the data for the isostructural nitrides and carbides, where there appears to be a significant differences in the hardnesses of the multilayers at longer wavelengths and in the monolithic films, suggesting the importance either of internal stresses or of microstructural changes to the individual layers of the multilayer, caused by making the material in a multilayer form.
8.4.1
Hardening due to internal stresses
It is well known from the work of Thornton, Hoffman and others that large residual stresses can be developed in thin films (Thornton and Hoffman, 1989; Hoffman, 1994). These may arise due to expansivity mismatch with the substrate or due to stresses induced during the growth of the coating, which can be varied depending on the conditions of temperature atmosphere and deposition method under which the film is grown. For more information on the origin of stresses in thin films, the reader is referred to the reviews by Windischmann (1987, 1992). Furthermore, because the coatings need typically be only a few microns in thickness, they can sustain internal stresses of the
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
231
order of gigapascals without peeling off the substrate (Kendall, 1975), although special precautions may be taken to improve the adhesion of the coating. There is considerable experimental work in both metallic and ceramic films, showing that such internal stresses can greatly increase the measured hardness. The simplest explanations are in terms of superimposing an inplane stress on the overall maximum shear stress under the indenter, although Pharr (Bolshakov et al., 1996; Tsui et al., 1996), looking at nanoindentation of Al films, considers the effect to arise due to pile-up around the indenter, causing the actual depth of penetration, and hence area of the indentation, to be greater than that calculated. However, regardless of the origin, residual compressive stresses in films are clearly associated with increases in the measured hardness of both monolithic and multilayer films. Figure 8.5 shows the results of Münz and his co-workers for different multilayers but all with a bilayer period of between 3 and 4 nm (Lewis et al., 1999; Münz et al., 2001). It can be seen that hardnesses of up to 70 GPa have been measured. Also plotted is data from TiAlN (Derflinger et al., 1999) and TiN (Martin et al., 1999) which shows surprisingly similar behaviour, suggesting that internal stresses might account for a substantial fraction of the hardening observed earlier in Fig. 8.2. This is consistent with the observation of differences in the hardnesses of monolithic materials and longer wavelength multilayer materials, by almost a factor of two. Furthermore the magnitude of such stresses is dependent on 12 Multilayers TiAlN/YN TiAlYN/VN TiAlN/VN TiAlN/CrN TiAlN/ZrN CrN/NbN Monoliths TiAlN TiN IAAD TiN FAD
Internal stress (GPa)
10 8 6 4 2 0 10
20
30
40 50 60 Hardness (GPa)
70
80
8.5 Variation of the internal stress and measured hardness for a variety of multilayer films compared with monolithic films. The multilayer films are shown with filled symbols and the monolithic films with unfilled ones. The TiN was made by either ion assisted arc deposition (IAAD) or filtered arc deposition (FAD). Note that hardnesses in the monolithic materials of up to 50 GPa appear to be achievable simply by internal stresses.
© Woodhead Publishing Limited, 2006
232
Ceramic matrix composites
the processing conditions, consistent with the variability in the observed behaviour. It is also consistent with the observation that heating can cause a reduction in the hardness in TiN/NbN multilayers. This is normally attributed to a decrease in the sharpness of the interface (Barnett et al., 2003) but also occurs in monolithic films (Jindal et al., 1999).
8.4.2
Deformation processes and microstructure of the film
In uniformly strained materials, deformation structures can be readily observed using transmission electron microscopy. However, it is much more difficult to prepare a similar sample where the deformation is more localized, as is the case of nanoindentation. Recently this situation has been revolutionized by the development of focused ion beam techniques for semiconductor processing, so that it is possible to select the region to be thinned to within 100 nm (Overwijk et al., 1993; Saka, 1998). Figure 8.6 shows an example of a cross-section made through a nanoindentation in a TiN/NbN multilayer grown on an MgO substrate (MolinaAldareguia et al., 2002). As expected, shear occurs on (101) and (10 1 ) planes inclined at approximately 45° to the surface of the film (marked with black arrows). However, other processes are also taking place. In particular, the crystal close to the indentation is deformed by rotation of the lattice planes and has been observed elsewhere both in multilayers and in monolithic materials (Hultman et al., 1999). Although the situation here is not yet understood, such rotations can occur by the development of an array of geometrically necessary dislocations, associated with the strain gradients under the indent (Fleck et al., 1994) and which have in themselves been associated with hardening (Ashby et al., 1989). In Fig. 8.6(a) it can be seen that flow under an indentation can occur by dislocation motion across the planes, whereas in Fig. 8.6(b) it occurs by compression of the layers: compare points A and B. Another feature that can be seen in Fig. 8.6(a) is the shearing that has occurred in the columnar grain boundaries in the film (marked with the white arrows). Sometimes these boundaries can grow through the complete thickness of the film and lead to a section of the film being punched out, with a consequent reduction in hardness. This weakness is associated with the pores that are found along these boundaries. Such effects are well known in monolithic films, where their number and size can be diminished by increasing the surface mobility of adatoms on the growing film, for instance by raising the substrate temperature or increasing the flux of bombarding ions (Este and Westwood, 1987; Thornton and Hoffman, 1989; Hoffman, 1994). However, the pore distribution can also be influenced by the multilayer structure, as shown in Fig. 8.7, which shows how the size and number of the pores is
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
233
Lattice rotations
200 nm
MgO (001) (a)
Bottom of indent
C C
A
B 25 nm (b)
8.6 Cross-sectional TEM images of a 10 mN indent in a TiN/NbN multilayer. White arrows point to a pre-existing columnar boundary along which shear can be observed. Black arrows point to concentrated shear along {101} crystallographic planes; (b) shows a high-magnification image under the tip of the indenter, where it can be seen that deformation has occurred by the layers at point B having decreased compared to those at A.
© Woodhead Publishing Limited, 2006
234
Ceramic matrix composites
(a)
(b)
8.7 Cross-sectional TEM images of (a) a CrN–AlN multilayer (the AlN is the lighter phase) and (b) monolithic CrN. Note that there are long pores present in the monolithic CrN, marked by the black arrow, but these are not present in the multilayer.
reduced as the bilayer thickness, Λ, decreases, giving an increase in the hardness. Molina-Aldareguia has also shown how the structure of TiN in a multilayer adopts the denser structure of the NbN, again suggesting that improvements in the layer structure might be important (Molina-Aldareguia, 2002). A rather extreme example of this is given by Wang et al. in a study of TiN/ AlN multilayers (Wang et al., 1998). For samples made with pulsed d.c. substrate bias, the improved hardness was retained down to the lowest values of Λ. However, when r.f. substrate bias was used the structure at Λ < 2 nm
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
235
was made up of columnar grains that had grown separately from their neighbours and led to a decrease in hardness from 20 to 12 GPa. Changes in hardness at low values of Λ in this system have also been obtained elsewhere (Setoyama et al., 1996). The coating appears to behave rather like a series a bed-springs where only that material directly under the indenter is pushed downwards into the substrate, with no lateral constraint as would occur if the film were intact. Whilst this can be envisaged for the case where the grains are completely separate, it cannot occur where the boundaries are porous but connected, as is more commonly observed (Hultman et al., 1992; MolinaAldareguia et al., 2000, 2002). In this case porosity presumably reduces the stress at which a columnar boundary will fracture, giving rise to a segment of multilayer being pushed into the film. However, there is no quantitative description of such a process or of the microstructural variables that influence it, despite its importance in limiting the hardness of the film. This increase in hardness has been associated with the formation of a rocksalt cubic (c) form of AlN, stabilized by the reduction of the interfacial energy with the rocksalt-structured TiN (Madan et al., 1997). As the thickness increases the effect is offset by the increase in the volume free energy, so that the cubic form is stable only at layer thicknesses of less than 2 nm, although some increase is possible by using rocksalt-structured compounds such as VN with a lower mismatch (Li et al., 2002, 2004). Stabilization with other materials such as W (Kim et al., 2001) or ZrN (Wong et al., 2000) is also possible, and other compounds such as CrN (Yashar et al., 1998) can also show stabilized cubic forms. The reason for the increase in hardness has been attributed to the bulk modulus of c-AlN being greater than that of the wurtzite (w) structure. The bulk modulus of c-AlN has been predicted to be 270 GPa compared with the measured value of 205 GPa for w-AlN (Christensen and Gorczyca, 1993). The effect is two-fold. The higher bulk modulus is likely to give rise to a higher hardness as well as an increased elastic mismatch, assuming the Poisson ratio is similar in both crystal structures. However, as TiN has a bulk modulus greater than both forms of AlN, it is clear that any transformation to the wurtzite structure will lead to a reduction rather than an increase in repulsive stresses due to any modulus mismatch. The lattice resistance of the two crystal forms of AlN is not known. However, some tentative conclusions can be made by assuming that the lattice resistance at room temperature is close to the Peierls stress (Peierls, 1940), which agrees with experimental observations of a wide range of materials to within a factor of 3 and is given by
τp 2π = 2 exp ⋅ d G 1–ν 1 – ν b
(8.13)
where d is the spacing of the atoms across the slip plane and b is Burgers’
© Woodhead Publishing Limited, 2006
236
Ceramic matrix composites
vector. B1 rocksalt-structured TiN shears on the {110}<110> slip system at room temperature. If cubic AlN slips on the same slip system at room temperature as TiN, then the ratio d/b will be 0.5. In w-AlN, like others with the same structure, slip takes place by the movement of partial dislocations on the more closely spaced {0001} planes (Delavignette et al., 1961; Yonenaga, 2002), the glide planes, where d/b is 0.354. This suggests that τp /G for the latter should be almost 10 times that of the cubic form of AlN, so that despite the increase in the elastic modulus due to the increased packing density, wAlN would be expected to show a higher lattice resistance and hence be harder than the cubic form. At best the differences are minimal. This suggests that the increase in hardness where Λ < 2 nm was due to the structure being epitaxial, whereas at higher Λ it was polycrystalline, being made up of grains of both c- and w-AlN.
8.4.3
Summary
In summary, it is clear that there are substantial effects that vary systematically with the wavelength of the multilayer due both to internal stresses and the microstructure of the coatings. It has also been seen that deformation can occur not just by dislocation flow, as the initial analyses have assumed, but by mechanisms such as lattice rotations and shear along column boundaries. In addition, the use of indentation complicates the deformation field, so that the assumption that equal strains in both layers are required need not be correct. These effects all influence the hardness but have not so far been included in analyses.
8.5
Conclusions
It can be seen that ceramic multilayer structures have been produced with increments of the hardness of up to 60 GPa, increasing the hardness by up to a factor of almost 3. Initial work in this area has developed a number of ideas, such as the effect of modulus mismatch, which in some cases give good agreement with the models suggested but in many others do not. It is suggested that at least some of this discrepancy can be accounted for by differences in the microstructure and residual stress-state of the film, both of which are often poorly characterized. Furthermore there is very little direct evidence about how these structures deform and in particular about how different layers must be strained in order to accommodate the indenter when it is pressed into the sample. Further advances in this area will require the greater use of numerical techniques to analyse the complex stress and strain behaviour under the indentation, coupled with the use of recently developed techniques that allow the localized deformation behaviour to be observed in detail.
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
8.6
237
Future trends
Multilayer structures offer the potential for great increases in hardness and, although this has not been discussed here, also in wear resistance. The greatest disadvantage of such structures is that the maximum in the hardness observed appears over a very narrow range of wavelengths. This is undoubtedly a challenge where structures with complex surface structures have to be coated. The best way forward is undoubtedly to understand how such hardening arises and how the peak that is observed may be smoothed out. This will no doubt require a complex combination of hardening mechanisms, as used in complex alloy systems. Such an understanding will require the detailed characterization of the deformation processes occurring in indentation and wear and the correlation of these structures, initially, with the deformation patterns that might be expected from continuum models of flow under an indenter, before developing more complex atomistic models.
8.7
Further reading
A list of references is given below which will allow the reader to investigate any points of detail. A good place to start is the collection of papers in the March 2003 MRS Bulletin (Volume 28, issue 3), which provides a snapshot of some of the more recent work on hard materials. For a general study of thin films, try M. Ohring, The Materials Science of Thin Films, 1992, published by Harcourt Brace Jovanovich, Boston, MA.
8.8
References
Ashby, M.F., Blunt, F.J. and Bannister, M. (1989), ‘Flow characteristics of highly constrained metal wires’, Acta Metallurgica, 37, 1847–1857. Barnett, S.A., Madan, A., Kim, I. and Martin, K. (2003), ‘Stability of nanometer-thick layers in hard coatings’, MRS Bulletin, 28, 169–172. Benlahsen, M., Lepinoux, L. and Grilhe, J. (1993), ‘Image forces on dislocations: the elastic modulus effect’, Materials Science and Engineering, A164, 428–432. Bolshakov, A., Oliver, W.C. and Pharr, G.M. (1996), ‘Influences of stress on the measurement of mechanical properties using nanoindentation: Part II. Finite element simulations’, Journal of Materials Research, 11, 760–768. Cahn, J.W. (1963), ‘Hardening by spinodal decomposition’, Acta Metallurgica, 11, 1275– 1282. Cheng, Y.T. and Cheng, C.M. (2000), ‘What is indentation hardness?’, Surface and Coatings Technology, 133–134, 417–424. Christensen, N.E. and Gorczyca, I. (1993), ‘Calculated structural phase-transitions of aluminium nitride under pressure’, Physics Review B, 47, 4307–4314. Chu, X. and Barnett, S.A. (1995), ‘Model of superlattice yield stress and hardness enhancements’, Journal of Applied Physics, 77, 4403–4411. Delavignette, P., Kirkpatrick, H.B. and Amelinckx, S. (1961), ‘Dislocations and stacking faults in aluminium nitride’, Journal of Applied Physics, 32, 1098–1100.
© Woodhead Publishing Limited, 2006
238
Ceramic matrix composites
Derflinger, V., Brandle, H. and Zimmermann, H. (1999), ‘New hard/lubricant coating for dry machining’, Surface and Coatings Technology, 113, 286–292. Este, G. and Westwood, W.D. (1987), ‘Stress control in reactively sputtered AlN and TiN films’, Journal of Vacuum Science and Technology A, 5, 1892–1897. Fleck, N.A., Muller, G.M., Ashby, M.F. and Hutchinson, J.W. (1994), ‘Strain gradient plasticity: theory and experiment’, Acta Metallurgica et Materialia, 42, 475–487. Gil-Sevillano, J. (1979), ‘On the yield and flow stress of lamellar pearlite’, in Strength of Metals and Alloys, Vol. 2 (ed. Haasen, P., Gerold, V. and Kostorz, G.), Pergamon Press, pp. 819–824. Helmersson, U., Todorova, S., Barnett, S.A., Sundgren, J.-E., Market, L.C. and Greene, J.E. (1987), ‘Growth of single-crystal TiN/VN strained-layer superlattices with extremely high mechanical hardness’, Journal of Applied Physics, 62, 481–484. Hoffman, D.W. (1994), ‘Perspective on stresses in magnetron-sputtered thin films’, Journal of Vacuum Science and Technology A, 12, 953–961. Högberg, H., Birch, J., Oden, M., Malm, J.O., Hultman, L. and Jansson, U. (2001), ‘Growth, structure and mechanical properties of transition metal carbide superlattices’, Journal of Materials Research, 16, 1301–1310. Hubbard, K.M., Jervis, T.M., Mirkarimi, P.B. and Barnett, S.A. (1992), ‘Mechanical properties of epitaxial TiN/(V0.6Nb0.4)N superlattices measured by indentation’, Journal of Applied Physics, 72, 4466–4468. Hultman, L., Wallenberg, L.R., Shinn, M. and Barnett, S.A. (1992), ‘Formation of polyhedral voids at surface cusps during growth of epitaxial TiN/NbN superlattice and alloy films’, Journal of Vacuum Science and Technology A, 10, 1618–1624. Hultman, L., Engström, C., Birch, J., Johansson, M.P., Odén, M., Karlsson, L. and Ljungcrantz, H. (1999), ‘Review of the thermal and mechanical stability of TiN-based thin films’, Zeitschrift für Metallkunde, 90, 803–813. Jayaweera, N.B., Downes, J.R., Frogley, M.D., Hopkinson, M., Bushby, A.J., Kidd, P., Kelly, A. and Dunstan, D.J. (2003), ‘The onset of plasticity in nanoscale contact loading’, Proceedings of the Royal Society London A, 459, 2049–2068. Jindal, P.C., Santhanam, A.T., Schleinkofer, U. and Schuster, A.F. (1999), ‘Performance of PVD TiN, TiCN and TiAlN coated cemented carbide tools in turning’, International Journal of Refractory Metals and Hard Materials, 17, 163–170. Johnson, K.L. (1970), ‘The correlation of indentation experiments’, Journal of Mechanics and Physics of Solids, 18, 115–126. Kato, M., Mori, T. and Schwartz, L.H. (1980), ‘Hardening by spinodal modulated structure’, Acta Metallurgica, 28, 285–290. Kelly, A. (1991), In 2nd International Conference on Advanced Materials and Technology. New Compo ’91 Hyogo, Kobe, Japan. Kelly, A., Groves, G.W. and Kidd, P. (2000), Crystallography and Crystal Defects, John Wiley & Sons, Chichester. Kendall, K. (1975), ‘The effects of shrinkage on interfacial cracking in a bonded laminate’, Journal of Physics D: Applied Physics, 8, 1722–1732. Kim, I.W., Madan, A., Gunz, M.W., Dravid, V.P. and Barnett, S.A. (2001), ‘Stabilization of zinc-blende cubic AlN in AlN/W superlattices’, Journal of Vacuum Science and Technology A, 19, 2069–2073. Kim, J.O., Achenbach, J.D., Mirkarimi, P.O., Shinn, M. and Barnett, S.A, (1992), ‘Elastic constants of single-crystal transition-metal nitride, films measured by line-focus scoustic microscopy’, Journal of Applied Physics, Vol. 72 5, 1805–1811. Koehler, J.S. (1970), ‘Attempt to design a strong solid’, Physics Review B, 2, 547–551.
© Woodhead Publishing Limited, 2006
Hardness of multilayered ceramics
239
Krzanowski, J.E. (1991), ‘The effect of composition profile shape on the strength of metallic multilayer structures’, Scripta Materialia, 25, 1465–1470. Krzanowski, J.E. (1992), In Materials Research Society Symposium Proceedings, Vol. 239 (ed. Nix, W.D., Bravman, J.C., Arzt, E. and Freund, L.B.), Materials Research Society, Boston, MA, pp. 509–515. Lehoczky, S.L. (1978a), ‘Retardation of dislocation generation and motion in thin-layered metal laminates’, Physical Review Letters, 41, 1814–1818. Lehoczky, S.L. (1978b), ‘Strength enhancement in thin-layered Al–Cu laminates’, Journal of Applied Physics, 49, 5479–5485. Lewis, D.B., Wadsworth, I., Münz, W.-D., Kuzel, R. and Valvoda, V. (1999), ‘Structure and stress of TiAlN/CrN superlattice coatings as a function of CrN layer thickness’, Surface and Coatings Technology, 116–119, 284–291. Li, G., Lao, J., Tian, J., Han, Z. and Gu, M. (2004), ‘Coherent growth and mechanical properties of AlN/VN multilayers’, Journal of Applied Physics, 95, 92–96. Li, Q., Kim, I.W., Barnett, S.A. and Marks, L.D. (2002), ‘Structures of AlN/VN superlattices with different AlN layer thicknesses’, Journal of Materials Research, 17, 1224–1231. Ljungcrantz, H. (1995), ‘Growth, microstructure and mechanical properties of Ti and TiN thin films, and TiN-based supertlattices’, In Department of Physics, Linköping University, Linköping, Sweden. Ljungcrantz, H., Engström, C., Hultman, L., Olsson, M., Chu, X., Wong, M.S. and Sproul, W.D. (1998), ‘Nanoindentation hardness, abrasive wear, and microstructure of TiN/ NbN polycrystalline nanostructured multilayer films grown by reactive magnetron sputtering’, Journal of Vacuum Science and Technology A, 16, 3104–3113. Madan, A., Kim, I.W., Cheng, S.C., Yashar, P., Dravid, V.P. and Barnett, S.A. (1997), ‘Stabilization of cubic AlN in epitaxial AlN/TiN superlattices’, Physical Review Letters, 78, 1743–1746. Madan, A., Wang, Y., Barnett, S.A., Engström, C., Ljungcrantz, H., Hultman, L. and Grimsditch, M. (1998), ‘Enhanced mechanical hardness in epitaxial nonisostructural Mo/NbN and W/NbN superlattices’, Journal of Applied Physics, 84, 776–785. Marsh, D.M. (1963), ‘Plastic flow in glass’, Proceedings of the Royal Society A, 279, 420–435. Martin, P.J., Bendavid, A., Netterfield, R.P., Kinder, T.J., Jahan, F. and Smith, G. (1999), ‘Plasma deposition of tribological and optical thin film materials with a filtered cathodic arc source’, Surface and Coatings Technology, 112, 257–260. Mirkarimi, P.B., Hultman, L. and Barnett, S.A. (1990), ‘Enhanced hardness in latticematched single-crystal TiN/V0.6Nb0.4N superlattices’, Applied Physics Letters, 57, 2654–2656. Molina-Aldareguia, J.M. (2002), ‘Processing and nanoindentation behaviour of nitride multilayers’, in Department of Materials Science and Metallurgy, University of Cambridge, Cambridge, UK. Molina-Aldareguia, J.M., Lloyd, S.J., Barber, Z.H., Blamire, M.G. and Clegg, W.J. (2000), In Materials Research Society Symposium Proceedings: Thin Film Stresses and Mechanical Properties VIII, Vol. 594 (ed. Vinci, R., Kraft, O., Moody, N., Besser, P. and Shaffer, E. II), Materials Research Society, Boston, MA, pp. 9–14. Molina-Aldareguia, J.M., Lloyd, S.J., Odén, M., Joelsson, T., Hultman, L. and Clegg, W.J. (2002), ‘Deformation structures under indentations in TiN/NbN single-crystal multilayers deposited by magnetron sputtering at different bombarding ion energies’, Philosophical Magazine A, 82, 1983–1992. Münz, W.-D., Lewis, D.B., Hovsepian, P.E., Schönjahn, C., Ehiasarian, A. and Smith, I.J.
© Woodhead Publishing Limited, 2006
240
Ceramic matrix composites
(2001), ‘Industrial scale manufactured superlattice hard PVD coatings’, Surface Engineering, 17, 15–27. Overwijk, M.H.F., Van der Hauvel, F.C. and Bulle-Lieuwma, C.W.T. (1993), ‘Novel scheme for the preparation of transmission electron microscopy specimens with a focused ion beam’, Journal of Vacuum Science and Technology A, 11, 202. Pacheco, E.S. and Mura, T. (1969), ‘Interaction between a screw dislocation and a bimetallic interface’, Journal of Mechanics and Physics of Solids, 17, 163–170. Peierls, R. (1940), ‘The size of a dislocation’, Proceedings of the Physical Society, 52, 34–37. Saka, H. (1998), ‘Transmission electron microscopy observation of thin foil specimens prepared by means of a focused ion beam’, Journal of Vacuum Science and Technology B, 16, 2522–2527. Setoyama, M., Nakayama, A., Tanaka, M., Kitagawa, N. and Nomura, T. (1996), ‘Formation of cubic-AlN in TiN/AlN superlattice’, Surface and Coatings Technology, 86–87, 225–230. Shinn, M. and Barnett, S.A. (1994), ‘Effect of superlattice layer elastic moduli on hardness’, Applied Physics Letters, 64, 61–63. Shinn, M., Hultman, L. and Barnett, S.A. (1992), ‘Growth, structure, and microhardness of epitaxial TiN/NbN superlattices’, Journal of Materials Research, 7, 901–911. Tabor, D. (1951), Hardness of Metals, Clarendon Press, Oxford, UK. Thornton, J.A. and Hoffman, D.W. (1989), ‘Stress-related effects in thin films’, Thin Solid Films, 171, 5–31. Tsui, T.Y., Oliver, W.C. and Pharr, G.M. (1996), ‘Influence of stress on the measurement of mechanical properties using nanoindentation: Part I. Experimental studies in an aluminium alloy’, Journal of Materials Research, 11, 752–759. Wang, Y.Y., Wong, M.S., Chia, W.J., Rechner, J. and Sproul, W.D. (1998), ‘Synthesis and characterization of highly textured polycrystalline AlN/TiN superlattice coatings’, Journal of Vacuum Science and Technology A, 16, 3341–3347. Williams, W.S. and Schaal, R.D. (1962), ‘Elastic deformation, plastic flow and dislocations in single crystals of titanium carbide’, Journal of Applied Physics, 33, 955–962. Windischmann, H. (1987), ‘An intrinsic stress scaling law for polycrystalline thin films prepared by ion beam sputtering’, Journal of Applied Physics, 62, 1800–1807. Windischmann, H. (1992), ‘Intrinsic stress in sputter deposited thin films’, Critical Reviews in Solid State and Material Sciences, 17, 547–596. Wong, M.-S., Hsiao, G.-Y. and Yang, S.-Y. (2000), ‘Preparation and characterization of AlN ZrN and AlN TiN nanolaminate coatings’, Surface and Coatings Technology, 133–134, 160–165. Yashar, P., Chu, X., Barnett, S.A., Rechner, J., Wang, Y.Y., Wong, M.S. and Sproul, W.D. (1998), ‘Stabilization of cubic CrN0.6 in CrN0.6/TiN superlattices’, Applied Physics Letters, 72, 987–989. Yashar, P.C. and Barnett, S.A. (1999), ‘Deposition and mechanical properties of polycrystalline Y2O3/ZrO2 superlattices’, Journal of Materials Research, 14, 3614– 3622. Yonenaga, I. (2002), ‘Hardness of bulk single-crystal GaN and AlN’, MRS Internet Journal Nitride Semiconductor Research, 7, 1–4.
© Woodhead Publishing Limited, 2006
Part III Nanostructured ceramic composites
© Woodhead Publishing Limited, 2006
9 Nanophase ceramic composites L Y O N G L I, Beijing University of Technology, China
9.1
Introduction
The application of ceramics has infiltrated almost all fields in the last 20 years, because of their advantages over metals due to their strong ionic or covalent bonding. But it is just this bonding nature of ceramics that directly results in their inherent brittleness and difficulty in machining. In other words, ceramics show hardly any macroscopic plasticity at room temperature or at low temperatures like metals. Hence, superplasticity at room temperature is a research objective for structural ceramics. In recent years, many researches have been carried out to investigate nanophase ceramic composites. Depending on the matrix grain size, nanophase ceramic composites can be classified in two fundamental groups. One is composed of micrometersized matrices dispersed with a nanometer second phase, which has attracted a lot of interest in the last 15 years. In this group, the second phase plays a crucial role that affects the microstructure and the properties. Niihara [1] has divided it into three types – intragranular, intergranular and intra-/intergranular – trying to relate the distribution of the second nanophase in the matrix. Niihara and co-workers have reported dramatic improvements in toughness, strength at room temperature and high temperatures, creep strength and thermal shock resistance by incorporating nanocrystalline dispersion in a microcrystalline matrix. The other group of nanophase ceramic composites is nanocrystalline matrix composites, also called nanoceramics, in which the matrix grain size is below 100 nm. The nano–nano type microstructure will be formed when the second phase is also nano-scaled. Nanoceramics exhibit promising properties due to the changes in deformation mechanisms when the grain size is reduced to the order of 100 nm. The superplasticity of nanocrystalline CaF2 and nanocrystalline TiO2 at low temperatures, reported by Karch et al. [2], indicates that ceramics are learning to ‘bend’ instead of fracture. Furthermore, nanoceramics also show high toughness, in which a novel toughening mechanism called 243 © Woodhead Publishing Limited, 2006
244
Ceramic matrix composites
Ferroelectric Domain Switching is recognized, different from that in micro– nano type ceramic composites.
9.2
Micro–nano type ceramic composites
In early nanocomposites, hard and strong dispersoids, such as SiC, Si3N4, TiC, etc., were mainly incorporated into the matrix to improve the mechanical properties. But in later years, enhancement of fracture strength was also achieved by addition of even soft and weak dispersoids like metals, graphite and h-BN [3–5]. The density, microstructure and mechanical properties of nano-sized particulate dispersion nanocomposites were strongly dependent on the volume fraction of particulate dispersion and sintering conditions.
9.2.1
Hard nanoparticle dispersed nanocomposites
Hard nanoparticles usually have a higher sintering temperature than that of the matrix, so that the sintering temperature is increased with increasing hard-particulate content. In Al2O3/SiC systems, only 5 vol% SiC incorporation can evidently cumber the densification process. The nearly full densities attained by hot-pressing (HP) were achieved at 1600°C for 5 vol% SiC, at 1700°C for 11 vol% and at 1800°C for up to 33 vol%, while 1400–1500°C was needed for Al2O3 monolithic ceramics [1]. At the same time, matrix grain growth was dramatically inhibited owing to the pinning action of dispersed particles. The sintering temperature for Al2O3 is 1500°C, in which the grain size grew up to 2.6 µm with uneven distribution. Al2O3 grain size in 5 vol% and 10 vol% SiC dispersed composites are 1.6 µm and 1.4 µm, respectively, with homogeneous distribution, even though sintering temperatures reach 1700°C or higher [3]. In general, particles disperse according to their grain size and the variety of the matrix. For Al2O3/SiC, finer particles disperse within the matrix grains and larger particles at the grain boundaries. The critical grain size is typically 200 nm. For the MgO/SiC, Al2O3/Si3N4 and natural mullite/SiC composites, nanoparticles homogeneously disperse within as well as at the grain boundaries, which were confirmed to be the intra/inter-type nanocomposites. It was found that dramatic improvements in toughness, strength, creep strength and thermoresistance could be achieved by incorporating nano-SiC dispersion in a microcrystalline matrix. Subsequently, similar improvements were found in other nanocomposites [1]. Niihara [1] considered the improved toughness was mainly attributed to the residual stress that results from differential thermal expansion coefficients of two phases. In Al2O3/SiC systems, the tensile hoop stress, thought to be over 1000 MPa around the nanoparticles within the matrix grains, was generated from the large thermal expansion mismatch. Thus, the material may be
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
245
toughened primarily by the crack deflection and bridging due to the nanosized SiC particles within the matrix grains. However, it was argued by other researchers that the incorporation of SiC into Al2O3 works on toughness in a very limited degree. Hoffman et al. [6] and Ferroni and Pezzotti [7] found that in Al2O3/5 vol% SiC systems, cracks were not distinctly deflected or bridged by nano-sized SiC, and no R-curve existed in Al2O3/5 vol% SiC nanocomposites. Zhao et al. [8] considered that the dramatically improved toughening mentioned by Niihara is actually attributed to surface compressive stresses. After relief annealing at high temperature, the toughness of Al2O3/ SiC nanocomposites consequently had a sudden decrease. The remarkable refinement of matrix grains by the dispersions is associated with the sub-grain boundary formation in the matrix grains. Sub-grain boundaries were found to be formed due to the pinning and pile-up of dislocations by intragranular hard particles, which were generated in the matrix during cooling from the sintering temperature by the highly localized thermal stress within and/or around the hard particles caused from the thermal expansion mismatch between the matrix and the dispersions. This thermal expansion mismatch, on the other hand, causes residual compressive stresses at the matrix grain boundaries. Both effects strengthen the composites. The sub-grain boundaries were more extensive for Al2O3/5 vol% SiC after annealing at 1300°C, and then the fracture strength was further improved from ~1050 MPa to 1550 MPa. The improvement in high-temperature hardness and brittle–ductile transition temperature (BDTT) must be due to the pinning of dislocations by the nanosized dispersions. The Al2O3/SiC, Al2O3/Si3N4 and MgO/SiC nanocomposites give a notable improvement in high-temperature strength up to and over 1000°C. In particular, the greatest improvement in high-temperature strength was observed for the MgO/SiC nanocomposites [1]. It is well known that grain boundary sliding and/or cavitation are responsible for the hightemperature strength degradation of oxide ceramics. Thus, the enhancement in strength at high temperatures is mainly due to the prohibition of the grain boundary sliding or cavitation by the dispersions within the matrix grains.
9.2.2
Metal nanoparticle dispersed nanocomposites
The mechanical properties of ceramics were improved by the addition of nano-sized metal particles that were dispersed in the ceramic matrix. Ceramic/ metal nanocomposites consisted of an oxide ceramic and either refractory metal such as in the Al2O3/W, Al2O3/Mo and ZrO2/Mo systems or a metal with a low melting point such as in Al2O3/Ni, Al2O3/Cr, Al2O3/Co, Al2O3/Fe and Al2O3/Cu systems. These composites were fabricated by hot-pressing fine ceramic and metal powder mixtures or by reducing and hot-pressing the matrix and metal oxide powders [9–15].
© Woodhead Publishing Limited, 2006
246
Ceramic matrix composites
For refractory metal dispersed systems, an outstanding improvement in toughness has been obtained in Al 2 O 3/5 vol% Mo – as high as 7.1 MPa.m1/2. At the same time, the fracture strength was just 306 MPa, much lower than that of Al 2O 3 monolithic ceramics. In Al 2 O 3/Mo nanocomposites, the strength decreased with increasing sintering temperature, while toughness was considerably improved with increasing Mo volume fractions or increasing sintering temperature. When sintered at low temperature, nano-sized Mo particles are dispersed within the matrix grains, which can improve the strength of grain boundaries and induce transgranular fracture. With an increase of sintering temperature or Mo volume fraction, nano-sized Mo particles agglomerated and grew to elongated grains. These elongated Mo grains make cracks deflect and bridge, which plays a key role in toughness improvement. Plastic deformation of large Mo particles at the crack tip also gives an important contribution to toughening. In low melting point metal dispersed systems, the Al2O3/Ni system has been studied to obtain the desired microstructure and improvement of mechanical properties by modification of the microstructure. Moreover, considering the magnetic properties of the composites, it was expected to improve both mechanical and magnetic properties by incorporating merely nanometer-sized Ni, Co, and Fe into an Al2O3 matrix. The average grain size of Al2O3/5 vol% Ni composites is finer (0.64 µm) than that of monolithic Al2O3 (1.2 µm), which is due to the growth restraint by the homogeneous dispersion of fine nickel particles. Fine nickel particles, less than 100 nm in size, disperse homogeneously at the matrix grain boundaries, forming the intergranular nanocomposite. The ferromagnetic properties of nickel, such as high coercive force, were observed because of the fine magnetic dispersions, which indicates a functional value of structural composites. The strength was enhanced up to 1090 MPa by dispersing only 5 vol% of nickel, almost doubling the strength of hotpressed monolithic Al2O3. The high strength value of the Al2O3/5 vol% Ni composites could be explained by the refinement of the matrix grains, because of the growth restriction caused by fine nickel dispersion. With increasing Ni content over 10 vol%, the strength decreased and attained a value of 700 MPa for Al2O3/20vol%Ni. This variation could be caused by the agglomeration of Ni particles dispersion at higher contents.
9.2.3
Soft and weak nanoparticle dispersed nanocomposites
The super-fine dispersoids with laminar structure and a low modulus could be expected to play a crucial role in improving the machinability of ceramics and their mechanical properties. Among the various nanocomposites, h-BN reinforced composites showed excellent corrosion resistance to molten metal
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
247
and high thermo shock resistance as well as good machinability [5, 16–19]. Vickers hardness of both micro- and nanocomposites decreased directly with an increase of BN content. The fracture strength was marginally higher than that of the monolithic counterparts by adding h-BN up to 5 vol% and then decreased gradually with an increase in BN for incorporation of low strength BN, caused by the aggregation of h-BN particles. Oku and co-workers [20] developed a chemical process to prepare nano-sized BN coatings on ceramic powders by reducing boric acid and urea in hydrogen gas. Kusunose et al. [16] reported that Si3N4/BN nanocomposites with homogeneously dispersed nano-sized h-BN in an amount of not less than 20 vol% possess both good machinability and outstanding strength as high as about 1100 MPa. To date, examples of such materials include Si3N4/BN, Sialon/BN, SiC/BN, Al2O3/ BN and 3Y-ZrO2/BN. The fracture strength of nanocomposites was considerably improved, in comparison to the corresponding microcomposites. This implies that a nanoBN coating on particle surfaces effectively inhibited the grain growth and avoided abnormal grains in the sintering procedure so that the nanocomposites possessed a fine microstructure. Although the strength of samples decreased with increasing BN content, the strength of samples containing 20 vol% BN remained at 487 MPa for Al2O3/BN and at 838 MPa for 3Y-ZrO2/BN. From another point of view, BN exerted a different influence on the fracture toughness of Si3N4/BN, Al2O3/BN and 3Y-ZrO2/BN. Si3N4/BN is toughened mainly by β-Si3N4 rod-like grains and 3Y-ZrO2/BN by martensitic phase transformation. There is no other toughening mechanism expected for BN in Al2O3/BN. So it is easy to understand that by incorporating nanosized BN the toughness of nanocomposites was marginally increased for the Al2O3/BN system and nearly the same for Si3N4/BN. Toughness of 3Y-ZrO2/ BN nanocomposites decreased with increasing BN content but was higher than that of conventional microcomposites. The decrease in the fracture toughness is due to the spontaneous tetragonal–monoclinic transformation during cooling of the composites from the fabrication temperature and therefore a decrease in the amount of the available transformable tetragonal zirconia present in the micro- and nanocomposites. However, the tetragonal–monoclinic transformation is relatively insensitive to BN content in the 3Y-ZrO2/BN nanocomposites. Although addition of BN no doubt introduces flaws and influences strength of the materials to some degree, the nanocomposites still keep more tetragonal ZrO2. All the above nanocomposites showed improvements of thermal resistance by h-BN addition. In particular, Si3N4/BN has a ∆TC (the temperature difference above which the residual strength decreases suddenly) as high as ~1500°C, 50% higher than that of Si3N4 monolithic ceramics. It is noted that, as in hard nanoparticle dispersed systems, h-BN incorporation also increases the consolidation temperature, and besides, there is a large
© Woodhead Publishing Limited, 2006
248
Ceramic matrix composites
thermal expansion mismatch between the Al2O3 matrix and BN (great contrast difference also exists in the ZrO2/BN system), which causes the tensile stress and cracks at the interfaces between the matrices and the weak BN grain. The sintered body with BN content over 20 vol% contained destructive cracks and hence resulted in a low bulk density and poor strength [17, 18]. Therefore, compounds with a low coefficient of thermal expansion such as Si3N4 and SiC should be added into the system to obtain fully dense bodies.
9.3
Nano–nano type ceramic composites
Nanocrystalline materials form an exciting area of materials research because bulk materials with grain size less than 100 nm have properties not seen in their microcrystalline counterparts. But the brittleness of nanoceramics has limited their potential for use in structural applications, namely, research on nanoceramics shows that they are not inherently tougher than their microcrystalline counterparts. Many strategies have been proposed to improve the mechanical properties of nanoceramics by using reinforcement by a second-phase addition and hybridization to develop nanocrystalline matrix composite materials.
9.3.1
Variety of hardness
Siegel et al. [21] pointed out that the shift in mechanical properties of metals and ceramics occurs in opposite directions as grains become nanocrystalline. At room temperature, the nanocrystalline metals are harder than coarsegrained metals because the dislocation movement is pinned by the grain boundaries. On the other hand, nanoceramics are softer than their micrometer counterparts because of possible grain boundary movement. Hardness of nanoceramics complies with an inverted Hall–Petch model. That is, hardness decreases with a reduced grain size. For example, the average room-temperature hardness value was 4.45 GPa for Al2O3/ZrO2 nanoceramics by HIP [22], which is one quarter of the value of a comparable conventional material. But interestingly, SPSed Al2O3/ZrO2 nanoceramics have a considerably higher hardness, 15.2 GPa [23]. There is still some controversy regarding the hardness of nanoceramics in some cases, i.e., hardness increases with reduced grain size, indicating a positive Hall–Petch relation. Hardness of TiO2 nanoceramics (100°C) reaches 12.75 GPa, much higher than the 1.96 GPa for traditional TiO2 ceramics [24].
9.3.2
Superplastic deformation at low temperature
Since the early report of superplasticity in a ceramic material in 1986, a variety of such materials have been shown to exhibit superplastic behavior
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
249
at high temperatures, usually close to their original sintering temperature. Plastic deformation of the yttrium-stabilized tetragonal zirconia polycrystal (Y-TZP) containing micro- or nano-scaled grain was systematically studied, which is consistent with a grain boundary sliding (GBS) mechanism. Similar studies on fully stabilized cubic ZrO2 (c-ZrO2) single crystals have received considerable attention in recent years, because the slip planes and directions of c-ZrO2 are readily identified. Creep behavior of monoclinic ZrO2 (mZrO2) was also studied. These studies were all done at high temperatures. It is well established that the plastic deformation of crystalline solids occurs by the movement of lattice dislocations and/or diffusional creep. The rate of diffusion is expressed as
ε˙ = Bσ Ω δ Db /( d 3 KT )
(9.1)
According to the equation, the diffusional creep rate of a polycrystal may be enhance by reducing the crystal size d, and by increasing the boundary diffusivity Db. Nanoceramics are therefore expected to exhibit enhanced diffusional creep for two reasons: first, the reduction of the crystal size from about 10 µm to ~10 nm enhances the creep rate by a factor of 109, and second, the enhanced boundary diffusivity may increase the creep rate by ~103, so that the total enhancement is ~1012. Significant plastic deformation will occur by a GBS mechanism if a polycrystalline ceramic is generated with a crystal size of a few nanometers, without having examined whether individual grains had been deformed. Karch and co-workers [2] obtained in 1987 the nanocrystalline CaF2 and TiO2 (8 nm) by applying high pressure in a high vacuum environment after collecting the powder in a mold. They observed that conventionally brittle ceramics became ductile, permitting large (~100%) plastic deformations at a low temperature (80°C for CaF2 and 180°C for TiO2) to follow the shape of a corrugated iron piston under pressure. Based on the fact that the CaF2 cubic structure has many slip systems available, the ductility seems to originate from the diffusional flow of atoms along the intercrystalline interfaces. It is noted, however, that both of the above CaF2 and TiO2 nanoceramics had some amount of porosity. This may account for an apparent soft behavior related to the superplastic deformation at low temperature, which does not yet reveal the plastic deformation characteristics in nanoceramics. Localized superplastic deformation under cyclic tensile fatigue tests was observed by Yan et al. on 3Y-TZP nanoceramics at room temperature [25]. The micromechanism behind this phenomenon is argued to be essentially governed by grain-boundary diffusion. The contribution of dislocation slip might be in operation as a parallel mechanism to develop slip band-like microfeatures. It was shown that, after cyclic tensile fatigue fracture at room temperature in a narrow region within a couple of micrometers of each side of the fractured surfaces, the morphology of grain elongation appeared to be general. Viewed
© Woodhead Publishing Limited, 2006
250
Ceramic matrix composites
on the fractured surface, the ratio of the long and short axes of the grains was in the range 4–5 to 8–10. When the imaging was manipulated to track down a few micrometers on the side surface, it could be seen that the original equiaxed grain morphology remained. It was also observed by AFM imaging that microfeatures such as slip bands were developed on the side face of the 3Y-TZP nanoceramics. For comparison, the microstructure of the surfaces of the specimens with an average grain size of 0.35 µm, after fatigue failure, showed equiaxed grains that had retained their original morphology. It is easy to believe that the grain-boundary diffusion mechanism was the major one to be considered, as described in equation (9.1). The direct effect of fine grain size alone on the rate of deformation can be obtained from the ε˙ ∝ d –3 relationship. Therefore the 100 nm sized 3Y-TZP material should exhibit some 40 times higher rates of deformation in comparison with the 0.35 µm grain-sized ones, under similar conditions.
9.3.3
Superplasticity of Si3N4 /SiC nanoceramics at high temperature
Si3N4 and SiC are promising structural materials for mechanical applications, especially in forming wear-resistant components such as engine parts, because of their excellent mechanical properties such as strength and hardness. Unlike in ionic crystals, plastic deformation of covalent compounds such as Si3N4 and SiC by dislocation glide is difficult because of their high Peierls force. Superplasticity has also been observed in some ionic crystals, such as YZrO2. Wang and Raj [26] pointed out that superplasticity might occur in liquid-phase-sintered Si3N4 by diffusional creep enhanced by solution and precipitation of crystals in a Si–O–N liquid phase at the grain boundaries. High ductility in compression has been observed, because cavitation at grain boundaries under tension and subsequent fracture occurs readily in the presence of an intergranular liquid phase. Wakai et al. [27] reported superplastic deformation of a covalent crystal composite, Si3N4/SiC nano–nano ceramics, which could be elongated by more than 150% at 1600°C. The mechanism of grain-boundary sliding depends on the structure of the grain boundary. A liquid phase present at grain boundaries will enhance sliding. The microstructural changes during deformation were as follows: (1) phase transformation from α-Si3N4 to β-Si3N4 and grain growth of βSi3N4, (2) crystallization of the intergranular liquid phase Si3N4·Y2O3, and (3) volatilization of the liquid phase in the Y–Si–Al–O–N system, causing weight loss. The constitutive equation for steady creep can generally be expressed as
ε˙ = Aσ n / d n
(9.2)
From the relationship between flow stress at a true strain of 0.1 and the strain
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
251
rate, n ≈ 2 was obtained. Models of diffusional creep enhanced by a liquid phase predict n = 1, and thus the superplasticity observed cannot be explained by such models. The superplasticity exhibited by these covalent polycrystals was characterized by non-Newtonian flow. The liquid-phase-enhanced creep model must be modified to accommodate this fact. A stress exponent of n = 2 has been observed in superplastic alloys that do not contain an amorphous phase, and in zirconia polycrystals (Y-TZP) with a very thin amorphous film at two-grain junctions, as well as in their Si3N4/SiC composites. Thus nonNewtonian flow was considered to be a common feature of superplasticity in fine-grained polycrystalline materials. The observed superplasticity was considered to be related to the presence of an intergranular liquid phase. Combined with its hardness, this property suggests several useful applications for the novel material, for example to form engine components. Superplasticity will make it readily moldable at high temperatures. It is suggested that the apparent strain hardening can be attributed to these microstructural changes.
9.3.4
High toughness and its toughening mechanism
Al2O3 /CNT nanoceramics Carbon nanotubes (CNTs) offer a kind of nano-sized reinforcement that is lightweight, has a hollow core, and has immense aspect ratio. Both theoretical and experimental studies showed that carbon nanotubes have exceptionally high mechanical properties such as strength, stiffness and flexibility, as well as electrical and thermal conductivity, i.e. CNTs have a Young’s modulus approaching 1 GPa, and exceptional tensile strength, in the range 20–100 GPa. Especially, single-wall carbon nanotubes (SWCN) have an expected elongation to failure of 20–30%, which combined with the stiffness (Young’s modulus of ~1.5 TPa) predict a tensile strength well above 100 GPa. The flexibility of SWCN is remarkable and the bending may be fully reversible up to a critical angle as large as 110°. Hence, CNTs are considered to be the utmost type of fiber-like reinforcements. The excellent toughness of CNTs should be helpful in solving the inherent brittleness of ceramics. In recent years, attempts have been made to develop advanced engineering materials with improved or novel properties through the incorporation of CNTs in selected matrices of polymers, metals and ceramics. The bridging effect on cracks and the pullout of CNTs from the matrix are possible mechanisms leading to the improvement of the fracture toughness. The contribution to toughness from the nanotube bridging for cracking transverse to the axis of the nanotubes was calculated to be similar to 5 MPa.m1/2. From the CNT bridging law, the CNT strength and interfacial frictional stress were estimated to range from 15 to 25 GPa and from 40 to 200 MPa, respectively [28]. These preliminary results demonstrate that
© Woodhead Publishing Limited, 2006
252
Ceramic matrix composites
nanotube-reinforced ceramics can exhibit the interfacial debonding/sliding and nanotube bridging necessary to induce nanoscale toughening, and suggest the feasibility of engineering residual stresses, nanotube structure, and composite geometry to obtain high-toughness nanocomposites. But actually, the use of CNTs to reinforce ceramic composites has not been very successful, in comparison to conventional ceramic fibers. Real toughness was not observed, or was fairly limited; even in situ or surface modification of CNTs solved their dispersion into the matrix. The fracture toughness for the aluminabased composite containing 10 vol% of the in situ multi-wall carbon nanotubes (MWNT) was 4.2 MPa.m1/2, giving an improvement of only 24% over that of the monolithic alumina [29]. Sun et al. [30] described a simple colloidal processing method to modify the surface of CNTs. The addition of 0.1 wt% CNTs in the alumina composite increases the fracture toughness from 3.7 to 4.9 MPa.m1/2, about a 32% improvement. It is indicated that the toughness of CNT-contained nanocomposites is directly dependent on the variety of CNTs and the sintering process. On the one hand, there are differences in the ability to transfer load from the matrix to the nanotubes between SWCN and MWCN, in addition to the difference of their mechanical properties. The internal shells of MWCN are unable to bond to the alumina matrix, and therefore tensile loads are carried entirely by the external shell. On the other hand, to be effective as reinforcing elements, high-quality, undamaged CNTs must be effectively bonded to the matrix so that they can actually carry the loads. Ceramic composites reinforced by CNTs consolidated by hot-pressing methods require higher temperatures and longer duration. These sintering parameters must damage the CNTs in the composites, leading to a decrease or total loss of reinforcing effects. For example, in composites of CNTs plus metal and ceramic, some of the hotpressing temperatures were as high as 1600°C. With both the Al2O3 and the MgAl2O4 matrices, a fraction of the CNTs seems to be destroyed during the hot-pressing at 1500°C; when using the MgO matrix, most CNTs are destroyed during hot-pressing at 1600°C [31]. It seems that the quantity of CNTs retained in the massive composite is more dependent on the treatment temperature than on the nature of the oxide matrix. CNT damage produces disordered graphene layers which gather at matrix grain junctions, and matrix grains grew up to submicrometer range without producing fully dense nanocomposites. Zhan et al. [32, 33] fabricated fully dense nanocomposites of SWCN with a nanocrystalline alumina matrix at sintering temperatures as low as 1150°C by spark-plasma sintering (SPS). The introduction of SWCN leads to refinement of grain size. Most of the α-Al2O3 grains were in the nanocrystalline range, around 200 nm. The fracture toughness of the Al 2O 3/5.7vol%SWCN nanocomposite is more than twice that of pure alumina and there is almost no decrease in hardness. A toughness of 9.7 MPa.ml/2, nearly three times that
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
253
of pure alumina, was achieved in the Al2O3/10vol%SWCN nanocomposite when sintered under identical conditions. Some interesting features of microstructure can be noted in Al2O3/SWCN nanocomposite. First, ropes of SWCN were distributed along grain boundaries to develop a network microstructure. Some nanotubes were entangled with alumina grains and some encapsulated nanoscale alumina grains. Second, intimate contact between SWCN and alumina was observed in this material, unlike in alumina nanocomposites reinforced by CNTs grown in situ where the cohesion between carbon nanotubes and the matrix was poor and pullouts of carbon nanotubes were observed. Stronger bonding of ropes with the matrix can be seen in the fully dense nanocomposite, suggesting that the extent of interfacial bonding should be a factor in increasing the toughness of the composites. Third, no other forms of carbon, such as graphite, were detected along the grain boundary, indicating that the nanotubes were not damaged during consolidation by SPS. The increase in the quality and quantity of SWCN may have resulted in easier transfer of the stress. It should be noted that the new work showed that the toughness of these nanocomposites can be severely overestimated when measured by the standard indentation method. For dense Al2O3/SWNT composites, Vickers (sharp) and Hertzian (blunt) indentation tests reveal that these composites are highly contact-damage resistant, as shown by the lack of crack formation. However, direct toughness measurements, using the single-edge V-notch beam method, show that these composites are as brittle as dense Al2O3 (having a toughness of 3.22 MPa.m1/2) [34]. This type of unusual mechanical behavior was also observed in SPS-processed, dense Al2O3/graphite composites and other soft dispersoids-contained systems such as Al2O3/BN composites. Al2O3 / ZrO2 nanoceramics Cottom and Mayo [35] claimed that no increase in toughness occurs in ZrO2 nanoceramics having density values close to theoretical unless the materials are heat-treated such that grains become susceptible to phase transformation. Taking the cue from the literature on superplastic metals, it is preferable to have a two phase in one at microstructure of nano size. The Al2O3/ZrO2 binary system was chosen as the candidate system. There are two reasons for this choice. First, there is very little miscibility between the two phases, as per the phase diagram. Secondly, in coarse-grained materials, there is evidence of enhanced stability of the grain boundary structure due to the presence of two phases. Recent studies have indicated that a different mechanism (rather than phase transformation) of toughening must be operative here. The Ferroelectric Domain Switching is responsible for the great increase of toughness in the Al2O3/ZrO2 nanoceramics, which is now well established as a mechanism for enhanced toughness without undergoing transformation in
© Woodhead Publishing Limited, 2006
254
Ceramic matrix composites
ZrO2. Physically speaking, ferroelasticity is similar to ferromagnetism or ferroelectricity. Drawing the analogy, ferroelasticity can be characterized by the existence of permanent strain and an energy dissipating hysteresis loop between the stress and strain axes. In such materials, new domains or twins can be nucleated depending on the state of stress. From the literature on superplastic deformation of metals, it was concluded that nano–nano composites are preferred, in which the constituent phases have similar grain sizes. Indeed, fine-grained (submicron but larger than nanocrystalline) Al2O3/ZrO2 composites have been superplastically deformed. It was believed that these results in coarser than nanocrystalline grained Al2O3/ZrO2 materials are applicable to nanoceramics as well. The average value of HIP toughness in Al2O3/ZrO2 nanoceramics was calculated to be 8.38 MPa.ml/2 [22]. A conventional ceramic material cracks up substantially under such a load. This means that the nano/nano composites were actually deforming plastically under load. Lange reported a toughness value of 6.73 MPa.m1/2 for a conventionally processed material. Furthermore, the conventional material is referred to as zirconia toughened alumina, where the controlled martensitic transformation of metastable tetragonal zirconia to the stable monoclinic phase should lead to phase transformation and microcrack toughening. In the nano/nano case, the toughness measurements were carried out after ensuring that zirconia was not of the type resulting from martensitic transformation by verifying with X-ray diffraction (XRD) after grinding the samples. It was determined that roughly 5% of zirconia was transforming, but about 95% of the zirconia was stable, presumably due to the nanocrystalline size as well as constraint by the alumina grains. Thus, the increase in toughness (compared to 4 MPa.ml/2 for pure alumina) has to be ascribed to the incorporation of nanocrystalline zirconia in alumina, which is also in a nanocrystalline size. It is pointed out that Al2O3/ZrO2 seems to be a versatile system for obtaining tailored properties by designing the microstructure. For example, it is possible to fabricate hybrid microstructures of nano–nano inter-type where transformation toughening may act in synergism with the present mechanism. Another strategy is to heat-treat the nano–nano composite such that the ZrO2 grains grow to such an extent that they transform during the propagation of cracks. In this suggested processing approach, transformation toughening will be a major mechanism and the primary use of these composites will be at temperatures closer to room temperature. Yet another interesting application can be at moderate temperatures (i.e., at somewhat lower than the processing temperatures, e.g. 1100°C) where there may be more ductility due to additional grain boundary sliding. The high temperature limit of application of such materials will be somewhat more than 1200°C, where there may be a ductile to brittle transition due to the growth of nanocrystalline grains to conventional grains.
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
255
A combination of very rapid sintering at a heating rate of 500°C/min and at a sintering temperature as low as 1100°C for 3 min by the spark plasma sintering (SPS) technique, and mechanical milling the starting γ-A12O3 nanopowder via a high-energy ball-milling (HEBM) process, can also result in a fully dense nanocrystalline alumina matrix ceramic nanocomposite. The grain sizes for the matrix and the toughening phase were 96 and 265 nm, respectively. In regard to toughening, a great improvement in fracture toughness (~8.9 MPa.m1/2) was observed in the fully dense nanocomposite. It was nearly three times as tough as the pure nanocrystalline alumina (152 nm, 3.03 MPa.m1/2) [23]. It should be pointed out that the XRD study does not indicate any phase transformation occurring during the 24 h HEBM period even though it is longer than the reported minimal time for the complete transformation (10 h). It is very interesting to note that the width of XRD for high-energy ball-milled γ-Al2O3 nanopowder became much greater than that for the starting nanopowder without HEBM. The residual stress induced by HEBM is likely to be responsible for this wider XRD peak. Moreover, HEBM can lead to high green density due to pore collapse from the high compressive and shear stresses during the milling.
9.4
Fabrication of nanoceramics
Research on processing fully dense bulk nanoceramics and nanocomposites is attracting more and more interest. There have been some low-cost but effective processes to obtain nano-sized ceramic powder and nanocomposite powders, such as sol-gel, micro emulsion, auto ignition, co-deposition and high-energy ball-milling (HEBM). One of the principal problems is the inability to consolidate nanopowders to high relative density without grain growth. To obtain the dense bulk nanoceramics, it is essential to decrease either sintering temperature or retaining time at the highest point, or both HP, HIP, high-pressure sintering and fast consolidation techniques such as microwave sintering and spark plasma sintering (SPS) have been employed. Among these techniques, high-pressure sintering seems to be the best way of obtaining fully dense nanoceramics at the present time. The application of high pressure over 1–8 GPa results in a decrease of the temperature ‘window’ within which fully dense compacts can be obtained without grain growth or with only very limited grain growth. For Al2O3 based nanoceramics, success in achieving such fine grain size can be mainly attributed to two factors. Firstly, lower sintering temperature no doubt leads to lower coarsening. Secondly, γ → α transformation before sintering is quite important. This transformation is known to be nucleation controlled, and formation of a vermicular structure during transformation hampers sintering transformation. Also, the presence of the second phase helps to restrain the grain growth.
© Woodhead Publishing Limited, 2006
256
Ceramic matrix composites
The average grain size of fully dense alumina specimens was less than 100 nm. The average grain size of Al2O3 /Nb was in the range 40–50 nm. This is the finest fully dense alumina-matrix composite reported so far. However, high-pressure sintering is limited to small and simple samples due to the high-pressure requirement [36]. Fast consolidation techniques, such as microwave sintering and plasma activated sintering (PAS), can enhance sintering and reduce the time available for grain growth. The advanced consolidation technique used in the present study to overcome this hurdle is SPS. SPS is a comparatively new technique. It allows very fast heating and cooling rates, very short holding times, and the possibility of obtaining fully dense samples at comparatively low sintering temperatures, typically a few hundred degrees lower than in normal hotpressing. Unlike the first-generation spark sintering and the second-generation PAS, SPS can result in better control of the microstructure and properties of materials in terms of sintering temperature and time. It is a pressure-sintering method based on high-temperature plasma (spark plasma) momentarily generated in the gaps between powder materials by electrical discharge at the beginning of on–off DC pulsing. This energizing method can generate spark plasma, spark impact pressure, joule heating, and an electrical field diffusion effect. In this process, powders are loaded into a graphite die and are heated by passing an electric current through the assembly. These processes have now been developed beyond the production of small objects with simple shapes, as continuous production of compacts of complex geometry and of pieces with diameter larger than 150 mm has been achieved. Despite the fact that a uniaxial pressure is applied, green bodies of complex geometry can be exposed to a ‘pseudo-isostatic’ pressure when embedded in free-flowing electrically conducting particulates that act as a pressure-transmitting medium inside the die. By designing the mold, a temperature gradient along the direct current can be obtained, which is advantageous to simultaneously consolidating components with different sintering temperatures [37]. By optimizing the sintering parameters, various oxide powders were successfully consolidated with nano- or submicron grain sizes, e.g. ZnO, Al2O3, ZrO2, YAG, Si3N4, SiC, Sialon and BaTiO3 [24, 38, 39]. The related grain growth factor is not more than 2. A fully dense Al2O3/3Y-TZP nanoceramic was obtained using SPS when the heating rate was increased to 500°C/min up to 1100°C for 3 min. The mean grain size for the alumina matrix was as small as 96 nm with nearly 100% theoretical density. A ZnO powder with particle size of 25 nm can be consolidated to 98.5% density with a grain size of ~100 nm. The heating rate is from 200°C up to 550°C, holding for one minute, while it is consolidated at a temperature as high as 900°C by microwave sintering.
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
9.5
257
Conclusions and future trends
Materials scientists now have the confidence and skills to conceive and develop ceramic materials with microstructures custom designed as micro– nano systems to obtain and improve the combination of strength, toughness, hardness, high temperature resistance, corrosion resistance and temperature creep resistance according to their application requirement. In situ reaction or coating is the main way to incorporate and homogeneously disperse nanoparticles into the matrix. Further research into enhancing the toughness, serviceability and machinability of nanocomposites is still required. Additionally, multifunctional ceramics incorporating the addition of a nanosized second phase, which integrates both strong mechanical properties and some electrical, magnetic, optical and calorific functions, will attract more interest. New features of ceramics, such as machinability and superplasticity, have been observed for the nano–nano composites. Today’s challenge is how to obtain dense nanoceramics in an effective way at a low cost.
9.6
References
1. Niihara, K., New design concept of structural ceramics: ceramic nanocomposites, J. Ceram. Soc. Japan, 1991, 99(10): 974. 2. Karch, J., Birringer, R., Gleiter, H., Ceramics ductile at low temperature, Nature, 1987, 330: 556. 3. Gao, L., Jin, X.H., Zheng, S., Ceramic Nanocomposites, Beijing: Chemical Engineering Publ., 2004. 4. Suganuma, K., Sasaki G., Fujita, T. et al., Mechanical properties and microstructures of machinable silicon carbide, J. Mater. Sci., 1993, 28(5): 1175. 5. Mizutani, T., Kusunose, T., Sando, M. et al, Fabrication and properties of nano-sized BN-particulate dispersed Sialon ceramics, Ceram. Eng. Sci. Proc., 1997, 18(4B): 669. 6. Hoffman, M., Rodel, J., Sternitzke, M. et al., Fracture toughness and subcritical crack growth in alumina/silicon carbide ‘nanocomposites’, Fracture Mechanics of Ceramics, 1996, 12: 179. 7. Ferroni, L.P., Pezzotti, G., Evidence for bulk residual stress strengthening in Al2O3/ SiC nanocomposites, J. Am. Ceram. Soc., 2002, 85(8): 2033. 8. Zhao, J.H., Stearns, L.C., Harmer M.P. et al., Mechanical behavior of alumina– silicon carbide ‘nanocomposites’, J. Am. Ceram. Soc., 1993, 76: 503. 9. Oh, S.T., Lee, J.S., Sekino, T. et al., Fabrication of Cu dispersed Al2O3 nanocomposites using Al2O3/CuO and Al2O3/Cu-nitrate mixtures, Scripta Mater., 2001, 44(8–9): 2117. 10. Nawa, M., Sekino, T., Niihara, K., Fabrication and mechanical behaviour of Al2O3/ Mo nanocomposites, J. Mater. Sci., 1994, 29(12): 3185. 11. Nawa, M., Yamazaki, K., Sekino, T. et al., A new type of nanocomposite in tetragonal zirconia polycrystal–molybdenum system, Mater. Lett., 1994, 20(5–6): 299. 12. Sekino, T., Niihara, K., Microstructural characteristics and mechanical properties for Al2O3/metal nanocomposites, Nanostruc Mater., 1995, 6(5–8): 663.
© Woodhead Publishing Limited, 2006
258
Ceramic matrix composites
13. Sekino, T., Niihara, K., Fabrication and mechanical properties of fine-tungstendispersed alumina-based composites, J. Mater. Sci., 1997, 32(15): 3943. 14. Ji, Y., Yeomans, J.A., Processing and mechanical properties of Al2O3–5 vol% Cr nanocomposites, J. Eur. Ceram. Soc., 2002, 22(12): 1927. 15. Sekino, T., Nakajima T., Satoru U. et al., Reduction and sintering of a nickeldispersed-alumina composite and its properties, J. Am. Ceram. Soc., 1997, 80(5): 1139. 16. Kusunose, T., Sekino, T., Choa, Y.H. et al., Fabrication and microstructure of silicon nitride/boron nitride nanocomposites, J. Am. Ceram. Soc., 2002, 85(11): 2678. 17. Li, Y.L., Qiao, G.J., Jin, Z.H., Machinable, Al2O3/BN composite ceramics with strong mechanical properties, Mater. Res. Bull., 2002, 38(7): 1401. 18. Li, Y.L., Zhang, J.X., Qiao, G.J. et al., Fabrication and properties of machinable 3YZrO2/BN nanocomposites, Mater. Sci. Eng. A, 2005, 397: 35. 19. Wang, X.D., Qiao, G.J., Jin, Z.H., Fabrication of machinable silicon carbide–boron nitride ceramic nanocomposites, J. Am. Ceram. Soc., 2004, 87(4): 565. 20. Oku, T., Hirano, T., Kuno, M. et al., Synthesis, atomic structures and properties of carbon and boron nitride fullerene materials, Mater. Sci. Eng. B, 2000, 74(1–3): 206. 21. Siegel, R.W., Chang, S.K., Ash, B.J. et al., Mechanical behavior of polymer and ceramic matrix nanocomposites, Scripta Mater., 2001, 44: 2061. 22. Bhaduri, S., Bhaduri, S.B., Enhanced low temperature toughness of Al2O3–ZrO2 nano/nano composites, Nanostruc. Mater., 1997, 8(6): 775. 23. Zhan, G.D., Kuntz, J., Wan, J. et al., A novel processing route to develop a dense nanocrystalline alumina matrix (less than or equal to 100 nm) nanocomposite material, J. Am. Ceram. Soc., 2003, 86(1): 200. 24. Gao, L., Li, W., Nanoceramics, Beijing: Chemical Engineering Publ., 2002. 25. Yan, D.S., Zheng, Y.S., Gao L. et al., Localized superplastic deformation of nanocrystalline 3Y-TZP ceramics under cyclic tensile fatigue at ambient temperature, J. Mater. Sci., 1998, 33(10): 2719. 26. Wang, J.G., Raj, R., Mechanism of superplastic flow in a fine-grained ceramic containing some liquid phase, J. Am. Ceram. Soc., 1984, 67(6): 399. 27. Wakai, F., Kodama, Y., Sakaguchi, S. et al., A superplastic covalent crystal composite, Nature, 1990, 344: 421. 28. Xia, Z., Curtin, W.A., Sheldon, B.W., Fracture toughness of highly ordered carbon nanotube/alumina nanocomposites, J. Eng. Mater. Tech., 2004, 126(3): 238. 29. Chang, S., Doremus, R.H., Ajayan, P.M. et al., Processing and mechanical properties of C-nanotube reinforced alumina composites, Ceram. Eng. Sci. Proc., 2000, 21(3): 653. 30. Sun, J., Gao, L., Li, W., Colloidal processing of carbon nanotube/alumina composites, Chem. Mater., 2002, 14(12): 5169. 31. Flahaut, E., Peigney, A., Laurent, C. et al., Carbon nanotube–metal-oxide nanocomposites: microstructure, electrical conductivity and mechanical properties, Acta Mater., 2000, 48: 3803. 32. Zhan, G.D., Kuntz, J.D., Wan, J. et al., Single-wall carbon nanotubes as attractive toughening agents in alumina-based nanocomposites, Nature Mater, 2003, 2: 38. 33. Zhan, G.D., Kuntz, J.D., Wan, J. et al., Plasticity in nanomaterials, Mat. Res. Soc. Symp. Proc., 2003, 740: 41. 34. Wang, X.T., Padture, N.P., Tanaka, H. et al., Contact-damage-resistant ceramic/ single-wall carbon nanotubes and ceramic/graphite composites, Nature Mater., 2004, 3(8): 539.
© Woodhead Publishing Limited, 2006
Nanophase ceramic composites
259
35. Cottom, B.A., Mayo, M.J., Fracture toughness of nanocrystalline ZrO2–3 mol% Y2O3 determined by Vickers indentation, Scripta Mater., 1996, 34(5): 809. 36. Mishra, R.S., Mukherjee, A.K., Processing of high hardness–high toughness alumina matrix nanocomposites, Mater. Sci. Eng. A, 2001, 301: 97. 37. Tokita, M., Mechanism of spark plasma sintering, Nyu Seramikkusu, 1997, 10: 43. 38. Nygren, M., Shen, Z.J., On the preparation of bio-, nano- and structural ceramics and composites by spark plasma sintering, Solid State Sciences, 2003, 5(1): 125. 39. Shen, Z.J., Zhan, Z., Peng, H. et al., Formation of tough interlocking microstructures in silicon nitride ceramics by dynamic ripening, Nature, 2002, 417: 266.
© Woodhead Publishing Limited, 2006
10 Nanostructured coatings on advanced carbon materials Y M O R I S A D A and Y M I Y A M O T O Osaka University, Japan
10.1
Introduction
Carbon is the most versatile element in the periodic table. Due to various bond structures such as sp3, sp2, sp hybrids, and multiple pπ–pπ bonds, it can form one-, two-, and three-dimensionally bond-structured substances and provide a wide range of applications.1 Carbon materials such as graphite, diamond, activated carbons, carbon fibers, and C–C composites have been extensively investigated and used for many years. Since the discovery of carbon nanotubes in 1997, carbon materials have been newly focused as frontier materials in various fields.2–15 However, carbon materials have a serious shortcoming. They are easily oxidized above 530°C in air. It is possible to protect graphite plates or carbon fibers with SiC coating by CVD or pyrolysis of polymer containing Si and C.16–18 SiC is known as an effective material to prevent oxidation and corrosion due to the strong covalent bond and the passive oxidation by forming a protective SiO2 layer on SiC.19–24 It is difficult, however, to coat fine carbon materials such as carbon nanotubes and fine diamond powders with SiC uniformly. The SiC coating on carbon nanotubes would improve not only the oxidation resistance, but poor adhesion with the matrix when they are used as nano-reinforcements. Many researchers indicate that the improvement of the adhesion between carbon nanotubes and the matrix is a critical issue to improve the mechanical properties of their composites.25–27 The SiC coating is very useful for fine diamond particles as well. Diamond is widely used for cutting, grinding, and polishing of various materials; however, the graphitization of diamond by the reaction with transition metals such as iron, cobalt, and nickel limits its applications. If diamond particles could be coated with an effective protective layer, they could be used at high temperatures under oxidizing and corrosive environments and the tool life could be extended. New composite formation of diamond with WC/Co would be possible. 260 © Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
261
In this chapter, a new and easy process for SiC coating on fine carbon materials is described28–30 and some applications of SiC-coated diamond particles and carbon nanotubes to create new composites are demonstrated.31–33
10.2
Coating method of nanostructured SiC
10.2.1 Coating assembly The SiC coating is processed based on the reaction of SiO vapor and carbon materials. Commercial SiO powders (99.9% pure) are provided as the silicon source. The carbon materials are placed on the SiO powder bed via a carbon felt as illustrated in Fig. 10.1. This assembly is covered with carbon sheets in an alumina crucible to keep the SiO gas pressure in the crucible, and heated in a vacuum furnace at various temperatures from 1150 to 1550°C in vacuum (about 0.03 Pa) for periods of time between 1 and 90 minutes. It is necessary to heat at a temperature greater than 1150°C for the vaporization of solid SiO.
10.2.2 SiC Coating on diamond particles Microstructure Diamond powders with a particle size of 1–30 µ m are used for the SiC coating. Figure 10.2 shows a TEM image of a SiC-coated diamond with a particle size of ~ 1 µm. Each diamond particle is completely covered with a polycrystalline SiC layer ~60 nm thick. The grain size of SiC is several nanometers. Although a large thermal expansion mismatch exists between SiC (α = 4.6 × 10–6/K) and diamond (α = 3.1 × 10–6/K), no crack or debonding Alumina crucible Graphite cover
Carbon felt Carbon materials Carbon sheet
SiO(s)
10.1 Assembly for the SiC coating of carbon materials.
© Woodhead Publishing Limited, 2006
262
Ceramic matrix composites β-SiC layer Diamond
200 nm
10.2 TEM image of the SiC-coated diamond particle.
occurs in the SiC layer or the interface. If there were a gradual change in composition from diamond to SiC at the interface, the thermal stress would be relaxed. Growth mechanism of SiC
0.04
1400
0.03
1000
0.02
600
0.01
200
0
100 Time (min)
Temperature (°C)
Degree of vacuum (Pa)
Figure 10.3 shows the relation between temperature and pressure in the furnace when the assembly of SiO powders, carbon sheets and carbon felts is heated. The increase of the total pressure in the furnace at about 1200°C results from the vaporization of SiO according to reaction (10.1). The evolution of CO gas by the formation of SiC according to reaction (10.2) causes the increase of the total pressure after vaporization of SiO.34–36 In this case, the
200
10.3 Temperature and degree of vacuum in the furnace.
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
263
surface of carbon sheets and carbon felts should react and transform to SiC, producing CO gas. Then, the diamond surface reacts with SiO(g) and forms a thin SiC layer on diamond. This SiC layer will act as a protective layer to limit reaction (10.2) to produce further, thus limiting the evolution of CO(g) from diamond. SiO(s) → SiO(g)
(10.1)
2C(s) + SiO(g) → SiC(s) + CO(g)
(10.2)
The SiO vapor is consumed within 30 min under these treatment conditions. Further treatment over 30 min causes thinning of the SiC layer, probably due to the active oxidation taking place according to reaction (10.3). The partial pressure of oxygen in the furnace is about 6.0 × 10–3 Pa and this value at the coating temperature belongs to the active-oxidation region.37–40 Oxygen is continuously supplied from the outer atmosphere. SiC(s) + O2(g) → SiO(g) + CO(g)
(10.3)
The SiC-coated diamond particles can be characterized by X-ray powder diffractometry. The diffraction peak appears at 35.6° which is assigned as the β-SiC (111) plane. The mechanism of SiC coating on diamond can be analyzed as follows. When the SiC-coated diamond particles are placed in an alumina container and heated to 1200°C in an airflow using a thermogravimetric apparatus, the sample weight decreases with increasing temperature mainly due to the oxidation of diamond, and reaches a minimum at about 1000°C where diamond is almost completely converted to CO2 gas, leaving SiC behind. When it is further heated above 1000°C, the minimum weight increases slightly due to the passive-oxidation of the SiC layer. Because the mass gain due to the silica formation on SiC below 1000°C is negligibly small, we can determine the minimum weight as the initial weight of the SiC layer on diamond. Let us consider a model structure consisting of a diamond sphere that is coated uniformly with SiC. Based on this model, the initial thickness of the SiC layer is expressed using the following equation:
Wf / n = 4π ri2 l ρSiC
(10.4)
where Wf is the minimum weight corresponding to the initial weight of the SiC layer, n is the number of diamond particles, ri is the average radius of a diamond particle, l is the thickness of the SiC layer, and ρSiC is the density of SiC (3.2 g/cm3). The number of diamond particles can be obtained as follows: n = (Wi – Wf)/(4/3)π ri3 ρdia
(10.5)
where Wi is the initial weight of a SiC-coated diamond, and ρdia is the density of diamond (3.5 g/cm3).
© Woodhead Publishing Limited, 2006
264
Ceramic matrix composites
Table 10.1 Thickness and mass gain of the SiC layer on diamond particle Coating time (min)
1 15 30 90
Coated at 1250°C
Coated at 1350°C
Coated at 1450°C
Thickness (nm)
Mass gain*
Thickness (nm)
Mass gain*
Thickness (nm)
Mass gain*
15 21 34 79
0.64 0.89 1.45 3.47
17 29 48 109
0.72 1.24 2.05 4.68
73 82 127 235
3.13 3.52 5.46 10.18
*Mass gain is expressed by (∆W/W0)106; W0 is the initial weight of diamond before SiC coating, and ∆W is the weight gain after coating.
By substituting equation (10.4) into (10.5), the thickness (l) of the SiC layer is calculated using the following equation: l = Wfri ρdia /3(Wi – Wf)ρSiC
(10.6)
Table 10.1 shows the calculated results on the thickness of the SiC layer and the mass gain due to the SiC formation depending on the coating temperature and time. The SiC coating for 90 min was obtained by repeating the 30 min coating three times. The thickness of the SiC layer increases with an increase in the coating time and temperature. The weight of the SiC layer increases linearly with time. This result suggests that the growth of the SiC layer is not controlled by the self-diffusion of Si or C atoms through SiC, but by precipitation or deposition of SiC from the vapor-phase reaction. The following vapor–solid reactions account for the linear growth of the SiC layer with coating time: SiO(g) + 3CO(g) → SiC(s) + 2CO2(g)
(10.7)
CO2(g) + C(s) → 2CO(g)
(10.8)
Based on these analyses on the SiC coating, the growth mechanism of the SiC layer on diamond is considered as follows. In the early stage of the SiC formation on diamond, a very thin SiC layer is formed on the diamond surface according to reaction (10.2) between diamond and SiO(g). Once the SiC layer is formed, this reaction does not proceed due to the protective layer of SiC. The carbon sheet and felt in an alumina crucible act as the carbon source. The reaction of CO2(g) with these carbon sources will produce further CO(g) and deposit SiC(s) by reaction (10.7). Thin β-SiC whiskers are observed on the surface of the SiC-coated diamond, suggesting the vapor growth of SiC. The apparent activation energy of the SiC formation reaction is obtained by an Arrhenius plot of the rate constants that can be calculated using the mass gain data as a function of the coating temperature using the least-
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
265
(b)
(a)
200 nm
200 nm (c)
(d)
200 nm
200 nm
10.4 SEM images of the SiC-coated diamond treated at 1350 °C for (a) 1 min, (b) 15 min, (c) 30 min, and (d) 90 min.
squares method. The calculated value is 100 ± 21 kJ/mol. Shimoo et al. calculated the apparent activation energy for the formation of a SiC layer on a graphite plate based on reaction (10.7) and obtained 97 kJ/mol.36 Both values show excellent agreement. Figure 10.4 shows SEM photographs of the surface of SiC-coated diamond particles coated at 1350°C. Tiny granules of SiC were deposited and aggregated with an increase in coating time. Even for samples treated for 1 min, the entire surface is considered to be covered with a thin SiC layer formed by the direct reaction of diamond and SiO(g) because the samples show good oxidation resistance, to be discussed later. EDX analysis shows a uniform distribution of Si atoms on the entire surface of the SiC-coated diamond particle. Therefore, the SiC layer on diamond is considered to grow in a two-step process as follows: 1. Formation of a very thin SiC layer by the direct reaction between SiO(g) and diamond. 2. Deposition of SiC on a thin SiC layer by reaction (10.7).
10.2.3 SiC Coating on carbon nanotubes Microstructure Multi-walled carbon nanotubes (MWCNTs) are coated with SiC because MWCNTs are more useful as reinforcements and more cost effective than
© Woodhead Publishing Limited, 2006
266
Ceramic matrix composites
SiC 2.57Å β-SiC (111)
3.54Å CNTs (002)
MWCNTs
5 nm
10.5 High-resolution TEM image of the SiC-coated MWCNTs treated at 1350 °C for 15 min with carbon source.
single-walled carbon nanotubes (SWCNTs). The surfaces of MWCNTs are covered with the SiC granules in the same way as the SiC-coated diamond particles. The size of SiC granules is influenced by the coating temperature and time. It is less than 50 nm for the sample treated at 1150°C for 15 min, while it is about 150 nm for the sample treated at 1550°C for 45 min. Therefore, the size of SiC granules can be controlled by adjusting the coating conditions. Figure 10.5 shows a high-resolution TEM photograph at the interface between the SiC coating and MWCNTs, which was treated at 1350°C for 15 min. The (111) plane of β-SiC and the (002) plane of MWCNTs are clearly observed in the image. Some parts of MWCNTs at the vicinity of the interface with β-SiC show an amorphous structure. The measured angle between the (111) plane of β-SiC and the (002) plane of MWCNTs as shown in Fig. 10.5 is 66–71°. This angle matches closely the angle between two different (111) planes of β-SiC (70.5°). These crystallographic relations suggest that the (111) plane of β-SiC is formed epitaxially on the (002) plane of MWCNTs and is grown toward the <111> direction. Growth mechanism Two types of assembly were used for the SiC coating to investigate the growth mechanism of the SiC layer. In the first method, the SiO powders are set on the bottom of an alumina crucible and MWCNTs are placed upon SiO powders via a carbon felt, as shown in Fig. 10.6(a). In the second method, an alumina plate with a center hole is used instead of the carbon felt to separate the MWCNTs from SiO powders, as shown in Fig. 10.6(b). These assemblies
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
267
Alumina lid
MWCNTs Carbon felt SiO(s) (b)
(a) Alumina crucible
Alumina plate
10.6 Assemblies for the SiC coating of MWCNTs (a) with carbon source, and (b) without carbon source.
are covered with an alumina lid to keep the SiO gas pressure inside the crucible, and heated at temperatures between 1250 and 1550°C in a vacuum of about 0.03 Pa for 15 min and 30 min. Figure 10.7(i) and (ii) show the XRD patterns of MWCNTs treated at various temperatures for 15 min. Figure 10.7(i) applies to the assembly of Fig. 10.6(a), and Fig. 10.7(ii) to Fig. 10.6(b). The diffraction peaks of β-SiC appear in all samples. For the samples treated in the assembly of Fig. 10.6(a), : MWCNTs : β-SiC : α-SiC
(i)
(iii)
1550°C
1550°C
1350°C 1250°C (ii) 1550°C
1450°C
Intensity (a. u.)
Intensity (a. u.)
1450°C
1350°C 1250°C (iv)
1450°C 1450°C 1350°C 1350°C 1250°C 20 30 40 50 60 70 80 Diffraction angle (2θ/degree, CuKα)
1250°C 20 30 40 50 60 70 80 Diffraction angle (2θ/degree, CuKα)
10.7 XRD patterns of the SiC-coated MWCNTs prepared (i) with carbon source, and (ii) without carbon source for 15 min, and (iii) with carbon source, and (iv) without carbon source for 30 min, at various coating temperatures.
© Woodhead Publishing Limited, 2006
268
Ceramic matrix composites
the peak of MWCNTs exists at 26.2°(2θ) which arises from the (002) graphite layers. However, the peak intensity is very small for the sample treated at 1450°C in the assembly of Fig. 10.6(b). The peak of MWCNTs disappears when the sample is treated at 1550°C. Above 1450°C, the MWCNTs are converted to SiC in the assembly of Fig. 10.6(b). Figure 10.7(iii) and (iv) show XRD patterns for samples treated at various temperatures for 30 min. Figure 10.7(iii) applies to the assembly of Fig. 10.6(a), and 10.7(iv) to Fig. 10.6(b). Both peaks of β-SiC and MWCNTs exist in the assembly of Fig. 10.6(a). However, the sample volatizes away when treated at 1550°C for 30 min in the assembly of Fig. 10.6(b). SiC is oxidized actively under a low oxygen potential at elevated temperature by a reaction called ‘active oxidation’ following reaction (10.3). In this case, the oxidation occurs continuously and SiC is decomposed to SiO(g) and CO(g). The degree of vacuum in the furnace is ~0.03 Pa. This coating condition at 1550°C belongs to the active-oxidation region as reported by Schneider et al.37 Referring to Fig. 10.6, in assembly (a) the carbon felt would be oxidized by the following reactions: C(s) + O2(g) = CO2(g) C(s) + 1 O 2 (g) = CO(g) 2
(10.9) (10.10)
A reducing atmosphere with very low oxygen content may not lead to a significant active-oxidation reaction for SiC. Since the treatment for 30 min appears to be ineffective for forming a good SiC layer on MWCNTs, the SiC-coated samples prepared for 15 min were utilized to investigate the coating mechanism. There are some granules on the surfaces of MWCNTs coated in assembly (a). On the other hand, the surfaces of the SiC-coated MWCNTs prepared in assembly (b) are smooth. These morphological differences in SiC coatings suggest that the formation process of the SiC layer is different. Reaction (10.7) would proceed when there is a rapid decrease in the partial pressure of CO2(g). Because there is carbon felt in the crucible in assembly (a), CO2(g) is converted to CO(g) by reaction (10.8). The CO(g) generated by this reaction will be supplied for reaction (10.7). In assembly (b), it is difficult to promote reaction (10.8) because no extra carbon source exists in the crucible. In this case, the surfaces of MWCNTs react directly with SiO(g) and convert to SiC by reaction (10.2). From the above results, the growth model of the SiC formation on MWCNTs can be proposed as illustrated in Fig. 10.8. The growth of SiC is influenced by the existence of the carbon source in the crucible. In the early stage of the reaction, SiO(g) reaches the surface of MWCNTs and forms a thin SiC layer
© Woodhead Publishing Limited, 2006
1
2
SiC formed by conversion (reaction 1 )
SiC formed by deposition (reaction 2 , 3)
3
SiC formed by deposition 2 , 3) (reaction
MWCNTs
(a) 1
2
SiC formed by conversion (reaction 1 )
SiC formed by deposition (reaction 2 )
3
SiC formed by deposition 2 ) (reaction
MWCNTs
SiC formed by conversion (reaction 1 ) (1) (2) (3)
SiO(g) + MWCNTs → SiC(s) + CO(g) SiO(g) + 3CO(g) → SiC(s) + 2CO2(g) CO2(g) + C(s) → 2CO(g)
SiC formed by conversion (reaction 1 )
(b)
10.8 Growth models of SiC layer on MWCNTs: (a) with carbon source; (b) without carbon source.
© Woodhead Publishing Limited, 2006
270
Ceramic matrix composites
due to reaction (10.2). This stage needs no carbon source. In assembly (a), SiC granules are deposited on the thin SiC layer by reaction (10.7). This reaction continues until the SiO(s) is consumed because the partial pressure of CO2(g) is decreased by reaction (10.8). The generation of CO(g) by reaction (10.8) would control reaction (10.7). MWCNTs can remain after the SiC coating in assembly (a). In assembly (b), reaction (10.2) is promoted and MWCNTs are converted to SiC. It is inferred that MWCNTs are changed to SiC nanorods by reaction (10.2) because the diffraction peaks of MWCNTs are not observed on XRD patterns of the sample prepared at 1550°C for 15 min. A small amount of nanometer-scale SiC granules are deposited by reaction (10.7) even in assembly (b). In this case, the MWCNTs are considered to play the role of carbon source.
10.2.4 Oxidation resistance SiC-coated diamond particles Table 10.2 shows starting temperatures of oxidation for SiC-coated diamond particles depending on the coating time and temperature. The coating at 1350°C shows superior oxidation resistance. When the total coating time is 90 min (30 min coating repeated three times), the starting temperature of oxidation reaches about 950°C, which is 400°C higher than that of diamond without SiC coating. Even for diamond particles treated for only 1 min at 1350°C, no oxidation starts below 750°C. This result suggests the existence of a thin SiC layer on the entire surface of diamond. The thickness of this SiC layer is estimated to be about 15 nm by interpolating the linear relation between the thickness of the SiC layer and the holding time at 1350°C. The diamond surface is converted to SiC by the reaction-diffusion of Si into diamond. Other coatings at 1250°C and 1450°C show lower oxidation resistance. It is reported that the transformation of diamond to graphite and the generation of cracks in diamond start at over 1400°C.41 The coating at 1250°C for 90 min exhibits no improvement against oxidation compared to the coating at the same temperature for 1 min. Table 10.2 Starting temperature of oxidation for the SiC-coated diamond particles Coating time (min)
Coated at 1250°C (°C)
Coated at 1350°C (°C)
Coated at 1450°C (°C)
1 15 30 90
747 752 753 761
742 798 848 926
794 786 808 831
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
271
The oxidation durability of SiC-coated diamond particles treated at 1350°C for 30 min and 90 min is evaluated at 700°C in an airflow of 50 ml/min. The uncoated diamond particles show a rapid weight loss, whereas the SiCcoated diamond particles treated for 90 min maintain over 70% of their weight after oxidation for 5 h. SiC-coated carbon nanotubes It is easy to form SiC nanorods in the assembly shown in Fig. 10.6(b). The SiC nanorod must show excellent oxidation resistance. However, the microstructure and superior properties of MWCNTs are lost. Therefore, only the SiC-coated MWCNTs treated in the assembly shown in Fig. 10.6(a) for 15 min are evaluated. Figure 10.9 shows the TG curves for SiC-coated MWCNTs which are heated at 650°C in air. As-received MWCNTs are oxidized completely within 5 min. The remaining mass detected (~2.5%) is attributed to iron because it is used as a catalyst for the synthesis of MWCNTs. Further heating increases the mass gain, probably due to the formation of Fe2O3. For the SiC-coated MWCNTs treated at 1550°C, about 90% of mass remains after heating at 650°C for 60 min. The SiC coating at higher temperature provides an improved oxidation resistance. The morphological change of MWCNTs before and after the oxidation at 650°C for 10 s is shown in Fig. 10.10(a) and (b). A tip of MWCNTs before oxidation is closed with a cap. On the other hand, the cap is removed after
100 1550°C 1450°C
Mass loss (%)
80
1350°C 60
1250°C
40
20 as-received 0 0
10
20 30 40 Holding time (min)
50
60
10.9 TG curves for the SiC-coated MWCNTs heated at 650°C in air. The coating was carried out at various temperatures from 1250°C to 1550°C.
© Woodhead Publishing Limited, 2006
272
Ceramic matrix composites
(a)
(b)
1 µm
1 µm (d)
(c)
1 µm
1 µm
10.10 SEM photographs of MWCNTs and SiC-coated MWCNTs: (a) as-received MWCNTs; (b): MWCNTs oxidized at 650°C for 10 s; (c) SiC-coated MWCNTs; (d) SiC-coated MWCNTs oxidized at 650°C for 60 min.
oxidation for 10 s. It is reported that the cap is not resistant to chemical reactions because it has a pentagonal shape.42 When the cap is lost, oxygen can enter the interplanar spaces between [002] planes of MWCNTs. Such planes of carbon atoms are bonded by Van der Waals forces. It is well known that the (002) plane of graphite has a higher oxidation rate than other planes. Therefore, the tip of MWCNTs is very important to prevent oxidation of MWCNTs. For the sample coated at 1350°C for 15 min, the surface and cap are covered with tiny SiC granules (Fig. 10.10(c)). The cap is held and about 70% of the mass remains after oxidation at 650°C for 1 h (Fig. 10.10(d)). A relatively strong (002) peak of MWCNTs remains on the XRD pattern even after oxidation tests. The oxidation rate must be controlled by the diffusion of oxygen through the SiC layer. The crystalline size of SiC can be estimated using Scherrer’s equation for the (111) peak of β-SiC and is plotted as a function of coating temperature in Table 10.3. The crystalline size of SiC gradually increases from 13 nm to 26 nm with an increase in coating temperature. The higher coating temperature can produce dense and thick SiC layers with larger crystalline size, resulting in a higher oxidation resistance. The shape of the SiC-coated MWCNTs is not changed by the oxidation test. No crack or exfoliation is observed on the surface.
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
273
Table 10.3 Relation between coating temperature and apparent crystalline size of SiC Coating temperature (°C)
Crystalline size (nm)
1250 1350 1450 1550
13.0 19.4 21.7 26.0
Note: Coating time is 15 min.
10.3
Applications of nanostructured SiC coatings in advanced composites
10.3.1 Development of SiC-coated diamond/WC–Co composites Cemented carbide, an alloy made of tungsten carbide (WC) and cobalt (Co), is widely used for cutting tools and wear-resistant tools because of its excellent hardness, strength, toughness, and Young’s modulus.43 Although sintered diamond has extremely high hardness and wear resistance,43 it is costly and limited in size and shape because of the need for using ultra-high pressure and difficulty in machining. Therefore, the composite formation of cemented carbide and diamond under lower pressure is very attractive because new wear-resistant tools can be produced at lower cost. However, diamond reacts with cobalt at the sintering temperature of cemented carbide, ~1150°C, and converts to graphite. This reaction must be prevented to develop the diamonddispersed cemented carbide. Fabrication SiC-coated diamond powders with a particle size of ~8–16 µm are mixed with fine WC and cobalt powders (10 wt% Co) and sintered in a vacuum at 1220°C at 30 MPa for 5 min by pulsed-electric current sintering (PECS).44,45 The diamond content is 20 vol%. The PECS method enables sintering of materials at a lower temperature and shorter time than the conventional sintering methods because it uses a high pulsed current of 1000–3000 A. This current is sent through the material directly, after placing the material in a graphite mold, under uniaxial loading. Microstructure Figure 10.11(a) is a SEM micrograph of the polished surface of a SiC-coated diamond dispersed WC-10wt%Co composite. It is well sintered and the
© Woodhead Publishing Limited, 2006
274
Ceramic matrix composites (a)
100 µm (b)
20 µm
10.11 SEM images of the SiC-coated diamond-dispersed cemented carbide composite: (a) polished surface, (b) crack propagation.
diamond particles are uniformly dispersed. The relative density reaches 99.5%. The dense composite suggests that the SiC layer protects the diamond from attack by molten cobalt and prevents the conversion of the diamond surface to graphite. In contrast, when uncoated diamond powders are mixed with WC–Co, well-sintered materials could not be obtained. Mechanical properties Values of Vickers hardness and indentation fracture toughness measured for the WC-Co with and without SiC-coated diamond are compared in Table 10.4. The Vickers hardness is measured under a 98 N load. The fracture toughness is evaluated under the same load using the indentation fracture method. Both sintered composites show almost the same hardness of ~15.5 GPa. However, the fracture toughness of the diamond-dispersed composite was 16.3 MPa.m1/2, which is nearly 200% higher than that of WC–10wt%Co itself. Because of the extremely high hardness (~110 GPa) and high Young’s modulus (~950 GPa) of diamond, the presence of diamond particles is expected
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
275
Table 10.4 Comparison of Vickers hardness and fracture toughness for cemented carbide sintered with and without SiC-coated diamond particles Materials
Vickers hardness (GPa)
Indentation fracture toughness (MPa.m1/2)
WC + 10wt%Co WC + 10wt%Co + 20vol% SiC-coated diamond
15.4 15.5
8.7 16.3
to impede the crack propagation. Such an impeding effect against crack propagation is seen in Fig. 10.11(b). The lower thermal expansion coefficient (~3 × 10–6/K)46 than that of the cemented carbide (5.7 × 10–6/K)47 and the high Young’s modulus of diamond would produce a high tensile stress around each diamond particle. Such a high tensile stress can further enhance the crack deflection.48 No increase in hardness of the cemented carbide composite with SiC-coated diamond dispersion could be attributed to weak bonding between the diamond and cemented carbide matrix. The diamond-dispersed cemented carbides have a wear resistance ten times higher than that of the conventional cemented carbides. Such products are commercialized as super wear-resistant tools in Japan.49
10.3.2 Development of SiC-coated carbon nanotubes/WC–Co composites MWCNTs have been tested to reinforce various matrices because they have many unique mechanical and physical properties.14,15 However, these nanotubes become corroded with metals (such as iron, cobalt, and aluminum) at temperatures above 850°C. These shortcomings limit the applications of MWCNTs as nano-reinforcements. The SiC coating can effectively protect the diamond from molten cobalt, thus allowing dense SiC-coated diamonddispersed cemented carbide composites to be successfully fabricated at lower pressures. If MWCNTs can be coated with the same SiC layer, more stable MWCNTs would be produced and expected to be used as nano-reinforcements for various matrices. The development of SiC-coated MWCNTs/WC-Co composites has potential to extend functions of both MWCNTs and WC–Co. Fabrication SiC-coated or uncoated MWCNTs are dispersed in isopropylalcohol (IPA) using ultrasonic vibration for 5 min. Then the SiC-coated or uncoated MWCNTs are mixed with fine WC and cobalt powders (10 wt% cobalt) in IPA using plastic balls (10–20 mm in diameter) for 5 h. After ball milling, the IPA solution is evaporated by stirring using an electric heater and then dried at
© Woodhead Publishing Limited, 2006
276
Ceramic matrix composites
100°C. The MWCNTs are uniformly dispersed without aggregation. The mixed powders are sintered at different temperatures from 950°C to 1200°C at 30 MPa for 5 min under vacuum using PECS. The content of the SiCcoated MWCNTs is 3 vol%. Sintering behavior The mixed powders of SiC-coated MWCNTs and WC–10wt%Co are charged in a graphite mold and sintered by PECS. The sample is heated from room temperature to sintered temperatures for 20 min. Table 10.5 shows the change of relative density of the WC–10wt%Co and the WC–10wt%Co with MWCNTs compacts depending on the sintering temperature. The density of WC– 10wt%Co increases with an increase in sintering temperature and reaches nearly 100% at 1150°C. On the other hand, the composites of WC–10wt%Co with MWCNTs are fully sintered at 1050°C, which is 100°C lower than the sintering temperature of WC–10wt%Co itself. The resistance heating of MWCNTs is capable of accelerating the sintering process. Microhardness The microhardness of the WC–10wt%Co compact increases by incorporating SiC-coated MWCNTs except for the coating at above 1450°C, as shown in Table 10.6. The microhardness is measured under a 19.6 N load. It is reasonable Table 10.5 Relation between relative density of WC-10wt%Co and sintering temperature Materials
1000°C (%) 1050°C(%)
1100°C(%) 1150°C(%)
WC + 10wt%Co WC + 10wt%Co + 3vol% SiC-coated MWCNTs
69.8 92.6
95.0 100.0
86.3 100.0
100.0 100.0
Table 10.6 Relation between microhardness of the WC-10wt%Co compacts with MWCNTs, SiC-coated MWCNTs, and without MWCNTs, and coating temperature Materials
Coating temperature (°C)
Microhardness (GPa)
WC + 10wt%CO WC + 10wt%Co + MWCNTs WC + 10wt%CO + SiC-coated MWCNTs
— — 1150 1250 1350 1450 1550
17.5 18.9 20.0 20.0 19.9 18.3 18.0
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
277
to assume that the hardness is enhanced probably by the elastic recovery of MWCNTs. The increase of microhardness by incorporating uncoated MWCNTs is lower than that by SiC-coated MWCNTs. This difference of hardness may be due to the corrosion of uncoated MWCNTs with molten Co and poor adhesion with the matrix as suggested by Laurent et al.25 The presence of the SiC coating overcomes these problems because SiC is chemically stable and the SiC granules can provide an anchor effect. The lower hardness obtained when SiC-coated MWCNTs prepared at over 1450°C are used for reinforcements could be attributed to the strength degradation of MWCNTs as a result of high coating temperatures.
10.3.3 Development of SiC-coated carbon nanotubes/SiC composites SiC has high heat and oxidation resistance. Therefore, various applications relating to space developments and efficient power generators are expected. However, the low reliability due to the brittle nature of SiC is a critical problem. MWCNTs may be good candidates to reinforce the SiC matrix if the original strength of MWCNTs is maintained. The SiC coating is expected to improve the weak adhesion between MWCNTs and the SiC matrix. Fabrication The SiC-coated or uncoated MWCNTs are dispersed in isopropyl alcohol (IPA) using ultrasonic vibration for 5 min. They are then mixed with nanometersized SiC (mean diameter 30 nm) and B4C (mean diameter 240 nm) powders in IPA using ultrasonic vibration for 10 min. The B4C is added at 2 wt% as a sintering aid. After the mixing, the IPA solution is evaporated and the mixed powders are dried at 100°C. The MWCNTs are uniformly dispersed without aggregation. The mixed powders are sintered at 1800°C at 40 MPa for 5 min under a vacuum by means of PECS. The content of the SiC-coated MWCNTs is varied between 1 and 5 vol%. Mechanical properties The microhardness of the SiC compact measured under a 19.6 N load increases by incorporating SiC-coated MWCNTs, as shown in Table 10.7. The hardness reaches 30.6 GPa for the content of 5 vol% SiC-coated MWCNTs. This high hardness is considered as an apparent value due to the elastic recovery of the indentation after loading. This interesting phenomenon is discussed later. On the other hand, the increment of hardness by incorporating uncoated MWCNTs is very low compared with the increment by incorporating SiC-coated MWCNTs. This behavior may be due to the poor adhesion with the matrix.
© Woodhead Publishing Limited, 2006
278
Ceramic matrix composites Table 10.7 Relation between microhardness and MWCNT content Content (vol%)
Without coating (GPa)
Coated at 1150°C (GPa)
Coated at 1250°C (GPa)
1 3 5
26.7 25.4 25.8
25.7 29.0 30.5
26.4 27.1 27.4
Note: Microhardness of monolithic SiC is 25.5 GPa.
Table 10.8 Relation between fracture toughness and MWCNT content Content (vol%)
Without coating (MPa.m1/2)
Coated at 1150°C (MPa.m1/2)
Coated at 1250°C (MPa.m1/2)
1 3 5
4.4 4.9 4.6
4.7 4.8 4.9
5.1 5.5 5.4
Note: Fracture toughness of monolithic SiC is 4.8 MPa.m1/2.
Table 10.9 Effect of indentation load on microhardness and fracture toughness Materials
Monolithic SiC MWCNTs (5 vol%)/SiC SiC-coated MWCNTs (5 vol%)/SiC
Microhardness (GPa)
Fracture toughness (MPa.m1/2)
<9.8 N>
<19.6 N>
<9.8 N>
<19.6 N>
26.1 20.5 34.3
25.5 25.8 30.6
4.0 4.4 7.1
4.8 4.6 5.4
The SiC coating acts as an adhesive to the SiC matrix and the SiC granules provide an anchor effect. The relatively lower hardness is obtained when SiC-coated MWCNTs prepared at over 1250°C are used for reinforcements. It is believed that the high coating temperature causes the strength degradation of MWCNTs due to the conversion of MWCNTs to SiC. Table 10.8 shows the values of fracture toughness measured under a 19.6 N load. The toughness increases to 5.4 MPa.m1/2 by the dispersion of SiC-coated MWCNTs, although the data are quite scattered. The mean values of hardness and fracture toughness of the compact measured under 9.8 N and 19.6 N loads are listed in Table 10.9. The results are attributed to the improvement of the adhesion between the MWCNTs/SiC matrix and the SiC coating. The hardness of the monolithic SiC and the uncoated MWCNTs dispersed SiC composite does not change in relation to the indentation load, while that of the SiC-coated MWCNTs/SiC composite increases up to 34.3 GPa when the hardness test is carried out
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials (c)
(b)
(a)
10 µm
279
10 µm
10 µm
10.12 SEM and 3D images of the indentation: (a) monolithic SiC, (b) MWCNTs/SiC composite, (c) SiC-coated MWCNTs/SiC composite.
under a 9.8 N load. The fracture toughness has a tendency to increase depending on the decrease of the indentation load. These results suggest that uncoated MWCNTs do not act as reinforcements for SiC matrix, due to weak interfacial adhesion between the surface of the MWCNTs/SiC matrix. Figure 10.12 shows the SEM and three-dimensional (3D) images of indentation prints marked by the hardness test under a 19.8 N load. The 3D images are composed on the same scale to compare the shape of indentation prints. These images are synthesized with signals of secondary electrons using four detectors in the 3D-SEM equipment. The indentation print of the monolithic SiC ceramic is very sharp, reflecting a square pyramidal shape of Vickers hardness tester. The cracks propagate outward from each corner of the indent. The MWCNTs/SiC composite shows somewhat similar fractography. On the other hand, the indentation print and the crack propagation of the SiC-coated MWCNTs/SiC composite are very indistinct and the square pyramidal print cannot be observed in the 3D image. However, the lateral sides of the indent indicate an elastic deformation. It is difficult to understand why such a high hardness of 34.3 GPa is obtained by incorporating only 5 vol% of MWCNTs into the SiC matrix. If we suppose that the SiC-coated MWCNTs are tough and exhibit an excellent load transformation effect, the elastic recovery of the indentation print would occur and show high hardness, apparently. Microstructure SEM images of the fractured surfaces of a monolithic SiC ceramic, uncoated MWCNTs/SiC, and SiC-coated MWCNTs/SiC are shown in Fig. 10.13. All samples were dense and pore-free. For the uncoated MWCNTs/SiC composite,
© Woodhead Publishing Limited, 2006
280
Ceramic matrix composites (a)
1 µm (b)
1 µm (c)
1 µm
10.13 SEM images of the fractured surface for (a) monolithic SiC, (b) MWCNTs/SiC composite, and (c) SiC-coated MWCNTs/SiC composite.
pullouts of MWCNTs are easily seen compared with those of the SiC-coated MWCNTs/SiC composite. These morphological differences of the fractured surface indicate that SiC layers on MWCNTs should improve the weak adhesion between MWCNTs and the SiC matrix.
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
10.4
281
Conclusions
Nanometer-sized β-SiC granules are coated on diamond particles and MWCNTs uniformly by a new simple method using SiO. The SiC-coated diamond particles and MWCNTs can act as effective reinforcements to produce new high-performance composites. The results can be summarized as follows. • The SiC layer grows in two steps. In the first step, a thin SiC layer is formed by the direct reaction between SiO(g) and diamond or MWCNTs. In the second step, nanometer-sized SiC granules are deposited on the SiC layer by the vapor phase reaction between SiO(g) and CO(g). • The oxidation resistance of diamond particles and MWCNTs is markedly improved by SiC coating. • Dense composites of cemented carbide containing SiC-coated diamond particles can be fabricated without conversion of diamond to graphite. The fracture toughness of the composite is double that of cemented carbide due to the deflection and blocking effects against crack propagation by the dispersed diamond particles. • The microhardness of WC-10wt%Co increases by the dispersion of SiCcoated MWCNTs. The SiC-coated MWCNTs treated at 1150°C to 1350°C can act as nano-reinforcements for WC–Co compacts. • The dispersion of SiC-coated MWCNTs increases the microhardness and fracture toughness of SiC. The SiC coating on MWCNTs at 1150°C is effective in improving the weak adhesion between MWCNTs and the SiC matrix. SiC-coated MWCNT/SiC composites show elastic behavior due to the crack-bridging effect of MWCNTs.
10.5
References
1. Inagaki, M., (1996), ‘Carbon materials’, Solid State Ionics, 86/88, 833–839. 2. Scharff, P., (1998), ‘New carbon materials for research and technology’, Carbon, 36, 481–486. 3. Dresselhaus, M.S., Dresselhaus, G., (1997), ‘Nanotechnology in carbon materials’, Nanostructured Materials, 9, 33–42. 4. Calvert, P., (1999), ‘Nanotube composites – a recipe for strength’, Nature, 399, 210– 211. 5. Treacy, M.M.J., Ebbesen, T.W., Gibson, J.M., (1996), ‘Exceptionally high Young’s modulus observed for individual carbon nanotubes’, Nature, 381, 678–680. 6. Wong, E.W., Sheehan, P.E., Lieber, C.M., (1997), ‘Nanobeam mechanics: elasticity, strength, and toughness of nanorods and Nanotubes’, Science, 277, 1971–1975. 7. Service, R.F., (1998), ‘Materials science: superstrong nanotubes show they are smart, too’, Science, 281, 940–942. 8. Salvetat, J.P., Bonard, J.M., Tomson, H.K., Kulik, A.J., Forro, L., Benoit, W., Zuppiroli, L., (1999), ‘Mechanical properties of carbon nanotubes’, Appl. Phys. A., 69, 255– 260.
© Woodhead Publishing Limited, 2006
282
Ceramic matrix composites
9. Falvo, M.R., Clary, C.J., Taylor, R.M., Chi, V., Brooks, F.P., Washburn, S., Superfine, R., (1997), ‘Bending and buckling of carbon nanotubes under large strain’, Nature, 389, 582–584. 10. Frank, S., Poncharal, P., Wang, Z.L., de Heer, W.A., (1998), ‘Carbon nanotube quantum resistors’, Science, 280, 1744–1746. 11. Batchtold, A., Strunk, C., Salvetant, J.P., Bonard, J.M.,Forro, L., Nussbaumer, T., Schonenberger, C., (1999), ‘Aharonov–Bohm oscillations in carbon nanotubes”, Nature, 397, 673–675. 12. Trans, S.J., Verschueren, A.R.M., Dekker, C., (1998), ‘Room-temperature transistor based on a single carbon nanotube’, Nature, 393, 49–52. 13. Kong, J., Franklin, N.R., Zhou, C.W., Chapline, M.G., Peng, S., Cho, K.J., Dai, H.J., (2000), ‘Nanotube molecular wires as chemical sensors’, Science, 287, 622–625. 14. Kuzumaki, T., Ujiie, O., Ichinose, H., Ito, K., (2000), ‘Mechanical characteristics and preparation of carbon nanotube fiber-reinforced Ti composite’, Adv. Eng. Mater., 2(7), 416–418. 15. Lao, K.T., Hui, D., (2002), ‘Effectiveness of using carbon nanotubes as nanoreinforcements for advanced composite structures’, Carbon, 40, 1605–1606. 16. Fergus, J.W., Worrell, W.L., (1995), ‘Silicon-carbide/boron-containing coatings for the oxidation protection of graphite’, Carbon, 4, 537–543. 17. Hatta, H., Aoki, T., Kogo, Y., Yarii, T., (1999), ‘High-temperature oxidation behavior of SiC-coated carbon fiber-reinforced carbon matrix composites’, Composites: Part A, 30, 515–520. 18. Cheng, L., Xu, Y., Zhang, L., Luan, X., (2002), ‘Oxidation and defect control of CVD SiC coating on three-dimensional C/SiC composites’, Carbon, 40, 2229–2234. 19. Costello, J.A., Tressler, R.E., (1986), ‘Oxidation of silicon carbide crystals and ceramics: I. In dry oxygen’, J. Am. Ceram. Soc., 69, 674–681. 20. Luthra, K.L., (1991), ‘Some new perspectives on oxidation of silicon carbide and silicon nitride’, J. Am. Ceram. Soc., 74, 1095–1103. 21. Jacobson, N.S., (1993), ‘Corrosion of silicon-based ceramics in combustion environments’, J. Am. Ceram. Soc., 76, 3–28. 22. Shimoo, T., Morisada, Y., Okamura, K., (2000), ‘Oxidation behavior of Si-C-O fibers (Nicalon) under oxygen partial pressures from 102 to 105 Pa at 1773 K’, J. Am. Ceram. Soc., 83, 3049–3056. 23. Shimoo, T., Morisada, Y., Okamura, K., (2002), ‘Active-to-passive oxidation transition for polycarbosilane-derived silicon carbide fibers heated in Ar–O2 gas mixtures’, J. Mater. Sci., 37, 1793–1800. 24. Shimoo, T., Morisada, Y., Okamura, K., (2002), ‘Oxidation behavior of Si-M-C-O fibers under wide range of oxygen partial pressures’, J. Mater. Sci., 37, 4361–4368. 25. Laurent, C., Peigney, A., Dumortier, O., Rousset, A., (1998), ‘Carbon nanotubes– Fe–alumina nanocomposites. Part II: Microstructure and mechanical properties of the hot-pressed composites’, J. Eur. Ceram. Soc., 18, 2005–2013. 26. Thostenson, E.T., Ren, Z.F., Chou, T.W., (2001), ‘Advances in the science and technology of carbon nanotubes and their composites: a review’, Comp. Sci. Technol., 61, 1899–1912. 27. Ma, R.Z., Wu, J., Wei, B.Q., Liang, J., Wu, D.H., (1998), ‘Processing and properties of carbon nanotubes–nano-SiC ceramic’, J. Mater. Sci., 33, 5243–5246. 28. Miyamoto, Y., Lin, J., Yamashita, Y., Kashiwagi, T., Yamaguchi, O., Moriguchi, H., Ikegaya, A., (2000), ‘Reactive coating of SiC on diamond particles’, Ceramic Eng. and Sci. Proc. 21, 185–192.
© Woodhead Publishing Limited, 2006
Nanostructured coatings on advanced carbon materials
283
29. Morisada, Y., Moriguchi, H., Tsuduki, K., Ikegaya, A., Miyamoto, Y., (2004), ‘Growth mechanism of nanometer sized SiC and oxidation resistance of SiC-coated diamond particles,’ J. Am. Ceram. Soc., 87, 809–813. 30. Morisada, Y., Moriguchi, H., Tsuduki, K., Ikegaya, A., Miyamoto, Y., (2004), “Oxidation resistance of multiwalled carbon nanotubes coated with silicon carbide’, J. Am. Ceram. Soc., 87, 804–808. 31. Miyamoto, Y., Kashiwagi, T., Hirota, K., Yamaguchi, O., Moriguchi, H. Tsuduki, K., Ikegaya, A., (2003), ‘Fabrication of new cemented carbide containing diamond coated with nanometer-sized SiC particles’, J. Am. Ceram. Soc., 86, 73–76. 32. Morisada, Y., Miyamoto, Y., (2004), ‘SiC-coated carbon nanotubes and their application as reinforcements for cemented carbides’, Mater. Sci. Eng. A, 381, 57–61. 33. Morisada, Y., Takaura, Y., Hirota, K., Yamaguchi, O., Miyamoto, Y., ‘Mechanical properties of SiC composites incorporating SiC-coated multi-walled carbon nanotubes’, J. Am. Ceram. Soc., submitted. 34. Miyata, M., Sawai, Y., Yasutomi, Y., Kanai, T., (1998), ‘Microstructure of Si3N4– SiC ceramics prepared from Si–SiO–C mixed powder’, J. Ceram. Soc. Japan, 106, 815–819. 35. Fujii, K., Nakano, J., Shindo, M., (1995), ‘Evaluation of characteristic properties of a newly developed graphite material with a SiC/C composition gradient’, Proc. 3rd Int. Symp. on Structural and Functionally Gradient Materials, ed. B. Ilschner and N. Cherradi, Lausanne, Switzerland, pp. 541–547. 36. Shimoo, T., Mizutaki, F., Ando, S., Kimura, H., (1988), ‘Mechanism of formation of SiC by reaction of SiO with graphite and CO’, J. Japan Inst. Metals 52, 279–287. 37. Schneider, B., Guette, A., Naslain, R., Cataldi, M., Costecalde, A., 1998), ‘A theoretical and experimental approach to the active-to-passive transition in the oxidation of silicon carbide’, J. Mater. Sci., 33, 535–547. 38. Narushima, T., Goto, T., Iguchi, Y., Hirai, T., (1991), ‘High-temperature active oxidation of chemically vapor-deposited silicon carbide in an Ar–O2 atmosphere’, J. Am. Ceram. Soc., 74, 2583–2586. 39. Shimoo, T., Takeuti, H., Okamura, K., (2001), ‘Thermal stability of polycarbosilanederived silicon carbide fibers under reduced pressures’, J. Am. Ceram. Soc., 84, 566–570. 40. Shimoo, T., Morisada, Y., Okamura, K., (2003), ‘Suppression of active oxidation of polycarbosilane-derived silicon carbide fibers by preoxidation at high oxygen pressure’, J. Am. Ceram. Soc., 86, 838–845. 41. Gargin, B.G., (1982), ‘Thermal destruction of synthesis diamond’, Advanced Materials, 2, 17–20. 42. Saito, Y., Mizushima, R., Hata, K., (2002), ‘Field ion microscopy of multiwall carbon nanotubes: observation of pentagons and cap breakage under high electric field’, Surface Science, 499, 119–123. 43. Suzuki, H., (1986), Cemented Carbide and Sintered Hard Material, Tokyo: Maruzen. 44. Omori, M., (2000), ‘Sintering, consolidation, reaction and crystal growth by the spark plasma system (SPS)’, Mater. Sci. Eng. A, 287, 183–188. 45. Takeuchi, T., Tabuchi, M., Kondoh, I., Tamari, N., Kageyama, H., (2000), ‘Synthesis of dense lead titanate ceramics with submicrometer grains by spark plasma sintering’, J. Am. Ceram. Soc., 83, 541–544. 46. Touloukian, Y.S., Kirby, R.K., Taylor, R.E., Lee, T.Y.R., (1977), in Thermophysical Properties of Matter, Vol. 13, Thermal Expansion. IFI/Plenum, New York, p. 19.
© Woodhead Publishing Limited, 2006
284
Ceramic matrix composites
47. Upadhyaya, G.S., (2001), ‘Materials science of cemented carbides: an overview’, Mater. Des., 22, 483–489. 48. Faber, K.T., Evans, A.G., (1983), ‘Crack deflection process-I. Theory, and II. Experiment’, Acta Metall., 31, 565–584. 49. Moriguchi, H., Tsuzuki, K., Itozaki, H., Ikegaya, A., Hagiwara, K., Takasaki, M., Yanase, Y., Fukuhara, T., (2001), ‘Fabrication and applications of high-toughness, highly wear-resistant diamond-and cBN-dispersed cemented carbide’, Sei Technical Review, 51, 121–125.
© Woodhead Publishing Limited, 2006
11 Processing and microstructural control of metal-reinforced ceramic matrix nanocomposites W D K A P L A N and A A V I S H A I, Technion – Israel Institute of Technology, Israel
11.1
Introduction
While modern ceramics have unique combinations of high strength, hardness, wear and corrosion resistance, and depending on the specific system functional properties, their poor resistance to crack propagation can lead to catastrophic failure under mechanical loading. As a result there has been a rich history of attempts to improve the fracture toughness while maintaining high fracture strengths by forming ceramic matrix composites, reinforced with secondary phases, fibers or whiskers. One of the more recent developments in the field of ceramic matrix composites is the subject of nanocomposites. The term nanocomposites usually refers to a material in which sub-micron secondary phase particles are dispersed in a polycrystalline matrix of micron grain size. The initial focus on ceramic matrix nanocomposites was based on polycrystalline α-Al2O3 reinforced with SiC nano-particles, which according to some researchers showed large increases in fracture strength (and some improvement in toughness) as compared to monolithic alumina.1 A separate series of nanocomposites are those based on a polycrystalline ceramic matrix, but which contain metallic nano-particles dispersed throughout the composite (a type of ceramic–metal composite, or cermet). This type of nanocomposite was first produced by the simple mixing of two oxide powders, for example α-Al2O3 and sub-micron sized NiO, followed by sintering in a reducing atmosphere to produce sub-micron Ni particles within an alumina matrix.2 Again, cermet nanocomposites were reported to have extremely high values of fracture strength, and additional functional (i.e. magnetic) properties.3 It is the goal of this chapter to explore various processing methods for the production of cermet nanocomposites, and to discuss in detail variations in the microstructure which can significantly influence the final properties.
11.2
Processing
The main goal of most ceramic sintering routes is to obtain near-theoretical 285 © Woodhead Publishing Limited, 2006
286
Ceramic matrix composites
density while controlling grain growth. This is a difficult task, since both processes depend on thermally activated mass transport mechanisms which often occur simultaneously.4,5 For several monolithic ceramic systems, dopants have been found which ‘enhance’ sintering, limit pore detachment from grain boundaries (pore occlusion), and limit grain growth, although the mechanism by which such beneficial dopants work is still not clear. In the case of α-Al2O3, MgO is the dopant of choice. Magnesium additions apparently increase the solubility limit of detrimental impurities, such as Si and Ca, which form glassy phases and amorphous intergranular films,6 and/or change the manner in which glassy phases affect grain growth.7,8 Abnormal or exaggerated grain growth is thus prevented, and a finer microstructure is obtainable after sintering to full density. The introduction of secondary phase particles, especially metallic particles, adds a new parameter to control the evolution of the sintered microstructure, and to obtain new or refined properties. Important microstructural parameters include density, particle size, shape, location and content, as well as matrix grain size which is influenced by the presence of the particles. At the same time the introduction of metallic particles necessarily complicates the processing route, including difficulties in slip processing, reaching homogeneous distributions of particles in the microstructure, and very real health hazards. In the sections below, various processing routes are described, including a discussion of these important issues.
11.2.1 Simple powder mixing The simplest concept for the preparation of cermet nanocomposites would be to use conventional mixing or milling processes, in a method analogous to that used to prepare ceramic nanocomposites, such as Al2O3 reinforced with SiC. The obvious problems in such a process would be the oxidation of the metal particles during the powder processing, prior to sintering, compounded with the need to deflocculate the particles. While nanometer-sized metallic particles can be produced, the high surface area means there is a very real danger of explosive oxidation. This effect, compounded with the relatively high cost of such starting materials, limits this process to laboratory-scale experiments. Even laboratory-scale processes of such sorts are restricted, due to the requirement to break up soft agglomerates, which usually requires the additions of deflocculants. As the particle size decreases and the total metal surface area increases, the amount of deflocculant necessarily increases, resulting in significant amounts of long-chain molecules in the green body, which must be removed prior to or during sintering. Since the deflocculation stage is very different for metallic particles compared to the ceramic matrix particles, there has been limited advance in this direction. Having said this, there have been experiments to produce ceramic matrix
© Woodhead Publishing Limited, 2006
Processing and microstructural control
287
composites reinforced with micron-sized metallic powders, prepared by simple powder mixing, followed by controlled atmosphere sintering. An example is the work by Tai et al., describing alumina reinforced with nanometer-sized Co, prepared by simple agate milling in a mixture of methanol and ethylene glycol, followed by hot-pressing.3 Hot-pressing is required to overcome the relatively low green densities. In this study it was found that Co additions from 30–50 wt% resulted in significant particle coalescence, reaching particle sizes of ~900 nm, while Co additions of up to 10 wt% resulted in Co particles of 500 nm or less (the starting Co particle size had an average of ~30 nm). This work nicely demonstrates the main problem in processing metal-reinforced ceramic matrix nanocomposites, i.e. reaching full density while retaining the nanometer-sized length scale of the reinforcing phase. This effect is amplified when using metals with a relatively low melting point temperature, such as copper.
11.2.2 Oxide reduction An alternative processing method would be to utilize conventional powder processing of nanometer-sized oxide particles, combined with the matrix phase.2,9,10 The particles are then reduced to the metallic state during sintering, or immediately prior to sintering by using an appropriate atmosphere. Since both phases are initially in the oxide state, conventional deflocculation techniques can be combined with conventional powder processing to produce green bodies with high densities and a homogeneous distribution of the reinforcing phase. After production of the green bodies a reduction stage is required to reduce the oxide particles to the metallic state. The accepted approach has been to sinter under reducing conditions such that the reinforcing phase reaches the metallic state during the initial stages of sintering, as the open pores in the ceramic matrix are closing. Once the ceramic matrix has reached the point that only closed pores exist, coarsening of the reinforcing metallic phase is limited by grain boundary diffusion kinetics through the ceramic matrix, which is slower than surface diffusion. This of course will not prevent coalescence and coarsening of particles during the course of matrix grain growth. Naturally the final size of the metallic reinforcing particles is limited by the initial size of the oxide particles. This is rather important, since there is a critical maximum particle size which can lead to degradation of the composite properties via thermal stress-induced cracking.2 The critical maximum particle size was evaluated by Kolhe et al.,11 both experimentally and via finite element analysis. Assuming perfectly spherical particles, the difference in thermal expansion coefficients between Ni and α-Al2O3 resulted in a critical particle size for an isolated Ni particle of 3.0 µm. Experimental observations
© Woodhead Publishing Limited, 2006
288
Ceramic matrix composites
for the same system showed cracking for (non-spherical shaped) particles with a diameter of 750 nm. Thus the oxide reduction process to form metalreinforced ceramic matrix nanocomposites still requires sub-micron or nanometer-sized oxide particles, encompassing health problems and relatively high costs for the raw materials. A relatively simple method to overcome the need for expensive and potentially dangerous nanometer-sized oxide particles is to chemically deposit the oxide particles during the powder processing stage.12–14 This can be accomplished quite easily by adding to water-based slips nitrates (i.e. nickel nitrate), which is calcined to the oxide state after drying. Such processes usually result in very fine oxide particle size distributions, which are reduced either prior to or during the final sintering process. While an extra sieving and milling stage is required after calcining, water-soluble metal salts are usually inexpensive, making the process more commercially feasible.
11.2.3 Sol-gel and gel-casting Another option increasingly being encountered in ceramic processing is the use of sol-gels. The use of sol-gels for processing of metal–ceramic composites introduces a wide range of new possibilities, including the preparation of complex shapes by gel-casting,15 and the option to obtain unique functional properties (electrical, optical, magnetic) by co-precipitation.16 This method offers a number of advantages, including the possibility for low temperature processing, better control over homogeneity and particle dispersion, and relatively low cost. However, this process suffers from rather low final densities, and if high-temperature sintering is involved then this usually results in coarsening of the microstructure, resulting in many cases with limited advantages over other processing methods. Sol-gel processing involves the use of a hydrolysis reaction to obtain a cross-linked network, which results in the formation of a gel. When preparing metal-ceramic composites, both components may be obtained in this way, or alternatively the metal reinforcement can be introduced by adding, for example, metal nitrates.17,18 The gel properties may be controlled by adjusting the pH level, water to metal ratio, and temperature. The possibility to obtain a uniformly dispersed composite powder was shown for the α-Fe–Al2O3 system where metal particles with an average size of 55 nm were formed in an amorphous/nano alumina matrix.18 Other studies attempting to obtain dense bulk composites based on the sol-gel route using conventional pressure-assisted sintering (~1400°C and an applied force of 10 MPa) resulted in a coarse microstructure.16 However, if reaching theoretical density is not a necessary requirement, a porous ceramic microstructure containing nanometer-sized metal particles can be used as a catalytic material.19 Certain combinations of composite materials demand
© Woodhead Publishing Limited, 2006
Processing and microstructural control
289
special attention during the annealing treatments with regard to temperature and atmosphere to avoid reduction of both species or even total loss of one of the components.20 Gel-casting Niihara et al.15 applied gel-casting to nanocomposite processing. Using a mixture of oxide powders prepared from an alumina powder and Ni nitrate described above, together with methacrylamide as the monomer and N, N′methylenebisacrylamide as the cross-linker, they obtained a viable gel, which was cast and sintered to a final density of ~99%. Due to the complex object shape, pressureless sintering was used. Niihara et al. reported a fracture strength of ~590 ± 50 MPa for these samples, which is slightly higher than that of monolithic alumina. This process has the advantage of obtaining a near net-shaped object with complicated geometries while avoiding the need for costly machining of a hard composite material.
11.2.4 Salt infiltration A simple variant of the various methods described above is based on metal salt infiltration into porous ceramic preforms, followed by reduction and sintering under controlled atmosphere. This method skips the more complicated stages of calcining, secondary milling, and sieving. The process begins with conventional ceramic powder processing to reach a green body. Partial sintering (firing) is used to induce necking at the particle contact points, which results in a minimal level of mechanical strength, required for subsequent handling. The fired body is then infiltrated with metal salts in a water-based solution. If the contact angle of the salt solution is low enough (nominally under ~50° depending on the geometry of the particles21), spontaneous infiltration is possible, although infiltration under vacuum is usually required to ensure complete penetration. After drying, the preform is heated under a reducing atmosphere to form metallic particles, and sintered to full density.22 While this process is extremely simple, and various sintering modes are possible, the amount of metal salt which can be introduced in a homogeneous manner is dependent on the solubility limit in the liquid medium. Nitrates are extremely soluble in water, thus water-based solutions are convenient for a number of different metallic salts. However, too high a concentration can result in large salt particles after drying, leading to micron-sized metal particles. Thus the preferred processing method is to conduct multiple infiltration– drying–reduction stages, where each stage adds ~2 wt% of metal particles to the fired body.23 In this way various particle concentrations are possible, while the small (nanometer or sub-micron) particle size is maintained.
© Woodhead Publishing Limited, 2006
290
Ceramic matrix composites
Furthermore, different salts can be used in the same process, allowing for the production of alloyed particles and/or functionally graded nanocomposites.
11.3
Microstructure
Since the very nature of nanocomposites depends on well-defined processing routes to achieve a specific microstructure, detailed characterization of the microstructural features is extremely important. This is particularly critical in identifying the role of the microstructure in defining the final bulk properties. The microstructural features in nanocomposites which have been linked to bulk properties include the matrix grain size, the reinforcing particle size, its distribution and location (grain boundaries or occluded within the matrix grains), segregation at the various interfaces, and residual stress fields. Figure 11.1 schematically illustrates the microstructural parameters most important to the processing and properties of metal-reinforced nanocomposites. There are two main characterization issues which must be addressed regarding these materials. First, it is clear that conventional methods often do not provide the relevant information required to understand the Chem
MTriple PTriple
MGB
Chem
Γ/IGF Γ/IGF
PGB POccluded
MOccluded
D
11.1 Schematic drawing illustrating the important microstructural parameters for processing and properties of metal-reinforced nanocomposites. These include occluded pores (POccluded) and metal particles (MOccluded), particles at grain boundaries (PGB) and triple junctions (PTriple), segregation at grain boundaries/interfaces and intergranular films at grain boundaries/interfaces (Γ/IGF), as well as bulk chemistry for both the ceramic matrix and reinforcing particles (Chem).
© Woodhead Publishing Limited, 2006
Processing and microstructural control
291
2µm
11.2 A conventional SEM (backscattered electrons) micrograph of a Ni–Al2O3 nanocomposite demonstrating problematic use of the BSE signal for microstructure characterization.
Cu
500 nm
11.3 SEM (secondary electrons) micrograph of a fracture surface from a Cu–Al2O3 nanocomposite. The Cu particles occupy intergranular positions and the fracture observed is mostly intergranular.
microstructures. For example, conventional backscattered electron (BSE) images in scanning electron microscopy (SEM) simply do not have the resolution required to measure nanometer (even tens of nanometers) particle size, for metallic particles in a non-conducting matrix. This is demonstrated in Fig. 11.2. Characterization of fracture surfaces by low-voltage SEM yields (in some cases) important morphological information regarding the location of the reinforcing particles (triple junctions and grain boundaries) and the mode of fracture (intergranular or transgranular). Figure 11.3 shows an example
© Woodhead Publishing Limited, 2006
292
Ceramic matrix composites
of a fracture surface of a Cu–Al2O3 nanocomposite. In this case the location of the metal particles could not be revealed by thermal or chemical etching. As a result of these limitations more indirect characterization methods, such as crystal size analysis from X-ray diffraction (XRD), or methods with limited statistical meaning but higher spatial resolution, such as transmission electron microscopy (TEM), must be applied. Secondly, given the enormous amount of interfacial area in nanocomposites, understanding the atomistic structure, chemistry, and energetics (adhesion) at the metal–ceramic interfaces is a fundamental issue for nanocomposites, which to a large part has been ignored. In the following we attempt to define some of the critical microstructural parameters and how they can be addressed.
11.3.1 Evolution and control While the particle size depends initially on the method used to introduce the particles, subsequent thermal cycles (i.e. sintering to obtain a dense composite) can lead to significant particle coarsening. This is especially so for metal particles in a ceramic matrix, for which the sintering temperature can be close to, if not above, the melting point of the particles. In general, when a ceramic preform is sintered, the open porosity closes, and then during the final sintering stages the remaining closed pores are removed by vacancy diffusion through the grain boundaries. As such, the critical stage for particle coarsening will be when open pores exist in the preform. Particle coarsening during the final stages of sintering requires metal (cation) diffusion through the ceramic grain boundaries, which can occur but will be significantly slower than surface diffusion in open pores. While particle size is usually considered an important factor in defining the bulk properties of the nanocomposite, particle shape can be equally important, especially for mechanical properties. Normally nanometer-sized metal particles will be equiaxed, with a certain degree of faceting dictated by the degree of surface energy anisotropy. However, since the particles are confined by interfaces with the ceramic phase, the shape will depend on the interface energy anisotropy, which is usually not known a priori, but can be defined with a modified Wulff construction.24 For FCC metals with a limited degree of surface anisotropy in an alumina matrix, the interface planes are usually dictated by the facet planes of the alumina matrix.25 Similar observations have been made regarding SiC particles in an alumina matrix.26 However, the interface energy will be affected by segregation effects, including gas species in the processing atmosphere. This can have a significant influence on the shape of metal particles in a ceramic matrix, which has been demonstrated for Cu particles in alumina due to adsorption of oxygen to the interface (see Fig. 11.4).27,28 There are several reports in the literature stating that matrix grain size
© Woodhead Publishing Limited, 2006
Processing and microstructural control
293
Cu
500 nm
11.4 TEM micrograph of a Cu–Al2O3 nanocomposite. The Cu particles are elongated and located at grain boundaries and triple junctions.
refinement, due to the existence of nanometer-sized particles at the grain boundaries, is the main reason for improved mechanical strength.14 While the exact mechanism for increased strength in bending is probably more complicated, depending upon a combination of microstructural features, there is no doubt that control of the matrix grain size by particle pinning is a possible advantage to nanocomposites in general. This is actually just a new label to a microstructural refinement mechanism which has been employed by both metallurgists and ceramicists for many years. Grain growth kinetics depend on the driving force for grain boundary movement, as well as the mobility of grain boundaries. During sintering (or any high-temperature annealing) the driving force for grain growth is the simple reduction of grain boundary energy (or area).4,5 Grain boundary mobility can be separated into intrinsic and extrinsic mobility. Intrinsic mobility will depend on the exact mechanism for atom transfer from one ‘side’ of the boundary to the other. Extrinsic mobility depends upon a variety of factors, which include solute drag due to grain boundaries (this is a result of segregation moving with the boundary), impurity (dopant) segregation, mobility due to the presence of a liquid phase at the boundary which increases local diffusion rates, and grain boundary pinning by secondary phases (known as Zener drag).29 Grain boundary pinning by secondary particles is naturally a point of interest for processing nanocomposites. Grain boundary pinning can be described as a drag force on the grain boundary, which negates the driving force for grain growth, and was elegantly described by Hsueh et al.30 via the dihedral angle ψ formed at the triple junction of a particle with radius r at a grain boundary with energy γGB: F = π r γGB(17.9 – 6.2ψ)
(11.1)
Assuming the particle remains at the grain boundary, and is not occluded by the growing crystal, the larger the particle and the smaller the dihedral angle, the larger the drag force negating grain growth. For nanocomposites it is the accumulated force of many small particles acting at each grain boundary
© Woodhead Publishing Limited, 2006
294
Ceramic matrix composites
which defines the drag force, compensating for the small particle radius. The dihedral angle is a measure of the interface energy, thus elongated particles (as shown in Fig. 11.4) result in a high drag force and more effectively limit matrix grain growth.
11.3.2 Interfaces Due to small particle size, the amount of metal–ceramic interface area in nanocomposites can be quite significant, even for metal particle additions of less than 10 wt%. As such, the nature of the metal–ceramic interfaces can influence the bulk properties to extents not normally achieved in conventional metal–ceramic composites. Relevant parameters include the interface energy, which defines the thermodynamic work of adhesion and has a significant and non-linear connection with interface toughness.31 The interface energy will in turn be influenced by segregation of constituent atoms and impurities/ dopants, including elements in the processing atmosphere. Basic studies of metal–ceramic interface thermodynamics have clearly demonstrated the importance of segregation on adhesion. An example pertinent to Ni-reinforced alumina matrix nanocomposites is the influence of Al segregation on the Ni–alumina interface energy, where additions of only a few atomic percent of Al to Ni can significantly increase the thermodynamic work of adhesion.32,33 Pre-doping the Ni particles is not necessary, since some dissolution of alumina is expected as the particle and alumina grains adjust their shape to reduce the total interface area through the formation of lenticular-shaped particles at grain boundaries.34 While this is an important issue in basic surface or interface science, it has yet to be investigated or applied to nanocomposites. One interesting example of possible modification of interfaces in metalreinforced nanocomposites is the presence of intergranular films. High resolution transmission electron microscopy (HRTEM) of grain boundaries and interfaces in such materials revealed the existence of thin, apparently amorphous films, with a constant thickness in the range of 1–2 nm.25 The existence of thin intergranular films at grain boundaries in ceramics is a well-established phenomenon, and has been characterized in detail at grain boundaries in alumina, silicon nitride, and silicon carbide.7,35,36 The formation of the film is usually a result of sintering additives or impurities, which form amorphous phases commonly found at triple junctions. Clarke and co-workers developed a model to calculate the thickness of the amorphous film observed in polycrystalline ceramics.37,38 The model is based on a force balance between an attractive van der Waals dispersion force that acts across the grain boundaries, any capillary forces present, and repulsive disjoining forces (such as steric forces and electrical double-layer forces) in the amorphous film.37,38 The repulsive steric force is based on the
© Woodhead Publishing Limited, 2006
Processing and microstructural control
295
assumption that the first monolayer of molecules of the glassy phase exhibits a heteroepitaxial arrangement with both grain surfaces and that this ordering extends over a certain distance within the amorphous film. The electrical double-layer forces occur if the interface between the grains and the film is charged by cations. While the van der Waals and capillary forces act to bring the grains closer together, the disjoining forces widen the film. The estimated values of the amorphous film thickness, which is dependent on the dielectric properties of the film and the grains, are in excellent agreement with the experimentally observed film width.37,39,40 Due to the balance of forces leading to a defined film thickness, such films are often termed equilibrium amorphous films. Recently, it was shown that equilibrium amorphous films also exist at interfaces between metals (Ni and Cu) and alumina, via model experiments based on nanocomposites. Cu–Al2O3 and Ni–Al2O3 nanocomposites were prepared using high-purity alumina powder to which predetermined amounts of Ca and Si dopants were added.25 Detailed HRTEM characterization of the specimens showed that all metal– ceramic interfaces in the two different nanocomposites had thin (~1 nm thick) amorphous films (see Fig. 11.5). In addition, occluded particles were found inside the alumina grains which also had thin amorphous films at their interfaces with alumina. Analytical microscopy showed the films to contain Ca, Si, and Al.41 Hamaker coefficients were calculated for metal–ceramic interfaces in the presence of a SiO2-based film, which indicated that a stronger attractive force is expected for intergranular films at metal–alumina interfaces, Ni
(111
)
(0001)
5 nm Al2O3
11.5 HRTEM micrograph of a Ni–Al2O3 interface taken from a Ni– Al2O3 glass-doped nanocomposite. A thin (~1 nm) amorphous film exists at the metal–ceramic interface, extending from a glass-pocket at the triple grain boundary junction.
© Woodhead Publishing Limited, 2006
296
Ceramic matrix composites
relative to alumina grain boundaries, correlating to the experimentally measured film thickness.40 The ability to modify the metal–ceramic interface in nanocomposites by the formation of intergranular films holds exciting prospects. From a thermodynamic point of view, the existence of a film at equilibrium indicates a lower interface energy than an interface without a film. This indicates the potential to increase the adhesion of interfaces, although experimental investigations are required to fully evaluate this effect. However, the promotion of particle occlusion due to the presence of the films has been shown,28 and this means that a new method to modify and control the microstructural evolution of nanocomposites is available, as discussed in the next section.
11.3.3 Particle occlusion As mentioned in Section 11.3.1, the particles in nanocomposites induce a drag force on the matrix grain boundaries, which in some cases can reduce grain boundary migration rates, resulting in a finer matrix microstructure. However, if the grain boundary migration rate is faster than that of the particles, particle detachment can occur, resulting in the particle being occluded within the growing matrix grain. Detachment (or occlusion) depends on the relative interface mobility for a given driving force. In the nanocomposites mentioned in the previous section, occluded metal particles were not found in nanocomposites without the glassy phase.28 However, when the nanocomposites were doped with glass-forming elements (Ca and Si), intergranular films were detected and a significant amount of metal particles was found to be occluded within the alumina grains (see Fig. 11.6). The occluded particles in the Cu–alumina and Ni–alumina composites had an average size of 250 ± 20 nm and 260 ± 90 nm respectively, while the particles located at alumina grain boundaries had an average size of 1400 ± 300 nm and 850 ± 350 nm, respectively. These measurements were performed using TEM in order to differentiate between particles found at alumina grain boundaries from the occluded particles. A similar type of behavior is seen in other studies. In composites prepared by Oh et al. using high-purity alumina, only a few occluded particles were observed.14,42 However, in a study by Chen and Tuan,43 a large number of Ni particles were reported to be occluded (~20–30%). Chen and Tuan observed a similar difference in the distribution of the reinforcing particle size between the occluded particles (limited to ~100 nm) and the particles found at grain boundaries and triple junctions. Based on the findings from the samples purposely doped with glass and containing intergranular films, resulting in 70–80% of the particles occluded in the alumina grains, the question is raised whether the samples prepared by Chen and Tuan43 contained small amounts of glass-forming impurities.
© Woodhead Publishing Limited, 2006
Processing and microstructural control
297
Occluded Ni
Occluded pores 500 nm
11.6 Bright field TEM micrograph showing the metal particle morphology in a Cu–Al2O3 glass-doped nanocomposite.
Two important points should be noted regarding metal particle occlusion: the fact that glass doping promotes particle occlusion, and the distinct size distribution of the occluded particles relative to particles found at grain boundaries. Unless a special orientation relationship exists between the matrix and an occluded particle, the occlusion process should be energetically unfavorable, since it will increase the surface energy of the system. The occlusion of particles (and pores) is therefore a kinetic result. For the particle (or pore) to stay attached to a grain boundary, its velocity should be the same as that of the grain boundary.44 The velocity (Vi) may be described as the product of two parameters – the driving force (Fi) and the mobility (Mi): Vi = Fi · M i
(11.2)
The driving force for grain growth (Fb) is the reduction in the total internal surface (grain boundary) energy. This can be expressed by the grain boundary energy (γGB) and average grain size ( G ):44 Fb ≈
3γ GB G
(11.3)
For a given grain boundary energy, the smaller the grain size the larger the driving force for grain growth. Due to the fact that a stable intergranular film forms during the initial stages of sintering, it may be assumed to a first approximation that the mobility would not change drastically during the sintering process. This would mean that the velocity of the grain boundaries will decrease proportional to 1/ G during the sintering. The drag force for a particle (or pore) was defined in equation (11.1). The
© Woodhead Publishing Limited, 2006
298
Ceramic matrix composites
mobility (Mp) of a particle/pore can be expressed in a similar way if the particle is assumed to have a dominant interfacial diffusion mechanism:29,44 M p(s) =
Ds δ s Ω Dδ Ω , M p (I) = I I 4 4 kT πr kT πr
(11.4)
where Ds and DI are the surface and interface diffusion coefficients respectively, δs is the boundary core width, Ω is the total ionic volume divided by the number of slow diffusing ions, k is Boltzmann’s constant, and T is the absolute temperature. For metal-ceramic interfaces, diffusion tends to be faster than for ceramic surfaces.45 However, in some cases volume diffusion dominates over interfacial diffusion, as was shown by Saiz et al.45 and Monchoux and Rabkin.46 In this case mobility is a function of 1/r3:29,46
M p(V) =
3DV c V Ω 4kT πr 3
(11.5)
where cV is the solubility of the diffusing atoms in the particle. Whichever diffusion mechanism is dominant, the velocity of small particles is expected to be higher than that of larger ones, since the velocity will depend on 1/r3 for surface/interface diffusion while for volume diffusion it will depend on 1/r2. This means the observed occluded particle size distribution (i.e. small particles), and larger particles found at the alumina grain boundaries, cannot be controlled by particle velocity dependence on particle size. Both the matrix grains and metal particles coarsen during sintering. During the initial stages of sintering, due to the presence of the thin liquid film at the alumina grain boundaries, the mobility of the boundaries is increased, and at the same time the driving force for grain growth is high since the alumina starts from a fine 0.3 µm size powder. This results in a high velocity of the grain boundaries, and relatively small occluded particle size. As the sintering process advances, the metal particles coarsen, which reduces their velocity and should result in their occlusion. However, simultaneous grain growth of the alumina grains reduces the driving force for grain growth and therefore the grain boundary velocity. This enables the metal particles to advance together with the moving alumina boundaries and results in the observed particle size distribution. This kinetically dependent mechanism provides a means to develop a nanocomposite microstructure with the particles (or the majority of the particles) occluded within the matrix grains. On the other hand, occlusion can be prevented, for the most part, if glass-forming impurity elements are not introduced into the material during the processing stage. As we will see in the next section, the position of the particles (i.e. occluded or at grain boundaries) can influence the microstructurally dependent properties of nanocomposites.
© Woodhead Publishing Limited, 2006
Processing and microstructural control
299
11.3.4 Residual stresses and fracture strength The measurement of residual stresses is usually associated with the analysis of mechanical properties, and not microstructure. However, residual stress fields in nanocomposites depend directly on microstructural parameters (particle size, position and spacing), as well as bulk material properties, such as differences in the coefficient of thermal expansion. The residual stresses in a composite can be of two types. The first is often called a macrostress, resulting from, for example, cutting, grinding or polishing. Macrostresses will result in changes in atomic spacing across a volume of material, and can be measured by the shifts (∆2θ) of XRD reflections. In composites microstresses are associated with thermoelastic mismatch between different phases or between anisotropic grains. Microstresses are inhomogeneous in nature, and will fluctuate from point to point in the material. While the average value of microstresses, across a given volume of material, will also contribute to XRD reflection shifts, fluctuations in the amplitude of the microstresses causes broadening of XRD reflections.47 Any nanocomposite material will have some degree of residual stress, due to the difference in thermal expansion coefficients between the particle and matrix phases. The paradigm for stress analysis in nanocomposites is the SiC-reinforced alumina system, where the smaller coefficient of thermal expansion of the SiC particles, relative to alumina, induces in the surrounding alumina a compressive radial stress and a tangential tensile hoop stress.48 Due to the anisotropic thermal expansion of alumina, ~100 MPa tensile stress fields are expected to exist along alumina grain boundaries.49 The thermally induced compressive stresses due to SiC particles adjacent to alumina grain boundaries may account for the reduced pullout of grains during polishing which has been observed for SiC-reinforced alumina nanocomposites.50 Furthermore, tensile hoop stresses due to occluded SiC particles would be expected to contribute to the transition from intergranular to transgranular fracture. Levin et al. explored the microstructurally dependent residual stress fields by analyzing their distribution and amplitude as a function of particle size and distribution (or content).51 The contribution of fluctuating microstress fields to changes in the material fracture toughness showed that an increase in fracture toughness is expected only for relatively small particle contents (3.5–5 wt% SiC), which can be optimized by reducing the particle size for the same volume fraction of occluded particles. For most metal-reinforced nanocomposites the thermal expansion coefficient of the metal phase will be larger than that of the matrix, reversing the expected stress fields compared to SiC-reinforced alumina. Thus while the tensile radial stresses surrounding occluded particles may induce transgranular cracking, the compressive hoop stresses may inhibit crack propagation if the particles are located at grain boundaries. Macrostresses in sub-micron Ni
© Woodhead Publishing Limited, 2006
300
Ceramic matrix composites
particles in a sintered alumina matrix have been measured using XRD techniques.23 The level of stress can be quite significant, even considering that the yield strength of bulk Ni at high temperatures is significantly reduced compared to room temperature (i.e. less than 10% of its room temperature value at ~1000°C).52 This is no doubt due to the limited plasticity in confined nanometer-sized particles.
11.4
Properties
The initial interest in ceramic matrix nanocomposites arose from reports by Niihara and co-workers indicating enhanced mechanical properties due to the presence of ceramic (SiC) particles.53 With the development of various processing routes to introduce nanometer-sized metal particles in a ceramic matrix, variations in functional (i.e. magnetic) properties are possible. In the following we briefly review the microstructurally dependent properties, with emphasis on the possible mechanisms leading to improved properties and using SiC-reinforced alumina as a point of comparison.
11.4.1 Mechanical properties Particle-reinforced ceramic matrix nanocomposites gained attention due to their enhanced mechanical properties, demonstrated by Niihara and coworkers.53 The initial focus was on ceramic particle (SiC)-reinforced alumina, with matrix grain size refinement due to grain boundary pinning being the initial explanation for the high fracture strengths. With the advent of metalreinforced nanocomposites, matrix grain size refinement was again used to explain the high fracture strengths (see Fig. 11.7).12 Upon examining the residual stress in SiC-reinforced alumina, it appears that thermally induced compressive stresses due to occluded particles induce transgranular cracking and reduce crack propagation along the matrix grain boundaries.1 According to this model, only occluded particles, within a specific particle concentration limit,51 contribute to the two combined strengthening mechanisms, and particles located at grain boundaries will reduce the fracture strength of the nanocomposite. This may be the reason for alternative processing routes leading to fracture strengths lower than the values reported by Niihara and co-workers, i.e. nanometer-sized particle occlusion was not achieved or a significant number of particles remained at the grain boundaries. In metal-reinforced ceramic matrix nanocomposites, matrix grain refinement has also been demonstrated by Niihara and co-workers.12 At the same time, as mentioned in the previous section, significant residual stress fields have been measured for Ni-reinforced alumina nanocomposites,23 although with reversed compression–tension fields compared to SiC-reinforced alumina. Thus for metal particles below the critical size for intrinsic cracking11 and
© Woodhead Publishing Limited, 2006
Processing and microstructural control
301
1200
Fracture strength (MPa)
1100
1000
900
800
700 600 0
5
10 15 Ni (vol%)
20
11.7 Effect of volume fraction of Ni particles on the fracture strength of Ni–Al2O3 nanocomposites, produced by the reduction of NiO particles (■), and the reduction of Ni-nitrate (•). Reproduced from Sekino et al.,12 with permission from the Journal of the American Ceramic Society.
located at grain boundaries, compressive hoop stresses exerted by the metal particles on the ceramic grain boundaries are expected to play some role in reducing crack propagation.22 While occluded metal particles will induce local tensile radial stresses in the alumina grains, it has yet to be directly shown that this mechanism results in a change in mode to transgranular cracking. It is most unlikely that matrix grain refinement alone can explain the high fracture strengths that have been reported. Using the reported maximum fracture strengths (σ = 1100 MPa)12 and the nominal fracture toughness of alumina (KIc = 3.5 MPa.m1/2), the critical flaw size, (c), can be estimated from
K c = Ic Yσ
2
(11.6)
to be of the order of 10 µm. This is an order of magnitude larger than the matrix grain sizes measured for the same nanocomposites. So while matrix grain size reduction may contribute to increases in fracture strength, it can only be part of the series of mechanisms leading to a reduced effective flaw size. Some reports have cited bridging across the relatively ductile metal particles as mechanisms which contribute to the increase in fracture strength.54 Other reports demonstrate cracks propagating at the metal–ceramic interfaces,
© Woodhead Publishing Limited, 2006
302
Ceramic matrix composites
precluding bridging effects.12 It is almost certain that metal particles in the size range from nanometers to a few hundred nanometers do not contribute to bridging effects and have extremely limited plasticity. It is far more likely that energy dissipation mechanisms of advancing cracks are defined by the metal–ceramic interfaces and the stress fields around the particles.
11.4.2 Functional properties A thorough review of the various functional properties obtained using metal– ceramic composites is beyond the scope of this work. However, a few cases and general trends will be briefly considered. Obtaining nanometer-sized metal particles dispersed in a ceramic matrix is attractive for a number of reasons. The ceramic matrix provides protection against corrosion or hightemperature oxidation of the normally oxidation-susceptible metal particles. Ferromagnetic particles in this size range can form single magnetic domains. Combining the latter with a conducting or semiconducting matrix which is stable at high temperatures can form a system with giant-magnetoresistance properties.23,55 The ratio of surface area to bulk volume of the reinforcing particles can have important implications on optical properties, where the contribution of surface states can result in unique properties.56,57 These surface states cause shifts in the plasmon absorption frequencies and can be manipulated by use of different combinations of metals and ceramics.56 Another possibility due to the high surface area of the metal particles is catalysis applications, provided the ceramic matrix contains open pores.19 The key to most of the functional properties reported is a fine microstructure of the metal particles (i.e. in the nanometer scale) which is uniformly dispersed within a ceramic matrix. In some cases the particle size needed is in the range of a few nanometers in order to enhance the surface properties, while in other cases optimization is needed between the demand for single domain particles while minimizing unwanted surface states.
11.4.3 Oxidation resistance Due to the potential high-temperature application of nanocomposites, as well as the fact that metal-reinforced ceramic nanocomposites combine metal and non-metal phases in equilibrium, it is important to understand the oxidation resistance of such materials. Using the Ni–alumina system as an example, and following Sekino et al.,12 the partial pressure of oxygen required to prevent the formation of nickel spinel (NiAl2O4) from a two-phase mixture of Ni and Al2O3 can be described as:58,59 PO 2 (atm) = exp
© Woodhead Publishing Limited, 2006
–58240 ± 480 + 17.94 T
(11.7)
Processing and microstructural control
303
10–5 10–7
NiAl2O4 + Al2O3
10–9
P(O2) (atm)
10–11 10–13 Ni +Al2O3 10–15 10–17 10–19 10–21 10–23 10–25 800
1000
1200 1400 Temperature (°C)
1600
1800
11.8 Graph showing the partial pressure of oxygen required to prevent the formation of nickel spinel (NiAl2O4) from a two-phase mixture of Ni and Al2O3 as a function of temperature.
Figure 11.8 graphically demonstrates this relationship, and shows that for conventional sintering temperatures (1400–1600°C), a reasonable partial pressure of oxygen (10–8–10–6 atm) is required to prevent the reaction Ni + Al 2 O 3 + 1 O 2 → NiAl 2 O 4 2
(11.8)
Such levels of oxygen are fairly simple to maintain during processing by using a slightly reducing atmosphere, and the cooling rates employed during processing are apparently fast enough to prevent spinel formation. However, during long exposures to medium-temperature operating conditions, e.g. 1000°C, spinel formation is certainly expected. Wang et al.60 demonstrated this for the Ni–alumina system, showing the diffusion of Ni atoms to the free surface of the nanocomposite, followed by the formation of a nickel spinel surface coating which then limits the kinetics of subsequent oxidation. In this case the formation of a spinel surface layer may be beneficial to mechanical properties, since the reaction results in a volume increase, and the formation of compressive residual stresses. An analogous behavior was reported for ceramic particle nanocomposites, where oxidation of SiC particles results in an increase in volume and compressive residual stresses.61 In an opposite trend, the addition of ZrO2 particles to Ni-reinforced alumina
© Woodhead Publishing Limited, 2006
304
Ceramic matrix composites
matrix nanocomposites degraded the oxidation resistance of the nanocomposite.62 Above the critical ZrO2 particle concentration for percolation, the ZrO2 phase provides a rapid route for oxygen diffusion, which enhances oxidation of the composite.
11.5
Future trends
It is very clear from the numerous studies on metal-reinforced ceramic matrix nanocomposites that controlling the microstructural features is critical to controlling and/or achieving specific properties. At the same time, our understanding of the mechanisms controlling microstructural evolution in such complicated systems is limited. Basic studies are required to understand the mechanisms, and specifically the role of dopants (including gas phases) and especially metal cations on sintering and grain boundary migration rates. In a similar manner, correlating interface energy with fracture energy at metal–ceramic interfaces is critical to the design of optimized mechanical properties, while interface structure and chemistry is important for functional properties. Alternative processing methods also offer the potential to control the microstructure and final properties of nanocomposites. Both self-propagating high-temperature sintering and spark plasma sintering offer means to obtain metastable yet dense nanocomposites. Subsequent heat treatments can then be used to approach equilibrium microstructures, where the properties will be a function of the heat treatment temperature and time. In this way a variety of microstructures, and thus variations of the composite properties, can become available. Additions of more than one type of metal particle, intermetallic phases, or graded particle concentrations offer a rich field for research into potential functional properties. Based on the processes discussed in this chapter, it is fairly simple to introduce different types of metallic particles into the ceramic matrix, which depending on the respective phase diagram would either remain as separate phases, form solid solutions, or form intermetallic compounds. Varying the partial pressure of gas components in the sintering atmosphere further expands the number of degrees of freedom to form different nanometersized phases. Finally, as demonstrated, intergranular films can form both at the matrix grain boundaries and between the matrix grains and the reinforcing particles, which can alter the processing kinetics and final properties of the metal-reinforced ceramic matrix nanocomposite.
11.6
References
1. Ferroni, L.P. and Pezzotti G., ‘Evidence for bulk residual stress strengthening in Al2O3/SiC nanocomposites’, J. Am. Ceram. Soc., 2002 85(8) 2033–2038.
© Woodhead Publishing Limited, 2006
Processing and microstructural control
305
2. Tuan, W.H., Lin, M.C. and Wu, H.H., ‘Preparation of Al2O3/Ni composites by pressureless sintering in H2’, Ceramics International, 1995 21 221–225. 3. Tai, W.P., Kim, Y.S. and Kim, J.G., ‘Fabrication and magnetic properties of Al2O3/ Co nanocomposites’, Mater. Chem. and Phys., 2003 82 396–400. 4. Barsoum, M., Fundamentals of Ceramics, Series in Materials Science and Engineering, London: McGraw-Hill, 1997. 5. Chaing, Y.M., Birnie, III D. and Kingery, W.D., Physical Ceramics – Principles for Ceramic Science and Engineering, MIT Series in Materials Science and Engineering, New York: Wiley, 1997. 6. Gavrilov, K.L., Bennison, S.J., Mikeska, K.R., Chabala, J.M. and Levi-Setti, R., ‘Silica and magnesia dopant distributions in alumina by high-resolution scanning secondary ion mass spectrometry’, J. Am. Ceram. Soc., 1999 82(4) 1001–1008. 7. Brydson, R., Chen, S.C., Riley, F.L., Milne, S.J., Pan, X. and Rühle, M., ‘Microstructure and chemistry of intergranular glassy films in liquid-phase-sintered alumina’, J. Am. Ceram. Soc., 1998 81(2) 369–379. 8. Park, C.W. and Yoon, D.Y., ‘Abnormal grain growth in alumina with anorthite liquid and the effect of MgO addition’, J. Am. Ceram. Soc., 2002 85(6) 1585–1593. 9. Tuan, W. H., Wu, H.H. and Chen, R.Z., ‘Effect of sintering atmosphere on mechanical properties of Ni/Al2O3 composites’, J. Eur. Ceram. Soc, 1997 17 735–741. 10. Li, G.J., Huang, X.X. and Guo, J.K., ‘Fabrication and mechanical properties of Al2O3–Ni composite from two different powder mixtures’, Mater. Sci. Eng. A, 2003 352 23–28. 11. Kolhe, R., Wi, C.Y.I., Ustandag, E. and Sass, S.L., ‘Residual thermal stresses and calculation of the critical metal particle size for interfacial crack extension in metal– ceramic matrix composites’, Acta Mater, 1996 44(1) 279–287. 12. Sekino, T., Nakajima, T., Ueda, S. and Niihara K., ‘Reduction and sintering of a nickel–dispersed-alumina composite and its properties’, J. Am. Ceram. Soc., 1997 80(5) 1139–1148. 13. Chen, R.Z. and Tuan, W.H., ‘Pressureless sintering of Al2O3/Ni nanocomposites’, J. Eur. Ceram. Soc, 1999 19 463–468. 14. Oh, S.T., Sekino, T. and Niihara, K., ‘Fabrication and mechanical properties of 5 vol% copper dispersed alumina nanocomposite’, J. Eur. Ceram. Soc., 1998 18 31– 37. 15. Niihara, K., Kim, B.S., Nakayama, T., Kusunose, T., Nomoto, T., Hikasa, A. and Sekino, T., ‘Fabrication of complex-shaped alumina/nickel nanocomposites by gelcasting process’, J. Eur. Ceram. Soc., 2004 24(12) 3419–3425. 16. Rodeghiero, E.D., Tse, O.K., Chisaki, J. and Giannelis, E.P., ‘Synthesis and properties of Ni–Al2O3 composites via sol-gel’, Mater. Sci. Eng. A, 1995 195 151–161. 17. Takahashi, R., Sato, S., Sodesawa, T., Suzuki, M. and Ichikuni, N., ‘Ni/SiO2 prepared by sol-gel process using citric acid’, Microporous and Mesoporous Materials, 2003 66(2–3) 197–208. 18. Huang, Y.L., Xue, D.S., Zhou, P.H., Ma, Y. and Li, F.S., ‘α-Fe–Al2O3 nanocomposites prepared by sol-gel method’, Mater. Sci. Eng., 2003 359(1–2) 332–337. 19. Sales, L.S., Robles-Dutenhefner, P.A., Nunes, D.L., Mohallem, N.D.S., Gusevskaya, E.V. and Sousa, E.M.B., ‘Characterization and catalytic activity studies of sol-gel Co–SiO2 nanocomposites’, Materials Characterization, 2003 50(2–3) 95–99. 20. Viart, N., Richard-Plouet, M., Muller, D. and Pourroy, G., ‘Synthesis and characterization of Co/ZnO nanocomposites: towards new perspectives offered by metal/piezoelectric composite materials’, Thin Solid Films, 2003 437 1–9.
© Woodhead Publishing Limited, 2006
306
Ceramic matrix composites
21. Trumble, K.P., ‘Spontaneous infiltration of non-cylindrical porosity: close packed spheres’, Acta. Mater., 1998 46(7) 2363–2367. 22. Lieberthal, M. and Kaplan, W.D., ‘Processing and properties of Al2O3 nanocomposites reinforced with sub-micron Ni and NiAl2O4,’ Mater. Sci. Eng. A., 2001 302(1) 83– 91. 23. Aharon, O., Bar-Ziv, S., Gorni, D., Cohen-Hyams, T. and Kaplan, W.D., ‘Residual stresses and magnetic properties of alumina–nickel nanocomposites’, Scripta Mater, 2004 50(9) 1209–1213. 24. Gabrisch, H., Dahmen, U. and Johnson, E., ‘In-situ observations of the interaction of liquid lead inclusions with grain boundaries in alumina’, Microscopy and Microanalysis, 1998 4 286–293. 25. Avishai, A., Scheu, C. and Kaplan, W.D., ‘Amorphous films at metal–ceramic interfaces’, Zeitschrift für Metallkunde, 2003 94 272–276. 26. Kaplan, W.D., Levin, I. and Brandon, D.G., ‘Significance of faceting on SiC nanoparticles in alumina’, Mater. Sci. Forum., 1996 207–209 733–736. 27. Ghetta, V., Fouletier, J. and Chatain, D., ‘Oxygen adsorption isotherms at the surfaces of liquid Cu and Au–Cu alloys and their interfaces with Al2O3 detected by wetting experiments’, Acta. Mater., 1996 44(5) 1927–1936. 28. Avishai, A. and Kaplan, W.D., ‘Intergranular films in metal–ceramic composites and the promotion of metal particle occlusion’, Zeitschrift für Metallkunde, 2004 95 266–270. 29. Gottstein, G. and Shvindlerman, L.S., ‘Theory of grain boundary motion in the presence of mobile particles’, Acta. Mater., 1993 41(11) 3267–3275. 30. Hsueh, C.H., Evans, A.G. and Coble, R.L., ‘Microstructure development during final/intermediate stage sintering – I. Pore/grain boundary separation’, Acta. Metall., 1982 30 1269–1279. 31. Lipkin, D.M., Clarke, D.R. and Evans, A.G., ‘Effect of interfacial carbon on adhesion and toughness of gold–sapphire interfaces’, Acta. Mater., 1998 46(13) 4835–4850. 32. Levi, G., Scheu, C. and Kaplan, W.D., ‘Segregation of aluminium at nickel–sapphire interfaces’, Inter. Sci., 2001 9 213–220. 33. Levi, G., Clarke, D.R. and Kaplan, W.D., ‘Free surface and interface thermodynamics of liquid nickel in contact with alumina’, Inter. Sci., 2004 12(1) 73–83. 34. Saiz, E., Tomsia, A.P. and Cannon, R.M., ‘Ridging effects on wetting and spreading of liquids on solids’, Acta. Mater., 1998 46(7) 2349–2361. 35. Kleebe, H.J., Hoffmann, M.J. and Rühle, M., ‘Influence of secondary phase chemistry on grain boundary film thickness in silicon nitride’, Zeitschrift für Metallkunde, 1992 83(8) 610–617. 36. Moberlychan, W.J., Cao, J.J, and De, Jonghe, L.C., ‘The roles of amorphous grain boundaries and the beta–alpha transformation in toughening SiC’, Acta. Mater., 1998 46(5) 1625–1635. 37. Clarke, D.R., ‘On the equilibrium thickness of intergranular glass phases in ceramic materials’, J. Am. Ceram. Soc., 1987 70(1) 15–22. 38. Clarke, D.R., Shaw, T.M., Philipse, A.P. and Horn, R.G., ‘Possible electrical doublelayer contribution to the equilibrium thickness of intergranular glass films in polycrystalline ceramics’, J. Am. Ceram. Soc., 1993 76(5) 1201–1204. 39. Choi, H.J., Kim, G.H., Lee, J.G. and Kim, Y.W., ‘Refined continuum model on the behavior of intergranular films in nitride ceramics’, J. Am. Ceram. Soc., 2000 83(11) 2821–2827.
© Woodhead Publishing Limited, 2006
Processing and microstructural control
307
40. Avishai, A. and Kaplan, W.D., ‘Intergranular films at metal–ceramic interfaces: Part II – Calculation of Hamaker coefficients’, Acta. Mater., 2005 53(5) 1571–1581. 41. Avishai, A., Scheu, C. and Kaplan, W.D., ‘Intergranular films at metal–ceramic interfaces: Part I – Interface structure and chemistry’, Acta. Mater., 2005 53(5) 1559–1569. 42. Oh, S.T., Sando, M., Sekino, T. and Niihara, K., ‘Processing and properties of copper dispersed alumina matrix nanocomposites’, Nanostructured Mater, 1998 10 267–272. 43. Chen, R.Z. and Tuan, W.H., ‘Pressureless sintering of Al2O3/Ni nanocomposites’, J. Eur. Ceram. Soc., 1999 19 463–468. 44. Powers, J.D. and Glaeser, A.M., ‘Grain boundary migration in ceramics’, Inter. Sci., (1998) 6(1–2) 23–39. 45. Saiz, E., Cannon, R.M. and Tomsia, A.P., ‘Energetics and atomic transport at liquid metal/Al2O3 interfaces’, Acta. Mater., 1999 47(15) 4209–4220. 46. Monchoux, J.P. and Rabkin, E., ‘Microstructure evolution and interfacial properties in the Fe–Pb system’, Acta. Mater., 2002 50 3159–3174. 47. Cullity, B.D., Elements of X-ray Diffraction, 2nd edn, London: Addison-Wesley, 1978. 48. Levin, I., Kaplan, W.D., Brandon, D.G. and Wieder., T., ‘Residual stresses in alumina– SiC nanocomposites’, Acta. Mater., 1994 42(4) 1147–1154. 49. Blendell, J.E. and Coble, R.L., ‘Measurement of stress due to thermal expansion anisotropy in Al2O3’, J. Am. Ceram. Soc., 1982 65(3) 174–178. 50. Zhao, J., Stearns, L.C., Harmer, M.P., Chan, H.M., Miller, G.A., and Cook, R.F., ‘Mechanical behavior of alumina–silicon carbide nanocomposites’, J. Am. Ceram. Soc., 1993 72(2) 503–510. 51. Levin, I., Kaplan, W.D., Brandon, D.G. and Layyous, A.A., ‘Effect of SiC submicrometer particle size and content on fracture toughness of alumina–SiC nanocomposites’, J. Am. Ceram. Soc., 1995 78(1) 254–256. 52. Rosenberg, S.J., ‘Nickel and its alloys’, National Bureau of Standards Monograph, 1968 106 38–39. 53 . Niihara, K. and Nakahira, ‘Particular-strengthened oxide ceramics’, Mater. Sci. Monographs, 1991 68 637–644. 54. Tuan, W.H. and Brook, R.J., ‘The toughening of alumina with nickel inclusions’, J. Eur. Ceram. Soc., 1990 6 31–37. 55. Cohen-Hyams, T., Plitzko, J.M., Hetherington, C.J.D., Hutchison, J.L., Yahalom, J. and Kaplan, W.D., ‘Microstructural dependence of giant-magnetoresistance in electrodeposited Cu–Co alloys’, J. Mater. Sci., 2004 39(18) 5701–5709. 56. Zakrzewska, K., Radecka, M., Kruk, A. and Osuch, W., ‘Noble metal/titanium dioxide nanocermets for photoelectrochemical applications’, Solid State Ionics, 2003 157(1-4) 349–356. 57. Selvan, S.T., Hayakawa, T., Nogami, M., Kobayashi, Y., Liz-Marzán, L.M., Hamanaka, Y. and Nakamura, A, ‘Sol-gel derived gold nanoclusters in silica glass possessing large optical nonlinearities’, J. Phys. Chem. B, 2002 106 10157–10162. 58. Elrefaie, F.A. and Smeltzer, W.W., ‘Thermodynamics of nickel–aluminum–oxygen system between 900 and 1400K’, J. Electrochemical. Soc., 1981 128(10) 2237– 2242. 59. Trumble, K.P. and Rühle, M., ‘The thermodynamics of spinel interphase formation at diffusion-bonded Ni/Al2O3 interface’, Acta. Mater., 1991 39(8) 1915–1924.
© Woodhead Publishing Limited, 2006
308
Ceramic matrix composites
60. Wang, T.C., Chen, R.Z. and Tuan, W.H., ‘Oxidation resistance of Ni-toughened Al2O3’, J. Eur. Ceram. Soc., 2003 23 927–934. 61. Yoshimura, M., Ohji, T. and Niihara, K., ‘Oxidation-induced toughening and strengthening of Y2O3/SiC nanocomposites’, J. Am. Ceram. Soc., 1997 80(3) 797– 799. 62. Wang, T.C., Chen, R.Z. and Tuan, W.H., ‘Effect of zirconia addition on the oxidation resistance of Ni-toughened Al2O3’, J. Eur. Ceram. Soc., 2004 24 833–840.
© Woodhead Publishing Limited, 2006
12 Carbon nanotubes-ceramic composites A P E I G N E Y and C H L A U R E N T, CIRIMAT, Université Paul-Sabatier, France
12.1
Introduction
Carbon nanotubes (CNTs) have recently emerged in the research world as one of the most promising nanomaterials. Both their characteristics and their properties lead one to think that their incorporation in materials could lead to new nanocomposites with new or enhanced physical or mechanical properties. In the present chapter, the results of researches which have been recently developed to prepare and characterize novel CNT-ceramic composites are reviewed. Firstly, in Section 12.2, the different structural forms of CNTs are described and the most common synthesis methods are presented as well as the physical and mechanical properties. In Section 12.3, the different processes which have been retained to obtain homogeneous dispersions of CNTs in ceramic powders are explained and the methods used to densify the composites are compared. In Section 12.4, the mechanical properties, and most particularly the results obtained for fracture toughness, are discussed, the effects of CNT addition on the electrical conductivity of insulating or semi-conducting ceramics are described, and works reporting the thermal conductivity of these composite materials are presented. Then, in the light of these very recent results, the key problems which require further researches will be discussed.
12.2
Structure, synthesis and properties of carbon nanotubes
Five years after the discovery of fullerenes, Iijima reported in 19911 a novel form of organized carbon which consists of hollow cylindrical structures, a few nanometers in diameter and some micrometers long. Although hollow carbon nanofibers had been prepared for several decades, their walls had never been resolved by High-Resolution Transmission Electron Microscopy (HRTEM). These HRTEM images allowed Iijima to conclude that the walls of the so-called multi-walled carbon nanotubes (MWCNTs) are made up of several concentric cylinders, each being formed by a graphene sheet rolled 309 © Woodhead Publishing Limited, 2006
310
Ceramic matrix composites
onto itself so that the open edges match exactly. Half fullerenes or more complex structures that include pentagons close the tips of each cylinder. In 1993, the Iijima and Bethune groups reported simultaneously2,3 the synthesis of CNTs composed of only one wall, the so-called single-walled carbon nanotubes (SWCNTs). For each cylindrical wall, the hexagons can present a lot of different orientations related to the tube axis, each giving a particular structure (Fig. r 12.1).4 Each orientation is represented by a helical vector Ch which is deduced r r ( a1 and a 2 ) of the graphene sheet by using a pair from the director vectors r r r of integers (n, m): Ch = na1 + ma 2 (Fig. 12.2). The limiting cases are referred to as zig-zag (n, 0) and armchair (n, n) SWCNTs (Figs 12.1(b) and 12.1(a) respectively), with a helical (or chiral) angle θ equal to 0° and 30°, respectively. Other SWCNTs have a helical (often termed chiral) structure (Fig. 12.1(c)). It will be seen later that the electronic properties of CNTs greatly depend on their helicity. The most common defects in the wall consist of pentagons or heptagons which can induce elbows on a tube or allow junctions between tubes of different structures. It is important to note that most SWCNTs are generally found in ropes (or bundles) of several tens of SWCNTs, arranged in a triangular lattice. The intertube distances are around 0.34 nm. In MWCNTs, the measured interlayer distance (0.34–0.39 nm) is close to that measured between graphene sheets in graphite and no particular correlation appears
…
(a)
…
(b)
…
(c)
12.1 Three examples of particular structures of SWCNTs, depending on the orientation of the hexagons related to the tube axis. (a) armchair-type tube (θ = 30°), (b) zigzag type tube (θ – 0°), and chiral tube (0 < θ < 30°). Reprint from Carbon, vol. 33, No. 7, Dresselhaus M.S., Dresselhaus G., Saito R., Physics of carbon nanotubes, pages 883–891, Copyright (1995) with permission from Elsevier.
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
311
y
B′
x
B θ
T A
a1 O
Ch
a2
r 12.2 Helical (or chiral) vector Ch defined from the director vectors r r (a1) and (a 2 ) of the graphene sheet by using a pair of integers (n, m) : r r r Ch = na1 + ma2 and chiral angle θ. Reprint from Carbon, vol. 33, No. 7, Dresselhaus M.S., Dresselhaus G., Saito R., Physics of carbon nanotubes, pages 883–891, Copyright (1995) with permission from Elsevier.
between the helicity of concentric layers. CNTs made up of two walls, the so-called double-walled nanotubes (DWCNTs), are of great interest, particularly for composite applications, because the outer wall provides the interface with the matrix and can be functionalized to control interface, whereas the inner wall is protected and thus retains its intrinsic properties. Three main methods are used to synthesize CNTs: arc-discharge evaporation of graphite electrode, laser sublimation of graphite rods, and catalytic chemical vapor deposition (CCVD).5 The first method produces short and wellcrystallized CNTs, either MWCNTs (without a catalyst) or SWCNTs (with a catalyst) with a narrow diameter distribution (around 1.2–1.4 nm). The SWCNTs are generally arranged in ropes. The main problem is the fairly low purity of the samples, which requires oxidative treatments resulting in some damage to the CNTs. The laser method produces also ropes of SWCNTs having a narrow diameter distribution but with a much higher purity. These ropes, however, have a larger diameter (several tens of nanometers) than those of the arc-CNTs and can be made up of hundreds of SWCNTs. CCVD routes derive from the methods used for decades for the synthesis of carbon nanofibers (CNF). The catalytic decomposition, at a high temperature (600– 1100°C), of a carbonaceous gas (generally a hydrocarbon or carbon monoxide) on nanoparticles (mainly Fe, Co or Ni) produces the formation of CNTs. Generally, the mechanism involved in the formation of a CNT is similar to that of a CNF, i.e. only one CNT from one nanoparticle. Depending mainly on the diameter of the active catalytic particle, and also on the reaction
© Woodhead Publishing Limited, 2006
312
Ceramic matrix composites
conditions, either a MWCNT, a SWCNT or a DWCNT can be formed. Thus, particular catalytic materials or methods are needed to obtain a sufficient selectivity, particularly to produce SWCNTs or DWCNTs, because it is necessary to limit the size of the catalytic particle to a few nanometers (<3 nm), at a high temperature. The SWCNTs or DWCNTs generally have a distribution of diameters larger than with the previous methods. They are individual or in small ropes, and can be up to 100 µm long. MWCNTs produced by CCVD, particularly those with many walls, often achieve less crystallization than those produced by arc or laser methods. They are generally very long (hundred of micrometers), and the presence of defects can induce important irregularities in the walls. Some are compartmented and sometimes have a bamboo shape. In this case, the designation of these carbon filaments as MWCNTs or as CNFs is debatable. CCVD methods are the most promising because they are economical and can be scaled up. Several companies currently sell MWCNT or SWCNT samples produced by such methods. But, before attempting to use the CNTs to prepare a composite, it is interesting to investigate the characteristics of the samples because each method produces CNTs with their own particularities and purities. Furthermore, particular attention must be paid to the eventual damage to CNTs, or to the functionalization, which could have been induced by purification treatments. CNTs are unique one-dimensional hollow objects with a very high aspect ratio (100–100 000) and a very high specific surface area (1300 m2/g for closed SWCNTs6). Their ends are closed and they have a low reactivity, but oxidative treatment is efficient in opening the ends and getting functional groups on the walls. They combine a high Young modulus (>1 TPa) with a high tensile strength (mean value close to 30 GPa)7 and great resilience due to the capacity to form kinks reversibly. These mechanical properties are related to well-crystallized individual CNTs, and are less for most MWCNTs produced by CCVD. Otherwise, inside a rope, the SWCNTs are weakly linked and allow easy sliding, which is detrimental to the tensile strength. The thermal conductivity of CNTs is very high (2000–6000 W.m–1K–1). The electronic properties of CNTs have been reviewed by Ahlskog et al.8 The most important characteristic to note for use of the CNTs in composites is that either a metallic or a semiconducting behavior may be observed, depending on the diameter and helicity. The electrical conductivity of a metallic CNT could reach 10 000 S.cm–1 and that of a semiconducting CNT is in the range 0.1–100 S.cm–1. But no method has currently been found to produce CNT samples containing only metallic or only semiconducting CNTs. It has only been reported that some separation methods tend to discriminate SWCNTs as a function of their helicity, but only at a microscopic scale.
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
12.3
313
Preparation of CNT-ceramic composites
After getting a sufficient quantity of CNTs and collecting their characteristics, the first difficulty in preparing dense CNT-ceramic composite materials is to obtain a composite powder in which the CNTs are well distributed, without forming CNT aggregates. Mixing by co-milling of CNT and ceramic powders, generally in wet media, has been used by several authors. The better efficiency of the synthesis of CNT in situ within the ceramic powders or, symmetrically, the synthesis of the ceramic in situ around the CNT, has been demonstrated. The second difficulty is to achieve good densification of the material. Both hot-pressing (HP) and spark plasma sintering (SPS) have been used to densify the composite.
12.3.1 Mixing the CNTs with the ceramic powders A simple ultrasonic agitation in alcohol of a mixture of long CCVD–MWCNTs with nano-SiC, nano-Si3N4 or SiO2 powders seems to be insufficient to disperse the CNT aggregates, which appear on SEM images of the densified composites.9–11 The use of an ultrasonic probe, more powerful than a bath, to disperse arc-MWCNTs, shorter than CCVD ones, in nano-Al2O3 powders is more efficient.12 To mix SWCNTs, most of which are included in ropes, with nanometric alumina, Zhan et al13,14 used, after agitation of an ethanol suspension in an ultrasonic bath, ball-milling for 24 h with zirconia media. SEM observations reveal only little damage to the CNTs. Wang et al.15 conducted TEM observations on powders obtained by the same preparation mode and reported good dispersion. Note that, in these works, the carbon filaments are ropes of SWCNTs, at least 10 nm in diameter, and not individual SWCNTs. A few authors have worked to improve the dispersion of CCVD-MWCNTs in oxide powders by using organic additives to modify the CNT and/or oxide surfaces.16–18 Sun et al.16 treated the CNTs in NH3 at 600°C for 3 h, resulting in a positive surface change, and added a small quantity of polyethyleneamine (PEI) which promotes the dispersion of the CNTs upon sonication. An aqueous suspension of nanometric alumina with polyacrylic acid is added, giving coated CNTs, which are themselves added to another concentrated alumina suspension. After drying and grinding, the composite contains 0.1 wt% of CNTs. TEM observations of the dense composites reveal a good dispersion of CNTs in the alumina matrix.16 In further works, Sun and Gao17 showed that the control of the surface nature (basic or acidic) of MWCNTs and the use of appropriate dispersants allow their efficient coating by surface-treated alumina or titania powders via a heterocoagulation process. These works have proved that the dispersion of MWCNTs in oxide powders can be significantly improved by an appropriate functionalization and the addition of an appropriate surface agent.
© Woodhead Publishing Limited, 2006
314
Ceramic matrix composites
12.3.2 In situ synthesis of the ceramic in CNT samples Some authors have prepared composite powders by the synthesis of the ceramic in situ around the CNT. Using short arc-made MWCNTs as templates, Hwang and Hwang19 prepared silica glass rods and then mixed them with a silica powder. More precisely, the templates were surfactant–MWCNT comicelles, prepared by using C16TMAB (cetyltrimethylammonium bromide), on which silicates were polymerized from sodium silicate by treatment at 110°C during 48 h in an autoclave. Ning et al.18 used a sol-gel process to prepare a composite made up of very long CCVD-made MWCNTs in a silica matrix. Several kinds of surfactants were tested – cationic (C16TMAB), anionic (polyacrylic acid) and nonionic (C16EO) with ultrasonic agitation to disperse the CNTs. TEOS (Si(C2H5O)4) was used as raw material of SiO2. The three kinds of surfactants were efficient at dispersing the CNTs in water. The SEM study of the dense samples gave evidence that CNTs are more homogeneously dispersed and have less agglomeration in composites prepared with C16TMAB than in composites prepared without any surfactant. Huang and Gao.20 prepared MWCNT–BaTiO3 composite materials in three steps. Firstly, rutile TiO2 particles were immobilized on the walls of CCVD-made MWCNTs. After treatment of CNTs in nitric acid reflux (140°C, 2 h) to functionalize the walls, this was performed by, the addition of TiCl4 to the suspension of functionalized CNTs at 90°C followed by a reaction time of 6 h. Secondly, barium acetate and NaOH were added and the reaction was conducted under hydrothermal conditions (160°C, 8 h). Finally, the obtained product was mixed and wet ball-milled with a BaTiO3 powder. The dispersion of CNTs was homogeneous, owing to the better compatibility with the BaTiO3 powder of the MWCNTs covered by BaTiO3 than the asprepared MWCNTs. Recently, Liu and Gao21 prepared MWCNT–NiFe2O4 composite materials by in situ chemical precipitation of metal hydroxides followed by hydrothermal processing. A long duration oxidation (10°C, 8 h) of the CNTs by a mixture of concentrated sulfuric acid and nitric acid was conducted to get their functionalization. Then, nickel and iron nitrates were added to an ethanol suspension of these CNTs, and after addition of sodium hydroxide (up to pH = 8.5) and a 2 h stirring, the product was treated in an autoclave at 110°C for a few hours. In comparison with composites prepared by the same route, but using non-oxidized CNTs, the dispersion was more homogeneous, as evidenced by SEM and TEM observations and also by an increase of the electrical conductivity of the corresponding dense materials. Jiang and Gao22 reported the preparation of MWCNT–TiN and MWCNT– Fe2N composites using CCVD CNTs oxidized in nitric acid (140°C, 24 h). A mixed solution of Ti(OC4H9)4 and ethanol was added to an aqueous suspension of CNTs. The dried precursor was calcinated in N2 (450°C, 2 h)
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
315
and then treated in flowing NH3 (800°C, 5 h) to lead to the CNT–TiN product. The CNT–Fe2N product was prepared similarly using a Fe(III)– urea complex as the precursor. Jiang and Gao23 also obtained a MWCNT– magnetite composite by an in situ solvothermal synthesis from a Fe(III)– urea complex. Besides these operations to produce composite powders with the aim of preparing dense materials, further work was devoted to coating CNTs with in situ synthesized oxide particles, for instance via a sol-gel process24,25 or the precipitation of a hydroxide.26 Thus many processes are efficient at synthesizing oxide or nitride particles in situ on functionalized MWCNTs to obtain composite powders in which the dispersion of MWCNTs has much better homogeneity than in powders obtained by mixing of MWCNTs and ceramic particles.
12.3.3 In situ synthesis of the CNTs in the ceramic powder The third way to prepare CNT–ceramic composite powders is via the synthesis of CNT by a CCVD process, in situ in the ceramic powder. A ceramic powder which contains catalytic metal particles at a nanometric size, appropriate to the formation of CNTs, is treated at a high temperature (600–1100°C), in an atmosphere containing a hydrocarbon or CO. In the method reported in 1997 by the present authors,27 iron nanoparticles are generated in the reactor itself, at a high temperature (>800°C), by the selective reduction in H2/CH4 (18% CH4) of an α-Al2O3 based oxide solid solution: Al2 – 2xFe2xO3 + 3xH2 → (1 – x)Al2O3 + 2xFe° + 3xH2O with x ≤ 0.10 Many clusters of Fe atoms are formed both inside and at the surface of each alumina grain and progressively grow. When the Fe nanoparticles located at the surface of the alumina grains reach a size around 0.7–2 nm, they immediately catalyze the decomposition of CH4 and the nucleation and growth of CNTs of very small diameter: CH4 → C(Fe°) → CNT(Fe/Fe3C)/Al2O3 The obtained CNT–Fe–Al2O3 powder is composed of clean isolated CNTs and small-diameter bundles of CNTs, surrounding all the oxide grains as a web. SEM and HRTEM studies have shown that most CNTs are SWCNTs or DWCNTs (80%), with only a small proportion of three- to six-walled CNTs, and have external diameters between 0.7 and 5 nm. The method has been extended to MgAl 2 O 4 (Fig. 12.3) and MgO based powders, using Mg(1–x)MxAl2O4 or Mg(1–x)MxO (M = Co, Fe or Ni) as starting materials.28,29 The products which were used to prepare dense composites were CNT–Fe–
© Woodhead Publishing Limited, 2006
316
Ceramic matrix composites
200 nm
(a)
200 nm
(b)
200 nm
(c)
200 nm
(d)
12.3 High resolution SEM images of SWCNT-Co-MgAl2O4 composite powders prepared by in situ synthesis of SWCNTs within the oxide powders, showing the high homogeneity of the distribution of CNT and the control of the CNT content. CNT contents: (a) 2.5 vol%, (b) 15.0 vol%, (c) 18.3 vol%, (d) 24.5 vol%. Reprint from Acta Materilia, vol, 52, No. 4, Rul S., Lefevre-Schlick F., Capria E., Laurent Ch. and Peigney A., Percolation of single-walled carbon nanotubes in ceramic matrix nanocomposites, pages 1061–1067, Copyright (2004) with permission from Elsevier.
Al2O 3, CNT–Co–MgAl2O4 and CNT–Co–MgO powders. In CNT–Co– MgAl2O4, most CNTs were SWCNTs. With such MgAl2O4 based powders modified by the additional use of a molybdenum compound, and of catalytic materials in the form of ceramic foams as opposed to powders, it was possible to synthesize in situ up to 25 vol% of SWCNTs.30,31 With submicronic starting oxide powders, the homogeneity of the distribution of CNTs inside the matrix is very high (Fig. 12.3), probably much better than that obtained with other methods. However, this method presents the disadvantage of being limited to a few matrices, i.e. those allowing the substitution of a catalytically active species in their cationic sub-lattice. It must also be pointed out that the obtained composite powders contain both a dispersion of CNTs and a dispersion of metal (and/or carbide) nanoparticles. Another way to prepare catalytic metal nanoparticles in a ceramic powder is to impregnate the powder by a solution of a precursor salt. Weidenkaff et al.32 impregnated a substituted LnCoO3 (Ln = Er, La) powder with a citrate,
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
317
precursor or the same phase, but containing an slight excess of cobalt. MWCNTs were formed in situ by treatment of this material at 700°C in acetylene. An et al.33 mixed alumina, a small quantity of MgO and some iron nitrate by planetary ball-milling in ethanol. After calcination, this material was treated in a C2H2/H2/N2 gas flow (750°C, 2 h) to form MWCNTs, well distributed in the alumina powder. These methods, based on the use of catalytic particles which are produced from the impregnation of the matrix by a precursor, offer a large choice of ceramic matrices, but generally lead to MWCNTs because catalytic particles become too large, at a high temperature, to catalyze the formation of SWCNTs. A similar method was also used to prepare CNT– MgO composite films.34 Iron and magnesium nitrates and molybdenum acetylacetone were added to a block polymer in ethanol suspension and deposited on a silica substrate. Treatment in CH4 (950°C, 15 min) led to the formation of bundles of CNTs, which were identified as SWCNTs, at the film surface. Another interesting process has been reported by Kamalakaran et al.35 and consists of the spray pyrolysis of a slurry of γ-Al2O3 and ferrocene in xylene, which was sprayed at 1000°C using argon as the carrier gas. They obtain large flakes composed of a large quantity of MWCNTs intricately matted in a glassy alumina matrix. This continuous process could be promising, if it can be optimized to control the quantity and the quality of CNTs and to obtain a crystallized matrix.
12.3.4 Densification of CNT-ceramic composites Hot-pressing (HP) is the most common method which has been used to densify CNT–oxide11,12,18,20,31,33,36–38 or CNT–SiC composites.9 The present authors showed that, by HP at temperatures between 1450 and 1530°C, under 43 MPa and in a primary vacuum, CNT–Fe–Al2O3 nanocomposites reach a lower densification than the corresponding carbon-free Fe–Al2O3 nanocomposites, and that both the matrix grain growth and densification are inhibited when the CNT content is increased.36–38 Moreover, some CNTs are damaged by HP at 1500°C, producing disordered graphene layers gathered at matrix grain junctions.38 Similar results are obtained on CNT–Fe/Co– MgAl2O4 composites, with only a densification of 90.6% for the composites containing 4.9 wt% of SWCNTs versus 98.2% for the corresponding Fe/Co– MAl2O4 material.38 The study of densification by HP (1300°C, 43 MPa, secondary vacuum) of 15 different SWCNT–Co/Mo–MgAl2O4 composites containing between 1.2 and 16.7 vol% CNT has shown that CNTs do influence the rearrangement step.39 For low quantities of CNT (up to 8 vol%), CNTs are favorable to this process, but for higher contents they are detrimental because they form too rigid a web structure.39 In the second step, where the main active process of densification is a plastic flow controlled by a Nabarro– Herring diffusion creep,40 CNTs are as detrimental to the process when their
© Woodhead Publishing Limited, 2006
318
Ceramic matrix composites
content increases, from 5.0 to 16.7 vol%,39 leading to a decrease in densification from 95% to 73%. For CNT–Co–MgO composites (2.8 vol% of SWCNTs or DWCNTs), it is necessary to increase the HP temperature up to 1600°C to reach a densification of 93%, which results in the destruction of most CNTs.38 The quantity of undamaged CNTs retained in the dense composite is more dependent on the treatment temperature than on the nature of the oxide matrix.38 In order to align the CNTs, the present authors have performed high-temperature extrusion of CNT-ceramic composites, at 1500°C with Al2O3 and MgAl2O4 matrices and at 1730°C for the MgO matrix.41 In the first two composites, CNTs were globally aligned with no more damage than in HP composites, showing their high resistance to the shear stress developed during the extrusion. But CNTs were destroyed in the MgO–matrix composite owing to the too high temperature. An et al.33 performed HP of MWCNT–Al2O3 composites (CNTs obtained by CCVD) at 1800°C, in argon at a pressure of 40 MPa. The negative influence of CNTs on the densification was reported. SEM images of fractures showed that the grain size of Al2O3 tends to decrease when the CNT content is increased (from 2.7 to 12.5 wt%), and that MWCNTs are located at grain boundaries. Thus, MWCNTs appear to be much more resistant at temperatures between 1500 and 1800°C than SWCNTs or DWCNTs. Siegel et al.12 reported the near-total densification by HP at 1300°C in argon, under an applied pressure of 60 MPa, of a mixture of nanophase alumina powders with 10 vol% of short MWCNTs obtained by the arc method. Both the high reactivity of the matrix and the particular characteristics of the arc-made MWCNTs explain this good result. Ning et al.11 prepared MWCNT–SiO2 materials (CNTs obtained by CCVD) by HP at 1300°C, in N2, and with an applied pressure of 25 MPa. The sintering mechanism was the viscous flow of amorphous silica. These authors reported the presence of agglomerates of CNTs and a decrease of the densification from 98% to 74% when the CNT content was increased from 5 to 30 vol%. In a further work,18 they showed that enhancement of homogeneity, by using surfactant addition to prepare the starting powders, is favorable to the densification. Huang and Gao20 studied HP, in N2 and at 25 MPa, of MWCNT–BaTiO3 composites (CNTs obtained by CCVD). They reported 1200°C as the optimal temperature (for 0.1 vol% CNT) and, for this HP temperature, a decrease of the densification from 99% to 86.5% when the CNT content increases from 0.1 to 3.0 wt%, which is correlated with a much smaller matrix grain size. Ma et al.9 studied the densification by HP at 2000°C (25 MPa, Ar atmosphere) of a mixture of 10 wt% MWCNTs, obtained by CCVD, with nanometric SiC (80 nm). The densification obtained was only 64.7%. Thus, the addition of 1 wt% B4C and the increase of the temperature up to 2200°C were necessary to reach a densification of 98.1%. These authors reported some images showing that at least some of the MWCNTs had not been destroyed by this very high
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
319
temperature treatment. The only work of hot-isostatic pressing (HIP) of CNT–ceramic composites so far was reported by Balazsi et al.,10 and used a mixture of 1 wt% CCVD-MWCNTs with Si3N4, treated at 1700°C. They showed that MWCNTs remained after HIP at 2 MPa for 1 h, but that they had completely disappeared after treatment at 20 MPa for 3h. An efficient method to achieve the total densification of CNT–ceramic composites without damaging the CNTs could be the spark plasma sintering (SPS) technique. Zhan et al.13 studied the SPS of a mixture of ropes of 5.7 or 10 vol% SWCNTs (obtain by CCVD) with nanocrystalline alumina (40 nm), containing both the α and γ forms. Relative densities of 100% by SPS at 1150°C, for only 3 min and under an applied pressure of 63 MPa, were obtained for pure alumina and for the two CNT–Al2O3 composites. Owing to the low temperature reached and the very short time of treatment (heating within a few minutes), CNTs were not damaged. The SPS technique is well known to combine the effects of rapid heating, pressure and powder surface cleaning and that could explain its efficiency for CNT-ceramic materials. An appropriate microstructure of the powder, i.e. highly reactive alumina and SWCNTs in ropes rather than individual, probably also favors a better densification. Wang et al.15 sintered the same materials using SPS, but at 1450–1550°C under an applied pressure of 40 MPa, resulting in no more damage of CNTs than in the former work13 but in lower densifications (95.1% for 10 vol% CNTs). Sun et al.16 reported the full densification by SPS at 1300°C (5 min) of a MWCNT–Al2O3 mixture prepared by a colloidal process, but the result is less significant because of the very low CNT content (0.1 wt%). The difficulty of achieving the densification of CNT–ceramic composites is a critical problem, particularly when good mechanical properties are essential. Generally, the CNTs inhibit the matrix grain growth and the densification processes. Hot-pressing is often not efficient to totally densify the materials, particularly when the CNT content is higher than a few vol%. Higher homogeneity of the CNT distribution, such as that obtained by the in situ formation of CNTs, seems to make the problem worse. Increasing the temperature treatment and/or the applied pressure can contribute to increasing the densification, but leads to damage to the SWCNTs or DWCNTs. However, large-diameter MWCNTs are generally much more resistant. In comparison with HP, the SPS method involves shorter treatments and lower temperatures to obtain a complete densification, when the microstructure of the green material is optimized. Thus SPS is attractive because it avoids or at least limits the damage to the CNTs due to high temperatures.
© Woodhead Publishing Limited, 2006
320
12.4
Ceramic matrix composites
Properties of CNT-ceramic composites
12.4.1 Mechanical properties Most research on CNT-ceramic composite materials has for its main purpose been directed at the reinforcement of the ceramic, i.e. obtaining a significant increase in the fracture toughness of these brittle materials. Most authors have also measured the fracture strength because a tougher material with lower fracture strength is not desirable. Microscopy studies were also conducted on damaged or fractured materials to investigate the behavior of the CNTs included in the ceramic matrix, with the aim of identifying possible reinforcement mechanisms. These works were conducted mainly with oxide matrices such as Al2O3,12,13,15,16,36,38,42,43 MgAl2O4,38 and glassy or crystallized SiO2,11,18,19 and a only few were related to SiC9 or Si3N4.10 Two groups have reported the tribological properties of CNT–alumina composites33 and of carbon/carbon composites covered by a film of CNTs.44 The first results on the mechanical properties of alumina–matrix composites containing CNTs were reported in 1998 by the present authors.36 Several CNT–Fe–Al2O3 composite powders with different CNT contents (0.7–6 wt% of carbon, mainly SWCNTs or DWCNTs) were prepared by reducing Al2(1–x)FexO3 solid solutions of different compositions (x = 0.02–0.20) in H2–CH4, at 900 or 1000°C. The hot-pressed composites had fracture strengths (measured by three-point bending) lower than the corresponding Fe–Al2O3 nanocomposites45 and fracture toughness (measured by the single-etched notched beam (SENB) technique) similar to or slightly lower than that of Al2O3 (3–5 MPa.m1/2 and 4.4 MPa.m1/2 respectively). The matrix grains were micrometric and the CNTs were located either at grain boundaries or at grain junctions, or within the grains (Fig. 12.4(c)). The pullout of some CNTs was evidenced, showing a possible reinforcement mechanism (Fig. 12.4(c)). The absence of real reinforcement was attributed to the uncompleted densification of the materials, a too low CNT volume fraction, and the presence, in some composites at least, of undesirable forms of carbon (carbon nanofibers, nano-ribbons). In a further work,37 an attrition-milled starting oxide was used to enhance the homogeneity of the distribution of the CNTs within the matrix (100–500 nm grains – Figs 12.4(a), (b)), but both the fracture strength and fracture toughness were lower than that of the previous materials. Then, the quality of CNT was improved and the preparation method was extended to MgAl2O4 and MgO matrices, without obtaining any improvement of these mechanical properties, in comparison with the pure matrix.38 As reported in the previous section, whatever the nature of the oxide matrix, CNTs induce a low densification of the HP materials, and this effect increases with the CNT content, resulting in relatively poor mechanical properties. Siegel et al.12 reported, in 2001, an improvement of the fracture toughness from 3.4 MPa.m1/2 for hot-pressed Al2O3 to 4.2 MPa.m1/2 for a hot-pressed
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
(a)
(b)
321
200 nm
100 nm
(c)
100 nm
12.4 SEM images of fracture surface of CNT-Fe-Al2O3 composites densified by hot-pressing of composites powders, which have been prepared by in situ synthesis of CNTs (mainly SWCNTs and DWCNTs) within the oxide powder. (a) and (b) specimen prepared by using submicronic oxide powders, (c) specimen prepared by using micronic oxide powders. Reprint from Ceramic International, vol. 26, No. 6, Peigney A., Laurent Ch, Flahaut E. and Roussel A., Carbon nanotubes in novel ceramic matrix nanocomposites, pages 1061– 1067, Copyright (2000) with permission from Elsevier.
MWCNT–Al2O3 composite (10 vol% of short MWCNTs prepared by the arc method). Note that the fracture toughness was calculated by using the Evans and Charles equation, from the lengths of cracks emanating from Vickers indentations (5 kg load). Sun et al.16 reported a fracture toughness (determined by the same method) of 4.9 MPa.m1/2 for a MWCNT–Al2O3 composite fully densified by the SPS method and containing only 0.1 wt% of CNTs (long MWCNTs prepared by CCVD), in comparison to 3.7 MPa.m1/2 for Al2O3. These authors underlined the tight bonding between the CNTs and the matrix and inferred a bridging effect on cracks and some pullout as possible reinforcement mechanisms. The results which have attracted most attention, however, were those
© Woodhead Publishing Limited, 2006
322
Ceramic matrix composites
1000 nm
12.5 SEM image of the fracture surface of 5.7 vol% SWCNT-Fe-Al2O3 composite densified by spark plasma sintering (SPS) of a mixture of nanometric alumina and ropes of SWCNTs. Reprint from Nature Materials, No. 2, 2002, pp. 38–42, Zhan G.-D., Kuntz J.D., Wan J. and Mukherjee A.K., ‘Single-wall carbon nanotubes as attractive toughening agents in alumina-based nanocomposites’, with the permission of Nature Materials and of the authors (http:// www.nature.com/nmat/index.html).
reported by Zhan et al.13 on SWCNT–Al2O3 composites, with an increase of the fracture toughness (determined from Vickers indentations using the Antis equation) from 3.3 MPa.m1/2 for Al2O3 to 7.9 MPa.m1/2 for 5.7 vol% CNTs, and up to 9.7 MPa.m1/2 for 9.7 vol% CNTs. The hardness simultaneously decreased from 20.3 to 20.0 GPa, and then to 16.1 GPa. Remember that these composites were fully densified by SPS at 1150°C, and that the starting materials were ropes of SWCNTs (obtained by CCVD) mixed by ball-milling with a nanocrystalline Al2O3 powder (40 nm), containing both the α and γ forms. The ropes of SWCNTs are located at intergranular positions and seemed to be undamaged (Fig. 12.5). Very recently, Wang et al.15 reported the results of measurements, by the single-edged V-notch beam (SEVNB) method, of the fracture toughness of SWCNT–Al2O3 composites similar to those of the previous reference:13 using the same starting material, with 10 vol% SWCNTs, but sintering by SPS at 1550°C instead of 1150°C, giving a densification of 95.1%. They reported a fracture toughness of only 3.32 MPa.m1/2, similar to that of Al2O3 or graphite– Al2O3 composites (3.22 and 3.51 MPa.m1/2 respectively). The ropes of SWCNTs were not damaged by the short SPS treatment at 1550°C but the matrix grains seemed slightly larger and the densification lower than that obtained by SPS treatment at 1150°C.13 By Hertzian indentations using a tungsten carbide ball, they showed that no Hertzian cracks occur on such composites, in the conditions in which they are widely formed on alumina. So, CNT–
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
323
alumina composites are highly contact-damage resistant, similarly to graphite– alumina composites, and could be used in all applications where contact loading is important. Moreover, these results showed that the fracture toughness of such composites can be severely overestimated when measured by the standard indentation method.46 Indeed, so far neither the SENB nor the SEVNB results have evidenced that CNT can significantly reinforce alumina ceramics. Hwang and Hwang19 studied the hardness of glassy SiO2 in which either SiO2 rods or CNT–SiO2 rods (synthesized using MWCNTs as templates) are dispersed. The CNT–SiO2 rod–SiO2 materials had a significantly higher hardness than the SiO2 rod–SiO2 composites. However, the real influence of CNT on the hardness was not proved, because the presence of CNT inside each rod in the dense composite is questionable. Probably, the higher hardness is a consequence of a better morphology of the rods prepared by using the CNT templates. Thus, using the words ‘Carbon nanotube reinforced ceramics’ as the title of the paper19 is debatable. Ning et al.11 reported an increase of bending strength from 50 to 85 MPa (three-point bending) in 5 vol% MWCNT– SiO2 composites densified at about 98%, in spite of the fact that the homogeneity of the distribution of CNT seemed not ideal (Figs 12.6(a), (b)). They also reported an increase of the fracture toughness from 1 to 2 MPa.m1/2, but these values were determined by the indentation method, and thus the second value could be overestimated. Moreover, CNTs provide very suitable conditions for the nucleation and crystallization of cristobalite,11 and that could contribute to the evolution of the mechanical properties. In further work, Ning et al.18 enhanced the homogeneity of the CNT distribution (see the previous section) and obtained full densification of the 5 vol% CNT-containing composite; they reported still higher bending strength (97 MPa) and fracture toughness (2.46 MPa.m1/2). From TEM images, they inferred that a good interface bonding exists between the MWCNTs and glassy SiO2. These results on glassy materials are to be considered, but the fracture toughness has to be confirmed by measurements using a SENB-type method. Ma et al.9 reported 10% increases, both in fracture strength (three-point bending) and in fracture toughness (SENB method), in MWCNT–SiC composites hot-pressed at 2000 or 2200°C. The homogeneity of the distribution of CNTs was low and the increases were too small to be significant. Balazsi et al.10 compared the elastic modulus and the bending strength of several different materials prepared by HIP: Si3N4 alone and Si3N4 with additions of carbon black (23 wt%), graphite (23 wt%), carbon fibers (1 wt%) or MWCNTs (1 wt%) which all decreased the apparent density of the material. The higher modulus and strength values were obtained for Si3N4 alone and, although the values decreased less with the addition of MWCNTs than with addition of other additives, these results did not justify inclusion of the words ‘carbon nanotube reinforced silicon nitride composites’10 in the title of the paper.
© Woodhead Publishing Limited, 2006
324
Ceramic matrix composites
(a)
(b)
12.6 SEM images at low (a) and high magnification (b) of surface fractures of 5 vol% MWCNT-SiO2 composites densified at about 98% by hot-pressing, showing a rather low degree of homogeneity of the CNTs distribution. Reprint from Materials Science & Engineering, A: Structural Materials: Properties, Microstructure and Processing A, Vol. 357, No. 1–2, Ning J., Zhang J., Pan Y. and Guo J., Fabrication and mechanical properties of SiO2 matrix composites reinforced by carbon nanotube, pages 392–396, Copyright (2003) with permission from Elsevier.
Very recently, Xia et al.43 reported microstructural investigations on MWCNTs which had been formed within the regular and well-aligned pores of an alumina membrane. The material was too thin (20–90 µm) to permit mechanical measurements, but different possible reinforcement mechanisms induced by the CNTs were evidenced on stressed and damaged materials,
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites (a)
325
(b)
12.7 SEM images of the deformation around an indentation in a 90 µm thick MWCNT-Al2O3 composite prepared by in situ formation of CNTs within the regular and very well aligned pores of an alumina membrane. (a) array of the ‘shear bands’ formed, (b) close up view of lateral buckling or collapse of the CNTs in one ‘shear band’. Reprint from Acta Materialia, Vol. 52, Xia Z., Ricster L., Curtin W.A., LiH., Sheldon B.W., Liang J., Chang B. and Xu J.M., Direct observation of toughening mechanisms in carbon nanotube ceramic matrix composites, pages 931–944, Copyright (2004) with permission from Elsevier.
such as crack deflection, crack bridging and CNT pullout. Moreover, a new mechanism of CNT collapse in shear bands occurs (Fig. 12.7), rather than crack formation, suggesting that these materials can have a multiaxial damage tolerance.43 These results showed that the key problem to solve in order to obtain real reinforcement of a ceramic is probably the preparation of macroscopic samples in which the CNTs would have been well organized within the matrix. An et al.33 hot-pressed MWCNT–Al2O3 composites at 1800°C and noted both a decrease of the matrix grain size when the CNT content increases and a poor cohesion between the CNTs and alumina. For CNT contents up to 4 wt%, the microhardness was enhanced and the wear loss decreased, maybe owing to a lower matrix grain size. At higher CNT contents, the evolution of these parameters was reversed. Thus, the influence of the CNTs was not clearly established. Lim et al.44 applied on carbon/carbon composites different coatings made of a carbon-based material containing MWCNTs (synthesized by a CCVD method). They showed that the wear loss decreases regularly (up to 100% for 20 wt% CNT) but also that the friction coefficient increases slightly when the CNT content in the coating increased, showing a significant influence of the CNTs on tribological properties.
12.4.2 Electrical properties Only a few authors have reported results concerning the electrical properties of CNT-ceramic composites and most have been concerned with the influence
© Woodhead Publishing Limited, 2006
326
Ceramic matrix composites
of CNT on DC electrical conductivity measured at room temperature. In 2000, Flahaut et al.38 reported that CNTs confer an electrical conductivity to ceramic–matrix composites, whereas the corresponding ceramics and metal– oxide nanocomposites are insulators. Owing to the percolation of the CNTs (mainly SWCNTs or DWCNTs), the DC electrical conductivity jumped from values lower than 10–10 S.cm–1 for Fe–Al2O3 of Fe/Co–MgAl2O4 composites to 0.4–4.0 S.cm–1 for CNT–Fe–Al2O3 composites, or to 1.5–1.8 S.cm–1 for CNT–Fe/Co–MgAl2O4 composites.38 These values are fairly well correlated to the relative quantity of CNTs. For CNT–Co–MgO composites, the value was lower (0.2 S.cm–1) because an important proportion of CNTs were damaged or destroyed during the hot-pressing at 1600°C. Rul et al.31 prepared, by in situ synthesis, 22 different SWCNT–Co/Mo– MgAl2O4 composites with a wide range of CNT content (between 0.11 and 24.5 vol%). Their DC electrical conductivity jumped from 10–10 S.cm–1 to 0.0040 S.cm–1 between 0.23 and 1.16 vol% and reached 8.5 S.cm–1 for the higher CNT content (Fig. 12.8(b)). It was shown that the electrical conductivity was well fitted by the scaling law of the percolation theory, σ = σ0(p – pc)t with a percolation at a low threshold, pc = 0.64 vol%, and an exponent t = 1.73 close to the theoretical value (t = 1.94) characteristic of a three-dimensional network (Fig. 12.8(a)). The low percolation threshold is a consequence of the very large aspect ratio (>10 000) of the SWCNT. These results showed that the electrical conductivity of such composites can be tailored in a wide scale (10–2–10 S.cm–1) through the CNT content. Peigney et al.41 showed that the alignment of CNTs within such matrices, obtained by hot extrusion, can lead to an important anisotropy of the electrical conductivity: 20 S.cm–1 in the extrusion direction versus 0.60 S.cm–1 in the transverse direction. Zhan et al.14 also measured the DC electrical conductivity of Al2O3-based composites containing ropes of SWCNTs. The DC electrical conductivity increased up to 33.45 S.cm–1 upon increasing the CNT content, a value significantly higher than that reported by Rul et al.31 with unorganized CNTs. Less damage to CNTs during the SPS sintering (1150°C for only a few minutes) than during hot-pressing (1300°C, heating rate 10°C.min–1) could explain this result. Some authors have reported results on the influence of CNTs on the electrical properties of non-insulating mixed oxides. Huang and Gao20 densified MWCNT–BaTiO3 composites by hot-pressing (1200°C, 35 MPa) of BaTiO3 powders containing MWCNTs covered by in situ synthesized BaTiO3 particles. The electrical conductivity decreased when the CNT content increased (from 6.9 S.cm–1 for BaTiO3 to 3.6 S.cm–1 for the composite containing 3 wt% CNT), and the n-type semi-conductivity of BaTiO3 was converted to p-type, even for a very low CNT content (0.1 wt%). This phenomenon has been attributed to a Schottky barrier constructed at the CNT–matrix contact, which could be promising for the fabrication of new-style ferroelectric and
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites 10+2
327
σ = k (p – pc)t with pc = 0.64 ± 0.02 t = 1.73 ± 0.02
1 10–2 Log10 σ (σ in S/cm)
σ (S/cm)
0.5
10–4 10–6 0.64 10–8
Log10 σ = 1.73 log10 (p – 0.64) – 1.758 R2 = 0.9921
0 –0.5 –1 –1.5 –2 –2.5 –0.5
0 0.5 Log10 (p – 0.64)
10–10 0
2
4 6 CNT (vol %)
8
1
10
(a) 10
σ (S/cm)
1
0.1
0.01
0.001
0
5
10 15 CNT (vol %)
20
25
(b)
12.8 DC electrical conductivity of SWCNT-Co-MgAl2O4 composites versus the CNT contents. (a) up to 10 vol% CNTs, the values are well fitted by the scaling law of the percolation theory σ = σo(p – pc)t with a percolation at a low threshold, pc = 0.64 vol%, and an exponent t = 1.73. (b) the value jumps from 10–10 S.cm–1 to 0.0040 S.cm–1 Reprint from Acta Materialia, vol. 52, No. 4, Rul S., Lefèvre-Schlick F., Capria F., Laurent Ch. and Peigney A., Percolation of single-walled carbon nanotubes in ceramic matrix nanocomposites, pages 1061– 1067, Copyright (2004) with permission from Elsevier.
© Woodhead Publishing Limited, 2006
328
Ceramic matrix composites
thermoelectric devices.20 Liu and Gao21 prepared MWCNT–NiFeO4 composites and studied the effect of a surface oxidation treatment of MWCNTs. The increase of electrical conductivity of the composites was larger by using preoxidized CNTs (from 6.8 × 10–4 to 82.2 S.cm–1 versus to 9.1 S.cm–1, for 10 wt% CNTs) and this was attributed to a better homogeneity of the distribution of CNTs within the matrix and also to a stronger adhesion between the CNTs and the matrix.21 Jiang and Gao23 showed that, in composites prepared by in situ synthesis of magnetite on MWCNTs produced by CCVD, the electrical conductivity increased from 1.9 S.cm–1 for the material without CNTs to 2.5 S.cm–1 for the material containing 32.95 wt% of CNTs. This increase seems moderate, taking into account both the large content and the electrical conductivity of CNTs, showing that the influence of interfaces is probably determinant in the conduction mechanisms of the material. Huang et al.47 embedded MWCNTs (arc-discharge, 3 wt%) into Bi2Sr2CaCu2O8+δ (a Bi-2212 superconductor) by a partial melting processing and measured an insulator–superconductor transition slightly lower in the composite than in the Bi-2212 material (87 K versus 95 K) and also an increase in current densities. Owing to the percolation of CNTs in insulating ceramics, materials with a DC electrical conductivity directly tailored in a wide range (0.01 to 10–100 S.cm–1) by the quantity of CNTs can be prepared. The high aspect ratio of CNTs allows percolation thresholds lower than 1 vol%, i.e. CNT contents for which full densification of the composites is easily obtained. However, to reach the higher conductivity values (> 10 S.cm–1), higher CNT contents (>10 vol%) are necessary, which requires the use of the SPS method to achieve the densification of the materials. One of the critical points is the homogeneity of the distribution of CNTs, and the in situ synthesis methods are promising in that respect. The other critical point is the characteristic of the interfaces between CNTs and the matrix, which can be greatly influenced by oxidative or more complex fuctionalization treatments. In a semiconducting oxide matrix, CNTs can either increase or decrease the electrical conductivity, as a function of the conducting mechanisms involved both at the interfaces and in the matrix. As shown in ferroelectric or superconductor ceramic matrices, the addition of CNTs could notably influence the transport properties of many ceramics used in electronics, and possibly lead to materials with novel or improved functional properties. The organization of CNTs within the matrix makes possible the preparation of materials with anisotropic functional properties.
12.4.3 Thermal properties The very high thermal conductivity of CNTs (2000–6000 W.m–1K–1) lets one envision that they could be used to manage the thermal properties of ceramics.
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
329
Yowell48 reported the preparation by tape-casting of films of partially stabilized zirconia containing 1 wt% SWCNTs. However, the SWCNTs did not survive the thermal treatment, and thus the slight increase of the measured thermal conductivity could have been due to the presence of residual metal catalyst particles.48 In contrast, the dispersion of 1 wt% vapor-grown carbon fibers (VGCF), which withstand the sintering, provoked a decrease of about 30% of the thermal conductivity. Seeger et al.49 incorporated MWCNTs into SiO2 by partial matrix melting caused by a laser. They reported that the presence of CNTs was crucial for the heat absorption and melting of the matrix, showing that the thermal transport of SiO2 was probably greatly enhanced by CNTs. Thus, further works are required to investigate more precisely the influence of CNTs on the thermal properties of materials and the corresponding heat transport mechanisms. If these studies confirm that CNTs can significantly increase the thermal conductivity of ceramic material, the crucial point will be to determine for which CNT content this could be achieved.
12.5
Conclusions and future trends
Because the first reports on CNT-ceramic composites date only from 1998, and because only a few teams have worked so far on these novel materials, it could be argued that we are at the infancy of the development of a new class of composite materials. Researches on these materials depend firstly on a better knowledge of the CNTs by their users. Depending on their microstructural characteristics (SWCNTs, individual or in ropes, MWCNTs, diameter, length, number of walls), but also on the synthesis methods which have been used, the properties of CNTs may greatly vary. Notably, the treatments involved in the control of the surface properties and reactivity of the CNTs need to be optimized for a particular form of CNTs synthesized by a particular method. To obtain good homogeneity by mixing CNTs with a ceramic powder, it is necessary to adapt the surface properties of both the CNTs and the ceramic particles, which will be preferably nanometric, and this requires the addition of organic additives. The in situ synthesis of ceramic onto CNTs also requires a fine and stable dispersion of CNTs in a suitable medium. The latter method can ensure a good adhesion between the CNTs and the ceramic, but is more complex and less flexible than the previous one. The in situ synthesis of CNT within the ceramic powder leads to very homogeneous materials but requires an in-depth knowledge of the CCVD process to avoid the formation of undesirable forms of carbon, and can be applied to only a few matrices. Thus, for each particular variety of CNT-ceramic composites, and with a view to developing a particular property, a dedicated preparation method could be preferred. To densify such materials, spark plasma sintering (SPS) has proved to be efficient, leading to good densifications with limited damage
© Woodhead Publishing Limited, 2006
330
Ceramic matrix composites
to the CNTs. The possibility of increasing the fracture toughness of ceramics through the addition of CNTs is still debatable. But it has been demonstrated that for these materials, SENB-type tests are to be preferred to indentationderived measurements which are unsuitable. The contact damage resistance of CNT-ceramic composite could be promising, but the use of CNTs instead of graphite will have to be justified by gains in the other mechanical properties. However, at the microscopic scale, several reinforcement modes have been evidenced, particularly when the CNTs are well aligned within the matrix. In spite of its very high aspect ratio and its intrinsic mechanical properties, the interfacial surface area developed by one CNT is small, due to its nanometric diameter. Thus, the effect of only one (or a few) isolated CNT(s) is probably too low to really influence the propagation of a critical crack. But the effect of many CNTs operating simultaneously could be useful, which requires a sufficient CNT content, and especially a high degree of organization of the CNTs to ensure a synergistic effect. Owing to the high aspect ratio of CNTs which facilitates their percolation, their addition to insulating ceramics is efficient in giving electronic conductivity to the material and in tailoring the conductivity value directly by the quantity of CNTs. Moreover, anisotropic conductivity is obtained when the CNTs are aligned within the composite. In semiconducting ceramics, CNTs could notably influence the transport properties of many ceramics used in electronics, and possibly lead to materials with novel or improved functional properties. The thermal properties of ceramics can be modified, and the thermal conductivity possibly enhanced, via the addition of CNTs. But, for most properties, the organization of the CNTs within the matrix is a prerequisite to taking full advantage of their dispersion in ceramic materials. While the viability of structural CNT-ceramic composites may be a long shot, the promising results reviewed here give strong hope that functional materials with tailored electrical/thermal characteristics will find their way into industrial applications.
12.6
Sources of further information
The Nanotube Site (http://www.nanotube.msu.edu/) has general information about CNTs, links to sources of nanotubes and nanotube-based products and to sites relevant to nanotube research. Note that research in the carbon nanotube field is progressing at a very fast pace (about 10 reviewed papers each day). Another source of information is the website of the research group of the authors, the Nanocomposites and Carbon Nanotubes Group (http://ncn.f2g.net/). Key books on carbon nanotubes are:
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
331
• Dresselhaus, M.S., Dresselhaus, G., Eklund, P.C. and editors, Science of Fullerenes and Carbon Nanotubes, Academic Press, San Diego, 1996. • Harris P.J.F., Carbon Nanotubes and Related Structures – New Materials for the Twenty-first Century, Cambridge University Press, Cambridge, 1999.
12.7
References
1. Iijima, S., ‘Helical microtubules of graphitic carbon’, Nature, 1991, 354, 56. 2. Iijima, S. and Ichihashi, T., ‘Single-shell carbon nanotubes of 1-nm diameter’, Nature (London), 1993, 363, 603–605. 3. Bethune, D.S., Kiang, C.H., de Vries, M.S., Gorman, G., Savoy, R., Vazquez, J. and Beyers, R., ‘Cobalt-catalyzed growth of carbon nanotubes with single-atomic-layer walls’, Nature (London), 1993, 363, 605–607. 4. Dresselhaus, M.S., Dresselhaus, G. and Saito, R., ‘Physics of carbon nanotubes’, Carbon, 1995, 33, 883–891. 5. Harris, P.J.F., Carbon Nanotubes and Related Structures – New Materials for the Twenty-first Century, Cambridge University Press, Cambridge, 1999. 6. Peigney, A., Laurent, Ch., Flahaut, E., Bacsa, R.R. and Rousset, A., ‘Specific surface area of carbon nanotubes and bundles of carbon nanotubes’, Carbon, 2001, 39, 507– 514. 7. Yu, M.-F., Files, B.S., Arepalli, S. and Ruoff, R.S., ‘Tensile loading of ropes of single wall carbon nanotubes and their mechanical properties’, Phys. Rev. Lett., 2000, 84, 5552–5555. 8. Ahlskog, M., Laurent, Ch., Baxendale, M. and Huhtala, M. (2004) In Encyclopedia of Nanoscience and Nanotechnology, Vol. 3 (ed. Nalwa, H.S.), Am. Sci. Pub., pp. 139–161. 9. Ma, R.Z., Wu, J., Wei, B.Q., Liang, J. and Wu, D.H., ‘Processing and properties of carbon nanotube/nano-SiC ceramic’, J. Mater. Sci., 1998, 33, 5243–5246. 10. Balazsi, C., Konya, Z., Weber, F., Biro, L.P. and Arato, P., ‘Preparation and characterization of carbon nanotube reinforced silicon nitride composites’, Materials Science and Engineering, C: Biomimetic and Supramolecular Systems, 2003, C23, 1133–1137. 11. Ning, J., Zhang, J., Pan, Y. and Guo, J., ‘Fabrication and mechanical properties of SiO2 matrix composites reinforced by carbon nanotube’, Materials Science and Engineering, A: Structural Materials: Properties, Microstructure and Processing, 2003, A357, 392–396. 12. Siegel, R.W., Chang, S.K., Ash, B.J., Stone, J., Ajayan, P.M., Doremus, R.W. and Schadler, L.S., ‘Mechanical behavior of polymer and ceramic matrix nanocomposites’, Scr. Mater., 2001, 44, 2061–2064. 13. Zhan, G.-D., Kuntz, J.D., Wan, J. and Mukherjee, A.K., ‘Single-wall carbon nanotubes as attractive toughening agents in alumina-based nanocomposites’, Nature Materials, 2003, 2, 38–42. 14. Zhan, G.-D., Kuntz, J.D., Garay, J.E. and Mukherjee, A.K., ‘Electrical properties of nanoceramics reinforced with ropes of single-walled carbon nanotubes’, Appl. Phys. Lett., 2003, 83, 1228–1230. 15. Wang, X., Padture, N.T. and Tanaka, H., ‘Contact-damage-resistant ceramic/SWCNT and ceramic/graphite composites’, Nature Materials, 2004, 3, 539.
© Woodhead Publishing Limited, 2006
332
Ceramic matrix composites
16. Sun, J., Gao, L. and Li, W.,‘Colloidal processing of carbon nanotube/alumina composites’, Chemistry of Materials, 2002, 14, 5169–5172. 17. Sun, J. and Gao, L., ‘Development of a dispersion process for carbon nanotubes in ceramic matrix by heterocoagulation’, Carbon, 2003, 41, 1063–1068. 18. Ning, J., Zhang, J., Pan, Y. and Guo, J., ‘Surfactants assisted processing of carbon nanotube-reinforced SiO2 matrix composites’, Ceramics International, 2004, 30, 63–67. 19. Hwang, G.L. and Hwang, K.C., ‘Carbon nanotube reinforced ceramics’, Journal of Materials Chemistry, 2001, 11, 1722–1725. 20. Huang, Q. and Gao, L., ‘Manufacture and electrical properties of MWCNT/BaTiO3 nanocomposite ceramics’, J. Mater. Chem., 2004, 14, 2536–2541. 21. Liu, Y. and Gao, L., ‘A study of the electrical properties of carbon nanotube–NiFe2O4 composites: effect of the surface treatment of the carbon nanotubes’, Carbon, 2005, 43, 47–52. 22. Jiang, L. and Gao, L., ‘Carbon nanotubes–metal nitride composites: a new class of nanocomposites with enhanced electrical properties’, J. Mater. Chem., 2005, 15(2), 260–266. 23. Jiang, L. and Gao, L., ‘Carbon nanotubes–magnetite nanocomposites from solvothermal processes: formation, characterization, and enhanced electrical properties’, Chemistry of Materials, 2003, 15, 2848–2853. 24. Seeger, T., Kohler, T., Frauenheim, T., Grobert, N., Rühle, M., Terrones, M. and Seifert, G., ‘Nanotube composites: novel SiO2 coated carbon nanotubes’, Chemical Communications, 2002, 34–35. 25. Vincent, P., Brioude, A., Journet, C., Rabaste, S., Purcell, S.T., Brusq, J.L. and Plenet, J.C., ‘Inclusion of carbon nanotubes in a TiO2 sol-gel matrix’, J. Non. Cryst. Solids, 2002, 311, 130–137. 26. Hernadi, K., Ljubovic, E., Seo, J.W. and Forro, L., ‘Synthesis of MWNT-based composite materials with inorganic coating’, Acta Materialia, 2003, 51, 1447–1452. 27. Peigney, A., Laurent, Ch., Dobigeon, F. and Rousset, A., ‘Carbon nanotubes grown in situ by a novel catalytic method’, J. Mater. Res., 1997, 12, 613–615. 28. Govindaraj, A., Flahaut, E., Laurent, Ch., Peigney, A., Rousset, A. and Rao, C.N. R., ‘An investigation of carbon nanotubes obtained from the decomposition of methane over reduced Mg1–xMxAl2O4 spinel catalysts’, J. Mater. Res., 1999, 14, 2567–2576. 29. Flahaut, E., Peigney, A., Laurent, Ch. and Rousset, A., ‘Synthesis of single-walled carbon nanotube–Co–MgO composite powders and extraction of the nanotubes’, J. Mater. Chem., 2000, 10, 249–252. 30. Rul, S., Laurent, Ch., Peigney, A. and Rousset, A., ‘Carbon nanotubes prepared in situ in a cellular ceramic by the gelcasting-foam method’, J. Eur. Ceram. Soc., 2003, 23, 1233–1241. 31. Rul, S., Lefèvre-Schlick, F., Capria, E., Laurent, Ch. and Peigney, A., ‘Percolation of single-walled carbon nanotubes in ceramic matrix nanocomposites’, Acta Mater. 2004, 52, 1061–1067. 32. Weidenkaff, A., Ebbinghaus, S.G. and Lippert, T., ‘Ln1–xAxCoO3 (Ln = Er, La; A = Ca, Sr)/carbon nanotube composite materials applied for rechargeable Zn/air batteries’, Chemistry of Materials, 2002, 14, 1797–1805. 33. An, J.W., You, D.H. and Lim, D.S., ‘Tribological properties of hot-pressed alumina– CNT composites’, Wear, 2003, 255, 677–681. 34. Du, C. and Pan, N., ‘Preparation of single-walled carbon nanotube reinforced magnesia films’, Nanotechnology, 2004, 15, 227–231.
© Woodhead Publishing Limited, 2006
Carbon nanotubes-ceramic composites
333
35. Kamalakaran, R., Lupo, F., Grobert, N., Lozano-Castello, D., Jin-Phillipp, N. Y. and Rühle, M., ‘In-situ formation of carbon nanotubes in an alumina–nanotube composite by spray pyrolysis’, Carbon, 2003, 41, 2737–2741. 36. Laurent, Ch., Peigney, A., Dumortier, O. and Rousset, A., ‘Carbon nanotubes–Fe– alumina nanocomposites. Part II: Microstructure and mechanical properties of the hot-pressed composites’, J. Eur. Ceram. Soc., 1998, 18, 2005–2013. 37. Peigney, A., Laurent, Ch., Flahaut, E. and Rousset, A., ‘Carbon nanotubes in novel ceramic matrix nanocomposites’, Ceram. Int., 2000, 26, 677–683. 38. Flahaut, E., Peigney, A., Laurent, Ch., Marlière, Ch., Chastel, F. and Rousset, A., ‘Carbon nanotube–metal-oxide nanocomposites: microstructure, electrical conductivity and mechanical properties’, Acta Mater., 2000, 48, 3803–3812. 39. Peigney, A., Rul, S., Lefèvre-Schlick, F. and Laurent, Ch., ‘Densification during hotpressing of carbon nanotube–metal–magnesium aluminate spinel composites’, to be submitted, 2006. 40. Ting, C.-J. and Lu, H.-Y., ‘Hot-pressing of magnesium aluminate spinel–I. Kinetics and densification mechanism’, Acta. Mater., 1999, 47, 817–830. 41. Peigney, A., Flahaut, E., Laurent, Ch., Chastel, F. and Rousset, A., ‘Aligned carbon nanotubes in ceramic–matrix nanocomposites prepared by high-temperature extrusion’, Chem. Phys. Lett., 2002, 352, 20–25. 42. Peigney, A., Laurent, Ch., Flahaut, E. and Rousset, A., ‘Carbon nanotubes as a part of novel ceramic matrix nanocomposites’, Adv. Sci. Technol. (Faenza, Italy), 1999, 15, 593–604. 43. Xia, Z., Riester, L., Curtin, W.A., Li H., Sheldon, B.W., Liang, J., Chang, B. and Xu, J.M., ‘Direct observation of toughening mechanisms in carbon nanotube ceramic matrix composites’, Acta Mater., 2004, 52, 931–944. 44. Lim, D.-S., An, J.-W. and Lee, H.J., ‘Effect of carbon nanotube addition on the tribological behavior of carbon/carbon composites’, Wear, 2002, 252, 512–517. 45. Devaux, X., Laurent, Ch., Brieu, M. and Rousset, A. (1992) In Composites Materials (ed., Benedetto, A.T.D., Nicolais, L. and Watanabe, R.), Elsevier Science, Amsterdam, pp. 209–214. 46. Sheldon, B.W. and Curtin, W.A., ‘Tough to test’, Nature Materials, 2004, 3, 505– 506. 47. Huang, S.L., Koblischka, M.R., Fossheim, K., Ebbesen, T.W. and Johansen, T.H., ‘Microstructure and flux distribution in both pure and carbon-nanotube-embedded Bi2Sr2CaCu2O8+δ superconductors’, Physica C, 1999, 311, 172–186. 48. Yowell, L.L., ‘Application of carbon nanotubes and fullerenes for thermal management in ceramics’, Mat. Res. Soc. Symp. Proc., 2001, 633, A17.4.1–A17.4.6. 49. Seeger, T., de la Fuente, G., Maser, W.K., Benito, A.M., Callejas, M. A. and Martinez, M.T., ‘Evolution of multiwalled carbon-nanotube/SiO2 composites via laser treatment’, Nanotechnology, 2003, 14, 184–187.
© Woodhead Publishing Limited, 2006
13 Machinable nanocomposite ceramics R W A N G, Arizona State University, USA
13.1
Introduction
Advanced ceramic materials have been successfully developed over the past few decades. They are widely used in a variety of applications for their superior properties in comparison to conventional materials, such as metals and polymers. A few of these properties include high strength-to-mass ratio, excellent wear resistance and exceptional corrosion resistance. Although the development of advanced ceramic materials has progressed tremendously over the years, barriers to their wide acceptance exist. One of these barriers is the inherent difficulty in materials processing. For instance, after sintering, ceramic materials become hard and brittle. In the processing of ceramic materials using traditional methods, such as machining and grinding, fracture occurs at stress-concentration locations which leaves cracks on and beneath the machined surfaces of ceramic components. These processinginduced damage areas degrade the quality of products and often lead to malfunction and/or catastrophic failure during the service life. Research on machinable advanced ceramics has concentrated on multiphase composites and materials design of ceramics. Most of these research efforts aim at improving the ceramic machinability while maintaining the material’s distinct features. Recently, nanocomposites in which nano-sized particles were dispersed within the matrix grains or at the grain boundaries show much better mechanical properties (hardness and strength) as well as machinability and superplasticity compared with monolithic materials. In this chapter, some machinable microcomposites and nanocomposites with weak or soft phase (h-BN, LaPO4 and Ti3SiC2) will be discussed based on the relationship between properties and microstructure.
13.2
Design principles of machinable ceramics
Ceramic composites may be defined as those ceramic materials that consist of two or more fundamentally different components that are able to act 334 © Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
335
synergistically to give properties superior to those provided by either component alone. Compound machinable ceramics are defined as those having a distinct weak interface phase or layered phase distributed throughout the bulk matrix ceramic. By varying the type and distribution of the weak interface phase or layered phase in the composite, it is possible to obtain a wide range of mechanical properties and machinability combinations. Such ceramic materials have a number of potential advantages for fabrication of complex-shape engineering components. Table 13.1 gives some basic materials properties for advanced ceramics and layered or soft materials. Based on the thermodynamic properties and sintering compatibility of these advanced ceramics and layered materials, a variety of machinable ceramic composites can be designed and constructed, such as Si3N4/h-BN, SiC/h-BN, AlN/hBN, Al2O3/h-BN, Al2O3/LaPO4, Al2O3/Ti3SiC2, Ce–ZrO2/CePO4 , etc. Unreactive or weak bonding is the main design principle of compound machinable ceramics. The addition of a layered or weak ceramic used as boundary phase in the ceramic–matrix composite can improve the machinability of advanced ceramics by deflecting crack propagation. The following energyabsorbing mechanisms have been identified from SEM or TEM images of areas in the vicinity of the indentation: diffuse microcracking, delamination, crack deflection, grain pullout and the buckling of individual grains. All of these might be attributed to the improved machinability of composites. Figure 13.1 shows the crystal structure and morphological microstructure of h-BN, Ti3SiC2 and LaPO4, so-called weak or layered phase. You can see that these materials possess a layered structure or step-way fracture characteristics, and form weak bonding with the matrix ceramics. Based on this design principle of machinable ceramics, oxide (Al2O3/LaPO4) and nonoxide (Si3N4/ h-BN) have been investigated and introduced considering the relationship between the mechanical properties, microstructure and machinability. In some research, nanocomposites showed potential machinability and superplasticity [1], as we will also discuss in this chapter.
13.3
Al2O3–LaPO4
Alumina (Al2O3) has been recognized as one of the most promising structural materials for many mechanical or thermomechanical applications because of its excellent high-temperature strength, good oxidation and wear resistance, high hardness, and low specific weight. Because of its strong covalent bonding character, extremely hard Al2O3 ceramics make conventional machining very difficult or even impossible. The addition of a weak interface phase or layered phase in the matrix to facilitate crack deflection and propagation during machining, named compound machinable ceramic, was used for improving the machinability of ceramic such as mica glass-ceramic. Lanthanum phosphate (LaPO4), also known as monazite, has been found to be a suitable and effective
© Woodhead Publishing Limited, 2006
Table 13.1 Basic material properties for advanced ceramics and layered or soft materials Advanced ceramics
Density (g cm–3)
Layered or soft materials
Si3N4
SiC
Al2O3
Y-PSZ
C
3.19
3.22
3.98
6.1
2.265 *
h-BN
Ti3SiC2
Mica
LaPO4
2.27
4.5
2.889
5.07
*
Hardness (GPa)
14–18
21–25
19.3
8–12
1–2
3–5
2–4
4~7
Elastic modulus (GPa)
280–320
450
400–410
180–220
2.5–10
50–80
320
—
100–220
Bending strength (MPa)
400–1000
640
550–600
650–1000
20–70
40–60
>250
170–360
90–140
Fracture toughness, KIC (MPa.m1/2)
3.4–8.2
5.7
3.8–4.5 RT
6–8 RT
<1
7
—
~1.0
Thermal expansion coefficient (× 10–6C–1)
2–3 0–1000°C
4.2 6.5–8.9 0–1500°C 200–1200°C
10.6 0–1000°C
1.2–28.3 20–800°C
0.59–10.51 8.6–9.7
—
10
Thermal conductivity (W m–1 K–1)
15–50
50
1–2
83.7–272
40–70
—
5
*Moh’s scale.
© Woodhead Publishing Limited, 2006
600°C
38.9
20°C
2
20–40
Machinable nanocomposite ceramics
337
C0 = 6.660 Å
B atom N atom
a0 = 2.054 Å
Max = 20.00 KX
2µm
2µm
EHT = 10.00 kV WD = 4 mm
(c)
Signal A = InLens Date: 28 Sep 2001
(b)
(a)
Mag =10.00KX
EHT = 10.00 kV WD = 8 mm
Signal A = SE2 Date: 22 Mar 2002
Mag = 20.00 KX
1µm
EHT = 10.00 kV WD = 6 mm
Signal A = SE2 Date: 22 Mar 2002
(d)
13.1 Crystal structure and morphology of some previously used ceramic interface phases: (a) h-BN; (b) Ti3SiC2 (courtesy of Y.M. Luo, Tsinghua University); (c) and (d) LaPO4.
debonding material for high-temperature oxide/oxide composites [2]. Fracture energy measurements have shown that the Al2O3/LaPO4 interface was weak enough, when compared to the Al2O3/Al2O3 interface, to satisfy the He and Hutchison debonding criteria [3] for bimaterial interfaces. Recent work by Davis and Marshall [4] has shown monazite to be a suitable weak bonding material in preparing composites with Al2O3. These composites show high toughness with an interface sufficiently weak for debonding to occur when cracks are deflected into monazite and away from the Al2O3 phase. LaPO4 in the Al2O3 composites is quite stable and no reaction occurs between the two phases up to 1600°C, provided the La:P ratio in the monazite is close to 1. Furthermore, composites of LaPO4 with some other ceramics are machinable as evaluated from information available in the literature. Interface properties of LaPO4 with Al2O3 and yttrium aluminum
© Woodhead Publishing Limited, 2006
338
Ceramic matrix composites
garnet (YAG) fibers have also yielded useful information on the design of the interfaces. Sudheendra et al. [5] have carried out a systematic investigation of the particulate composites of Al2O3 and TiO2 with a variety of monazites such as LnPO4 (Ln = La, Pr, Nd or Gd) by means of scanning electron microscopy in order to check the chemical compatibility of LnPO4 with alumina and titanium oxide. The studies reveal that all the monazites are compatible with the oxide ceramics and have favorable mechanical properties in terms of damage tolerance and increase in fracture toughness.
13.3.1 Experimental procedure The nanocrystalline LaPO4 powders were synthesized by mixing phosphoric acid with lanthanum oxide in a water bath [6]. The composite powders of different LaPO4 composition were ball-milled in ethyl alcohol with agate balls for 24 hours and then dried, and sieved using 100-mesh sift. Details of material preparation have been reported elsewhere [7, 8]. XRD was carried out using an X-ray diffractometer (Cu-Kα, Model Automated D/Max B, Japan). Hardness values were determined on polished samples using a 50 N load in a microhardness tester fitted with a Vickers square pyramidal indenter. Five indents were made on each sample. The hardness was deduced from the diagonal width (2a) of the indentation and the contact load (P) by the following equation:
H=
1854.4 P (2a ) 2
(13.1)
The flexural strength was measured by the three-point bending method with specimen dimensions of 36 × 4 × 3 mm3, a bending span of 30 mm, and a crosshead speed of 0.5 mm/min at room temperature. The tensile surface of the sample was polished with diamond paste down to 0.5 µm and the long edges of the tensile surface were rounded. The elastic modulus was measured by the load–displacement curve on specimens 36 × 4 × 3 mm3 using an A2000 Shimadzu universal materials testing machine with a crosshead speed of 0.05mm/min. Fracture surfaces of the composites were observed by scanning electron microscopy, using SEM LEO 1530, Germany. Samples were evaporatively coated with carbon or gold to obtain conductivity without affecting observed surface morphology.
13.3.2 Mechanical properties Figure 13.2 shows the influence of LaPO4 content on the bending strength of Al2O3/LaPO4 composites prepared by different sintering routes (pressureless sintering, hot pressing and spark plasma sintering). It is obvious that
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
339
600 Pressureless sintering at 1600°C Hot pressing at 1450°C Spark plasma sintering at 1350°C
550
Bending strength (MPa)
500 450 400 350 300 250 200 150 100 0
20
40 60 Content of LaPO4 (wt%)
80
100
13.2 Effect of LaPO4 content on the bending strength of Al2O3/LaPO4 composites.
incorporating LaPO4 particles remarkably reduced the fracture strength of the composites. From the result, for hot-pressed samples, the monolithic Al2O3 sintered at 1450°C exhibited a maximum strength of 550 ± 32 MPa, while the strength of 40 wt% Al2O3/LaPO4 composite and LaPO4 ceramic were 331 ± 41 MPa and 137 ± 18 MPa, respectively. Davis and Marshall [4] noted that the Al2O3/LaPO4 composites are completely unreactive at high temperature (below their eutectic temperature) and bond poorly to one another. An added benefit was the ability of activating dislocation slip systems in monazite (LaPO4) under critical shear stresses. Due to the weak bonding between Al2O3 and LaPO4 and the activating dislocation slip systems in monazite (LaPO4), it was evident that microstructure greatly affected the mechanical properties. Figure 13.3 shows the variation of the elastic modulus of the composites with increasing LaPO4 addition done by two different sintering routes. The decrease in elastic modulus attributed to LaPO4 concentration can be explained by the rule-of-mixture. The most important advantage of Al2O3/LaPO4 composites is that they possess excellent machinability. Hardness is an important parameter as an indication of ceramic machinability. Generally, lower hardness means better machinability. Figure 13.4 shows the effect of LaPO4 content on the hardness of Al2O3/LaPO4 composites made by three sintering routes. Hardness dropped rapidly with increased LaPO4 content. Large amounts of LaPO4 may produce a relatively high volume fraction of the grain-boundary phase, which has a
© Woodhead Publishing Limited, 2006
340
Ceramic matrix composites 340
Pressureless sintering at 1600°C Hot pressing at 1450°C
320 300
Elastic modulus (GPa)
280 260 240 220 200 180 160 140 120 100 80 0
20
40 60 Content of LaPO4 (wt%)
80
100
13.3 Effect of LaPO4 content on the elastic modulus of Al2O3/LaPO4 composites.
18 Pressureless sintering at 1600°C Hot pressing at 1450°C Spark plasma sintering at 1350°C
17 16
Vickers hardness (GPa)
15 14 13 12 11 10 9 8 7 6 5 4 3 0
20
40 60 Content of LaPO4 (wt%)
80
100
13.4 Effect of LaPO4 content on the Vickers hardness of Al2O3/LaPO4 composites.
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
341
lower hardness than Al2O3 grains. All of the sample density is higher than 99% theoretical density, so the decrease of hardness is not due to the low densification caused by addition of LaPO4. The resistance of a material to the formation of a permanent surface impression by an indenter is termed hardness. The formation of weak interfacial bonding between Al2O3 and LaPO4 phases decreases the resistance of the composite to the plastic deformation. In this system, the crack can propagate along the weak bonding interface of Al2O3 and LaPO4 and activate the dislocation slip system of the LaPO4 phase [9]. Therefore, the improved crack deflection ability may contribute to the machinability improvement in the composites. The Vickers hardness of the 40 wt% Al2O3/LaPO4 composite and the LaPO4 ceramic by hot pressing was 4.69 ± 0.01 GPa and 4.48 ± 0.18 GPa, respectively. These values match the requirement of engineering machining using cemented carbide tools. The indicated LaPO4 ceramic and 40 wt% Al2O3/LaPO4 composite could possess excellent machinability, which mainly arises from the ease of chipping during cutting because of weak interfacial bonding between Al2O3 and LaPO4.
13.3.3 Microstructure Figure 13.5 shows the microstructure evolution of the composites fabricated by pressureless sintering. It is evident that the addition of LaPO4 in composites retards the grain growth of Al2O3. The back-scattered electron images also show even distribution of the LaPO4 phase in the Al2O3 matrix. Figure 13.6 shows the fracture surface of the 40 wt% Al2O3/LaPO4 composites prepared by the same sintering method. A layered LaPO4 phase surrounding the Al2O3 grains is the main fracture surface feature for these composites. It is worth noting from the SEM images that the secondary phases are primarily located at the grain boundaries. Since the thermal expansion coefficient of Al2O3 (~9 × 10–6/°C) is lower than that of LaPO4 (~10 × 10–6/°C), a residual tensile stress would be generated at the interface between Al2O3 and LaPO4 during cooling down after sintering. The existence of such a stress may weaken the interfacial bonding, leading to regional particle pullout during machining.
13.3.4 Machinability Using cemented carbide drills, it is possible to successfully machine 40 wt% Al2O3/LaPO4 specimens fabricated by pressureless sintering and hot pressing. Normally, due to their high hardness, diamond tools are used for the machining of advanced ceramics. However, pure Al2O3 cannot be machined using cemented carbide drills. As stated above, the layered LaPO4 phase and the weak interface at the Al2O3/LaPO4 grain boundaries are the main reason for the improvement of the machinability. Both enhance crack deflection and
© Woodhead Publishing Limited, 2006
342
Ceramic matrix composites
10 µm
10 µm (a) 10 wt% LaPO4/Al2O3
10 µm
10 µm (b) 20 wt% LaPO4/Al2O3
10 µm
10 µm (c) 30 wt% LaPO4/Al2O3
10 µm
10 µm (d) 40 wt% LaPO4/Al2O3
13.5 Surface morphology of Al2O3/LaPO4 composites by pressureless sintering.
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
343
2µm (a)
2µm (b)
13.6 SEM images of fracture surface for 40 wt% Al2O3/LaPO4 composite by pressureless sintering.
help avoid catastrophic failure of the material during drilling. Figure 13.7 shows the effect of LaPO4 on the drilling rate and mass loss of Al2O3/LaPO4 composites. It is obvious that the machinability of the composites can be improved by increasing the LaPO4 content. Other important factors still under study include the wear on the tool, the cutting force vs drilling rate, the surface roughness of the workpiece, and the diffusion mechanism between the tool and the workpiece. The reduced hardness and improved machinability are attributed primarily to the crack deflection process. It can be seen in Fig. 13.8 that the composite showed obvious particle pullout and significant crack deflection along interphase boundaries due to the weak interface bonding. The crack deflection mechanism (absorbing fracture energy and blunting crack tip) could lead to an increase in machinability. As described above, the thermal expansion
© Woodhead Publishing Limited, 2006
344
Ceramic matrix composites 0.25 0.24 0.22
0.15
Drilling rate (mm/s) Mass loss (g) Drilling time: 30 second Drilling force: 49 N
0.20 0.18 0.16 0.14 0.12
0.10
0.10
Mass loss (g)
Drilling rate (mm/s)
0.20
0.08 0.05
0.06 0.04 0.02
0
0 0
20
40 60 LaPO4 content (wt%)
80
100
13.7 Effect of LaPO4 content on the drilling rate and mass loss of Al2O3/LaPO4 composites.
mismatch between Al2O3 and the second LaPO4 phase can induce residual tensile stresses at the interface between Al2O3 and LaPO4 upon cooling down from sintering temperatures. This may result in microcracking to dissipate the strain energy and shield the main crack, thus avoiding catastrophic failure and leading to an increase in machinability.
13.4
Si3N4/h-BN
Silicon nitride (Si3N4) has great promise for engineering applications due to its excellent thermomechanical properties such as high bending strength, high thermal shock resistance and high fracture toughness [10]. However, it is quite difficult to machine after sintering. This disadvantage leads to high machining cost, and limits Si3N4 applications greatly. A variety of methods have been utilized to improve the machinability of Si3N4 ceramics, including porous Si3N4 ceramic [11], microstructure design [12–14], Si3N4/h-BN nanocomposites [15], etc. Reinforcement by nanoparticles is one of the methods to obtain good mechanical reliability. Kusunose et al. [15] reported that the Si3N4/h-BN nanocomposite materials possess good machinability.
13.4.1 Experimental procedure Starting powders (Si3N4, 0.7–1.2 µm, Founder Co., China; h-BN, 0.53 µm, GY Powder Industries, China) were used as raw materials. According to the
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
Mag = 20.00 K X
1µm
EHT = 10.00 kV WD = 10 mm
345
Signal A = SE2 Date: 8 Oct 2001
(a)
Mag=20.00KX
1µm
EHT = 10.00 kV WD = 10 mm
Signal A = SE2 Date: 8 Oct 2001
(b)
13.8 Fracture surface of 30 wt% and 40 wt% Al2O3/LaPO4 composites by hot pressing.
different proportions of Si3N4 and h-BN with 8wt%Y2O3–2wt%Al2O3 (Y2O3– Al2O3, >99.5%, GRINM, China) as sintering aids, the mixtures were carefully weighed, mixed and milled in a roller mill with ethanol for 48 h, then dried at 80°C and sieved through a 100 mesh sieve. More details of material preparation and characterization have been reported elsewhere [13, 14].
© Woodhead Publishing Limited, 2006
346
Ceramic matrix composites
13.4.2 Mechanical properties As is known, h-BN, it is difficult to completely densify at the sintering temperature that Si3N4 reaches its theoretical density, due to the chemical inertness and plate-like structure. Figure 13.9 shows the density change of Si3N4/h-BN composites with different h-BN concentrations. The reduction in bulk density with increased h-BN content is caused by two factors: the low density of h-BN (ρh–BN = 2.27 g/cm3; ρSi 3 N 4 = 3.19 g/cm3), and the lowchemically active nature of h-BN in Si3N4/h-BN composites. Therefore, hBN reduced the sinterability of Si3N4/h-BN composites, resulting in decrease in bulk density. The effect of h-BN content on Vickers hardness, flexural strength, and elastic modulus of Si3N4/h-BN composites was investigated. Figure 13.10 shows the relationship between composite hardness and h-BN content. The hardness of Si3N4 with 25% h-BN volume content is as high as HV = 5.67 GPa, which matches the request of engineering machining. The hardness is close to that of machinable mica glass–ceramic (3 GPa) and layered ternary compounds Ti3SiC2 (4~5 GPa). The Vickers hardness of Si3N4/h-BN composite decreased with increasing h-BN volume fraction. Easy cleavage of basal plane h-BN platelets causes hardness and fracture strength to decrease with h-BN addition as shown in Fig. 13.11. Along with the addition of h-BN, a steep decrease in the hardness of Si3N4/h-BN composites occurs due to the formation of the weak interface, leading to good machinability. Of course, an increase in porosity will also cause a decrease in hardness. Sometimes, Theoretical values Measured values
3.20
Density (g/cm3)
3.15 3.10 3.05 3.00 2.95 2.90 2.85 0
5
10 15 h-BN content (vol%)
20
13.9 Effect of the h-BN content on the density of Si3N4/h-BN composites.
© Woodhead Publishing Limited, 2006
25
Machinable nanocomposite ceramics
347
20 18 Si3N4/h-BN 1750°C (2h)
16
Hv (GPa)
14 12 10 8 6 4 0
5
10 15 h-BN content (vol%)
20
25
13.10 Effect of the h-BN content on the hardness of Si3N4/h-BN composites.
Bending strength (MPa)
750
Si3N4/h-BN 1750°C (2h)
700
650
600
550
500
0
5
10 15 h-BN content (vol%)
20
25
13.11 Effect of the h-BN content on the bending strength of Si3N4/hBN composites.
© Woodhead Publishing Limited, 2006
348
Ceramic matrix composites 260 250 Si3N4/h-BN 1750°C (2h)
Elastic modulus (GPa)
240 230 220 210 200 190 180 170
0
5
10 15 h-BN content (vol%)
20
25
13.12 Effect of the h-BN content on the elastic modulus of Si3N4/h-BN composites.
this is more important than a weak interface. However, another 25%h-BN/ Si3N4 sample sintered by spark plasma sintering still possesses excellent machinability even with high densification (99.6% theoretical density). So, when the addition of the weak phase is above 20%, it seems that the weak bonding has a greater effect than increased porosity in the sample, and dominates the improved machinability. As a weak phase, the addition of hBN will benefit the crack deflection and reduce the composite hardness. The reduction of composite sinterability also leads to decrease of flexure strength and increased porosity. It has long been reported that the bending strength of ceramics is governed by porosity, and the same result was observed in this study. An important reason is that Si3N4/h-BN is hard to densify due to the h-BN addition, which produces many pores in the microstructure. The SEM image of Si3N4/h-BN graded composites proved that the amount of pores in the specimen increases and the size of pore enlarges with increasing h-BN content [14]. At the same time, the pore orientation parallel to the HP direction is similar to those observed by other authors [16]. Figure 13.12 shows the dependence of elastic modulus of Si3N4/h-BN ceramic composites on h-BN content. From the result, high h-BN content resulted in reduction of elastic modulus of Si3N4/h-BN ceramic composites.
13.4.3 Microstructure and machinability Figure 13.13 shows the microstructures of sintered Si3N4 with and without the addition of 5, 10, 15, 20 and 25 vol% h-BN. SEM micrographs of the
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
6 µm
(a) Si3N4
6 µm (b) 5 vol% h-BN/Si3N4
6 µm (c) 10 vol% h-BN/Si3N4
6 µm (d) 15 vol% h-BN/Si3N4
6 µm (e) 20 vol% h-BN/Si3N4
6 µm (f) 25 vol% h-BN/Si3N4
13.13 SEM micrographs of Si3N4/h-BN composites with different addition of h-BN, showing: (a) Si3N4; (b) 5 vol% h-BN/Si3N4; (c) 10 vol% h-BN/Si3N4; (d) 15 vol% h-BN/Si3N4; (e) 20 vol% h-BN/ Si3N4; (f) 25 vol% h-BN/Si3N4. → ← direction corresponds to the hot-pressing direction.
© Woodhead Publishing Limited, 2006
349
350
Ceramic matrix composites
fracture surface perpendicular to the pressing direction confirm the preferred orientation of h-BN plates in the hot-pressed h-BN composites. A similar preferred orientation of h-BN grains in hot-pressed BN/Al2O3 [17], SiC/BN [18], BN/oxide ceramic [19], and BN/B2O3 has previously been observed. Figure 13.13 also shows that the microstructures of specimens of different hBN are different in comparison to that of monolithic Si3N4 ceramic. Elongated β-Si3N4 grains were observed in monolithic Si3N4 and 5 vol%Si3N4/h-BN composite, but were not observed in other samples. h-BN grew anisotropically in a plate-like configuration. The addition of h-BN retarded the growth of elongated β-Si3N4 grains as seen in the SEM photographs. A scanning electron micrograph of the resulting microstructure clearly shows the presence of acicular β-Si3N4 grains with high aspect ratios compared to the monolithic Si3N4. Every composite microstructure shows that the layered crystal structure and the preferential orientation of the flaked h-BN grains are perpendicular to the hot-pressing direction, which may be due to the rotation of the h-BN flakes during the viscous flow of the glass phase under hot pressing. h-BN has a layered structure similar to graphite, with strong bonding within each layer and weak bonding between the layers. When a crack tip meets an h-BN grain, it will propagate either along the interface between the Si3N4 and hBN grains or along the interlayer within the h-BN grains. In contrast, the crack deflection and propagation across an h-BN flake are difficult because of the strong bonding within each layer of the h-BN grains. The fracture surface of the composite shows quite a large area of crack deflection, pullout and cleavage cracking of the h-BN grains. All of these explain the improvement in machinability of Si3N4/h-BN composites. The mechanism of materials removal of 25 vol% Si3N4/h-BN composite during drilling using a cemented carbide drill seems to rely on the cleavage of layered h-BN crystals. During drilling, the cleavage of h-BN crystals localizes the fracture of composite at a microscopic scale and allows powdered chips to form easily, giving rise to good machinability. The intercrystal porosity can provide another reason for the improvement of machinability by preventing growth of cracks associated with machining. Moreover, it appears that the more porous a ceramic, the weaker, softer and more machinable it becomes. According to Fig. 13.9, the relative density decreased and porosity increased with the increase of h-BN content, which also resulted in the good machinability of Si3N4/h-BN composites. Figure 13.14 shows the TEM images of fracture surfaces for the 25 vol% Si3N4/h-BN composite. One can see that the plate-like h-BN grains were not wetted by the glass phase and had no obvious strong bonding with the Si3N4 matrix. In some areas, one can even see layered h-BN showing a bending direction perpendicular to the c-axis.
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
351
h-BN
Si3N4
Glass phase
150 nm (a)
C axis (002)
h-BN
150 nm
(b)
13.14 TEM observation of machinable 25 vol% Si3N4/h-BN composite.
13.5
Machinable nanocomposites
As shown above, the improved machinability with increasing concentration of a second weak phase normally will cause a decrease in fracture strength, except for the Al2O3/Ti3SiC2 system (here weak means that the hardness of Ti3SiC2 is less than that of Al2O3) [20]. How to combine high strength with good machinability is an inevitable consideration for the wide application of advanced ceramics, especially for structural ceramics. In recent years, many attempts have been made to develop strong machinable ceramics. Among them, a machinable nanocomposite is one of the most promising routes [21, 22].
© Woodhead Publishing Limited, 2006
352
Ceramic matrix composites
Because nanocomposites are made from different phases with different thermal expansion coefficients and elastic moduli, they inevitably develop residual thermal stress during cooling after sintering. Assuming the dispersion phase is spherical particulate in the matrix material, residual stresses can be developed due to differences in the thermal expansion and elastic constants between the matrix and the particles [23]: ( α m – α f ) E m ∆T σf = (13.2) E (1 + ν m ) + (1 + ν f ) m Ef σf (13.3) σt = – 2 where σf is the radial stress and σt is the tangential stress; αm and αf, Em and Ef, and νm and νf are thermal expansion coefficients, elastic moduli and Poisson’s ratios of the matrix and the dispersion phase, respectively; and ∆T is the temperature range during cooling down. Due to the stress concentration, crack propagation may be pinned by the nanoparticles near the crack tip. The schematic diagrams and TEM images in Fig. 13.15 show crack pinning by a nanoparticle and transgranular fracture induced by an intragranular particle. These experiments also show that crack pinning can give rise to nanoparticle pullout. When crack growth is pinned by a nanoparticle, the crack can penetrate through the nanoparticle by breaking it or can propagate along the nanoparticle and matrix. But due to the residual stress, the crack dominates over propagation along the phase boundary. The main mechanisms involve deflection of cracks or secondary cracking caused by the presence of a weak layer, weak interfaces, residual stress, or other microstructural defects. If the crack propagation is along the weak interface between the matrix and the soft phase, the main materials removal mechanism is grain pullout. This will avoid catastrophic fracture during machining. The weak bonding and weak phases are the main sources of energy dissipation and damage tolerance. In Al2O3/LaPO4, the interfacial toughness is sufficiently low to satisfy the criterion of He and Hutchinson for a normally incident crack to deflect along the interface rather than cross it. This system is stable for long periods in air at temperatures at least as high as 1600°C. Dislocations are the only line defect in a solid material. As we known, the movement of dislocations is the elementary mechanism of plastic deformation of many crystalline materials, especially in ceramics. A more detailed investigation is still needed to understand the relation between the improved machinability and formation of dislocations induced by residual thermal stress.
13.6
Conclusions
Design and fabrication of strong machinable ceramics is emerging as an inevitable requirement for flexible use of advanced ceramics, especially for
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics LaPO4 particle
353
Al2O3 matrix
Crack
(a)
100 nm
(c)
100 nm
(b)
100 nm
13.15 Schematic and TEM image of crack pinning by a nanoparticle and transgranular fracture induced by intragranular particle.
structural ceramics. However, the extremely high hardness of ceramics makes conventional machining very difficult or even impossible. In the past 10 years, many researchers have focused on the improvement of ceramic machinability. Normally, compound design of machinable ceramics is used to improve the machinability of ceramic materials. This method introduces a weak interphase or layered structure material into the matrix to facilitate crack deflection and propagation during machining and therefore avoid brittle fracture. However, the improved machinability with increasing concentration of a second weak phase normally will cause a decrease in fracture strength. At present, machinable nanocomposite design is considered one of the promising methods to combine high strength with good machinability. More
© Woodhead Publishing Limited, 2006
354
Ceramic matrix composites
investigation is necessary on processing procedures, sintering thermodynamics, residual thermal stress and microstructure details to understand the real reason for the improved machinability.
13.7
References
1. Niihara, K., New design concept of structural ceramics – ceramic nanocomposites, J. Ceram. Soc. Jap., 99(10): 974–982 (1991). 2. Parthasarathy, T.A., Boakye, E., Cinibulk, M.K., and Perry, M.D., Fabrication and testing of oxide/oxide microcomposites with monazite and hibonite as interlayers, J. Am. Ceram. Soc., 82(12): 3575–3583 (1999). 3. He, M.Y., and Hutchinson, J.W., Crack deflection at an interface between dissimilar elastic materials, Int. J. Solids Structures, 25(9): 1055–1067 (1989). 4. Davis, J.B., and Marshall, D.B., Machinable ceramics containing rare-earth phosphates, J. Am. Ceram. Soc., 81(8): 2169–2175 (1998). 5. Sudheendra, L., Renganathan, M.K., and Raju, A.R., Bonding of monazite to Al2O3 and TiO2 ceramics, Mater. Sci. Eng. A, 281(1–2): 259–262 (2000). 6. Wang, R.G., Pan, W., Chen, J., Fang, M.H., Cao, Z.Z., and Luo, Y.M., Synthesis and sintering of LaPO4 powder and its application, Mater. Chem. Phys., 79(1): 30–36 (2003). 7. Wang, R.G., Pan, W., Chen, J., Jiang, M.N., Luo, Y.M., and Fang, M.H., Properties and microstructure of machinable Al2O3/LaPO4 ceramic composites, Ceramic International, 29(1): 19–25 (2003). 8. Wang, R.G., Pan, W., Chen, J., Fang, M.H., Jiang, M.N., and Cao, Z.Z., Microstructure and mechanical properties of machinable Al2O3/LaPO4 composites by hot pressing, Ceramic International, 29(1): 83–89 (2003). 9. Davis, J.B., et al., Influence of interfacial roughness on fiber sliding in oxide composites with La-monazite interphases, J. Am. Ceram. Soc., 86(2): 305–316 (2003). 10. Rouxel, T., High temperature mechanical behavior of silicon nitride ceramics, J. Ceram. Soc. Jap., 109(6): 89–98 (2001). 11. Kawai, C., and Yamakawa, A., Machinability of high-strength porous silicon nitride ceramic, J. Ceram. Soc. Jap., 106(11): 1135–1137 (1998). 12. Kawai, C., and Yamakawa, A., Effect of porosity and microstructure on the strength of Si3N4: designed microstructure for high strength, high thermal shock resistance, and facile machining, J. Am. Ceram. Soc., 80: 2705–2708 (1997). 13. Wang, R.G., Pan, W., Chen, J., Jiang, M.N., and Fang, M.H., Fabrication and characterization of machinable Si3N4/h-BN functionally graded materials, Mater. Res. Bull., 37(7): 1269–1277 (2002). 14. Wang, R.G., Pan, W., Jiang, M.N., Chen, J., and Luo, Y.M., Investigation of the physical and mechanical properties of hot-pressed machinable Si3N4/h-BN composites and FGM, Mater. Sci. Eng. B, 90(3): 261–268 (2002). 15. Kusunose, T., et al., in Innovative Processing and Synthesis: Ceramic, Glass and Composite, edited by Bansal, N.P., Logan, K.V., and Singh, J.P., Westerville, OH: American Ceramic Society, 1997, pp. 443–454. 16. Oliveira, F.J., Carrapichano, J.M., Silva, R.F., and Vieira, J.M., Sintering of Si3N4BN composites, Sil. Ind., 63(5–6): 69–72 (1992). 17. Goeuriot-Launay, D., Brayet, G., and Thevenot, F., Boron nitride effect on the thermal shock resistance of an alumina based ceramic composite, J. Mater. Sci. Lett., 5: 940– 942 (1986).
© Woodhead Publishing Limited, 2006
Machinable nanocomposite ceramics
355
18. Ruh, R., Bentsen, L.D., and Hasselman, D.P.H., Thermal diffusion anisotropy of SiC/BN composites, J. Am. Ceram. Soc., 67: C-83 (1984). 19. Trice, R.W., and Halloran, J.W., Investigation of the physical and mechanical properties of boron nitride/oxide ceramic composites, J. Am. Ceram. Soc., 82(9): 2563–2565 (1999). 20. Luo, Y.M., Li, S.Q., Chen, J., Wang, R.G., Li, J.Q., and Pan, W., Effect of composition on properties of alumina/titanium silicon carbide composites, J. Am. Ceram. Soc., 85(12): 3099–3101 (2002). 21. Kusunose, T., Sekino, T., Choa, Y.H. and Niihara, K., Machinability of silicon nitride/ boron nitride nanocomposites, J. Am. Ceram Soc., 85(11): 2689–2695 (2002). 22. Wang, X.D., Qiao, G.J., and Jin, Z.H., Fabrication of machinable silicon carbide– boron nitride ceramic nanocomposites, J. Am. Ceram Soc., 87(4): 565–570 (2004). 23. Li, R.T., PhD dissertation, Tsinghua University, Beijing, 2002.
© Woodhead Publishing Limited, 2006
Part IV Refractory and speciality ceramic composites
© Woodhead Publishing Limited, 2006
14 Magnesia–spinel (MgAl2O4) refractory ceramic composites C A K S E L, Anadolu University, Turkey and F L R I L E Y, University of Leeds, UK
14.1
Introduction
The spinels are a class of double oxide of general formula AB2O4: industrially important members of this class include aluminates (e.g. MgAl2O4), ferrites (e.g. MgFe2O4) and chromites (e.g. MgCr2O4). Magnesium aluminate spinel (MgAl2O4) is an important constituent of magnesia-based refractory materials. The melting point of MgAl2O4 is 2135°C. There are no natural deposits of MgAl2O4, which is therefore normally obtained by reaction of mixtures of magnesium and aluminium oxides. The theoretical stoichiometric composition of magnesium aluminate spinel is 71.68% Al2O3 and 28.32% MgO by weight, but compositions can vary because of a limited range of solid solution. Its density is 3.579 Mg m–3, approximately the same as MgO (3.583 Mg m–3). Commercial sintered magnesia–spinel refractory materials are divided into three categories: magnesia rich, stoichiometric, and alumina rich.1 Typical properties of magnesia rich spinel bricks are given in Table 14.1.2 Spinel is used as an additive in magnesia rich bricks, for example for cement kiln linings. Magnesia rich spinel bricks are used in the cooling zone and in the upper side of the sintering zone of the cement kiln.3,4 In addition, spinel particles are added in various proportions to MgO in order to improve its thermal shock resistance. Magnesia–spinel materials give significantly (two to three times) longer service lives than other basic bricks such as conventional magnesia chrome.2 The reason for the improvement in thermal shock resistance is linked to the large difference in thermal expansion coefficient5 between magnesium oxide (mean value ~13.5 MK–1) and spinel (~7.6 MK–1). This difference leads to the development of large tensile hoop stresses, and eventually extensive microcracking, around spinel grains during cooling from fabrication temperatures in excess of 1600°C. Thermal shock is particularly severe during the kiln heating and cooling periods, when temperature changes can be large and rapid, leading to significant thermal 359 © Woodhead Publishing Limited, 2006
360
Ceramic matrix composites Table 14.1 Typical properties of magnesia rich spinel bricks2 Properties
Magnesia–spinel brick with alumina
Magnesia–spinel brick with sintered spinel
MgO (%) Al2O3 (%) Fe2O3 (%) CaO (%) SiO2 (%) Bulk density (Mg m–3) Apparent porosity (%) Cold crushing strength (N mm–2) Refractoriness under load (°C) Thermal shock resistance 950°C/air Thermal conductivity (W m–1 K–1) 500°C 1000°C
96 – 86 3–8 0.1 – 4 0.4 – 2 0.2 – 4 2.85 – 2.95 16 – 20 30 – 90
90 – 80 9 – 18 0.1 – 0.5 0.4 – 1.0 0.2 – 0.6 2.90 – 3.00 14 – 18 40 – 80
1550 – 1700 40 – 100
>1700 >100
3.0 – 4.0 2.4 – 3.0
3.0 – 4.0 2.8 – 3.7
stresses. Some transient healing of the cracks may occur on subsequent reheating during use of the refractory.1,6–8 The stoichiometric type is similarly used in cement kiln linings, as well as an alumina-based castable in ladles, which can also have good resistance to corrosion and erosion.9 The alumina rich type is being studied for use in alumina-based castables in order to improve resistance to slag* penetration; 85–90 wt% Al2O3 is suggested as the optimum. A higher alumina content increases the spinel bonding and improves strength, and increased spinel bonding results in better spalling resistance.10 Alumina rich magnesia–spinel is also used in steel and aluminium production, petrochemical applications, and glass melting furnaces.3 A major reason for the use of MgAl2O4 compared to other spinels (such as magnesia–chrome) is its better resistance to thermal shock and alkali attack2 (Fig. 14.1). Magnesia–chrome refractories were used for many years as high strength hot-face refractories in a range of systems, including rotary cement kilns and steel-making vessels,11 but there is risk of the contamination of ground water by hexavalent (VI) chromium ions leached from waste materials. Hence, a second significant reason for the escalating interest in MgAl2O4 is to avoid the use of refractories containing chromium oxide. Magnesium aluminate spinel is therefore being used as an alternative second phase in magnesia–spinel and alumina–spinel refractories, allowing a wide *Slag is the product of reaction between fluxing materials and a refractory furnace lining116
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites Resistance to
Dolomite
Magnesia–chromite
361
Magnesia–spinel
Reducing atmosphere Free SO2/O3 CO2 Free K2O/Na2O Clinker liquids K2SO4 KCl Thermal shocks Stress (kiln shell) Abrasion = very good
= good
= average
= low
14.1 Comparison of resistance of dolomite, magnesia–chromite and magnesia–spinel to environmental attack.2
range of compositions and types for a large number of applications to be produced.9,12,13 The environmental hazard posed by the conversion of insoluble trivalent (III) chromium to the soluble hexavalent state allows Cr(VI) ion leaching from waste magnesite–chrome refractories. Cr(VI) has been associated with allergic skin ulceration and carcinomas in humans.11 When chrome ore reacts with alkalis to form potassium chromate or potassium dichromate (containing Cr(VI)), this can result in the destruction of the brick.5,14 Hexavalent chromium diffuses from the refractory into the cement clinker, and increases the risk of toxic reactions during processing of the cement.2 In cement kilns, low viscosity clinker can also react with magnesia–chrome spinel to form relatively low melting point compounds, especially monticellite (CaO.MgO.SiO2).15,16 There are eutectic points at 1360 and 1490°C in the monticellite region.17 In the CaO–Al2O3 binary system near the composition CaO.6Al2O3, there is a eutectic point at 1395°C.18 The spinel content must therefore be kept to a minimum, and it is necessary to establish the optimum amount. One widely used type of refractory is based on material consisting of coarse (mm) grain size magnesite, bonded with a (µm) grain-size magnesite– spinel phase. The main advantages6,7,13,19-22 of using magnesia–spinel bricks in cement kilns can be summarised as: • Low thermal expansion coefficient of magnesia–spinel bricks • High resistance to thermomechanical stress
© Woodhead Publishing Limited, 2006
362
Ceramic matrix composites
• Chemical resistance to oil and ash deposits • High resistance to corrosion and changes in kiln atmosphere • Low content of secondary oxides which results in minimal alteration in structure of the hot face in service • Elimination of chromite that makes the brick less susceptible to alkali attack in service • No toxic Cr(VI) ions leached from waste materials • White cements can be made without discoloration problems caused by transition metal cations.
14.2
Crystal structures
MgO has the NaCl crystal structure;23,24 each magnesium ion is coordinated by six O2– ions and each O2– by six Mg2+. The structure of α-Al2O3 consists of close-packed sheets of O2– ions stacked in the sequence A–B–A–B, forming a hexagonal close-packed array of anions.25 The cations are located in twothirds of the octahedral sites. The structure of α-Al2O3 results in coordination numbers of 6 and 4 for the cation and the anion, respectively.25 Spinels of the general formula AB2O4, such as magnesium aluminate spinel, MgAl2O4, are based on a face-centred cubic (fcc) array of oxygen ions.26 There are two types of spinel: in normal spinels the A2+ ions are on tetrahedral sites and the B3+ ions are on octahedral sites (MgAl2O4); in inverse spinels, the A2+ ions and half the B3+ ions are on octahedral sites, with the other half of the B3+ ions on tetrahedral sites, B(AB)O4: an example is FeMgFeO4.26 There are two kinds of interstice in a close-packed lattice:27 octahedral sites, defined by six ions, and tetrahedral sites, defined by four ions. In MgAl2O4, the oxygen ions are in a face-centred cubic, or cubic close packed, array. A unit cell contains eight Mg2+ ions, 16 Al3+ and 32 O2– ions; there are 32 octahedral interstices and 64 tetrahedral interstices.26 In the normal spinel, the octahedral sites are occupied by the smaller trivalent ions, and the tetrahedral sites are occupied by the larger divalent ions. There are twice as many filled octahedral sites as tetrahedral sites, corresponding to the numbers of trivalent and divalent ions in MgAl2O4: one-eighth of the tetrahedral and half of the octahedral sites are filled.26
14.3
Production of MgAl2O4
Commercially synthetic spinel is produced in four main district forms:12,28 • Electrofused spinel • Sintered spinel clinker* *Clinker is a material that has been fully pre-reacted at high temperature in granulated form.116
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
363
• Sintered magnesia–spinel clinker • Calcined spinel. In the fusion process calcined Bayer Al2O3 and MgO powders are mixed in stoichiometric ratio, and melted in an electric arc furnace at 2200°C. After cooling and milling a dense spinel is obtained. Sintered spinel clinker has been developed as a more economical raw material than fused spinel, which it replaces in many applications. Al2O3, MgO and a selected binder (organic polymer or inorganic) are pulverised and mixed in a tube mill. After forming and drying, the mixture is sintered at temperatures above 1600°C in a rotary kiln until fully dense. MgO–spinel clinker is obtained by the same procedure, but with a substantial proportion of free MgO (75–85%). MgO–spinel is produced by sintering at temperatures of 1600–1800°C and a dense product is obtained. MgO–spinel refractory has better high temperature mechanical properties in arc furnace roof application compared to magnesia–chrome refractories, but it has lower strength than a high-alumina refractory. High-density spinel refractory brick is made by calcining (1200–1300°C for about 3 h) compacted powders and then sintering at 1700°C. The extent of spinel formation increases by calcination; however, the powder mixture is not completely converted to spinel. Typically 10–15% of α–Al2O3 and 5– 10% of MgO are observed by XRD, depending on the conditions of temperature, time and particle size of the reactants.
14.4
Densification
Grain size and boundaries, impurities, additives, particle size, porosity, sintering temperature and time, heating and cooling rate, and shaping practice play an important role in controlling many physical, mechanical and chemical properties of magnesia-based bricks.29
14.4.1 Sintering of MgO The main difficulty in sintering of MgO is its chemical instability with respect to water and atmospheric humidity. Hydration of MgO causes extensive swelling. Dehydration during drying and in sintering is therefore connected with an extremely large shrinkage, which can be associated with the formation of cracks in sintered materials. Precalcination of the MgO powder (especially submicrometre powder) makes the initial material less susceptible to hydration. The use of alcohol rather than water as a mixing or milling medium also reduces hydration.30 The sinterability of MgO powder depends on powder purity, particle size, precalcination stage and temperature.31 MgO of purity ranging from 98.5%
© Woodhead Publishing Limited, 2006
364
Ceramic matrix composites
to 99.3%, precalcined at 600–1300°C and sintered at 1650–1870°C, gives material of density 96–98%.30,32,33 Hot-pressing >50 nm MgO powder at 1600°C and ~15 MPa for 1 h gives material of relative density ~98%.34 A 10–50 nm MgO powder containing 0.5–1 µm agglomeration and of purity >99.9% sintered at 1450°C for 5 h gives 95% dense material of grain size 20–30 µm.35 Rapid grain growth leads to the entrapment of pores; pores far from the grain boundaries will disappear more slowly than those on the boundaries. For this reason, the location and size of the closed porosity play an important role in controlling the maximum sintered density.
14.4.2 Sintering in the MgO–Al2O3 system MgO and Al2O3 powders of ~1 µm particle size react to form spinel and sinter, but it is difficult to obtain full sintered density even at temperatures as high as 1750°C. If high density is to be obtained, precalcination at a lower temperature (1300°C for 3 h) is essential.32,36,37 This stage allows almost full reaction to spinel to occur. For 55–79% of pre-reacted spinel, final densities of 95–96% can be obtained at 1700–1750°C in 15–60 min. The minimum temperature required to obtain a fully dense product increases with Al2O3 content from the stoichiometric spinel (71.68 wt% Al2O3) to 85 wt% Al2O3, requiring from 1650°C to ~1680°C. The sinterability of spinel thus depends markedly on stoichiometry; it then decreases steadily to 1650°C with further increased alumina content (93 wt%).38
14.4.3 Solid state reactions The formation of spinel from magnesia and α-Al2O3 is associated with a volume expansion of ~8%.39 Solid state reaction between MgO and Al2O3 to form spinel occurs at the Al2O3–MgAl2O4 and MgO–MgAl2O4 interfaces. The reaction between these is by counterdiffusion of the Al3+ and Mg2+ ions through the rigid oxygen lattice or the spinel phase. Three Mg2+ ions diffuse for every two Al3+ ions that diffuse in the opposite direction. Three moles of spinel are formed at the Al2O3–MgAl2O4 interface for every mole formed at the MgO–MgAl 2 O 4 interface because of ionic charges and stoichiometry.32,33,40,41 This is illustrated by the following schematic section through a growing MgAl2O4 layer. I
II
MgO
MgAl2O4
→ 3Mg → 2Al + 4MgO 2+ 3Mg + 4Al2O3 → 2+
At I At II
3+
© Woodhead Publishing Limited, 2006
Al2O3
2Al ← MgAl2O4 + 3Mg2+ 3MgAl2O4 + 2Al3+ 3+
Magnesia–spinel (MgAl2O4) refractory ceramic composites
365
14.4.4 Liquid phase sintering in the MgO–spinel system In the MgO–spinel system, the diffusion rates of Mg2+ and Al3+ are accelerated by adding components of the liquid-forming system containing CaO, SiO2 and Fe2O3; and densification can be achieved with 2–3 wt% of additive at 1600°C. The rate of grain growth is decreased because these additions restrict discontinuous grain growth.42,43 Using CaO and SiO2, the densities obtained at 1600°C for ~30 min were equal to those obtained without addition at 1800°C for ~30 min.36
14.4.5 Grain growth in the MgO–spinel system Approximately 200 nm 99.8% purity spinel powder begins to densify at temperatures above 1100°C. Neck growth between grains takes place at 1100°C and continues up to ~1500°C because of the slow changes in grain size distribution. Grain growth begins to take place at around 1100°C and increases significantly from 1500°C up to 1625°C; however, the grains stop growing on further heating to 1700°C.44 The average grain size at 1500°C is twice as large as at 1100°C. A halt of grain growth above 1625°C can be assumed to be because the migration of grain boundaries is restricted by the presence of pores.44 Spinel particles are precipitated on cooling largely at the MgO grain boundaries, but some are trapped within growing MgO grains.28
14.5
In situ formed/preformed spinel-based refractories
In situ formed spinel-based refractories use the addition of an alumina source to form the spinel in situ by reaction with the magnesia matrix during sintering. A strong bond between spinel grains and the magnesia matrix is formed.45 Initial spinel formation occurs around the periphery of the alumina particle and proceeds towards the particle centre.46 The strongly bonded peripheral spinel and a hollow core in these granules is claimed to give better fracture toughness.46 The origin of the hollow core is not clear. The strength and high-temperature fracture characteristics depend strongly on the level of impurities in the magnesia, and the type and distribution of secondary phases.45,47 The complete conversion of the granular alumina is slow during manufacture of the spinel. Residual-free alumina will continue to form spinel with associated expansion in service, at a rate that depends on temperature and time. A smaller alumina particle size gives faster spinel formation, and the product magnesia–spinel composite also has better strength.33,45,46,48,49 The second type, which is technologically more advanced, is usually made by mixing 10–25% sintered or fused synthetic spinel with 75–90%
© Woodhead Publishing Limited, 2006
366
Ceramic matrix composites
sintered magnesia clinker. Sintered magnesia–spinel clinker is formed by heating at ~1700°C.50 In contrast to the in situ reaction of alumina, which develops a strong spinel bond with magnesia, preformed spinel grain has little tendency to bond to the magnesia. Preformed spinel in a magnesia matrix has a peripheral gap around the spinel grain, which appears to act as a crack initiator, and cracks propagate into the surrounding magnesia matrix.46 Cracks often emanate from sharp corners. Fracture is arrested when the crack meets the peripheral gap of another spinel granule.46 The properties of MgO–spinel refractories containing preformed spinel are more strongly influenced than the in situ formed spinel refractory by sintering temperature. Fracture toughness increases significantly with increasing sintering temperature; however, the Young’s modulus decreases.46 The hightemperature behaviour of spinel is explained in Section 14.6.4.
14.6
Industrial applications and properties of magnesia–spinel materials
14.6.1 Type and effect of the additives in magnesia– spinel refractories (CaO, SiO2 and Fe2O3) Keeping the level of CaO low prevents the formation of low melting point calcium aluminate, which can lead to fractures and destruction of the lining, in the magnesia–spinel brick structure during firing.51 A carefully adjusted CaO content (as shown in Table 14.2) forms a protective coating in the sintering zone. A larger amount of silicate in the microstructure forms interconnected silicate networks, which can act as crack propagation paths and decrease the resistance to spalling.6 The optimum value of the CaO/SiO2 ratio and reasons are given in Table 14.2.4 Table 14.2 Brick components of basic magnesia–spinel bricks, their most important reactions with the kiln feed and volatile components from the kiln atmosphere and the resulting demands on the bricks4 CaO/SiO2 (C/S) ratio (wt%)
I. C/S >> 1.87 CaOfree + MgAl2O4 → calcium aluminates CaOfree + SO3 → CaSO4
Decrease of refractoriness,* abrasion† Infiltration
II. C/S = 1.87
Formation of highly refractory compounds; few reactions
III. C/S << 1.87 2MgO.SiO2 + 2CaO.SiO2 → 2(CaO.MgO.SiO2)
Decrease of refractoriness
CaO/SiO2 in the brick approximately 1.7–2.2 if S(C+S) > 1.25
*Refractoriness is the ability of a material to withstand high temperatures that is evaluated in terms of the Pyrometric Cone Equivalent.116 †Abrasion is wear caused by the mechanical action of a solid.116
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
367
A reducing/oxidising atmosphere is important for the performance of the bricks, especially in kilns fired with solid fuel, since the reactions seen in Table 14.2 are combined with dramatic changes in volume, which result in the destruction of the brick where unfired particles can lead to reducing atmosphere. This condition affects the brick severely and may lead to premature wear. A magnesia–spinel brick for the cement industry should therefore have as little Fe2O3 as possible (preferably under 1%).52 Service life in the transition and the cooling zone of the rotary cement kiln compared to other basic bricks is longer. This type of magnesia–spinel brick can also be used in the upper side of the sintering zone of the cement kiln.53 The effect of phase changes from MgO–FeO to MgO–Fe2O3 on the bonded texture might be disadvantageous in developing high thermal spalling resistance.54 A reducing atmosphere accelerates the decomposition of alkaline salts, which causes their corrosive power to be increased.55 When MgO–FeO is oxidised, the volumetric expansion is about 8–23%, depending on the amount of FeO. The dissolution of Fe2O3 in MgAl2O4 leads to the formation of cation vacancies, which introduces defects in the host lattice. During ferrite formation the reaction rate depends on temperature, grain size distribution, density and also the partial oxygen pressure because the concentration gradients over the reaction layer increase with decreasing partial oxygen pressure. The ferrous content of the final product depends on the homogeneity of the starting mixture.56,57
14.6.2 Applications in rotary cement kilns A rotary cement kiln can be divided into five zones,2,58–60 which are taken sequentially from the beginning (inlet cone) where the raw material is introduced. Temperatures in this zone are in the 800–1000°C range and the refractory installed is ordinarily an aluminous firebrick. The calcining zone uses the low heat transfer properties of thermally insulating MgO–spinel brick, and high-alumina brick. This is where the corrosive effects of volatiles and alkaline salts manifest themselves. Semi-insulating brick may also be used for lining in this zone, depending upon its mechanical properties. The (upper) transition zone is characterised by mechanical stresses, thermal shock and chemical attack. The chemical attack occurs by reaction between the refractory and the silica in the clinker melt, and also alkali–sulphate interactions with volatile species in the kiln atmosphere. Unstable clinker coating leads to thermal spalling of the refractory lining. The sintering zone (the burning zone) is characterised by high-temperature (1500–1800°C) corrosion and extreme wear on the refractory lining. For this zone magnesia–spinel or magnesia–chrome bricks, or less commonly magnesia-enriched dolomite, are used. The burning zone of a rotary cement kiln is normally lined with basic bricks. In this area, the lining quality must be capable of retaining a
© Woodhead Publishing Limited, 2006
368
Ceramic matrix composites
stable clinker coating, which is built up on the lining to afford protection, to create a barrier between the hostile elements and the refractory to prevent erosion. Although the kiln temperature is highest in this region, the thermal gradient is not excessive and the refractories lining this zone are therefore not subjected to the same degree of thermal shock. Dolomite-based refractories are the dominant form of brick used in this zone, due to their ability to maintain a clinker coating but also for their high degree of refractoriness. Less frequently magnesia–spinel or magnesia–chrome bricks are employed. The cooling zone (lower transition zone) also depends on the abrasion and spalling resistance of MgO–spinel brick. It is believed that conditions are more severe in the transition zones where the coating is thinner and less stable, because thermal gradients are higher than in the other zones, resulting in extreme thermal shock damage. The service life in the lower transition zone is particularly severe due to the higher operating temperatures. Service life using the magnesite–spinel refractory in the lower transition zones has improved. For example, magnesia–spinel brick lost at least 60% of the original lining thickness in a one-year campaign, however, service life using magnesite– chrome bricks is approximately eight months.61,62 Brick wear in solid-fuel firing is accelerated by falling of the coating caused by the instability of the flame. Coating loss is considered to be a frequent occurrence in rotary cement kiln transition zones. When coating loss occurs, the refractory lining is subject to a very rapid rise in temperature, which can cause thermal shock damage to the refractory lining.61 Accumulation of alkalis from the fuel and reducing atmosphere is caused by unburned carbon from the precalciner. The alkali minerals volatilise in the high temperature burning zone and are carried by the gas stream to the feed end of the rotary kiln where they condense and return to the feed. This cycle within the cement kiln has a concentrating effect on the alkalis. It is common to find considerable concentrations of alkali salts deposited in the brick pores when examining basic brick taken from cement kiln burning zones. These alkali salts penetrate through the pore structure and freeze towards the cold end of the brick. When chrome is present in the basic brick, it is also possible to form alkali chromates. Alkali attack of magnesite– chrome brick deteriorates the bonding, causing loss of strength. MgO–spinel bricks are better than other basic refractory bricks in solid-fuel and oil-fired cement kilns. The life of MgO–spinel bricks was more than 1.5–2 times that of direct-bonded magnesia–chrome bricks in each zone, especially in the cooling zone and upper transition zone. The use of MgO–spinel refractory materials in the sintering zone of cement kilns is also being considered.61
14.6.3 Problems in the rotary cement kiln Service conditions, selection and installation of refractories, quality of raw materials and refractory materials are important parameters in increasing the
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
369
lifetime of refractory materials. Generally, problems in a rotary cement kiln can be classified into three groups:51,63–65 • Thermal problems: thermal shock, excessive thermal load, infiltration of silicates, migration of silicates • Mechanical problems: thermal expansion, concentric stress cracks, displacements, deformation of the kiln shell, formation of grooves, forces originating from retaining rings • Chemical wear problems: infiltration of alkali salts, redox effects, hydration cracks, corrosion of chrome ore. The function of the refractory lining in the rotary kiln is to resist thermal stresses and withstand the abrasive effects of the cement clinker. Thermal stress fracture has been attributed to rapid heating and cooling of the refractory lining component. The temperature gradient during rapid heating or cooling develops thermal stresses. The lining reduces maximum tensile stress during heating. Fracture of the refractory component results in lining deterioration. During cooling of a hot lining, faster contraction is observed on the hot face of the lining than on the centre of the brick. The maximum tensile stress occurs at the (cooled) surface, whereas the maximum tensile stress during heating is in the centre and is half the maximum tensile stresses during cooling.66 Further cracks are more stable when produced by cooling than those produced by heating. Therefore a higher temperature difference is needed to initiate crack propagation on heating but crack propagation is more catastrophic.46 Stress begins to build near the hot face at low temperatures. The Young’s modulus of a brick is lower at the maximum hot face temperature of ~1450°C. Thus the stress is reduced near the hot face. The maximum stress is transferred towards the interior of the brick as the hot face temperature rises. High stress is developed towards the hot face of the brick. The lower the thermal conductivity, the higher the rate of thermal expansion, and the higher the Young’s modulus, the greater the stress.67 During the steady state heat flow conditions in cement kiln operation, high levels of thermal stress should not occur. Critical levels of thermal stress for fracture would be associated with heat flow conditions of the heating and cooling periods. Sudden changes in temperature produce thermal tensions contributing to breakage of bricks. Temperature distributions in non-steady-state conditions are mostly noticeable during the initial stage of heating when temperature differences through the lining are greatest. These can become particularly severe when the initial heating by oil is changed to solid fuel, which can result in a sudden and marked increase in the heating rate, sometimes resulting in thermal shock damage and expansion. This problem can be minimised by the practice of holding the kiln at about 800°C for a few hours during warm-up to allow a near-uniform temperature distribution to be
© Woodhead Publishing Limited, 2006
370
Ceramic matrix composites
obtained. Coating formation will decrease the temperature distribution through the lining, and loss of coating causes thermal shock due to flame impingement.68 Methods used to improve brick lifetime are the production of stable coatings in the burning zone, the control of ovality of the tyres and kiln alignment, avoidance of frequent kiln stoppages, and slow heating and cooling rates.62,69,70
14.6.4 Performance Effect of expansion mismatch on Young’s modulus at high temperatures Fired MgO–spinel brick has high resistance to thermomechanical* stress, which is a result of a combination of direct crystal bonding and the thermal expansion mismatch between magnesia and spinel grains. Direct bonding provides high temperature strength. The mismatch is produced by the difference in thermal expansion between the magnesia and the spinel.6,71 As spinel is added to MgO, a thermal expansion mismatch is introduced into the system and microcracking occurs. This lowers the Young’s modulus and high-temperature strength.72–75 There is a slightly descending trend when the temperature rises and a small peak is shown at ~1200°C, which is the result of expansion mismatch between the silicate phases and spinel. The later decrease in Young’s modulus is the result of softening of silicate phases. On further cooling, the expansion mismatch causes cracks, which result in a decrease in Young’s modulus and strength. On reheating from room temperature, the cracks close up and partial healing can occur.8,72 Thermomechanical behaviour of stoichiometric spinel at high temperatures Young’s modulus, fracture toughness and strength of stoichiometric spinel have been studied from room temperature up to 1200–1500°C.76–78 In the low-temperature region, this material behaves elastically, whereas in the high-temperature (>800°C) region, plastic deformation is possible. At less than 800°C, the fracture toughness (KIC) decreases with increasing temperature; however, KIC increases rapidly with increasing temperature above 800°C. The decrease in KIC for the low-temperature (<800°C) elastic region is explained by the decrease of Young’s modulus with increasing temperature. In the high-temperature region, strength decreases with temperature, while KIC increases from 800°C up to 1200–1500°C.76–79 This indicates that the fracture behaviour of spinel at temperatures greater than 800°C is governed by nonlinear elastic processes. Several explanations can be given for transition to *Thermomechanical properties, e.g. softening under load, creep in compression, refractoriness under load and thermal shock resistance.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
371
totally intergranular fracture over 800°C: for example, the loss of grainboundary coherency from the presence of impurities concentrated at the grain boundaries, and a crack-tip plasticity enhanced crack-propagation mechanism e.g. secondary cracking in the primary crack tip.76,77,80 At the sintering temperature, liquid phases are formed and would remain as glass at room temperature due to high cooling rates. Grain boundary microcracking was observed because of dislocation* and grain boundary sliding due to secondary glassy phase softening. Grain boundary microcracking can lead to a decrease of strength during fracture, even if toughness increases. The fracture path changes from intergranular at low loading rates to transgranular for high loading rates.79 Hot strength Hot strength of MgO–spinel is reduced in the presence of commercial spinel due to the low melting point of the impurity phases (CaO.6Al2O3 and monticellite). As the spinel content increases up to 40%, the hot strength increases almost linearly due to the decrease in the impurity content of MgO. However, for spinel amounts over 40%, there was no marked change in the hot strength, even though the impurity content in MgO (>2%) continued to decrease with further addition of spinel up to 80%.28 Slag resistance and corrosion Spinel-bonded magnesia shows better slagging and spalling resistance due to the development of strong solid solution bonds between MgO and spinel on firing at 1600–1800°C. The arc furnace slag resistance improves with the presence of 10% spinel in the magnesia, but no further improvement occurs until the spinel content exceeds 40%. The resistance to basic slag deteriorates sharply with the addition of more than 40% spinel.28 Magnesia–spinel bricks perform better than basic slags when compared with direct bonded magnesia– chromite and mullite refractories.81 Since the slag is the most corrosive component in the melt, its composition has a critical effect on the corrosion mechanism. Corrosion resistance in the rotary slag test is evaluated on thickness loss and the depth of penetration. The stoichiometry of the spinel used strongly influences the corrosion behaviour; for example, alumina-rich spinel addition increases the resistance to slag penetration and wear. Furthermore, the size of the spinel particles added also affects the wear rate: fine particle additions are more effective in *Dislocations and line defects are relatively easy to move by low stresses, making the crystals weaker and more susceptible to plastic flow.
© Woodhead Publishing Limited, 2006
372
Ceramic matrix composites
resisting corrosion than coarse-grained additions. However, spalling resistance is improved with coarse (300 µm) spinel addition, since fine additions result in excessive sintering shrinkage and consequent stress.82
14.7
Thermal shock
A major factor determining the importance of ceramics lies in their usefulness at high temperatures. Two types of experimental techniques are generally used in order to determine the minimum shock required to nucleate fracture (cracking), and the amount of the damage caused by thermal shock:83 the number of quenching cycles, and strength as a function of quench temperature and crack patterns. One of the common techniques for evaluating thermal shock resistance is the thermal cycling of materials until fracture occurs. The strength of the materials after thermal cycling is then compared with the original strength. The number of cycles necessary to cause a defined damage or weight loss is used as a measure of thermal shock resistance. The results of a thermal cycling test can be useful only when used with another thermal shock test, which indicates the relative degree of difficulty of crack initiation and propagation.84 The traditional standardised method used to characterise the thermal shock resistance of commercial refractory materials normally deals with quenching of bricks from high temperatures by using water, oil or air as a cooling medium. A slow decline in strength occurs in weaker, low Young’s modulus, materials; however, a sharp drop in strength is shown by a strong material subjected to greater than the critical shock.85 The critical quench temperature, ∆Tc, can be defined as the temperature drop required to produce cracking in half the specimens tested. Cracking is usually detected by the dye-penetrant method. The depth of cracking increases slowly with increasing quenching temperature, and this gives rise to a gradual fall-off in strength. The deepest cracks control the strength after quenching.83
14.7.1 Thermal shock parameters On the basis of these tests, two types of parameter have been proposed: thermal stress resistance, and thermal shock damage.52,86 Since the term ‘thermal shock resistance’ is used to describe both the nucleation of fracture by the thermal shock and the degree of damage by thermal shock, it is proposed here to refer to the resistance to nucleation by thermal shock as ‘thermal shock fracture resistance’ or ‘thermal stress resistance’ and to the resistance damage by thermal shock as ‘thermal shock damage resistance’.52,86 The first method expresses the difficulty of crack initiation. The second expresses the degree of possibility for further damage by crack propagation
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
373
after crack nucleation is measured. Correlations were then sought between these parameters and the observed thermal shock behaviour. It is clear that these methods are useful only for evaluating the relative degree of damage (i.e. the fraction of retained strength) by thermal shock.84 Thermal stress resistance parameters For material initially undamaged, the appropriate parameter expressing the tendency for cracks to be developed, and therefore strength to be lost, can be considered to be that for crack initiation. This has been expressed in terms of thermal stress resistance parameters.25,30,52,86–88 Kingery used the infinite slab symmetrically heated or cooled with a constant heat transfer coefficient to derive thermal shock fracture resistance parameters R, R′ and R′′ using the equations: R=
σ f (1 – ν ) Eα
(14.1)
R′ =
σ f (1 – ν ) k Eα
(14.2)
R ′′ =
σ f (1 – ν ) A Eα
(14.3)
where σf is the strength (normally taken to be the bend strength), E is the Young’s modulus, α is the mean thermal expansion coefficient of the composite, ν is Poisson’s ratio, k is the thermal conductivity, and A is a stress reduction term. The parameter R is applicable for the case of instantaneous change in surface temperature (infinite h) for conditions of rapid heat transfer; R′ is for a relatively low Biot modulus (β < 2) for conditions of slow heat transfer; R′′ is for a constant heating or cooling rate.88 R defines the minimum temperature difference to produce fracture under conditions of infinite heat-transfer coefficient, i.e. A = 1. The parameter R is inversely proportional to α. A low value of α is therefore essential for good thermal stress resistance. The coefficient of thermal expansion normally increases with increasing temperature; however, thermal conductivity decreases. Hasselman’s approach89 to thermal shock fracture is based on the conversion of released elastic energy into surface energy, which gives the thermal conditions for fracture initiation. The length of the crack and the conditions are also important for further crack propagation. Resistance to crack initiation can be maximised by achieving high strength and thermal conductivity, with low values of thermal expansivity and Young’s modulus. However, avoiding thermal fracture by increasing strength in order to make initiation difficult is dangerous because once initiated, crack propagation will be fast and catastrophic.
© Woodhead Publishing Limited, 2006
374
Ceramic matrix composites
Thermal shock damage resistance parameters Hasselman approached the problem of thermal shock resistance to damage by comparing the degree of damage in materials fractured by thermal shock, rather than fracture initiation. Once cracking is initiated, the maximum surface area (Smax) of the fracture face is limited by Smax ≤ U/γWOF, where U is the elastic stored energy per unit volume and γWOF is the effective surface energy or work of fracture per unit projected area of fracture face. He suggested that a thermal shock damage or toughness parameter involving U/γWOF should be useful for comparing amounts of cracking.83 The Griffith view is that a crack will start propagating and will continue to propagate as long as the elastic energy released from the stress field surrounding the crack is greater than the fracture surface energy.90 The extent of crack propagation is directly proportional to the elastic energy stored at fracture. In contrast, the R′′′ and R″″ parameters (see below) are inversely proportional to the elastic strain energy, and directly proportional to the effective surface energy.88 Hasselman derived the thermal shock damage resistance parameters R′′′ and R′′′′, expressing the ability of the material to resist crack propagation and further damage and loss of strength on thermal shocking:52,86
R ′′′ = E2 ⋅ 1 σf 1 – ν
(14.4)
E γ WOF R′′′′ = σ 2 ⋅ 1 – ν f
(14.5)
The parameter R′′′′ can be applied to compare the degree of damage of materials with widely different values of γWOF, such as brittle and ductile materials. R′′′ can also be used to compare the relative degree of damage of materials with similar crack propagation properties, i.e. the same values of γWOF.91 R′′′ is a simplified formula derived from R′′′′ by eliminating the term of γWOF energy. The criteria for minimising the extent of crack propagation, and for obtaining a low degree of damage,84 are high values of the Young’s modulus, Poisson’s ratio and surface energy, and low values of strength. For example, when a body with near-zero strength undergoes only a minimal amount of damage, it will still have a low value of strength after thermal shock, simply because of its initial low value of strength before thermal shock. Although high values of the thermal shock damage resistance parameter R′′′ and R′′′′ are desirable, it is clear that these parameters cannot be maximised by letting the strength (σf) approach zero. There must be some intermediate value of strength and a resulting degree of damage such that the strength (after thermal shock) remains within acceptable limits.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
14.8
375
Mechanical properties and thermal shock behaviour of magnesia–spinel composite refractory materials
There have been a number of reports regarding the similarities and differences of the fracture toughness of single-crystal (KS) and polycrystalline (KP) magnesium aluminate spinel ceramics. Polycrystalline values are larger than single-crystal values, e.g. KP may be 2–10 times KS.92 The values of singlecrystal bars gave overall averages of 1.0 ± 0.2 MPa m1/2. Toughness of polycrystalline MgAl2O4 was measured by various techniques; indentation fracture and various notched beam values range from 1.4 to 2.2 MPa m1/2, averaging ~1.9 ± 0.2 MPa m1/2.93 The polycrystalline to single-crystal fracture toughness (KP/KS) ratios are in the range > 1.5 to < 2.5 for Al2O3 and ≥ 2.4 to ≤ 3.2 for MgO.93 The single crystals have a KIC region at approximately 1 MPa m1/2, up to 800°C. However, at high temperatures (>800°C) they have a second region of rapidly increasing KIC values (~3 MPa m1/2) with temperature, because of increased crack tip dislocation mobility.77 At intermediate temperatures the polycrystalline data become indistinguishable from single-crystal values; however, above 1000°C they fall significantly (55%) below the rapidly rising single-crystal curves. This suggests in the fracture process zone fracture may occur by other mechanisms, such as grain boundary plasticity, which may cause homogeneous crack tip plasticity in the spinel phase. It is reported1 that fracture toughness and hardness values of sintered magnesium aluminate spinel are in the ranges of 1.9–2.7 MPa m1/2 and 9.3– 14.9 GPa respectively, with the additions of various amounts of alumina from 49% to 75%. For example, in magnesia–spinel refractory materials containing 72 wt% Al2O3, the room-temperature fracture toughness value has been reported to be 1.9 MPa m1/2. The fracture toughness of dense MgO (ρ = 3.5 Mg m–3) measured using the single-edge notched beam (SENB) technique is given as ~2.05 ± 0.05 MPa m1/2 for materials with grain sizes ranging from 6 to 8 µm.94 Furthermore, thermal conductivity values of commercial MgO and spinel are given as ~50 W m–1 K–1 and 15 W m–1 K–1 at 25°C and ~7 W m–1 K–1 and 5 W m–1 K–1 at 1000°C, respectively.1 In addition, room-temperature values of strength and Young’s modulus for commercial sintered spinel are also reported as 135 MPa and 260 GPa, respectively.1 The room-temperature strength of the hot-pressed pure MgO95,96 is found to be ~225 and 230 MPa for ~32 and 25 µm MgO grain sizes, respectively. The Young’s modulus of the pure hot-pressed MgO at room temperature is stated as ~290 GPa95 (~25 µm grain size), and 258 GPa94 (~7 µm grain size). The room-temperature fracture surface energy of MgO obtained by using SENB (for a grain size ~100 µm) has been reported to be 14 J m–2,97,98 and 15 J m–2.99 The fracture surface energy of dense spinel has
© Woodhead Publishing Limited, 2006
376
Ceramic matrix composites
been reported to be 4 ± 0.5 J m–2,77 and 5–7 J m–2.78,100 The measured work of fracture (obtained from the areas under the stress-strain curve) is also reported to be 35 J m–2 for polycrystalline fully dense MgO (3.56 Mg m–3) of ~100 µm average grain size.101 Some of the thermomechanical properties of MgO–spinel refractories are given in Table 14.3,28 though there is no information about thermal shock parameters, fracture surface energy, work of fracture and strength values. For example, the coefficient of thermal expansion decreases almost linearly with increasing spinel content. Hot strength increases significantly with the addition of spinel content up to 40% but further additions of spinel in general do not change the hot strength values markedly (Section 14.6.4). Young’s modulus falls to a minimum with increasing spinel content at 30–40% addition; however, Young’s modulus increases significantly with spinel content exceeding 40% (e.g. back to the MgO value). As the spinel become the continuous phase, Young’s modulus stayed approximately constant between 50 and 80%.28 Thermal shock damage caused by water quenching from 800°C decreases with increasing spinel content up to 40%.28 Compositions with higher than 40% spinel show less improvement due to the increase in initial Young’s modulus, in spite of a decreasing thermal expansion coefficient. As shown in Table 14.3, the best combination of properties is assumed to be achieved by the addition of 40% spinel.28 Using preformed sintered spinel higher than 40% leads to a reduction in refractoriness; however, spinel contents less than 10% Al2O3 lead to deterioration in thermal shock resistance.102 Various studies based on the industrial work have been made over the last 15 years, with the objective of developing magnesia–spinel materials of improved resistance to thermal shock and alkali attack. The work done so far has been mostly phenomenological, and little quantitative understanding of the function of the system variables has been developed. It is therefore still difficult to specify optimum compositions with confidence, and materials development is based largely on trial and error. Much of the magnesia–spinel refractory currently produced is used for cement kiln linings, where there are two conflicting requirements: for thermal shock resistance the optimum spinel content should be fairly high; for reduced calcium oxide attack, involving reaction with aluminium oxide and the formation of low-melting calcium aluminates, the spinel content must be as low as possible. It is clearly necessary to keep the spinel content to a minimum, while developing the maximum resistance to thermal shock.4 The optimum spinel content has been claimed to be as high as ~40% by previous researchers.46 Subsequent researchers have used a range of model, high purity, magnesia–spinel composites to examine in detail the effects of spinel particle size and quantity on thermomechanical behaviour, and suggested an optimum composite composition for a maximum resistance to further damage by thermal shock of ~20% of spinel.74,75,103 In spite of the importance of this refractory system,
© Woodhead Publishing Limited, 2006
Table 14.3 Properties of impure MgO–spinel systems28 Properties
Values of properties at nominal spinel content % 0 –3
Bulk density (Mg m ) Apparent porosity (%) Theoretical density (%) Hot strength (3-point loading at 1600°C) Failure stress (MPa) Coefficient of thermal expansion (MK–1) 20–1000°C 20–1500°C Young’s modulus (E) at 20°C (GPa) Thermal shock resistance Progressive quench test % drop in E quenching from 800°C into water *n.d. = not determined.
© Woodhead Publishing Limited, 2006
10
20
30
40
50
60
70
80
3.43 0.0 95.8
3.42 1.2 95.6
3.37 1.7 94.3
3.33 4.3 93.2
3.31 3.5 92.7
3.44 0.0 96.5
3.24 0.0 90.9
3.36 0.0 94.4
3.44 0.0 96.7
7.2
4.6
6.7
9.2
11.6
11.3
9.8
11.0
12.8
13.9 14.9 274
13.1 14.3 264
12.5 13.3 194
11.7 11.7 151
11.5 11.9 152
n.d* n.d* 256
9.6 10.0 213
9.6 9.7 239
9.1 9.5 247
80.8
72.0
38.3
24.8
22.7
58.9
41.7
44.5
54.6
378
Ceramic matrix composites
few fundamental underpinning scientific studies have been reported so far, apart from the most recent researches.45,71,73,74,103,104 It has already been reported that the predicted thermal shock parameters103,105 and γWOF/γi ratios105 are compared with the actual thermal shock behaviour, and the R′′′′ parameter and γWOF/γi ratios are found to be good indicators for a quantitative evaluation of the retained strengths of MgO–spinel composites. It is also stated that resistance to thermal shock damage in terms of the effect of particle size distribution of spinel particles on the basis of thermal shock parameters can be more strongly favoured with materials containing a significantly broader distribution of spinel particles, rather than with narrowly distributed spinel particles for which a much larger volume percentage is required to achieve a similar improvement.106 The precise conditions required for the generation of microcracks in this system, their mechanism of operation, relationships between microstructure and those mechanical properties likely to affect thermal shock behaviour, through a detailed examination of hot-pressed model composite materials, will be examined in depth in the following sections of this chapter. Emphasis is placed on establishing the nature and extent of microcrack development, and the relationships between composition and microcracking, and between mechanical properties and thermal shock behaviour. In general, attention is given to the different methods of forming the spinel (as in situ and preformed), including with the different sizes of the spinel particles. It is expected that this investigation will provide a platform for detailed modelling of both mechanical properties and thermal shock behaviour and high-temperature properties of magnesia–spinel composite materials, and develop guidelines to allow the optimisation of commercial magnesia–spinel refractory compositions and microstructures.
14.8.1 Strength and MgO grain size Figure 14.2 shows that 0.5 µm in situ spinel composites prepared from Al2O3 powder demonstrate a ~20% decrease in strength for up to 10% addition, but further increases do not significantly affect strength. Composites prepared from the preformed spinel powder (3, 11 and 22 µm) show a more marked decrease in strength with increasing amounts of spinel, for all particle sizes. Although the overall pattern of behaviour is similar to that shown by the in situ spinel composites, there is a significant (~55 to 75%) decrease in strength of MgO at 30% addition, probably because of more extensive microcracking in the composites.73,104,105 The larger the spinel particles, the more the decline in strength (Fig. 14.2). Spinel content, but predominantly spinel particle size, significantly affect the strength. It is reported45,71 that in all composites the mean grain size of MgO increases by a factor of 2–3 in the presence of 5–10% spinel; however,
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
Strength (MPa)
250
0.5 µm 11 µm
200
379
3 µm 22 µm
150 100 50 0 0
5
10
15 20 Spinel (%)
25
30
14.2 Bend strength as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
14.3 SEM micrograph: MgO containing 0.5 µm 5% in situ spinel.
further additions of spinel decrease the MgO grain size to that of pure MgO. The deceleration in grain growth with further additions of spinel (≥ 10%) is most probably because of the dominance of grain boundary pinning effects by the spinel grains.45 Both acceleration and deceleration are most marked for the 0.5 µm spinel particles, which are located at the grain boundaries (Fig. 14.3). There may be a minor effect of MgO grain size in influencing strength, but crack length, spinel particle size and content seem likely to be the major factors influencing strength.33
14.8.2 Young’s modulus Young’s moduli of spinel composites decrease with additions of spinel, for all particle sizes, where the influence of the spinel is greater, the larger the
© Woodhead Publishing Limited, 2006
380
Ceramic matrix composites
Young’s modulus (GPa)
300 250 200 150 100 0.5 µm 11 µm
50
3 µm 22 µm
0 0
5
10
15 20 Spinel (%)
25
30
14.4 Young’s modulus as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
14.5 SEM micrograph showing the microstructure of a 20% 22 µm preformed spinel composite.
particle size (Fig. 14.4). SEM observations show (in Figs 14.3 and 14.5) that coarse particles are associated with longer crack formation, compared to finer particles. The crack lengths also increase with increasing concentration of spinel.73,104 The greater the extent of cracking, the greater the decrease in Young’s modulus. The changes in Young’s modulus can be explained in terms of crack development and interlinking, and these pre-existing cracks are assumed to be the result of thermal expansion mismatch between the MgO and MgAl2O4 phases.33 When the volume fraction of spinel increases, more microcracking occurs and lower Young’s modulus values are obtained.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
381
2.5
KIC (MPa.m1/2)
2.0 1.5 1.0 0.5
0.5 µm 11 µm
0 0
5
3 µm 22 µm 10
15 20 Spinel (%)
25
30
14.6 Fracture toughness (KIC) as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
14.8.3 Fracture toughness and fracture surface energy Figure 14.6 shows that composites prepared from preformed spinel show a general marked decrease in toughness with increasing spinel content for all particle sizes. Values of KIC of preformed spinel composites decrease with up to 10% addition, and level out at ~1 MPa m1/2 with further additions. It appears that K1C is certainly much more sensitive to spinel content than spinel particle size, and within experimental error may be independent of particle size. On the contrary, K1C for the 5% 0.5 µm in situ spinel composites initially decreases, but further spinel additions bring the values back to those of pure MgO (Fig. 14.6). This can be attributed to the special microstructure of this material in which the spinel particles were located only along the MgO grain boundaries (Fig. 14.3), as compared to preformed spinel composites. (Fig. 14.5). The significant change in fracture path from transgranular to intergranular mode (explained in Section 14.8.5) is possibly the explanation for the marked increase in K1C (e.g. back to the MgO value) at higher spinel contents.45 As will be expected, this pattern of behaviour in fracture surface energy values calculated using experimental Young’s modulus values in general is found to be similar to that of K1C.45
14.8.4 Critical defect size The mean critical defect size (c), calculated from the Griffith equation,107,108 increases markedly with increasing spinel content (Fig. 14.7). For 0.5 µm in situ spinel composites, the defect size increases up to ~20%, and then decreases slightly with further additions of spinel. However, for the preformed spinel composites there is a gradual increase in defect size up to 30% spinel additions. The increase in defect size is in general much more marked at 20%, in
© Woodhead Publishing Limited, 2006
382
Ceramic matrix composites 60
0.5 µm 3 µm 11 µm 22 µm
50
c (µm)
40 30 20 10 0 0
5
10
15 20 Spinel (%)
25
30
14.7 Critical defect size (c) as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
comparison to the pure MgO, and the defect sizes of the in situ formed 0.5 µm spinel and preformed 22 µm spinel composites increase approximately by factors of 2 and ≥ 1.5 respectively. The main result obtained from Fig. 14.7 is that the general trend is for the defect size to increase with additions of spinel. The highest increase in the defect size of 0.5 µm composites is possibly because of a grain boundary pinning effect of the spinel particles located at the MgO grain boundaries (Fig. 14.3). It is suggested45,71 that spinel content (predominantly) and spinel particle size (weakly) are important factors determining the critical defect size. MgO cleavage, grain boundary strength, pre-existing cracks controlling effective grain boundary energy, and pores, may be other parameters determining critical flaw size. However, particle interaction coupled with thermal expansion mismatch causing microcracking and interlinking that leads to a marked increase in the critical defect size, seems most likely to be the important parameter.71
14.8.5 Work of fracture For the in situ formed 0.5 µm spinel composites, the work of fracture (γWOF) value increases only slightly with up to 10% additions, but at 20% it increases by a factor of 2.5 (Fig. 14.8). There is a general but less marked increase in γWOF, by a factor of ~2, at 30% additions of preformed spinel (Fig. 14.8). It seems that the effect of preformed spinel content is more important than particle size, within the larger scatter data. Fracture surfaces of pure MgO show a large proportion of transgranular cracks, with a few intergranular cracks (Fig. 14.9). At low additions of spinel, transgranular cracks are still present with some intergranular cracks, in the fracture surfaces of each spinel composite. However, at higher additions of spinel (≥ 20%), mostly intergranular fracture occurs. For example, 20%
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
383
Work of fracture (J m–2)
100 80 60 40 0.5 µm 3 µm 11 µm 22 µm
20 0 0
5
10
15 20 Spinel (%)
25
30
14.8 Work of fracture (γWOF) as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
14.9 Fracture surface of dense MgO.
22 µm spinel composite (Fig. 14.10) shows mostly intergranular fracture with some transgranular. It therefore appears that higher values of γWOF are associated with the occurrence of more intergranular fracture with increasing spinel additions that requires more energy to propagate a crack completely through the specimen. The fracture of the magnesia–spinel composites is either semi-stable or stable, but never catastrophic.105 It may be concluded that crack propagation is a much greater energy-consuming process than crack initiation in these materials. For many industrial applications, the initiation of fracture is less
© Woodhead Publishing Limited, 2006
384
Ceramic matrix composites
14.10 Fracture surface of composite containing 20% 22 µm preformed spinel.
important than γWOF and the degree of damage (e.g. further loss of mechanical integrity by strength and material loss through large-scale fracture behind the hot face).109 Large values of the γWOF/γi ratio are obtained in these composites.105 This is a basic requirement for refractory materials to show good thermal shock damage resistance.110
14.8.6 R and R ′′′′ parameters Figure 14.11 shows that the R parameter for 22 µm spinel composites decreases with additions of up to ~20%, and then increases with further additions of
R parameter (K)
80 0.5 µm 11 µm
60
3 µm 22 µm
40 20 0 0
5
10
15 20 Spinel (%)
25
30
14.11 R parameter as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
385
spinel back to the value for pure MgO, within experimental scatter. For the composites containing lower particle sizes of spinel, the R parameter initially decreases slightly with spinel content, and reaches a minimum at about 10% addition: further additions of spinel result in increases in the R parameter. For example, 30% of 0.5 µm in situ formed spinel composite gives an R parameter ~50% greater than for pure MgO. It appears that both strength and Young’s modulus are controlled by the extent of microcracking with strength being influenced more strongly, until very high spinel contents are reached. It cannot be expected on this basis that spinel additions will improve the thermal shock resistance of MgO through improvement of resistance to crack initiation.74,103 Refractories are not very resistant to crack initiation, but have a significant resistance to crack propagation or extension. Crack propagation is much more difficult in refractories than the initiation of cracks.111 The fracture mechanism in magnesia–spinel refractories relies on the development of microcracks that allows easy crack initiation but makes propagation, in which fracture occurs in a quasi-static manner, more difficult.46 The main concern by assessing the service performance of materials (e.g. strength loss) is the resistance to crack propagation and to extension of damage caused by thermal shock, rather than resistance to crack initiation.112–115 Therefore, there is a basic requirement to investigate how the R′′′′ parameter varies for each composition. Figure 14.12 shows that R′′′′ increases with additions of 0.5 µm in situ formed spinel to a maximum at 20% loading, and an improvement by a factor of ~4, as compared to MgO. The R′′′′ parameter for the 3 µm and 11 µm preformed spinel composites shows similar values to the 0.5 µm spinel composites up to 20% additions, but is more sensitive to particle size with further additions. The 22 µm preformed spinel composites shows a similar trend but there is a larger effect on R′′′′ above 10% additions, as
R ′′′′ parameter (mm)
2.5
0.5 µm 3 µm 11 µm 22 µm
2.0 1.5 1.0 0.5 0 0
5
10
15 20 Spinel (%)
25
30
14.12 R′′′′ parameter as a function of in situ formed (0.5 µm) and preformed (3, 11, 22 µm) spinel content.
© Woodhead Publishing Limited, 2006
386
Ceramic matrix composites
compared to the other composites. It will be expected that the 20% 22 µm spinel composite will have the highest resistance to thermal shock damage, with an improvement by a factor of ~9, as compared to pure MgO. This increase in R′′′′ parameter appears mainly to be the result of the marked increase in γWOF with spinel additions (Fig. 14.8), and the greater sensitivity of strength to composition and particle size than Young’s modulus.103 Basically the R′′′ and R′′′′ parameters exhibit similar patterns in terms of variations in spinel content.33,74 On the basis of the R plot (Fig. 14.11) spinel composites in general should not be more resistant to crack initiation than pure MgO. As can be expected on the basis of the R′′′ and γWOF values, the R′′′′ values predicts that composites containing the coarser spinel particles should in general be best at resisting further thermal shock damage.33 The fracture mechanism in magnesia–spinel composites therefore appears to rely on the development of microcracks; though the composites are not more resistant to crack initiation than is MgO, they have a stronger resistance to crack extension.74 The thermal shock damage resistance of 0.5 µm in situ formed composites will be expected to be much less than that of preformed composites. For 3 µm and 11 µm spinel composites, addition of 30–40% spinel may provide some improvements. Further deterioration of strength in the coarsest spinel composite as a result of thermal shock should be a minimum at 20% (Fig. 14.12). In general coarser (22 µm) preformed spinel powders appear to be more beneficial than finer powders, but there is no obvious advantage with additions of more than 30%.74,103
14.8.7 Relative strength One of the common procedures for evaluating thermal shock resistance is to compare relative strength values, which are the strengths of bars after quenching relative to initial strengths, as a function of quench temperature. This provides a direct indication of resistance to further damage caused by thermal shock.84 Relative strengths of shocked MgO, and preformed spinel composites, are shown in Fig. 14.13, together with the initial strength values of each material. Figure 14.13 illustrates that values for pure MgO were almost constant up to ~575°C, but further increases in the quench temperature result in a sharp decrease (~80%) in strength. Above this quench temperature, the strength values remain almost at the same level until 1000°C. In contrast, the spinel composites for all spinel volume fractions have a higher relative strength than pure MgO from ~600°C up to the maximum quench temperature used. The absence of any corresponding change in strength for the 20% 22 µm spinel composite suggests that the maximum defect size is unchanged. The 20% 22 µm spinel composite does not lose its strength further after the quench tests, and in this respect is the most stable of the composites tested.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
387
Relative strength
1.2 1.0 0.8 0.6 0.4
MgO – 233 MPa 20% 22 µm – 65.1 MPa 20% 11 µm – 111 MPa
0.2 0 0
200
400 600 800 Quench temperature (°C)
1000
14.13 Relative strengths of pure MgO and preformed spinel composites, as a function of quench temperature.
Spinel composites should be more useful than pure MgO in terms of resistance to thermal shock damage and further loss of strength, because the relative strengths of all composite materials after shocking remain much higher than those of MgO (Fig. 14.13). These results show clearly that the resistances to thermal shock damage for all these materials have the trends expected on the basis of the calculated R′′′ and R′′′′ parameters.74,103
14.8.8 Degree of damage in MgO and spinel composites The amounts and distributions of cracking after quenching in fully dense MgO and in composites containing 20% 22 µm spinel are shown in Figs 14.14(a) and 14.14(b), respectively. Observations of pure MgO quenched into oil from <600°C, on polished surfaces and cross-sections through the centres of the bars, show no cracks; however, at the critical quench temperature (∆Tc), ~600°C, a large number of cracks appear on the polished surfaces, and a small number progress towards the centres of the bars (Fig. 14.14(a)). The amount of cracks increases markedly at quenching from 1000°C, and cracks are now visible on both polished surfaces and cross-sections through to the centres of the bars; most cracks penetrate more than one-quarter of the distance through the bar. Figure 14.14(b) demonstrates that at a quench temperature of 600°C for 20% 22 µm spinel composite, cracks start developing, and the number on polished surfaces increases smoothly up to a quench temperature of 1000°C. Cracks do not progress into the bar centres, and the distance of crack penetration through the bar remains very small, even at the highest quench temperature of 1000°C. No particular quench temperature at which crack density increased markedly is observed using dye penetrant for the spinel composites, in contrast to the MgO materials. It is clearly seen that there is a gradual development of a dense microcrack network in the spinel composites, with increase in the quench temperature.
© Woodhead Publishing Limited, 2006
388
Ceramic matrix composites
600°C
1000°C 2 mm
(a)
600°C
1000°C 2 mm
(b)
14.14 (a) Thermal shock crack patterns revealed by dye penetrant, on the polished surfaces and cross-sections of MgO bars. (b) Thermal shock crack patterns revealed by dye penetrant, on the polished surfaces and cross-sections of 20% 22 µm preformed spinel bars.
14.8.9 Comparison of fraction retained strength and R ′′′′ parameter
Fraction retained strength
In order to compare the thermal shock damage resistance of MgO and MgO– spinel composites, all the values of strength after shock are compared to initial strengths, i.e. retained strength after shock, rather than absolute strength values.74 The calculated R′′′′ parameter for the MgO and spinel composites shows a good correlation with retained strength after thermal shock testing (Fig. 14.15). Figure 14.15 for a quench temperature of 800°C shows that for pure MgO there is very severe mechanical property degradation at low values of R′′′′: 1.0 0.8 0.6
MgO 20% 11 10% 22 20% 22 30% 22
0.4 0.2 0 0
1
2
µm µm µm µm
Spinel Spinel Spinel Spinel 3
R ′′′′ (mm)
14.15 Fraction retained strengths of MgO, and spinel composites, after a quench from 800°C, as a function of R′′′′ parameter.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
389
pure MgO retained only ~20% of its strength. For the best (20–30% 22 µm) magnesia–spinel composites, the R′′′′ parameter is approximately five times that of the pure MgO. The other composites retain about half of their original strength. There is a clear and approximately linear decrease in the extent of thermal shock damage, with increasing R′′′′. The retained strength reaches a maximum and levels out with increasing R′′′′ parameter above the value of ~2 mm. This comparison shows the general relationship that materials with larger values of the damage resistance parameter, R′′′′, suffer less thermal shock damage, and have higher retained strength after quenching. This investigation shows that the R′′′′ parameter, calculated from strength and Young’s modulus values measured at room temperature, can be used to predict the loss in strength of MgO and spinel composites, after thermal shock.74 The observed changes in strength with thermal shock for these materials are in accord with those predicted.
14.9
Conclusions
The coarser spinel particles cause greater loss of strength: spinel content, and more importantly spinel particle size, are factors determining strength. The Young’s modulus values of spinel composites in general decrease with increasing spinel content. Young’s modulus appears to be more sensitive to spinel content than to spinel particle size. The relationships between the KIC and fracture surface energy are similar to each other, and generally decrease with increasing spinel content, apart from the in situ formed composites which show different microstructures and fracture paths. It is therefore clear that the development of thermal shock resistance in the magnesia–spinel composites cannot be linked to any increased fracture toughness. The calculated critical defect sizes, and also actual crack lengths, increase with spinel additions. MgO grain size varies with spinel content but this appears to have little if any influence on strength. The thermal expansion mismatch, causing microcracking, longer interlinked cracks, and spinel particle size and content, are likely to be the major factors determining strength. There is a general increase in γWOF with increasing spinel additions, and the effect of spinel content is greater than that of particle size. The R parameter generally decreases with spinel content for all particle sizes. Magnesia– spinel composites are not more resistant to crack initiation than MgO, because of the pre-existing cracks, but they show greater resistance to further crack propagation than MgO. For this reason, predictions made from the calculated R′′′′ parameter must be taken into account in characterising thermal shock damage and retained strength of materials, rather than the crack initiation resistance parameter, R.
© Woodhead Publishing Limited, 2006
390
Ceramic matrix composites
In the MgO – spinel composites a large amount of cracking is created during cooling, as a result of high tensile hoop stresses generated around the spinel particles, due to the large thermal expansion difference between MgO and spinel. The improved resistance to thermal shock observed in these model magnesia–spinel composites can therefore be attributed to the microcrack networks developed around the spinel particles. Finer spinel particles result in shorter initial crack propagation distances from the spinel particles; the coarser spinel particles are the origins of longer cracks. It might therefore have been expected that resistance to thermal shock damage, resulting from resistance to microcrack propagation and interlinking, would be higher with materials containing the coarser particle spinel. This prediction can be confirmed by the R′′′′ parameter for these model composite materials, which suggests an optimum composition of 20% 22 µm spinel particles, indicating that resistance to further thermal shock damage is about nine times higher than in MgO. The predictions are similar to the results obtained from strength loss values. The relationships between the R′′′ and R′′′′ parameters are similar to each other, and one is as good as the other; they are good indicators for a quantitative evaluation of the retained strengths. After quenching at ~600°C, a large number of cracks appear on the polished surfaces of MgO bars, and a small proportion of cracks progress towards the centres of the bars. In contrast, the 22 µm spinel composites do not show a critical quench temperature: a gradual development of microcrack networks is observed with increasing quench temperature, and cracks do not progress into the cross-section of the bars. For pure MgO, cracks propagate catastrophically after shocking from above the critical quench temperature, with a large decrease in strength. However, spinel composites of all types have higher relative strengths after quenching than pure MgO. The 20% 22 µm spinel composites show a significant improvement in shock damage in terms of increased difficulty of crack propagation. In general, coarser spinel particles are more beneficial than fine, but there is no advantage with additions of >30% for all the particle sizes.
14.10 Future trends The multiple disadvantages of bricks containing high (>90%) levels of magnesia, and the intrinsic hydration susceptibility of dolomite-based products, seem likely to limit their future application to the specific kiln areas where the conditions pertaining do not expose their limitations.11 Spinel is added to magnesia so that the brick is able to cope with the mechanical stresses exerted during kiln operation. Stress factors prevailing in cement kilns, such as thermal shocks, thermal expansion and kiln shell ovality, demand high
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
391
elasticity from bricks – behaviour that a pure magnesia brick is not able to offer.4 Magnesia–spinel bricks are more resistant to reducing atmosphere, attack by sulphur oxides, CO2, and alkali compounds than are magnesia– chromite and dolomite bricks, which lack resistance to moisture, leading to destruction of the brick due to an increase in volume.4 With these properties, magnesia–spinel brick is the most suitable brick for peripherical areas of the sintering zone, which lacks a protective coating. Because of its better thermal shock resistance combined with good abrasion strength, it can also be recommended for the discharge zone to nose ring kiln sections.63 The magnesite–chrome and chrome–magnesite ranges of refractory materials are also intensively used for applications requiring a high hot-strength and resistance to attack by basic slags and molten metals.10 An important example of this type of environment is the rotary cement kiln lining, with centre zone temperatures exceeding 1600°C and the presence of semi-liquid and corrosive calcium aluminosilicates. It is well known that increasing concern over the toxicity of the Cr(VI) produced from Cr2O3 under alkaline conditions has meant that the handling and disposal of waste refractories containing chrome are now subject to strict European Union regulations,117 and alternative refractory materials which do not contain Cr2O3 are needed. Magnesite and dolomite refractories are satisfactory in many respects for the types of application that previously used magnesite–chrome materials, but lack their good resistance to thermal shock. It has been found that this deficiency can be overcome by incorporating into a magnesite matrix 9–30% of particulate magnesium aluminate spinel (MgAl2O 4), to form the magnesia–spinel materials.12,28 Magnesia–spinel refractories were first evaluated more than 30 years ago,10 but it has only been during the last decade that marked efforts have been made to use them as alternative refractories to magnesia–chrome materials. Fortuitously, magnesia–spinel refractories also have the additional important advantage of a longer (1.5–2 times) life than an otherwise equivalent magnesite–chrome refractory, particularly in locations where they are exposed to high temperatures and severe thermal shock.53 Various studies based on the industrial work have been made over the last 15 years, with the objective of developing magnesia–spinel materials with improved resistance to thermal shock and alkali attack. The work done so far has been mostly phenomenological, and little quantitative understanding of the function of the system variables has been developed. It is therefore still difficult to specify optimum compositions with confidence, and materials development is based largely on trial and error. Much of the magnesia–spinel refractory currently produced is used for cement kiln linings, where there are two conflicting requirements: for thermal shock resistance the optimum spinel content should be fairly high; for reduced calcium oxide attack, involving reaction with aluminium oxide and the formation of low-melting calcium aluminates, the spinel content must be as low as possible. It is clearly necessary
© Woodhead Publishing Limited, 2006
392
Ceramic matrix composites
to keep the spinel content to a minimum, while developing the maximum resistance to thermal shock.4 The other technical problems still to be solved in the magnesia–spinel system are the two most difficult problems of achieving both good clinker adhesion and low thermal conductivity.63 These must be achieved without sacrificing resistance to chemical and thermo-physical damage and at a cost comparable with other conventional materials. The process parameters such as chemistry, granulometry and design of products producing changes in porosity, texture, strength and heat-transfer characteristics must be investigated in order to obtain maximum cost-effectiveness.63 The system most capable of long-term development combines magnesia with preformed spinel. It has also been suggested118,119 that mechanical properties and thermal shock of magnesia–spinel refractory materials may be improved by adding small amounts of TiO2 or ZrO2, but full details have not been given. This conclusion needs to be confirmed by further work, and any basic underlying mechanism established. Thermal shock and mechanical tests should be carried out on fully characterised materials to determine the optimum amount of additives and their composition, and the optimum MgO grain size, for the best thermal shock resistance.
14.11 Acknowledgements The contributions of Dr P.D. Warren, who provided suggestions and guidance regarding the investigation of the study, and Professor B. Rand, are gratefully acknowledged. P. Bartha, S. Plint, and M.W. Roberts are thanked for helpful discussions. The contributions of the late Professor R.W. Davidge to the planning of this investigation are acknowledged. The authors also wish to acknowledge P. Bartha,2,4 H.J. Klischat,4 S.C. Cooper28 and P.T.A. Hodson28 for all published sources used by modifying in this chapter (Fig. 14.1 and Tables 14.1, 14.2 and 14.3).
14.12 Sources of further information • Alcoa Corporate Center, 201 Isabella Street, Pittsburgh, PA 15212-5858, USA http://www.alcoa.com • Baker Refractories, Steetley Works, Nottinghamshire S80 3EA, UK http:/ /www.emnet.co.uk/baker-refractories/ • Calbex Mineral Trading Inc., Wenhua 5#, Zhengzhou, Henan 450000, China http://www.calbex.com • Capital Refractories Ltd, Station Road, Clowne, Derbyshire S43 4AB, UK http://www.capital-refractories.com • C.E. Minerals, 901 East 8th Avenue, King of Prussia, PA 19406, USA http://www.ceminerals.com
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
393
• Coating and Crystal Technology, RD #4 Box 113-B, Cadogan Road, Kittanning, PA 16201, USA http://www.coatingandcrystal.com • Custom Technical Ceramics, Inc., 8041 North I-70 Frontage, Unit 6 Arvada, CO 80002, USA http://www.customtechceramics.com • China Industrial Resources Co Ltd, No. 64-302, Wanlian Villa, 2nd Ave. TEDA, Tianjin 300457, China http://www.circogroup.com • Didier Werke AG, RHI AG, Wienerbergstraße 11, A-1100 Vienna, Austria http://www.rhi-ag.com • Magneco/Metrel, Inc., 223 Interstate Road, Addison IL 60101, USA http:/ /www.magneco-metrel.com • Refractarios Peruanos S.A., Casilla 2828, Lima, Peru http:// www.refractoriesinstitute.org • Refratechnik Cement GmbH, Rudolf Winkel Strasse 1, D 37079, Göttingen, Germany http://www.refratechnik.com/ • Resco Products, Inc., 2 Penn Center Blvd. Suite 430, Pittsburgh, PA 15276, USA www.rescoproducts.com • Roskill Information Services, 27a Leopold Road, London SW19 7BB, UK http://www.roskill.com/ • Saint-Gobain C.R.E.E., Research and Development Center, 550 Avenue Alphonse Jauffret, BP 224, 84306 Cavaillon, France http:// www.refractories.saint-gobain.com • Sanac SPA, Viale Certosa 249, 20151 Milano, Italy www.sanac.com • Tiger Industrial Ceramics Co. Ltd, Zhangli Village, Xiangdong Town, Pingxiang City, Jiangxi Province, China http://www.alibaba.com • Vesuvius Dyko GmbH, Group Sachon, Wiesenstrasse 61, 40549 Düsseldorf, Germany http://www.sachon-exportadressbuch.de • Vesuvius Group sa/nv, Mechelsesteenweg 455/1, 1950 Kraainem, Belgium http://www.vesuvius.com/index.htm • Vrag (Veitsch Radex AG), RHI Refractories, Technology Center, RHI AG, Magnesitstr. 2, A-8700 Leoben, Austria http://www.veitsch-radex.com • VRW Refractories, A Division of South India Corporation (Agencies) Ltd, 1513 GIDC, Kerala Industrial Estate, near Bavla, Ahmedabad 382 220, 600 095, Tamil Nadu, India http://www.vrwrefractories.com
14.13 References 1. Wilson, D.R., Evans, R.M., Wadsworth, I. and Cawley, J., ‘Properties and applications of sintered magnesia alumina spinels’, UNITECR ’93 CONGRESS, São Pãulo, Brazil, 1993. 2. Bartha, P., ‘Magnesia spinel bricks – properties, production and use’, Proc Int Symp Refractories, in X. Zhong et al., Pergamon, Hangzhou Refractory Raw Materials and High Performance Refractory Products, 1989 661–74. 3. Laurich-McIntyre, S.E. and Bradt, R.C., ‘Room temperature strengths of individual tabular alumina and sintered spinel grains (aggregates)’, UNITECR ’93 Congress, São Paulo, Brazil, 1993.
© Woodhead Publishing Limited, 2006
394
Ceramic matrix composites
4. Klischat, H.J. and Bartha, P., ‘Further development of magnesia spinel bricks with their own specific properties for lining the transition and sintering zones of rotary cement kilns’, World Cement, 1992 52–8. 5. Tabbert, W. and Klischat, H-J., ‘Magnesia spinel bricks for the cement industry’, Beijing China Symposium, 1992 424–30. 6. Evans, R.M., ‘Magnesia–alumina spinel raw materials production and preparation’, Am. Ceram. Soc. Bull., 1993 72(4), 59–63. 7. ‘Steetley Magnesia Products Limited’, Steetley Co., January 1993. 8. Kimura, M., Yasuda, Y. and Nishio, H., ‘Development of magnesia spinel bricks for rotary cement kilns in Japan’, Interceram Special Issue, 1984 33 Proc. 26th Int Col. Ref., Aachen 1983 344–76. 9. Reyes Sanchez, J.A. and Toledo, O.D., ‘New developments of magnesite–chrome brick and magnesite–spinel for cement rotary kilns higher thermal shock resistance and higher coating adherence’, UNITECR 89, 1989. 10. Eusner, G.R. and Hubble, D.H., ‘Technology of spinel-bonded periclase brick’, J. Am. Ceram. Soc., 1960 43(6) 292–6. 11. Moore, B., Frith, M. and Evans, D., ‘Developments in basic refractories for cement kilns’, World Cement, 1991 5–12. 12. Dal Maschio, R., Fabbri, B. and Fiori, C., ‘Industrial applications of refractories containing magnesium aluminate spinel’, Industrial Ceramics, 1988 8(3) 121–6. 13. Gonsalves, G.E., Duarte, A.K. and Brant, P.O.R.C., ‘Magnesia–spinel brick for cement rotary kilns’, Am. Ceram. Soc. Bull., 1993 72(2) 49–54. 14. Kuennecke, M., Wieland, K. and Faizullah, M., ‘The correlation between burning zone linings and operation of cement rotary kilns – Part 2’, World Cement, 1986 247–53. 15. Gabis, V. and Graba, L., ‘Microstructure of reaction-sintered spinel/corundum refractories prepared from various alumina–magnesia mixtures’, Euro. Ceramics., 1991 2593–8. 16. Lee, W., ‘Microscopy of refractory bricks’, Ceramic Technology Int., 1992 113–22. 17. Goto, K. and Lee, W.E., ‘The ‘Direct Bond’ in magnesia chromite and magnesia spinel refractories’, J. Am. Ceram. Soc., 1995 78(7) 1753–60. 18. Walter, H. and Weibel, G., ‘Lining recommendation for high thermally stressed lime recovery kilns’, World Cement, 1992 34–8. 19. Carbone, T.J., ‘Characterization and refractory properties of magnesium aluminate spinel raw materials’, Interceram. Spec. Issue, 1985 91–4. 20. Bartha, P., ‘Using magnesia-spinel bricks to prevent the formation of rings in rotary cement kilns’, World Cement, 1990 98–100. 21. Jain, M.K. and Richter, T., ‘Magnesia–alumina spinel refractories based on residue from a plasma dross recovery process’, UNITECR ’93 CONGRESS, São Paulo, Brazil, 1993. 22. Modern Refractory Practice, Harbison-Walker Refractories Company, Pittsburgh, PA, 1961. 23. Van Vlack, L.H., Physical Ceramics for Engineers, Addison-Wesley, Reading, MA, 1964. 24. Chesters, J.H., Refractories: Production and Properties, Iron and Steel Institute, London, 1973. 25. Dorre, E. and Hubner, H., Alumina: Processing, Properties and Applications, SpringerVerlag, Berlin, Heidelberg, New York, 1984.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
395
26. Kingery, W.D., Bowen, H.K. and Uhlmann, D.R., Introduction to Ceramics, ed. Burke, E., Chalmers, B. and Krumhansl, J.A., John Wiley, New York, 1976. 27. Chiang, Y.M., Birnie, D.P. and Kingery, W.D., Physical Ceramics – Principles for Ceramic Science and Engineering, John Wiley, New York, 1997. 28. Cooper, S.C. and Hodson, P.T.A., ‘Magnesia–magnesium aluminate spinel as a refractory’, Trans. J. Brit. Ceram. Soc., 1982 81 121–8. 29. Aksel, C., Kasap, F. and Sesver, A., ‘Investigation of parameters affecting grain growth of sintered magnesite refractories’, Ceram. Int., 2005 31(1) 121–7. 30. Ryshkewitch, E. and Richerson, D.W., Oxide Ceramics, General Ceramics Inc., Haskell, N.J., 1985. 31. Sajgalik, P., Panek, Z. and Haviar, M., ‘Sintering map for MgO’, J. Mat. Sci. Letters, 1985 4 1533–5. 32. Aksel, C., ‘Spinel formation, reaction conditions and densification properties of magnesia–spinel composites’, Key Eng. Mater, 2004 264–8 1071–4. 33. Aksel, C., ‘Thermal Shock Behaviour and Mechanical Properties of Magnesia– Spinel Composites’, PhD Thesis, Department of Materials Engineering, University of Leeds, Leeds, UK, 1998. 34. Ramakrishnan, P., ‘The hot-pressing of magnesium oxide’, Trans. J. Brit. Ceram. Soc., 1968 67 135–45. 35 Itatani, K., Nomura, M., Kishioka, A. and Kinoshita, M., ‘Sinterability of various high-purity magnesium oxide powders’, J. Mat. Sci., 1986 21 1429–35. 36. Kostic, E. and Momcilovic, I., ‘Reaction sintered MgAl2O4 bodies from different batch compositions’, Ceramurgia International, 1977 3(2) 57–60. 37. Bailey, J.T. and Russell, Jr. R., ‘Sintered spinel ceramics’, Am. Ceram. Soc. Bull., 1968 47(11) 1025–9. 38. Bailey, J.T. and Russell, Jr. R., ‘Preparation and properties of dense spinel ceramics in the MgAl2O4-Al2O3 system’, Trans. J. Brit. Ceram. Soc., 1969 68 159–64. 39. Tsuchiya, I., Takahashi, H., Kawakami, T. and Kadota, Y., ‘Proof testing of refractories for cement rotary kiln’, Interceram. Spec. Issue, 1983 33 Proc. 26th Int. Col. Ref., Aachen., 148–65. 40. Carter, R.E., ‘Mechanism of solid-state reaction between magnesium oxide and aluminium oxide and between magnesium oxide and ferric oxide’, J. Am. Ceram. Soc., 1961 44(3) 116–20. 41. Beretka, J. and Brown, T., ‘Effect of particle size on the kinetics of the reaction between magnesium and aluminium oxides’, J. Am. Ceram. Soc., 1983 66(5) 383– 8. 42. Serry, M.A., Mandour, M.A. and Weisweiler, W., ‘Phase equilibria, microstructure and properties of some MgO–Al2O3 refractories’, Brit. Ceram. Trans., 1993 92(6) 227–32. 43. Bakker, W.T. and Lindsay, J.G., ‘Reactive magnesia spinel, preparation and properties’, Am. Ceram. Soc. Bull., 1967 46(11) 1094–7. 44. Itatani, K., Sakai, H., Scott Howell, F., Kishioka, A. and Kinoshita, M., ‘Sinterability of spinel (MgAl2O4) powder prepared by vapour-phase oxidation technique’, Brit. Ceram. Trans., 1989 88 13–16. 45. Aksel, C., Warren, P.D. and Riley, F.L., ‘Magnesia–spinel microcomposites’, J. Eur. Ceram. Soc., 2004 24(10–11) 3119–28. 46. Soady, J.S. and Plint, S., ‘A quantitative thermal shock approach to the development of magnesia–spinel refractories for the cement kiln’, UNITECR ’91, Aachen, Germany, 1991.
© Woodhead Publishing Limited, 2006
396
Ceramic matrix composites
47. Singh, M. and Lewis, D., In Situ Composites: Science and Technology, Pittsburgh, PA, Proceedings of a Symposium, 1993. 48. Aksel, C., Davidge, R.W., Warren, P.D. and Riley, F.L., ‘Mechanical properties of model magnesia–spinel composite materials’, in Key Engineering Materials, Euro Ceramics V, Part 3, Versailles, France, 1774–7, 1997. 49. Aksel, C., Davidge, R.W., Knott, P. and Riley, F.L., ‘Mechanical properties of magnesia– magnesium aluminate spinel composites’, in III Ceramic Congress Proceedings Book, Engineering Ceramics, Istanbul, Turkey, 2, 172–9, 1996. 50. Roy, D.M., Roy, R. and Osborn, E.F., ‘The system MgO–Al2O3–H2O and the influence of carbonate and nitrate ions on the phase equilibria’, Am. J. Science, 1953 251 337– 61. 51. Harburg, H.K.F., ‘Experience with magnesium–aluminium–spinel bricks in a 3000 t/d rotary kiln’, Zement-Kalk-Gibs International, 1993 (3/4) 446–54. 52. Hasselman, D.P.H., ‘Thermal stress resistance parameters for brittle refractory ceramics: a compendium’, Am. Ceram. Soc. Bull., 1970 49(12) 1033–7. 53. Tokunaga, K., Kozuka, H., Honda, T. and Tanemura, F., ‘Further improvement in high temperature strength, coating adherence, and corrosion resistance of magnesiaspinel bricks for rotary cement kiln’, UNITECR ‘91, 1991. 54. Sato, A., Tsuchiya, I., Takahashi, H., Ishii, T., Takebayashi, K. and Kawakami, T., ‘Effect of thermal shock on the structural changes of the basic refractories for cement rotary kiln’, Taikabutsu Overseas, 1986 8(1) 37–9. 55. Hara, K., Kusunose, H., Kenmochi, I. and Tokunaga, K., ‘Study for improvement of spinel bricks’, Taikabutsu Overseas, 1986 8(1) 31–2. 56. Reijnen, P., ‘The formation of ferrites from the metal oxides’, in Science of Ceramics, ed. Stewart, G.H., Academic Press, London and New York, 3, 245–61, 1967. 57. Hulbert, S.F., Wilson, H.H. and Venkatu, D.A., ‘Kinetics of the reaction between MgO and Fe2O3 in powder compacts’, Trans. J. Brit. Ceram. Soc., 1970 69 9–13. 58. Benbow, J., ‘Cement kiln refractories – down to basics’, Industrial Minerals, 1990 37–45. 59. Bilham, M.A., ‘Magnesia–spinel: the chrome free solution’, International Cement Review, 1991 40–1. 60. Uchikawa, H., Hagiwara, H., Shirasaka, M. and Watanabe, T., ‘Application of periclase– spinel bricks to cement rotary kiln in Japan’, Interceram. Spec. Issue, 1984 33 386– 406. 61. Macey, C.L., ‘Evaluation of magnesite–spinel refractories for mineral processing kilns’, Industrial Heating, 1992 28–9. 62. Olbrich, M., ‘Fully automated thermal shock test method for testing fired refractory brick’, Radex-Rundschau, 1990 (2/3) 268–74. 63. Hobrecht, E.J., Daldrup, H.G. and Bartha, P., ‘Development in basic bricks’, Cements, Betons, Platres, Chaux, 1988 (773) 219–25. 64. Refratechnik GmbH Report (1994), FO-034, 1–21. 65. Barthel, H. and Kaltner, E., ‘The basic lining of cement rotary kilns to conform to changed requirements’, Proc. 2nd Int. Conf. on Refractories, Tokyo, 2, 1987. 66. Chandler, H.W., ‘Thermal stress in ceramics’, Trans. J. Brit. Ceram. Soc., 1981 80(6) 191–5. 67. Geisler, T.A., ‘Finite element analysis of thermal stresses in cement kiln brick’, UNITECR ‘89, 1989. 68. Schacht, C.A., ‘Influence of lining restraint and non-linear material properties in predicting thermal shock fracture of refractory linings’, UNITECR ‘89, 1989.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
397
69. Kingery, W.D., ‘Factors affecting thermal stress resistance of ceramic materials’, J. Am. Ceram. Soc., 1955 38(1) 3–15. 70. Sveda, M. and Gomolova, Z., ‘Increasing the thermal resistance of brick’, Am. Ceram. Soc. Bull., 1995 74(11) 93–5. 71. Aksel, C., Rand, B., Riley, F.L. and Warren, P.D., ‘Mechanical properties of magnesia– spinel composites’, J. Eur. Ceram. Soc., 2002 22(5) 745–54. 72. Rigby, G.R., ‘The effect of expansion mismatch on the mechanical properties of ceramic materials’, Trans. Indian. Ceram. Soc., 1972 31(1) 18–30. 73. Aksel, C. and Riley, F.L., ‘Young’s modulus measurements of magnesia–spinel composites using load–deflection curves, sonic modulus, strain gauges and Rayleigh waves’, J. Eur. Ceram. Soc., 2003 23(16) 3089–96. 74. Aksel, C., Rand, B., Riley, F.L. and Warren, P.D., ‘Thermal shock behaviour of magnesia–spinel composites’, J. Eur. Ceram. Soc., 2004 24(9) 2839–45. 75. Aksel, C., Davidge, R.W., Warren, P.D. and Riley, F.L., ‘Investigation of thermal shock resistance in model magnesia–spinel refractory materials’, in IV Ceramic Congress, Proceedings Book, Part 1, Eskiflehir, Turkey, 1998. 76. White, K. and Kelkar, G.P., ‘Fracture mechanism of a coarse-grained, transparent MgAl2O4 at elevated temperatures’, J. Am. Ceram. Soc., 1992 75(2) 3440–4. 77. Stewart, R.L. and Bradt, R.C., ‘Fracture of polycrystalline MgAl2O4’, J. Am. Ceram. Soc., 1980 63(11) 619–23. 78. Stewart, R.L. and Bradt, R.C., ‘Fracture of single crystal MgAl2O4’, J. Mater. Sci., 1980 15 67–72. 79. Baudin, C., Martinez, R. and Pena, P., ‘High-temperature mechanical behaviour of stoichiometric magnesium spinel’, J. Am. Ceram. Soc., 1995 78(7) 1857–62. 80. Ghosh, A., Ehite, K.W., Jenkis, M.G., Kobayasi, A.S. and Bradt, R.C., ‘Fracture resistance of a transparent magnesium aluminate spinel’, J. Am. Ceram. Soc., 1990 74(7) 1624–30. 81. Cortes, G., ‘Magnesia alumina spinels for the refractory industry’, Ceramic Technology International, 1994 109–12. 82. Korgul, P., Wilson, D.R. and Lee, W.E., ‘Microstructural analysis of corroded alumina– spinel castable refractories’, J. Eur. Ceram. Soc., 1997 17 77–84. 83. Davidge, R.W. and Tappin, G., ‘Thermal shock and fracture in ceramics’, J. Brit. Ceram. Soc., 1967 66 405–22. 84. Hasselman, D.P.H., ‘Elastic energy at fracture and surface energy as design criteria for thermal shock’, J. Am. Ceram. Soc., 1963 46(11) 535–40. 85. Morrell, R., Handbook of Properties of Technical and Engineering Ceramics, Part 1, Her Majesty’s Stationery Office, London, 1985. 86. Hasselman, D.P.H., ‘Unified theory of thermal shock fracture initiation and crack propagation in brittle ceramics’, J. Am. Ceram. Soc., 1969 52(11) 600–4. 87. Hulbert, S.F., ‘Models for solid state reactions in powdered compacts’, J. Brit. Ceram Soc., 1968 1–32. 88. Chaklader, A.C.D. and Bradley, F., ‘Thermal shock resistance parameters and their application to refractories’, UNITECR ’89, 1989. 89. Hasselman, D.P.H. and Singh, J.P., ‘Analysis of thermal stress resistance of microcracked brittle ceramics’, Am. Ceram. Soc. Bull., 1979 58(9) 856–60. 90. Lutz, E.H., Swain, M.V. and Claussen, N., ‘Thermal shock behaviour of duplex ceramics’, J. Am. Ceram. Soc., 1991 74(1) 19–24. 91. Nakayama, J. and Ishizuka, M., ‘Experimental evidence for thermal shock damage resistance’, Am. Ceram. Soc. Bull., 1966 45(7) 666–9.
© Woodhead Publishing Limited, 2006
398
Ceramic matrix composites
92. Stewart, R.L., Iwasa, M. and Bradt, R.C., ‘Room-temperature KIC values for singlecrystal and polycrystalline MgAl2O4’, J. Am. Ceram. Soc., 1981 64(2) C-22–3. 93. Rice, R.W., Wu, C.C. and Mckinney, K.R., ‘Fracture and fracture toughness of stoichiometric MgAl2O4 crystals at room temperature’, Journal of Materials Science, 1996 31 1353–60. 94. Llorca, J. and Ogawa, T., ‘Crack wake effects on MgO fracture resistance’, in Bradt, R.C., Hasselman, D.P.H., Munz, D., Sakai, M., Ya Shevchenko, V., Fracture Mechanics of Ceramics, 1992 9 305–17. 95. Davidge, R.W., Mechanical Behaviour of Ceramics, Cambridge University Press, 1979. 96. Rice, R.W., ‘Strength and fracture of hot-pressed MgO’, Proc. Brit. Ceram. Soc., 1972(20) 329–63. 97. Evans, A.G., ‘Energies for crack propagation in polycrystalline MgO’, Phil. Mag, 1970 22 841–52. 98. Davidge, R.W., ‘The texture of special ceramics with particular reference to mechanical properties’, Proc. Brit. Ceram. Soc., 1972(20) 364–78. 99. Unchno, J.J., Bradt, R.C. and Hasselman, D.P.H., ‘Fracture surface energies of magnesite refractories’, Am. Ceram. Soc. Bull., 1976 55(7) 665–8. 100. Swanson, G.D., ‘Fracture energies of ceramics’, J. Am. Ceram. Soc., 1972 55(1) 48–9. 101. Clarke, F.J.P., Tattersall, H.G. and Tappin, G., ‘Toughness of ceramics and their work of fracture’, Proc. Brit. Ceram. Soc., 1996(6) 163–72. 102. Olbrich, M. and Dobrowsky, F., ‘Periclase spinel bricks in the cement industry’, Zement-Kalk-Gibs, 1989 12 307–9. 103. Aksel, C. and Warren, P.D., ‘Thermal shock parameters (R, R′′′ and R′′′′) of magnesia– spinel composites’, J. Eur. Ceram. Soc., 2003 23(2) 301–8. 104. Aksel, C., Warren, P.D. and Riley, F.L., ‘Fracture behaviour of magnesia and magnesia–spinel composites before and after thermal shock’, J. Eur. Ceram. Soc., 2004 24(8) 2407–16. 105. Aksel, C. and Warren, P.D., ‘Work of fracture and fracture surface energy of magnesia–spinel composites’, Comp. Sci. Technol., 2003 63(10) 1433–40. 106. Aksel, C. and Riley, F.L., ‘Effect of particle size distribution of spinel on the mechanical properties and thermal shock performance of MgO–spinel composites’, J. Eur. Ceram. Soc., 2003 23(16) 3079–87. 107. Griffith, A.A., ‘The phenomena of rupture and flow in solids’, Phil. Trans. Roy. Soc. London, 1920 A221 163–98. 108. Griffith, A.A., ‘The theory of rupture’, Proc. First Int. Congr. for Applied Mechanics, 1924 55–63. 109. Hasselman, D.P.H., ‘Strength behaviour of polycrystalline alumina subjected to thermal shock’, J. Am. Ceram. Soc., 1970 53(9) 490–5. 110. Nakayama, J., Abe, H. and Bradt, R.C., ‘Crack stability in the work-of-fracture test: refractories applications’, J. Am. Ceram. Soc., 1991 64(11) 671–5. 111. Bradt, R.C., ‘Fracture testing of refractories, past present and future’, Proc. 2nd Int. Conf. on Refractories, Tokyo 1, 1987. 112. Semler, C.E., Hawisher, T.H. and Bradt, R.C., ‘Thermal shock of alumina refractories: damage-resistance parameters and the ribbon test’, Am. Ceram. Soc. Bull., 1981 60(7) 724–9. 113. Semler, C.E. and Bradt, R.C., ‘Thermal shock damage of magnesite chrome refractories in the ribbon test’, Am. Ceram. Soc. Bull., 1984 63(4) 605–9.
© Woodhead Publishing Limited, 2006
Magnesia–spinel (MgAl2O4) refractory ceramic composites
399
114. Coppola, J.A. and Bradt, R.C., ‘Thermal shock damage in SiC’, J. Am. Ceram. Soc., 1973 56(4) 214–17. 115. Ainsworth, J.H. and Herron, R.H., ‘Thermal shock damage resistance of refractories’, J. Am. Ceram. Soc. Bull., 1974 53(7) 533–8. 116. Dodd, A.E. and Murfin, D., Dictionary of Ceramics, Institute of Materials, London, 1994. 117. Bray, D.J., ‘Toxicity of chromium compounds formed in refractories’, Bull. Amer. Ceram. Soc., 1985 64(7) 1012–16. 118. Mosser, J., Buchebner, G. and Dosinger, K., ‘New high-quality MgO–Cr2O3 bricks and Cr-free alternatives for the lining of RH/DH-vessels’, Veitsch-Radex Rundschau, 1997 1 11–23. 119. Schulle, W., Khanh, P.G. and Anh, V.T., ‘Periclase–spinel products with improved properties by effective addition of TiO2’, Veitsch-Radex Rundschau, 1995 1 563–4.
© Woodhead Publishing Limited, 2006
15 Thermal shock of ceramic matrix composites C K A S T R I T S E A S, P S M I T H and J Y E O M A N S, University of Surrey, UK
15.1
Introduction
The aim of this chapter is to present an overview of the performance of ceramic matrix composites (CMCs) under conditions of thermal shock, i.e. when they are subjected to sudden changes in temperature during either heating or cooling. Such conditions are possible in the high-temperature applications for which these materials are targeted (e.g. Ohnabe et al., 1999). For example, while thermal shock is not a concern during steady-state operation of a gas turbine, it becomes of great importance during emergency shut-downs, when cool air drawn from the still spinning compressor is driven through the hot sections and can result in a temperature decrease of more than 800°C within one second at the turbine inlet (Baste, 1993). The fact that such a situation may arise about 100 times during the lifetime of a modern gas turbine engine shows how important it is to assess, and possibly model, the effect of thermal shock on the mechanical and thermal properties of CMCs. Another example comes from the nuclear industry, where SiC reinforced with SiC fibres has been proposed as structural material for the first wall and blanket in several conceptual design studies of future fusion power reactors (Jones et al., 2002). In this case, apart from the moderate shocks inflicted during start-up and shut-down of the system, the plasma-facing material can suffer rapid heating due to plasma discharges. The description of the thermal shock behaviour of CMCs is given with reference to the thermal shock resistance of monolithic ceramic materials. Monolithic ceramics have greater thermal shock sensitivity than metals and can even suffer catastrophic failure due to thermal shock because of an unfavourable ratio of stiffness and thermal expansion to strength and thermal diffusivity, and their limited plastic deformation. The structure of the chapter is as follows: the stress field developed in a thermally shocked component is described in Section 15.2 and maximum stresses are identified and quantified. Section 15.3 contains an overview of 400 © Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
401
the experimental methods used to simulate thermal shock conditions in the laboratory and the means utilised to assess its impact on ceramics and CMCs. The behaviour of monolithic ceramics is described in Section 15.4 with reference to the main mechanical and thermal properties that affect it. In addition, methods to model this behaviour are presented. Section 15.5 concentrates on the thermal shock behaviour of particle- and whisker-reinforced CMCs, while Section 15.6 contains an extensive review of damage modes sustained in fibre-reinforced CMCs due to thermal shock, their effect on properties, the role of the interface and attempts to analyse and model the situation. It should be noted that this review concentrates on thermal shock (i.e. a single thermal cycle) and no attempt is made to incorporate and describe the effects of cyclic thermal loading (cyclic thermal shock, thermal shock fatigue, etc.) on the behaviour of CMCs. For information regarding cyclic thermal loading of ceramics and CMCs the reader is advised to consult the extensive review of Case (2002). Additionally, recent studies have shown that laminated ceramic–metal systems (Sherman, 2001) and layered ceramic–structures (Vandeperre et al., 2001) exhibit better resistance to thermal shock compared with monolithic materials. However, such systems are also beyond the scope of this contribution.
15.2
Thermal shock of brittle materials: the induced stress field
When a body is subjected to a rapid temperature change such that non-linear temperature gradients appear, stresses arise due to differential expansion of each volume element at a different temperature. The temperature at each point changes with time at a rate dependent on the coefficient of surface heat transfer (HTC) between the medium of different temperature and the body, the shape of the body, and its thermal conductivity. High HTCs, large dimensions, and low thermal conductivities result in large temperature gradients and, thus, large stresses. This leads to the establishment of a dimensionless parameter, the ‘Biot modulus’, for the description of the heat transfer condition (Kreith, 1986):
Bi = l h k
(15.1)
where l is a characteristic material dimension (e.g. the half-thickness of a plate), h is the HTC between the body and the medium, and k is the thermal conductivity of the body. The larger the value of Bi, the larger is the rate of heat transfer between a medium of different temperature and the body. The sudden temperature change (∆T) that generates non-linear temperature gradients in a body and, as a consequence, thermal stresses is termed ‘thermal
© Woodhead Publishing Limited, 2006
402
Ceramic matrix composites
shock’. If ∆T is positive (i.e. the temperature change is downward) the material is subjected to a ‘cold’ shock, whereas if ∆T is negative the material is subjected to a ‘hot’ shock. The term refers to a single thermal cycle (N = 1) in contrast to terms such as ‘thermal cycling’, ‘cyclic thermal shock’, and ‘thermal fatigue’ that apply to multiple thermal cycles (N > 1). For the calculation of the thermal shock-induced stresses, we consider the plate shown in Fig. 15.1 with Young’s modulus E, Poisson’s ratio ν, and coefficient of thermal expansion (CTE) α, initially held at temperature Ti. If the top and bottom surfaces of the plate come into sudden contact with a medium of lower temperature T∞ they will cool and try to contract. However, the inner part of the plate initially remains at a higher temperature, which hinders the contraction of the outer surfaces, giving rise to tensile surface stresses balanced by a distribution of compressive stresses at the interior. By contrast, if the surfaces come into contact with a medium of higher temperature T∞, they will try to expand. As the interior will be at a lower temperature, it will constrain the expansion of the surfaces, thus giving rise to compressive surface stresses balanced by a distribution of tensile stresses at the interior. If perfect heat transfer between the surfaces and the medium is assumed (i.e. if Bi → ∞) the surface immediately adopts the new temperature while the interior of the plate remains at Ti. Following Munz and Fett (1999), this case corresponds to having a plate that can expand freely in the z-direction with suppressed expansion in the x- and y-directions. In the absence of displacement restrictions, the plate would expand along the x- and y-directions by thermal strains of:
εx = α (T∞ – Ti)
(15.2)
εy = α (T∞ – Ti)
(15.3)
Since thermal expansion in both directions is completely suppressed, elastic strains are created that compensate the thermal strains, i.e.
εel,x + εth,x = 0
(15.4)
εel,y + εth,y = 0
(15.5)
From equations (15.2)–(15.5) we have: TS
σx σyTS
y
z
x 2H
15.1 Schematic of a plate of thickness 2H subjected to thermal shock.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
403
εel,x = –εth,x = –α (T∞ – Ti) = α (Ti – T∞) = α ∆T
(15.6)
εel,y = –εth,y = –α (T∞ – Ti) = α (Ti – T∞) = α ∆T
(15.7)
The elastic strains cause ‘thermal stresses’ along the x- and y-axes and can be written as:
ε el,x =
σ xTS νσ yTS – E E
(15.8)
ε el,y =
σ yTS νσ xTS – E E
(15.9)
By substituting (15.6) and (15.7) in (15.8) and (15.9) respectively and solving first for σ xTS and then for σ yTS we can obtain the thermal shock-induced stresses along the x- and y-axes as:
σ xTS = σ yTS =
Eα ∆T 1–ν
(15.10)
Equation (15.10) shows that thermal shock induces a biaxial stress field, whose maximum value depends on the elastic properties of the material and the imposed temperature differential. However, if the rate of heat transfer is not infinite the thermal shockinduced stresses will gradually build up and after some time reach a peak value that will be a fraction of the value given by equation (15.10). The solution requires transient stress analysis such as those of Cheng (1951) and Manson (1966) with the assumption of the plate of Fig. 15.1 being infinite. Following Lu and Fleck (1998), the plate is initially held at temperature T0 and at time t = 0 its top and bottom faces (at z = ± H) are suddenly exposed to a convective medium of temperature T∞. The surface heat flow is assumed to satisfy kz
∂T = m h ( T∞ – T ), at z = ± H ∂z
(15.11)
where kz is the thermal conductivity in the z-direction and T(z, t) is the temperature of the material. The plate is assumed to be a uniform, linear thermoelastic solid and is analysed under the constraint that it is free to expand with vanishing axial force
∫
H
σ xTS dz =
–H
∫
H
σ yTS dz = 0
(15.12)
–H
and vanishing normal stress in the through-thickness direction, i.e. σz = 0. The transient thermal shock-induced stresses, σx(z, t) = σy(z, t), associated with the temperature distribution T(z, t) are then given by:
© Woodhead Publishing Limited, 2006
404
Ceramic matrix composites
σ xTS ( z , t ) = σ yTS ( z , t ) = –
Eα ( T – Ti ) Eα + 1–ν (1 – ν )2 H
∫
H
( T – Ti )dz
(15.13)
–H
To obtain the temperature distribution T(z, t) the heat flow in the throughthickness direction needs to be considered. This is governed by: ∂ 2 T = 1 ∂T , | z| ≤ H k z ∂t ∂z 2
(15.14)
This equation is solved with heat transfer boundary condition (15.11) by a standard separation-of-variables technique, to give: ∞ T( z ,t ) – Ti sin β n cos( β n z / H ) k t = –1 + 2 Σ exp – β n2 z 2 × n=1 Ti – T∞ β n + sin β n cos β n H
(15.15) where βn are the roots of βn tan βn = Bi. The thermal shock-induced stresses are obtained from (15.13) and (15.15), and are written in non-dimensional form as:
σ xTS = σ yTS = =
σ TS σ xTS ( z , t ) y ( z, t ) = –1 Eα (1 – ν ) ( Ti – T∞ ) Eα (1 – ν ) –1 ( Ti – T∞ ) T( z , t ) – Ti – 1 Ti – T∞ 2H
∫
H
–H
T( z , t ) – Ti dz Ti – T∞
∞ sin β n k t = 2 Σ exp – β n2 z2 n=1 H β n + sin β n cos β n
sin β n × cos β n z – H β n
(15.16)
The evolution of dimensionless stresses is then plotted against dimensionless time ( t = k z t / H 2 ) at selected locations (z/H) through the thickness of the plate and for various values of the Biot modulus. An example of such a plot is given in Fig. 15.2. The plots show that under cold shock and for all values of Bi, the maximum tensile stress is achieved at the surfaces while the maximum compressive stress is achieved at the centre of the plate. The opposite is true for hot shock TS , achieved at the surface during conditions. The maximum tensile stress, σ max cold shock and at the centre during hot shock, is then plotted against 1/Bi, as shown in Fig. 15.3. TS increases with increasing Bi for both cold and hot It can be seen that σ max
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
405
0.8
σ (z , t )
0.6
E α (Ti – T ∞ )
Bi = 10
z =1 H
0.9
0.4
0.8 0.2
0.7 0.6 0.5 0.4 0.2
0
– 0.2
0 – 0.4
0
0.1
0.2
0.3
0.4
0.5
kt H2
15.2 The evolution of dimensionless thermal shock-induced stress with dimensionless time at various locations through the plate thickness for Bi = 10 (reprinted from Acta Materialia, 46, Lu and Fleck, ‘The thermal shock resistance of solids’, 4755–4768, copyright 1998, with permission from Elsevier).
1 Numerical Curve fit
σ max
E α (T1 – T ∞ )
0.8
0.6
0.4 Surface 0.2 Centre 0
0
2
4
6
8
10
1/Bi
15.3 The maximum tensile thermal shock-induced stress achieved at the surface in cold shock and in the centre of the plate in hot shock as a function of 1/Bi. Also shown are curve fits expressed described by equations [15.17] and [15.18] (reprinted from Acta Materialia, 46, Lu and Fleck, ‘The thermal shock resistance of solids’, 4755–4768, copyright 1998, with permission from Elsevier).
© Woodhead Publishing Limited, 2006
406
Ceramic matrix composites
shock, whereas the maximum tensile stress developed at the surface during cold shock is always much higher than the peak tensile stress developed at the centre of the plate during hot shock. This observation, combined with the fact that brittle materials usually contain a distribution of surface flaws, means that cold shock is a much more dangerous condition for a brittle material. The maximum surface stress in the infinite plate under cold shock is adequately described by the formula: TS σ max ( ± H , t *) = 1.5 + 3.25 – 0.5e –16/ Bi Bi
–1
(15.17)
where t* is the time taken for this value to be reached. The maximum tensile stress at the centre of the plate for hot shock is given by: TS σ max (0, t *) =
0.3085 1 + (2/ Bi )
(15.18)
Equation (15.17) can be written through (15.10) as: TS σ max =
–1 Eα ∆T 1.5 + 3.25 – 0.5e –16/ Bi Bi 1–ν
(15.19)
and subsequently as: TS σ max =A
Eα ∆T 1–ν
(15.20)
In equation (15.20) the parameter A is termed the ‘stress reduction factor’ and is given by: A=
1 = 1 f ( Bi ) 3.25 –16/ Bi 1.5 + – 0.5e Bi
(15.21)
Equation (15.20) is the classic formula used to characterise thermal shockinduced stresses at the surfaces of brittle components during cold shock. The function f (Bi) can be written more generally as: f ( Bi ) = a + b – ce d / Bi Bi
(15.22)
where the values of a, b, c and d depend on the shape of the component and are determined by using analyses similar to the one presented above for an infinite plate. For example, it was shown that for an infinite plate a = 1.5, b = 3.25, c = 0.5 and d = –16, whereas for an infinite rod a = 1.5, b = 4.67, c = 0.5 and d = –51 (Manson, 1966). The value of f (Bi) (and consequently A) depends on the Biot modulus, i.e. on the HTC, thermal conductivity and material dimensions. For severe shocks
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
407
Bi becomes very large, so f (Bi) becomes f (Bi) ≈ a, which leads to the maximum value of the stress reduction factor being given by A ≈ 1/a (Wang and Singh, 1994). Experimental evidence suggests that there is a critical value of the characteristic specimen dimension, lc, above which Bi (and consequently A and the value of the shock-induced stress) becomes independent of material dimensions (Wang and Singh, 1994). Becher and Warwick (1993) showed graphically that this value may be approximated by:
lc ≈ b k ah
(15.23)
For example, for an infinite rod, the critical dimension is given by lc ≈ 3.1k/h.
15.3
Experimental methods
15.3.1 Introduction This section aims to present briefly the experimental methods used to evaluate the performance of ceramics and CMCs under conditions of thermal shock. Reference is made to techniques used to impose the actual thermal shock condition as well as the destructive and non-destructive methods employed to assess damage morphologies and changes in residual properties.
15.3.2 Thermal shock simulation methods The methods used to simulate thermal shock can be classified into two categories, depending on the sign of the temperature differential to which the material is exposed: 1. Quench tests, when the material is subjected to a sudden temperature decrease (∆T < 0) 2. Fast heating methods, if a sudden increase in temperature (∆T > 0) is involved. In a quench test, the specimen is heated to a pre-determined temperature in a furnace and is held at that temperature for a certain period of time (~10– 20 min) to allow for the furnace and specimen temperatures to reach equilibrium. A sudden temperature decrease is then brought about by bringing the heated specimen into contact with a cooling agent. The difference in temperatures between the specimen and the cooling agent is defined as the ‘quenching temperature difference’. The process is repeated for different furnace temperatures until the temperature at which fracture and/or property degradation is just initiated can be determined. The difference in temperature
© Woodhead Publishing Limited, 2006
408
Ceramic matrix composites
between this and the medium is the ‘critical quenching temperature difference’, ∆Tc. Methods to cause rapid temperature decrease of the heated specimen include immersing the specimen into a quenching medium, subjecting it to a flow of high velocity cold air (Faber et al., 1981) or water (Absi and Glandus, 2004), and contacting the specimen with a cold metal rod (Rogers and Emery, 1992). The most popular method has been the first, while the most commonly used quenching medium is room-temperature water (Wang and Singh, 1994). Other quenching media include boiling water (Becher, 1981; Tiegs and Becher, 1987), room-temperature air (Boccaccini et al., 1998, 1999), glycerine oil (Ishitsuka et al., 1989; Uribe and Baudin, 2003), silicone oil (Evans et al., 1975; Konsztowicz, 1990, 1993; Tancret and Osterstock, 1997), ethylene glycol (Thompson and Rawlings, 1991), methyl alcohol (Ishitsuka et al., 1989), liquid nitrogen (Lee et al., 1993; Tancret and Osterstock, 1997), liquid metals (Hencke et al., 1984), pre-heated salt (Soboyejo et al., 2001), or fluidized beds (Morrel, 1993; Schneibel et al., 1998). The advantages of the quench test include its simplicity and the welldefined temperature difference between sample and cooling agent. However, a major drawback is that the value of the HTC is often difficult to assess, especially for quenching into water where h is affected by different boiling phenomena (Kreith, 1986). In addition, the HTC is not a constant for a certain quenching medium, as it changes with specimen temperature and is affected by the surface finish of the specimen (Becher et al., 1980; Becher, 1981; Becher and Warwick, 1993). Quenching into media other than water results in significantly lower values of h (Lee et al., 1993). For these reasons, quenching experiments are suitable for comparing materials but not for measuring absolute values (Pompe et al., 1993; Morrel, 1993). In a fast heating test, usually the central area of a specimen is quickly heated up by a heating source. Heating sources used include plasma jets, lasers, energetic electron beams, hot gas jets, arc discharges and hydrogen– oxygen flames (Pompe et al., 1993; Schneider and Petzow, 1993). The fast heating test causes a different stress distribution in the specimen to the quench test, which results in the activation of a different population of flaws. The thermal shock resistance of a material can be evaluated by measuring the critical temperature difference for crack initiation (i.e. as in the quench test), by measuring the critical power of the heating source for failure, or by measuring the temperature gradient as a function of time and calculating the corresponding energy input and stress intensity factor (Wang and Singh, 1994). Generally, the thermal shock induced by a heating source is considered to be much less severe than that imparted during a quench test, especially when room-temperature water is the quenching medium (Case, 2002).
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
409
15.3.3 Assessment of thermal shock damage The impact of thermal shock on the properties of a ceramic or a CMC is assessed by means of both destructive and non-destructive testing methods. Flexural or tensile (mainly for CMCs) tests of suitably-sized thermally shocked specimens are usually employed to measure retained mechanical properties as a function of the temperature difference. The temperature differential for which a significant drop in property values is observed is the ∆Tc. For monolithic ceramics and particle- or whisker-reinforced CMCs the property under investigation is usually strength, whereas in fibre-reinforced CMCs a drop in Young’s modulus is usually a better indication of the onset of damage. Alternative approaches, termed ‘indentation thermal shock tests’, with pre-cracks of known sizes have been used by several authors to assess thermal shock damage in monolithic ceramics. Knoop (Hasselmann et al., 1978; Faber et al., 1981) or Vickers (Gong et al., 1992; Osterstock, 1993; Andersson and Rowcliffe, 1996; Tancret and Osterstock, 1997; Collin and Rowcliffe, 1999, 2000; Lee et al., 2002) indentations were made on rectangular bars, which were then heated to pre-determined temperatures and quenched into water. Crack extensions from the indentations were measured as a function of quench temperature differential, and the critical temperature for spontaneous crack growth (failure) was determined for the material. Fracture mechanics analyses, which took into account measured resistance-curve (R-curve) functions, were then used to account for the data trends. In addition, it has been shown (Boccaccini et al., 2001; Chlup et al., 2001) that the chevron-notched specimen flexural technique (the CN-technique) can be a reliable method of assessing fracture properties (fracture toughness, work of fracture) in thermally shocked brittle matrix composites reinforced by brittle fibres. As an alternative to destructive methods, various non-destructive techniques have been employed to assess damage caused by thermal shock. These include the determination of the post-shock Young’s modulus using ultrasonics or through the identification of the mechanical resonant frequencies of the material (Carter et al., 1988; Lee and Case, 1989, 1990; Wang and Singh, 1994; Wang et al., 1994, 1996; Boccaccini et al., 1997, 1998), the monitoring of the change in the spectra of ultrasonic pulses passed through a specimen before and after thermal shock (Thompson and Rawlings, 1991), the acoustic emission technique (Evans et al., 1975; Konsztowicz, 1990, 1993; Rogers and Emery, 1992), the measurement of the change in specific damping capacity (Q–1) (Lee and Case, 1989, 1990; Boccaccini et al., 1997, 1998, 1999), and the measurement of thermal diffusivity before and after the shock using the ‘flash diffusivity’ method (Ellingson, 1995; Graham et al., 2003). In addition, optical microscopy (e.g. reflected light microscopy) and scanning electron microscopy have been used extensively for direct observation of
© Woodhead Publishing Limited, 2006
410
Ceramic matrix composites
crack patterns on suitably polished surfaces of thermally shocked samples. More recently, a thermal shock testing technique has been developed (Wereszczak et al., 1999) that uses a high-resolution, high-temperature infrared camera to capture the surface temperature distribution of a test specimen at fracture.
15.4
Thermal shock of monolithic ceramics
The behaviour of ceramic materials under conditions of thermal shock is characterised by a number of parameters (figures-of-merit) that concentrate on either the initiation of cracking due to thermal shock or the resistance of a material to crack propagation during thermal shock (Kingery 1955; Hasselman, 1970, 1978, 1985). The first parameter is derived by considering equation (15.10) and assuming that fracture occurs when the thermally induced stress, σTS, becomes equal to the strength of the material, σt. By solving (15.10) for ∆T (= ∆Tc) we obtain the ‘maximum allowable quenching temperature difference’ for the onset of cracking under severe thermal shock conditions, i.e. conditions that approximate perfect heat transfer between the material surface and the quenching medium (e.g. in water quench), as: R=
σ t (1 – ν ) Eα
(15.24)
Equation (15.24) shows that, for high resistance to crack initiation, high strengths combined with low stiffness and CTE are required. Under mild thermal shock conditions (e.g. in a boiling water quench) the thermal conductivity also becomes important, and (15.24) is modified to give: R′ =
σ t (1 – ν ) k Eα
(15.25)
The resistance to crack propagation is characterised by the following parameter: R″″ =
σ t2
EG (1 – ν )
(15.26)
where G is the surface fracture energy. Equation (15.26) shows that for better resistance to crack propagation, high values of stiffness and toughness are required, combined with low strengths. Different parameters impose different requirements on ceramic materials depending on whether fracture resistance or crack propagation resistance is of prime importance. The values of the above parameters for a range of ceramic materials are presented in Table 15.1, where the property dependence of thermal shock behaviour can be observed. A variety of other parameters
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
411
Table 15.1 Values of the thermal shock resistance parameters R, R′, R′′′′ for a range of ceramic materials where HPSN is hot pressed silicon nitride and RBSN is reaction bonded silicon nitride (reprinted from Table 11.1 on p 213 of ‘Ceramics: Mechanical Properties, Failure Behaviour, Materials Selection’ by Munz and Fett, 1999, published with permission from Springer-Verlag GmbH) Al2O3
R (K) R ′(kW m–1) R ″′′(mm)
73 2.19 0.23
MgO
46 3.9 0.28
ZrO2
SiC
324 2.7 0.11
Bi3N4 HPSN
RBSN
206 495 66 75 0.12 0.11
342 20 0.10
BeO
Al2TiO5
47 36 0.71
962 9.6
have also been proposed that characterise the thermal shock behaviour of brittle materials under a range of different conditions, and these can be found in reviews such as that of Wang and Singh (1994). Alternative analyses, aiming to combine the two different approaches, have been performed using fracture mechanics concepts. Hasselman (1969) considered a brittle solid that contained circular, uniformly distributed Griffith microcracks. Crack instability due to thermal shock was assumed to take place by the simultaneous radial propagation of N cracks of radius l in a unit volume. Hasselman proposed that the driving force for crack propagation is derived from the elastic energy stored in the body at the instant of fracture. The total energy per unit volume of a body is the sum of the elastic energy and the fracture energy of the cracks, i.e.:
Wt =
3( L ∆T ) 2 E 16(1 – ν 2 ) N L3 1+ 9(1 – 2ν ) 2(1 – 2ν )
–1
+ 2π NL2 γ
(15.27)
Cracks are unstable between those limits for which:
dWt =0 dL
(15.28)
Combining (15.27) and (15.28), we get the critical quenching temperature difference as:
π G (1 – 2ν ) 2 ∆Tc = 2 2 2 Eα (1 – ν ) L
1/2
128π G (1 – ν 2 ) N 2 L5 + 81Eα 2
1/2
(15.29) The analysis showed that a material containing short cracks, much smaller than a characteristic length Lm, would propagate in an unstable manner at ∆Tc, due to the released elastic energy being converted into kinetic energy, towards a final crack length Lf and cause a drastic reduction in strength. For initially longer cracks, i.e. L > Lm, or for short cracks that have reached Lf,
© Woodhead Publishing Limited, 2006
412
Ceramic matrix composites
quenching at higher values of ∆T causes propagation in a stable, quasi-static manner and the material shows a gradual strength decrease above the ∆Tc associated with that particular crack size. In addition, it was shown that thermal shock resistance increased with increasing initial microcrack density. Evans (1975), Evans and Charles (1977), and Emery (1980) performed more refined fracture mechanics studies regarding the onset and arrest conditions; Bahr et al. (1988) and Pompe (1993) extended this work and considered the propagation of multiple cracks; while Swain (1990) found that materials showing non-linear deformation and R-curve behaviour have a better resistance to thermal shock. More specifically, the behaviour of a crack in the thermal shock-induced stress field was deduced from the dependence of the crack length on the stress intensity factor. Unstable propagation of a flaw in a brittle material under conditions of thermal shock was assumed to occur when the following criteria were satisfied: K > Ki ,
dK dK i > dL dL
(15.30)
where K is the thermal stress intensity factor, Ki is the material fracture toughness (Ki) or the crack length-dependent critical stress intensity factor (KR) for materials that exhibit R-curve behaviour, and α is the crack length. In the case where K > Ki ,
dK dK i < dL dL
(15.31)
the flaw will propagate in a stable manner. Thus, by superimposing the Kicurve of a material onto curves that describe the K-curve behaviour generated for a given thermal shock treatment, conditions of crack propagation and arrest were predicted. Such analyses verified Hasselman’s findings but were able to define onset and arrest conditions with better accuracy. The analyses were also found to correlate well with experimental findings. High-performance engineering ceramics usually have high stiffnesses combined with low values of toughness. Due to careful processing conditions the number and size of flaws they contain are limited, ensuring high strengths. The extent of the temperature differentials that they can sustain without cracking is then dictated mainly by the values of CTE and, to a lesser degree, by the values of thermal conductivity. Materials with lower CTE and high thermal conductivities can sustain higher values of ∆T. However, all such materials suffer large and abrupt losses of strength at ∆Tc as crack propagation occurs in an unstable fashion. By contrast, refractory or porous ceramic materials usually have low stiffnesses and contain a lot of large flaws or pores. These materials do not show a definite ∆Tc but exhibit a gradual reduction in strength starting at low values of ∆T. Examples of both types of materials are given in Fig. 15.4.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
413
∆Tc′
∆Tc
Strength (MPa)
400
200
0 0
200
400 ∆T(°C)
600
800
(a) 50
Strength (MPa)
40
30
20
Sintering temperature 2100°C 2000°C 1900°C
10
0
0
200
400 600 ∆T(°C)
800
(b)
15.4 The thermal shock behaviour of (a) monolithic alumina, and (b) porous SiC at different sintering temperatures (reprinted from Figure 11.6 on p 212 of ‘Ceramics: Mechanical Properties, Failure Behaviour, Materials Selection’ by Munz and Fett, 1999, published with permission from Springer-Verlag GmbH.
15.5
Thermal shock of particle- and whisker-reinforced CMCs
The attraction of reinforcing ceramic matrices with particles or whiskers is that, with appropriate microstructural design and property tailoring, materials with property combinations not possible in monolithic ceramics can be obtained. In addition, the materials remain effectively isotropic and can be manufactured by well-established techniques already in use for the manufacture of monolithic ceramics (Hansson and Warren, 2000).
© Woodhead Publishing Limited, 2006
414
Ceramic matrix composites
It is the capability of property tailoring that gives particle- and whiskerreinforced CMCs the edge over monolithic ceramics under conditions of thermal shock. By choosing carefully the properties of the reinforcement, reductions in Young’s modulus and CTE combined with increases in thermal conductivity compared with the unreinforced matrix material can be realised. Strict microstructural control during processing can result in fully dense, finely grained materials with good adhesion between reinforcement and matrix that ensure high strengths. In this way, high critical temperature differentials for crack initiation can be achieved. In addition, the presence of the reinforcement results in the introduction of a number of energy-dissipating mechanisms such as crack deflection, crack bridging, etc., which significantly improve toughness and damage tolerance. Thus, better resistance to crack propagation compared with monolithic ceramics is also possible. The result is a material that can sustain higher values of ∆T and, in addition, retain a higher percentage of its initial strength at ∆T > ∆Tc compared with its monolithic equivalent. A number of experimental studies support the above analysis. Aghajanian et al. (1989) reported that the ∆Tc of low-porosity alumina–matrix CMCs reinforced with aluminium particles increased compared with unreinforced alumina, while porous CMCs with the same constituents behaved as refractory ceramics, i.e. displayed a low, but not definite, ∆Tc and gradual reduction in strength with increasing ∆T. Similar refractory-type behaviour was observed by Aldridge and Yeomans (1999) in the case of a sintered alumina–matrix composite reinforced with 20 vol% iron particles that contained increased levels of porosity. However, a similar hot-pressed CMC with low porosity exhibited much higher ∆Tc and higher strength retention at ∆Tc compared with monolithic alumina (Fig. 15.5). Jin and Batra (1999) showed theoretically that crack bridging by metal particles resulted in a significant reduction of the thermal shock-induced stress intensity factor. Bannister and Swain (1990) and Swain (1991) investigated the thermal shock behaviour of ZrO2-particle-reinforced Al2O3 and AlN- and BN-particlereinforced TiB2 and reported higher thermal shock resistance compared with the respective monoliths as well as no significant reduction in post-shock flexural strength. This was attributed to the materials exhibiting R-curve behaviour due to the formation of microcracks around the reinforcing phases. Wang et al. (2001) found improved resistance to thermal shock (by ~70°C) of a 6 vol% tungsten carbide particle reinforced alumina compared with the unreinforced material, which was consistent with higher toughness, reductions in Young’s modulus and CTE, as well as the strong bonding of the reinforcement particles to the matrix (so that they did not act as strength-reducing flaws). Similar reasons were put forward by Uribe and Baudin (2003) to explain the increased thermal shock resistance of an alumina–matrix CMC reinforced with 10 vol% aluminium titanate particles, and also by Nieto et al. (2004),
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
415
800 Hot-pressed Al2O3 Hot-pressed Al2O3 – Fe Sintered Al2O3 – Fe
Retained strength (MPa)
700 600 500 400 300 200 100 0 0
100
200
300 400 500 600 Temperature differential (°C)
700
800
15.5 The thermal shock behaviour of hot-pressed alumina, hotpressed and sintered alumina reinforced with iron particles (reprinted from Journal of the European Ceramic Society, 19, Aldridge and Yeomans, ‘The thermal shock behaviour of ductile particle toughened alumina composites’, 1769–1775, copyright 1999, with permission from Elsevier).
who observed increased resistance to crack initiation and stable crack propagation in an alumina containing 10 vol% sub-micron-sized AlN particles. In the case of alumina/silicon carbide-particle nanocomposites, Maensiri and Roberts (2002) observed superior resistance to thermal shock compared with the matrix material. However, since no changes in thermal or mechanical properties could be identified, the improvement was associated with a change in crack path (intergranular in alumina, transgranular in the nanocomposite). Similar observations have been made regarding the thermal shock behaviour of whisker-reinforced CMCs. Tiegs and Becher (1987) reported no decrease in flexural strength following quenches into boiling water for a 20 vol% SiC whisker-reinforced Al2O3. The authors noted that a small increase in thermal conductivity and a slight decrease in CTE of the composite, compared with the matrix material, could not account for the extent of the improvement in thermal shock resistance and attributed it to the interaction of microcracks with the reinforcement (i.e. crack arrest, deflection, etc.), which resulted in increased toughness. Similar conclusions were drawn by Collin and Rowcliffe (2001) who tested the same material using the indentation-quench method. Pettersson and Johnson (2003) identified a clear correlation between increased toughness and pronounced R-curve behaviour with improved thermal shock resistance to explain the behaviour of alumina reinforced with Ti (C, N)
© Woodhead Publishing Limited, 2006
416
Ceramic matrix composites
whiskers. Zhao et al. (2002) investigated the thermal shock resistance of β– sialon matrix composites reinforced with titanium carbonitride whiskers and noticed that the addition of whiskers had no influence on the matrix microstructure, but their presence improved both the hardness and the fracture toughness of the CMCs. No unstable crack extension occurred in the composites for ∆T = 90–700°C, but above 700°C performance deteriorated as a result of severe oxidation of the whiskers. Studies have also shown that, since the amount of reinforcement added affects all mechanical and thermal properties, there is an optimum volume fraction of particle or whisker reinforcement that should be added to the matrix material to ensure superior resistance to thermal shock (Becher, 1981; Jia et al., 1996; Sbaizero and Pezzotti, 2003; Pettersson and Johnson, 2003). The shape of the reinforcement also plays an important role in determining behaviour under thermal shock. Sbaizero and Pezzotti (2003) showed that the use of coarse and elongated particles resulted in better CMC performance compared with the use of fine-grained particles. The importance of careful tailoring of the constituents to achieve improved thermal shock resistance is highlighted by the study of Jia et al. (1996). The incorporation of SiC whiskers in Si3N4 resulted in the CMC having a lower ∆Tc than monolithic Si3N4, as the CMC had a slightly lower thermal conductivity but a much larger CTE compared with the unreinforced matrix. However, stable crack growth occurred in the CMC in contrast to unstable crack growth in the monolithic material, which was attributed to the presence of the reinforcement. It was concluded that unreinforced Si3N4 is more suitable for use under mild thermal shock conditions, where the objective is to avoid fracture, while the CMC should be used under severe thermal shock conditions, where initiation of cracking is unavoidable and resistance to crack propagation and post-shock strength retention become important. The behaviour of particle- and whisker-reinforced CMCs under conditions of thermal shock can be modelled successfully using the fracture mechanics methods outlined in the previous paragraph (e.g. Aldridge and Yeomans, 2001) while the thermal shock parameters (figures-of-merit) can also be useful for initial material comparison.
15.6
Thermal shock of fibre-reinforced CMCs
15.6.1 Introduction Although particle- and whisker-reinforced CMCs can exhibit better thermal shock behaviour compared with monolithic ceramics, generally they still show a step decrease in their strength at ∆Tc. A combination of the properties of high-performance engineering ceramics with high ∆Tc and gradual strength reduction above ∆Tc (i.e. refractory-type behaviour) can only be realised
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
417
with the incorporation of continuous ceramic fibres into ceramic matrices. With optimum selection of fibres and matrices, favourable residual stress conditions can be established in the matrix, which lead to increased ∆Tc. Above ∆Tc, matrix cracks appear but the presence of crack-deflecting fibrematrix interfaces ensures minimal effect on mechanical properties as the fibres remain largely unaffected. As damage is also confined mostly to the surface of the materials, changes in mechanical and thermal properties are more readily identified by means other than mechanical testing. In the following paragraphs an overview of damage due to thermal shock and its effect on the mechanical properties of CMCs with different fibre architectures is provided for a number of different reinforcement architectures. Subsequently, the effect of thermal shock on interfacial properties is discussed, followed by a description of attempts to analyse and model the thermal shock behaviour of these materials.
15.6.2 Thermal shock damage and its effect on mechanical and thermal properties Unidirectional (UD) CMCs Bhatt and Phillips (1990) reported that thermal shock reduced the flexural mechanical properties of a UD composite comprising SiC fibres in a reactionbonded Si3N4 matrix but that it did not affect its tensile properties (Young’s modulus, ultimate strength, matrix cracking stress). It was suggested that the loss in flexural strength was caused by the loss of inter-ply integrity of the composite after matrix fracture and the failure mode changing from a tensile fracture to delamination driven by shear stress. Matrix cracking due to thermal shock and its effect on the flexural properties of UD Nicalon™ fibre-reinforced composites with borosilicate glass (Pyrex™) and lithium aluminosilicate (LAS) matrices was described by Kagawa et al. (1989, 1993). Damage was confined to the surface (two to three fibre diameters deep) and was independent of ∆T. The Pyrex™–matrix system exhibited multiple matrix cracking perpendicular to the fibre axis at ∆Tc = 600°C, which coincided with a notable decrease in Young’s modulus E and flexure strength. The decrease in E was attributed to matrix crack formation on the specimen surface, but the reduction in flexure strength was explained as a change in failure mode to interlaminar shear failure, caused by a reduction in interfacial shear strength due to thermal shock. In the LAS–matrix system matrix cracks parallel to the fibre axis were observed at ∆Tc = 800°C, accompanied by a reduction in Young’s modulus, although flexure strength seemed to remain unaffected by thermal shock treatment. This was attributed to the difference in the direction of matrix crack propagation in the two composites due to the formation of α-spodumene–
© Woodhead Publishing Limited, 2006
418
Ceramic matrix composites
silica solution in the LAS matrix during thermal shock, which could have acted as a source of microcracking because of thermal expansion mismatch. Multiple matrix cracking perpendicular to the fibre axis was also reported by Blissett et al. (1997) for a UD Nicalon™/CAS (calcium aluminosilicate) (Fig. 15.6). The density of these cracks increased with increasing ∆T but showed a reduction for ∆T > 800°C, which seemed to be consistent with the formation of strong silica bridging between the matrix and the fibres. In addition to matrix cracking perpendicular to the fibre axis, matrix cracks also occurred parallel to the mid-plane of the laminate. These cracks were first seen on the end faces of the composite at ∆Tc = 400°C (Fig. 15.7). The depth the cracks penetrated into the matrix increased and their path geometries changed with increasing ∆T. These effects were attributed to the interaction of increasing applied thermal stresses with simultaneous reductions in the interfacial shear strength due to oxidation of carbon. Similar damage modes, termed ‘thermal debond cracks’, were observed by Graham et al. (2003) on the end face of a thermally shocked UD Nicalon™/LAS II composite. The authors highlighted the presence of high tensile radial stresses across the fibre–matrix interface, which favoured the appearance of such cracks, and noted that they tended to run through fibre-rich regions where these stresses are highest. Reductions in thermal diffusivity due to thermal debond crack formation were measured. The appearance of such damage modes may also
20 µm
15.6 Multiple matrix cracking perpendicular to the fibre axis due to thermal shock in UD Nicalon™/CAS (reprinted from Journal of Materials Science 32(2) 1997, ‘Thermal shock behaviour of unidirectional silicon carbide reinforced calcium aluminosilicate’ Blissett, Smith and Yeomans, Figure 2, with kind permission of Springer Science and Business Media).
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
419
50 µm (a)
(b)
15.7 (a) Photomicrograph, and (b) schematic of matrix cracking on end face of UD Nicalon™/CAS (reprinted from Journal of Materials Science 32(2) 1997, ‘Thermal shock behaviour of unidirectional silicon carbide reinforced calcium aluminosilicate’, Blissett, Smith and Yeoman, Figure 1a, with kind permission of Springer Science and Business Media).
be responsible for the change in failure mode under flexure reported for other systems (Bhatt and Phillips, 1990; Kagawa et al., 1993). Blissett et al. (1998) reported that thermal shock effects on the residual flexural properties of the Nicalon™/CAS were more evident at intermediate temperature differentials, i.e. ∆T = 450–600°C, and this was attributed to the observed matrix cracking.
© Woodhead Publishing Limited, 2006
420
Ceramic matrix composites
Cross-ply CMCs Blissett (1995) studied cross-ply Nicalon™/CAS laminates of two different lay-ups and reported that thermal shock damage within individual plies was similar to that seen in UD specimens of the same material (Blissett et al., 1997). Initial damage in a [0 °2 /90 °4 ]s laminate was sustained at ∆Tc = 400°C in the eight central 90° plies, and consisted of a single thermal debond crack similar to the ones observed on the end faces of UD Nicalon™/CAS (Fig. 15.8). As this damage mode is not observed under monotonic tensile or fatigue loading applied along the axis of the longitudinal fibres in the 0° plies (e.g. Pryce and Smith, 1992), its appearance is indicative of the biaxial nature of the thermal shock-induced stress field. Short cracks just crossing the interface between plies as a result of thermal shock treatment were also reported. It was noted that both damage modes became more pronounced at higher values of ∆T. The second laminate, [0°/90°]3s, exhibited only slightly different cracking features, attributed to the difference in the stacking sequences of the laminates. A major thermal debond crack appeared at ∆Tc = 350°C and was confined to the two central 90° plies. Similar cracks were observed in some of the adjacent 90° plies but were less pronounced. At ∆T = 400°C matrix cracks perpendicular to the longitudinal fibres appeared in the 0° plies. For higher values of ∆T, debond cracks were observed in most of the other 90° plies while the perpendicular matrix cracks were seen crossing to the adjacent 90° plies before being arrested by the horizontally running debond cracks. However, the outer plies and the thinner 0° plies seemed to remain intact up to ∆T =
500 µm
15.8 Photomicrograph of thermal debond crack in the eight central 90° plies of a [02° / 90°4 ]s , Nicalon™/CAS laminate (Blissett, 1995, reprinted courtesy of Dr M.J Blissett).
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
421
720°C. Flexure strength and Young’s modulus were found to decrease with increasing ∆T (Blissett et al., 1998). 2-D and 3-D woven CMCs CMCs with 2-D woven fibre reinforcements have been found to possess higher resistance to thermal shock than unidirectional or cross-ply CMCs of the same constituents (Nicalon™ fibres and SiC matrices) and prepared by the same method (Chemical Vapour Infiltration-CVI) (Wang et al., 1997). Only a slight drop in the flexural strength of a woven Nicalon™/Al2O3 composite was observed by Fareed et al. (1990) after quenching through ∆T = 1000°C and 1200°C. This was attributed to the effectively engineered weak fibre/matrix interface. Lamicq et al. (1986) reported that the bending strength of water-quenched woven SiC/SiC (CVI) specimens decreased slightly in the quench range ∆T = 300–750°C, and then remained unchanged up to ∆T = 1200°C. The composite also seemed to exhibit a steep R-curve behaviour. Wang et al. (1994, 1996) reported on the thermal shock behaviour of 2-D woven Nicalon™/SiC CMCs manufactured by CVI and polymer impregnation and pyrolysis (PIP), as well as that of a Nextel™–312/SiC (CVI) composite system. The Nextel™/SiC (CVI) system failed in post-quench flexure tests by fracture through the 2-D fibre planes and showed different critical temperature differentials for the onset of decrease in each of its macroscopic properties. Reduction in ultimate strength, σu, began at ∆Tc(σu) = 400°C, matrix cracking stress, σmc, started to decrease at ∆Tc(σmc) = 600°C, while the work of fracture (WOF) decreased continuously as ∆T increased. Reductions in thermal diffusivity with increasing values of ∆T were also reported for this system by Ellingson (1995). The properties of the Nicalon/SiC (PIP) system followed a similar pattern (∆Tc(σu) = 400°C, ∆Tc(σmc) = 500°C), though this system failed through an interlaminar shear failure process (delamination) and the property reduction saturated at ∆T = 600°C. The Nicalon™/SiC (CVI) system failed by fracture through fibre planes but its properties (σu, σmc, WOF) had the same critical temperature difference, ∆Tc = 700°C. The pre- and post-quench stress– displacement curves for this material can be seen in Fig. 15.9. However, measurement of the Young’s modulus of this system before and after quenching by means of a dynamic mechanical resonance technique showed the onset of decrease at ∆Tc(E) = 400°C, i.e. significantly lower than the ∆Tc of the other properties. An assessment of the thermal shock damage of woven Nicalon™/SiC (CVI) composite specimens was performed by Webb et al. (1996), the results being confirmed subsequently by Kagawa (1997). It was noted that the many pores and irregularities in the matrix inherent to this particular composite
© Woodhead Publishing Limited, 2006
422
Ceramic matrix composites 600
Stress (MPa)
500 400 300 200 100 0 0
0.2
0.4
0.6
0.8 1 1.2 1.4 Displacement (mm)
1.6
1.8
2
15.9 Effect of increasing ∆T on stress–displacement curves of Nicalon™/SiC (CVI) – solid line corresponds to unshocked sample (reprinted from Wang et al. 1996, ‘Thermal shock behaviour of two-dimensional woven fiber-reinforced ceramic composites’, Journal of the American Ceramic Society, with kind permission of Blackwell Publishing).
geometry provide stress concentrators that amplify the thermal loading and create preferential sites for crack formation. For this reason, CVI-SiC composites exhibit lower ∆Tc for the onset of cracking than monolithic SiC (Kagawa, 1997). Three types of thermal shock-induced damage on the material surface were reported: • Matrix cracks that originated from the corners of uninfiltrated pores in regions outside fibre bundles. These cracks appeared at ∆T = 250°C and did not penetrate deeply into the fibre bundles, though the penetration depth increased with increasing ∆T. • Matrix cracks between fibres within a fibre bundle. These occurred at ∆T = 1000°C and were similar to thermal shock damage observed by Kagawa et al. (1993) in UD CMCs. • Degradation of the fibre–matrix interface and removal of fibres. This type of damage appeared at ∆T = 600°C but was attributed to both thermal shock and/or oxidation effects. In addition, matrix cracks that severed ligaments between cloths were seen at ∆T ≥ 600°C in the interior of thermally shocked specimens. The mechanism of formation of these cracks is not clear as thermal shock loading induces mainly high stresses at or near the surface. However, Kastritseas et al. (2004a) observed such cracks on polished parallel surfaces of similar SiC/SiC CMCs as well. Webb et al. (1996) reported that further increases in ∆T increased the severity of all types of thermal shock damage.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
423
Property
σ
E Q–1
ε ∆Tc
∆T
15.10 Schematic diagram showing the variation of fracture strength (σ), Young’s modulus (E), internal friction (Q–1), and microcracking density (ε) with increasing shock severity (reprinted from Scripta Materialia, 38, Boccaccini, ‘Predicting the thermal shock resistance of fibre reinforced brittle matrix composites’, 1211–1217, copyright 1998, with permission from Elsevier).
Correlation of these observations with property measurements by Wang et al. (1996) led to the postulation that surface matrix cracks that appear at low ∆T (= 250°C) are not strength-controlling but are responsible for the reduction in Young’s modulus observed at ∆Tc(E) = 400°C. On the other hand, the interior cracks that severed links between fibre cloths at ∆T ≥ 600°C seem to affect the strength of the composite, which decreases after ∆Tc = 700°C. Such behaviour was summarised by Boccaccini (1998) in the graph of Fig. 15.10. It has to be noted that the behaviour of E is also typical of some thermal properties of CMCs (Ellingson, 1995; Graham et al., 2003). Note that there is no abrupt change in any property above ∆Tc. However, if the fibre–matrix interface is strong, fibre-reinforced CMCs revert to behaviour typical of monolithic ceramics (Twitty et al., 1995). Damage modes resulting from thermal shock were identified by Kastritseas et al. (2004b) for a woven Nicalon™/CAS. The surface of the CMC was described as an assembly of alternating ‘plies’ containing either longitudinal or transverse fibres, with matrix-rich regions in between. Multiple matrix cracks confined to the surface appeared perpendicular to the fibre direction at ∆Tc for the plies with longitudinal fibres and for the matrix-rich regions, while debond cracks running parallel to the longitudinal fibres could be seen in the plies with transverse fibres located towards the centre of the material face. With increasing values of ∆T, debond cracks grew significantly in length and depth (Fig. 15.11) but their number did not change significantly. By contrast, perpendicular matrix cracks did not change in morphology (i.e. they remained surface features of small depth) but increased moderately in length and significantly in number. Crimp regions were not observed to affect either crack initiation or development.
© Woodhead Publishing Limited, 2006
424
Ceramic matrix composites
(a)
(b)
15.11 Matrix cracking due to thermal shock in a 90° ply of a woven Nicalon™/CAS at (a) ∆T = 700°C, and (b) ∆T = 800°C (after Kastritseas et al, 2004b).
The thermal shock behaviour of a 3-D carbon fibre-reinforced SiC-matrix CMC manufactured by CVI was assessed using the air-quench method by Yin et al. (2002). Damage consisted of matrix cracks that induced a reduction in Young’s modulus, strength, and work of fracture for ∆T > 700°C. Studies of the interface It appears that the strength of the fibre–matrix bond in thermally shocked CMCs remains unaffected unless high-temperature oxidation processes are involved. Boccaccini et al. (1999) did not identify any significant changes in the properties of the fibre–matrix interface of SiC/borosilicate glass composites as a result of thermal shock, while Chawla et al. (2001) observed only a slight decrease in the interfacial shear stress of a thermally shocked Nicalon™fibre SiC-whisker BMAS (barium magnesium aluminosilicate)–matrix hybrid composite. If heating and soaking at temperatures harmful to the integrity of the interface are involved prior to quenching, degradation due to oxidation processes occurs (Blissett et al., 1997, 1998). In this case, changes in properties are explained as a combination of both oxidation and thermal shock (Graham et al., 2003). The oxidation of the carbon interface in Nicalon™-reinforced glass ceramic–matrix CMCs leads to cracks in the matrix, causing fibre
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
425
failure due to the resulting strong interfacial bond (Blissett et al., 1997; Kastritseas et al., 2004b).
15.6.3 Theoretical considerations Only a few studies have appeared in the literature regarding the analysis and modelling of the thermal shock behaviour of fibre-reinforced CMCs. Wang and Chou (1991) studied numerically the 3-D transient thermal stress in angle-ply laminated composites caused by sudden changes in the thermal boundary conditions. The study showed that ∆Tc would be reduced if the fibre volume fraction, Vf, the CTE or the Young’s modulus of the composite increased, while it would increase with increasing thermal conductivity. By contrast, Boccaccini (1998) showed that increasing Vf increased the ∆T for the onset of matrix cracking in glass and glass–ceramic matrix composites. Wang and Chou (1991) also demonstrated that the change in CTE had the biggest effect on ∆Tc while the change in thermal conductivity had the least influence. In addition, as the fibre orientation angle deviates from 45° towards 90° or 0° the interlaminar normal stress decreased while the in-plane thermal stress transverse to the fibre direction increased. This resulted in the initial failure mechanism changing from delamination to matrix micro-cracking. Wang et al. (1996) performed a 1-D qualitative analysis using the stresses generated due to thermal shock and the residual stresses associated with the thermal expansion mismatch between the fibres and the matrix. The analysis showed that if (CTE)f > (CTE)m then the matrix is under tension only in the radial direction and possible matrix cracking will be circumferential, while the fibre is under tension in all directions (longitudinal, radial, and circumferential), which may promote fibre damage. If (CTE)f < (CTE)m, then the matrix is under tension in both longitudinal and circumferential directions; hence, radial and normal-to-fibre matrix cracking will be possible. Moreover, the fibre is under compression in all three directions, so fibre damage will be limited. Debonding would also be possible in both cases. As a confirmation, they applied their analysis to the results of Kagawa et al. (1993). In the Nicalon™/Pyrex™ composite, where (CTE)f < (CTE)m, the matrix is under a tensile stress in the longitudinal direction, which dictates that cracks will be perpendicular to the fibre axis, as observed in the experiment. Conversely, in the LAS–matrix composite (CTE)f > (CTE)m, i.e. the matrix is under tension in the radial direction, which results in cracks parallel to the fibre. Particular interest has been paid to the analytical prediction of the ∆Tc for the onset of matrix cracking. Blissett et al. (1997) and Boccaccini (1998) considered the residual stresses present in the composite due to thermal expansion mismatch between fibre and matrix which, when superimposed to
© Woodhead Publishing Limited, 2006
426
Ceramic matrix composites
the applied thermal stresses, could lead to matrix cracking. Their approach was based on the assumption that the stress that produces matrix cracking would be the same whether applied mechanically or thermally. Hence, the matrix cracking stress (σmu) was equated with the critical thermal shockinduced stress (σTS), which is the thermal stress required to produce matrix cracking, taking also into account the effect of residual stress (σr), i.e.
σ mu = σ cTS + σ r
(15.32)
Following equation (15.18), the critical thermal shock-induced stress is given as:
σ cTS =
AEα ∆Tc 1–ν
(15.33)
For Blissett et al. (1997), E, a, and v are matrix properties whereas Boccaccini (1998) defines ‘effective’ values calculated using the rule of mixtures. Blissett et al. (1997) used the concentric cylinder model of Powell et al. (1993) to obtain residual stresses, whereas Boccaccini (1998) utilised the results of a simple force balance in 1-D performed by Wang et al. (1996), which gives the residual thermal stresses in the matrix along the axial direction as:
σ rmatrix =
E m ∆α ∆ T 1 + E m (1 – Vf ) E f Vf
(15.34)
Different models were also used to obtain the matrix cracking stress (σmu), with Blissett et al. (1997) using the classic Aveston et al. (1971) (ACK) analysis and Boccaccini (1998) using the model of Pagano and Kim (1994), which gives σmu as:
σ mu =
K IC 2 r+s π
(15.35)
where KIC is the fracture toughness of the matrix, r is the fibre radius, and s is the fibre spacing. The model assumes that there is no interaction between cracks, which Boccaccini (1998) explains as a plausible assumption in the early stages of thermal shock damage. By solving the resulting expressions for ∆Tc, the values of the critical temperature differentials are obtained. These are given as:
∆ Tc = 1 – v AE m α m
© Woodhead Publishing Limited, 2006
1 2 3 6τ Γm E f E m Vf – σ (Blissett et al., 1997) r E1 rVm (15.36)
Thermal shock of ceramic matrix composites
427
E m ∆α ∆T 1 – v e K IC,m (Boccaccini, 1998) ∆ Tc = – AE e α e E (1 – Vf ) 2 r + s 1 + m E f Vf π (15.37) In (15.36) Γm is the matrix fracture energy, τ is the interfacial shear strength, and E1 is the axial modulus of the composite. In (15.37) e refers to the effective properties of the composite, which, for unidirectional fiber reinforcement, can be calculated with good approximation by the rule of mixtures. Although the two approaches are very similar, the value of ∆Tc in Boccaccini’s model does not depend on the interfacial shear strength τ, as a result of the model chosen for the value of matrix cracking stress. Blissett et al. (1997) suggested that their method was valid for the UD material providing that some key parameters (interfacial shear stress, matrix fracture energy) were determined independently. A recent analysis by Kastritseas et al. (2004c) suggested that in both cases the magnitude of the thermal shock-induced stresses was overestimated as the anisotropic character of the materials was not taken into account. If material anisotropy is accounted for, then both (15.36) and (15.37) cannot predict ∆Tc accurately even for the largest possible value of the thermal shock-induced stresses (corresponding to a maximum value of the stress reduction factor, A = 0.66). To explain the discrepancy, it was proposed that the interfacial properties may be affected by the shock due to the biaxial nature of the induced stress field, which dictates that a tensile thermal stress component that acts perpendicular to the fibre–matrix interface is present for the duration of the shock.
15.7
Concluding remarks
This chapter has reviewed the performance of CMCs under conditions of thermal shock. It has been shown that CMCs exhibit superior resistance to thermal shock, compared with their monolithic counterparts, as catastrophic failure can always be avoided. Resistance to higher temperature differentials and property retention after the onset of thermal shock cracking (especially in fibre-reinforced CMCs) can be realised, provided that the mechanical and thermal properties of CMCs are optimised by careful choice of their constituents. The behaviour of particle- and whisker-reinforced CMCs can be adequately described by using and adapting the models and methodology developed for monolithic ceramics. By contrast, analysis and modelling of the performance
© Woodhead Publishing Limited, 2006
428
Ceramic matrix composites
of fibre-reinforced CMCs is a subject still in its infancy that requires further attention. The situation is very complex due to the variety of damage mechanisms developed in these materials (especially 2-D CMCs) and is further complicated due to their anisotropic character, the scarcity of experimental results, and the variety of manufacturing methods that result in materials with different design philosophies.
15.8
References
Absi, J., Glandus, J.C. (2004), ‘Improved method for severe thermal shocks testing of ceramics by water quenching’, J. Eur. Ceram. Soc., 24(9), 2835–2838. Aghajanian, M.K., Macmillan, N.H., Kennedy, C.R., Luszcz, S.J., Roy, R. (1989), ‘Properties and microstructure of Lanxide® Al2O3–Al ceramic composite materials’, J. Mater. Sci., 24, 658–670. Aldridge, M., Yeomans, J.A. (1999), ‘The thermal shock behaviour of ductile particle toughened alumina composites’, J. Eur. Ceram. Soc., 19, 1769–1775. Aldridge, M., Yeomans, J.A. (2001), ‘Thermal shock behaviour of iron-particle-toughened alumina’, J. Am. Ceram. Soc., 84(3), 603–607. Andersson, T., Rowcliffe, D.J. (1996), ‘Indentation thermal shock test for ceramics’, J. Am. Ceram. Soc., 79(6), 1509–1514. Aveston, J., Cooper, G.A., Kelly, A. (1971), ‘Single and multiple fracture’, in The Properties of Fiber Composites, IPC Science & Technology Press, Guildford, UK, 15–22 Bahr, H.A., Weiss, H.J., Maschke, H.G., Meissner, F. (1988), ‘Multiple crack propagation in strip caused by thermal shock’, Theor. Appl. Fract. Mech., 10, 219–226. Bannister, M.K., Swain, M.V. (1990), ‘Thermal shock of a titanium diboride based composite’, Ceram. Int., 16, 77–83 Baste, U. (1993), ‘Thermal shock and cyclic loading of ceramic parts in stationary gas turbines’, in Schneider G. A. and Petzow G. (editors), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic, 87–97 Becher, P.F. (1981), ‘Transient thermal stress behavior in ZrO2-toughened Al2O3’, J. Am. Ceram. Soc., 64(1), 37–39. Becher, P.F., Warwick, W.H. (1993), ‘Factors influencing the thermal shock behavior of ceramics’, in Schneider, G.A. and Petzow, G. (editors), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic 37–48 Becher, P.F., Lewis, D.III., Carman, K.R., Gonzalez, A.C. (1980), ‘Thermal shock resistance of ceramics: size and geometry effects in quench tests’, Am. Ceram. Soc. Bull., 59(5), 542–545 Bhatt, R.T., Phillips, R.E. (1990), ‘Thermal effects on the mechanical properties of SiC fibre reinforced reaction-bonded silicon nitride matrix composites’, J. Mater. Sci., 25, 3401–3407. Blissett, M.J. (1995), Aspects of the Thermal Shock Behaviour of Continuous FibreReinforced Glass-Ceramic Composites, PhD Thesis, University of Surrey. Blissett, M.J., Smith, P.A., Yeomans, J.A. (1997), ‘Thermal shock behaviour of unidirectional silicon carbide fibre reinforced calcium aluminosilicate’, J. Mater. Sci., 32, 317–325. Blissett, M.J., Smith, P.A., Yeomans, J.A. (1998), ‘Flexural mechanical properties of thermally treated unidirectional and cross-ply Nicalon-reinforced calcium aluminosilicate composites’, J. Mater. Sci., 33, 4181–4190.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
429
Boccaccini, A.R. (1998), ‘Predicting the thermal shock resistance of fibre reinforced brittle matrix composites’, Scripta. Mat., 38(8), 1211–1217. Boccaccini, A.R., Pearce, D.H., Janczak, J., Beier, W., Ponton, C.B. (1997), ‘Investigation of cyclic thermal shock behaviour of fibre reinforced glass matrix composites using non-destructive forced resonance technique’, Mater. Sci. Tech., 13, 852–858. Boccaccini, A.R., Strutt, A.J., Vecchio, A.S., Mendoza, D., Chawla, K.K., Ponton, C.B., Pearce, D.H. (1998), ‘Behaviour of Nicalon-fiber-reinforced glass–matrix composites under thermal cycling conditions’, Composites, 29A, 1343–1352. Boccaccini, A.R., Pearce, D.H., Janczak, J., Beier, W., Ponton, C.B. (1997), ‘Investigation of cyclic thermal shock behaviour of fibre reinforced glass matrix composites using non-destructive forced resonance technique’, Mater Sci Tech, 13, 852–858. Boccaccini, A.R., Strutt, A.J., Vecchio, A.S., Mendoza, D., Chawla, K.K., Ponton, C.B., Pearce, D.H. (1998), ‘Behaviour of Nicalon™-fiber-reinforced glass–matrix composites under thermal cycling conditions’, Composites, 29A, 1343–1352. Boccaccini, A.R., Janczak-Rusch, J., Pearce, D.H., Kern, H. (1999), ‘Assessment of damage induced by thermal shock in SiC-fiber-reinforced borosilicate glass composites’, Comp. Sci. Tech., 59, 105–112. Boccaccini, A.R., Kern, H., Dlouby, I. (2001), ‘Determining the fracture resistance of fibre-reinforced glass matrix composites by means of the chevron-notch flexural technique’, Nater, Sci. Eng., A308 (1/2), 111–117. Carter, S., Ponton, C.B., Rawlings, R.D., Rogers, P.S. (1988), ‘Microstructure, chemistry, elastic properties and internal friction of Silceram glass-ceramics’, J . Mater. Sci., 23, 2622– 2630. Case, E.D. (2002), ‘The saturation of thermomechanical fatigue damage in brittle materials’, in Aliabadi, M.H. (editor), Thermomechanical Fatigue and Fracture, WIT Press, Southampton, Boston, 137–208. Chawla, N., Chawla, K.K., Koopman, M., Patel, B., Coffin, C., Eldridge, J.I. (2001), ‘Thermal-shock behaviour of a Nicalon-fiber-reinforced hybrid glass–ceramic composite’, Comp. Sci. Tech., 61, 1923–1930. Cheng, C.M. (1951), ‘Resistance to thermal shock’, J. Am. Rocket. Soc., 21, 147–153. Chlup, Z., Dlouhy, I., Boccaccini, A.R. (2001), ‘Fracture toughness of thermally shocked SiC-fibre reinforced glass matrix composite’, in Krenkel, W., Naslain, R., Schneider, H. (editors), High Temperature Ceramic Matrix Composites, Wiley, 463–468. Collin, K.M., Rowcliffe, D.J. (1999), ‘Analysis of indentation-quench test for ceramics’, Ceram. Eng. Sci. Proc., 20(3), 301–308. Collin, K.M., Rowcliffe, D.J. (2000), ‘Analysis and prediction of thermal shock in brittle materials’, Acta Mater, 48, 1655–1665. Collin, K.M., Rowcliffe, D.J. (2001), ‘Influence of thermal conductivity and fracture toughness on the thermal shork resistance of alumina-silicon-carbide-whisker compositer’ Journal of the American Ceramic Society, 84(6), 1334–1340. Ellingson, W., as cited in Singh, R.N., Webb, J.E., Wang, H., Anadakumar, U. (1995), ‘Thermal shock, damage is major concern’, CFCC News: A communication of the continuous fiber ceramic composite program, Oak Ridge National Laboratory, 6, 12– 15. Emery, A.F. (1980), ‘Thermal stress fracture in elastic–brittle materials’, in Hasselman, D.P.H. and Heller, R.A., Thermal Stresses in Severe Environments, Plenum Press, New York, 95–121. Evans, A.G. (1975), ‘Thermal fracture in ceramic materials’, Proc. Br. Ceram. Soc., 25, 217–237.
© Woodhead Publishing Limited, 2006
430
Ceramic matrix composites
Evans, A.G., Charles, E.A. (1977), ‘Structural integrity in severe thermal environments’, J. Am. Ceram. Soc., 60(1/2), 22–28. Evans, A.G., Linzer, M., Johnson, H., Hasselman, D.P.H., Kipp, M.E. (1975), ‘Thermal fracture studies in ceramic systems using an acoustic emission technique’, J. Mater. Sci., 10, 1608–1615. Faber, K.T., Huang, M.D., Evans, A.G. (1981), ‘Quantitative studies of thermal shock in ceramics based on a novel test technique’, J. Am. Ceram. Soc., 64(5), 296–301. Fareed, A.S., Sonuparlak, B., Lee, C.T., Fortini, A.J., Schiroky, G.H. (1990), ‘Mechanical properties of 2-D NicalonTM fiber-reinforced LANXIDETM aluminum oxide and aluminum nitride matrix composites’, Ceram. Eng. Sci. Proc., 11(7–8), 782–794. Gong, J.H., Guan, Z.D., Jiang. D.C. (1992), ‘Analysis of strength degradation of indented specimens due to thermal shock’, in Bradt, R.C., Hasselman, D.P.H., Munz, D., Sakai, M., Shevchenko, V.Y., Fracture Mechanics of Ceramics, Plenum Press, New York, 605–610. Graham, S., McDowell, D.L., Lara-Curzio, E., Dinwiddie, R.B., Wang, H., Porter, W. (2003), ‘Nondestructive characterization of thermal shock and oxidation-induced damage by flash diffusivity’, J. Comp. Mater., 37(1), 73–87. Hansson, T., Warren, R. (2000), ‘Particle and whisker reinforced brittle matrix composites’, in Kelly, A., Zweben, C. (editors-in-chief), Comprehensive Composite Materials, Vol. 6 (R Warren, editor), Elsevier. Hasselman, D.P.H. (1969), ‘Unified theory of thermal shock fracture initiation and crack propagation in brittle ceramics’, J. Am. Ceram. Soc., 52(11), 600–604. Hasselman, D.P.H. (1970), ‘Thermal stress resistance parameters for brittle refractory ceramics: a compendium’, Ceramic. Bulletin., 49(12), 1033–1037. Hasselman, D.P.H. (1978), ‘Figures-of-merit for the thermal stress resistance of hightemperature brittle materials: a review’, Ceramurgia, 4(4), 147–150. Hasselman, D.P.H. (1985), ‘Thermal stress resistance of engineering ceramics’, Mater. Sci. Eng., 71, 251–264. Hasselman, D.P.H., Chen, E.P., Urick, P.A. (1978), ‘Prediction of the thermal fatigue resistance of indented glass rods’, Am. Ceram. Soc. Bull., 57(2), 190–192. Hencke, H., Thomas, J.R., Hasselman, D.P.H. (1984), ‘Role of material properties in the thermal stress fracture of brittle ceramics subjected to conductive heat transfer’, J. Am. Ceram. Soc., 67, 393–398. Ishitsuka, M., Sato, T., Endo, T., Shimada, M. (1989), ‘Thermal shock fracture behaviour of ZrO2 based ceramics’, J. Mater. Sci. Lett., 24, 4057–4061. Jia, D.C., Zhou, Y., Lei, T.C. (1996), ‘Thermal shock resistance of SiC whisker reinforced Si3N4 ceramic composites’, Ceram. Int., 22, 107–112. Jin, Z-H., Batra, R.C. (1999), ‘Thermal shock cracking in a metal-particle-reinforced ceramic matrix composite’, Eng. Fract. Mech., 62, 339–350. Jones, R.H., Giancarli, L., Hasegawa, A., Katoh, Y., Kohyama, A., Riccardi, B., Snead, L.L., Weber, W.J. (2002), ‘Promise and challenges of SiCf/SiC composites for fusion energy applications’, J. Nucl. Mater., 307–311, 1057–1072. Kagawa, Y. (1997), ‘Thermal shock damage in a two-dimensional SiC/SiC composite reinforced with woven SiC fibers’, Comp. Sci. Tech., 57, 607–611. Kagawa, Y., Kurosawa, N., Kishi, T. (1989), ‘Thermal shock behaviour of SiC fiber(Nicalon ) reinforced glass’, Ceram. Eng. Sci. Proc., 10(9/10), 1327–1336. Kagawa, Y., Kurosawa, N., Kishi, T. (1993), ‘Thermal shock resistance of SiC fibrereinforced borosilicate glass and lithium aluminosilicate matrix composites’, J. Mater. Sci., 28, 735–741.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
431
Kastritseas, C., Smith, P.A., Yeomans, J.A. (2004a), ‘Thermal shock behaviour of woven fibre-reinforced SiC/CAS and SiC/SiC composites’, un published data. Kastritseas, C., Smith, P.A., Yeomans, J.A. (2004b), ‘Damage characterisation of thermallyshocked woven fibre-reinforced ceramic matrix composites’, Proceedings of the 11th European Conference in Composite Materials (ECCM-11), Rhodes, Greece, Vol. 2. Kastritseas, C., Smith, P.A., Yeomans, J.A. (2004c), ‘The onset of thermal shock damage in unidirectional fibre-reinforced ceramic matrix composites’, Proceedings of the 5th International Conference in High Temperature Ceramic Matrix Composites (HTCMC5), Seattle, WA, 235–240. Kingery, W.D. (1955), ‘Factors affecting thermal stress resistance of ceramic materials’, J. Am. Ceram. Soc., 38, 3–15. Konsztowicz, K.J. (1990), ‘Crack growth and acoustic emission in ceramics during thermal shock’, J. Am. Ceram. Soc., 73(3), 502–508. Konsztowicz, K.J. (1993), ‘Acoustic emission amplitude analysis in crack growth studies during thermal shock of ceramics’, in Schneider, G.A. and Petzow, G. (editors), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic, 429–441. Kreith, F. (1986), Principles of Heat Transfer, 4th Edition, Intext Educational Publishers, New York and London. Lamicq, P.J., Bernhart, G.A., Dauchier, M.M., Mace, J.G. (1986), ‘SIC/SIC Composite Ceramics’, American Ceramic Society Bulletin, 65(2), 336–338. Lee, S.K., Moretti, J.D., Readey, M.J., Lawn, B.R. (2002), ‘Thermal shock resistance of silicon nitrides using an indentation-quench test’, J. Am. Ceram. Soc., 85(1), 279–281. Lee, W.J., Case, E.D. (1989), ‘Cyclic thermal shock in SiC-whisker-reinforced alumina composite’, Mater. Sci. Eng., A119, 113–126. Lee, W.J., Case, E.D. (1990), ‘Thermal fatigue in polycrystalline alumina’, J. Mater. Sci., 25, 5043–5054. Lee, W.J., Kim, Y., Case, E.D. (1993), ‘The effect of quenching media on the heat transfer coefficient of polycrystalline alumina’, J. Mater. Sci., 28, 2079–2083. Lu, T.J., Fleck, N.A. (1998), ‘The thermal shock resistance of solids’, Acta. Mater., 46, 4755–4768. Maensiri, S., Roberts, S.G. (2002), ‘Thermal shock resistance of sintered alumina/silicon carbide nanocomposites evaluated by indentation techniques’, J. Am. Ceram. Soc., 85(8), 1971–1978. Manson, S.S. (1966), Thermal Stress and Low-cycle Fatigue, McGraw-Hill, New York. Morrel, R. (1993), ‘Thermal shock testing and the problem of standardisation’, in Schneider, G.A. and Petzow, G. (editors), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic, 27–33. Munz, D., Fett, T. (1999), Ceramics: Mechanical Properties, Failure Behaviour, Materials Selection, Springer. Nieto, M.I., Martinez, R., Mazerolles, L., Baudin, C. (2004), ‘Improvement in the thermal shock resistance of alumina through the addition of submicron-sized aluminium nitride particles’, J. Eur. Ceram. Soc., 24, 2293–2301. Ohnabe, H., Masaki, S., Onozuka, M., Miyahara, K., Sasa, T. (1999), ‘Potential application of ceramic matrix composites to aero-engine components’, Composites, 30A, 489– 496. Osterstock, F. (1993), ‘Contact damage submitted to thermal shock: a method to evaluate and simulate thermal shock resistance of brittle materials’, Mater. Sci. Eng., A168, 41–44.
© Woodhead Publishing Limited, 2006
432
Ceramic matrix composites
Pagano, N.J., Kim, R.Y. (1994), ‘Progressive microcracking in unidirectional brittle matrix composites’, Mech. Compos. Mater. Struct., 1(1), 3–29. Pettersson, P., Johnsson, M. (2003), ‘Thermal shock properties of alumina reinforced with Ti(C, N) whiskers’, J. Eur. Ceram. Soc., 23, 309–313. Pompe, W.E. (1993), ‘Thermal shock behaviour of ceramic materials – modelling and measurement’, in Schneider, G.A. and Petzow, G. (editors), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic, 3–14. Pompe, W., Bahr, H-A., Schneider, G., Weiss, H-J. (1993), ‘Modelling and measuring of the thermal shock behaviour of ceramics’, Cfi/Ber DKG, 70(3), 79–84. Powell, K.L., Smith, P.A., Yeomans, J.A. (1993), ‘Aspects of residual thermal stresses in continuous-fibre-reinforced ceramic matrix composites’, Comp. Sci. Tech., 47, 359– 367. Pryce, A.W., Smith, P.A. (1992), ‘Behaviour of unidirectional and crossply ceramic matrix composites under quasi-static tensite loading’, Journal of Materials Science, 27(10), 2695–2704. Rogers, W.P., Emery, A.F. (1992), ‘Contact thermal shock test of ceramics’, J. Mater. Sci., 27, 146–152. Sbaizero, O., Pezzotti, G. (2003), ‘Influence of molybdenum particles on thermal shock resistance of alumina matrix ceramics’, Mater. Sci. Eng., A343, 273–281. Schneibel, J.H., Sabol, S.M., Morrison, J., Ludeman, E., Carmichael, C.A. (1998), ‘Cyclic thermal shock resistance of several advanced ceramics and ceramic composites’, J. Am. Ceram. Soc., 81(7), 1888–1892. Schneider, G.A., Petzow, G. (editors) (1993), Thermal Shock and Thermal Fatigue Behavior of Advanced Ceramics, Dordrecht: Kluwer Academic. Sherman, D. (2001), ‘Alumina/NiCu laminate under thermal shock up to 1000°C: I, Experimental’, J. Am. Ceram. Soc., 84(12), 2819–2826. Soboyejo, W.O., Mercer, C., Schymanski, J., van der Laan, S.R. (2001), ‘Investigation of thermal shock in a high-temperature refractory ceramic: a fracture mechanics approach’, J. Am. Ceram. Soc., 84(6), 1309–1314. Swain, M.V. (1990), ‘R-curve behaviour and thermal shock resistance of ceramics’, J. Am. Ceram. Soc., 73, 621–628. Swain, M.V. (1991), ‘Quasi-brittle behaviour of ceramics and its relevance for thermalshock’, Eng. Fract. Mech., 40, 871–877. Tancret, F., Osterstock, F. (1997), ‘The Vickers indentation technique used to evaluate thermal shock resistance of brittle materials’, Scripta. Mater., 37(4), 443–447. Thompson, I., Rawlings, R.D. (1991), ‘Monitoring thermal shock of alumina and zirconiatoughened alumina by acoustic techniques’, J. Mater. Sci., 26, 4534–4540. Tiegs, T.N., Becher, P.F. (1987), ‘Thermal shock behaviour of an alumina-SiC whisker composite’, J. Am. Ceram. Soc., 70(5), C109-C111. Twitty, A., Russell-Floyd, R.S., Cooke, R.G., Harris, B. (1995), ‘Thermal shock resistance of Nextel/silica–zirconia ceramic–matrix composites manufactured by freeze-gelation’, J. Eur. Ceram. Soc., 15, 455–461. Uribe, R., Baudin, C. (2003), ‘Influence of a dispersion of aluminum titanate particles of controlled size on the thermal shock resistance of alumina’, J. Am. Ceram. Soc., 86(5), 846–850. Vandeperre, L.J., Kristofferson, A., Carlstrom, E., Clegg, W.J. (2001), ‘Thermal shock of layered ceramic structures with crack-deflecting interfaces’, J. Am. Ceram. Soc., 84(1), 104–110. Wang, H., Singh, R.N. (1994), ‘Thermal shock behaviour of ceramics and ceramic composites’, Int. Mat. Rev., 39(6), 228–244.
© Woodhead Publishing Limited, 2006
Thermal shock of ceramic matrix composites
433
Wang, H., Singh, R.N., Lowden, R.A. (1994), ‘Thermal shock behaviour of continuous fiber ceramic composites’, Ceram. Eng. Sci Proc., 15(4), 292–302. Wang, H., Singh, R.N., Lowden, R.A. (1996), ‘Thermal shock behaviour of two-dimensional woven fiber-reinforced ceramic composites’, J. Am. Ceram. Soc., 79(7), 1783–1792. Wang, H., Singh, R.N., Lowden, R.A. (1997), ‘Thermal shock behaviour of unidirectional, 0/90, and 2-D woven fibre-reinforced CVI SiC matrix composites’, J. Mater. Sci., 32, 3305–3313. Wang, L., Shi, J.-L., Gao, J.-H., Yan, D.-S. (2001), ‘Influence of tungsten carbide particles on resistance of alumina matrix ceramics to thermal shock’, J. Eur. Ceram. Soc., 21, 1213–1217. Wang, Y.R., Chou, T.-W. (1991), ‘Thermal shock resistance of laminated ceramic matrix composites’, J. Mater. Sci., 26, 2961–2966. Webb, J.E., Singh, R.N., Lowden, R.A. (1996), ‘Thermal shock damage in a two-dimensional woven-fiber-reinforced-CVI SiC-matrix composite’, J. Am. Ceram. Soc., 79(11), 2857– 2864. Wereszczak, A.A., Scheidt, R.A., Ferber, M.K., Breder, K. (1999), ‘Probabilistic thermal shock testing using infrared imaging’, J. Am. Ceram. Soc., 82(12), 3605–3608. Yin, X., Cheng, L., Zhang, L., Xu, Y. (2002), ‘Thermal shock behavior of 3-dimensional C/SiC composite’, Carbon, 40, 905–910. Zhao, Z., Johnson, M., Shen, Z. (2002), ‘Microstructure and mechanical properties of titanium carbonitride whisker reinforced β–sialon composites’, Materials Research Bulletin, 37, 1175–1187.
© Woodhead Publishing Limited, 2006
16 Superplastic ceramic composites A DOMÍNGUEZ-RODRÍGUEZ D G Ó M E Z - G A R C Í A, Universidad de Sevilla, Spain and F W A K A I, Tokyo Institute of Technology, Japan
16.1
Introduction
Superplasticity is macroscopically defined as the ability of a polycrystalline material to exhibit large elongations at elevated temperatures and relatively low stresses. It is commonly found in a wide range of materials from metals to ceramics (bioceramics or high-temperature superconductors, among others) when the grain size is small enough: a few micrometres for metals and less than a micron in ceramics. From a microscopic point of view, a superplastically deformed polycrystal is characterized by a microstructure, i.e., grain size and form factor, almost unchanged after deformation, and it is generally accepted that these features are mainly achieved through grain boundary sliding (GBS). The mechanisms accounting for the relaxation of the stresses created during GBS (what is commonly known as the accommodation process), especially in ceramics, are not yet clearly elucidated and remain controversial. However, it is accepted that atomic diffusion, either between the grain boundaries and the bulk, or along grain boundaries or during the non-conservative movement of dislocations, is among the accommodation processes controlling superplasticity. Superplasticity is a very promising property, not only because, like in metals, the superplastic formation opens a way for the manufacturing of complex ceramic pieces for industrial applications, but also because the combination of GBS and diffusional processes makes superplasticity an interesting tool for joining ceramic pieces in shorter times and lower temperatures than the diffusional joining technique. Several parameters can influence strongly the superplastic behaviour of ceramics, i.e. the strain rate at which the material can be superplastically deformed. Between them can be mentioned the grain size, second phases and segregation of impurities at the grain boundaries, etc. This chapter discusses the following leading topics: • Analysis of both the macroscopic and microscopic features of superplasticity 434 © Woodhead Publishing Limited, 2006
Superplastic ceramic composites
• • • •
435
The different accommodation processes controlling superplastic output A critical analysis of the parameters improving superplasticity Applications: forming and joining Future tendencies in the field, with a special emphasis on up-to-date information about the most outstanding and promising superplasticity in related ceramic materials.
16.2
Macro- and microscopic superplastic characteristics
Although superplasticity in metals has been extensively studied since the 1960s, it was only in the 1980s when superplasticity in ceramics (monolithic and composites) started to become a very active research field, and it has expanded rapidly since then. The reason for this development has been the need to improve the structural properties of these materials due to the demands of industry, asking for materials to be used in more severe working conditions. Ceramics can be a good candidate if their mechanical behaviour, especially at low temperatures, can be improved. Several techniques have been developed; probably the most widely used is based on the transformation toughness in zirconia, a keystone in ceramics investigation since Garvie et al.1 published their paper ‘Ceramic steel?’. Recently, new techniques have emerged through the processing and sintering of ceramic powders, leading to the attainment of fully dense ceramics with equiaxed grain sizes below 1 µm, primordial for superplastic behaviour. The first observation of superplasticity in a 3 mol% yttria-stabilized tetragonal zirconia polycrystal ceramic (YTZP) with a grain size of 0.4 µm was reported by Wakai et al.2 in 1986 (Fig. 16.1). Since then, a large number
16.1 First demonstration of superplastic deformation of 3Y-TZP: specimens before and after deformation at 1450°C and 3 × 10–4 s–1.
© Woodhead Publishing Limited, 2006
436
Ceramic matrix composites
of fine-grained polycrystalline ceramics and ceramic composites have been shown to have superplastic behaviour.3–8 Today, from an engineering point of view, the name superplasticity is ascribed to a polycrystalline material pulled out to very high tensile elongations prior to failure with necking-free strain. This phenomenon is usually found in many metals, alloys, intermetallics, composites and ceramics (recently in high-temperature superconductor ceramics) when the grain size is small enough, less than 10 µm for metals and less than 1 µm for ceramics. From a microscopic point of view, when a polycrystalline material is deformed at high temperatures, grain boundary sliding (GBS) takes place in two different ways: • The deformation is due to the flow of point defects, then GBS occurs to maintain grain coherency. This is called diffusional creep: Nabarro–Herring if the diffusion takes place along the bulk, or Coble if it takes place along grain boundaries. In these cases, each individual grain suffers almost the same deformation as that imposed on the specimen and the grains which are nearest neighbours remain nearest neighbours. This is termed ‘Lifshitz grain boundary sliding’ (Fig. 16.2(a)). • A different situation happens when GBS is responsible for the deformation. In order to release stresses created during GBS, deformation may be accompanied by intergranular slip throughout adjacent grains, by localized slip adjacent to the boundaries or by diffusional process of point defects. Sometimes, formation of triple-point folds or the opening up of cracks at the triple points can also accommodate GBS; however, as soon as there is coalescence of voids or cracks, the material fails, and that happens at low strains. This type of GBS, accommodated by the process described above, is termed ‘Rachinger grain boundary sliding’ (Fig. 16.2(b)). When GBS is accommodated by some of the mechanisms involving dislocation movement or diffusion of point defects, the grains retain almost the original size and shape even after large deformations. This GBS, as the primary mechanism for deformation, is the basis for the high ductility exhibited σ
σ
σ
σ (a)
(b)
16.2 (a) Schema of the Lifshitz GBS. The thick line shows the GBS after deformation. (b) Schema of the Rachinger GBS, showing that the grain shape remains constant after deformation.
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
437
by some materials at high temperatures and therefore for their structural superplastic behaviour. Although superplasticity is defined as the ability of a polycrystalline material to exhibit large elongations, in many ceramics-related materials and ceramic composites superplasticity is also said to occur even though the polycrystal is deformed in compression, or in three- or four-point bending conditions, as long as GBS is the primary deformation process.4–7 The commonly used equation for the steady-state strain rate ε˙ characterizing superplastic behaviour is written as:
Gb b σ D kT d G p
ε˙ = A
n
(16.1)
where A is a dimensionless constant, G is the shear modulus, b is the magnitude of a Burgers vector, k is the Boltzmann constant, T is the absolute temperature, d is the grain size, σ is the stress, D = D0 exp(–Q/RT) is the appropriate diffusion coefficient involved in the accommodation process (D0 is the frequency factor, Q the activation energy and R the gas constant), and p and n are the grain size and stress exponents respectively. The values of the creep parameters (p, n and Q) identifying the superplastic behaviour of ceramic-related materials are not unique to such materials, nor to the same type of materials. As shown in the review papers, these parameters are very similar in tension as in compression in zirconia-based materials (probably the most widely studied ceramics in the widest experimental conditions), although that depends strongly on the purity of the ceramics;5,7 however, their behaviour seems to be very different in compression than in tension when an aid-sintering phase is necessary during the processing, as in silicon carbide and silicon nitride ceramics.8 For the sake of clarity, we will mention the reasons for the discrepancies in the creep parameters in materials for which most studies have been performed. In the case of YTZP, values of p between 1 and 3, of n between 2 and higher than 5, and of Q between 450 and 700 kJ/mol, have been reported during creep.7 Three different explanations have been developed to account for the discrepancies in the experimental creep data: • These differences have been interpreted on the basis of two sequential mechanisms: at high stresses, deformation occurs by GBS which is the slower process and therefore controls plasticity, whereas at low stresses the deformation is controlled by an interface-reaction process. In both cases the activation energy is the same.9,10 However, this explanation fails for several reasons: the experimental values found for the activation energy are not constant (it is high at low stresses and becomes equal to 450 kJ/ mol at high stresses) and the n values decrease gradually to 2 when the stress increases.2,9
© Woodhead Publishing Limited, 2006
438
Ceramic matrix composites
• Another explanation was proposed by Berbon and Langdon11 using a modified Coble mechanism developed by Arzt et al.12 In this model, they assume the existence of perfect grain-boundary dislocations evenly spaced in the boundary planes so that they can all climb at the same speed. This model predicts that n values cannot be higher than 3 and the activation energy is constant and equal to the energy for grain boundary diffusion of Zr. However, both predictions are contrary to the experimental observations. Finally, during Coble creep, grains of polycrystals change shape and size to reflect the overall strain within the sample, again contrary to the microstructural observations of these deformed materials, in which grains remain almost unchanged. • Probably the most plausible explanation for the scatter of the creep parameters is based on a single mechanism involving GBS with a threshold stress (σ0).7,10,13 When a threshold stress is introduced into the creep equation (16.1), all the creep parameters in YTZP become n = 2, p = 2 and Q = 460 kJ/mol whatever the stress or temperature of the test. The value of this σ0 was found experimentally:13
σ 0 = 5 × 10 –4
exp
120 kJ/ mol RT d
(16.2)
with d in µm. Recently, a model has been developed on the basis of the segregation of the yttrium atoms at the grain boundaries to account for σ0. These are responsible for an electric field to appear, which influences the grain displacement of each other (i.e. GBS). The model is able to explain quantitatively the dependence of σ0 on both temperature and grain size.14 Another explanation for the threshold stress has been pointed out, which takes into consideration the stress required for intragranular dislocations to be activated, estimated to be 3.6–10.7 MPa;15 however, the role these intragranular dislocations play is not yet clear, because the flow stress required to activate dislocations in yttria-tetragonal zirconia single crystals at temperatures as high as 1400ºC is over 400 MPa, much higher than the stress used in superplasticity at these conditions.16 For second-phase sintered ceramics, these phases control the plasticity and they are responsible for the asymmetric behaviour when deformed in tension or compression, because there is a crucial difference in the microstructure evolution associated with tension and compression creep. There are few explanations for this asymmetry. It is a well-established fact that under tension, the formation and nucleation of cavities contributes much more significantly to the macroscopic strain than under compression.17 Experimental results have shown that the difference
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
439
between tensile and compressive creep can be approximately explained by consideration of the volume fraction of cavities.18 Another explanation of the creep asymmetry is based on the rate of approach (during compression) or separation (during tension) of adjacent grain facets controlled by the viscous secondary phase and the fact that the number of grain facets supporting compressive stresses is less than those supporting tensile stresses.19 When the material is composed of a soft phase with rigid inclusions, the strain rate is written in the form:
ε˙ = A1 (1 – Vf ) q σ n
(16.3)
where A1 is a stress-independent constant and Vf the volume fraction of inclusions. The exponent q depends exclusively on the stress exponent, the aspect ratio of the rigid inclusions and the orientation with respect to the stress axis. This term has been proposed as the origin of the asymmetric behaviour between tension and compression.20 None of these accommodation processes is diffusion-controlled, since a diffusional process cannot account for the asymmetric behaviour between tension and compression, due to its inherent symmetry in the sign of the stress.
16.3
Accommodation processes controlling superplasticity
As mentioned above, several mechanisms can be responsible for the grain boundary sliding accommodation; however, so far there is no consensus on a general single mechanism to accommodate GBS, nor one concerning a particular ceramic. In this section the different mechanisms for accommodation will be analysed. For the sake of clarity, the accommodation process will be described for each type of ceramic, whether monolithic, with secondary glassy phases or composite. For monolithic ceramics and ceramic composites, during superplastic flow, the relative motion of two adjoining grains has components parallel and perpendicular to their common grain boundary. GBS is the component parallel to the grain boundary and is responsible for 70–80% of the deformation in superplasticity of fine-grained polycrystals,21–23 as has been shown by measurements of the grain aspect ratio, by both SEM and atomic force microscopy as displayed recently by Duclos24 in YTZP. During GBS, rigid grains inevitably generate cavities and cracks, and in order to avoid fracture and allow the material to deform superplastically, accommodation processes involving non-conservative motion of the grains (movement perpendicular to the grain boundary) are necessary. This non-conservative motion involves either diffusion of point defects, or dislocation glide and climb and grain boundary migration when grain growth occurs.
© Woodhead Publishing Limited, 2006
440
Ceramic matrix composites
16.3.1 GBS accommodated by diffusional flow Ashby and Verrall (thereafter A-V) developed a model for GBS accommodated by a diffusional process which integrates quantitatively the essential topological features found in superplasticity.25 The principle of this model for grain rearrangement is shown in Fig. 16.2(b). This type of grain rearrangement retains the equiaxed grains after large deformation as shown in the typical microstructural features of the polycrystals superplastically deformed.7,24 A modified A-V model accounting for a more realistic symmetrical diffusion path was developed by Spingarn and Nix when diffusional flow occurs only along grain boundaries.26 Several modifications of the original A-V model have been made by various authors.27 However, the main features of the original A-V model remain unchanged. The constitutive equation of the A-V model, when lattice and grain boundary diffusion are taken into account, is written:
ε˙ =
3.3δ Dgb 0.72γ 100 Ω σ– DL 1 + 2 d dDL kTd
(16.4)
where Ω is the atomic volume of the diffusion controlling species, 72γ /d is a threshold stress for the grain-switching event, γ is the grain boundary free energy, δ is the thickness of the boundary, and DL and Dgb are the lattice and grain boundary diffusion coefficients, respectively. As can be observed in Eq. (16.4), the strain rate is a linear function of the stress (n = 1) in disagreement with experimental observations for many monolithic ceramics5,7,28 for which n is 2. For YTZP, the grain size dependence fits to 2 and the activation energy corresponds to that of cationic lattice diffusion. In the case of YTZP, on which a large number of studies have been performed, the data could be fitted to a constitutive equation, which is identical to that found in metals when lattice diffusion is the rate-controlling mechanism:29 460kJ mol –1 (σ – σ 0 ) 2 ε˙ = 3 × 10 10 exp – 2 RT Td 2
= 2 × 10 7
2
Gb σ – σ 0 b D Zr kT G d L
(16.5)
16.3.2 GBS accommodated by dislocation movement When GBS is accommodated by the movement of dislocations, the strain rate can be deduced in a similar way as for recovery creep: the dislocations generated at the grain boundaries move either along the grain boundaries or
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
441
across the grain until they are piled up at an obstacle, which can be overcome when the stress at the head of the pile-up is high enough to promote the climb of the head dislocation. From this picture the strain-rate-controlling GBS can be written:30
ε˙ ≈
2 DLσ 2 b 3 3 3 h 2 GkTd
(16.6)
with D an appropriate diffusion coefficient, L the length of the pile-up and h the climb distance. Depending on the grain size, two limit cases have been analysed.30 At low grain size, it happens that L ~ d, h ~ 0.3d and Eq. (16.6) becomes:
ε˙ ≈
A2 Dgb Gb b 2 σ 2 kT d G
(16.7)
where A2 is a dimensionless constant with a value of 10. Conversely, when the grain size is large, L ≈ 20Gb/σ, h ≈ bG/16σ (for more details see reference 30), and by substituting these parameters, Eq. (16.6) becomes:
A D Gb b σ 3 ε˙ ≈ 2 L kT d G
(16.8)
where A2 is a dimensionless constant having a value of the order of 103. Recently, the activity of dislocations has been observed during the superplasticity of YTZP.15 The authors suggest that the rate of deformation is controlled by the recovery of the intragranular dislocations in the highstress region where n is 2.7 and p is between 2 and 3. As can be inferred from the equations outlined above, none of the different models can adjust the creep parameters for all the different ceramics, especially in the case of YTZP,7 explaining why there is still controversy over the accommodation process controlling superplasticity. The same conclusions can be outlined for ceramic composites, although more experimental work should be done.20,31 For ceramics with secondary glassy phases, the accommodation processes are governed by these phases. Although diffusion may occur, the glassy phase viscosity controls accommodation mainly in different ways: • These glassy phases may act as a lubricant for grain boundary sliding. In this case, the accommodation mechanism is the viscous motion of these secondary phases. • The secondary glassy phases can improve the diffusivity pathways throughout the grain boundaries. Accommodation is controlled by diffusion along them.
© Woodhead Publishing Limited, 2006
442
Ceramic matrix composites
• Finally, the secondary phase can provide a preferential location for the nucleation and growth of cavities during deformation, producing fracture of the material and reducing its superplasticity ability. At high temperatures the glassy phase may become less viscous and even liquid and as a consequence may account for the plastic deformation. However, viscous flow creep is not regarded as a viable creep mechanism for superplasticity due to its limited deformation, which corresponds to the redistribution of the glassy phase and therefore to the squeeze of these secondary phases from grain boundaries subjected to compression.8 In the next section, we will analyse the mechanisms that are considered to control superplasticity in ceramics with secondary phases.
16.3.3 Solution–precipitation creep In this case, the secondary phases melt at temperatures lower than the matrix and, provided that the crystals are at least partially soluble in the glassy phase, creep may take place by: • solution of the crystal in the liquid phase at grain boundaries under compression, • diffusion along the liquid phase, and • precipitation of the crystalline material at grain boundaries under traction. The solution–precipitation creep model was first proposed by Raj and Chyung;32 they assumed two cases: • The strain rate is controlled by the diffusion along the glassy phase. Thus:
ε˙ = 1 13 σ ηd
(16.9)
with η the viscosity of the secondary phase. • The strain rate is controlled by the reactions of solution and precipitation at the interfaces. If this is the case:
ε˙ = 1 k d σ d
(16.10)
kd being a reaction constant. This model postulates that the glassy phase in compression is able to support normal stresses because of the existence of islands, thus avoiding the complete squeeze of this intergranular liquid (Fig. 16.3). However, it has been shown that these islands are not necessary for the grain boundaries to support normal stresses. Several modifications and revisions of this first model have been made.8
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
443
Upper grain Island
Glassy phase
Glassy phase h
Lower grain
16.3 Schema of the solution–precipitation model with the island structure supporting normal stresses.
An important modification of this model was performed by Wakai.33 The main assumptions are that the solution and precipitation reactions take place at line defects as ‘kinks’ in steps formed at the grain boundaries (Fig. 16.4), and the spacing between kinks is small enough for the step to be considered as an ideal source or sink of solute particles. Thus, the solution and precipitation of crystalline materials at these steps produces their movement, and consequently strain and strain rate will have an expression analogous to Orowan’s equation for dislocation movement:
ε˙ =
ρS av S d
(16.11)
with ρS the density of surface steps per unit length, a the height of the steps and vS the velocity of the steps. This velocity depends on the process of integration into the crystal at a kink, the diffusion in the absorption layer and
Grain (i) Solute transport along the glassy phase Glassy phase
(iv) Integration at a kink
‘Kink’ (iii) Diffusion along this layer
(ii) Deposition at the adsorption layer
Grain
16.4 Schema of the solution–precipitation in Wakai’s step model.
© Woodhead Publishing Limited, 2006
444
Ceramic matrix composites
the diffusion in the liquid film. The rate-controlling process will be the more resistant to the step movement. Three different situations were analysed by Wakai for the density of steps: • The constant density of surface steps is independent of the applied stress. • If the initial density of steps is very low, two-dimensional nucleation of the surface steps occurs. • If the continuous source of steps is a screw dislocation, then a spiral step is generated. The combinations between the rate-controlling process and the density of steps give the different equations of Wakai’s model, which are summarized in Table 5 of reference 8. The case of constant density of steps modelled by Wakai is equivalent to the diffusion-controlled creep modelled by Raj and Chyung32 and is also consistent in terms of the stress, temperature and grain size dependence of the strain rate for interface-reaction-controlled creep predicted by Raj and Chyung.32 However, in the two cases of bidimensional nucleation of step and spiral step, the creep parameters differ from those predicted by wakai.33 In particular, for two-dimensional nucleation, there is a divergence of the creep parameters which has been recently solved34 by considering in detail the precipitation or solution of the crystalline material at the step, which changes significantly the free enthalpy involved in the process. The essential key of this new modification is the free energy change per unit volume for the precipitation mechanism, making a strong correction to the apparent parameters n and Q measured in mechanical tests.
16.3.4 Shear thickening creep This phenomenon has been postulated as an explanation for the compressive superplastic deformation of a SiAlON which undergoes a transition from n = 1 (Newtonian behaviour) to n = 0.5 for a characteristic critical stress (σc) independent of both temperature and composition of the secondary phase.35 The model is based on the idea that the glassy phase is composed of two layers – a normal glassy phase layer behaving in a Newtonian way, embedded into an over-condensed layer with non-Newtonian behaviour. Thus, for stresses lower than the critical stress, the creep is controlled by the normal glassy phase (n = 1), and when the stress exceeds a critical value, the squeeze of this phase makes the two over-condensed layers come into contact, thus the material creeps in a non-Newtonian way (n = 0.5). The creep rate is written:
ε˙ = (1 – Vf ) 2.5 σ η′
(16.12)
with η′ an apparent viscosity depending on temperature, grain size, phase
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
445
composition and liquid content, and Vf the volumetric fraction of the overcondensed rigid phase, which for σ ≥ 23 σ c yields a value: Vf = 1 –
1 + σc 4 2σ
(16.13)
explaining the transition from n = 1 to n = 0.5. Again a review of the model can be found in the literature.8
16.4
Parameters improving superplasticity
The strategy to enhance superplasticity is twofold: refinement of the microstructure, i.e. decrease of the grain size, or improvement of the accommodation process needed to relax the stresses created during GBS throughout the appropriate diffusion coefficient. Although both means are independent of each other, in a few cases the reduction of the grain size may induce a decrease in the diffusion coefficient involved in the accommodation process, giving rise to a compensating effect. These different strategies will be analysed in the following sections.
16.4.1 Refinement of the microstructure As observed in the different equations accounting for superplasticity in ceramicbased materials, the strain rate is an inverse function of the grain size; in consequence, the grain size should be stable and as small as possible to attain high-strain-rate superplasticity. If grain growth occurs during deformation, the level of stresses for successive deformation will increase, inducing the formation of intergranular cavities leading to failure. Several techniques have been developed to achieve ceramics and composites with fine microstructure: • One of the techniques to suppress grain growth is based upon the inclusion of dispersed phases into the ceramics. With this regard, a multi-phase ceramic composite containing 40 vol% ZrO2, 30 vol% spinel and 30 vol% Al2O3 has been sintered,36 which was superplastically deformed to 1050% at 1650ºC at a strain rate of 0.4 s–1. Other zirconia-based composites have also been fabricated with inhibition of grain growth.31,37 With this technique it is possible to deform the ceramics, with their microstructure unchanged, at a temperature at which grain growth would be important in ceramics without dispersed phases. • Another technique makes use of ultra-fine powders sintered under stressassisted conditions so the sintering temperature is reduced,38 for instance hot-isostatic pressing, hot pressing, sinter forging, or techniques with fast
© Woodhead Publishing Limited, 2006
446
Ceramic matrix composites
heating and cooling ramp, such as spark plasma sintering (SPS) or microwave sintering, which avoid grain growth by reducing the time or temperature of sintering. Very fast densification has been reported in oxides and Si3N4-based ceramics by SPS.39 Microwaves have also been used to sinter PSZ40 and alumina.41 • The sintering of Y2O3 nanocrystalline ceramics (d = 60 µm) has been achieved through a two-step sintering method. The first step is pre-sintering at high temperatures to obtain ceramics with intermediate density values between 70 and 80%. Secondly, suppression of grain growth is achieved by sintering at lower temperatures than those used during the first step, exploiting the difference in kinetics between grain-boundary diffusion, which controls sintering, and grain-boundary migration, which controls grain growth (second step).42
16.4.2 Improvement of the processes controlling diffusion in superplasticity The diffusion coefficients of the process controlling superplasticity may be enhanced or retarded by the addition of impurities or solute atoms or by the addition of secondary phases, normally used as sintering aids, which distribute along the grain boundaries and triple-point junctions of the grains. There are a great number of papers dealing with the influence of the grain boundary segregation on superplasticity in YTZP. It has been shown that the superplastic flow stress at 1400ºC of a 3YTZP doped with different cations is correlated with the ionic radius of the dopant.43,44 Cations with smaller ionic sizes decrease the flow stress, whereas those with large ionic sizes increase the flow stress. The authors suggest that the flow stress is determined by the grain boundary diffusivity, which is affected by the segregation of the dopant. The same improvement of superplasticity has been respectively found45,46 in a 0.3 mol% SiO2 doped 3YTZP with d = 0.35 µm and a 0.18 mol% Al2O3 doped 3YTZP with d = 0.4 µm. This behaviour also seems to account for fine-grained Al2O3 with ZrO2 as a dopant.47,48 However, this explanation conflicts with other results in Al2O3 doped with different impurities, in which the mechanical behaviour is explained in terms of the change of the ionic bonding strength between Al and O and the covalent bonding between Al and the surrounding cations, thus affecting the grain boundary diffusivity.49–51 This explanation has been outlined in SiO2– TZP doped with several kinds of metal oxides.51–53 This change in the bond strength has also been used to explain the superplasticity of SiC doped with small amounts of boron.54 The doped boron segregates at grain boundaries, removing silicon from its site, thence forming bonds in a local environment, similar to that in the B4C structure.55 This fact enhances the deformation by grain boundary diffusion.
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
447
An improvement of the diffusivity and in consequence of the flow stress and elongation to failure is also found when a secondary glassy phase is added. This behaviour has been reported in barium and borosilicate doped 3Y-TZP.7,56 As mentioned above, non-oxide ceramics such as SiC and Si3N4 are fabricated with sintering aids, thence generating a two-phase material, with a hard phase surrounded by a soft secondary glassy phase. This secondary phase is the one controlling the mechanical behaviour of these materials, as mentioned when referring to the accommodation processes controlling superplasticity. In the case of Si3N4, glass pockets and thin glass film with thickness of about 1 µm often remain at grain boundaries.57 Its plasticity is controlled by the viscosity of the intergranular glassy phase and the solubility of the crystalline phase in the liquid and have been reviewed in several contributions.8,58,59 The solid solution of silicon nitride with some aluminium-based compounds or mixture form the so called α′ or β′-SiAlON compounds which are superplastic by the addition of secondary glassy phase; for example, Rosenflanz and Chen60 reported that Li-doped SiAlON deforms 10 times faster than Si3N4. A revision of the mechanical properties of these compounds can be found in the literature.8,58 The enhancement of superplastic deformation by intergranular glass phase was also applicable to liquid-phase sintered SiC.61,62 As mentioned in the last two paragraphs, to improve superplasticity it is necessary to reduce the grain size or to enhance the diffusion process controlling it; however, as was recently shown in YTZP, the reduction of grain size may produce a reduction of the diffusivity of the species controlling superplasticity. It has been successively shown that the yttrium in YTZP polycrystals segregates at grain boundaries and this segregation was the possible cause of the threshold stress (σ0) and could explain quantitatively the dependence of this σ0 with temperature and grain size.14 The segregation of yttrium atoms whose electric charge is different from that of the parent ions induces a local density of negative charge produced by the YZr′ defects accounting for a local electric field which is screened by the gradient of oxygen vacancies between the bulk and the boundaries. When the grain size of the polycrystal becomes close to the screening length (nanoscale length), the electric field can influence the diffusional processes and the creep equation (16.1) will be multiplied by the following factor:63
α=
1 – z eV ( R ) λ 1 + 4 λ exp D – 1 d 3ε r kT d
(16.14)
with λ the Debye attenuation length, zD the valence of yttrium, V(R) the electrical potential and εr the relative dielectric constant. A plot of α versus the average grain size from (16.14) is shown in Fig. 16.5. From this it can be seen that the effect of the nanostructured character
© Woodhead Publishing Limited, 2006
448
Ceramic matrix composites
α
1
10–1
λ λ λ λ 10–2
0
40
80 120 Grain size (nm)
= = = =
5 nm 10 nm 15 nm 20 nm
160
200
16.5 Plot of α versus grain size for different λ values.
of the YTZP specimen is more and more pronounced the bigger λ is. The creep resistance increases up to a factor of 10 for a grain size around 50 nm, when λ is equal to 20 nm. Whereas an ample number of publications have dealt with submicronsized YTZP, the number of publications decreases when going down to nanometre size, due mainly to the fact that only recently have fully-densified nanocrystalline YTZP become available. Recently, improved creep resistance has been reported in 50 nm YTZP deformed at 1200ºC in agreement with prediction with the model.64,65 Few papers have been published 66–68 on nanocrystalline monoclinic ZrO2; however, as mentioned, in monoclinic materials there is no segregation at the grain boundaries and the predictions of the model developed when segregation occurs cannot be tested. At this point, it is necessary that more systematic work on well-defined systems be conducted in order to verify the importance of an electric field in the diffusion process when the grain size decreases down to the Debye length scale.
16.5
Applications of superplasticity
The increasing applications of advanced ceramics in technical areas including aerospace, energy, electronics, biology, etc., often require complex shapes to be manufactured at low prices. The extensive potential applications, together with the possibility of processing dense ceramics and forming complex structures by superplasticity, have been the driving force for the appearance and fast development of a large number of ceramic systems with superplastic capabilities. Several industrial processes in the metal and polymer industries
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
449
have already made use of these high-ductile ceramics. The processing of dense ceramics includes sheet forming, blowing, stamping, forging and joining. A good example of superplastic forming of different ceramics can be found in Figure 1 of reference 3. In this figure, flat 1 mm thick discs of Si3N4, 2YTZP:Al2O3, 2Y-TZP:Mullite and 2Y-TZP + 0.3% doped with Mn, Fe, Co, Cu and Zn, were stretched with a 6.5 mm radius punch at temperatures and forming times depending on the ceramics. Sinter forging is a promising technique because densification and netshaping are achieved simultaneously.69 This technique avoids cavities and voids because sinter forging is produced by compression. High strength and high fracture toughness of Si3N4 have been achieved by superplastic sinter forging due to the reduction of flaw size and grain alignment.70 As previously mentioned, superplasticity is due to GBS accommodated by diffusional processes; a novel technique to join ceramics, which takes advantage of both processes, has been developed. When two ceramics in contact are deformed in the superplastic regime (i.e. as soon as GBS is activated), the grains of one part interpenetrate those of the other, producing a rapid and perfect junction of both parts in shorter times and at lower temperatures than those commonly required in other conventional processes for ceramics joining.71 An example of this and further proof that GBS is the mechanism of superplasticity is displayed in Fig. 16.6. This figure is composed of two parts: Fig. 16.6(a) displays a set of two pieces of 3Y-TZP joined at 1400ºC for 15 min; Fig. 16.6(b) displays an analogous set of two 3Y-TZPs with different grain sizes. In the Fig. 16.6(a), the arrows show the interface of the two pieces and it is easy to see how the grains of both parts interpenetrate into each other. On the other hand, superplasticity is grain size dependent (Eq. (16.1)): materials with coarse grains will be more creep resistant than those with smaller grains. Based on this behaviour, it was possible to join zirconia layers of different grain sizes to obtain a multilayer composite with clean and strong interface and heterogeneous mechanical behaviour at room and high temperatures depending on the stress application.72 The same behaviour may be obtained when two layers of different compositions are joined, because in yttria-stabilized zirconia (YSZ), grain size is a function of the yttria content.73 This fact is displayed in Fig. 16.7. Figure 16.7(a) is a micrograph showing two different layers joined together; Fig. 16.7(b) shows a room-temperature Vickers indentation along the interface, where the crack does not propagate along the interface but inside the brittle material; and Fig. 16.7(c) is a plot of the high-temperature plastic deformation at 1400ºC with the compression axis parallel and perpendicular to the interface. Using this technique, a multilayer made of four different layers 0.5 mm thick of 3YTZP with grain sizes of 0.3, 0.5, 0.8 and 1.0 µm, obtained by annealing the as-sintered ceramics (0.3 µm), labelled from ‘a’ to ‘d’, have
© Woodhead Publishing Limited, 2006
450
Ceramic matrix composites
2 µm
(a)
2 µm
(b)
16.6 SEM micrographs of junctions of (a) two layers with the same grain size (0.3 µm), and (b) two layers with different grain sizes (0.3 and 1 µm).
been joined together in a sequence ‘abcdabcd’ (Fig. 16.8(a)) at 1400ºC in order to form a compound with anisotropic mechanical behaviour depending on the compression axis parallel or perpendicular to the interface.74 When the compression axis is perpendicular to the interface, all the layers are submitted to the same stress (thereby an isostress test), and the layer with the smaller grain size controls superplasticity. When the compression axis is parallel to the interface, all the layers suffer the same strain (thereby an isostrain test) and the layer with the biggest grain size controls the plasticity (Fig. 16.8(b)). Using the creep model for composites with duplex microstructure developed by French et al.,75 it is possible to fabricate a composite with a
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
451
3Y 6Y
2 µm
3Y
100 µm
6Y (b)
(a) 40
Stress (MPa)
30 Isostrain
3Y–6Y
20 3Y–6Y Isostress 10
T = 1400°C 0
0
2
4 6 Strain (%) (c)
8
10
16.7 (a) SEM micrograph of two layers of different composition, 3 mol% YTZP (3Y) and 6 mol% YTZP (6Y) (b). Optical micrograph of the cross-section of the 3Y–6Y junction with a Vickers indentation and the crack pattern. (c) Stress–strain curve of the 3Y–6Y junction deformed at 1400°C with the stress parallel (isostrain) and perpendicular (isostress) to the interface.
defined creep resistance by controlling the grain size, the width of the layers and the compression axis; in conclusion, with this technique, it is possible to obtain a defined functionally graded material (FGM). The use of nano-ceramics as interlayers can reduce drastically the temperature of joining. A good example can be observed in layers of YTZP, which were joined with an interlayer of 20 nm of YTZP at 1150ºC, 200ºC lower than the temperature used without the nano-layer.76
© Woodhead Publishing Limited, 2006
452
Ceramic matrix composites Isostress
a b c d a
Isostrain
b c d
(a)
Engineering stress (MPa)
30 Isostrain 20 Isostress
10
0 0
10 20 Engineering strain (%)
30
(b)
16.8 (a) Schema of the composite obtained with four different layers labelled a to d. Stress-strain curve of the composite deformed at 1400°C in the isostrain and isostress regime.
16.6
Future trends
In order to outline the future trends in superplasticity in ceramics, first of all it is necessary to give an answer to the following question: why is superplasticity in ceramics so important? The potential use of these materials in more and more severe applications makes superplasticity in ceramics an important tool for their processing, as happened with metals at the beginning of the 1960s.
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
453
16.6.1 What properties should be controlled in superplasticity? The nature of the grain boundaries when segregation or glassy phases exist It is well known that the impurities can segregate or precipitate at the grain boundaries; however, it is not known what role they play in superplasticity. For instance, what is the importance of the ionic radius of the impurities, the charge effect or the binding energy between the host atoms and the impurities in the superplastic behaviour? As indicated in the text, the charge effect has been used to justify the threshold stress in YTZP and the binding energy has been used to justify the different behaviour of monolithic ceramics like YTZP or alumina when doped with different impurities, although the ionic radius has also been used for the same explanation by the same authors. Type of defects, if any, created during GBS For instance, dislocations have been shown to play a key role in the accommodation process in YTZP, justifying the threshold stress in YTZP, in contrast with the hypothesis that this threshold stress is due to the electric field created by impurity segregation. However, dislocations are not systematically observed in YTZP; furthermore it was shown that in yttriastabilized tetragonal zirconia single crystals, the stress necessary to activate dislocations at 1400ºC was over 400 MPa, one order of magnitude higher than the stresses used during superplastic deformation of YTZP at the same temperature. It will be necessary to conduct a systematic study of the microstructure of the monolithic ceramics such as YTZP before and after deformation and to correlate their relationship with the superplastic features. Grain boundary sliding, electron density and binding energy It is necessary to advance in the knowledge of the superplastic equations, at least for pure monolithic ceramics and composites with glassy phases, to predict the behaviour for a given material. At this point, probably firstprinciples simulations of grain boundary sliding and first-principles calculation of electron density and the binding energy between the guest and host atoms in grain boundaries will be of great help in this task.77,78 Interatomic potentials using the Embedded Atom Method in conjunction with molecular static and dynamics calculations have been used to study the sliding and migration of (110) symmetric tilt grain boundaries in aluminium;79 although in ceramics the grain boundaries are more complicated, it could be interesting to attempt the same task in order to get better insights about GBS.
© Woodhead Publishing Limited, 2006
454
Ceramic matrix composites
16.6.2 Future trends in superplastic ceramics In the immediate future, the main objective in ceramic superplasticity will be the search of the right conditions to achieve ‘high strain rate superplasticity’ (HSRS) ( ( ε˙ ≥ 10 –2 s –1 ). Although this phenomenon has been found in several ceramic compounds and several inputs have been outlined to achieve it, we are still far from knowing what to do to obtain this effect systematically. This HSRS will enlarge the applications for ceramics. On the other hand, improvements in ceramic powder processing technology, the routine preparation of high-purity ceramics of nanometre scale, and the new techniques for the processing of these powders such as HIP, SPS, microwave furnace, etc., will be the driving forces for a very active study on nanoceramics in the near future, probably opening up new phenomena and new applications. Finally, the forthcoming comprehension of superplasticity will demand a well-settled justification of the basic equations for this phenomenon. In order to achieve this goal, first-principles calculations should be conducted. This is a very challenging task, because the mechanical behaviour of grain boundaries requires an understanding of the physics involved at many different scales. At this point, simulations at microscopic as well as mesoscopic levels can become a useful tool.
16.7
Acknowledgements
The financial support awarded by the Spanish Ministerio de Educación y Ciencia through the Project CICYT no. MAT2003-04199-C02-02 is acknowledged.
16.8
References
1. Garvie, R.C., Hannink, R.C., and Pascoe, R.T., ‘Ceramic steel?’, Nature, 1975, 258, 703–4. 2. Wakai, F., Sakaguchi, S., and Matsuno, Y., ‘Superplasticity of yttria-stabilized tetragonal ZrO2 polycrystals’, Adv. Ceram. Mater., 1986, 1, 259–63. 3. Wei Chen, I., and Xue, L.A., ‘Development of superplastic structural ceramics’, J. Am. Ceram. Soc, 1990, 73, 2585–609. 4. Nieh, T.G., Wadsworth, J., and Wakai, F., ‘Recent advances in superplastic ceramics and ceramic composites’, Int. Mater. Rev., 1991, 36, 146–61. 5. Chokshi., A.H., ‘Superplasticity in fine-grained ceramic and ceramic composites: current understanding and future prospects’, Mater. Sci. Eng., 1993, A166, 119–33. 6. Nieh, T.G., Wadsworth, J., and Sherby, O.D., in Superplasticity in Metals and Ceramics, Cambridge University Press, 1997. 7. Jimenez-Melendo, M., Dominguez-Rodriguez, A., and Bravo-Leon, A., ‘Superplastic flow in fine-grained yttria-stabilized zirconium polycrystals: constitutive equation and deformation mechanisms’, J. Am. Ceram. Soc., 1998, 81, 2761–76.
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
455
8. Melendez-Martinez, J.J., and Domínguez-Rodríguez, A., ‘Creep of silicon nitride’, Progress Mat. Sci., 2004, 49, 19–107. 9. Owen, D.M., and Chokshi, A.H., ‘The high temperature mechanical characteristics of superplastic 3 mol% yttria stabilized zirconia’, Acta Mater, 1998, 46, 667–79. 10. Jimenez-Melendo, M., and Dominguez-Rodriguez, A., ‘High temperature mechanical characteristics of superplastic yttria stabilized zirconia. An examination of the flow process’, Acta Mater, 2000, 48, 3201–10. 11. Berbon, M.Z., and Langdon, T.G., ‘An examination of the flow process in superplastic yttria-stabilized tetragonal zirconia’, Acta Mater, 1999, 47, 2485–95. 12. Arzt, E., Ashby, M.F., and Verrall, R.A., ‘Interface-controlled diffusional creep’, Acta Metal., 1983, 31, 1977–89. 13. Dominguez-Rodriguez, A., Bravo-Leon, A., Ye, J.D., and Jiménez-Melendo, M., ‘Grain size and temperature dependence of the threshold stress for superplastic deformation in yttria-stabilized zirconia polycrystals’, Mater. Sci. Eng, 1998, A247, 97–101. 14. Gomez-Garcia, D., Lorenzo-Martin, C., Muñoz, A., and Domínguez-Rodriguez, A., ‘Correlation between yttrium segregation at the grain boundaries and the threshold stress for plasticity in yttria-stabilized tetragonal zirconia polycrystals’, Phil. Mag, 2003, 83, 93–108. 15. Morita, K., and Hiraga, K., ‘Critical assessment of high-temperature deformation and deformed microstructure in high-purity tetragonal zirconia containing 3 mol% yttria’, Acta. Mater., 2002, 50, 1075–85. 16. Muñoz, A., Gomez-Garcia, D., Domínguez-Rodriguez, A., and Wakai, F., ‘High temperature plastic anisotropy of YTZ-single crystals’, J. Eur. Ceram. Soc, 2002, 22, 2609–13. 17. Wiederhorn, S.M., Roberts, D.E., and Chuang, T.J., ‘Damage-enhanced creep in a siliconized silicon-carbide-phenomenology’, J. Am. Ceram. Soc, 1988, 71, 602–8. 18. Lofaj, F., Okada, A., and Kawamoto, H., ‘Cavitation strain contribution to tensile creep in vitreous bonded ceramics’, J. Am. Ceram. Soc, 1997, 80, 1619–23. 19. Dryden, J.R. and Wilkinson, D.S., ‘Three-dimensional analysis of the creep due to a viscous grain boundary phase’, Acta Mater, 1997, 45, 1259–73. 20. Morita, K., Hiraga, K., and Sakka, Y., ‘High-strain rate superplasticity in Y2O3stabilized tetragonal ZrO2 dispersed with 30 vol% MgAl2O4 spinel’, J. Am. Ceram. Soc, 2002, 85, 1900–2. 21. Ishihara, S., et al., ‘Stereographic analysis of grain boundary sliding in superplastic deformation of alumina–zirconia two phase ceramics’, Mater. Trans. JIM, 1999, 40, 1158–65. 22. Duclos, R., Crampon, J., and Carry, C., ‘Grain-boundary sliding and accommodation mechanism during creep of yttria-stabilized zirconia’, Phil. Mag. Lett., 2002, 82, 529–33. 23. Morita, K., and Hiraga, K., ‘Reply to comment on the role of intragranular dislocations in superplastic yttria-stabilized zirconia’, Scripta Mater, 2003, 48, 1403–7. 24. Duclos, R., ‘Direct observation of grain rearrangement during superplastic creep of a fine-grained zirconia’, J. Eur. Ceram. Soc., 2004, 24, 3103–10. 25. Ashby, M.F., and Verrall, R.A., ‘Diffusion-accommodated flow and superplasticity’, Acta Metal.; 1973, 21, 149–63. 26. Spingarn, J.R., and Nix, W.D., ‘Diffusional creep and diffusionally accommodated grain rearrangement’, Acta Metal., 1978, 26, 1389–98.
© Woodhead Publishing Limited, 2006
456
Ceramic matrix composites
27. Langdon, T.G., ‘The significance of grain boundary sliding in creep and superplasticity’, Metal Forum, 1981, 4, 14–23. 28. Ruano, O.A., Wadsworth, J., and Sherby, O.D., ‘Deformation of fine-grained alumina by grain boundary sliding accommodated by slip’, Acta Mater. 2003, 51, 3617–34. 29. Jimenez-Melendo, M., and Dominguez-Rodriguez, A., ‘Like-metal superplasticity of fine-grained Y2O3-stabilized zirconia ceramics’, Phil. Mag, 1999, A 79, 1591– 608. 30. Langdon, T.G., ‘A unified approach to grain boundary sliding in creep and superplasticity’, Acta. Metall. Mater., 1994, 42, 2437–43. 31. Wang, J., Taleff E.M.. and Kovar, D., ‘High-temperature deformation of Al2O3/YTZP particulate composites’, Acta Mater., 2003, 51, 3571–83. 32. Raj, R., and Chyung, C.K., ‘Solution-precipitation creep in glass ceramics’, Acta Metal., 1981, 29, 159–66. 33. Wakai, F., ‘Step model of solution-precipitation creep’, Acta Mater., 1994, 42, 1163– 72. 34. Melendez-Martinez, J.J., Gómez-García, D., and Domínguez-Rodríguez, A., ‘A critical analysis and a recent improvement of the two-dimensional model for solutionprecipitation creep: application to silicon nitride ceramics’, Phil. Mag., 2004, 84, 2305–16. 35. Chen, I.W., and Hwang, S.L., ‘Shear thickening creep in superplastic silicon nitride’, J. Am. Ceram. Soc., 1992, 75, 1073–9. 36. Kim, B.N., Hiraga, K., Morita, K., and Sakka, Y., ‘A high-strain-rate superplastic ceramics’, Nature, 2001, 413, 288–91. 37. Yonn, C.K., and Chen, I.W., ‘Superplastic flow of two-phase ceramics containing rigid inclusions–zirconia/mullite composites’, J. Am. Ceram. Soc., 1990, 73, 1555– 65. 38. Mayo, M.J., ‘Processing of nanocrystalline ceramics from ultrafine particles’, Int. Mater. Rev., 1996, 41, 85–115. 39. Shen, Z., Peng, H., and Nygren, M., ‘Formidable increase in the superplasticity of ceramics in the presence of an electric field’, Adv. Mater., 2003, 15, 1006–9. 40. Wilson, J., and Kunz, S.M., ‘Microwave sintering of partially stabilized zirconia’, J. Am. Ceram. Soc., 1988, 71, C40-1. 41. Mizuno, M., Obata, S., Takayama, S., Ito, S., Kato, N., Hirai, T., and Sato, M., ‘Sintering of alumina by 2.45 GHz microwave heating’, J. Eur. Ceram. Soc., 2004, 24, 387–91. 42. Chen, I.W., and Wang, X.H., ‘Sintering dense nanocrystalline ceramics without final-stage grain growth’, Nature, 2000, 404, 168–71. 43. Mimurada, J., Nakano, M., Sasaki, K., Ykuhara, Y., and Sakuma, T., ‘Effect of cation doping on the superplastic flow in yttria-stabilized tetragonal zirconia polycrystals’, J. Am. Ceram. Soc, 2001, 84, 1817–21. 44. Nakatani, K., Nagayama, H., Yoshida, H., Yamamoto, T., and Sakuma, T., ‘The effect of grain boundary segregation on superplastic behavior in cation-doped 3YTZP’, Scripta Mater., 2003, 49, 791–5. 45. Morita, K., Hiraga, K., and Kim, B.N., ‘Effect of minor SiO2 addition on the creep behaviour of superplastic tetragonal ZrO2’, Acta Mater., 2004, 52, 3355–64. 46. Sato, E., Morioka, H., Kuribayashi, K., and Sundararaman, D., ‘Effect of small amount of alumina doping on superplastic behaviour of tetragonal zirconia’, J. Mater. Sci., 1999, 34, 4511–18.
© Woodhead Publishing Limited, 2006
Superplastic ceramic composites
457
47. Yoshida, H., Okada, K., Ikuhara, Y., and Sakuma, T., ‘Improvement of high-temperature creep resistence in fine-grained Al2O3 by Zr4+ segregation in grain boundaries’, Phil. Mag. Lett., 1997, 76, 9–14. 48. Wakai, F., Nagano, T., and Iga, T., ‘Hardening in creep of alumina by zirconium segregation at the grain boundary’, J. Am. Ceram. Soc., 1997, 80, 2361–6. 49. Yoshida, H., Yamamoto, T., Ikuhara, Y., and Sakuma, T., ‘A change in the chemical bonding strength and high-temperature creep resistance in Al2O3 with lanthanoid oxide doping’; Phil. Mag, 2002, A82, 511–25. 50. Yoshida, H., Ikuhara, Y., and Sakuma, T., ‘Grain boundary electronic structural related to the high-temperature creep resistance in polycrystalline Al2O3’, Acta Mater., 2002, 50, 2955–66. 51. Ikuhara, Y., Yoshida, H., and Sakuma, T., ‘Impurity effects on grain boundary strength in structural ceramics’, Mater. Sci. Eng., 2001, A319-321, 24–30. 52. Thavorniti, P., Ikuhara, Y., and Sakuma, T., ‘Microstructural characterization of superplastic SiO2-doped TZP with a small amount of oxide addition’, J. Am. Ceram. Soc., 1998, 81, 2927–32. 53. Ikuhara, Y., Yamamoto, T., Kuwabara, A., Yoshida, H., and Sakuma, T., ‘Structure and chemistry of grain boundaries in SiO2-doped TZP’, Sci. Tech. Adv. Mater, 2001, 2, 411–24. 54. Shinoda, Y., Nagano, T., Gu, H., and Wakai, F., ‘Superplasticity of silicon carbide’, J. Am. Ceram. Soc., 1999, 82, 2916–18. 55. Gu, H., Shinoda, Y., and Wakai, F., ‘Detection of boron segregation to grain boundaries in silicon carbide by spatially resolved electron energy-loss spectroscopy’, J. Am. Ceram. Soc., 1999, 82, 469–72. 56. Imamura, P.H., Evans, N.D., Sakuma, T., and Mecaryney, M.L., ‘High temperature tensile deformation of glass-doped 3Y-TZP’, J. Am. Ceram. Soc., 2000, 83, 3095– 99. 57. Clarke, D.R., ‘On the equilibrium thickness of intergranular glass phases in ceramic materials’, J. Am. Ceram. Soc., 1987, 70, 15–22. 58. Wilkinson., D.S., ‘Creep mechanisms in multiphase ceramic materials’, J. Am. Ceram. Soc., 1998, 81, 275–99. 59. Wakai, F., Kondo, N., and Shinoda, Y., ‘Ceramics superplasticity’, Curr. Opin. Sol. Sta. Mater. Sci., 1999, 4, 461–5. 60. Rosenflanz, A., and Chen, I.W., ‘Classical superplasticity of SiAlON ceramics’, J. Am. Ceram. Soc., 1998, 81, 713–16. 61. Wang, C.M., Mitomo, M., and Emoto, H., ‘Microstructure of liquid phase sintered superplastic silicon carbide ceramics’, J. Mater. Res. 1997, 12, 3266–70. 62. Nagano, T., Gu, H., Shinoda, Y., Zhan, D., Mitomo, M., and Wakai, F., ‘Tensile ductility of liquid-phase sintered β-silicon carbide at elevated temperatures’, Mater. Sci. Forum, 1999, 304–6, 507–12. 63. Gomez-Garcia, D., Lorenzo-Martin, C., Muñoz, A., and Domínguez-Rodriguez, A., ‘Model of high-temperature plastic deformation of nanocrystalline materials: Application to yttria tetragonal zirconia’, Phys. Rev., 2003, B67, 144101–1–8. 64. Gutierrez-Mora, F., Jimenez-Melendo, M., Domínguez-Rodriguez, A., and Chaim, R., ‘High temperature mechanical behavior of YSZ nanocrystals’, Key Eng. Mater, 2000, 171–4, 787–92. 65. Gutierrez-Mora, F., Domínguez-Rodriguez, A., and Jimenez-Melendo, M., Chaim, R., and Hefetz., M., ‘Creep of nanocrystalline Y-SZP ceramics’, Nanostructured Mater., 1999, 11, 531–7.
© Woodhead Publishing Limited, 2006
458
Ceramic matrix composites
66. Roddy, M.J., Cannon, W.R., Skandan, G., and Hahn, H., ‘Creep behaviour of nanocrystalline monoclinic ZrO2’, J. Eur. Ceram. Soc. 2002, 22, 2657–62. 67. Yoshida, M., Shinoda, Y., Akatsu, T., and Wakai, F., ‘Deformation of monoclinic ZrO2 polycrystals and Y2O3-stabilized tetragonal ZrO2 polycrystals below the monoclinic–tetragonal transition temperature’, J. Am. Ceram. Soc., 2002, 85, 2834– 6. 68. Yoshida, M., Shinoda, Y., Akatsu, T., and Wakai, F., ‘Superplasticity-like deformation of nanocrystalline monoclinic zirconia at elevated temperatures’, J. Am. Ceram. Soc., 2004, 87, 1122–5. 69. Venkatachari, K.R., and Raj, R., ‘Enhancement of strength through sinter forging’, J. Am. Ceram. Soc., 1987, 70, 514–20. 70. Kondo, N., Suzuki, Y., and Ohji, T., ‘Superplastic sinter-forging of silicon nitride with anisotropic microstructure formation’, J. Am. Ceram. Soc., 1999, 82, 1067–9. 71. Ye, J., and Dominguez-Rodriguez, A., ‘Joining of Y-TZP parts’, Scripta Metall. Mater., 1995, 33, 441–5. 72. Domínguez-Rodriguez, A., Guiberteau, F., and Jiménez-Melendo, M., ‘Heterogeneous junction of yttria partially stabilized zirconium by superplastic flow’, J. Mater. Res., 1998, 13, 1631–6. 73. Lee, I.G., and Chen, I.W., in ‘Sintering 87’, ed. Somiya, S., Yoshimura, M., and Watanabe, R., Elsevier, London, 1988, vol. 1, pp. 340–5. 74. Domínguez-Rodriguez, A., Jiménez-Pique, E., and Jiménez-Melendo, M., ‘High temperature mechanical properties of a multilayer Y-TZP processed by superplastic flow’, Scripta mater.; 1998, 39, 21–5. 75. French, J.D., Zhao, J., Harper, M.P., Chan, H.M., and Millar, G.A., ‘Creep of duplex microstructures’, J. Am. Ceram. Soc., 1994, 77, 2857–65. 76. Gutiérrez-Mora, F., Domínguez-Rodriguez, A., Routbort, J.L., Chaim, R., and Guiberteau, F., ‘Joining of yttria-tetragonal stabilized zirconia polycrystals using nanocrystals’, Scripta Mater., 1999, 41, 455–60. 77. Molteni, C., Francis, G.P., Payne, M.C., and Heine, V., ‘First principles simulation of grain boundary sliding’, Phys, Rev. Lett., 1996, 76, 1284–7. 78. Yoshida, H., Ikuhara, Y., and Sakuma, T., ‘Grain boundary electronic structure related to the high-temperature creep resistance in polycrystalline Al2O3’, Acta Mater., 2002, 50, 2955–66. 79. Chandra, N., ‘Mechanisms of superplastic deformation at atomic scale’, Mater. Sci. Forum, 1999, 304–6, 411–20.
© Woodhead Publishing Limited, 2006
Part V Non-oxide ceramic composites
© Woodhead Publishing Limited, 2006
17 Interfaces in non-oxide ceramic composites S T U R A N, Anadolu University, Turkey and K M K N O W L E S, University of Cambridge, UK
17.1
Introduction
An interface is a meeting surface between two dissimilarly oriented perfect crystals or between two chemically different crystals, i.e., a crystallographic or chemical discontinuity. In terms of crystal structure and chemistry of adjacent crystals, interfaces can be classified into homophase and heterophase boundaries (Cahn and Kalonji, 1982; Finnis and Rühle, 1993). Homophase boundaries form between grains of identical crystal structure and composition, e.g., grain boundaries, twin boundaries, domain boundaries and stacking faults. Heterophase boundaries form between regions of different crystal structure and/or chemical composition, e.g., interfaces between co-existing polytypes, between metal-ceramic joints and most interfaces between the matrix and the reinforcing agents in composite materials. Hence, the term ‘interface’ is a general term, whereas the terms ‘grain boundary’ and ‘interphase boundary’ are more specialised terms used to distinguish between the two types of interface. In contrast to grain boundaries and interphase boundaries in metals, in which grains make intimate contact at the atomic level regardless of the orientation relationship across the boundary (Sutton and Balluffi, 1995), grain boundaries in non-oxide engineering ceramics and interphase boundaries in composites invariably contain thin films of extraneous material which arise through the presence of impurities and additives present in the starting powders used to make the ceramics and composites (Kleebe et al., 1992; Turan and Knowles, 1995). Here, we will use the term ‘intergranular film’ to describe a phase at an interface. In the context of the non-oxide particulate ceramic composites that we will discuss here, such films are typically 1–2 nm thick. This range of thickness contrasts with the thickness of chemically distinct regions between the matrix and the reinforcing fibres in fibre-reinforced ceramics where such regions can have thicknesses between 50 nm and a few micrometres (see, for example, Kumar and Knowles, 1996a, 1996b). If this intergranular film is amorphous, a distinction has to be made between 461 © Woodhead Publishing Limited, 2006
462
Ceramic matrix composites
segregation of chemical species and a genuine intergranular film. Rühle et al. (1984) have suggested that an intergranular film forms when the density of any segregated impurities is larger than one atomic layer, otherwise the term segregation is more appropriate. Terms such as ‘glassy films’ to describe intergranular films should be used with caution, since there is some evidence that these films are more ordered than glass at triple junctions (Marion et al., 1987). Intergranular films have traditionally been viewed as a problem for the mechanical properties of engineering ceramics and composites. Typically these films have a lower fracture toughness than the surrounding grains, with a correspondingly lower fracture stress. In addition, they tend to have lower softening points or melting points than the more refractory matrix grains. The more amorphous phase there is present in an engineering ceramic or an engineering composite, the more it will be expected to creep at high temperatures, because of the exponential decrease in the viscosity of amorphous phase with increasing temperature. For example, Choi et al. (2004) found that silicon carbide (SiC) ceramics are able to retain their strength at 1500°C when AlN and either Sc2O3 or Lu2O3 are present in the starting powders because the grain boundaries which form are free from intergranular films. However, intergranular films can be beneficial. At temperatures below 1000°C, Pezzotti (1993), Becher et al. (1998) and Sun et al. (1998) have all reported that intergranular films can be used very effectively to enhance the matrix fracture toughness by controlling the interface chemistry. For example, it has been shown that the fracture toughness of in-situ reinforced silicon nitride (Si3N4) ceramics can be optimised by a suitable choice of sintering aids to control the composition of the grain boundary films as well as the surface chemistry of the adjoining grains (Becher et al., 1998; Sun et al., 1998). If the grain boundaries between adjacent matrix grains in in-situ reinforced silicon nitride are too strong, cracks will propagate through the matrix silicon nitride grains rather than be deflected along the more tortuous grain boundaries. Such a silicon nitride engineering ceramic will not have an enhancement of its fracture toughness over more conventional dense silicon nitride with equiaxed grains. However, if the grain boundaries between adjacent matrix grains are too weak, the silicon nitride will have a low strength, even if it has a higher fracture toughness than a more conventionally made silicon nitride. Thus, interfacial structure and interfacial chemistry both play important roles in the toughening and strengthening response of the material. Amorphous intergranular films also play an important role in both the β (or 3C) → α phase transformation in SiC (Moberlychan et al., 1998) and the reverse α → β (or 3C) phase transformation (Turan and Knowles, 1996a). In this overview, we will first discuss how transmission electron microscopy (TEM) techniques can be used to determine the presence or absence of intergranular amorphous phases at interphase boundaries in structural
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
463
engineering ceramics and non-oxide composites. We will show how the observed distribution of thicknesses of the order of 1−2 nm for such films can be understood in terms of theoretical models of the attainment of an equilibrium film thickness from suitable competing attractive and repulsive forces at interphase boundaries. We will then discuss the evidence for and against the development of preferred orientation relationships and good lattice matching at intergranular and interphase boundaries, because such considerations are also relevant for the development of models for the strength and toughness of engineering ceramics and composites. Finally, we will examine future trends within this research area.
17.2
Assessment of the accuracy of TEM techniques for the detection and measurement of film thickness at interfaces
Detailed characterisation of interfaces requires high magnifications and high instrumental resolutions. Usually, this precludes the use of techniques such as scanning electron microscopy, requiring instead more specialist TEM or scanning transmission electron microscopy (STEM) techniques. TEM techniques used to date for the characterisation of interfaces in non-oxide ceramic composites are bright field imaging, dark field imaging, diffuse dark field imaging from the part of reciprocal space into which there is scattering from the intergranular amorphous phase, weak beam dark field imaging, diffraction pattern analysis, Fresnel defocus imaging and high resolution transmission electron microscopy (HRTEM) (Clarke, 1979; Cinibulk et al., 1993a, 1993b; Turan, 1995; Kleebe, 1997, 2002). A recent elegant demonstration of the capabilities of STEM has been shown by Shibata et al. (2004) who used aberration corrected Z-contrast STEM to examine the segregation of lanthanum atoms to intergranular films at grain boundaries in silicon nitride. For these techniques, the interface under examination is oriented edge-on so that it is parallel to the electron beam. For HRTEM, grains at either side of the interface must be diffracting strongly, ideally with the electron beam parallel to a low-index zone of each grain, whereas in Fresnel fringe analysis the two grains must be diffracting weakly. Diffuse dark field imaging, HRTEM and Fresnel defocus imaging are particularly useful for characterising interfaces containing intergranular films, but, in addition, analytical electron microscopy techniques such as energy dispersive X-ray (EDX) and parallel electron energy loss spectroscopy (PEELS) need to be used to obtain chemical information about the intergranular films. There is no single experimental technique capable of characterising an interface structure fully. Therefore, it is desirable to use techniques which are complementary in order to maximise the available information as well as
© Woodhead Publishing Limited, 2006
464
Ceramic matrix composites
the reliability of the results. A comprehensive study by Cinibulk et al. (1993a) in which diffuse dark field imaging, HRTEM and Fresnel defocus imaging were all compared concluded that HRTEM gives an accuracy of ±1 Å for intergranular film thickness measurements if grains either side of the interphase boundary have at least a one-dimensional lattice image to distinguish them from the intergranular amorphous phase. By comparison, the Fresnel technique predicts a thickness 20−35% higher than HRTEM, and diffuse dark field measurements predict thicknesses some 50−100% higher than those measured by HRTEM. Similar trends were first noted by Clarke (1979). An example of the use of Fresnel defocus imaging, diffuse dark field imaging and HRTEM to examine the same SiC–SiC interface is shown in Fig. 17.1 (Turan, 1995). All three techniques clearly indicate the presence of (a)
3C SiC α-SiC 3C SiC (b)
(c)
21 Å
20 nm
17.1 (a) Underfocus, (b) overfocus Fresnel fringe images obtained from an edge-on interface between SiC grains, (c) a diffuse dark field image, (d) an HRTEM image from the same interface, and (e) a plot of Fresnel fringe spacing against defocus for a part of the interface where two 3C SiC grains are separated with an intergranular film shown in (a) and (b). The contamination and damage evident in (a)–(d) arises from the time required for the edge-on alignment of the interface and also because of the high voltages used in HRTEM.
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
465
3C SiC
α-SiC
3C SiC
10 nm (d) 5.5
Fringe spacing (nm)
4.5
3.5
2.5
1.5 0.5 –6000
–4000
–2000 0 2000 Defocus (nm)
4000
6000
(e)
17.1 Continued
an intergranular amorphous film at this interface. The images shown in Fig. 17.1 were taken on two different transmission electron microscopes, since the microscope on which the HRTEM work was undertaken did not have a small enough objective aperture to obtain high-contrast Fresnel and diffuse dark field images. Using the methodology described by Cinibulk et al. (1993a), the thickness of the intergranular film was measured to be approximately 13 ± 1 Å from HRTEM images well away from the small triple junction between two 3C grains and α-SiC, but it will be apparent from Fig. 17.1(d) that the thickness varies as a function of position along the interface, particularly in the vicinity of the triple junction.
© Woodhead Publishing Limited, 2006
466
Ceramic matrix composites
One of the advantages of HRTEM over the other techniques in measuring the intergranular film thickness accurately is that an internal calibration is possible for every single image, whether originally in the form of a negative or a digitised image. Other, more general, calibration methods are usually needed for the two other techniques. However, if phases are present with a very large interplanar spacing which can be suitably oriented, such as the (0001) planes of 6H polytype of α-SiC which have an interplanar spacing of ~15 Å, as in Fig. 17.1(d), lattice fringes from these (0001) planes will be readily visible in any of the recent generation of transmission electron microscopes at the typical magnifications of 100 000–150 000 used for Fresnel imaging. Thus, an image at the magnification used for Fresnel imaging conditions with these lattice fringes present can be used to calibrate the magnifications used during subsequent Fresnel imaging of the interfaces and during the eventual image processing. The through-focal series for Fresnel imaging analysis for the interface in Fig. 17.1 contained 21 negatives in total taken from underfocus (∆F = −3830 nm) to overfocus (∆F = +3830 nm). Six of these images were used to find the exact position of the Gaussian focus, where ∆F = 0 nm, through an examination of the intensity of the Fresnel fringes at the interface. Unfortunately, low fringe visibility around Gaussian focus prevents accurate measurements of the Fresnel fringe spacings at low defocus values. The remaining 15 fringe spacings were then measured and plotted against their corresponding defocus values (Fig. 17.1(e)). The film thickness obtained from Fig. 17.1(e) by extrapolating the high defocus values to low defocus values so that the underfocus and overfocus curves meet corresponds to a film thickness of ~16 Å. This is ~23% higher than the value obtained from HRTEM measurements. The film thickness was determined to be ~21 Å from the diffuse dark field image which is ~ 60% larger from that of HRTEM measurements and is also consistent with the results obtained by Clarke (1979) and Cinibulk et al. (1993a). Diffuse dark field imaging is also the technique most likely to produce artefacts, most notably from preferential etching from ion beam thinning of the interfaces and subsequent sputter deposition or damage in these regions. The question of whether modern-day SEMs can be used to infer the presence or absence of intergranular films at grain boundaries has recently been addressed by Kleebe (2002). Kleebe characterised two different nonoxide ceramics, Si3N4 and SiC, with respect to their grain boundary structure using both SEM and TEM. SEM imaging of plasma etched surfaces revealed a characteristic bright contrast along the interfaces for both ceramics, suggesting the presence of an intergranular amorphous film. HRTEM studies of the Si3N4 sample confirmed that these fine bright lines along grain boundaries represented intergranular amorphous films separating Si3N4 matrix grains.
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
467
However, when HRTEM was employed on the SiC samples, which showed a similar contrast variation across SiC grain boundaries in the SEM, the presence of residual intergranular films was not detected even at the triple junctions. Hence, Kleebe concluded that SEM imaging and Fresnel fringe TEM imaging alone do not enable a safe conclusion to be drawn about interface wetting in ceramic polycrystals. A further aspect of interfaces in engineering ceramics was addressed by Gu and Shinoda (2000). These workers used spatially resolved electron energyloss spectroscopy (EELS) analysis and spatially resolved energy-loss nearedge structures (ELNES) analysis to characterise interfaces in a hotisostatically-pressed SiC material whose powders contained ~3 wt% free carbon and 1 wt% amorphous boron. They demonstrated that the structural width and the chemical width of general boundaries were not the same. Their HRTEM observations did not detect the existence of any amorphous film at such grain boundaries. Instead, each grain boundary had a core structure 1–2 atomic planes wide in which B–C and Si–O bonds were detected by EELS and ELNES, as well as Si–C bonds subtly different from the Si–C bonds in the grain interiors. The chemical width of the grain boundaries that they inferred from EELS analysis of elemental profiles was visibly wider than this core region. ELNES analysis distinguished a third extended grain boundary width larger than the chemical width within which the Si–C bonding was modified from that of the grain interiors, being most strongly modified at the grain boundary and decreasing in modification with increasing distance from the grain boundary. Thus, in these materials, they concluded that intergranular films were not seen, but rather segregation of boron and oxygen to the grain boundaries, with modification of the Si–C bonding extending several lattice planes into each adjacent SiC grain.
17.3
Wetting, non-wetting and dewetting behaviour of interphase boundaries in non-oxide ceramic composites
TEM studies using Fresnel, diffuse dark field, HRTEM, Z-contrast imaging and spatially resolved electron energy-loss spectroscopy techniques have revealed that, while most of the interfaces in non-oxide engineering ceramics and composites are covered with intergranular films (Clarke, 1979; Cinibulk et al., 1993a, 1993b; Turan, 1995; Gu et al., 1995; Turan and Knowles, 1995; Kleebe, 2002), some interfaces are clearly not (Schmid and Rühle, 1984; Turan, 1995; Knowles and Turan, 1996; Turan and Knowles, 1997, 1999). Interfaces without intergranular films will be considered in more detail in Section 17.6. Examples of wetting and non-wetting behaviour of grain boundaries are shown in Fig. 17.2. The converse behaviour to wetting, dewetting of a liquid from an interface
© Woodhead Publishing Limited, 2006
468
Ceramic matrix composites
(a)
(b)
Si3N4
α-SiC
3C SiC
Si3N4
2 nm
5 nm (c) Si3N4 Si3N4
Si3N4 10 nm
17.2 HRTEM examples of non-wetting behaviour at interfaces in (a) SiC and (b) Si3N4, and wetting behaviour at an interface in (c) Si3N4. (a) and (c) reprinted from Mat. Sci. Forum, Turan S and Knowles KM, ‘Wetting and non-wetting behaviour of SiC grain boundaries’, 294–296, 313–316 (1999) with kind permission of Trans. Tech. Publications.
to the sample surface, is also occasionally reported in liquid-phase sintered ceramics. For example, such behaviour has been observed on aluminium nitride (AlN) ceramics when annealed at high temperature in a highly reducing atmosphere (Ueno and Horiguchi, 1989) and in SiAlON ceramics after vacuum heat treatment (Mandal and Thompson, 2000). Kleebe and Pezzotti (2002) characterised the grain-boundary structure of a model SiAlON polycrystal with nominal composition Si5AlON7 by TEM both in an equilibrium (as-
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
469
processed) state at room temperature and after quenching from elevated temperature. They found that in the equilibrated low-temperature microstructure amorphous phases existed only at multigrain junctions, and not at grain boundaries. However, samples of this model polycrystal heated to 1380°C and rapidly quenched showed wetted grain boundaries.
17.4
Equilibrium film thickness at interphase boundaries
Other than the wetting, non-wetting and dewetting behaviour of interfaces, the other striking TEM observation is that the thickness of intergranular films seems to be relatively constant from one interface to another in any given sample, but that it varies from material to material. Theoretical considerations by Clarke and co-workers (Clarke, 1987; Clarke et al., 1993) show that an equilibrium film thickness arises from the competition between attractive dispersion forces determined by the dielectric properties of the grains and repulsive disjoining forces which can be steric forces and/ or double-layer forces. Wetting will occur when the solid–solid boundary energy, γb, is less than that of the wetted boundary, 2γl, where γl is the liquid– solid interfacial energy (Clarke, 1985), provided that there is a suitable source of liquid, for example as a consequence of liquid-phase sintering at high temperatures. The most general equilibrium condition which applies to a thin film sandwiched between two phases when there is an applied pressure, P, and a capillary pressure, PCAP, is P + PCAP + ΠDISP + ΠST + ΠEDL + ΠADS + ΠHB = 0
(17.1)
where ΠDISP is an attractive dispersion force per unit area of interface arising from van der Waals forces, ΠST is a repulsive steric force per unit area, ΠEDL is a repulsive electrical double-layer force per unit area, ΠADS is a repulsive force per unit area arising from the effects of any solute absorption, and ΠHB is an attractive force per unit area arising from any hydrogen bonding present (Clarke, 1987; Clarke et al., 1993). Clarke (1987) examined the form of Eq. (17.1) for the situation where ΠEDL = ΠADS = ΠHB = 0, in which case the repulsive force per unit area enabling an equilibrium film thickness to arise is simply ΠST. Subsequently, Clarke et al. (1993) examined the situation where ΠEDL ≠ 0, ΠADS = ΠHB = 0 for zero and finite values of ΠST, concluding that it is only under certain restricted conditions that it is plausible for an electrical double-layer force to contribute significantly to the total repulsive force. If the only significant repulsive force per unit area is ΠST, as is likely to be the case for interphase boundaries between non-oxide engineering ceramics such as SiC, Si3N4 and h-BN, the net force, F, per unit area pushing two grains together becomes
© Woodhead Publishing Limited, 2006
470
Ceramic matrix composites
F = P + PCAP + Π DISP + Π ST = P + PCAP +
aη A – 3 6π L sinh 2 ( L /2ξ ) (17.2) 2
where L is the thickness of the film, A is the Hamaker constant determining the magnitude of the attractive dispersion force, ξ is a molecular correlation distance and aη2 is a constant in the term representing the repulsive steric force per unit area which is the free energy difference between ordered and disordered states of the film (Clarke, 1987). If L is the equilibrium thickness, F is simply zero. The trend which arises from a consideration of Eq. (17.2) is that the lower the value of the Hamaker constant, A, the higher the equilibrium film thickness (Clarke, 1987). Striking confirmation experimentally of such a trend has come from work in which the local Hamaker constants in silicon nitride ceramics have been determined from spatially resolved-valence electron energyloss spectroscopy (French et al., 1998). Conversely, if A is too large, then there will be no thickness L for which Eq. (17.2) is satisfied. For a suitably high critical value of A, this theoretical model predicts a lower limit on the equilibrium thickness that can be observed. This lower limit on L, Lmin, is defined by the conditions F = 0 and dF/dL = 0 since for a stable film F = 0 and dF/dL ≥ 0 (Clarke, 1987). Various solutions to these conditions have been examined by Knowles and Turan (2000). In the absence of capillary pressure and external pressure, Lmin = 2.58ξ. Using reasonable estimates for ξ, Knowles and Turan estimate Lmin to be ≥6.50 Å. That in practice the observed intergranular film thicknesses are typically of the order of 1–2 nm in non-oxide engineering ceramics indicates that the relevant Hamaker constants for ceramics are significantly lower than the critical value. Israelachvili (1992) discusses the various approximate and analytic formalisms for A, with the conclusion that where two macroscopic isotropic phases 1 and 2 interact across an isotropic medium 3, a suitable approximation to the relevant Hamaker constant, A132, valid for L greater than molecular dimensions is:
( ε – ε 3 )( ε 2 – ε 3 ) A132 = 3 k B T 1 4 ( ε 1 + ε 3 )( ε 2 + ε 3 ) +
( n12 – n32 )( n 22 – n32 ) 3hν e 8 2 ( n12 + n32 )1/2 ( n 22 + n32 )1/2 ( n12 + n32 )1/2 + ( n 22 + n32 )1/2
[
]
(17.3) where kB is Boltzmann’s constant, T is absolute temperature, ε1, ε2 and ε3 are the static dielectric constants of the three phases, νe is the characteristic absorption frequency taken to be the same for all three materials, h is Planck’s
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
471
constant and n1, n2 and n3 are the refractive indices of the three materials, extrapolated to zero energy, or equivalently, zero frequency. Of the two terms on the right-hand side of Eq. (17.3), the first term can be at most 3 4 kBT. For most situations of practical interest, this term is significantly smaller than the second term. It is usual in calculations using Eq. (17.3) for the common characteristic absorption frequency νe to be assigned a value of 3 × 1015 s–1 (see, for example, Israelachvili, 1992; French, 2000). Horn and Israelachvili (1981) have derived a slightly more complex form of Eq. (17.3) for the situation where materials 1 and 2 have the same absorption frequency but material 3 has a different value of absorption frequency, and Prieve and Russel (1988) have derived a form of A132 for the most general situation where the three materials have different absorption frequencies ν1, ν2 and ν3. Knowles and Turan (2000) and Knowles (2005) have used the approach of Parsegian and Weiss (1972) to examine the effect of anisotropy on Hamaker constants. Such calculations are relevant for materials such as h-BN and rutile, TiO2, which exhibit strong anisotropy in their refractive indices because of their crystal structure. They make little difference to predicted values of Hamaker constants as a function of grain orientation for materials such as Si3N4 and SiC which, by comparison, exhibit modest values of birefringence. The analysis of Knowles and Turan (2000) of h-BN–amorphous silica–3C SiC interfaces showed that Eq. (17.3) could be used to calculate values of the Hamaker constant as a function of the orientation of h-BN with respect to a planar interface containing a thin amorphous silica film, provided that the effective values of static dielectric constant and refractive index for h-BN, εh–BN and nh–BN respectively, were taken to be
sin 2 θ ε h–BN = ( ε o )1/2 ε e + ( ε o – ε e ) 2
1/2
(17.4)
and 2 2 n h–BN = 1 n o2 + 1 n e2 + ( n o2 – n e2 ) sin θ 2 2 4
(17.5)
where θ is the angle between the normal to the interface plane and the normal to the (0001) h-BN planes, no is the refractive index of the ordinary rays in hBN, and n e is the refractive index of the extraordinary ray. For h-BN no >> ne and εo >> εe. However, the dominance of the second term on the right in Eq. (17.3) means that A is most sensitive to the difference between the effective refractive index, nh-BN, of h-BN and the refractive index of amorphous silica. Since nh-BN is always greater than the refractive index of amorphous silica for all values of θ, it follows that for h-BN–amorphous silica–3C SiC interfaces A is predicted to increase as θ increases from 0° to 90°. Detailed
© Woodhead Publishing Limited, 2006
472
Ceramic matrix composites (a) h-BN
12 Å
3C SiC
(b) h-BN
12 Å
3C SiC
(c) h-BN
8.5 Å
3C SiC
17.3 HRTEM observations of three differently misoriented interphase boundaries between hexagonal boron nitride (h-BN) and 3C silicon carbide (3C SiC) grains showing an orientation dependence on equilibrium film thickness. In (a) and (b) the (0001) of the highly anisotropic h-BN are parallel to the interface, whereas in (c) they make an angle of 68° with the interphase boundary (reprinted from Ultramicroscopy, Knowles KM and Turan S, ‘The dependence of equilibrium film thickness on grain orientation at interphase boundaries in ceramic–ceramic composites’, 83(3/4) 245–259 (2000) with kind permission of Elsevier Science).
calculations by Knowles and Turan showed that this increase was significant: at θ = 0°, A was calculated to be 113 zJ, increasing to 139 zJ at θ = 90°. The limited experimental observations they were able to make on h-BN–3C SiC interfaces containing amorphous silica-rich intergranular films were consistent with this trend in A, i.e., those interfaces parallel to (0001) h-BN planes had thicker intergranular films of 12 ± 1 Å, whereas the intergranular film at an interface where the (0001) h-BN planes made an angle of 68° with respect to the interface plane was noticeably smaller, ~ 8.5 Å (Fig. 17.3).
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
17.5
473
Effect of intergranular film composition on equilibrium film thickness
As we have already noted in Section 17.4, an examination of the behaviour of Eq. (17.2) shows that the equilibrium film thickness, L, increases with decreasing Hamaker constant. Furthermore, the dominant term in Eq. (17.3) for the Hamaker constant for two isotropic media interacting across a third isotropic medium is the one dependent on the refractive indices of the three media, n1, n2 and n3 respectively. Thus, if n1 and n2 are both greater than n3, as will be the case for silica-rich films between grains of either SiC or Si3N4, it is obvious from Eq. (17.3) that the higher the refractive index of the intermediate third medium, the lower the Hamaker constant will be, and, in turn, the higher L will be. Conversely, a decrease in the molecular correlation length, ξ, will cause a decrease in L. The refractive index of the intermediate third medium is determined by its chemical composition. Although the exact chemical compositions of intergranular films have to be known for accurate Hamaker constant calculations, experimental limitations of both interference with adjacent grains and beam broadening during chemical analysis in the electron microscopes have in practice meant that it has not been possible to ascertain these with confidence. The chemical analysis that has been undertaken has shown that most of the films in ceramics are amorphous silica mixed with a very small amount of metal ion contamination from the starting powders. In addition, there are some indications that intergranular silica films can, and will, contain chemical elements from sintering additives and incomplete chemical reactions during processing. For example, nitrogen and yttrium enrichment would be expected in a predominantly silica-rich intergranular film between Si3N4 and Y2Si2O7 grains, Si–C–O glasses rather than amorphous silica would be expected between SiC grains, and Si–C–O–N glasses rather than amorphous silica would be expected between SiC and Si3N4 grains. The effect of the nitrogen content on the refractive index of M-SiAlON glasses reported in Lewis (1989) clearly shows that when the nitrogen content increases, the refractive index of the glass also increases. It is also believed (Lewis, 1989) that by replacing oxygen, nitrogen increases the density of SiAlON materials, because nitrogen has a superior ability to oxygen for cross-linking the network structure of oxynitride glass − nitrogen atoms are capable of bridging three network tetrahedral groups, whereas oxygen atoms are capable of bridging only two. In this context, it is relevant to examine a situation where two Si3N4 grains are separated by a supposedly pure silica film and to assume instead that 10 mol% nitrogen is in it, undetected, as has been claimed by Vetrano et al. (1993). From Lewis (1989), the effect of 10 mol% nitrogen is predicted to increase the refractive indices of the SiAlON glasses by ~4%. This increase, while modest, increases the refractive index
© Woodhead Publishing Limited, 2006
474
Ceramic matrix composites
of silica from 1.45 to 1.51 and decreases the predicted Hamaker constant from 82 to 65 zJ. Thus, a small amount of nitrogen enrichment in the film can radically affect the Hamaker constant and therefore the equilibrium film thickness. Kleebe (1997) has summarised experimental results on equilibrium film thicknesses from a number of studies on interfaces in Si3N4 ceramics and discussed them in the light of the competing attractive and repulsive forces present across the films and their chemical composition. One of these studies in which calcium dopant was deliberately introduced into the processing procedure is of particular note. In this study, originally reported by Tanaka et al. (1994), the equilibrium film thickness at Si3N4 grain boundaries was shown to be very sensitive to the addition of small quantities of calcium ions. The film thickness was found to be 10 Å for undoped material and 7, 11 and 15 Å for samples doped with 80, 220 and 450 ppm Ca respectively. As expected, calcium was detected at the grain boundaries in the doped specimens. Tanaka et al. (1994) were able to account qualitatively for their results by considering the effect of calcium ions on the silica network structure and also the development of a repulsive electrical double-layer force, ΠEDL, through the incorporation of calcium ions as adsorbed species on the grain surfaces. The qualitative argument they advocated was that for small additions of calcium, any repulsive electrical double-layer force is more than compensated by the disruption of the silica network which lowers the magnitude of the molecular correlation distance, ξ, so that as a result the thickness of the film, L, decreases relative to calcium-free Si3N4 grain boundaries. For larger additions of calcium, Tanaka et al. (1994) argue that ΠEDL will increase, so that L will begin to increase with calcium concentration. With the increase in the speed and capacity of modern computers, it is now possible to undertake atomistic simulations of the structure of intergranular films. The recent work of Garofalini and Luo (2003) and Su and Garofalini (2004a, 2004b) on calcium silicate intergranular films in silicon nitride is directly relevant to the experimental work of Tanaka et al. (1994). Thus, for example, Su and Garofalini (2004a) examined film compositions of 1.5 mol% calcium, equivalent to 80 ppm calcium doping in the bulk material. Their work suggests that an electrical double layer does not form at this level of calcium, as they found no evidence of calcium segregation to the grain surfaces. However, their simulations do show evidence for ordering within the intergranular film induced by the crystal surfaces and the incorporation of nitrogen into the films. The study of Garofalini and Luo (2003) on higher concentrations of calcium in the intergranular layer also failed to show evidence for segregation of calcium to the grain surfaces, although it did find evidence for alignment of the calcium ions within the films parallel to the grain surfaces, which weakened the interface by causing a decrease in the number of Si–O– Si bonds across the interface. Interestingly, none of these recent atomistic
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
475
studies attempted to interpret the results of their atomistic simulations quantitatively in terms of the various terms in the force balance of Eq. (17.1) – this suggests that much still remains to be done to understand how the detail of the intergranular film chemical composition at an atomic level determines the equilibrium film thickness.
17.6
Crystallography of interphase boundaries
In contrast to the number of research studies concerning the thickness and chemistry of intergranular films in non-oxide ceramics and ceramic composites, there have been relatively few studies on the crystallography of either grain boundaries or interphase boundaries in these materials. On the basis of TEM evidence Niihara et al. (1990) suggested that low energy interfaces develop preferentially in Si3N4 grains containing small SiC precipitates, without specifying uniquely the orientation relationships and interface planes that they observed. A more detailed two-part TEM study by Unal et al. (1992) and Unal and Mitchell (1992) on chemically vapour deposited Si3N4 grown on single crystal SiC found two characteristic orientation relationships between the Si3N4, which deposited itself as α-Si3N4, and the substrate SiC, which near the surface was twinned 3C SiC. The dominant orientation relationship that they reported can be described approximately as [101] 3C SiC || [0001] α-Si3N4 with ( (11 1 ) ) 3C SiC || ( (10 1 0) ) α-Si3N4. Unal et al. accounted for their observations in terms of the need to match SiN4 and SiC4 tetrahedra favouring ( (11 1 ) ) 3C SiC || (10 1 0) α-Si3N4 || {111} 3C SiC. They accounted qualitatively for small rotations away from this ideal orientation relationship in terms of the relief of structural mismatch arising from the large misfit between the corresponding crystal planes of the two crystal structures. This orientation relationship was also reported by Pan et al. (1996a) between 3C SiC and β-Si3N4 for the ‘type A’ SiC nanoparticles they observed surrounded by β-Si3N4 matrix. β-Si3N4, which has a hexagonal structure with a = 7.6044 Å and c = 2.9075 Å (JCPDS-ICDD No. 41-0360) is very similar in structure to α-Si3N4, which has a trigonal structure with a = 7.7541 Å and c = 5.6217 Å (JCPDS-ICDD No. 33-1160) and twice the c lattice repeat. Pan et al. also observed that amorphous intergranular films were only seen in some parts of the interfaces for the ‘type A’ nanoparticles. In contrast to these ‘type A’ nanoparticles, Pan et al. also found other nanoparticles which they designated ‘type B’. These nanoparticles had random orientation relationships with respect to the surrounding β-Si3N4 matrices and evidence of substantial amorphous intergranular phase present at the SiC−β-Si3N4 interfaces. A more detailed study of the crystallography of SiC−Si3N4 interphase boundaries was undertaken by Turan and Knowles (2000). The composites that they investigated containing either 10 or 20 wt% Si3N4, with the balance SiC. Composites were prepared by hot isostatic pressing. Two different types
© Woodhead Publishing Limited, 2006
476
Ceramic matrix composites
of interphase boundary were found in these composites as a consequence of the bimodal size distribution of the β-Si3N4 grains formed in the composites during the processing operation. In general, interphase boundaries between small intragranular β-Si3N4 precipitates and surrounding SiC grains were found to be relatively free of intergranular films, whereas interphase boundaries between large β-Si3N4 grains and adjacent SiC grains were invariably covered with thin intergranular films. Orientation relationships approximating to [110] 3C SiC || [0001] β-Si3N4 and (001) 3C SiC || ( (10 1 0) ) β-Si3N4 were found to dominate between the 3C SiC grains and the small intragranular β-Si3N4 precipitates, the interfaces of which were either clean or relatively devoid of amorphous intragranular materials (Fig. 17.4). In marked contrast to this, there was no evidence of any favoured orientation relationship between the large β-Si3N4 grains and adjacent 3C SiC grains. This dominant approximate orientation relationship is the same as one of the orientation relationships reported by Pan et al. (1996b) between 3C SiC particles and a surrounding β-Si3N4 grain. It is closely related to the orientation relationship reported by Unal et al. (1992), Unal and Mitchell (1992) and Pan et al. (1996a): a relatively small rotation of 5.26° about the common [110] 3C SiC || [0001] β-Si3N4 direction is required to bring (1 1 1) 3C SiC || (01 1 0) β-Si3N4. (a)
3C SiC
(b)
Si3N4
100 nm
17.4 (a) A dark field image of an interphase boundary between a SiC and Si3N4 precipitate when the electron beam is parallel to [110], (b) its composite diffraction pattern (c) schematic overlapping diffraction pattern, (d) a low magnification HRTEM image showing a Si3N4 precipitate embedded in a 3C SiC grain, (e) its composite diffraction pattern, (f) schematic overlapping diffraction patterns, (g)-(i) HRTEM images from IB1, IB6 interphase boundaries and the region between IB1 and IB6 in (d) (reprinted from Interface Science, Turan S and Knowles KM, ‘The crystallography of interface bondaries between silicon carbide and silicon nitride in silicon nitride–silicon carbide particulate composites’, 8(2/3) 279–294 (2000) with kind permission of Springer Science and Business Media).
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
477
002
002
111 3C SiC IB1 (c) IB6
IB2
Si3N4
IB5 IB3
IB4 20 nm (d)
(e)
17.4 Continued
© Woodhead Publishing Limited, 2006
478
Ceramic matrix composites m1
220
(2420) 111
11 1 (1210)
002
002 (10 1 0)
1 11
m2 (3030)
111
Si3N4 3C SiC 220
(f)
Si3N4
3C SiC
5 nm
(g)
5 nm (h)
5 nm (i)
17.4 Continued
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
479
At present there is no suitable atomistic modelling algorithm which can be used to examine structures and energies of interfaces between crystalline phases where one or more phase is covalently bonded. Therefore, Turan and Knowles explored the rationale for ‘special orientation relationships’ arising when there is no evidence for the presence of an intergranular film present at SiC−β-Si3N4 interfaces geometrically using the near-coincidence site lattice model. In such a purely geometric approach, the details of atomic bonding at the interfaces are necessarily of secondary importance, even though ultimately the adoption of any particular three-dimensional orientation relationship and the energetics of the interface will be determined by the way in which the atoms bond at the interface. Their computations showed that, relative to all possible orientation relationships between 3C SiC and β-Si3N4 that could be adopted, the dominant orientation relationships between the 3C SiC grains and the intragranular β-Si3N4 precipitates have low misfits. While obvious caution must be exercised in the use of purely geometric criteria for low interfacial energies (see, for example, Sutton and Balluffi (1987) who conclude that no geometric criterion for low interfacial energy can be regarded as wholly reliable), it must be regarded as significant that the dominant observed orientation relationships between the 3C SiC grains and the intragranular β-Si3N4 precipitates have low misfits, even though it is not possible to infer on the basis of geometry alone that low-energy interfaces occur when these orientation relationships are adopted. Turan and Knowles (1997) and Knowles and Turan (2002) have also explored the crystallography of interfaces between β-Si3N4 and hexagonal boron nitride, h-BN, and interfaces between 3C SiC and h-BN, respectively. In the samples they investigated, sub-micron sized h-BN platelets bounded by (0001) h-BN planes occurred as a contaminant. Most of these h-BN platelets were aligned with respect to SiC grains so that (111) 3C SiC and (0001) α-SiC planes were parallel to (0001) h-BN planes, with [1120] h-BN parallel to either [1 1 0] 3C SiC or [1120] α-SiC as appropriate (Fig. 17.5). Clear evidence of thin 12 ± 1 Å amorphous intergranular films were found at the SiC−h-BN interfaces when there was sufficient liquid available during the high temperature processing operation used to manufacture the samples (Turan and Knowles, 1996b). However, in contrast to interfaces between SiC and Si3N4, attempts to rationalise the occurrence of the dominant orientation relationships between h-BN and SiC were unsuccessful: unrealistically large misfits would have to be generated within parallel (111) 3C SiC and (0001) h-BN planes in the absence of any amorphous intergranular phase. Instead Knowles and Turan argued that the {111} 3C SiC planes act as templates upon which the h-BN basal plane meshes deposit and grow. Clear orientation relationships were also seen by Turan and Knowles (1997) for interfaces between h-BN and β-Si3N4 (Fig. 17.6). As for interfaces between h-BN and SiC, interphase boundaries were dominated by (0001)
© Woodhead Publishing Limited, 2006
480
Ceramic matrix composites
Amorphous phase
6H SiC
3C SiC
h-BN
100 nm (b)
(a)
12 ± 1 Å
12 ± 1 Å
3C
h-BN (c)
h-BN
6H (d)
17.5 (a) A h-BN particle sandwiched between 3C SiC and 6H SiC grains in a Si3N4–SiC composite made using as-received powders. (b) The diffraction pattern obtained when both SiC grains and the h-BN grain were inside the selected area aperture, showing that in the h-BN grain and 6H SiC grain the beam direction is parallel to [1120] and that in 3C SiC the beam direction is parallel to [110]. Double arrowed spots are 0001 and 1 1 03 2H SiC reflections respectively in the [1120] 2H SiC zone. Weak 1 1 01 and 1 1 03 h-BN reflections arise at the single arrowed positions. (c) The 3C SiC–h-BN interphase boundary, and (d) the h-BN–6H SiC interphase boundaries showing thin amorphous intergranular films at both interfaces (Figs 17.5(a), (c) and (d) represented from J. Am. Ceram. Sci., Turan S and Knowles KM, ‘Effect of boron nitride on the phase stability and phase transformations in silicon carbide’, 79(12), 3325–3328 (1996) with kind permission of Blackwells Publishing Ltd).
h-BN planes, but, significantly, HRTEM suggested that interfaces between h-BN and β-Si3N4 were free of amorphous intergranular films, such as the
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
481
(a) h-BN
Si3N4 4 nm (b) Si3N4
h-BN 4 nm
(c) h-BN
Si3N4 4 nm
17.6 (a) An example of HRTEM image of an interphase boundaries between h-BN particles and adjacent β-Si3N4 grains, showing that the amorphous phase at the triple junction does not seem to extend along the interphase boundary, (b) and (c) other examples of intergranular film free interphase boundaries between h-BN and βSi3N4 (reprinted from J. Eur. Ceram. Soc., Turan S and Knowles KM, ‘Interphase boundaries between hexagonal boron nitride and betasilicon nitride in silicon nitride–silicon carbide particulate composites’, 17(15/16), 1849–1854 (1997) with kind permission of Elsevier Science).
© Woodhead Publishing Limited, 2006
482
Ceramic matrix composites
example shown in Fig. 17.6(c). Knowles and Turan (2002) showed that the two orientation relationships found by Turan and Knowles, (i) [11 2 0] h-BN || [ 1 2 1 3] β-Si3N4 and (0001) h-BN || (10 1 0) β-Si3N4, and (ii) [1120] h-BN || [0001] β-Si3N4 with (0001) h-BN 3.5–4° from (10 1 0) β-Si3N4, could both be rationalised in terms of low misfits on the basis of the near-coincident site lattice approach. Thus, overall, there is strong evidence that non-random orientation relationships can be adopted in ceramic composites between dissimilar phases. This is most likely when one phase is formed inside the other during hightemperature processing. If there is evidence of remnant intergranular liquid phase at the interface between the two phases, such as at h-BN− SiC interfaces, there is no reason to expect that the interface is of low energy, but that another explanation is more appropriate, such as the templating explanation offered by Knowles and Turan (2002). However, if the interfaces are free of intergranular film, or if the intergranular film coverage is relatively incomplete, then the analyses of Turan and Knowles (2000) and Knowles and Turan (2002) show that the observed orientation relationships do correlate with those expected on the basis of relatively low misfits.
17.7
Future trends
The importance that interfaces have in determining structural and functional properties of engineering ceramics and composites and the trend towards nanostructured materials, and thus materials with large areas of grain boundary per unit volume, will both mean that interfaces will continue to be the object of intensive research interest. Ultimately, knowledge of the atomistic and electronic properties of interfaces in non-oxide engineering ceramics and composites should lead to the ability to design and control interfacial structure, leading to improvements in mechanical properties such as creep, strength and fracture toughness. Substantial progress has already been made experimentally and through the use of continuum models in understanding equilibrium film thicknesses of silica-rich films at grain boundaries in engineering ceramics such as Si3N4 and SiC. Progress is also beginning to be made in atomistic modelling of such films, as the recent work of Garofalini and co-workers (Garofalini and Luo, 2003; Su and Garofalini, 2004a, 2004b) has shown. Experimentally, the technique of measuring Hamaker constants in situ using spatially resolved-valence electron energy loss (SR-VEEL) spectroscopy in the STEM with a 0.6 nm probe (French et al., 1998) represents an important new tool for dispersion force and wetting studies (French, 2000). In addition, local variations in intergranular film chemistry and the dispersion forces throughout the microstructure of individual grains can now be determined using these methods. Other electron microscope-based techniques which are
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
483
being developed, and either are already being used or have the potential for future use in the study of interfaces in ceramics, include Z-contrast imaging (Shibata et al., 2004; Ziegler et al., 2004), Fourier filtering to remove the lattice fringes from the image and enhance the visibility of intergranular films (Maclaren, 2004), electron diffraction with a convergent probe focused on the intergranular film to obtain information about the local atomic structure of the film (Doblinger et al., 2004), focus-variation phase-reconstruction methods in HRTEM to image the atomic structure of interfaces at subÅngstrom resolution (Ziegler et al., 2003) and electron tomography to obtain three-dimensional structures (Midgley and Weyland, 2003; Weyland and Midgley, 2004). In addition, advances in atomic force microscopy now enable this technique to image surfaces of conductors and insulators in vacuum at atomic resolution (Giessibl 2003). Each of these techniques will be able to make significant contributions to our understanding of interfaces in engineering ceramics and composites. Modelling studies on interfaces have already made significant advances with the ability to undertake molecular dynamics studies (such as the work of Garofalini et al. referred to above) and first-principles calculations (see, for example, Painter et al., 2004). Progress is now being made in understanding the degree to which intergranular films are ordered and how cations introduced into the silica-rich films position themselves within the film. Painter et al. (2004) have shown recently that the origin of grain growth anisotropy in silicon nitride ceramics doped with rare earth elements lies in the site competition between the rare-earth cations and silicon for bonding at β-Si3N4 interfaces and within the intergranular film. We can reasonably surmise that there will be further advances in our understanding from such computer modelling. An area in which progress is still to be made is that of examining structures and absolute energies of interfaces between covalently bonded crystalline phases, such as between grains of silicon nitride or silicon carbide devoid of any intergranular film, in sharp contrast to interfaces in metallic materials and ionic solids. Such progress will help in the understanding of the crystallography of interphase boundaries in engineering ceramics and composites discussed in Section 17.6. Overall, therefore, we can confidently predict that the need for detailed knowledge about the behaviour of interfaces in non-oxide ceramics will lead to advances in a number of areas both experimentally and computationally. The relationship between macroscopic continuum models of the wetting behaviour of intergranular films and the forces keeping intergranular films present at interfaces, and the atomic-level picture gained from high resolution electron microscope studies and computer simulations, could be a particularly fascinating topic – it is apparent that there is substantial scope for progress in this area.
© Woodhead Publishing Limited, 2006
484
17.8
Ceramic matrix composites
Further reading
The references in this section are not meant to be an exhaustive list of further research papers and books relevant to the work we have described in this chapter. Instead, they are a representative selection of material relevant to the work we have described, but have been unable to reference in the text, because of the need for conciseness and brevity. Ahn, C.C., (2004), Transmission Electron Energy Loss Spectrometry in Materials Science and the EELS Atlas, New York, John Wiley & Sons. Angelescu, D.E., Harrison, C.K., Trawick, M.L., Chaikin, P.M., Register, R.A. and Adamson, D.H., (2004), ‘Orientation imaging microscopy in two-dimensional crystals via undersampled microscopy’, Appl. Phys. A – Materials Science and Processing, 78 (3): 387–392. Belousov, V.V., (2004), ‘Wetting of grain boundaries in ceramic materials’, Colloid Journal, 66 (2): 121–127. Bonnet, R., Cousineau, E. and Warrington, D.H., (1981), ‘Determination of near-coincident cells for hexagonal crystals. Related DSC lattices’, Acta Crystallogr. A, 37 (2), 184– 189. Buseck, P.R., Cowley, J.M. and Eyring, L., (1988), High-Resolution Transmission Electron Microscopy and Associated Techniques, Oxford, Oxford University Press. Cahn, J.W., (1977), ‘Critical point wetting’, J. Chem. Phys, 66 (8), 3667–3772. Cannon, R.M., Rühle, M., Hoffmann, M.J., French, R.H., Gu, H., Tomsia, A.P. and Saiz, E., (2000), ‘Adsorption and wetting mechanisms at ceramic grain boundaries’, Ceramic Transactions, 118, 427–444. Cinibulk, M.K. and Kleebe, H.-J., (1993), ‘Effects of oxidation on intergranular phases in silicon nitride ceramics’, J. Mater. Sci., 28 (21), 5775–5782. Clarke, D.R., (1979), ‘High resolution techniques and application to nonoxide ceramics’, J. Am. Ceram. Soc., 62 (5/6), 236–246. Clarke, D.R., (1987), ‘Grain boundaries in polycrystalline ceramics’, Ann. Rev. Mat. Sci., 17, 57–74. Clarke, D.R., (1989), ‘High-temperature microstructure of a hot-pressed silicon nitride’, J. Am. Ceram. Soc., 72 (9), 1604–1609. Clarke, D.R., (1990), ‘Perspectives concerning grain boundaries in ceramics’ Am. Ceram. Soc. Bull., 69 (4), 682–685. Clarke, D.R. and Gee, M.L., (1992), ‘Wetting of surfaces and grain boundaries’, in Wolf, D. and Yip, S., (eds), Materials Interfaces, London, Chapman & Hall, 255–272. Dietrich, S., (1991), ‘Fluid interfaces − wetting, critical adsorption, van der Waals tails, and the concept of the effective interface potential’, in Taub, H., Torzo, G., Lauter, H.J. and Fain, S.C., (eds), Phase. Transitions in Surface Films 2, NATO Advanced Science Series, Physics, Vol. 267, 391–423. Falk, L.K.L., (1998), ‘Electron spectroscopic imaging and fine probe EDX analysis of liquid phase sintered ceramics’, J. Eur. Ceram. Soc., 18 (15), 2263–2279. Flewitt, P.E.J. and Wild, R.K., (2004), Grain Boundaries: Their Microstructure and Chemistry, Chichester, John Wiley & Sons. Forwood, C.T. and Clarebrough, L.M., (1991), Electron Microscopy of Interfaces in Metals and Alloys, Bristol and Philadelphia, Institute of Physics Publishing. French, R.H., Cannon, R.M., DeNoyer, L.K. and Chiang, Y.-M., (1995), ‘Full spectral calculation of non-retarded Hamaker constants for ceramic systems from interband transition strengths’, Solid State Ionics, 75, 13–33. Gu, H., (2002), ‘Variation of width and composition of grain-boundary film in a highpurity silicon nitride with minimal silica’, J. Am. Ceram. Soc., 85 (1), 33–37.
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
485
Gu, H., (2004), ‘Electron energy-loss spectroscopy characterization of ~1 nm-thick amorphous film at grain boundaries in Si-based ceramics’, Mater. Trans., 45 (7), 2091−2098. Gu, H., (2004), ‘Evolution of intergranular boundaries and phases in SiC and Si3N4 ceramics under high temperature deformation: case studies by analytical TEM’, Zeitschrift für Metallkunde, 95 (4), 271−274. Jin, Q., Wilkinson, D.S. and Weatherly, G.C., (1999), ‘High-resolution electron microscopy investigation of viscous flow creep in a high-purity silicon nitride’, J. Am. Ceram. Soc., 82 (6), 1492–1496. Keyse, R.J., Goodhew, P.J., Garrattt-Reed, A.J. and Lorimer, G.W., (1998), Introduction to Scanning Transmission Electron Microscopy, Bios Scientific Publishers. Kleebe, H.-J. and Cinibulk, M.K., (1993), ‘Transmission electron microscopy characterization of a ceria-fluxed silicon nitride’, J. Mater. Sci. Lett., 12 (2), 70–72. Kleebe, H.-J., Corbin, N., Willkens, C. and Rühle, M., (1990), ‘Transmission electron microscopy studies of silicon nitride/silicon carbide interfaces’, Mat. Res. Soc. Symp. Proc., 170, 79−84. Kleebe, H.-J., Cinibulk, M.K., Cannon, R.M. and Rühle, M., (1993), ‘Statistical analysis of the intergranular film thickness in silicon nitride ceramics’, J. Am. Ceram. Soc., 76 (8), 1969–1977. Knowles, K.M., Turan, S., Kumar, A., Chen, S.H. and Clegg, W.C., (1999), ‘Transmission electron microscopy of interfaces in structural ceramic composites’, J. Microscopy, 196(2) 194–202. Krivanek, O.L., Shaw, T.M. and Thomas, G., (1979), ‘Imaging of thin intergranular phases by high-resolution electron microscopy’, J. Appl. Phys., 50 (6), 4223−4227. Lee, W.E. and Rainforth, W.M., (1994), Ceramic Microstructures: Property Control by Processing, London, Chapman & Hall. Luo, J. and Chiang, Y-M., (2000), ‘Stabilization of surface films in ceramics’, Ceramic Transactions, 118, 419–426. Luo, J. and Chiang, Y-M., (2000), ‘Existence and stability of nanometer-thick disordered films on oxide surfaces’, Acta Mater., 48 (18/19), 4501–4515. More, K.L., Koester, D.A. and Davis, R.F., (1991), ‘Microstructural characterization of a creep-deformed SiC whisker-reinforced Si3N4’, Ultramicroscopy, 37, 263−278. Ness, J.N., Stobbs, W.M. and Page, T.F., (1986), ‘A TEM Fresnel diffraction-based method for characterizing thin grain-boundary and interfacial films’, Phil. Mag. A, 54 (5), 679−702. Nishimura, T., Mitomo, M., Ishida, A. and Gu, H., (2000), ‘Improvement of high temperature strength and creep of alpha-sialon by grain boundary crystallization’, Key Engineering Materials, 171, 741−746. Norton, M.G. and Carter, C.B., (1990), ‘Interfaces in structural ceramics’, MRS Bulletin, 15 (10), 51−59. Norton, M.G. and Carter, C.B., (1990), ‘Grain and interphase boundaries in ceramics and ceramic composites’, in Wolf, D. and Yip, S., (eds), Materials Interfaces, London, Chapman & Hall, 151−189. Pan, X., Gu, H., van Weeren, R., Danforth, S.C., Cannon, R.M. and Rühle, M., (1996), ‘Grain-boundary microstructure and chemistry of a hot isostatically pressed highpurity silicon nitride’, J. Am. Ceram. Soc, 79 (9), 2313–2320. Pandit, R., Schick, M., and Wortis, M., (1982), ‘Systematics of multilayer adsorption phenomena on attractive substrates’, Phys. Rev. B, 26 (9), 5112–5140. Pezzotti, G. and Ota, K., (1997), ‘Grain-boundary sliding in fluorine-doped silicon nitride’, J. Am. Ceram. Soc., 80 (3), 599–603.
© Woodhead Publishing Limited, 2006
486
Ceramic matrix composites
Pezzotti, G., Tanaka, I. and Nishida, T., (1993), ‘Intrinsic fracture energy of polycrystalline silicon nitride’, Phil. Mag. Lett., 67 (2), 95–100. Pezzotti, G., Ota, K., Kleebe, H.-J., Okamoto, Y. and Nishida, T., (1995), ‘Viscous behavior of interfaces in fluorine-doped Si3N4/SiC composites’, Acta. Metall. Mater., 43 (12), 4357–4370. Proceedings of the International Congresses on Intergranular and Interphase Boundaries in Materials. Reimer, L., (1997), Transmission Electron Microscopy: Physics of Image Formation and Microanalysis, 4th edn, Berlin and Heidelberg, Springer-Verlag. Reimer, L., Zepke, U., Moesch, J., Schulze-Hillert, S., Ross-Messemer, M., Probst, W. and Weimer, E., (1992), EELSpectroscopy, Oberkochen, Germany, Carl Zeiss. Rühle, M., (1984), ‘TEM observations of grain boundaries in ceramics’, Mat. Res. Soc. Symp. Proc., 31, 317−323. Sakuma, T., Sheppard, L.M., and Ikuhara, Y., (eds) (2000), Grain Boundary Engineering in Ceramics: From Grain Boundary Phenomena to Grain Boundary Quantum Structures, Ceramic Transactions, 118, Westerville, OH, The American Ceramic Society. Schick, M., (1990), ‘Introduction to wetting phenomena’, in Charvolin, J., Joanny, J.F. and Zinn-Justin, J., Liquids at Interfaces, Les Houches Summer School Lectures, Session XLVIII, Amsterdam, Elsevier, 415–497. Sigl, L.S. and Kleebe, H.-J., (1993), ‘Core/rim structure of liquid-phase-sintered silicon carbide’, J. Am. Ceram. Soc., 76 (3), 773–776. Simpson, Y.K., Carter, C.B., Morrissey, K.J., Angelini, P. and Bentley, J. (1986), ‘The identification of thin amorphous films at grain boundaries in alumina’, J. Mater. Sci., 21 (8), 2689−2696. Smart, R., St C., and Nowotny, J., (1998), Ceramic Interfaces: Properties and Applications, London, Maney Publishing. Spence, J.C.H., (1988), Experimental High Resolution Electron Microscopy, 2nd edn, Oxford, Oxford University Press. Susnitzky, D.W. and Carter, C.B., (1990), ‘Structure of alumina grain boundaries prepared with and without a thin amorphous intergranular film’, J. Am. Ceram. Soc., 73 (8), 2485−2493. Tanaka, I., (2001), ‘Intergranular glassy films in ceramics, J. Ceram. Soc. Japan, 109 (8), S127−S134. Uematsu, K., Moriyoshi, Y., Saito, Y. and Nowotny, J., (eds) (1995), Interfaces of Ceramic Materials: Impact on Properties and Applications, Key Engineering Materials, 111– 112, Uetikon-Zurich, Trans Tech Publications. Wang, C-M., Pan, X., Hoffmann, M.J., Cannon, R.M. and Rühle, M., (1996), ‘Grain boundary films in rare-earth-glass-based silicon nitride’, J. Am. Ceram. Soc., 79 (3), 788–792. Williams, D.B. and Carter, C.B., (1996), Transmission Electron Microscopy, New York, Plenum Press. Wolf, D., (1992), ‘Atomic-level geometry of crystalline interfaces’, in Wolf, D. and Yip, S. (eds), Materials Interfaces, London, Chapman & Hall, 1−57. Zhang, X.F. and De, Jonghe, L.C., (2003), ‘Thermal modification of microstructures and grain boundaries in silicon carbide’, J. Mater. Res., 18 (12), 2807−2813.
17.9
References
Becher, P.F., Sun, E.Y., Plucknett, K.P., Alexander, K.B., Hsueh C.H., Lin, H.T., Waters, S.B., Westmoreland, C.G., Kang, E.S., Hirao, K. and Brito, M.E., (1998), ‘Microstructural
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
487
design of silicon nitride with improved fracture toughness: I, effects of grain shape and size’, J. Am. Ceram. Soc., 81 (11), 2821–2830. Cahn, J.W., and Kalonji, G., (1982), ‘Symmetry in solid state transformation morphologies’, in Aaronson, H.I., Laughlin, D.E., Sekerka, R.F. and Wayman, C.M., (eds), Proceedings of an International Conference on Solid → Solid Phase Transformations, Warrendale, PA., The Metallurgical Society of AIME, 3–14. Choi, H-J., Kim, Y.W., Mitomo, M., Nishimura, T., Lee, J-H. and Kim, D-Y., (2004), ‘Intergranular glassy phase free SiC ceramics retains strength at 1500°C’, Scripta Mater, 50 (9), 1203–1207. Cinibulk, M.K., Kleebe, H.-J. and Rühle, M., (1993a), ‘Quantitative comparison of TEM techniques for determining amorphous intergranular film thickness’, J. Am. Ceram. Soc., 76 (2), 426–432. Cinibulk, M.K., Kleebe, H.-J., Schneider, G.A. and Rühle, M., (1993b), ‘Amorphous intergranular films in silicon nitride ceramics quenched from high temperatures’, J. Am. Ceram. Soc., 76 (11), 2801–2808. Clarke, D.R., (1979), ‘On the detection of thin intergranular films by electron microscopy’, Ultramicroscopy, 4 (1), 33–44. Clarke, D.R., (1985), ‘Grain boundaries in polyphase ceramics’, J. de Physique, Coll. C4, 46, C4-51–C4-59. Clarke, D.R., (1987), ‘On the equilibrium film thickness of intergranular glassy phases in ceramic materials’, J. Am. Ceram. Soc., 70 (1), 15–22. Clarke, D.R., Shaw, T.M., Philipse, A. and Horn, R.G., (1993), ‘Possible electrical doublelayer contribution to the equilibrium thickness of intergranular glassy phases in polycrystalline ceramics’, J. Am. Ceram. Soc., 76 (5), 1201–1204. Doblinger, M., Marsh, C.D., Nguyen-Manh, D., Ozkaya, D. and Cockayne, D.J.H., (2004), ‘Intergranular films in Si3N4 studied by TEM’, Inst. Phys. Conf. Ser., 179, 401–404. Finnis, M.W. and Rühle, M., (1993), ‘Structures of interfaces in crystalline solids’, in Cahn, R.W., Haasen, P. and Kramer, E.J., (eds), Materials Science and Technology: A Comprehensive Treatment, Volume 1: Structure of Solids, New York, VCH Publishers, 533–605. French, R.H., (2000), ‘Origins and applications of London dispersion forces and Hamaker constants in ceramics’, J. Am. Ceram. Soc., 83 (9), 2117–2146. French, R.H., Müllejans, H., Jones, D.J., Duscher, G., Cannon, R.M. and Rühle, M. (1998), ‘Dispersion forces and Hamaker constants for intergranular films in silicon nitride from spatially resolved-valence electron energy loss spectrum imaging’, Acta Mater., 46 (7), 2271–2287. Garofalini, S.H., and Luo, W.W., (2003), ‘Molecular dynamics simulations of calcium silicate intergranular films between silicon nitride crystals’, J. Am. Ceram. Soc., 86 (10), 1741–1752. Giessibl, F.J., (2003), ‘Advances in atomic force microscopy’, Rev. Mod. Phys., 75 (3), 949–983. Gu, H. and Shinoda, Y., (2000), ‘Structural and chemical widths of general grain boundaries: modification of local structure and bonding by boron-doping in β-silicon carbide’, Interface Science, 8 (2/3), 269–278. Gu, H., Ceh, M., Stemmer, S., Müllejans, H. and Rühle, M., (1995), ‘A quantitative approach for spatially resolved electron-energy-loss spectroscopy of grain boundaries and planar defects on a subnanometer scale’, Ultramicroscopy, 59 (1/4), 215–227. Horn, R.G. and Israelachvili, J.N., (1981), ‘Direct measurement of structural forces between two surfaces in a nonpolar liquid’, J. Chem. Phys., 75 (3), 1400–1411.
© Woodhead Publishing Limited, 2006
488
Ceramic matrix composites
Israelachvili, J.N., (1992), Intermolecular and Surface Forces, 2nd edn, London, Academic Press. Kleebe, H.-J., (1997), ‘Structure and chemistry of interfaces in Si3N4 ceramics studied by transmission electron microscopy’, J. Ceram. Soc. Japan, 105 (6), 453–475. Kleebe, H.-J., (2002). ‘Comparison between SEM and TEM imaging techniques to determine grain-boundary wetting in ceramic polycrystals’, J. Am. Ceram. Soc., 85 (1), 43–48. Kleebe, H.-J. and Pezzotti, G., (2002), ‘Grain-boundary wetting–dewetting in z = 1 SiAlON ceramic’, J. Am. Ceram. Soc., 85 (12), 3049–3053. Kleebe, H.J., Hoffmann, M.J. and Rühle, M., (1992), ‘Influence of secondary phase chemistry on grain boundary film thickness in silicon nitride’, Z. Metallk, 83 (8), 610– 617. Knowles, K.M., (2005), ‘Dispersion forces at planar interfaces in anisotropic ceramics’, J. Ceram. Process. Res., in press. Knowles, K.M. and Turan, S., (1996), ‘Atomic structure of a Σ = 9 <011> silicon nitride tilt grain boundary’, Mater. Sci. Forum, 207-209, 353–357. Knowles, K.M. and Turan, S., (2000), ‘The dependence of equilibrium film thickness on grain orientation at interphase boundaries in ceramic–ceramic composites’, Ultramicroscopy, 83 (3/4), 245–259. Knowles, K.M. and Turan, S., (2002), ‘Boron nitride−silicon carbide interphase boundaries in silicon nitride−silicon carbide particulate composites’, J. Eur. Ceram. Soc., 22 (9/ 10), 1587–1600. Kumar, A. and Knowles, K.M., (1996a), ‘Microstructure–property relationships of SiC fibre-reinforced magnesium aluminosilicates – I. Microstructural characterisation’, Acta Mater., 44 (7), 2901–2921. Kumar, A. and Knowles, K.M., (1996b), ‘Oxidation behaviour of a Si–C–O–fiber-reinforced magnesium aluminosilicate’ J. Am. Ceram. Soc., 79 (9), 2364–2374. Lewis, M.H., (1989) Glasses and Glass-Ceramics, New York, Chapman and Hall. MacLaren, I., (2004), ‘Imaging and thickness measurement of amorphous intergranular films using TEM’, Ultramicroscopy, 99 (2/3), 103–113. Mandal, H. and Thompson, D.P., (2000), ‘New heat treatment methods for glass removal from silicon nitride and sialon ceramics’, J. Mater. Sci., 35 (24), 6285–6292. Marion, J.E., Hsueh, C.H. and Evans, A.G., (1987), ‘Liquid phase sintering of ceramics’, J. Am. Ceram. Soc., 70 (10), 708–713. Midgley, P.A. and Weyland, M., (2003), ‘3D electron microscopy in the physical sciences: the development of Z-contrast and EFTEM tomography’, Ultramicroscopy, 96 (3/4), 413–431. Moberlychan, W.J., Cao, J.J. and De Jonghe, L.C., (1998), ‘The roles of amorphous grain boundaries and the β–α transformation in toughening SiC’, Acta Mater., 46 (5), 1625– 1635. Niihara, K., Suganuma, K., Nakahira, A. and Izaki, K., (1990), ‘Interfaces in Si3N4–SiC nano-composite’, J. Mater. Sci. Lett., 9 (5), 598–599. Painter, G.S., Becher, P.F., Shelton, W.A., Satet, R.L. and Hoffmann, M.J., (2004), ‘Firstprinciples study of rare-earth effects on grain growth and microstructure in β-Si3N4 ceramics’, Phys. Rev. B, 70, article 144108. Pan, X., Mayer, J., Rühle, M. and Niihara, K., (1996a), ‘Silicon nitride based ceramic nanocomposites’, J. Am. Ceram. Soc., 79 (3), 585–590. Pan, X., Rühle, M. and Niihara, K., (1996b), ‘Microstructure of grain boundaries and phase boundaries in Si3N4–SiC nanocomposites’, Mater. Sci. Forum., 207-209, 761– 764.
© Woodhead Publishing Limited, 2006
Interfaces in non-oxide ceramic composites
489
Parsegian, V.A. and Weiss, G.H., (1972), ‘Dielectric anisotropy and the van der Waals interaction between bulk media’, J. Adhesion, 3 (4), 259–267. Pezzotti, G., (1993), ‘Si3N4/SiC-platelet composite without sintering aids – a candidate for gas-turbine engines’, J. Am. Ceram. Soc., 76 (5), 1313–1320. Prieve, D.C. and Russel, W.B., (1988), ‘Simplified predictions of Hamaker constants from Lifshitz theory’, J. Colloid and Interface Science, 125 (1), 1–13. Rühle, M., Bischoff, E. and David, O., (1984), ‘Structure of grain boundaries in ceramics’, Ultramicroscopy, 14 (1/2), 37–46. Schmid, H. and Rühle, M., (1984), ‘Structure of special grain boundaries in Sialon ceramics’, J. Mater. Sci., 19 (2), 615–628. Shibata, N., Pennycook, S.J., Gosnell, T.R., Painter, G.S., Shelton, W.A. and Becher, P.F., (2004), ‘Observation of rare-earth segregation in silicon nitride ceramics at subnanometre dimensions’, Nature, 428 (6984), 730–733. Su, X. and Garofalini, S.H., (2004a), ‘Atomistic structure of calcium silicate intergranular films between prism and basal planes in silicon nitride: a molecular dynamics study’, J. Mater. Res., 19 (3), 752–758. Su, X. and Garofalini, S.H., (2004b), ‘Role of nitrogen on the atomistic structure of the intergranular film in silicon nitride: a molecular dynamics study’, J. Mater. Res., 19 (12), 3679–3687. Sun, E.Y., Becher, P.F., Plucknett, K.P., Hsueh, C.H., Alexander, K.B., Waters, S.B., Hirao, K. and Brito, M.E., (1998), ‘Microstructural design of silicon nitride with improved fracture toughness: II effects of yttria and alumina additives’, J. Am. Ceram. Soc., 81 (11), 2831–2840. Sutton, A.P. and Balluffi, R.W., (1987), ‘On geometric criteria for low interfacial energy’, Acta Metall., 35 (9), 2177–2201. Sutton, A.P. and Balluffi, R.W., (1995), Interfaces in Crystalline Materials. Oxford, Oxford Science Publications, Clarendon Press. Tanaka, I., Kleebe, H.-J., Cinibulk, M.K., Bruley, J., Clarke, D.R. and Rühle, M., (1994), ‘Calcium-concentration dependence of the intergranular film thickness in siliconnitride’, J. Am. Ceram. Soc., 77 (4), 911–914. Turan, S., (1995), ‘Microstructural characterisation of silicon nitride–silicon carbide particulate composites’, Ph.D. thesis, University of Cambridge. Turan, S. and Knowles, K.M., (1995), ‘A comparison of the microstructure of silicon nitride–silicon carbide composites made with and without deoxidized starting material’, J. Microscopy, 177 (3), 287–304. Turan, S. and Knowles, K.M., (1996a), ‘α → β reverse phase transformation in silicon carbide in silicon nitride–particulate-reinforced-silicon carbide composites’, J. Am. Ceram. Soc., 79 (11), 2892–2896. Turan, S. and Knowles, K.M., (1996b), ‘Effect of boron nitride on the phase stability and phase transformations in silicon carbide’, J. Am. Ceram. Soc., 79 (12), 3325–3328. Turan, S. and Knowles, K.M., (1997), ‘Interphase boundaries between hexagonal boron nitride and beta silicon nitride in silicon nitride–silicon carbide particulate composites’, J. Eur. Ceram. Soc., 17 (15/16), 1849–1854. Turan, S. and Knowles, K.M., (1999), ‘Wetting and non-wetting behaviour of SiC grain boundaries’, Mater. Sci. Forum., 294-296, 313–317. Turan, S. and Knowles, K.M., (2000), ‘The crystallography of interphase boundaries between silicon carbide and silicon nitride in silicon nitride–silicon carbide particulate composites’, Interface Science, 8 (2/3), 279–294. Ueno, F. and Horiguchi, A., (1989), ‘Grain boundary phase elimination and microstructure
© Woodhead Publishing Limited, 2006
490
Ceramic matrix composites
of aluminium nitride’, in de With G., Terpstra, A. and Metselaar, R., (eds), EuroCeramics. Volume I: Processing of Ceramics (Proceedings of the First European Ceramic Society Conference, Maastricht, The Netherlands, 18–23 June 1989), New York, Elsevier Applied Science, 383–387. Unal, O. and Mitchell, T.E., (1992), ‘CVD Si3N4 on single crystal SiC: Part II. High resolution electron microscopy and atomic models of the interface’, J. Mater. Res., 7 (6), 1445–1454. Unal, O., Petrovic, J.J. and Mitchell, T.E., (1992), ‘CVD Si3N4 on single crystal SiC: Part I. Characterization and orientation relationship at the interface’, J. Mater . Res., 7 (1), 136–147. Vetrano, J.S., Kleebe, H.-J., Hampp, E., Hoffmann, M.J., Rühle, M. and Cannon, R.M., (1993), ‘Yb2O3-fluxed sintered silicon nitride: Part I. Microstructure characterization’, J. Mater. Sci., 28 (13), 3529–3538. Weyland, M. and Midgley, P.A., (2004), ‘Electron tomography’, Materials Today, 7 (12), 32–40. Ziegler, A., Kisielowski, C., Hoffmann, M.J. and Ritchie, R.O., (2003), ‘Atomic resolution transmission electron microscopy of the intergranular structure of a Y2O3-containing silicon nitride ceramic’, J. Am. Ceram. Soc., 86 (10), 1777–1785. Ziegler, A., Idrobo, J.C., Cinibulk, M.K., Kisielowski, C., Browning, N.D. and Ritchie, R.O., (2004), ‘Interface structure and atomic bonding characteristics in silicon nitride ceramics’, Science, 306 (5702), 1768–1770.
© Woodhead Publishing Limited, 2006
18 Sialons Z B Y U, Queen’s University, Canada and D P T H O M P S O N, University of Newcastle, UK
18.1
Introduction
Sialons (Oyama and Kamigaito, 1971; Jack and Wilson, 1972; Hampshire et al., 1978) are a family of advanced structural ceramics that exhibit a good combination of properties such as high strength at elevated temperatures, dimensional stability, good corrosion and erosion behaviour, high elastic modulus, low mass density, good thermal shock resistance and high hardness. These combined properties make them useful as structural materials. However, as with other structural ceramics, they suffer from the disadvantage of low fracture toughness (3.0–8.0 MPa m1/2) with no R-curve behaviour (Cao and Metselaar, 1991; Shen et al., 1996a,b,c). Although they can be used for engineering components, their reliability is still not adequate in many cases for industry to commit to production. The most attractive and effective way of increasing toughness of sialons is through the use of a second-phase reinforcement, thus generating a sialon matrix composite. In composite form, significant improvements can be achieved such as increased fracture toughness, less strength variability, less flaw sensitivity, reduced crack propagation and better reliability; and even more significantly, the failure manner of sialon composites can be changed and controlled. Recently, several new techniques have been developed, such as α ⇔ β sialon transformation (Mandal et a1., 1993; Thompson, 1994; Yu et al., 1998b) and the retention of elongated αor β-sialon grains in dense α/β-sialon composites (Shen et al., 1996a; Chen and Rosenflanz, 1997; Yu et al., 2001a, b). Attempts have also been made by Nordberg et al. (1993, 1997a; Nordberg and Ekström, 1995) to reinforce Yα-sialon ceramics by incorporating a second phase, e.g. MoSi2 particles or SiC-whiskers. More recently, continuous fibre reinforcement of sialons has been explored and this has proved to be an effective way of overcoming some of the deficiencies of sialon materials. Research by Zhang and Thompson (1995, 1997), Demir and Thompson (2001) and Yu and Thompson (1998), Yu et al., 2002a, 2002b) has given very encouraging results using highperformance carbon fibres and Nicalon SiC fibres for reinforcing oxynitride glass, α- and β-sialons. 491 © Woodhead Publishing Limited, 2006
492
Ceramic matrix composites
Section 18.2 gives a very brief introduction to sialons. Section 18.3 outlines the challenges to be overcome in order to make toughened and strengthened sialon products. Progress in developing sialon composites forms the main part of this chapter and is discussed in section 18.4, which deals with α/βsialon composites, particle/whisker-reinforced sialons, and fibre-reinforced sialons. In the final section, several conclusions and suggestions for future work are summarised.
18.2
Sialons
Sialon is the generic name for the large family of silicon nitride solid solutions containing the basic elements Si, Al, O and N. Over the last three decades, the matrix sialon phases (α-, β-, and O-) have been developed and various excellent reviews are available (Cao and Metselaar, 1991; Ekström and Nygren, 1992; Izhevskiy et al., 2000). In the sialon family, α- and β-sialons offer most interest as engineering ceramics because of their excellent combination of mechanical and high-temperature properties.
18.2.1 β-Sialon β-Sialon ceramics were concurrently discovered by Jack and Wilson (1972) in the UK and by Oyama and Kamigaito (1971) in Japan, who reported that up to two-thirds of the Si and N in β-Si3N4 could be substituted by Al and O without change of structure to form a Si6–zAlzOzN8–z solid solution with the β-Si3N4 structure, where the z-value shows the degree of solid solubility and varies continuously from zero to about 4.2 (Gauckler et al., 1975). β-Sialon ceramics dominated the early interest because they could be pressurelessly sintered into complex shapes, and the resulting materials had a good combination of properties; for example, some pressurelessly sintered β-sialon ceramics have strengths of up to 1000 MPa and fracture toughnesses of up to 8 MPa m1/2 (Jack, 1976).
18.2.2 α-Sialon α-Sialon is a sialon derivative of α-Si3N4, with the general composition MxSi12–(m+n)Alm+nOnN16–n, where, relative to the α unit cell contents of Si12N16, m (Si–N) bonds are replaced by m (Al–N) bonds and n (Si–N) bonds by n (Al–O) bonds (Ekström, 1992a,b). The charge discrepancy caused by the replacement mechanism is compensated by the introduction of the metal ion M, with m = vx, where v is the metal ion valency. Because of the particular atomic arrangement in α-Si3N4, the unit cell, containing Si12N16, has two interstitial sites large enough to accommodate the M atoms, and therefore x ≤ 2. This offers the possibility of reducing the amount of residual glassy
© Woodhead Publishing Limited, 2006
Sialons
493
phase if sintering additives are used which can subsequently be incorporated into the α-sialon structure. The existence of metal-doped α-sialons was demonstrated by Hampshire et al. (1978), even though the technological importance of α-sialon compared with β-sialon was not appreciated at that time. Their main advantage is that fully dense α-sialon ceramics have an excellent hardness in excess of ~20 GPa (Nordberg et al., 1993; Shen et al., 1996a, d, e; Yu et al., 2001b).
18.2.3 α/β-Sialon and transformation Dense β-sialon ceramics with low z values (z ~ 1) have microstructures consisting of elongated crystals and grain boundary glass, and of all the various sialon materials, show the highest observed fracture toughness values at room temperature (Ekström, 1989); however, their hardness is relatively low (14–15 GPa) partly because of the residual grain boundary glass (~10 GPa). In contrast, α-sialon ceramics generally have equiaxed grains and show an excellent hardness (~22 GPa), much higher than that of any other sialon ceramics. The properties of mixed α/β-sialon ceramics can be optimised by adjusting the proportion of the two phases, and this can be done by changing the starting composition (Ekström, 1989). A high content of βsialon yields high strength and toughness, whereas a high proportion of the α-sialon phase gives excellent hardness. Mixed α/β-sialon ceramics can be produced with very little intergranular glass and so their wear resistance, strength and especially creep resistance at high temperatures are greatly improved. More recently, it has been demonstrated by Mandal et al. (1993), Thompson (1994), and Yu et al. (2000) that some densified (α + β)-sialons can undergo in situ reversible α ⇔ β sialon transformation, merely by heat treatment at temperatures below the sintering temperature. Careful control of the heat treatment schedule enables predetermined values of hardness, strength, and toughness to be achieved in the final (α + β) composite.
18.3
Challenges in toughening and strengthening sialons
As with other ceramic composites, the combination of α- and/or β-sialon with reinforcement agents results in sialon composites. This simple and obvious statement encompasses many factors which must be taken into account for successfully fabricating composites with a designed microstructure and improved properties (Prewo, 1989). For sialon matrix composites, the most important factors are physical compatibility including Young’s modulus, elastic strain (Kerans and Parthasarathy, 1991) and thermal expansion coefficient (Sambell et al., 1972a, b), and chemical compatibility between sialon matrix
© Woodhead Publishing Limited, 2006
494
Ceramic matrix composites
and reinforcement agents; the special requirements for processing and fabricating sialon composites are also a challenge. As is well known, the development of α- and/or β-sialon composites has not proceeded quickly, even though some of the earliest ceramic matrix composites investigated were based on a Si3N4 matrix (Guo et al., 1982). The difficulties in achieving a fully dense sialon composite at relative low temperatures are greater than previously encountered for either glass or glass– ceramic materials because (1) the formation temperature of sialons is approximately 1500°C (Kuang et al., 1990; Yu and Thompson, 1998); (2) the densification temperature of sialons is higher than 1650°C (Cao and Metselaar, 1991); and (3) liquid phases formed during the formation and densification of sialons are substantially accommodated into the sialon structure as sintering proceeds, especially for α-sialons, thereby making final densification difficult. Therefore, formation and densification of sialons are not easy, and high densities cannot be easily achieved at low temperatures, short sintering times, and under pressureless sintering conditions; Furthermore, the incorporation of reinforcement agents, especially fibres, into a sialon matrix makes the densification even more difficult, and the high hot-pressing temperatures and pressures needed inevitably tend to create a very strong bond between the sialon matrix and the reinforcement agents or to degrade the reinforcement agents. Therefore, all these factors make it exceptionally difficult to achieve a fully dense sialon composite.
18.4
Sialon composites
Over the last decade, considerable efforts have been committed to the toughening of sialons and substantial progress has been achieved using various reinforcements. According to the form of reinforcement, sialon composites can be classified as either particle reinforced, discontinuous fibre (whiskers/ short fibres) reinforced, or continuous fibre reinforced.
18.4.1 α / β-Sialon composites Toughening mechanisms in α ⁄β-sialon composites are similar to those operative in second-phase particle reinforced composites, but, rather than the deliberate addition of a second phase, α⁄β-sialon composites are fabricated by simultaneous crystallisation of the two solid solutions α- and β-sialon from a eutectic composition liquid. This requires careful design of the starting composition which is usually located within the (α + β)-sialon region of the α-sialon plane as illustrated in Fig. 18.1. The α/β phase ratio can be controlled just by changing the overall composition in most sialon systems. β-Sialon grains normally grow in elongated shape with a high aspect ratio (Lange, 1979; Wötting et al., 1986), and this
© Woodhead Publishing Limited, 2006
Sialons
β-sialon line
495
4 (Al2O3·AIN) 3
n = 3.0
n = 2.0
α + β-sialon region
n = 1.0 Si3N4
α-sialon region
m = 1.0 (M2O3·9AIN)
m = 2.0 m = 3.0 MN.3AIN
18.1 Schematic illustration of α- and β-sialon phase regions on the αsialon plane (M: metal cation with a valence of 3+).
promotes mechanisms such as crack bridging and deflection. α-Sialon with very high hardness used to be considered as always occurring with equiaxed grains and thus its fracture toughness is inferior to that of β-sialon. These observations led to work programmes being focused on the design of α⁄βsialon composites that combine the strength and fracture toughness of βsialon and the hardness of α-sialon to give improved mechanical properties. For example, it has been reported that the composition containing 50% βsialon and 50% α-sialon showed a hardness (HV10) of ~22 GPa and a fracture toughness of 5.5 MPa m1/2 (Ekström, 1997). Jones et al. (2003) chose Y-α⁄βsialon composites with compositions lying on the Si3N4–9AlN.Y2O3 line and produced sialon composites with different α/β-sialon ratios (75–51%), the bimodal microstructures of the resulting sialon composites showing improved wear properties compared to their monolithic counterparts. More recently, it has been observed that dense α-sialons containing elongated grains can also be prepared by carefully selecting the starting composition (Shen et al., 1996a; Chen and Rosenflanz, 1997; Nordberg et al., 1997b, Wood et al., 1999; Yu et al., 2001a); these are encouraging results and show the good potential for obtaining materials with high hardness and toughness. Therefore, α/β composites have become an important research area, and details of the starting powder, composition, processing route, etc., should be carefully considered in order to obtain the desired phases, grain shape and size to optimise the resulting bimodal microstructure. Composition design Different starting powders and chemical compositions result in different phases, grain boundary states, and microstructures (see Table 18.1). For
© Woodhead Publishing Limited, 2006
496
Ceramic matrix composites
Table 18.1 Effects of starting powder, composition on phase present, microstructure and mechanical properties of α -sialons Composition Starting (m, n) powder*
α ⁄β Sialon ratio
Morphology of grains
HV10 (GPa)
KIc (MPa m1/2)
(0.5, 2.0)
α
41/59
Equiaxed + elongated
17.9 ± 0.5
6.8 ± 0.5
(1.0, 1.0)
α
100/0
Equiaxed
20.3 ± 0.4
4.9 ± 0.3
(1.0, 1.0)
β
100/0
Equiaxed
20.2 ± 0.9
4.2 ± 0.4
(1.0, 2.0)
α
100/0
Equiaxed + elongated
18.3 ± 0.5
5.0 ± 0.4
(1.0, 2.0)
β
100/0
Equiaxed + elongated
18.0 ± 0.2
4.4 ± 0.3
(2.0, 2.0)
α
100/0
Equiaxed
15.3 ± 0.7
4.7 ± 0.7
*α and β stand for α- and β-Si3N4 respectively. Reprinted from J. Mater. Sci., 36(14), Yu Z B, Thompson D P and Bhatti A R, ‘Selfreinforcement in Li-α-sialon ceramics’, Fig. 2, 3343–3353, (2001). Copyright 2001, with kind permission of Kluwer Academic Publishers.
instance on the α-sialon plane, for a general α-sialon composition MxSi12–(m+n)Alm+nOnN16–n, designated as (m, n), the microstructures of both the low (m, n) composition (1.0, 1.0) and the high (m, n) composition (2.0, 2.0) consist mainly of equiaxed α-sialon grains, whereas (0.5, 2.0) and (1.0, 2.0) results show a bimodal microstructure consisting of equiaxed and elongated grains of the two main sialon phases. Also, samples prepared with high βphase content silicon nitride powder show a lower toughness than those prepared from a high α-phase starting silicon nitride powder. Volume fraction of elongated grains According to the results of Faber and Evans (1983a,b), the majority of the toughening from crack deflection develops for a volume fraction of reinforcement of <20% for a given morphology of the reinforcing phase. Figure 18.2 shows typical microstructures for sialon samples with different compositions. The volume of the elongated grains in the microstructure shown in Fig. 18.2(b) is over 20%, and an excess of elongated grain in composition (0.5, 2.0) does not add to the fracture toughness. So, it might be expected that the contribution of the volume fraction of elongated grains in these sialons to crack deflection would be comparable. However, as Table 18.1 shows, the (0.5, 2.0) composition gives an approximately 50% higher fracture toughness compared to the (1.0, 2.0) sample; this can be attributed to other factors as discussed below.
© Woodhead Publishing Limited, 2006
Sialons
(a)
497
(b)
18.2 (a) Microstructures of sialons with different compositions: (a) (1.0, 1.0), (b) (0.5, 2.0) (reprinted from J. Eur. Ceram. Soc., 21(13), Yu Z B, Thompson D P and Bhatti A R, ‘In-situ growth of elongated grains in Li-α-sialon ceramics’, 2423–2434 (2001). Copyright 2001, with permission of Elsevier).
The aspect ratio effect The length:diameter (L/D) ratio of the elongated grain plays an important role in toughening. The average L/D of elongated sialon grains in sample (0.5, 2.0) is about 10–12, which is higher than that in the (1.0, 2.0) sample with an average L/D of ~6. It was predicted (Shalek et al., 1986) that rodshaped obstacles with large aspect ratios should impart maximum toughness because high aspect ratios result in a large twist angle of the crack. The maximum effect which can be achieved with rod-like particles is: Gc/Gm ≈ 4.0 (when L/D = 12)
(18.1)
Gc/Gm ≈ 3.0 (when L/D = 3)
(18.2)
where Gc = crack resistance force of the composite, and Gm = crack resistance force of the matrix. Therefore, the contribution of the elongated sialon grains in (0.5, 2.0) samples towards the fracture toughness is ~1.2 times higher than that of elongated sialon grains in (1.0, 2.0) samples as deduced from the L/D ratio. Thermal and Young’s modulus mismatches Elongated α- and β-sialon grains have different thermal expansion coefficients and slightly different Young’s moduli, which cause local stress fields to form around the elongated grains, and these play a large role in the crack deflection process. Compared to pure (1.0, 2.0) α-sialon, the multiphase α ⁄β-sialon of composition (0.5, 2.0) showed higher fracture toughness. This can be partially explained by the physical mismatches. Assuming plane strain and isotropic elastic behaviour, the thermal residual stresses P are proportional to the
© Woodhead Publishing Limited, 2006
498
Ceramic matrix composites
difference between the thermal expansion coefficients of the equiaxed and elongated sialon grains estimated according to the analysis of Budiansky et al. (1986) and Yu et al. (2001b): P=
E eq (1 – f )ε 2λ 1 (1 – ν α )
(18.3)
where ε = (αel – αeq )∆T λ1 = 1 – (1 – 2ν)[1 – f + (1 – f )Eeq/Eel]/[2(1 – ν)] f = volume fraction of the elongated grains αel, αeq = linear thermal expansion coefficients Eeq, Eel = Young’s moduli νeq, νel = Poisson’s ratios subscripts el and eq refer to the elongated and equiaxed sialon grains, respectively, and here it is assumed that νeq = νel = ν. The residual stress plays an important role in the deflection process since it actually determines the type of crack–microstructure interactions. The (1.0, 2.0) sample consists of pure α-sialon reinforced with in situ formed elongated α-sialon grains, and the local residual stress induced by thermal and elastic mismatch around the elongated α-sialon grains is nearly zero, i.e. αelα – αeqα = 0 and Εelα – Eeqα = 0. Therefore, the residual stress makes little contribution to the crack deflection, and the crack tilting and twisting is relatively small because they are only caused by the elongated grains. However, in the (0.5, 2.0) composite consisting of α-sialon grains reinforced with in situ formed elongated β-sialon needles, there exist local residual stresses around the elongated β-sialon grains induced by thermal and elastic mismatch, because the thermal expansion coefficients and Young’s moduli between the matrix, the equiaxed α-sialon grains (αeq, Eeq) and the elongated grains (αelβ, Eelβ) are slightly different, i.e. αelβ – αeqα ≠ 0, and Eelβ – Eeqα ≠ 0. Therefore, the residual stress field in sample (0.5, 2.0) makes an additional contribution to crack deflection. So the crack tilting and twisting are more marked in the (0.5, 2.0) sample than in the (1.0, 2.0) sample and the fracture toughness reflects this.
18.4.2 Particle-reinforced sialons Because conventional processing routes can be used, particle reinforcement is a more attractive fabrication route for making α⁄β-sialon composites, and offers other advantages, such as cheapness and the avoidance of health hazards associated with the use of whiskers or short fibres. Apart from α⁄β-sialon composites, various kinds of particles including ZrO2, SiC, TiN, and MoSi2, among others, have been suggested and explored as the reinforcing phase.
© Woodhead Publishing Limited, 2006
Sialons
499
Toughening mechanisms using ZrO2 particles in silicon nitride or sialonbased matrices can be tailored to take advantage of the phase transformation and this is still a popular research area (Ekström et al., 1991; Yu et al., 2002a). A slight increase in both strength and toughness can be observed for both silicon nitride and sialons. However, ZrO2 and sialons are not chemically compatible and there is some difficulty in controlling the reactions between them at sintering temperatures. SiC particles have been used as reinforcements for Si3N4 and sialon ceramics because of the good compatibility between nitrides and SiC. Buljan et al. (1987) found that an increase in toughness was observed for coarse rather than fine SiC particles since the coarser particles can promote large crack deflection. Niihara et al. (1990) reported that Si3N4/32vol%SiC nanocomposites with 8 wt% of Y2O3 sintering additive exhibited strengths of 970 MPa and 840 MPa at 1400°C and 1500°C, respectively, and indicated that the nanosize SiC particles were mainly dispersed within Si3N4 grains with little additional grain boundary material. More recently, the preparation of βsialon composites reinforced by nano-SiC particles generated by in situ chemical techniques has been explored by Li et al. (1997), who found that the chemical route is favourable for the sintering of the composite. Liu et al. (1999) demonstrated that when 10–20 wt% SiC nanoparticles were added to a series of rare-earth doped Ln-α-sialons (Ln = Nd, Dy, Yb) of composition Ln0.33Si9.3Al2.7O1.7N14.3, hot-pressed at 1800°C and then heat-treated at 1450°C to modify the microstructure and properties by α ⁄ β-sialon transformation, an enhanced hardness was found in proportion to the content of SiC particles and with an increase in the α:β-sialon ratio, with a maximum hardness of 22 GPa (HV10) being achieved; however, a decrease in fracture toughness was observed. It was suggested that the significant increase in high temperature strength could be attributed to residual microstresses around the SiC particles due to the difference in thermal expansion coefficients between Si3N4/sialons and SiC. The strengthening and toughening mechanism of nano-SiC particle reinforced sialon ceramics has generated many current research programmes. MoSi2 particulate reinforced α-sialon composites have been studied by Nordberg and Ekström (1995) who found that the hardness of fully dense samples changed from 22.5 to 15.3 GPa with KIc increasing from 3.4 to 5.2 MPa m1/2, when up to 30 vol% MoSi2 was present. Particle-reinforced sialon composites are a new direction and some progress has been made; however, to achieve a significant toughening effect, more work is needed in this area.
18.4.3 Whisker-reinforced sialons Whiskers provide a better reinforcement than particles from the point of view of the effect of dispersoid shape on toughness based on theoretical predictions (Shalek et al., 1986). The most commonly used whisker
© Woodhead Publishing Limited, 2006
500
Ceramic matrix composites
reinforcement for sialons is SiC. SiC whiskers have been widely used in toughening and strengthening Si3N4 ceramics (Buljan et al., 1987; Campbell et al., 1990) and it has been observed that the strength and fracture toughness increased by 25% and 40% respectively, compared with monolithic Si3N4 again depending on sintering additives, sintering temperatures and whisker type. The incorporation of SiC whiskers into Y-doped α-sialon ceramics has been explored by Nordberg et al. (1993), for composites containing 30 wt% SiC whiskers. Unfortunately, only a minor part of the SiC whisker/sialon matrix interfaces consisted of a 4–5 nm thick amorphous layer (which facilitated pullout), whereas a direct and strong bond was the most commonly occurring contact. The hardness of the composites increased to ~21 GPa, but the fracture toughness increased only slightly because of the strong bonding between whiskers and matrix. Although fracture toughness can be increased, particle- or whisker-reinforced sialon composites generally show brittle behaviour and low damage tolerance; this is in contrast to fibre-reinforced sialons which exhibit non-catastrophic failure.
18.4.4 Carbon fibre (Cf)-reinforced sialons Carbon fibres offer an excellent combination of strength, modulus, and toughness compared with all other reinforcing fibres, and among available materials, carbon fibres can retain good mechanical properties above 1000°C. Carbon fibres have been widely used in ceramic composite systems (Guo et al., 1982; Prewo, 1982; Phillips et al., 1990; Zhang and Thompson, 1995, 1997; Yu and Thompson, 1998; Yu et al., 2002a). Yu and Thompson (1998) demonstrated a new two-step process in developing carbon fibre-reinforced α-sialons to deal with the problems encountered in fabricating sialon composites. In their process, first of all, α-sialon was pre-prepared by pressureless sintering and a lithium aluminium oxide glass (LAS) sintering additive was prepared by conventional melting followed by crushing into a fine powder. Second, continuous fibre-reinforced sialon matrix composites, with the asprepared α-sialon powder and sintering additives (present as 25% by volume of the matrix phase), were produced using slurry infiltration followed by hotpressing and achieved apparent densities of up to 96% of theoretical density below 1550°C for composites containing as much as 60 vol% of carbon fibres. In the next section, effects such as thermal mismatch, fibre distribution and processing parameters in carbon fibre-reinforced sialons are discussed. Effects of processing parameters on Cf /sialon composites As shown in Table 18.2, both the volume content of Cf and the sintering temperature affect the properties and fracture behaviour of Cf /α-sialon
© Woodhead Publishing Limited, 2006
Sialons
501
Table 18.2 Effect of sintering temperature on the properties of Cf /sialon composites Sample
Properties
1350°C
1400°C
1450°C
1500°C
1550°C
α + 60 vol% (Cf)
KIc (MPa m1/2)
—
7.3 ± 1.0
12.7 ± 2.0
9.4 ± 2.0
3.6 ± 1.0
α + 60 vol% (Cf)
σ (MPa)
—
165 ± 10
223 ± 20
224 ± 20
93 ± 10
α + 40 vol% (Cf)
σ (MPa)
189 ± 10
—
337 ± 20
—
147 ± 10
Reprinted from J. Eur. Ceram. Soc., 22(2), Yu Z B, Thompson D P and Bhatti A R, ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, 225–235 (2002). Copyright 2002, with permission of Elsevier.
composites. The typical load–displacement curves and fracture surfaces for these composites are given in Fig. 18.3. Below 1500°C, a typical load–displacement curve for a Cf /sialon composite indicates a characteristic first matrix cracking event followed by a region of intermittent cracking non-linear behaviour until eventual failure occurs at some maximum load. The load drop occurs at a characteristic value followed by a long ‘tail’. Fracture at this maximum load is generally associated with fibre bundle fracture, and the tail region of intermittent cracking is associated with fibre bridging and pullout (Evans, 1985; Evans et al., 1991). Cracking, debonding and pullout are the main toughening processes operating in these materials during failure. However, sialon composites sintered at high temperatures (e.g. 1550°C) give poor mechanical performance (Table 18.2)
Load (N)
200 150 100
1.86
1.50
1.14
0.42
0.78
0.24
0.31
0.10
0.02
0
0.17
50
Displacement (mm) (a)
(b)
18.3 Fracture behaviour of Cf/α-sialon hot-pressed at 1500°C: (a) load–displacement curve; (b) fracture surface (reprinted from J. Eur. Ceram. Soc., 22(2), Yu Z B, Thompson D P and Bhatti A R, ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, 225–235 (2002). Copyright 2002, with permission of Elsevier).
© Woodhead Publishing Limited, 2006
502
Ceramic matrix composites
and usually show a linear load–displacement behaviour up to a maximum value followed by a sudden decrease in load with a very limited continued deformation, suggesting that serious degradation of the carbon fibre occurs at this sintering temperature. A significant improvement in fracture toughness can be achieved in Cf/α-sialon ceramics sintered at 1450°C, with KIc values as high as 12.7 MPa m1/2. However, the bending strength is low (~222 MPa), which can be in part attributed to thermal mismatch between the matrix and the fibre and also to non-homogeneous distribution of the fibres, which is often found in carbon fibre-reinforced ceramic matrix composites. Thermal mismatch in Cf /α-sialon composites The properties of fibrous composite materials are strongly dependent on microstructural parameters such as fibre diameter, fibre length, fibre distribution, volume fraction of the fibres and the alignment and packing arrangement of the fibres. It is important to control these parameters for effective design and manufacture of these composites. An ideal situation is that all the fibres should be aligned parallel to each other and distributed homogeneously, but this is not easy to do in practice. An example is given in Fig. 18.4, which shows micro-cracks occurring in Cf/α-sialon composites containing ~60 vol% of carbon fibres induced by non-homogeneous distribution of the fibres and the thermal mismatch between Cf and the sialon matrix.
18.4 Micro-crack induced by non-uniform distribution of carbon fibres and thermal mismatch between Cf and sialon in the Cf /α-sialon composite (reprinted from J. Eur. Ceram. Soc., 22(2), Yu Z B, Thompson D P and Bhatti A R, ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, 225–235 (2002). Copyright 2002, with permission of Elsevier).
© Woodhead Publishing Limited, 2006
Sialons
503
Assuming that continuous carbon fibre reinforced α-sialon composites are composed of two materials, namely carbon fibre layers and α/LAS matrix layers, forming alternate sheets with linear thermal expansion coefficients αf, αm, elastic moduli Ef, Em, and Poisson’s ratios νf, νm, then the boundary stresses in the laminate can be estimated in the same way as given by Kingery et al. (1976). When the composite is sintered at a relatively high temperature (Ts ~ 1450°C), and in a stress-free condition, and is cooled down to room temperature Tr, the temperature difference ∆T = Tc – Tr,, where Tc is the temperature below which the matrix ceases to behave viscoplastically, will result in thermal expansion changes in the reinforcing carbon fibres of αf∆T and in the α-sialon/LAS matrix of αm∆T. These expansions are not the same and the composite must adopt an intermediate overall expansion, determined by the relative elastic moduli and volume fractions of fibres and matrix, but with the net compressive force on the carbon fibres equal to the net tensile force in the α-sialon matrix. If σ is this stress, V the volume fraction, and ε the actual strain, then when the composite is cooled down to room temperature, for a composite in which the continuous fibres are distributed homogeneously (Kingery et al., 1976): σmVm + σfVf = 0
(18.4)
Based on this model, the thermal stress in the matrix has been deduced by Yu et al. (2002a) to be: E – Ef σm = m V ∆α∆T 1 – νf f
(18.5)
From equation (18.5), the stress caused by thermal mismatch in the matrix is proportional to the volume fraction of the fibres. Therefore, large thermal stresses in some local regions, where the distribution of carbon fibres is not uniform, will result in some cracks in the composite. It is worth pointing out that carbon fibre itself has anisotropic thermal expansion properties, and therefore this mismatch between the carbon fibres and the α-sialon matrix should be considered in both the radial and axial directions when carbon fibres are unidirectionally aligned in the composite. The thermal stress caused by thermal expansion differences between the carbon fibres and the matrix in the radial (σr) and axial (σa) directions can be estimated from the formulae (Chawla, 1993; Kerans and Parthasarathy, 1991):
σr =
– qE f E m [α f ⋅ r r – α m ) ∆T + A / r ] E f (1 + ν m ) + E m (1 – ν f )
σa = (αm – αf·a)∆T
© Woodhead Publishing Limited, 2006
E f Vf Vf [ E f / E m ) – 1] + 1
(18.6) (18.7)
504
Ceramic matrix composites
where ∆T = temperature change during cooling q = an adjustable parameter (normally taken as unity) αf.r, αf.a = thermal expansion coefficients of the fibres in radial and axial directions respectively αm = matrix thermal expansion coefficient A = amplitude of fibre roughness r = fibre radius Ef, Em = elastic moduli νf, νm = Poisson’s ratios subscripts f and m refer to the fibre and matrix, respectively. In Cf /sialon composites, because αf.r (~8 × 10–6 °C–1) is larger than αm (~5.6 × 10–6 °C–1) in the radial direction, the thermal mismatch between fibres and matrix is acceptable. However, in the axial direction, αf.a (~0) is much smaller than αm, and on cooling from the hot-pressing temperature will put the matrix in tension, generating stresses of ~466 MPa and 700 MPa for a Cf /sialon composite containing 40 vol% and 60 vol% of carbon fibres respectively (Yu et al., 2002a), which in both cases is greater than the strength of the matrix and therefore results in the appearance of transverse cracks. Compensation in thermal mismatch To reduce residual stresses, careful control of the slurry infiltration processing parameters is necessary, but this may not be enough. Alternative procedures such as incorporating additions of ZrO2 (Guo et al., 1982; Zhang and Thompson, 1995; Yu et al., 2002a) as well as introducing other phases into the matrix have been explored. By making use of the 3–5% volume increase accompanying the t → m ZrO2 transformation, Yu et al. (2002a) demonstrated that addition of ZrO2 to a sialon matrix is an effective route for compensating for the thermal expansion differences in the matrix. As shown in Fig. 18.5, for Cf /sialon composites, when 20 wt% ZrO2 was added to the sialon matrix, all microcracks were eliminated. When these cracks were eliminated, the strength and fracture toughness of the Cf /α-sialon composites were significantly improved (see Table 18.3). The KIc value for carbon fibre-reinforced α-sialons containing 20 wt% ZrO2 reached a value of ~12 MPa m1/2, with a bending strength of ~410 MPa, which is an excellent combination of these key properties. The introduction of ZrO2 plays an important role in changing the thermal behaviour of the composite. However, whether the t → m ZrO2 phase transformation occurs or not is conditional on other factors, such as the particle size of ZrO2, chemical composition and the local stress situation. Briefly, it is important to ensure that the t → m transformation successfully occurs immediately below the softening temperature of the glass during
© Woodhead Publishing Limited, 2006
Sialons
505
18.5 Back-scattered SEM photo showing the elimination of microcracks in Cf/α-sialon composites with the addition of 20 wt% ZrO2 (black strips: carbon fibres; white dots: ZrO2) (reprinted from J. Eur. Ceram. Soc., 22(2), Yu, Z B, Thompson D P and Bhatti A R, ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, 225–235 (2002). Copyright 2002, with permission of Elsevier). Table 18.3 Properties of carbon fibre/α-sialon composites with and without added ZrO2 Temperature (°C)
ZrO2 content (wt%)
Strength (MPa)
1400
0 20 0 20 0 20
190 336 218 410 105 120
1450 1500
± ± ± ± ± ±
10 10 5 10 5 10
Toughness (MPa m1/2) 7.0 12.5 5.6 11.3 4.0 3.5
± ± ± ± ± ±
1.0 1.5 0.5 1.0 0.2 0.2
Young’s modulus (GPa) — 171.4 — 182.0 — 180.0
(Reprinted from J. Eur. Ceram. Soc, 22(2), Yu Z B, Thompson D P and Bhatti A R, ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, 225–235 (2002). Copyright 2002, with permission of Elsevier).
cooling, and that the amount of ZrO2 addition required is not unreasonable (Yu et al., 2002a).
18.4.5 SiC fibre-reinforced sialons SiC fibres (SiCf) are some of the more refractory ceramic fibres, and perform very satisfactorily in oxidising environments. A key problem in the development of SiC fibre composites is thermal degradation of the SiC fibre (Johnson et al., 1987). The properties of SiC fibre start degrading above 600°C because
© Woodhead Publishing Limited, 2006
506
Ceramic matrix composites
of the thermodynamic instability of the composition and the microstructure (Chawla, 1993). Another problem is chemical reactions with the matrix. In this section, new processing techniques such as heat-treatment of SiC fibres and low-temperature processing routes involving selection and introduction of low-temperature sintering additives combined with careful control of process parameters, are shown to be useful for successful fabrication of α- and βsialon composites. Effect of heat-treatment of SiC fibres on SiCf /sialon composites The effect of high-temperature heat-treatment on SiC fibre strength has been examined for a variety of environments (Mah et al., 1984; Pysher et al, 1989; Bibbo et al., 1991; Bodet and Tressler, 1995). More recently, Demir and Thompson (2001) showed that SiC fibres heat-treated at 1600°C for 30 min under 45 bar of CO atmosphere gained 50% in strength while at the same time a thin carbon coating was deposited on the fibre. These changes offer significant advantages for the use of SiC fibres to reinforce sialons at high temperatures. These authors used the as-received Nicalon SiC fibres (NL207) and then the heat-treated Nicalon SiC fibres as reinforcements for the strengthening and toughening of β-sialons at the z = 1 and z = 2 compositions, Table 18.4 Mechanical properties of as-received and heat-treated SiCf /β-sialon composites β-Sialon Fibre* Sintering z-value additives
Temperature (°C)
TD (%)
Strength (MPa)
Toughness (MPa m1/2)
1
A
MgO, Sm2O3
1600
91
349 ± 31
11.71 ± 1.8
1
H
MgO, Sm2O3
1600
98
687 ± 52
11.04 ± 1.5
1
A
MgO, Y2O3, Sm2O3
1550
90
375 ± 112 10.66 ± 2.1
1
H
MgO, Y2O3, Sm2O3
1550
95
623 ± 53
13.36 ± 1.8
3
A
MgO, Y2O3, Sm2O3
1550
98
419 ± 55
11.93 ± 1.9
3
H
MgO, Y2O3, Sm2O3
1550
99
627 ± 25
11.87 ± 0.7
3
A
Li2O, MgO, SiO2
1450
94
408 ± 32
11.03 ± 2.5
3
H
Li2O, MgO, SiO2
1450
98
512 ± 24
10.50 ± 0.4
*A: as-received fibre; H: heat-treated fibre Reprinted from J. Eur. Ceram. Soc., 21(5), Demir A and Thompson D P, ‘Highperformance SiC-fibre reinforced β-sialon CMCs prepared from heat-treated Nicalon fibres’, 639–647 (2001). Copyright 2001, with permission of Elsevier.
© Woodhead Publishing Limited, 2006
Sialons
507
and found a marked difference in mechanical properties of hot-pressed samples for the as-received and heat-treated fibres as summarised in Table 18.4. Table 18.4 shows that in the selected compositions all the samples show high fracture toughness, whereas for bending strength most samples show an enormous strength increase for the heat-treated fibres, demonstrating that the latter achieved toughening and strengthening simultaneously. In addition, it seems that the heat-treated fibres also played a role in facilitating the densification of the sialon composites; moreover, heat-treatment of the fibres allows high-temperature processing of samples. Effect of sintering additives on SiCf /sialon composites Sintering additives have a strong effect on the interface between matrix and fibres. As shown in Table 18.5, composites hot-pressed with different sintering additives – lithium aluminosilicate glass (LAS), nitrogen-containing lithium aluminosilicate glass (NLAS), and mixed Y2O3 + Al2O3 + ZrO2 (YAZ) additives – show different fibre pullout effects. Fibre pullout can be achieved in composites with LAS and NLAS as sintering additives sintered at low temperatures and but not with the YAZ additive. The microstructures of fracture surfaces of the SiCf /α-sialon composites with different sintering additives are shown in Fig. 18.6. The compositional change from the matrix to the fibre centre in SiCf/sialon composites with different sintering additives is shown in the form of the X-ray line scan (Fig. 18.7). YAZ as a sintering additive causes some chemical reaction at the interface Table 18.5 Effect of sintering conditions and additives on fibre pullout in SiCf /α-sialons Fibre
Matrix/sintering additives
SiC
α ⁄ LAS
SiC
α /NLAS
SiC
α /YAZ
Hot-pressing conditions
T (°C)
P (MPa)
t (min)
1400 1450 1500 1350 1400 1450 1400 1450 1500
15 15 15 15 15 15 15 15 15
10 10 10 10 10 10 10 10 10
Maximum fibre pullout length (µm) ~25 ~5 ~0 ~80 ~40 ~10 ~0 ~0 ~0
Reprinted from Composites: Part A, 33(5),Yu Z B, Thompson D P and Bhatti A R, ‘Fabrication and characterisation of SiC fibre reinforced lithium-α-sialon matrix composites’, 621–629 (2002). Copyright 2002, with permission of Elsevier.
© Woodhead Publishing Limited, 2006
508
Ceramic matrix composites
(a)
(b)
(c)
18.6 Micrographs of fracture surfaces of continuous SiCf/α-sialon composites sintered with (a) YAZ, (b) LAS, and (c) NLAS additives (reprinted from Composites: Part A, 33(5), Yu Z B, Thompson D P and Bhatti A R, ‘Fabrication and characterisation of SiC fibre reinforced lithium-α-sialon matrix composites’, 621–629 (2002). Copyright 2002, with permission of Elsevier).
between the fibre and the matrix. Aluminium and oxygen from the matrix have diffused into the fibre, and silicon has diffused out of the fibre, forming a reaction layer about 1.5 µm in thickness, which results in such a strong interface that neither debonding nor pullout can take place during crack propagation in the composite. When LAS was used as the sintering additive (Fig. 18.7(b)), there was no obvious chemical reaction between the fibre and the matrix, which is further confirmed by the X-ray line scan result. When NLAS was used as the sintering additive, the interface between the fibres and the matrix was very clear (Fig. 18.6(c)) and there is no obvious chemical reaction on the surface (Fig. 18.7(c)), showing good chemical compatibility between fibre and matrix. Residual stresses in SiCf /α-sialon composites Thermal mismatch between SiC fibres and a sialon matrix is also an issue which can cause poor mechanical performance, even though there may be excellent chemical and physical compatibility between the fibres and the
© Woodhead Publishing Limited, 2006
Sialons OKa, 13
OKa, 12
Alka, 91
Alka, 41
SiKa, 563
SiKa, 201
509
(b)
(a) OKa, 13
Alka, 63
SiKa, 276
(c)
18.7 X-ray line scan of O, Al and Si in continuous SiCf/α-sialon composites sintered with (a) YAZ, (b) LAS, and (c) NLAS additives respectively (reprinted from Composites: Part A, 33(5), Yu Z B, Thompson D P and Bhatti A R, ‘Fabrication and characterisation of SiC fibre reinforced lithium-α-sialon matrix composites’, 621–629 (2002). Copyright 2002, with permission of Elsevier).
matrix. For instance, in SiCf/α-sialon composites with LAS as the sintering additive, the coefficient of thermal expansion of the SiC fibre, αSiCf (3.4 × 10–6°C–1), is lower than that of the α /LAS matrix (~5.6 × 10–6°C–1). During cooling from high temperatures, radial compressive stresses build up because the matrix shrinks more round the fibres. The magnitude of this stress can be approximately calculated from equation (18.6). A calculated radial compressive stress of 246 MPa on the fibre (corresponding to a frictional stress as large as 24.6 MPa) was obtained by Yu et al. (2002b), which was big enough to cause the SiC fibres to break when they were torn from the α-sialon matrix. Therefore, in SiC fibre-reinforced sialon composites, thermal treatment of fibres, thermal mismatch and chemical reactions between the fibre and the sialon matrix are significant factors affecting the interfacial bonding in these composites. The sintering additive plays an important role in controlling the nature of the interface and requires careful selection.
© Woodhead Publishing Limited, 2006
510
18.5
Ceramic matrix composites
Future trends
Sialon composites can be produced by in situ growth of elongated α- or βsialon grains, providing a self-reinforcement mechanism involving crack deflection and bridging. The reinforcement effect is mainly determined by the overall composition which in turn determines the morphology and the type of matrix sialon phase present; α ⁄β-sialon matrix offers the best option for producing a final composite with the desired microstructures and properties. Sialons reinforced by a secondary particulate phase or short fibres/whiskers gives some toughening effect. The most marked strengthening and toughening effects (achieved simultaneously) in sialons are obtained using fibre-reinforced composites, and these show controlled fractured behaviour and high damage tolerance. Avoiding physical and chemical incompatibilities between the sialon matrix and the fibres is very important, and initial calculations are essential to determine whether it is necessary to tailor the properties of the sialon matrix or to modify the properties of the fibres or both. There is also scope for the development of new techniques such as chemical vapour infiltration (CVI) (Caputo and Lackey, 1984; Caputo et al., 1985), normal chemical reaction bonding processes, laminar sialon composites, etc. More recently, laminated composites in non-oxide and sialons have demonstrated very promising results for strengthening (Goto and Kato, 1998) and even achieved a non-brittle failure behaviour accompanied by high damage tolerance (Yu and Krstic, 2003; Yu et al., 2005).
18.6
References
Bibbo, G.S., Benson, P.M. and Pantano, C.G., (1991), ‘Effect of carbon monoxide partial pressure on the high-temperature decomposition of Nicalon fibre’, J. Mater. Sci., 26(18), 5075–5080. Bodet, R. and Tressler, R.E., (1995), ‘Thermomechanical stability of Nicalon fibres in a carbon monoxide environment’, J. Eur. Ceram. Soc., 159(10), 997–1006. Budiansky, B., Hutchinson, J.W. and Evans, A.G., (1986), ‘Matrix fracture in fibrereinforced ceramics’, J. Mech. Phys. Solids., 34(2), 167–189. Buljan, S.T., Baldon, J.G. and Juchabee, M.L., (1987), ‘Si3N4–SiC composites’, Am. Ceram. Soc. Bull., 66(2), 347–352. Campbell, G.H., Rühle, M., Dalkgleish, B.J. and Evans, A.G., (1990), ‘Whisker toughening: a comparison between alumina and silicon nitride toughened with silicon carbide’, J. Am. Ceram. Soc., 73(3), 521. Cao, G.Z. and Metselaar, R., (1991), ‘α-Sialon ceramics: a review’, J. Chem. Mater., 3, 242–252. Caputo, A.J. and Lackey, W.J., (1984), ‘Fabrication of fibre-reinforced ceramic composites by chemical vapour infiltration’, Ceram. Eng. Sci. Proc., 5(7–8), 654–667. Caputo, A.J., Lackey, W.J. and Stinton, D.P., (1985), ‘Development of new, faster process for the fabrication of ceramic fibre reinforced ceramic composites by chemical vapour infiltraton’, Ceram. Eng. Sci. Proc., 6(7–8), 694–706. Chawla, K.K., (1993), Ceramics Matrix Composites, London, Chapman and Hall.
© Woodhead Publishing Limited, 2006
Sialons
511
Chen, I.W. and Rosenflanz, A., (1997), ‘A tough SiAlON ceramic based on alpha-Si3N4 with a whisker-like microstructure’, Nature, 389, 701–704. Demir, A. and Thompson, D.P., (2001), ‘High-performance SiC-fibre reinforced β-sialon CMCs prepared from heat-treated Nicalon fibres’, J. Eur. Ceram. Soc., 21(5), 639– 647. Ekström, T., (1989), ‘Effect of composition, phase content and microstructure on the performance of Yttrium–Si–Al–O–N ceramics’, Mater. Sci. Eng., A109, 341–349. Ekström, T., (1992a), ‘Preparation and properties of alpha-sialon ceramics’, J. Hard. Mater., 3, 109–118. Ekström, T., (1992b), ‘SiAlON ceramics sintered with yttria and rare earth oxides’, in Chen, I.W., Pedier, P.F., Mitomo, M., Pezow, G. and Yen, T.S., Silicon Nitride Ceramic Scientific and Technological Advances, Proc. Mat. Res. Soc. Symp., 287, 121–132. Ekström, T., (1997), ‘α-Sialon and α/β-sialon composites: recent research’, in Babini, G.N., et al., Engineering Ceramics ’96: Higher Reliability Through Processing, Dordrecht, Kluwer Academic, 147–167. Ekström, T. and Nygren, M., (1992), ‘SiAlON ceramics’, J. Am. Ceram. Soc., 75, 259– 276. Ekström, T., Falk, L.K.L. and Knutson-Wedel, E.M., (1991), ‘Si3N4–ZrO2 composites with small Al2O3 and Y2O3 additions prepared by HIP’, J. Mater. Sci., 26(16), 4331– 4340. Evans, A.G., (1985), ‘Engineering property requirements for high performance ceramics’, Mater. Sci. Eng., 71, 3–21. Evans, A.G., Zok, F.W. and Davis, J. (1991), ‘The role of interfaces in fibre-reinforced brittle matrix composites’, Comp. Sci. and Tech., 42, 3–24. Faber, K.T. and Evans, A.G., (1983a), ‘Crack deflection process: I. Theory’, Acta. Metall., 67, 565–576. Faber, K.T. and Evans, A.G., (1983b), ‘Crack deflection process: II. Experiment’, Acta. Metall., 67, 577–584. Gauckler L.J., Lukas, H.L. and Petzow, G. (1975), ‘Contribution to the phase diagram Si3N4–AlN–Al2O3–SiO2’, J. Am. Ceram. Soc., 58(7/8), 346–347. Goto, Y. and Kato, F.M., (1998), ‘Strength of β-sialon/Si3N4 layered composites’, J. Mater. Sci., 33, 423–427. Guo, J.K., Mao, Z.Q., Bao, C., Wang, T. and Yen, D.S., (1982), ‘Carbon fibre reinforced silicon nitride composite’, J. Mater. Sci., 17, 3611–3616. Hampshire, S., Park, H.K., Thompson, D.P. and Jack, K.H., (1978), ‘α-Sialon ceramics’, Nature, 274, 880–883. Izhevskiy, V.A., Genva, L.A., Bressiani, J.C. and Aldinger, F., (2000), ‘Progress in SiAlON ceramics’, J. Eur. Ceram. Soc., 20(3), 2275–2295. Jack, K.H., (1976), ‘Review: sialon and related nitrogen ceramics’, J. Mater. Sci., 11, 1135–1158. Jack, K.H. and Wilson, W.J., (1972), ‘Ceramics based on the Si–Al–O–N and related systems’, Nature Phys. Sci., 238, 28–29. Johnson, L.F., Hasselman, D.P.H. and Minford, E., (1987), ‘Thermal conductivity of carbon fibre-reinforced borosilicate glass’, J. Mater. Sci., 22, 3111–3117. Jones, M.I., Hirao, K., Hyuga, H., Yamauchi, Y. and Kanzaki, S., (2003), ‘Wear properties of Y-α /β composite sialon ceramics’, J. Eur. Ceram. Soc., 23(10), 1730–1750. Kerans, R.J. and Parthasarathy, T.A., (1991), ‘Theoretical analysis of the fibre pullout and push out test’, J. Am. Ceram. Soc., 74(7), 1585–1596. Kingery, W.D., Bowen, H.K. and Uhlmann, D.R., (1976), Introduction to Ceramics, New York, John Wiley & Sons.
© Woodhead Publishing Limited, 2006
512
Ceramic matrix composites
Kuang, S.F., Huang, Z.K., Sun, W.Y., and Yen, T.S., (1990), ‘Phase relationships in the Li2O-Si3N4-AIN system’, J. Mater. Sci Lett. (9), 72–74. Lange, F.F., (1979), ‘Fracture toughness of Si3N4 as a function of the initial alpha phase content’, J. Am. Ceram. Soc. 62(7–8), 428–430. Li, Q., Gao, L., Jiang, D., Zhang, C. and Yan, D.S., (1997), ‘Preparation of β-SiAlON/ nano-SiC composites’, J. Mater. Sci. Lett., 16, 1620–1621. Liu, Q., Gao, L., Yan, D.S. and Thompson, D.P., (1999), ‘Hard sialon ceramics reinforced with SiC nanoparticles’, Mater. Sci. Eng., A269, 1–7. Mah, T., Hecht, N.L., McCullen, D.E., Honigman, J.R., Kim, H.M., Katz, A.P. and Lipsitt, H.A., (1984), ‘Thermal stability of SiC fibre (Nicalon)’, J. Mater. Sci., 19(4), 1191–1201. Mandal H., Thompson D.P. and Ekström T. (1993), ‘α ⇔ β Transformation in heattreated sialon ceramics’, J. Eur. Ceram. Soc., 12(6), 421–429. Niihara, K., Izaki, K. and Kawakami, T., (1990), ‘Hot-pressed Si3N4–32%SiC nanocomposite from amorphous Si–C–N powder with improved strength above 1200°C’, J. Mater. Sci. Lett., 10, 112–114. Nordberg, L.O. and Ekström, T., (1995), ‘Hot-pressed MoSi2-particulate-reinforced αSiAlON composites’, J. Am. Ceram. Soc., 78(3), 797–800. Nordberg, L.D., Esktröm, T. and Wen, S., (1993), ‘Hot-pressed SiC whisker reinforced αSiAlON composites’, J. Hard. Mater., 4, 121–135. Nordberg, L.O., Ekström, T. and Xu, F.F. (1997a), ‘Simultaneously MoSi2 and SiC whisker reinforced α-SiAlON composites’, J. Mater. Sci. Lett., 16, 917–920. Nordberg, L.O., Shen, Z.J., Nygren, M. and Ekström, T., (1997b), ‘On the extension of the α-SiAlON solid solution range and anisotropic grain growth in Sm-doped αSiAlON ceramics’, J. Eur. Ceram. Soc., 17(4), 575–580. Oyama, Y. and Kamigaito, O., (1971), ‘Solid solubility of some oxides in Si3N4’, Jpn. J. Appl. Phys., 10, 1637–1642. Phillips, D.C., Park, N. and Lee, R.J., (1990), ‘The impact behaviour of high performance ceramic matrix fibre composites’, Comp. Sci. and Tech., 37(1–3), 249–265. Prewo, K.M., (1982), ‘A compliant, high failure strain, fibre reinforced glass–matrix composite’, J. Mater. Sci., 17, 3549–3563. Prewo, K.M., (1989), ‘Fibre reinforced ceramics: new opportunities for composite materials’, Am. Ceram. Bull., 68(2), 395–400. Pysher, D.J., Goretta, K.C., Hodder, R.S. and Tressler, R.E. (1989), ‘Strength of ceramic fibre at elevated temperatures’, J. Am. Ceram. Soc., 72(2), 284–288. Sambell, R.A.J., Briggs, A., Phillips, D.C. and Bowen, D.H., (1972a), ‘Carbon fibre composites with ceramic and glass matrices, Part I. Discontinuous fibres’, J. Mater. Sci., 7, 663–675. Sambell, R.A.J., Briggs, A., Phillips, D.C. and Bowen, D.H., (1972b), ‘Carbon fibre composites with ceramic and glass matrices, Part II. Continuous fibres’, J. Mater. Sci., 7, 676–681. Shalek, P.D., Petrovic, J.J., Hurley, G.F. and Gac, F.D., (1986), ‘Hot pressed SiC whisker/ Si3N4 matrix composites’, Am. Ceram. Bull., 65(2) 351–356. Shen, Z.J., Ekström, T. and Nygren, M., (1996a), ‘Homogeneity region and thermal stability of neodymium-doped α-sialon ceramics’, J. Am. Ceram. Soc., 79(3), 721–732. Shen, Z.J., Ekström, T. and Nygren, M., (1996b), ‘Ytterbium-stabilized α-sialon ceramics’, J. Phys. D – Appl. Phys., 29(3), 893–904. Shen, Z.J., Ekström, T. and Nygren, M., (1996c), ‘Temperature stability of Samariumdoped α-sialon ceramic’, J. Eur. Ceram. Soc., 16(1), 43–53.
© Woodhead Publishing Limited, 2006
Sialons
513
Shen, Z.J., Ekström, T. and Nygren, M., (1996d), ‘Reactions on occurring in post heattreated α /β-sialons: on the thermal stability’, J. Eur. Ceram. Soc., 16(8), 873–883. Shen, Z.J., Nordberg L.O., Nygren, M. and Ekström, T., (1996e), ‘α-Sialon grains with high aspect ratio – Utopia or reality’, in Babini G.N., Engineering Ceramics ’96: Higher Reliability Through Processing, Dordrecht, Kluwer Academic, 169–178. Thompson, D.P., (1994), ‘α ⇔ β Sialon transformation’, in Hoffmann, M.J. and Petzow, G., Tailoring of Mechanical Properties of Si3 N4 Ceramics, Dordrecht, Kluwer Academic, 125–136. Wood, C.A., Zhao, H. and Cheng, Y.B., (1999), ‘Microstructural development of Ca-αsialon ceramic’, J. Am. Ceram. Soc., 82(2), 421–428. Wötting, G., Kanka, B. and Ziegler, G., (1986), ‘Microstructural development, microstructural characterization and relations to mechanical properties of dense silicon nitride’, in Hampshire S, Non-oxide Technical and Engineering Ceramics, London and New York, Elsevier Science, 83–96. Yu, Z.B. and Krstic, V.D., (2003), ‘Fabrication and characterisation of laminated SiC ceramics with self-sealed ring structure’, J. Mater. Sci., 38, 4735–4738. Yu, Z.B. and Thompson, D.P., (1998), ‘Preparation of carbon fibre reinforced Li-alphasialon composites’, in Gibson, G., Consolidating New Applications, Seventh International Conference on Fibre Reinforced Composite, Cambridge, UK, Woodhead, 264–270. Yu, Z.B., Thompson, D.P. and Bhatti, A.R., (1998a), ‘Preparation and homogeneity region of Li-alpha-sialon ceramics’, Brit. Ceram. Trans., 97(2), 41–46. Yu, Z.B., Thompson, D.P., and Bhatti, A.R., (1998b), ‘Reverse α ⇔ β transformation in Li-alpha-sialon ceramic’, in Hampshire S, International Symposium on Nitrides II, Zürich, Trans. Tech., 264–268. Yu, Z.B., Thompson D.P. and Bhatti, A.R. (2000), ‘Transformation and thermal stability of Li-alpha-sialon ceramics’, J. Eur. Ceram. Soc., 20(11), 1815–1828. Yu, Z.B., Thompson, D.P. and Bhatti, A.R. (2001a), ‘In-situ growth of elongated grains in Li-α-sialon ceramics’, J. Eur. Ceram. Soc., 21(13), 2423–2434. Yu, Z.B., Thompson, D.P., and Bhatti, A.R., (2001b), ‘Self-reinforcement in Li-α-sialon ceramics’, J. Mater. Sci., 36(14), 3343–3353. Yu, Z.B., Thompson, D.P. and Bhatti, A.R., (2002a), ‘Synergistic roles of carbon fibres and ZrO2 particles in strengthening and toughening Li-α-sialon composites’, J. Eur. Ceram. Soc., 22(2), 225–235. Yu, Z.B., Thompson, D.P. and Bhatti, A.R., (2002b), ‘Fabrication and characterisation of SiC fibre reinforced lithium-α-sialon matrix composites’, Composites: Part A, 33(5), 621–629. Yu, Z.B., Krstic, Z. and Krstic, V.D., (2005), ‘Laminated silicon nitride/silicon carbide composites with self-sealed structure’, Key Engineering Materials 280–283, 1873– 1876. Zhang, E. and Thompson, D.P., (1995), ‘Elimination of cracks in carbon fibre reinforced nitrogen glass composites’, in Bellosi A, Fourth Eur. Ceram., Italy, Gruppo Editoriale Faenza Editrice, S.p.A., 193-198. Zhang, E. and Thompson, D.P., (1997), ‘Carbon fibre reinforced nitrogen glass composites’, Composites: Part A, 28A, 581–586.
© Woodhead Publishing Limited, 2006
19 Carbon-ceramic alloys C B A L Á Z S I, Research Institute of Technical Physics and Materials Science, Hungary
19.1
Introduction
Engineering ceramics based on silicon nitride are well known as low-density materials with high strength and toughness. With the combination of these properties, silicon nitride ceramics are an ideal candidate for several structural and functional applications. However, because of the relatively high brittleness of ceramics there is a continual need to improve their mechanical characteristics. At the same time, intensive research is being performed to improve the thermal and electrical properties of ceramics. In general, there are two ways to improve the mechanical properties of ceramics: controlling the microstructure, and preparation of the composite. In connection to the microstructure–property relationship, new observations have been performed on structural and morphological development of silicon nitride ceramics [1, 2]. It was found that the development of an interlocking microstructure of elongated grains is vital to ensure that this family of ceramics has good damage tolerance. A fast (within minutes) in situ formation of a tough microstructure has been observed by a so-called dynamic Ostwald ripening process that results from the rapid heating rate. In this way, through formation of a tough interlocking microstructure (e.g. elongated β-Si3N4 grains), mechanical properties may be improved. On the other hand, physical and mechanical properties of ceramics can be improved through nanocomposite processing [3–5]. To increase the fracture toughness, various energy-dissipating components have been incorporated into ceramic matrices [4]. These secondary constituents can be introduced in whisker, platelet, particle or fiber forms. Depending on processing route, micro/nano or nano/nano type microstructures can be synthesized, and silicon nitride–silicon carbide nanocomposites have been developed, which retained high strength and good oxidation resistance up to 1400°C [6]. Moreover, through nanocomposite processing not only the high-temperature properties may be improved, but high-performance structural materials can be synthesized [7–9]. From the literature, several methods are known for porous ceramic 514 © Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
515
preparation, such as partial sintering [8] or using sintering additives with high melting points to hinder the sintering process [9]. It has been reported that by the addition of small amounts of carbon, high-porosity silicon nitride composites with low shrinkage have been realized [10]. This process involved the formation of SiC particles at the grain boundaries, which inhibited the rearrangement of Si3N4 particles during sintering, assuring high porosity. A method for producing porous silicon nitride with high strength and low thermal conductivity by adding β-Si3N4 crystal seeds was presented by Toriyama et al. [11, 12]. Careful control of the oxidation process of Si3N4 that contains organic binder as a source of carbon can lead to porous ceramics with high strength [13]. As carbon nanotubes present exceptional mechanical, superior thermal and electrical properties in general, by using them as reinforcing elements there are high expectations for improvement of quality of nano- and microcomposites [14–18]. As shown from earlier measurements, through carbon nanotube addition a 15–37% improvement of mechanical properties (elastic modulus and strength) can be achieved in comparison to other carbonfilled samples [19]. This chapter describes the preparation and examination of ceramic matrix composites realized by the addition of different carbon polymorphs (carbon black nanograins, graphite micrograins, carbon fibers and carbon nanotubes) to silicon nitride matrices. In the following sections, structural, morphological and mechanical characteristics of carbon-containing silicon nitride ceramics are presented.
19.2
Carbon as fugitive additive for porous silicon nitride processing
Several methods are known for the production of porous ceramics, such as partial sintering [8] or using specific sintering additives [9]. An alternative method is to use fugitive additives such as carbon or other carbon-containing materials as additions, which has proven to be a successful technique for the preparation of high-performance porous ceramics [20]. The first technological step is mixing of powders, which can be monitored step by step by infrared observations. Infrared spectroscopy results are presented in Fig. 19.1. Samples from batches with carbon black and graphite additions are characterized mainly by Si–N vibration modes (Figs 19.1(a) and (b). In the case of the batch with excess oxygen (Fig. 19.1(c)), due to surface oxidation of alpha silicon nitride powders, vibration modes of Si–O bonds appeared at 1250 cm–1 and 1100 cm–1. However, some of the vibration modes of the ethanol that serves as mixing agent can be recognized as OH, H2O, CH3 and C = O stretching (ν) and bending (δ) vibrations [21, 22]. After performing the mechanical activation of powder mixtures, rectangular
© Woodhead Publishing Limited, 2006
516
Ceramic matrix composites
Relative intensity
Si–N
Si–O c
CO2
b
a
4000
3000 2000 Wavenumber (cm–1)
1000
0
19.1 FTIR spectra of milling products: (a) resulting mixture with carbon black after milling; (b) resulting mixture with graphite after milling; (c) mixture after milling, batch with oxidized starting powder.
12
Carbon content (wt%)
10 8
Carbon black
Graphite
6 4 2 Oxidized starting powder 0 400
450
500
550 600 650 700 Temperature (°C)
750
800
850
19.2 Samples from batches with carbon black, graphite and oxidized starting powder with different carbon content after oxidation in atmosphere. Average values of four samples are presented.
bars were obtained by dry uniaxial pressing at 220 MPa. The as-obtained samples were oxidized for 2 h. Oxidation at different temperatures resulted in samples with different carbon contents as presented in Fig. 19.2. Carbon
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
517
removal was very fast in the case of the batch with the oxidized surface. The behavior of the carbon black and graphite added silicon nitride matrix was found to be significantly different regarding the oxidation process. In the temperature range from 450°C up to 600°C the carbon black content decreased from 12 wt% to 0.4 wt%. At this stage the graphite content was around 10 wt%. Above this temperature the graphite content also had a tendency to decrease with increasing temperature. Continuing the oxidation process above 800°C, according to weight losses we obtained almost a carbon-free structure. During the oxidation process carbon is released as CO or CO2 gas while the surface of α-Si3N4 in the compacts may also be oxidized, but we have not observed Si–O bonds evolving in the infrared spectra of compacts. A de-sintering process (i.e. decreasing of density) was observed during the two-step HIP sintering process, as shown in Fig. 19.3. The starting points of arrows represent the first sintering step without any pressure applied. The ends (tips) of arrows represent the second sintering step with 2 MPa pressure applied. The de-sintering process was observed for the reference sample and for batches with carbon black (samples containing ~0.1 wt% carbon after oxidation) and graphite (e.g. samples with no graphite content after oxidation at 800°C) as well. A very similar de-sintering phenomenon was described by Hwang et al. in the case of gas-pressure sintering [23]. The decrease of density was ascribed to chemical dissolution of nitrogen into oxynitride
Modulus of elasticity (GPa)
300 Reference silicon nitride
250
With graphite 200 With carbon black 150
100
50 2.0
2.5 3.0 Apparent density (g/cm3)
3.5
19.3 Effect of de-sintering observed during two-step sintering process (without and with pressure applied) on reference sample, two samples with ~0.1 wt% carbon content after oxidation at 800°C from batch with carbon black, and one sample from batch with graphite and with no graphite content after oxidation at 800°C.
© Woodhead Publishing Limited, 2006
518
Ceramic matrix composites
300
BS4 BS3
With graphite
Bending strength (MPa)
250
200 With carbon black
Oxidized starting powder
150
100
50
0 0
1
8
12
0 10 12 0 Carbon content (wt%)
6
10
12
15
19.4 Relation between carbon content and four-point (BS4) and threepoint bending strength (BS3).
melts assisted by high pressure and presence of boron nitride phases. An interesting remark should be added to the de-sintering observation presented in Fig. 19.3. In the case of the reference sample we found a similar decrease of density as in the batches with carbon black and graphite, but the effect on modulus of elasticity is reversed. A comprehensive view on the carbon content effect on strength can be seen in Fig. 19.4. Samples with added graphite show higher values of strength as compared to carbon black and oxidized samples.
19.3
Comparison of silicon nitrides with carbon additions prepared by hot isostatic pressing and pressureless sintering
In addition to dense monolithic ceramics, porous silicon nitrides are gaining more importance in technological applications [24]. Some porous silicon nitrides with high specific surface area have already been applied as catalysis supports, hot gas filters and biomaterials [25]. There is an emerging tendency to facilitate silicon nitride as biomaterial, because of specific mechanical properties that are important for medical applications [25]. Moreover, in a recent study it was shown that silicon nitride is a non-toxic, biocompatible ceramic which has the ability to propagate human bone cells in vitro [25]. Bioglass and silicon nitride composites have already been realized to combine
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
519
their advantageous properties, such as excellent bioactivity, biocompatibility and favorable mechanical characteristics, particularly the required fracture toughness and wear resistance of components that are prerequisites for clinical applications [26, 27]. Porous silicon nitride rods inserted into rabbit femurs have been found to promote bone ingrowth. In this way a possible new application of silicon nitride is suggested, namely to promote an ideal candidate for bone-substituting implants. Details of the composition of the starting powder mixtures and preparation can be seen in Table 19.1. Si3N4 (Ube, SN-ESP), Al2O3 (Alcoa, A16) and Y2O3 (H. C. Starck, grade C) were used as starting powders. For composite processing in addition to batches, carbon black (Taurus Carbon black, N330, average particle size ~50–100 nm) and graphite (Aldrich, synthetic, average particle size 1–2 µm) were added. The powder mixtures were milled in ethanol in a planetary-type alumina ball mill for 3 h. It was found from weight measurements that each batch contained approximately 1 wt% alumina as contamination from balls and jars. The batches were dried and sieved. Green samples were obtained by dry uniaxial pressing at 220 MPa. Before sintering an oxidation was applied at very low heating rates up to 400°C, to eliminate the compaction agent, polyethyleneglycol (PEG), from samples. Hot isostatic pressing (HIP) was performed at 1700°C under a pressure of 20 bar in high-purity nitrogen by a two-step sinter-HIP method using BN embedding powder (Ceramic and Composites Laboratory, Budapest). The heating rate did not exceed 25°C/min. Pressureless sintering (PLS) was performed in a graphite furnace (Thermal Technology GmbH) at 1700°C for 2 hours under 1 bar flowing nitrogen (Ceramic Research Unit, Limerick). The dimensions of the as-sintered specimens were 3.5 × 5 × 50 mm. After sintering, the weight change of the samples was determined. All surfaces of the samples were finely ground on a diamond wheel, and the edges were chamfered. The density of the sintered materials was measured by the Archimedes method. Phase compositions were determined by Philips PW 1050 diffractometer. The morphology of the solid products was studied by field emission scanning electron microscope, LEO 1540 XB. For HIPed samples the elastic modulus, four-point and three-point bending strengths were determined by a bending test with spans of 40 and 20 mm. Hardness measurements were performed in a micro Vickers Model LL, Tukor Tester, by applying 1 kg load. Fracture surfaces of samples prepared by pressureless sintering are shown in Fig. 19.5. As can be observed, the preparation conditions assured a microstructure with pores of 2–5 µm in dimensions even in the case of the reference sample. Using carbon black or graphite addition, the pore size could be increased (Fig. 19.5(a), (b). At the same time, by increasing the carbon addition the morphology (and microstructural aspects) could be changed. Large silicon carbide granules can be observed in the center of Fig. 19.5(b),
© Woodhead Publishing Limited, 2006
Table 19.1 Starting compositions and preparation conditions of sintered samples Treatment*
Sample
PLS
PLS-1 PLS-3 PLS-5 PLS-7 PLS-9
90 90 90 90 90
4 4 4 4 4
HIP
HIP-1 HIP-2 HIP-3 HIP-4 HIP-5
90 90 90 90 90
4 4 4 4 4
Starting powder (wt%) ———————————————— Si3N4 Al2O3 Y 2 O3
*PLS pressureless sintering, HIP hot isostatic pressing. † CB carbon black, G. graphite.
© Woodhead Publishing Limited, 2006
Carbon Ball milling black/graphite (in ethanol) addition time (wt%)†
Sintering conditions —————————————————— Temp. Holding Pressure (°C) time (bar)
6 6 6 6 6
— 1CB 10CB 1G 10G
3 3 3 3 3
h h h h h
1700 1700 1700 1700 1700
2 2 2 2 2
6 6 6 6 6
– 1CB 10CB 1G 10G
3 3 3 3 3
h h h h h
1700 1700 1700 1700 1700
— — — — —
h h h h h
1 1 1 1 1 20 20 20 20 20
Carbon-ceramic alloys
Mag = 1.00 K X
10µm
EHT = 1.00 kV WD = 8 mm
Signal A = SE2 Photo No. = 7605
521
Date: 15 Jun 2004 Time: 9:28
(a)
Mag = 1.00 K X
20µm
Date: 15 Jun 2004 EHT = 2.00 kV Signal A = SE2 Photo No. = 7620 Time: 10:49 WD = 7 mm
(b)
19.5 (a) Sample PLS-5, 10% CB with large pores; (b) sample PLS-9, 10% G with pores and large platelets of SiC.
© Woodhead Publishing Limited, 2006
522
Ceramic matrix composites 1700°C, 1 bar, PLS PLS-1
Relative intensity
PLS-7, 1%G
PLS-9, 10%G
PLS-3, 1%CB
PLS-5, 10%CB
0
10
20
30
40 50 2θ, Cukα
60
70
80
1700°C, 20 bar, HIP HIP-1
Relative intensity
HIP-4, 1%G
HIP-5 10%G
HIP-2, 1%CB HIP-3, 10%CB
0
10
20
30
40 50 2θ, Cukα
60
70
80
19.6 X-ray measurements for samples resulting from PLS and HIP.
and in this connection we found that the structural characteristics had been changed, according to Fig. 19.6. As resulting from XRD measurements, the lines of α-sialon (PDF-JCPDS 42-0251), together with α-Si3N4 (PDF-JCPDS 41-0360), mellilite (Y2Si3N4O3, PDF-JCPDS 30-1460) and small additions of β-Si3N4 (PDF-JCPDS 33-1160), can be observed (sample PLS-1). After 1% carbon addition (Fig. 19.6, samples PLS-3 and PLS-7) the same main structural lines could be observed as in the reference sample (PLS-1). After we increased the carbon to 10% (PLS-5 and PLS-9) the α-sialon and melilite lines disappeared. In this case the main constituents of the microstructures are α-Si3N4, β-Si3N4 and SiC (PDF-
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
523
Three-point bending strength (MPa)
600 0% 1700°C 1 bar 0% 1700°C 20 bar 1%CB 1700°C 1 bar 1%CB 1700°C 20 bar 1%G 1700°C 1 bar 1%G 1700°C 20 bar 10%CB 1700°C 1 bar 10%CB 1700°C 20 bar 10%G 1700°C 1 bar 10%G 1700°C 20 bar
500
400
300
200
100
0 0
0.5
1.0
1.5 2.0 2.5 Apparent density (g/cm3)
3.0
3.5
19.7 Three-point bending strength of reference sample and composites as function of apparent density. At each point the average value of five samples is shown.
JCPDS 31-1231). For samples prepared by the HIP process we found only α-Si3N4 and β-Si3N4 as constituents for all of the structures (Fig. 19.6). A comparison of three-point bending strengths is presented in Fig. 19.7. By increasing the carbon content an increase of porosity and a decrease of bending strength can be observed for PLSed (open symbols) and HIPed samples (filled symbols). Contrary to observations made in hot pressing [28], we found an increase of hardness with an increase in the carbon content (Table 19.2). The scatter (related to density and strength as well) of PLSed samples is smaller than for HIPed samples. In the small density ranges (to 2.7 g/cm3), by changing from PLS to HIP samples with higher densities and higher strengths have been obtained. At higher density levels the change from PLS to HIP is accompanied by lower densities, but with higher strengths.
Table 19.2 Hardness values for samples prepared by PLS Samples
HV (GPa)
PLS-5, PLS-3, PLS-9, PLS-7, PLS-1,
5.6 15.9 21.2 18.0 11.0
10%CB 1%CB 10%G 1%G 0%
© Woodhead Publishing Limited, 2006
± ± ± ± ±
0.9 0.7 0.6 0.8 0.7
524
19.4
Ceramic matrix composites
In situ processing of Si3N4/SiC composites by carbon addition
Several research groups reported a new type of silicon nitride–silicon carbide nanocomposite with improved high-temperature strength and fracture toughness [29, 30]. Niihara and coworkers [31] were the first research group to study the fabrication of Si3N4/SiC nanocomposites in detail. Their production route consisted of hot pressing an amorphous SiCN powder obtained via a chemical vapor deposition process. These composites consist of a submicron-sized Si3N4 matrix that contains a dispersion of nano-sized and globular-shaped SiC particles. The particles are located both at Si3N4 grain boundaries and within Si3N4 grains. Si3N4/SiC nanocomposites exhibit excellent mechanical properties at both ambient and elevated temperatures. Different strategies for the synthesis of silicon carbide–silicon nitride composites from pre-ceramic polymers, polymer-derived powders or other novel methods have been presented [32, 33]. The polymeric precursor route to synthesize ceramic composites offers processing at low temperature and the possibility of optimizing the new composite materials production by tailoring the composition and molecular structure of the polymers. Recently, a low-cost silicon carbide–silicon nitride nanocomposite processing route has been reported [34] in which case, the bulk silicon nitride-based nanocomposite is formed by the carbothermal reduction of SiO2 by carbon in the Y2O3–SiO2 system at the sintering temperature, with SiC nanoparticles as the result. Although the mechanical properties of as-prepared samples should be further optimized, this process seems to be a prospective choice for silicon nitride–silicon carbide nanocomposite production.
19.4.1 Size effects in micro- and nano-carbon added C/Si3N4 composites prepared by hot pressing The powder mixtures (Table 19.3) with carbon black additions were milled in ethanol in a planetary-type alumina ball mill (Fritsch GmBH). For graphiteadded mixtures a highly efficient attritor mill (Union Process, type 01-HD/ HDDM) was employed. This apparatus allowed a higher rotation speed and a contamination-free mixing process, because of silicon nitride parts (tank, arm, balls). Samples for hot pressing (HP, CENTORR Vacuum Industries) were prepared as follows. After 10–4 torr vacuum, at 20°C, nitrogen gas was introduced at 1 atm. The heating rate was 20°C/min up to 1800°C. At 1000°C a uniaxial pressure of 2 MPa was applied and kept at this pressure until 1800°C. At 1800°C, the pressure was applied for 2 h. The samples were cooled together with the furnace (cooling rate was 30°C/min). In Fig. 19.8(a) and Fig. 19.8(c) the fine microstructures of hot-pressed reference samples resulting from ball and attritor milling and sintering are
© Woodhead Publishing Limited, 2006
Table 19.3 Preparation conditions and composition of starting powder mixtures Batch
Composition (wt%) —————————————— Si3N4 Al2O3 Y 2O 3
Carbon black/graphite added to batch (wt%)
Ball milling (ethanol) at 360 rpm
Attritor milling (ethanol) at 600 rpm
Hot pressing 1800°C, 2 h nitrogen atm. (MPa)
CB0% CB1% CB10%
90 90 90
4 4 4
6 6 6
— 1 10
3h 3h 3h
— — —
2 2 2
G0% G1% G10%
90 90 90
4 4 4
6 6 6
— 1 10
— — —
3h 3h 3h
2 2 2
© Woodhead Publishing Limited, 2006
526
Ceramic matrix composites
(a)
(b)
(c)
(d)
19.8 Micrographs of fracture surfaces: (a) reference sample CB0%, (b) sample CB10%; (c) reference sample G0%, (d) sample G10%. Bar: 1 µm for all micrographs.
shown. Both microstructures consisted of submicron grains, though in the case of the reference sample (CB0%) small hexagonal-shaped β-Si3N4 grains had appeared. Indeed, as presented in Fig. 19.9, in the microstructure of sample CB0% a considerable amount of β-Si3N4 lines appeared. A porous microstructure and an enhanced grain (β-Si3N4) growing process can be observed in the CB10% scanning micrograph (Fig. 19.8(b). Graphite platelets were included in the microstructure even after sintering, as shown in Fig. 19.8(d), sample G10%. By adding 1 wt% carbon black the microstructure remained the same as for the reference sample: α-Si3N4 (PDF-JCPDS 410360) and β-Si3N4 (PDF-JCPDS 33-1160) were the main phases (Fig. 19.9, sample CB1%), but with an increase in carbon black (CB10%) SiC lines had appeared (PDF-JCPDS 31-1231). This observation was confirmed by prompt gamma activation analysis (PGAA) as well. This method is based on the detection of prompt gamma rays originating from the (n, γ) reaction, and gives average elemental composition of the total volume of the sample [28, 35]. The compositions of some sintered samples are presented in Table 19.4.
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
527
SiC
Relative intensity
c
b
a
0
10
20
30
40 50 2θ, Cukα
60
70
80
Relative intensity
c
b
a
0
10
20
30
40 50 2θ, Cukα
60
70
80
19.9 X-ray diffractograms of sintered samples. Top panel: (a) reference sample CB0%, (b) sample CB1%, (c) sample CB10%. Bottom panel: (a) reference sample G0%, (b) sample G1%, (c) sample G10%.
By comparing the Si/N mass ratios of the monolithic reference sample and the composite sample, the effect of carbon content on the complex sintering process can be determined. The results show that at low pressure in the presence of 10 wt% carbon the Si/N mass ratio had considerably increased (to 1.91 ± 0.06 for CB10%)
© Woodhead Publishing Limited, 2006
528
Ceramic matrix composites Table 19.4 Compositions of sintered (at 2 MPa) reference sample CB0% and CB10% Reference sample CB0% Elements H B C N F Al Si S Cl Y Si/N = 1.49 ± 0.04
Sample CB10% Elements H B C N F Al Si S Cl Y Si/N = 1.91 ± 0.06
Wt%
Rel. unc.%*
0.0048 0.0093 — 36.6853 1.0272 2.3548 54.7314 0.0240 0.0096 5.1402
12.9 1.0 — 2.1 29.1 2.5 2.5 33.1 3.6 1.9
Wt%
Rel. unc.%*
0.0263 0.0178 12.952 24.192 9.4493 2.7066 46.423 — 0.0014 4.2161
5.9 1.0 10.4 2.3 8.4 3.0 2.5 — 24.1 1.9
*Because of high relative uncertainty oxygen was not taken into consideration.
compared to the reference sample (from 1.49 ± 0.04 for CB0%). The Si/N fraction for other samples was close to the theoretical value for an Si3/N4 ratio equal to 1.5. In accordance with X-ray measurements, which show the formation of SiC at low pressure and 10 wt% carbon content, the increase of Si/N ratio revealed a nitrogen exhaust from the structure according to the following reaction taking place around 1435ºC [28]: 3C(s) + Si3N4 (s) → 3SiC(s) + 2N2(g). In the case of graphite additions, α-Si3N4, β-Si3N4 and βY2Si2O7 (PDF 21-1454 JCPDS) lines can be observed (Fig. 19.9, samples G1% and G10%). The SiC phase was not formed in graphite composites. Graphite addition assured an increase of hardness in comparison to sample with carbon black added (Table 19.5). Addition of graphite resulted in more densified samples than for samples containing carbon black (Fig. 19.10). All modulus values have been found to be higher for graphite than for carbon
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
529
Table 19.5 Hardness of composites Batch
Hardness, HV (GPa)
Batch
Hardness, HV (GPa)
CB0% CB1% CB10%
13.07 ± 0.6 9.22 ± 0.7 1.21 ± 0.5
G0% G1% G10%
16.27 ± 0.6 14.72 ± 1.2 8.94 ± 2.1
300
Modulus (GPa)
250 200
2MPa 2MPa 2MPa 2MPa 2MPa 2MPa
+ + + + + +
0% Graphite 1% Graphite 10% Graphite 0% Carbon black 1% Carbon black 10% Carbon black
150 100 50 0 1.8
2.0
2.2
2.0
2.2
2.4 2.6 2.8 Apparent density (g/cm3)
3.0
3.2
Three-point bending strength (MPa)
600 500 400 300 200 100 0 1.8
2.4 2.6 2.8 Apparent density (g/cm3)
3.0
3.2
19.10 Modulus of elasticity and three-point bending strength as functions of density.
black added composites at the same carbon level. As regards bending strengths, although the reference sample has higher strength, for graphite added composites a higher strength was obtained than for carbon black added samples (Fig. 19.10). We found an increase in strength for G1% in comparison to reference G0%. It seems that, because of the high specific surface area of
© Woodhead Publishing Limited, 2006
530
Ceramic matrix composites
carbon black nanograins, agglomeration occurs that causes a more porous microstructure after sintering, with lower mechanical properties.
19.5
Silicon nitride ceramics reinforced with carbon fibers and carbon nanotubes
Carbon nanotubes (CNTs) present exceptional mechanical, superior thermal and electrical properties; therefore, in general, there are high expectations for improvement of quality of carbon nanotube nano- and microcomposites [14–18]. Despite this, only modest improvements have been reported in relation to electrical or mechanical properties of carbon nanotube silicon carbide-, polymer- or metal oxide-matrix composites [36–39]. These works emphasized that research concerning CNT-reinforced ceramic matrix composites should deal with increase of dispersion grade of CNTs in the ceramic matrix, assurance of good CNT/matrix interconnection, and hightemperature protection of CNTs. One of the possible approaches is to modify the shape factor of CNTs [40]. It was shown that an optimization of the stress distribution and the reinforcement of mechanical bonding can be achieved by using non-circular or hollow carbon fibers in an epoxy matrix [41]. In a recent study, high-temperature extrusion has been used to align the CNTs in bulk metal oxide nanocomposites, which could lead to an increase of electrical conductivity [42]. Many efforts have been made to modify the surface properties of carbon nanotubes. Using an acidic treatment in a cc.H2SO4/cc.HNO3 mixture results in nanotubes covered by carboxyl groups at their ends and/ or their shells. These groups can easily be converted into carbonyl chloride groups simply by reacting them with SOCl2. The resulting material is very reactive towards amines, so by choosing the appropriate reactant any kind of functional group can be generated [43]. A simple mechano-chemical functionalization of the CNT surface [44] or application of different coatings [45] may be prospective choices to enhance nanotube–matrix bonding. An alternative way to enhance the interfacial strength is to use a physical activation of CNT by spark plasma sintering (SPS) processing. As already shown, good interface contact was achieved in alumina–CNT composites by SPS [1]. This good contact may be attributed to a surface activation, which is assumed to be due to the external electrical field application in SPS. While the surface activation was shown in different materials, no systematic study has been performed. The surface activation and enhanced densification in field sintering is most noticeable at lower temperatures when the discharge effects are operational. The highest temperatures achieved in the necks provide the highest diffusion rates and thus enhanced matter transport towards the neck area. This is the area in which most of the matter transport is required for sintering. Therefore, field application intensifies the sintering rate. Beyond the electrical discharge stage, the sintering parameters in SPS have a similar
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
531
Table 19.6 Mechanical properties (elastic modulus (E), shear modulus (G), Poisson’s ratio (ν), hardness (HV), and fracture toughness (KIC) of composites sintered by SPS Samples/ SPS
ρ (g/cm3)
E (GPa)
G (GPa)
ν
HV (GPa)
KIC (MPa.m1/2)
Reference CNT/Si3N4
3.23 3.17
326.21 285.73
130.18 115.10
0.25 0.24
20.1 ± 0.9 16.6 ± 0.4
5.2 5.3
effect as in conventional pressure sintering. Diffusion processes and plastic flow are the main contributors to the densification. Diffusion studies indicate the extensive contribution of an applied electrical current (e.g., electromigration) in all sintering stages. In the initial sintering stages, the local discharge processes, which disrupt the surface layers, induce various types of defects that enhance surface and grain boundary diffusion. This way, both densification and grain growth are accelerated. Basic studies showed that field application is likely to enhance grain boundary velocity due to impurity elimination. The manufacture of a ceramic composite generally requires high temperatures. Destruction of carbon nanotubes using a hot-press technique was reported by Flahaut et al. [39]. Therefore, the high-temperature degradation process of CNTs has to be further optimized to achieve proper protection of CNTs in high-temperature processes. Concerning carbon nanotube-reinforced silicon nitride matrices, only a few reports have so far been published [19]. In this case, hot isostatic pressing has been used for composite processing. The carbon nanotubes remained in the microstructure only under low pressures (2 MPa); they connect the silicon nitride grains and produce a 15–37% improvement of the mechanical properties as compared with other carbon-filled samples (Fig. 19.11). Increase of pressure 350 684
Bending strength (MPa)
300
4 point bending strength 3 point bending strength
250 645 200 150 629
644
100 50 0
2,315
2,415
2,42 2,467 Apparent density (g/cm3)
19.11 Carbon nanotube-ceramic composites (684) showing 37% increase in bending strength compared with other carbon-filled samples (629 carbon fiber, 644 carbon black, 645 graphite), from Ref. 19.
© Woodhead Publishing Limited, 2006
532
Ceramic matrix composites
and sintering time resulted in the disappearance of the nanotubes. SPS sintering has a proven ability to sinter powders at lower temperatures than other sintering methods. The CNTs are located mainly in the intergranular places and present good adherence to silicon nitride grains (Fig. 19.12(b)). Pullouts can also be noted on the fracture surface, though as can be seen, the CNTs are connected
(a)
(b)
19.12 Scanning electron micrographs of fracture surfaces of SPS samples: (a) reference sample sintered at 1500ºC for 3 min at 50 MPa; (b) sample with MWNTs dispersed and located between grains, sintered at 1500ºC for 5 min at 100 MPa.
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
533
to each other and can be found in groups and nano- or micron–sized islands even after sintering. Therefore, samples with added CNT are characterized by greater porosity than reference samples, and the local reinforcements noted as pullouts are not enough for significant global strengthening. As can be observed in Table 19.6, the reference samples with higher densities show higher modulus, hardness and toughness. Keeping in mind the complexity of the consolidation process of silicon nitride, more precise control is needed to quantify the contribution of carbon nanotubes to the end properties of composites.
19.6
Concluding remarks
Based on the new types of ceramic matrix discussed above, composites with improved properties can be realized by carbon addition. Silicon nitrides with controlled porosity and in situ silicon carbide/silicon nitride composites were obtained with the help of carbon additions and by tailoring the complex reaction paths within sintering processes. Carbon nanotubes have been successfully applied as ceramic matrix reinforcements. The manufacturing processes have been optimized in order to thoroughly disperse the carbon nanotubes in the matrix, to assure good nanotube–silicon nitride contact, and to keep intact the nanotubes during high-temperature processing. An ideal processing–microstructure–property relationship was developed and demonstrated.
19.7
References
1. Zhan, G.-D., Kuntz, J.D., Garay, J.E., Mukherjee, A.K., Electrical properties of nanoceramics reinforced with ropes of single-wall carbon nanotubes, Appl. Phys. Lett. 83(6), 2003, 1228–1230. 2. Shen, Z., Zhao, Z., Peng, H., Nygren, M., Formation of tough interlocking microstructures in silicon nitride ceramics by dynamic ripening, Nature, 417, 16 May 2002, 266–269. 3. Dobedoe, R.S., West, G.D., Lewis, M.H., Spark plasma sintering of ceramics, Bull. Eur. Ceram. Soc., 1, 2003, 19–24. 4. Sternitzke, M., Structural ceramic nanocomposites, J. Eur. Ceram. Soc., 17, 1997, 1061–1082. 5. Derby, B., Ceramic nanocomposites: mechanical properties, Cur. Op. in Solid State Mat. Sci., 3, 1998, 490–495. 6. Niihara, K., New design concept of structural ceramics – ceramic nanocomposites, J. Jpn. Ceram. Soc., 99(10), 1991, 974–982. 7. Rendtel, A., Hübner, H., Hermann, M., Schubert, C., Silicon nitride/silicon carbide nanocomposite materials: II, Hot strength, creep and oxidation resistance, J. Am. Ceram. Soc., 81(5), 1998, 1109–1120. 8. Arató, P., Besenyei, E., Kele, A., Wéber, F., Mechanical properties in the initial stage of sintering, J. Mat. Sci., 30, 1995, 1863–1871.
© Woodhead Publishing Limited, 2006
534
Ceramic matrix composites
9. Yang, J.-F., Deng, Z.-Y., Ohji, T., Fabrication and characterisation of porous silicon nitride ceramics using Yb2O3 as sintering additive, J. Eur. Ceram. Soc., 23, 2003, 371–378. 10. Yang, J.-F., Yhang, G.-J., Ohji, T., Fabrication of low-shrinkage, porous silicon nitride ceramics by addition of a small amount of carbon, J. Am. Ceram. Soc., 84(7), 2001, 1639–1641. 11. Toriyama, M., Hirao, K., Brito, M.E., Kanzaki, S., Shigegaki, Y., US Patent no. 5,935,888. 12. Hirao, K., Brito, M.E., Toriyama, M., Kanzaki, S., Imamura, H., Hirai, T., Shigegaki, Y., US Patent no. 5,968,426. 13. Kawai, C., Matsuura, T., Yamakawa, A., US Patent no. 5,780,374. 14. Rochie, S., Carbon nanotubes: exceptional mechanical and electrical properties, Ann. Chim. Sci. Mat., 25, 2000, 529–532. 15. Thostenson, E.T., Ren, Z., Chou, T.W., Advances in the science and technology of carbon nanotubes and their composites: a review, Comp. Sci. and Techn., 61, 2001, 1899–1912. 16. Lau, K.T., Hui, D., The revolutionary creation of new advanced materials – carbon nanotube composites, Composites: Part B, 33, 2002, 263–277. 17. Lourie, O., Wagner, H.D., Evidence of stress transfer and formation of fracture clusters in carbon nanotube-based composites, Comp. Sci. and Techn., 59, 1999, 975–977. 18. Lau, K.T., Hui, D., Effectiveness of using carbon nanotubes as nano-reinforcements for advanced composite structures, Carbon, 40, 2002, 1597–1617. 19. Balázsi, Cs., Kónya, Z., Wéber, F., Biró, L.P., Arató, P., Preparation and characterization of carbon nanotube reinforced silicon nitride composites, Mat. Sci. Eng. C, 23(6–8), 2003, 1133–1137. 20. Balázsi, Cs., Cinar, F.S., Addemir, O., Wéber, F., Arató, P., Manufacture and examination of C/Si3N4 nanocomposites, J. Eur. Ceram. Soc., 24 (12), 2004, 3287–3294. 21. Balázsi, Cs., Wéber, F., Arató, P., Investigation of C/Si3N4 nanocomposites, Mat.wiss. u. Werkstofftech, 34, 2003, 332–337. 22. Nakamoto, K., Infrared Spectra of Inorganic and Coordination Compounds, WileyInterscience, New York, 1963. 23. Hwang, S.-L., Becher, P.E., Lin, H.T., Desintering process in the gas pressure sintering of silicon nitride, J. Am. Ceram. Soc., 80(2), 1997, 329–335. 24. Diaz, A., Hampshire, S., Characterisation of porous silicon nitride materials produced with starch, J. Eur. Ceram. Soc., 24, 2004, 413–419. 25. Kue, R., Sohrabi, A., Nagle, D., Frondoza, C., Hungerford, D., Enhanced proliferation and osteocalcin production by human osteoblast-like MG63 cells on silicon nitride ceramic discs, Biomaterials, 20, 1999, 1195–1201. 26. Amaral, M., Lopes, M.A., Silva, R.F., Santos, J.D., Densification route and mechanical properties of Si3N4–bioglass biocomposites, Biomaterials, 23, 2002, 857–862. 27. Guedes e Silva, C.C., Higa, O.Y., Bressiani, J.C., Cytotoxic evaluation of silicon nitride-based ceramics, Mat. Sci. Eng. C, 24, 2004, 643–646. 28. Balázsi, Cs. Cinar, F.S., Kasztovszky, Zs., Cura, M.E., Yesilcubuk, A., Wéber, F., Investigation of hot pressed C/Si3N4 nanocomposites, Silicates Industriels, 69 (7–8), 2004, 293–298. 29. Pezzotti, G., Tanaka, I., Okamoto, T., Si3N4/SiC-whisker composites without sintering aids: III, High-temperature behavior, J. Am. Ceram. Soc., 74(2), 1991, 326–332. 30. Goto, Y., Tsuge, A., Mechanical properties of unidirectionally oriented SiC-whisker-
© Woodhead Publishing Limited, 2006
Carbon-ceramic alloys
31. 32.
33.
34.
35. 36.
37.
38. 39.
40.
41. 42.
43.
44.
45.
535
reinforced Si3N4 fabricated by extrusion and hot-pressing, J. Am. Ceram. Soc., 76(6), 1993, 1420–1424. Niihara, K., Suganuma, K., Nakahira, A., Izaki, K., Interfaces in Si3N4–SiC nanocomposite, J. Mater. Sci. Lett., 9, 1990, 598–599. Sajgalik, P., Hnatko, M., Lofaj, F., Hvizdos, P., Dusza, J., Warbichler, P., Hofer, F., Riedel, R., Lecomte, E., Hoffmann, M.J., SiC/Si3N4 nano/micro composite – processing, RT and HT mechanical properties, J. Eur. Ceram. Soc., 20, 2000, 453–462. Poorteman, M., Descamps, P., Cambier, F., Plisnier, M., Canonne, V., Descamps, J.C., Silicon nitride/silicon carbide nanocomposite obtained by nitridation of SiC: fabrication and high temperature mechanical properties, J. Eur. Ceram. Soc., 23, 2003, 2361–2366. Hnatko, M., Galusek, D., Sajgalik, P., Low-cost preparation of Si3N4–SiC micro/ nano composites by in-situ carbothermal reduction of silica in silicon nitride matrix, J. Eur. Ceram. Soc., (2), 24 2004, 189–195. Révay, Zs., Molnár, G.L., Standardisation of the prompt gamma activation analysis method, Radiochim. Acta, 91, 2003, 361–369. Laurent, Ch., Peigney, A., Dumortier, O., Rousset, A., Carbon nanotubes–Fe–alumina nanocomposites. Part II: Microstructure and mechanical properties, J. Eur. Ceram. Soc., 18, 1998, 2005–2013. Laurent, Ch., Peigney, A., Flahaut, E., Rousset, A., Synthesis of carbon nanotubes– Fe–Al2O3 powders. Influence of the characteristics of the starting Al1.8Fe0.2O3 oxide solid solution, Mat. Res. Bull., 35, 2000, 661–663. Peigney, A., Laurent, Ch., Flahaut, E., Rousset, A., Carbon nanotubes in novel ceramic matrix composites, Ceram. Int., 26, 2000, 677–683. Flahaut, E., Peigney, A., Laurent, Ch., Marliere, Ch., Chastel, F., Rousset, A., Carbon nanotube–metal-oxide nanocomposites: microstructure, electrical conductivity and mechanical properties, Acta Mater., 48, 2000, 3803–3812. Park, S.J., Seo. M.K., Shim, H.B., Effect of fiber shapes on physical characteristics of non-circular carbon fiber-reinforced composites, Mat. Sci. Eng. A, 352, 2003, 34– 39. Wagner, H.D., Nanotube–polymer adhesion: a mechanics approach, Chem. Phys. Lett., 361, 2002, 57–61. Peigney, A., Flahaut, E., Laurent, Ch., Chastel, F., Rousset, A., Aligned carbon nanotubes in ceramics–matrix nanocomposites prepared by high-temperature extrusion, Chem. Phys. Lett., 352, 2002, 20–25. Kiricsi, I., Kónya, Z., Niesz, K., Koós, A.A., Biró, L.P., Synthesis procedures for production of carbon nanotube junctions, Proceedings of SPIE – The International Society for Optical Engineering, 5118, 2003, 280–287. Kónya, Z., Vesselényi, I., Niesz, K., Kukovecz, A., Demortier, A., Fonseca, A., Delhalle, J., Mekhalif, Z., Nagy, J.B., Koós, A.A., Osváth, Z., Kocsonya, A., Biró, L.P., Kiricsi, I., Large scale production of short functionalized carbon nanotubes, Chem. Phys. Lett., 360, 2002, 429–435. Hernadi, K., Ljubovici, E., Seo, J.W., Forró, L., Synthesis of MWNT-based composite materials with inorganic coating, Acta. Mater., 51, 2003, 1447–1452.
© Woodhead Publishing Limited, 2006
20 Silicon nitride and silicon carbide-based ceramics Y Z H A N G, New York University College of Dentistry, USA
20.1
Introduction
Despite the excellent combination of thermal mechanical properties of Si3N4 and SiC ceramics, the wide application of these materials is still hampered by their relatively low fracture toughness (Lange, 1973; Jack, 1976; Swain, 1994; Kleebe et al., 1999). Improvements in fracture toughness of Si3N4 and SiC ceramics are mainly achieved by crack deflection and bridging mechanisms (Fairbanks et al., 1987; Bennison and Lawn, 1989; Kelly et al., 1991; Chen and Engineer, 1999). To enhance crack deflection toughening, high aspect ratio acicular grains and a weak interface between the grains and grain boundary phases are desirable; to enhance crack bridging toughening, high aspect ratio grains with large diameters are required. However, weak interfaces and subsequently the debonding of large elongated grains result in the development of fracture origins and loss of strength of the material. Therefore, to develop a superior material with both high toughness and strength, optimization of the microstructure is essential. Accordingly, toughness and strength have been areas of intensive research over the past two decades. The outcomes have been quite impressive: materials with much improved fracture toughness and strength, exceeding 12 MPa m1/2 and 1 GPa respectively, have been reported (Hirao et al., 1995; Ohji et al., 1995; Becher et al., 1998). Although it has long been recognized that microstructure has a significant effect on the erosion performance of ceramic materials, the role of microstructure in the erosion process is not understood in detail. Most of the previous studies on the effect of microstructure on erosion response used alumina ceramics as the model material. However, alumina ceramics, in general, contain an equiaxed grain morphology and high residual stresses at the grain boundaries and are, therefore, not representative of other engineering ceramics, especially the self-reinforced ceramics. In addition, many of the previous studies were carried out on commercial materials in which there is little scope for independent control of microstructures. Thus in the current research, the role of microstructure in the erosion process has been studied 536 © Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
537
using a variety of nitride and carbide ceramics with different microstructures, including several in-house prepared sialon ceramics. Sialon ceramics are genetically associated with Si3N4, but have greater phase complexity and more degrees of freedom for tailoring of microstructures. This offers an ideal system for studying the role of microstructure on erosion of advanced ceramics.
20.2
Material selection
20.2.1 Sialons α-Sialons are isostructural with α-Si3N4 and can be described by the formula
M vx+ Si12–(m+n)Alm+nOnN16–n
(20.1)
where M is the compensating cation with a valency ν, and m and n are substitution numbers referring to m (Al–N) and n (Al–O) bonds replacing (m + n) (Si–N) bonds in each unit cell (Cao and Metselaar, 1991). x is related to m according to the relationship x = m/ν and has a value ≤2 (Hampshire et al., 1978). Two Ca α-sialon compositions, namely CA1005 and CA2613, with nominal x (= m/2 = n) values of 0.5 and 1.3, respectively, were selected for this study. Composition CA1005 lay inside the single-phase α-sialon forming region of the Ca α-sialon-phase diagram, while composition CA2613 was situated just outside the single-phase region on the Al-rich side. Details of the compositional design and powder preparation procedures of these materials are described in a previous paper (Zhang and Cheng, 2003). In this study, both compositions were first pressureless-sintered (PLS-ed) at 1800°C for 3 h followed by hot pressing (HP) at 1700°C for 1 h. Typical microstructures of polished sections of the Ca α-sialon materials imaged in a SEM using the secondary electron mode are shown in Fig. 20.1. Sample CA1005 contained almost equiaxed α-sialon grains with a fraction of the grains being slightly elongated (Fig. 20.1(a)). The average diameters of the α-sialon grains were 0.52 µm. Image analysis revealed that the volume fraction of grain boundary glass was ~2%. No secondary crystalline phase was detected in these materials according to the XRD analysis. XRD analysis showed that material CA2613 consisted mainly of the α-sialon phase coupled with a small amount of AlN-polytypoid. AlN-polytypoids are AlN defect structures that result from the incorporation of silicon and oxygen atoms into the AlN structure (van Tendeloo et al., 1983; Wood and Cheng, 2000). SEM examination revealed two distinct crystalline phases: the α-sialon phase with a smooth trait and the AlN-polytypoid phase with speckled features (Fig. 20.1(b)). The speckled appearance of the AlN-polytypoid phase was a result of the faster etching rate of this phase compared to that of α-sialon when NaOH etchant was used (Wood and Cheng, 2000). The α-sialon grains appeared mainly in an elongated shape with an average diameter of 0.51 µm
© Woodhead Publishing Limited, 2006
538
Ceramic matrix composites AIN-polytypoid
α-sialon 1 µm (a)
1 µm (b)
20.1 SEM micrographs of the two-stage sintered sialon samples: (a) CA1005, (b) CA2613. Specimens were etched in molten NaOH at 410°C for 10 s.
and an apparent aspect ratio of 4.1. Image analysis revealed that the area fractions of the AlN-polytypoid phase and the intergranular glassy phase in sample CA2613 were ~4% and ~3%, respectively.
20.2.2 Silicon nitride Two gas pressure sintered (GPS-ed) Si3N4 ceramics, namely SN-F and SNC, were prepared by the National Industrial Research Institute, Nagoya, Japan. The starting powders were Si3N4 (E-10 grade, ≥95% α-phase, Ube Industries, Japan), Y2O3 (purity >99.9%, Shinetsu Chemicals, Japan) and Al2O3 (purity >99.9%, Taimei Chemicals, Japan). The composition of the densification additives was 5 wt% Y2O3 + 2 wt% Al2O3. Material SN-F was prepared by cold isostatic pressing the powder mixture followed by gas pressure sintering, while material SN-C was fabricated using the same powder mixture with an additional 5 vol% of elongated β-Si3N4 seeds and was prepared by tape casting, stacking, debinding and gas pressure sintering. The β-Si3N4 seeds were rod-like single crystal particles with a typical diameter of ~0.5 µm and a length of ~2 µm. Both samples were densified at 1850°C for 6 h under a nitrogen gas pressure of 0.9 MPa. The processing procedures employed are described in greater detail in a paper by Hirao et al. (1995). Microstructures of the two GSP-ed Si 3N 4 ceramics are shown in Fig. 20.2. As can be seen, both materials contain elongated reinforcing β-Si3N4 grains. However, the unseeded sample SN-F (Fig. 20.2(a)) displays a fine-grained microstructure containing randomly oriented fine elongated grains. The distribution of its grain diameter appears to lie in a transition region between the distinctly bimodal and the broad monomodal distribution. The average grain diameters of the coarse and fine grains are approximately 0.3 and 0.6 µm, respectively. The seeded sample SN-C (Fig. 20.2(b)) exhibits
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
539
(b)
(a)
10 µm
5 µm Casting direction
20.2 Microstructures of the two GPS-ed silicon nitride materials: (a) SN-F (not seeded, cold pressed), (b) SN-C (seeded, tape cast). Plasma etching highlights the epitaxial growth of β-sialon on β-Si3N4 cores (indicated by the arrows).
a distinct bimodal distribution of grain diameter in which large elongated βgrains are evenly distributed in a matrix of finer β-grains and an amorphous grain boundary phase. More significantly, these large elongated β-grains, or whiskers, appear to lie mainly in the tape casting plane and are oriented in the casting direction. The average diameters of the large elongated grains and fine matrix grains are approximately 2 µm and 0.3 µm, respectively.
20.2.3 Silicon carbides The two SiSiC materials, namely SiC-C and SiC-S (supplied by Concord Engineering, Australia and Schunk, Germany, respectively) are two-phase ceramics which consist of high-purity SiC and Si. Fig. 20.3 shows details of (b)
(a)
20 µm
20 µm
20.3 Microstructures of the two siliconized silicon carbide materials, (a) SiC-C and (b) SiC-S, observed with reflected light under an optical microscope, showing different reflective indexes between Si (light) and SiC (dark).
© Woodhead Publishing Limited, 2006
540
Ceramic matrix composites
the microstructures of polished sections of the as-received materials imaged in an optical microscope using reflected light. The light Si phase in Fig. 20.3 is due to the higher reflectivity of Si in comparison to that of SiC (Lee and Rainforth, 1994). As can be seen, both materials possess a duplex microstructure with angular shaped SiC grains of a bimodal size distribution evenly dispersed in a matrix of fine β-SiC, formed from reaction of the carbon with liquid Si, and free silicon. The average grain size is approximately 50 µm and 6–10 µm for large and small SiC grains, respectively, in material SiC-C. The corresponding data for material SiC-S are 30 µm and 3–4 µm. The volume fractions of large SiC grains, small SiC grains and free Si, as determined using image analysis, are approximately 49%, 35% and 16%, respectively, for material SiC-C, and 58%, 31% and 11%, respectively, for material SiCS. Note that the volume fraction of small SiC grains includes the original fine-grained α-SiC particles as well as newly reacted elongated β-SiC grains. The microstructure of the SiSiC material reflects an interesting processing history. It involved the mixing of α-SiC particles, usually with a bimodal size distribution, with carbon and a thermosetting resin to form a green body. The green compact was then charred to carbonize the resin binder and to drive off the volatiles. Finally, the resulting porous body was infiltrated with molten Si at temperatures greater than 1500°C under either vacuum or an inert atmosphere. Liquid Si penetrated the porous body by capillary force and reacted with the carbon to form fine-grained β-SiC grains, epitaxial βSiC deposits on the α-SiC grains as well as large β-SiC grains (Lee and Rainforth, 1994). The reacted SiC along with the residual Si bonded the body together to form a final product with good strength.
20.3
Material characterization
20.3.1 Property evaluation The bulk densities of all the materials were determined using Archimedes’ method (AS 1774.5, 1979). The Vickers indentation technique was used to measure the hardness in each case. The applied load in the Vickers hardness tests was 10 kg for silicon nitrides and sialons. However, using the same load produced severe lateral cracking in silicon carbides around indents, which prevented the accurate measurement of the diagonals of indents. Therefore the load was reduced to 0.3 kg for silicon carbide samples. Fracture toughness of these materials was also estimated using the Vickers indentation technique by measuring the well-developed radial cracks emanating from the four corners of the indent. The indentation load was 0.3 kg for silicon carbides, 10 kg for sialons, and 20 kg for silicon nitrides. The reason for using different loads for the different materials was to produce welldeveloped radial cracks of length 2c which were twice as long as the diagonal
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
541
Table 20.1 Physical and microstructural properties of target materials Material
Density* (kg/m3)
Porosity§ (vol%)
Grain diameter (µm)
CA1005, Ca α-sialon, PLS-ed + HP-ed CA2613, Ca α-sialon, PLS-ed + HP-ed SN-F, Silicon nitride without seeds
3150 3208 3247
~2 <1 <1
SN-C, Silicon nitride with seeds SiC-C, Reaction bonded silicon carbide SiC-S, Reaction bonded silicon carbide
3238 3020 3101
<1 0 0
0.52 0.51 0.3/0.6, nearly bimodal 0.3/2, bimodal 8/50, bimodal 4/30, bimodal
*The bulk density of target materials was measured using the water immersion method. §The porosity of the sialon ceramics was determined using a combination of the water immersion method and the image analysis technique. The porosity of the rest of the materials was either obtained from the manufacturers data sheet (SiC-C and SiC-S) or quoted from literature (SN-F and SN-C) (Becher et al., 1998).
length 2a of the indent (Anstis et al., 1981). In addition, since the orientation of the large elongated β-grains, or whiskers, in sample SN-C is highly directional, being approximately aligned in the tape casting direction (Fig. 20.2(b)), toughness measurements were carried out in directions both parallel and perpendicular to the orientation of the whiskers. It is important to note that in many of the test materials, fracture toughness is not unique, but rather displays a strong dependence on the crack size, the so-called T-curve or Rcurve behaviour (Lawn, 1993). The values reported here thus correspond to the plateau toughness of the T-curve. The physical and mechanical properties of the materials studied in this work are presented in Tables 20.1 and 20.2.
20.3.2 Erosion tests Samples for erosion tests were cut and machined to dimensions approximately 24 mm diameter × 5 mm (sialons), 25 mm × 25 mm × 5 mm (Si3N4), and 25 mm × 25 mm × 10 mm (SiC). The surfaces to be eroded were prepared by finish grinding with 800-mesh SiC abrasive paper. Erosion tests were performed at room temperature in a gas-blast type erosion test rig described in detail elsewhere (Zhang et al., 2000). Mild steel was employed as the control material in each test. The test conditions used are as follows: Sample to nozzle distance: 13.8 mm; Particle velocity 20 m s–1; Erodent particles SiC (particle size, d50 = 388 µm) Impact angles 30° and 90° Dosage ~200 g for 30° impact, ~100 g for 90° impact
© Woodhead Publishing Limited, 2006
Table 20.2 Mechanical properties of target materials Sample
Hardness (GPa)a
CA1005 CA2613 SN-F SN-C
18.6 18.3 16.4 16.7
SiC-C SiC-S
20.5 ± 0.3 22.7 ± 0.1
± ± ± ±
0.4 0.2 0.3 0.2
a
Toughness (MPa m1/2)b 4.3 5.6 5.9 7.9 5.9 2.4 2.4
± ± ± ± ± ± ±
0.4 0.4 0.1 0.5¶ 0.2§ 0.6 0.3
Young’s modulus (GPa)c
Four-point flexure strength (MPa)c
Comments
235–245 235–245 304 312
— — ~1000 ~1400† ~700‡ 230–300 200–300
Mainly equiaxed grains, aspect ratio ~2 Elongated grains, aspect ratio ~4 Elongated grains 12 MPa m1/2(KIC, SEPB method)¶ 7 MPa m1/2(KIC, SEPB method)§ Mainly equiaxed grains Mainly equiaxed grains
350–400 300–390
The applied load in Vickers hardness tests was 0.3 kg for silicon carbides, 10 kg for silicon nitrides and sialons. The applied load in fracture toughness determination was 0.3 kg for silicon carbides, 10 kg for sialons, and 20 kg for silicon nitrides. c The values of Young’s modulus and four-point flexure strength of target materials were obtained from literature and suppliers. ¶ Toughness in the direction perpendicular to the orientation of the large elongated β-grains (whiskers). § Toughness in the direction parallel to the orientation of the large elongated β-grains (whiskers). † When tensile stress was applied parallel to the tape casting direction (whisker orientation). ‡ When tensile stress was applied perpendicular to the tape casting direction (whisker orientation). b
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
543
Mass loss from the samples was measured using an analytical balance with an accuracy of ±0.1 mg. Wear volume was calculated from the mass loss and the bulk density of each material. Cumulative volume loss was plotted as a function of the amount of erodent impacting on the surface. The steady state erosion rate, defined as the volume loss from the specimen per unit mass of erodent used, was determined from the slope of the linear part of the plot of volume loss against mass of erodent. The eroded surfaces of all target materials were examined using a JEOL FE6300 scanning electron microscope (SEM) equipped with a field emission gun. Prior to examination, the specimens were ultrasonically cleaned in ethanol for 5 minutes and then sputter coated with carbon to prevent charge accumulation on the samples during examination. The accelerating voltage used was 10 kV. To facilitate comparison of damage sustained by different samples under different impingement angles, the micrographs, unless specified, were taken at or near the center of the erosion crater.
20.4
Erosion response
20.4.1 Erosion performance The steady state erosion rates of the target materials followed by SiC erosion at 30° and 90° impact are shown in Fig. 20.4. Erosion rates for all samples were higher for 90° impact compared with 30°. This is consistent with the 120
Erosion rate ∆E (m3/kg) × 1010
30° 90° 100
80
60 40
20 I II 0 CA1005
CA2613
SN-F
SN-C
SiC-C
SiC-S
20.4 Steady state erosion rates of all target materials eroded by SiC particles at impingement angles of 30° and 90°. Solid bars labeled I and II indicate the erosion rates of material SN-C corresponding to 30° erosion in the direction perpendicular and parallel, respectively, to the whisker orientation.
© Woodhead Publishing Limited, 2006
544
Ceramic matrix composites
expected trend for brittle materials (Sheldon and Finnie, 1966). The ranking of the materials, in descending order of erosion resistance, was the Ca αsialon ceramics, the two types of self-reinforced silicon nitride materials, and the two SiSiC materials, irrespective of the impact angle. Note that, for the two self-reinforced Si3N4 materials, SN-C exhibited a higher erosion rate than SN-F, while for the two sialon ceramics, material CA1005 possessed a higher erosion rate than CA2613. In order to elucidate the effect of the whisker orientation on the erosion behavior of material SN-C, erosion tests were carried out in directions both parallel and perpendicular to the whisker orientation. It is apparent that in the highly directional whisker-reinforced silicon nitride material, solid particle erosion in the direction parallel to the whisker orientation resulted in a faster rate of material removal compared to that in the perpendicular direction (Fig. 20.4).
20.4.2 Examination of eroded surfaces Sialon The representative features of the eroded surfaces of the two sialon ceramics are shown in Figs 20.5–20.7. The damaged surfaces of sample CA1005 generated by erosion using SiC erodent at normal impact are presented in Fig. 20.5. The damage patterns consisted mainly of grain ejection with some limited plastically deformed materials, indicating that the dominant mechanism of material removal involved intergranular microfracture leading to dislodgment of individual grains. The micrographs in Fig. 20.6 show the types of damage produced by erosion at normal impact in sample CA2613. Fig. 20.6(a) depicts the general view of eroded surfaces of sample CA2613, while Fig. 20.6(b) is the higher magnification views of regions containing faceted grain morphology. It can 1 µm
20.5 SEM micrograph of steady state erosion surface of sample CA1005 after erosion with SiC at normal impact.
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics (a)
(b)
5 µm
1 µm
545
20.6 Surface morphology of sample CA2613 following erosion using SiC particles at 90° impact angle: (a) low-magnification general microstructure of the eroded surface; (b) higher-magnification image of regions with grainy morphology.
4 µm
20.7 Surface morphology of sample CA2613 following erosion using SiC particles at 30° impact angle. The particle impact direction is from top to bottom of the micrograph.
be seen that the damaged surface of sample CA2613 contained mainly plastically smeared materials with some small-scale grainy regions (Fig. 20.6(a)). High-magnification SEM observation revealed that the process of material removal in sample CA2613 involved the propagation and intersection of microcracks of grain-facet dimensions incorporating both transgranular and intergranular fracture (indicated by arrows in Fig. 20.6(b)). The micrograph in Fig. 20.7 shows the types of damage produced by erosion using SiC erodent at 30° impact in sample CA2613. As can be seen, damage features consisted mainly of plastically deformed materials originating from the cutting and ploughing actions of the hard, sharp SiC particles. Isolated brittle-fracture regions, characterized by faceted grainy morphology, were also observed.
© Woodhead Publishing Limited, 2006
546
Ceramic matrix composites
Silicon nitride Erosion damage in materials SN-F and SN-C has been investigated in detail in a previous paper (Zhang et al., 2005) to which interested readers are referred. Here we outline the critical findings. It was found that, in material SN-F, damage features induced by 30° and 90° impact were similar; both indicated brittle fracture and plastic deformation of the materials, except that the 30° impacts produced a higher area fraction of plastically deformed zones with long ploughing trajectories. In the brittle-fracture region, intergranular fracture and transgranular fracture as well as the smearing of exposed surface grains were observed. In addition, some wear debris, typically submicrometer in size and irregular in shape, were seen on the worn surface, suggesting that they were derived either from the fine, more equiaxed grains or from the fragmentation of the large, more elongated grains. Upon repeated impacts, the surface grains along with some loosely bonded wear debris were crushed and smeared to form highly deformed flakes. At the same time, crack networks also developed in the target material. Subsequent impacts resulted in fracture of the underlying silicon nitride grains as well as the spalling or fragmentation of the smeared flakes, leaving a brittle-fracture zone with faceted morphology. Damage patterns observed in material SN-C showed a strong dependence on the impact direction at 30° erosion. In the direction parallel to the tape casting direction (Fig. 20.8(a)), damage sustained by the reinforcing whiskers was mainly intergranular fracture with a few incidences of transgranular fracture, resulting in pullout of the whiskers, or at least large portions of the whiskers. In addition, microfracture of the fine matrix grains was also seen. However, in the direction perpendicular to the tape casting direction (Fig. 20.8(b)),
(a)
(b)
A
B
1 µm
1 µm
20.8 Worn surfaces of material SN-C after 30° SiC erosion with different impact directions. Arrows indicate the tape casting direction, while the particle impact direction is from top to bottom of the micrographs. (a) Parallel to the tape casting direction; (b) perpendicular to the tape casting direction. Arrows labeled A show the transgranular fracture, while arrows labeled B indicate the intergranular fracture (after Zhang et al., 2005).
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
547
more wear debris was observed than that resulting from erosion in the parallel direction. Although pullout still remained one of the dominant damage features sustained by the reinforcing whiskers, an increased incidence of microchipping of the whiskers as a result of the impact was clearly evident. Many parallel whiskers were fractured into a number of sections in the area of contact with the impinging particles, exhibiting a severely crushed morphology consisting of many fragments. The eroded surface resulting from 90° impact exhibited a higher area fraction of brittle-fracture zones compared to that produced by 30° impacts. Pullout and transgranular chipping of the whiskers as well as microfracture of the fine matrix grains were observed. Silicon carbides In gas-blast erosion, solid particles are accelerated along a parallel walled nozzle, but exit the nozzle as a diverging plume. The distribution of particle trajectories within the plume is well defined (Shipway, 1997), and the probability of a particle striking the target becomes smaller as the angle of particle trajectory with respect to nozzle axis increases. Thus the examination of the edge of the erosion crater permits the investigation of the damage induced by single particle impact. The early stage erosion surface of material SiC-C resulting from SiC erosion at 90° impact is shown in Fig. 20.9. A few isolated pits similar to those shown in Fig. 20.9(a), marked by arrows, were found in these regions. Interestingly, detailed examination revealed that damage features associated with individual impact sites were quite different. In the impact crater marked
(b)
(a)
SiC
Si
B Si A 10 µm
10 µm
20.9 SEM micrographs showing surface morphology of isolated particle impacts on material SiC-C following 90° SiC erosion: (a) secondary electron (SE) image; (b) back-scattered electron (BSE) image where atomic number contrast between Si (light) and SiC is perceived. The images were taken near the edge of the erosion crater.
© Woodhead Publishing Limited, 2006
548
Ceramic matrix composites
by the arrow labeled A, classical chipping or conchoidal fracture resulting from lateral cracking was evident. In contrast, in the nearby crater labeled B, dislodgment of fine SiC grains was observed. A back-scattered electron (BSE) image of the same area in Fig. 20.9(a) is shown in Fig. 20.9(b) where the bright phase indicates the regions rich in Si. The light grey phase presents the two-phase SiC–Si region while the dark grey phase, highlighted by the frame, shows the large SiC grains. As can be seen from Fig. 20.9(b), conchoidal fracture from lateral crack chipping occurred within individual coarse SiC grains, while the grain dislodgment took place in the two-phase region. In addition, small-scale chipping damage was also observed on the surface of large SiC grains, suggesting that not every impacting particle could effectively produce a significant lateral crack. This is probably due to the threshold effect on lateral crack initiation where only those particles with high kinetic energy are capable of producing lateral cracking or even multiple impacts are necessary to accumulate the requisite levels of stress to generate lateral cracking. The steady state erosion surface of material SiC-C produced by SiC erosion at 90° impact is shown in Fig. 20.10. The worn surface was much rougher than that resulting from early stage erosion. Areas of smeared-looking or deformed material were also observed in addition to the brittle-fracture regions (Fig. 20.10(a)). Furthermore, in the brittle-fracture zones, two distinct types of damage were present: (1) large-scale transgranular chipping resulting from the link-up of lateral cracks; and (2) dislodgment of the fine SiC grains and occasional small-scale transgranular chipping of the elongated β-SiC grains, highlighted by elliptical frames in Fig. 20.10. The BSE image of the same area in Fig. 20.10(a) is shown in Fig. 20.10(b). The effect of microstructure on erosion damage is immediately evident. The erosion damage of large SiC grains was predominately transgranular chipping
(b)
(a)
SiC
10 µm
10 µm
20.10 SEM micrographs showing surface morphology of material SiC-C following 90° SiC erosion: (a) SE image, (b) BSE image. The images were taken near the center of the erosion crater, where steady state erosion occurred.
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
549
of lateral cracks (see the dark grey regions), indicated by the frame with dashed lines, in Fig. 20.10(b). On the other hand, the damage features observed in the two-phase region were grain dislodgment, limited small-scale transgranular chipping as well as highly plastically deformed materials, as seen in regions containing the scattered bright Si phase in Fig. 20.10(b). The deformed material is believed to contain free Si as well as fine SiC debris and thus should have a bright appearance relative to SiC grains under the BSE mode. However, such differences are often diminished due to the rough erosion surface and the non-uniform distribution in thickness of the surface deformed materials. Steady state erosion surfaces of material SiC-S generated by SiC particles at 30° and 90° impact are shown in Figs 20.11(a) and 20.11(b), respectively. Remarkably, damage features were very similar in the two cases, both involving transgranular chipping associated with lateral cracks and intergranular fracture of fine SiC grains as well as plastic smearing of the deformed materials. Additionally, there was a noticeable lack of ploughing marks on the 30° eroded surface, suggesting that the plastic cutting mechanism was suppressed even when the hard, sharp SiC erodent was used. The Vickers indentation impressions of the two SiSiC materials, at least at 10 kg load or greater, were surrounded by regions of widespread lateral cracking and associated surface chipping (Fig. 20.12(a)). Figure 20.12(b) is a BSE image showing the damaged surface located at the edge of an indentation impression in material SiC-S, while Fig. 20.12(c) is a higher-magnification view of Fig. 20.12(b) illustrating a typical trajectory of the indentation induced crack. As can be seen, the advancing crack tip propagates through the SiC grains with no sign of deflection or bridging, indicating a low fracture toughness of these materials (Table 20.2).
(a)
(b)
10 µm
10 µm
20.11 Steady state erosion surfaces of material SiC-S following SiC erosion at (a) 30° impact, and (b) 90° impact. The particle impact direction for 30° impact is from top to bottom of the micrograph.
© Woodhead Publishing Limited, 2006
550
Ceramic matrix composites (b)
(a)
30 µm
50 µm
(c)
15 µm
20.12 SEM micrographs of Vickers indentation-induced damage in material SiC-S (peak load 10 kg): (a) SE image, (b) and (c) BSE images.
20.5
Microstructure and mechanical properties
It has been long recognized for Si3N4- and SiC-based materials that there exists a close relationship between microstructural parameters and mechanical properties (Becher et al., 1993; Padture and Lawn, 1994; Hoffmann, 1994; Zhao et al., 1997; Chen and Rosenflanz, 1997; Becher et al., 1998; Ellen et al., 1998; Kleebe et al., 1999). Compared to Si3N4 and SiC, sialon ceramics have greater phase complexity and more degrees of freedom for tailoring of microstructures and consequently mechanical properties. Thus, the possible microstructural tailoring of sialon ceramics means that these materials offer an excellent modeling system in any investigation of the effect of various microstructural aspects on the mechanical properties and the erosion behavior of ceramic materials.
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
551
The microstructural–mechanical properties relation of the Ca α-sialon materials has been studied in detail previously (Zhang et al., 2001; Zhang and Cheng, 2003). Some interesting trends were observed. It was found that the hardness of the materials decreases as the x-value increases. The decrease in the amount of α-sialon phase and the increase in the intergranular glass content with increasing x-value may account for this. It was also found that a high α-sialon content coupled with low intergranular glass and pore contents gave an optimized hardness (Zhang et al., 2001; Zhang and Cheng, 2003). In contrast, the apparent aspect ratio of the α-sialon grains appeared to have little influence on the hardness value of these materials. As tabulated in Table 20.2, despite a dramatic difference in the grain aspect ratio between the HP-ed samples CA1005 and CA2613, the hardness values of the two materials are virtually the same. It may be argued that the relatively high porosity in sample CA1005 can result in a decline in hardness; however, the higher glass content in sample CA2613 is also known to inhibit hardness. Considering that the porosity and the glass content in these materials are quite low, being less than 2 and 3 vol%, respectively, it is reasonable to conclude that the grain aspect ratio in α-sialon ceramics has little impact on their hardness values. The present finding is consistent with previous observations where α-sialons containing various amounts of elongated grains with very different aspect ratios possessed almost identical hardness values (Chen and Rosenflanz, 1997; Kim et al., 2000). On the other hand, a coarse grain size and a high aspect ratio can both give rise to the fracture toughness. The effect of grain aspect ratio on the fracture toughness is evidenced in HP-ed samples CA1005 and CA2613 where fracture toughness increases as the grain aspect ratio increases. The toughening effect observed in this study is mainly attributed to the crack bridging mechanism. However, to further improve the toughness, cracks must propagate along grain interfaces rather than through the grains. This interface debonding process appears to be governed by the chemistry of the oxynitride glass at the grain boundaries (Hoffmann, 1994; Kleebe et al., 1999; Becher et al., 1994). The chemistry of the grain boundary glass probably holds the key to understanding why the fracture toughness of α-sialon ceramics is significantly lower than that of silicon nitride. Current best room-temperature values of fracture toughness for α-sialon and Si3N4 ceramics were ~6 MPa m1/2 (Zhao et al., 1997; Chen and Rosenflanz, 1997) and >10 MPa m1/2 (Table 20.2), respectively. The Nd- and Y-stabilized α-sialon ceramics fabricated by Chen and Rosenflanz (1997) and later Kim et al. (2000) exhibited a microstructure consisting of large elongated α-sialon grains imbedded in a matrix of finegrained α-sialon. This microstructure is very similar to that of Si3N4 which was described to have the best properties where large elongated β-sialon grains were evenly dispersed in a matrix of fine-grained β-Si3N4 and an
© Woodhead Publishing Limited, 2006
552
Ceramic matrix composites
amorphous grain boundary phase. However, the fracture toughness of these α-sialons attained only half of the fracture toughness value of silicon nitride ceramics. One possible reason is that in α-sialon, unlike in β-sialon and βSi3N4, both grain boundary glass and α-sialon grains contain identical elements, resulting in a strong bonding between α-sialon grains and the glassy matrix. Therefore to further improve the fracture toughness of α-sialon ceramics, investigations into compositional design of the intergranular glass that leads to a weakened bonding strength between α-sialon grains and the glassy matrix is necessary.
20.6
Microstructure and erosion mechanisms
Traditionally, hardness and toughness are often used to model the erosion behavior of ceramic materials (Evans et al., 1978; Wiederhorn and Lawn, 1979). With the development of indentation fracture mechanics, two elastic– plastic theories have been developed to model the erosion behavior of brittle materials. Evans et al. (1978) considered the dynamic elastic–plastic response of an impacting particle, while Wiederhorn and Lawn (1979) assumed the particle-target contact velocity was slow (relative to sonic velocity), i.e. the quasi-static condition, and the kinetic energy of the erodent particle was absorbed completely by the plastic flow of the target. However, both theories assumed that the lateral cracks were responsible for material removal and both predicted a power-law dependence for the erosion rate, ∆E, on erodent and target properties given by ∆E ∝ v n D 2/3ρ p K C–4/3 H q
(20.2)
where v, D and ρ are the velocity, mean size and density of the erodent particles, and KC and H are the fracture toughness and hardness of the target material. The exponents n, p and q differ in the two models, being 3.2, 1.3 and –0.25 for the dynamic and 2.4, 1.2 and 0.11 for the quasi-static model, respectively. The models indicate that the erosion rate of ceramic materials should have a strong inverse dependence on the fracture toughness, but a much weaker dependence on the hardness of the material. However, such a prediction seems to be contrary to the present findings. Mechanical property evaluations showed that material SN-C exhibited a significantly higher toughness, in the direction perpendicular to the tape casting direction, than material SN-F, while the hardness of the two materials was almost identical (Table 20.2). According to the theoretical models (eq. (20.2)), material SNC should have a better erosion resistance, especially when eroded in a direction perpendicular to the tape casting direction, than material SN-F. Nevertheless, erosion tests showed that material SN-F exhibited better erosion resistance than material SN-C under both 30° and 90° impacts (Fig. 20.4).
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
553
20.6.1 Sialon Erosion tests showed that sample CA1005 exhibited a higher material removal rate than sample CA2613 at both 30° and 90° impacts. SEM examination of surfaces of the two samples generated by erosion at 90° impact revealed that in material CA1005 the dominant material removal mechanism was grain ejection caused by grain boundary microcracking, while in material CA2613 the mechanisms involve combined intergranular fracture of low-aspect-ratio grains and transgranular chipping of large interlocking grains. Consider these results in terms of the different microstructures in the two materials. Samples CA1005 and CA2613 were fabricated from different batches and were both PLS-ed first and then HP-ed at the high temperatures. A low content of intergranular glassy phase, or liquid phase above the eutectic temperature, can hinder the densification process of composition CA1005. As a result, sample CA1005 possesses higher porosity, by a factor of ~3, than sample CA2613 – a composition containing a considerably higher amount of intergranular glass. In material CA1005 a low intergranular glass content coupled with a high porosity can result in a weakened bonding between grains. In addition, pores, especially sintering pores at grain boundary triple points, can be regarded as a stress-concentrating flaw system and hence are preferred sites for microcrack initiation (Lawn, 1993). Therefore, upon impact, microcracks readily initiate at the multi-grain junctions and consequently propagate along the two-grain interfaces and link up to from crack networks. The fine, equiaxed structure of material CA1005 offers limited resistance to crack propagation. As a result, material is removed from the target surface via severe grain dislodgment (Fig. 20.5). In material CA2613, an elongated grain morphology coupled with a small amount of grain boundary glass can give rise to enhanced erosion resistance. Grain boundary glass fills into the pores and facilitates the particle rearrangement during the sintering and also has the potential to absorb or cushion the stress induced from solid particle impact via viscous flow. The elongated grain morphology can have at least two beneficial effects on erosion resistance of ceramic materials (Zhang et al., 2001): it can enhance crack deflection and bridging by blocking the crack path, forcing the crack to detour; and it can hinder grain dislodgment due to an interlocking effect. Both effects suggest that more energy is required to propagate and link up cracks between neighboring impact sites to form crack networks and to remove the interlocked grains from the target surface.
20.6.2 Silicon nitride Detailed discussion in respect to the erosion behavior of self-reinforced or in-situ reinforced silicon nitride materials can be found elsewhere (Zhang et al. 2005). Here we summarize the important arguments.
© Woodhead Publishing Limited, 2006
554
Ceramic matrix composites
In the erosion of silicon nitride with large directional reinforcing whiskers, the reinforcing whiskers are seen to undergo pullout and fragmentation. More significantly, the material removal mechanism of the whiskers is found to vary with the solid particle impacting direction, i.e. in a direction parallel or perpendicular to the tape casting direction. When solid particles attack the surface in the direction parallel to the casting direction, the whiskers are removed primarily by debonding, pullout, or fracture of a large portion of the whisker. Conversely, when erodent particles strike the surface in the direction perpendicular to the casting direction, the whiskers experience an increased incidence of multiple fractures in addition to pullout. In many cases, the multiple fractured sections are chipped away or crushed by the impacting particles. Since erosion in the direction perpendicular to the whisker orientation entails multiple fracture and chipping, which requires higher energy compared to the whisker-dislodgment in the parallel direction, the erosion loss is less in the direction perpendicular to the whisker orientation than that parallel to the whisker orientation. The present study indicates that self-reinforced silicon nitride ceramics displaying a high steady-state toughness might not exhibit good erosion resistance. This finding may be explained by the following reasons. In a fracture toughness test, when the propagating crack tip interacts with the reinforcing whiskers, the weak interface between the whiskers and the matrix promotes crack bridging, resulting in a crack-size dependent toughness. This is particularly the case for SN-C material where highly directional, large elongated grains are well dispersed in a fine, submicrometer grain sized matrix. Such a microstructure ensures a high steady-state fracture toughness owing to the large number of reinforcements and the large length of the debonded interfaces between the whiskers and the matrix. In an erosion test, the situation where a crack grows at equilibrium over a long distance and encounters a number of bridging grains does not exist. Rather, multiple microcracks develop simultaneously in the vicinity of the impact site and linking up of microcracks between the neighborhood impact sites forms a network of short cracks, leading to material loss from the target surface. The weak interface, which is essential for the R-curve behavior, can facilitate the formation of interfacial microcracking and thus result in the dislodgment of the large reinforcing whiskers, leading to a high rate of material removal. More significantly, reinforcing whiskers for SN-C are well dispersed and oriented parallel to the eroding surface. During erosion, the whiskers tend to get plucked out or sheared out of the matrix rather than undergo the partial pullout which leads to crack bridging. On the other hand, the randomly oriented reinforcing grains in material SN-F provide an interlocking effect. Upon erosion, the interlocking hinders the dislodgment of the large reinforcing grains, resulting in a relatively low material removal rate.
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
555
It has been reported in the literature (Srinivasan and Scattergood, 1991; Marrero et al., 1993) that when fracture toughness values corresponding to short cracks relevant to the erosion phenomenon are used, better correlation between the fracture toughness and the erosion results can be obtained. While the R-curve behavior or the operative short-crack toughness may give some indications of the erosion response of brittle materials, the use of shortcrack toughness to quantitatively predict the erosion rate must be made with caution. The key features for a pronounced R-curve behavior are large directional reinforcing whiskers coupled with weak interfaces (Becher et al., 1998). The present findings clearly point out that both directional whiskers and a weak interface could deteriorate the erosion resistance owing to the substantial dislodgment of the large reinforcing grains.
20.6.3 Silicon carbides The erosion mechanism in SiSiC materials apparently involves the linking up of lateral cracks, dislodgment of the fine SiC grains, occasional smallscale transgranular fracture of the fine, elongated β-SiC grains, and plastic smearing of the free Si phase and fine debris. The dominant mechanisms of material removal are, however, conchoidal fracture within individual large SiC grains and dislodgment of the fine-grained SiC particles in the twophase region. Despite the considerable differences in microstructure of the two SiSiC materials, e.g. the size and volume fraction of the bimodal SiC grains as well as the amount of the free Si phase, the erosion rates of the two materials vary by less than 10% for both 30° and 90° impacts. The lack of a microstructural effect on erosion rate under the current experimental conditions suggests that although there exists a distinct difference in mechanisms of material removal between the large SiC grains and the two-phase regions, the actual rate of material removal may be very similar. Indeed, SEM observations of the eroded surfaces showed a uniform type of surface morphology; there is no noticeable preferential wear of particular regions (Figs 20.10 and 20.11). A previous study reported in the literature showed that pure Si erodes much faster, by approximately two orders of magnitude, than single-phase SiC (Routbort et al., 1980). However, preferential erosion of the two-phase SiC–Si region was not observed in the current investigation. This can be attributed in part to the fact that the fine SiC particles and the newly reacted elongated β-SiC grains act as reinforcement to the residual silicon phase, which has effectively improved the erosion resistance of the two-phase regions, and in part to the severity of the SiC erosion which can effectively cause lateral chipping of the large SiC grains. In fact, previous studies of erosion of SiSiC materials using fine Al2O3 erodent particles (Routbort et al., 1980)
© Woodhead Publishing Limited, 2006
556
Ceramic matrix composites
and coal slurry (Shetty et al., 1982) have observed the preferential erosion of the two-phase region resulting in a surface morphology exhibiting mesa-like protrusions of the large SiC grains. This is because both Al2O3 and coal are much softer than SiC and therefore are not capable of causing significant fragmentation of the large SiC grains. As a result, the two-phase SiC–Si region between the large SiC grains erodes relatively fast, leaving the flattopped mesa-like protrusions of the large grains separated by valleys of the two-phase region (Routbort et al., 1980). Finally, no clear evidence of dislodgment of the large SiC grains was revealed in the two SiSiC materials under the current experimental conditions, suggesting that strong bonding exists between the large SiC grains and the mixed SiC–Si phase. This scenario is further supported by SEM observations of the interaction between a propagating crack and SiC grains of these materials, where the advancing crack tip cuts through the large SiC grains with no sign of deflection or bridging (Fig. 20.12(c)).
20.7
Conclusions
The mechanical properties of Ca α-sialon ceramics were found to be closely associated with their microstructures. It was observed that hardness was dependent on a combination of α-sialon content, grain size, porosity and the amount of glassy phase in the material, while the fracture toughness increased with the increasing grain size and grain aspect ratio. However, to further improve the toughness of these materials, a weakened interface between the α-sialon grains and intergranular glass is necessary. The Ca α-sialon and silicon nitride ceramics in the current study best illustrate the effect of microstructural features, such as grain size, grain morphology and grain boundary glass, on the erosion response of ceramic materials. The present results showed that materials consisting of randomly oriented, elongated grains coupled with a small amount of intergranular glass, i.e. CA2613 and SN-F, exhibited the best erosion resistance. This is mainly due to the fact that the elongated grain morphology gives rise to a better resistance to crack propagation/networking and has an interlocking effect that prevents material removal via grain dislodgment. On the other hand, an optimum amount of grain boundary glass can also improve the erosion resistance of ceramic materials by absorbing the impact stress via viscous flow as well as by enhancing the bonding strength between the grains and the matrix. The highly directional whisker-reinforced silicon nitride material, SN-C, showed a lower erosion resistance than the finer grained self-reinforced silicon nitride material, SN-F, and the Ca α-sialon material, CA2613. This can be attributed to substantial pullout of the reinforcing whiskers during the erosion of material SN-C. The significant pullout of the whiskers is a direct
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
557
consequence of the highly directional orientation of the whiskers, which is practically parallel to the erosion surface, and a weak interface between the whiskers and the matrix. The equiaxed sialon material CA1005 displayed a predominantly intergranular mode of fracture, associated with a relatively high erosion rate compared to its elongated counterpart CA2613. This is because composition CA1005 contains a very small amount of intergranular glass, leading to higher porosity and consequently a weak grain boundary in these materials. The weak grain boundary promotes grain boundary cracking under the external stress. In addition, the almost equiaxed grain morphology of these materials has little resistance to grain dislodgment. As a result, the dominant erosion mechanism of composition CA1005 is grain ejection. The relatively low erosion resistance of the two SiSiC materials in comparison to silicon nitride and sialon ceramics is due in part to the lateral chipping of the large SiC grains and in part to the relatively low erosion resistance of the soft Si phase.
20.8
References
Anstis, G.R., et al. (1981), ‘A critical evaluation of indentation techniques for measuring fracture toughness: I, Direct crack measurements’, J. Am. Ceram. Soc., 64(9), 533–8. AS 1774.5 (1979), ‘The determination of density, porosity and water absorption’, Standards Association of Australia, 1–3. Becher, P.F. et al. (1993), ‘The influence of microstructure on the mechanical behaviour of silicon nitride ceramics’, in Chen I-W, Silicon Nitride – Scientific and Technological Advances, MRS Symposium Proceedings, 287, MRS, Pittsburgh, PA, p. 147. Becher, P.F. et al. (1994), ‘Microstructural contribution to the fracture resistance of silicon nitride ceramics’, in Hoffmann M. J. and Petzow G., Tailoring of Mechanical Properties of Si3N4 Ceramics, NATO ASI Series, Series E: Applied Science, 276, Dordrecht, Kluwer Academic, 87–100. Becher, P.F. et al. (1998), ‘Microstructural design of silicon nitride with improved fracture toughness: I, Effects of grain shape and size’, J. Am. Ceram. Soc., 81(11), 2821–30. Bennison, S.J. and Lawn, B.R. (1989), ‘Role of interfacial grain-bridging sliding friction in crack-resistance and strength properties of non-transforming ceramics’, Acta Metall. Mater., 37(10), 2659–71. Cao, G.Z. and Metselaar, R. (1991), ‘α-Sialon ceramics: a review’, Chem. Mater., 3, 242– 52. Chen, I.-W. and Engineer, M. (1999), ‘Model for fatigue crack growth in grain-bridging ceramics’, J. Am. Ceram. Soc., 82(12), 3549–60. Chen, I.-W. and Rosenflanz, A. (1997), ‘A tough SiAlON ceramic based on α-Si3N4 with a whisker-like microstructure’, Nature, 389, 701-4. Ellen, Y.S. et al. (1998), ‘Microstructural design of silicon nitride with improved fracture toughness: II, Effects of yttria and alumina additives’, J. Am. Ceram. Soc., 81(11), 2831–40. Evans, A.G. et al. (1978), ‘Impact damage in brittle materials in the elastic–plastic response regime’, Proc. R. Soc. London, Ser. A, 361, 343–65.
© Woodhead Publishing Limited, 2006
558
Ceramic matrix composites
Fairbanks, C.J. et al. (1987), ‘Crack-interface grain bridging as a fracture resistance mechanism in ceramics’, J. Am. Ceram. Soc., 70, 279–94. Hampshire, S. et al. (1978), ‘α-Sialon ceramics’, Nature, 274, 880–2. Hirao, K. et al. (1995), ‘Processing strategy for producing highly anisotropic silicon nitride’, J. Am. Ceram. Soc., 78(6), 1687–90. Hoffmann, M.J. (1994), ‘Analysis of microstructural development and mechanical properties of Si3N4 ceramics’, in Hoffmann, M. J. and Petzow, G., Tailoring of Mechanical Properties of Si3N4 ceramics, NATO ASI Series, Series E: Applied Science, 276, Dordrecht, Kluwer Academic, 59–71. Jack, K.H. (1976), ‘Review − sialons and related nitrogen ceramics’, J. Mater. Sci., 11, 1135–58. Kelly, J.F. et al. (1991), ‘In situ measurements of bridged crack interfaces in the scanning electron microscope’, J. Am. Ceram. Soc., 74, 3154–7. Kim, J. et al. (2000), ‘Microstructure control of in-situ-toughened α-sialon ceramics’, J. Am. Ceram. Soc. 83(7), 1819–21. Kleebe, H.-J. et al. (1999), ‘Microstructure and fracture toughness of Si3N4 ceramics: combined roles of grain morphology and secondary phase chemistry’, J. Am. Ceram. Soc., 82(7), 1857–67. Lange, F.F. (1973), ‘Relation between strength, fracture energy, and microstructure of hot-pressed silicon nitride’, J. Am. Ceram. Soc., 56(10), 518–22. Lawn, B.R. (1993), Fracture of Brittle Solids, Cambridge, Cambridge University Press. Lee, W.E. and Rainforth, W.M. (1994), Ceramic Microstructures, Property Control by Processing, London, Chapman & Hall. Marrero, M. et al. (1993), ‘Solid-particle erosion of in situ reinforced Si3N4’, Wear, 162– 164, 280–4. Ohji, T. et al. (1995), ‘Fracture resistance behaviour of highly anisotropic silicon nitride’, J. Am. Ceram. Soc., 78(11), 3125–28. Padture, N.P. and Lawn, B.R. (1994), ‘Toughness properties of a silicon carbide with an in situ induced heterogeneous grain structure’, J. Am. Ceram. Soc., 77(10), 2518–22. Routbort, J.L. et al. (1980), ‘The erosion of reaction-bonded SiC’, Wear, 59, 363–75. Sheldon, G.L. and Finnie, I. (1966), ‘On the ductile behaviour of nominally brittle materials during erosive cutting’, Trans. ASME, 88B, 387–92. Shetty, D.K. et al. (1982), ‘Coal slurry erosion of reaction-bonded SiC’, Wear, 79, 275– 9. Shipway, P.H. (1997), ‘The effect of plume divergence on the spatial distribution and magnitude of wear in gas-blast erosion’, Wear, 205, 169–77. Srinivasan, S., and Scattergood, R.O. (1991), ‘R curve effects in solid particle erosion of ceramics’, Wear, 142, 115–33. Swain, M. (1994), Structure and Properties of Ceramics, Weinheim, VCH. van Tendeloo, G. et al. (1983), ‘Characterization of AlN ceramics containing long-period polytypes’, J. Mater. Sci., 18, 525–32. Wiederhorn, S.M. and Lawn, B.R. (1979), ‘Strength degradation of glass impacted with sharp particles: I, Annealed surfaces’, J. Am. Ceram. Soc., 62(1–2), 66–70. Wood, C.A. and Cheng, Y.-B. (2000), ‘Phase relationships and microstructures of Ca and Al-rich α-sialon ceramics’, J. Eur. Ceram. Soc., 20, 357–66. Zhang, Y. et al. (2000), ‘Erosion of alumina ceramics by air- and water-suspended garnet particles’, Wear, 240, 40–51. Zhang, Y. et al. (2001), ‘Influence of microstructure on the erosive wear behaviour of Ca α-sialon materials’, J. Eur. Ceram. Soc., 21(13), 2435–45.
© Woodhead Publishing Limited, 2006
Silicon nitride and silicon carbide-based ceramics
559
Zhang, Y. and Cheng, Y.-B. (2003), ‘Microstructural design of Ca α-sialon ceramics: effects of starting compositions and processing conditions’, J. Euro. Ceram. Soc., 23 (9), 1531–41. Zhang, Y. et al. (2005), ‘Erosion response of highly anisotropic silicon nitride’, J. Am. Ceram. Soc., 88(1), 114–20. Zhao, H. et al. (1997), ‘Elongated α-sialon grains in pressureless sintered sialon ceramics’, J. Eur. Ceram. Soc., 18, 1053–7.
© Woodhead Publishing Limited, 2006
21 Oxynitride glasses – glass ceramics composites S H A M P S H I R E, University of Limerick, Ireland
21.1
Introduction
Oxynitride glasses are special types of silicate or alumino-silicate glasses in which oxygen atoms in the glass network are partially replaced by nitrogen atoms (Hampshire, 2003a). They occur in M-Si-O-N, M1-M2-Si-O-N, M-SiAl-O-N and M1-M2-Si-Al-O-N systems where M, M1 and M2 are modifying cations such as the alkali metals (Li, Na, K), the alkaline earths (Mg, Ca, Ba, Sr) or Y, La and the rare earth lanthanides. Oxynitride glasses exhibit higher refractoriness, elastic modulus, viscosity and hardness compared with the corresponding oxide glasses as a result of the extra cross-linking provided by nitrogen within the glass structure. Oxynitride glasses may be heat treated to form glass-ceramics, effectively multi-phase composites. The process involves heat treatment at two different temperatures, firstly to induce nucleation, then to allow crystal growth of the nuclei. The crystalline phases formed depend on both the composition of the parent glass and the temperatures used for heat treatment. The extent of their formation and growth, the relative amounts and distributions of different phases (including residual glass) and their characteristics will determine the overall properties of the particular composite. The formation of these types of materials and their properties is outlined below. Other investigations have incorporated SiC nano-phase inclusions into oxynitride glass matrices in order to form composites with improved mechanical properties.
21.2
Potential applications
Speciality or ‘new’ glasses and glass-ceramics with novel functions are being used in modern industrial sectors such as optoelectronics, microelectronics, communications technologies, biomedical devices and niche areas of the automotive and architectural sectors. Oxynitride glasses and glass-ceramics and their composites expand the range of ‘new’ glasses and, in view of their 560 © Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
561
durability, higher refractoriness, tailored thermal properties, etc., there are the following areas of potential application: • • • •
Passive coatings on electronic substrates Novel glaze systems for refractory protection High-temperature joining of ceramics and of ceramics to metals Higher temperature range glass-ceramics and composites for structural applications.
21.3
Oxynitride glass/glass-ceramic composites
21.3.1 Oxynitride glasses Various reviews on oxynitride glasses by Leng-Ward and Lewis (1990), Thompson (1992), Sakka (1995) and Hampshire (2003a) have given insights into the understanding of structure–property relationships in these materials. Interest in oxynitride glasses developed following the discovery that silicon nitride-based ceramics contained oxynitride glassy phases at their grain boundaries as a result of the use of oxide sintering additives, which react with silica on the surface of the nitride particles and the nitride itself to form liquid phases that cool as intergranular glasses (Hampshire, 1994). The composition and volume fraction of these glass phases determine the properties of the material, particularly their high-temperature mechanical behaviour (Hampshire and Pomeroy, 2004). The extent of the glass-forming regions in various M-SiAlON systems (M = Mg, Y, Ca, Nd, etc.) has been studied (Hampshire et al., 1985) and represented using the Jänecke prism, with compositions expressed in equivalents instead of atoms or gram-atoms. Thus, equivalent percentage (e/o) of cation M is given by: e/o M =
v M [M] × 100 v M [M] + v Si [Si] + v Al [Al]
(21.1)
where [M], [Si] and [Al] are, respectively, the atomic concentrations of M, Si and Al, VM, VSi, VAl are respectively, the normal valencies for M, Si and Al. Similarly, for the anions, the equivalent percentage (e/o) of nitrogen is given by: (3[N]) × 100 (21.2) 2[O] + 3[N] where [O] and [N] are, respectively, the atomic concentrations of oxygen and nitrogen. Figure 21.1 shows the glass-forming region in the Y-Si-Al-O-N Jänecke prism (Drew et al., 1981). The region is seen to expand initially as nitrogen is introduced and then diminishes above approximately 10 e/o N until the solubility limit for nitrogen is exceeded. e/o N =
© Woodhead Publishing Limited, 2006
562
Ceramic matrix composites 4YN
2Y2O3
4AlN 6
Si3N4
/7 Y2Si2O7 a
3
50
b 10
2Al2O3
eq
%
/2 Si2N2O
N
*Y2O3–SiO2 eutectic
3SiO2
21.1 Glass-forming region in the Y-Si-Al-O-N Jänecke prism on cooling from 1700°C (after Drew et al., 1981).
Further investigations on oxynitride glass formation, structure and properties have shown that oxynitride glasses have higher glass transition temperatures Tg, elastic moduli, viscosities and hardness values compared with the corresponding oxide glasses (same cation ratio) as a result of the extra crosslinking provided by nitrogen within the glass structure (Hampshire, 2003a). Properties typically increase linearly with nitrogen content. Viscosity increases by more than 2 orders of magnitude from a value of 1010.3 Pa.s for the oxide glass to 1012.6 (~Tg) as 18 e/o oxygen is replaced by nitrogen at 950°C. Changes in the cation ratios result in smaller changes in properties than changes in N:O ratio. Ohashi and Hampshire (1991) Ohashi et al., (1995) studied glasses in various LnSiON systems while Ramesh et al. (1997) investigated a range of LnSiAlON glasses. For glasses with the same Ln-Si-Al-O-N ratios but different rare earth lanthanide cations (Ln), properties such as density, hardness, Tg, elastic modulus and viscosity increase linearly with increasing cation field strength (CFS) or decreasing cationic radius. Substitution of one Ln cation by another appears to cause no change in the overall glass structure.
21.3.2 Crystallisation of oxynitride glasses to form multi-phase glass-ceramic composites As with other silicate glasses, oxynitride glasses may be heat treated at appropriate temperatures to crystallise as glass-ceramics (Leng-Ward and Lewis, 1990; Thompson, 1992; Hampshire, 1993). The crystalline phases formed depend on both the composition of the parent glass and the heat-
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
563
treatment process, and the extent of their formation and growth will determine the properties of the particular material. The conventional process to produce a glass-ceramic involves two steps: a first-stage heat treatment of the glass at a temperature just above Tg to induce nucleation, followed by heating to a second higher temperature (the crystallisation temperature, Tc) to allow crystal growth of the nuclei. Many glasses require the addition of a nucleating agent to promote the crystallisation process (Tredway and Risbud, 1984), but oxynitride glasses appear to be self-nucleating because nucleation occurs heterogeneously on FeSi particles as found by Dinger et al. (1988) and Besson et al. (1993) in Y2O3-SiO2-AlN glasses.
21.3.3 Glass-ceramic composites in the Y-Si-Al-O-N system Many studies of crystallisation of oxynitride glasses have concentrated on the Y-Si-Al-O-N system. Thompson (1989) outlined the various crystalline phases that exist in this system as shown in the Jänecke prism in Fig. 21.2. Leng-Ward and Lewis (1985) studied the crystallisation at 1250°C of a series of glasses of constant cation composition (in e/o) 26Y:42Si:32Al with 0 to 30 e/o nitrogen. In the oxide glass, yttrium disilicate is formed as the primary phase on heat treatment. As the nitrogen content increased, gradual Y4 O 6 m 2410
Y4 N4 N-apatite
m > 2700
N-YAM
m 1700-1800
m 1900-2000
Y2SiO5
N-α-wollastonite
m 1980
d ~ 1400
N-melilite m ~ 1900
B
M
Y2Si2O7 m 1775
Y2SiAlO5N D ~ 1200
Si3O6 m 1720
Si3N4 d ~ 1900
YAG m 1930
Si2N2O D ~ 1900
α′ β′ d > 2000
Al4O6 m 2050
Al4N4 s ~ 2200
21.2 Y-Si-Al-O-N Jänecke prism showing crystalline phases and solid solution ranges.
© Woodhead Publishing Limited, 2006
564
Ceramic matrix composites
replacement of yttrium disilicate phase by yttrium aluminium garnet (YAG) was observed and nitrogen was mainly incorporated into Si2N2O. The glass-ceramic transformations in a glass of composition (in e/o) 28Y:56Si:16Al:83O:17N were studied by Ramesh et al. (1998) using both classical and differential thermal analysis techniques. These two methods were found to be in close agreement. Optimum nucleation and crystallisation temperatures were determined in relation to the glass transition temperature. The major crystalline phases present are mixtures of silicon oxynitride and different forms of yttrium disilicate which exists as α, β,γ, δ and y polymorphs depending on heat treatment times and temperatures as shown in Table 21.1. Bulk nucleation was observed to be the dominant nucleation mechanism. The activation energy for the crystallisation process was found to be 834 KJ/ mol. Hampshire et al. (1994) studied the crystallisation behaviour of a glass of composition (in e/o) 35Y:45Si:20Al containing 23 e/o nitrogen which resulted in formation of multi-phase composites, depending on the temperature of heat treatment. B-phase (Y2SiAlO2N), Iw-phase (Y2Si3Al2(O,N)10, i.e. 10 e/o N) and N-wollastonite (YSiO2N) are formed at temperatures below 1200°C, while α-yttrium disilicate (Y2Si2O7), N-apatite (Y5Si3O12N) and YAG (Y3Al2O12) are formed at higher temperatures. At relatively low heat treatment temperatures of ~950–1100°C, the nucleation and growth of N-wollastonite and the intermediate phases B and Iw are kinetically favoured over the more stable equilibrium phases YAG and Si2N2O. In a study of the initial stages of crystallisation of this same glass (Besson et al., 1997), it was found that the creep rate is higher than for the parent glass, since the residual glass in the composite following nucleation has a lower viscosity due to a decrease in yttrium and an increase in impurity cations. Following complete crystallisation, the creep rate was very low, showing that very little glass remains in the composite. Further studies on the crystallisation of B and Iw phase composites from Y-Si-Al-O-N (and Er-Si-Al-O-N) glasses with similar compositions to those Table 21.1 Crystalline phases observed in 28Y:56Si:16Al:83O:17N glassceramic composite after two-stage heat treatment process HT temperature (Tg = 985°C)
Crystalline phases
Tg Tg Tg Tg Tg Tg
α-Y2Si2O7, β-Y2Si2O7, Si2N2O α-Y2Si2O7, β-Y2Si2O7, Si2N2O, YAG*/AlYO3* β-Y2Si2O7, α-Y2Si2O7, Si2N2O, YAG*/AlYO3* β-Y2Si2O7, α-Y2Si2O7, Si2N2O, YAG*/AlYO3* β-Y2Si2O7, Si2N2O, γ-Y2Si2O7, YAG*/AlYO3* β-Y2Si2O7, Si2N2O, γ-Y2Si2O7 YAG*/AlYO3*
+ + + + +
20 K 40 K 60 K 80 K 100 K
*Trace amounts.
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
565
B-phase (002)
20 e/o N
Intensity
17 e/o N
lw-phase (020) 14 e/o N
12 e/o N
10 e/o N
8 e/o N 15
16
17
18 2θ (°)
19
20
21
21.3 XRD traces of the 35Y:45Si:20Al:xO:yN glass-ceramics heat treated at 1050°C showing the major peaks for the B and Iw phases.
reported above have been undertaken (Lemercier et al., 1997; Young et al., 2000; MacLaren et al., 2001; Diaz and Hampshire, 2002; Díaz et al., 2003). Glasses with 5–20 e/o N heat-treated in the range 1000–1250°C show development of different phase assemblages depending on N content, temperature and modifying cation. As shown in Fig. 21.3 (Menke and Hampshire, 2005), at low nitrogen contents (up to 10 e/o), Iw-phase predominates. At 1050°C and a nitrogen content of 12 e/o, B-phase appears in addition to Iw-phase. A further increase in nitrogen content diminishes the Iw-phase, and B-phase predominates, becoming the only phase in the glassceramic at higher nitrogen contents (17–20 e/o N). In the Y-Si-Al-O-N system, although B-phase is the major phase at 1050°C, at 14 to 17 e/o N, Iw-phase predominates at temperatures above 1100°C and traces of β-Y2Si2O7 are observed above 1150°C. At 20 e/o N, B-phase predominates at 1050–1100°C but is balanced by formation of Iw and other phases such as N-apatite at higher temperatures. At crystallisation temperatures high enough to nucleate the stable equilibrium phases (~1250°C), the crystallisation process involves a partitioning of the glass into oxide phases (Y2Si2O7, Y3Al5O12, Al6Si2O13 and Al2O3) and a high nitrogen phase silicon
© Woodhead Publishing Limited, 2006
566
Ceramic matrix composites
oxynitride, with its structure made up of SiON3 tetrahedra. At low crystallisation temperatures (~1100°C), this partitioning is inhibited by the much lower atomic diffusion rates, and the crystallisation process involves the nucleation and growth of phases that are kinetically preferred owing to their compositional similarity to structural units shown to be already present in the parent glass. Properties of Y-Si-Al-O-N glass-ceramic composites Crystallisation of Y-Si-Al-O-N glasses results in glass-ceramic composites which will offer greater refractoriness and better mechanical properties than the starting glass. As shown above, the crystallisation of these glasses is complex, and different composite products form depending on initial glass composition and on the temperature of the heat treatments (Thomas et al., 1982; Leng-Ward and Lewis, 1985; Dinger et al., 1988; Besson et al.,1993; Hampshire et al., 1994; Hampshire, 1994; Menke et al., 2005). Properties of glass-ceramic composites with cation composition (in e/o) 35M = Y or Er:45Si:20Al have been measured (Menke and Hampshire, 2005). A multiphase Y-Si-Al-O-N (14 e/o N) glass-ceramic composite of Iw and YAG formed by heat-treatment at 1200°C exhibits very good mechanical properties with Young’s modulus of 188 GPa. An equivalent composition ErSi-Al-O-N (20 e/o N) glass-ceramic composite containing ErAG, Nwollastonite and N-apatite exhibits a Young’s modulus of 204 GPa. Besson et al. (1993) found that, in general, the thermal expansion coefficient of the Y-Si-Al-O-N glass-ceramic composite containing Y2Si2O7, YAG and Si2N2O varies linearly with temperature between 20 and 800°C and is similar to that of the parent glass. At higher temperatures the thermal expansion coefficient increases slightly. The hardness of Y-Si-Al-O-N glass-ceramic composites was 9–10 GPa. This is quite high for a glass-ceramic and may be compared with the hardness values for β-sialon (15 GPa) and SiC (25 GPa). Unlike many glass-ceramics, the composite exhibited a positive temperature coefficient of electrical resistivity. The fracture behaviour of glasses of composition (in e/o) 35Y:45Si:20Al with 17 e/o N was assessed by means of flexural strength measurements and using a fractographic approach (Hampshire, 2003b). A first-stage heat treatment at Tg + 20°C (960°C) carried out on polished flexural specimens, in order to round out the crack tips produced during the surface-finishing step, resulted in substantial improvements in strength. A two-stage glass-ceramic heat treatment (960°C for 1 hour and 1050°C for 5 hours), which allows crystal growth to form the B-phase, was carried out and a mean flexural strength of 549 MPa was measured for the glass-ceramic.
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
567
21.3.4 Glass-ceramic composites in Ln-Si-O-N and Ln-Si-Al-O-N systems Crystallisation to form composites in the Ln-Si-O-N and Ln-Si-Al-O-N systems has been reported. The disilicate, metasilicate, N-apatite and N-wollastonite phases found in the Y-Si-O-N system (see Fig. 21.2) also exist in the Ln-SiO-N systems (Leng-Ward and Lewis, 1990; Mandal et al., 1992a; Sun et al., 1995; Liddell et al., 1998). The melilite phases, of compositions Ln2Si3O3N4, occur across the entire range of rare earths with the exception of lanthanum (Mandal et al., 1992a). J-phases, of composition Ln4Si2O7N2, are similar, but the La-J-phase does occur even though it has a limited stability (Mandal et al., 1992a). N-apatites (Y5Si3O12N) occur across the whole series of lanthanides; at the low atomic number (low Z) end, there is a triangular solid solution range extending between Ln8(SiO4)6, Ln9.33(SiO4)6O2 and Ln10(SiO4)6N2 (Morrissey et al., 1990; Mandal et al., 1992a; Hampshire et al., 1992). The N-wollastonites of composition LnSiO2N are the least stable of the Ln-Si-O-N oxynitrides, extending from La only as far as Sm, but the La, Ce and Nd end-members have much better thermal stability than the yttrium N-wollastonite (Mandal et al., 1992a). The higher-Z rare-earth cations (La, Ce, Nd) form compounds originally thought to be of composition Ln2O3.2Si3N4, but more recently shown to be of composition Ln3Si8N11O4 (Mandal et al., 1992b); analogues of these do not occur in the Y-Si-Al-O-N system. The extension of all these phases into the five-component Jänecke prism by simultaneous substitution of Si by Al and of N by O has been reported, especially for the wollastonite series. Korgul and Thompson (1989) explored crystallisation of the end-member LaSiO 2N, CeSiO 2N and NdSiO2N wollastonites in considerable detail. Work has also been reported on the sialon U-phases (La3Si3–xAl3+xO12+xN2–x) (Spacie et al., 1988; Leng-Ward and Lewis, 1990; Mandal et al., 1992a; Ramesh et al., 1996) and the sialon W-phases (Mandal et al., 1992a). During crystallisation of Ln-Si-Al-O-N glasses to form glass-ceramic composites, for larger radius cations (La, Nd, Sm, etc.) the W-phase (Ln4Si9Al5O30N) forms, whilst for smaller radius cations (Y, Er, etc.) the B-phase (Ln2SiAlO5N) or disilicates (Ln2Si2O7) are more stable. With respect to disilicates, Liddell and Thompson (1986) evaluated the effects of cation radius on the stability of various yttrium and lanthanide disilicates and endorsed earlier work which shows that, in general, αpolymorphs are stable for large radius cations whilst β-polymorphs are stable for small radius cations. Properties of some Ln-Si-Al-O-N glass-ceramic composites (of original composition 28Ln:56Si:16Al:83O:17N; Ln = Ho, Er, Yb, Y) after heat treatment at 1200°C for 5 hours to form β-Ln2Si2O7 and residual glass are compared with those for the parent glasses in Table 21.2.
© Woodhead Publishing Limited, 2006
568
Ceramic matrix composites
Table 21.2 Comparison of density, ρ, microhardness, Hv, Young’s modulus, E, and shear modulus, G, for LnSiAlON glasses and glass-ceramic composites GlassBefore heat treatment After heat treatment ceramic (1200°C for 5 h) composite ————————————— ———————————————————— system ρ Hv E G ρ Hv E G Crystalline (g cm–3) (GPa) (GPa) (GPa) (g cm–3) (GPa) (GPa) (GPa) phases present HoSiAlON ErSiAlON YbSiAlON YSiAlON
5.05 5.10 5.15 3.73
9.67 10.2 10.15 10.13
145 146 142 145
56 57 55 56
5.06 5.04 5.20 3.78
9.95 10.5 10.4 10.25
166 162 175 175
65 59 68 69
Ho2Si2O7 Er2Si2O7 Yb2Si2O7 Y2Si2O7
Crystallisation of the Ln2Si2O7 results in substantial increases in elastic moduli (tensile and shear), with Yb and Y-Si-Al-O-N glass-ceramics exhibiting Young’s moduli of 175 GPa. For M1-M2-Si-Al-O-N glasses where M1 = La or Nd and M2 = Y or Er, temperatures at which crystallisation exotherms arise have also been determined (Pomeroy et al., 2005) as well as crystalline phases present after the glasses had been heat treated to 1300°C in nitrogen. The results clearly demonstrate that glass properties vary linearly with effective cation field strength of the combined modifiers (M1, M2) which is calculated from the atomic fractions of M1 and M2 and their associated cation field strengths. Glass stability in both the La–Y and La–Er systems reaches a maximum at M1 and M2 fractions of 0.5 because of the relative stability of different oxynitride and disilicate phases with changes in ionic radius. Furthermore, La appears to stabilise the α-polymorph of yttrium disilicate because of combined La-Y ionic radius effects. Studies conducted by Weldon et al. (1996) showed that, for mixed modifier (La:Er = 1:1) glass, devitrification to apatite occurred rather than a two-phase mixture of La–W phase and yttrium disilicate which were the primary devitrification products of the single modifier (La or Er) Si-Al-O-N glasses. Chen et al. (1997), in a study of a mixed modifer La-Y-Si-O-N glass, also observed the crystallisation of apatite.
21.3.5 Glass-ceramic composites in the Nd-Mg-Si-O-N system Crystallisation of glasses in the Nd-Mg-Si-O-N system to form glass-ceramic composites has been investigated. Morrissey et al. (1990) showed that heat treatment at a single temperature resulted in only a small increase in hardness for a 12:24:64 Nd:Mg:Si composition, but two-stage heat treatments resulted in a much higher increase. They found that the optimum nucleation temperature was related to the glass transition temperature of the materials (usually ~ Tg
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
569
+ 40 K). The major phases formed included apatite, which can form a range of solid solution from an oxide form containing Mg to the N-apatite, Nd5Si3O12N. Lonergan et al. (1992) optimised the heat-treatment schedule of Nd-MgSi-O-N glass-ceramic composites with a cationic composition in equivalent percent of 36 (Nd+Mg):64Si and different nitrogen contents. They found as well that a two-stage heat treatment has to be applied to optimise the morphology of phases in the composite and therein the mechanical properties. The nucleation temperature was related to the glass transition temperature. The crystallisation temperatures increased initially with nitrogen content but there seemed to be a levelling off at higher nitrogen contents. These glasses were also used as matrices for glass–SiC composites as described in the next section.
21.4
Oxynitride glass–silicon carbide composites
Friend et al. (1990) were the first to report on oxynitride glass and glassceramic composites reinforced with SiC whiskers and showed that Young’s modulus increases in line with the volume fraction of SiC additions. However, at the temperatures used for hot-pressing, interaction of SiC with the oxynitride glass (represented by SiO2) occurs according to the following: 6SiC + 3SiO2 + 6N2 = 3Si3N4 + 6CO
(21.3)
and with a higher glass:SiC ratio: 3SiC + 3SiO2 + 2N2 = Si3N4 + 3CO + 3SiO
(21.4)
The fracture toughness increased in line with increases in Young’s modulus, E, but not as a consequence of increases in fracture surface energy, γf. Hampshire and Ramesh (1996) and Schneider et al. (1999) studied glasses and glass-ceramics in the Y-Si-Al-O-N-C and Ln-Si-Al-O-N-C systems. Table 21.3 shows the compositions investigated to form glasses by melting at 1700°C under nitrogen atmosphere as for previously reported oxynitride glasses and the phases crystallised after heat treatments at 1200°C for 1 h. The major phases formed were various polymorphs of yttrium disilicate (principally y- and β-Y2Si2O7) and mullite (3Al2O3.2SiO2). The maximum solubility of carbon in these glasses is less than 5 e/o, so no carbide phases were apparent except in the composition containing 10 e/o C. The composition containing 10 e/o N had not crystallised, showing that nitrogen improves glass stability. Rouxel and Verdier (1996) studied the viscoplastic forming and the crystallization ranges of SiC particle reinforced Y-Mg-Si-Al-O-N glass composites produced by hot-pressing at ~1050°C. Crystallization starts beyond 1050°C, with spinel, MgAl2O4, enstatite, MgSiO3, and y- and δ-Y2Si2O7 as
© Woodhead Publishing Limited, 2006
570
Ceramic matrix composites
Table 21.3 Phase assemblage of glass-ceramics with constant cation ratios (16.5Y:56Si:27.5Al e/o) and increasing levels of nitrogen, N, and carbon, C, after heat treatment at 1200°C for 1 h (Hampshire and Ramesh, 1996) Composition (eq %)
Crystalline phase assemblage
16.5Y:56Si:27.5Al:100O 16.5Y:56Si:27.5Al:98O:2C 16.5Y:56Si:27.5Al:98O:2N 16.5Y:56Si:27.5Al:90O:10C 16.5Y:56Si:27.5Al:90O:5C:5N 16.5Y:56Si:27.5Al:90O:10N
y-Y2Si2O7, 3Al2O3.SiO2, β-Y2Si2O7, δ-Y2Si2O7 y-Y2Si2O7, β-Y2Si2O7, 3Al2O3.SiO2 y-Y2Si2O7, β-Y2Si2O7, 3Al2O3.SiO2 β-Y2Si2O7, y-Y2Si2O7, 3Al2O3.SiO2, Y2C3, SiC y-Y2Si2O7, 3Al2O3.SiO2, β-Y2Si2O7 Amorphous
the major phases. A non-Newtonian flow behaviour was observed from 800 to 900°C, i.e. above and below Tg (Tg = 863°C), with n approximately equal to 0.6 (shear thickening behaviour). The viscosity versus temperature curves (deduced from creep tests) show that the temperature must be higher than 950°C for the viscosity to be lower than 1010 Pa s. In the light of these results, the feasibility for viscoplastic forming of a 40 vol% SiC composite was demonstrated by shaping a parabolic shell at 980°C within 30 min. Rouxel et al. (2000) reported on SiC particle reinforced Y-Mg-Si-Al-O-N oxynitride glass composites and found them to have remarkable mechanical properties and to be suitable for viscoplastic forming. In order to better understand the complex nature of flow in these composites, the stress relaxation and creep behaviour were characterized for SiC particle sizes of 3 to 150 µm and volume fractions from 0 to 40% SiC. The viscosity coefficient, as calculated from relaxation data, is very close to the creep viscosity, as determined from the stationary creep regime. The presence of rigid particles results in significant decreases in the relaxation kinetics and creep rates. The smaller the particle size or the higher the particle volume fraction, the lower the flow kinetics becomes. Furthermore, the apparent viscosity of the composite exhibits a strain-hardening behaviour, and a critical strain, at which flow is apparently blocked (depending on both particle size and volume fraction), has been successfully introduced to interpret the data. Baron et al. (2000) studied the fracture behaviour of YMgSiAlON glass– SiC composites; it was assessed by means of flexural strength measurements and a fractographic approach. The composites were produced by hot pressing a mixture of glass and SiC powder in a graphite cell lined with boron nitride, between two molybdenum sheets. The pressure was limited to 15 MPa and the composites were hot-pressed at a sintering temperature of 1040°C for 30 minutes. Then the glass composite was annealed at a temperature of 830°C for 30 minutes. Three-point bending test bars were prepared with different polished surfaces. Some samples were submitted to a flame polishing heat treatment. Results of flexural strength tests are shown in Fig. 21.4 (Baron et al., 2000).
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
571
350 after grinding polished 3 µm polished 1/4 µm flame polished
σr (MPa)
300 250 200 150 100 50
Matrix
3 µm
6 µm
16 µm
31 µm
150 µm
21.4 Dependence of the fracture strength of the YMgSiAlON glass– SiC composites (28 vol% SiC) on the SiC particle size and the surface finish (after Baron et al., 2000).
Fracture originated from the particles and their strength changed with surface finish only when the defects created by machining were bigger than the defects introduced by the particles. The consequence is that composites are less sensitive to machining flaws, but, when the surface finish was good, the strength of the matrix was only increased for small particle sizes. In glass matrix particulate composites, the size of critical defects is related to the size of the particles leading to a decrease of strength when the particle size increases as observed in Fig. 21.4. The increase in strength with increasing volume fraction of particles is shown in Fig. 21.5 (Baron et al., 2000) and is mainly due to the resultant increase in Young’s modulus. For low particle volume fraction, a reduction of the fracture strength was observed after flame polishing compared to the matrix. In this case, the increase of Young’s modulus was not sufficient to compensate for the increase of critical defect size due to the introduction of the particles. The quality of the surface finish influenced the fracture strength of the composites only when the flaws introduced by the polishing became smaller than the flaws introduced by the particles. This means that the larger the particle size the lower the fracture strength sensitivity of the composites to machining flaws. 350
σr (MPa)
300 250 200 150 polished 3 µm flame polished
100 50 Glass
10 vol%
28 vol%
21.5 Dependence of the fracture strength of the composites on the volume fraction of particles (6 µm SiC).
© Woodhead Publishing Limited, 2006
572
21.5
Ceramic matrix composites
Conclusion
Oxynitride glasses are silicate or alumino-silicate glasses in which oxygen atoms in the glass network are partially replaced by nitrogen atoms. As nitrogen increases, glass transition temperature, elastic modulus, viscosity and hardness increase while thermal expansion coefficient decreases. Many studies on crystallisation of oxynitride glasses have been carried out. Some of these have identified suitable two-stage heat treatments for nucleation and growth of crystal phases to form glass-ceramic composites with significant increases in mechanical properties over the parent glass. Overall, the thermal, electrical and mechanical properties of the oxynitride glass-ceramic composites differ from system to system and have to be established for every composition and every heat-treatment schedule applied. However, by using established techniques for processing of glass-ceramics, with a first-stage heat treatment to provide nucleation and a second-stage heat treatment for crystal growth, microstructures and properties of multiphase composites can be optimised. SiC particle reinforced oxynitride glass composites have been investigated and found to have higher values of mechanical properties which are related to the volume fraction of SiC inclusions, provided that SiC–glass reactions can be avoided. These composites have been shown to be suitable for viscoplastic forming.
21.6
References
Baron, B., Lemercier, H., Veyrac, C., Pomeroy, M., Hampshire, S. (2000), ‘Fracture of oxynitride glasses and SiC particulate composites’, Mater. Sci. Forum, 325, 295–301. Besson, J.L., Billiers, D., Rouxel, T., Goursat, P., Flynn, R., Hampshire, S. (1993), ‘Crystallization and properties of a Si-Y-Al-O-N glass-ceramic’, J. Am. Ceram. Soc., 76, C2103–5. Besson, J.-L., Lemercier, H., Rouxel, T., Troillard, G. (1997), ‘Yttrium sialon glasses: Nucleation and crystallisation of Y35Si45Al20O83N17’, J. Non-Cryst. Sol., 211, 1–21. Chen, J., Wei, P., Huang, Y. (1997), ‘Formation and properties of La-Y-Si-O-N oxynitride glasses’, J. Mater. Sci. Lett., 16, 1486–8. Diaz, A., Hampshire, S. (2002), ‘Crystallisation of M-SiAlON glasses to Iw-phase glassceramics: preparation and characterization’, J. Mater. Sci., 37, 723–30. Diaz, A., Dolekcekic, E., Pomeroy, M.J., Hampshire, S. (2003), ‘Effect of composition and processing conditions on the Formation of Y and Er-SiAlON B and Iw Phase Glass-ceramics’, Key Eng. Mater., 237, 247–252. Dinger, T.R., Rai, R.S., Thomas, G. (1988), ‘Crystallization behaviour of a glass in the Y2O3–SiO2–AlN system’, J. Am. Ceram. Soc., 71, 236–44. Drew, R.A.L., Hampshire, S., Jack, K.H. (1981), ‘Nitrogen Glasses’, Proc. Brit. Ceram. Soc., 31, 119–32. Friend, S.J., Piller, R.C., Briggs, A., Davidge, R.W., Lonergan, J.M., Hampshire, S. (1990), ‘Sintering of neodymia-containing oxynitride glasses and glass-ceramics and effects of additions of silicon carbide whiskers’, Silicates Industriels, (Journal of the Belgium Ceramic Society) LV, 303–8.
© Woodhead Publishing Limited, 2006
Oxynitride glasses – glass ceramics composites
573
Hampshire, S. (1993), ‘Oxynitride glasses and glass-ceramics’, in Chen I.W., Becher, P.F., Mitomo, M., Petzow, G., Yen, T-S. (eds), Silicon Nitride Ceramics – Scientific and Technological Advances, Mater. Res. Soc. Symp. Proc., 287, 93–100. Hampshire, S. (1994), ‘Nitride ceramics’, in Swain, M.V. (ed.) Structure and Properties of Ceramics, Materials Science and Technology Series, Vol. 11, Weinheim, VCH, Chapter 3, 119. Hampshire, S. (2003a), ‘Oxynitride glasses – a review’, J. Non-Cryst. Sol., 316, 64–73. Hampshire, S. (2003b), ‘SiAlON glasses, their properties and crystallisation’, Key Eng. Mater., 237, 239–46. Hampshire, S., Pomeroy, M.J. (2004), ‘Effect of composition on viscosities of rare earth oxynitride glasses’, J. Non-Cryst. Sol., 344, 1–7. Hampshire, S., Ramesh, R. (1996), ‘Oxynitride liquids, glasses and glass-ceramics’, invited presentation at 1st International Workshop on Synergy Ceramics, Nagoya, Japan, November, 1996. Hampshire, S., Drew, R.A.L., Jack, K.H. (1985), ‘Oxynitride glasses’, Phys. Chem. Glass., 26, 182–6. Hampshire, S., Flynn, R., Lonergan, J., O’Riordan, A. (1992), ‘Oxynitride glass systems and subsequent glass-ceramic heat treatments’, in Carlsson, R. (ed.), Ceramic Materials and Components for Engines, Proc. 4th Int. Symp. Göteborg, June 1991, London, Elsevier Applied Science. Hampshire, S., Nestor, E., Flynn, R., Besson, J.-L., Rouxel, T., Lemercier, H., Goursat, P., Sebai, M., Thompson, D.P., Liddell, K. (1994), ‘Yttrium oxynitride glasses: properties and potential for crystallisation to glass-ceramics’, J. Eur. Ceram. Soc., 14, 261–73. Korgul, P., Thompson, D.P. (1989), ‘The transparency of oxynitride glasses’, J. Mater. Sci., 28, 506–12. Lemercier, H., Ramesh, R., Besson, J.-L., Liddell, K., Thompson, D.P., Hampshire, S. (1997), ‘Preparation of pure B-phase glass-ceramic in the yttrium–sialon system’, Key Eng. Mater., 132–136, 814–7. Leng-Ward, G., Lewis, M.H. (1985), ‘Crystallisation in Y-Si-Al-O-N glasses’, J. Mater. Sci. Eng., 71, 101–11. Leng-Ward, G., Lewis, M.H. (1990), ‘Oxynitride glasses and their glass-ceramic derivatives’, in Lewis, M.H. (ed.), Glasses and Glass-ceramics, London, Chapman & Hall. Liddell, K., Thompson, D.P. (1986), ‘X-ray diffraction data for yttrium silicates’. Brit. Ceram. Trans. J., 85, 17–22. Liddell, K., Thompson, D.P., Wang, P.L., Sun, W.Y., Gao, L., Yan, D.S. (1998), ‘J-phase solid solution series in the Dy-Si-Al-O-N system’, J. Eur. Ceram. Soc., 18, 1479–92. Lonergan, J., Morrissey, V., Hampshire, S. (1992), ‘Optimisation of heat-treatment schedules for oxynitride glass-ceramics’, Brit. Ceram. Proc., 49, 57–62. MacLaren, I., Falk, L.K.L., Diaz, A., Hampshire, S. (2001), ‘Effect of composition and crystallization temperature on microstructure of Y- and Er-SiAlON Iw-phase glassceramics’, J. Am. Ceram. Soc., 84, 1601–8. Mandal, M., Thompson, D.P., Ekström, T. (1992a), ‘Heat treatment of Ln-Si-Al-O-N glasses’, Key. Eng. Mater., 72-74, 187–203. Mandal, M., Thompson, D.P., Ekström, T. (1992b), ‘Heat treatment of sialon ceramics densified with higher atomic number rare earth and mixed yttrium/rare earth oxides’, Brit. Ceram. Proc., 49, 149–62. Menke, Y., Hampshire, S. (2005), ‘Ln-Si-Al-O-N glass-ceramics: crystallisation and properties’, submitted to J. Eur. Ceram. Soc.
© Woodhead Publishing Limited, 2006
574
Ceramic matrix composites
Menke, Y., Falk, L.K.L., Hampshire, S. (2005), ‘The crystallisation of Er-Si-Al-O-N Bphase glass-ceramics’, accepted by J. Mater. Sci., in press. Morrissey, V., Lonergan, J., Pomeroy, M.J., Hampshire, S., (1990), ‘Crystallisation treatments for neodymia-containing glasses and glass-ceramics’, Brit. Ceram. Proc., 45, 23–9. Ohashi, M., Hampshire, S. (1991), ‘Formation of Ce-Si-O-N glasses’. J. Am. Ceram. Soc., 74, 2018–20. Ohashi, M., Nakamura, K., Hirao, K., Kanzaki, S., Hampshire, S. (1995), ‘Formation and properties of Ln–Si-O-N glasses’, J. Am. Ceram. Soc., 78, 71–6. Pomeroy, M.J., Nestor, E., Ramesh, R., Hampshires (2005), ‘Properties and crystallisation of rare earth SiAlON glasses containing mixed trivalent modifiers: J. Amer. Ceram. Soc., 88(4), 875–881. Ramesh, R., Nestor, E., Pomeroy, M.J., Hampshire, S., Liddell, K., Thompson, D.P. (1996), ‘Potential of NdSiAlON Glasses for Crystallisation to Glass-Ceramics’, J. Non-Cryst. Sol., 196, 320–5. Ramesh, R., Nestor, E., Pomeroy, M.J., Hampshire, S. (1997), ‘Formation of Ln-Si-Al-ON glasses and their properties’, J. Eur. Ceram. Soc., 17, 1933–9. Ramesh, R., Nestor, E., Pomeroy, M.J., Hampshire, S. (1998), ‘Classical and DTA studies of the glass-ceramic transformation in a YSiAlON glass’, J. Am. Ceram. Soc., 81, 1285–97. Rouxel, T., Verdier, P. (1996), ‘SiC particle reinforced oxynitride glass and glass-ceramic composites: crystallization and viscoplastic forming ranges’, Acta Mater., 44, 2217– 25. Rouxel, T., Sangleboeuf, J.C., Verdier, P., Laurent, Y. (2000), ‘Elasticity, stress relaxation and creep in SiC particle reinforced oxynitride glass’ Key Eng. Mater., 171, 733–40. Sakka, S. (1995), ‘Structure, properties and application of oxynitride glasses’, J. NonCryst. Sol., 181, 215–24. Schneider, N.K., Mooney, C., Baron, B., Hampshire, S. (1999), ‘Oxynitride glass composites containing nano-size SiC’, Brit. Ceram. Proc., 60(1), 401–2. Spacie, C.J., Liddell, K., Thompson, D.P., (1988), ‘The U-phase in heat-treated sialon ceramics’, J. Mater. Sci. Lett., 7, 95–6. Sun, W.Y., Yan, D.S., Gao, L., Mandal, H., Liddell, K., Thompson, D.P. (1995), ‘Subsolidus phase relationship in the systems Ln2O3–Si3N4–AlN–Al2O3 (Ln = Nd, Sm)’, J. Eur. Ceram. Soc., 65, 15, 1435–8. Thomas, G., Ahn, C., Weiss, J., (1982), ‘Characterization and crystallization of Y-Si-AlO-N glass’, J. Am. Ceram. Soc., 65, C185–8. Thompson, D.P. (1989), ‘Alternative grain boundary phases for heat treated Si3N4 and βsialon ceramics’, Brit. Ceram. Proc., eds. Davidge, R.W. and Thompson, D.P., Vol. 44, 1–14. Thompson, D.P. (1992), ‘Oxynitride glasses’, in Cable, M. and Parker, J.M. (eds) High Performance Glasses, Glasgow, Blackie and Sons, Chapter 5. Tredway, W.K., Risbud, S.H. (1984), ‘Influence of atmospheres and TiO2 nucleant on the crystallisation of Mg-SiAlON glasses’, J. Mater. Sci. Lett., 4, 31–3. Weldon, L., Pomeroy, M.J., Hampshire, S. (1996), ‘Glasses in the rare-earth sialon systems’, Key Eng. Mater., 118–119, 241–8. Young, W., Falk, L.K.L., Lemercier, H., Peltier-Baron, V., Menke, Y., Hampshire, S. (2000), J. Non-Cryst. Sol., 270, 6–12.
© Woodhead Publishing Limited, 2006
22 Functionally graded ceramics G A N N É, J V L E U G E L S and O V A N D E R B I E S T, Katholieke University Leuven, Belgium
22.1
Introduction
Functionally graded materials (FGMs) are multifunctional materials, which contain a spatial variation in composition and/or microstructure for the specific purpose of controlling variations in thermal, structural or functional properties. Also in the ceramics composites field, a wide range of functionally graded (FG) ceramics are available. Hence, a possible classification of the different classes is made in this chapter. Compared to homogeneous materials a lot of advantages exist. However, there would be little point in developing such a FGM material unless it could compete commercially with existing homogeneous ceramic components currently on the market. A wide variety of processing routes are available for FGMs and it is important to choose the right processing technique for the right application. Therefore, a large part of this chapter deals with the processing of FGM, more specifically for bulk FGMs. The gradient composition in FGMs not only results in a spatial variation in properties but will also generate residual stresses, which will affect the mechanical properties. One of the potential advantages of FG components is the positive influence of compressive residual surface stresses on the strength and wear resistance. A correct design of the gradient for an optimal distribution of the residual stresses is therefore important, as discussed in this chapter. Special attention will be given to structural FGM applications, where the operating conditions are severe. More specifically, alumina/zirconia and WC/ Co FGM components will be discussed.
22.2
Functionally graded ceramics concept
As a new concept in advanced materials research and development, introduced in the mid-1980s, functionally graded materials (FGMs) are defined as materials whose spatial distribution of microstructure and/or composition is tailored and quantitatively controlled in order to achieve an improvement in properties 575 © Woodhead Publishing Limited, 2006
576
Ceramic matrix composites
of the final component (Mortensen and Suresh, 1995; Neubrand and Rödel, 1997; Gasik, 1995). The concept of FGMs derives from the consideration that the service conditions and required materials performance vary with the location in a large number of structural components. Consider for example a turbine blade, which must withstand high non-stationary heat fluxes and centrifugal acceleration. An ideal structure for this application would consist of a tough metal core and a corrosion resistant ceramic at the hot surface of the blade. If the ceramic is directly bonded to the metal, spalling may occur during thermal cycling because of the high thermal stresses at the interface. A graded material with a smooth transition from the ceramic surface to the metal core can avoid the thermo-mechanical stress concentration at the interface, and thus has better performance. There are many other applications where the requirement for properties cannot be attained by a single material (Neubrand and Rödel, 1997). In all these cases, graded materials offer the possibility to combine two materials properties, avoiding most of the disadvantages of a bi-material. Fundamental and applied research in the recent past has clearly shown that the introduction of compositional step and profile gradients in FGMs can be a benefit on many accounts (Suresh and Mortensen, 1997; Munz et al., 1998; Bao and Cai, 1997). In a broad sense, the concept of FGMs is not new. In China, an ancient proverb is used frequently in people’s daily tasks: ‘the best steel only for cutting edges’. If we eliminate its philosophical sense, this proverb from ancient blacksmiths expressed clearly the idea of FGMs. In fact, this idea has been exploited for hundreds of years in iron and steel production (Gasik, 1995). Although it is not difficult to find examples throughout recorded history of using the idea of graded materials, as a systematic engineering approach it is far from being the norm. Instead, current engineering practice generally involves the design of a part using one of the large number of generally mass-produced uniform materials. When a structure requires vastly different materials, different uniform materials are joined along sharp boundaries using a variety of joining or coating methods. The task of the materials engineer is defined as seeking to improve materials, while the task of the structural part designer is using the handbooks of available materials. With the innovative FGM concept, component design and fabrication are based not on a list of existing materials but on a choice of available basic material ingredients and material processes, combined with three-dimensional mechanical analysis of graded structures. These two engineering disciplines are combined to synergistically design both the component and its processing. Here lies a second definition of the functionally graded materials concept, as an approach in engineering rather than a physical entity (Mortensen and Suresh, 1995).
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
577
Toughness
Hardness
Corrosion resistance Tool Turbine blades
Thermal conductivity
Substrate
Electrical insulation
Sensors
Acoustic impedance
Actuators
Piezoelectric properties
Dielectric properties
22.1 Examples of applications of graded materials (Neubrand and Rödel, 1997).
22.3
Classification of FG ceramics
There are a number of possibilities to classify graded materials (Neubrand and Rödel, 1997): • According to the material classes which the graded component combines, e.g. metal/ceramic, polymer/ceramic, metal/metal, ceramic/ceramic, etc. • According to the relative extension of the gradient, i.e. to what extent the gradient is distributed across a component. In functionally gradient coatings and joints, the gradient extends only over a part of the component close to its surface or in the interior. In functionally graded bulk materials, the gradient comprises the entire part. • According to the function in a component. Figure 22.1 gives some examples of applications of graded materials, which allow unusual combinations of properties.
22.4
Processing of FGMs
One characteristic of the fabrication of FGMs is certainly the very wide variety of available processing methods. Functionally graded materials include materials with a gradient in composition, grain size and/or porosity. The general goal of processing of FGMs is to realize a spatial distribution in the microstructure and/or composition in the final part. When selecting the processing method, differences between the properties of the two constituent phases of the FGM are of primary importance. In a compositional FGM, for example, the difference in heat resistance between
© Woodhead Publishing Limited, 2006
578
Ceramic matrix composites
the two phases is a key factor. If the two phases have a significantly different melting point, as in ceramic/metal FGMs, the composition gradient can be formed by producing a porosity gradient preform of the refractory phase subsequently infiltrated by the molten second phase to get a dense final product. If the two phases have a similar melting point, infiltration cannot be used because the skeleton cannot keep its strength during infiltration. The dimensions and geometry of the FGM has to be considered as well. It is feasible to produce FGMs in many systems with thermal coating technologies, but their low efficiency makes them useless for the production of threedimensional bulk FGMs.
22.4.1 Shaping and consolidation processes for FGMs For the fabrication of bulk FGMs, powder metallurgical processing is most economic and suitable for mass production. In order to produce a FGM by conventional powder processing, a green body with the desired gradient in phase volume fraction is first fabricated. After shaping and consolidation, this green body has to be densified by sintering. The gradation methods can be divided into two groups: dry and wet processes (Fig. 22.2) (Neubrand and Rödel, 1997). Dry processes are fast, but in general only allow the generation of step-graded profiles. In wet processing, a drying step is required for the removal of the liquid but continuous mixing is facilitated and smoother continuous gradients may be produced. Furthermore, transport processes occurring in suspensions, e.g. sedimentation and electrophoresis, may be used to produce gradients at low cost. The main challenge associated with powder processing is frequently the densification Powder metallurgy Dry processing Conventional Centrifugal PM PM
Wet processes Wet Centrifugal Slip casting spraying casting
Preparation of mixtures
Continuous mixing of powders
Preparation of slip
Sequential stacking
Continuous stacking
Sequential casting
EPD
Preparation of suspension
Spray
Centrifugation Sedimentation
Predensification Debinding
Dry
Sintering
22.2 Powder processing routes for FGMs.
© Woodhead Publishing Limited, 2006
Gravity sedimentation
EPD
Functionally graded ceramics
579
of the graded powder compact. Sintering rates differ with position and uneven shrinkage may lead to warping and cracking, unless sophisticated sintering techniques are employed. A widely used technique for ceramic/ceramic gradient materials is sequential slip casting where slips of different compositions are cast one after and over another (Requenna et al., 1993). By using a premixing system, the casting composition can be tailored continuously (Chu et al., 1993). In a process called wet spraying (Schindler et al., 1998), suspensions of two powders are created, mixed and sprayed under computer control on a heated substrate. After forming, the green body is removed from the substrate, for bulk FGMs, or bonded with the substrate, for FGM foils. A throughthickness composition gradient can be created by controlling the ratio of the two powders in the mixed suspension. Centrifugal casting (Fukui et al., 1994; Watanabe et al., 1998) is another FGM consolidation method using suspension mixing to realize the gradient. When suspensions of two powders of different density or different grain size are mixed and injected into a cylindrical cavity, which is rotating at high speed, the centrifugal forces cause a compositional or porosity gradient in the growing powder compact in the radial direction. Before stopping the rotation, wax is injected into the system to bind the powders in order to increase the green strength for body handling. The porous FGMs with a gradient distribution in porosity can be used as a preform for filters, or for ceramic membranes. A process similar to centrifugal casting is gravitational sedimentation (Bernhardt, 1999). Centrifugal casting can only be used for cylindrically shaped parts, whereas gravitational sedimentation is suitable for flat FGMs. Among the different colloidal processing techniques, electrophoretic deposition (EPD) is a very promising method (Put et al., 2003a, 2003c, 2002; Vleugels et al., 2003; Anné et al., 2004) because it is a fairly rapid, low-cost process for the fabrication of ceramic coatings, monoliths, composites, laminates and functionally graded materials varying in thickness from a few nanometres to centimetres (Van der Biest and Vandeperre, 1999). Electrophoretic deposition is a two-step process (Fig. 22.3). In the first step, particles having acquired an electric charge in the liquid in which they are suspended are forced to move towards one of the electrodes by applying an electric field to the suspension (electrophoresis). In the second step (deposition), the particles collect at one of the electrodes and form a coherent deposit on it. The deposit takes the shape imposed by this electrode. After drying and removal from the electrode, a shaped green ceramic body is obtained. Firing this green body then results in a ceramic component. Gradient materials can be obtained since the composition of the next powder layer that deposits is determined by the composition of the suspension at that moment (Fig. 22.3). Judiciously adapting the powder concentration in the suspension allows one to generate a well-controlled gradient profile in a continuous shaping step.
© Woodhead Publishing Limited, 2006
580
Ceramic matrix composites EPD set-up for FGM
Gradient profile
Composition 100%
Gradient thickness Scale: nanometres to millimetres
22.3 Electrophoretic deposition process for FGM materials.
The process is not material specific, since a wide variety of materials have already been deposited such as metal powders, ceramics, glasses, and polymers (Van der Biest and Vandeperre, 1999). In general, the only shape limitation is the feasibility to remove the deposit from the electrode after deposition. Continuously graded materials in the Al2O3/ZrO2 (Vleugels et al., 2003), ZrO2/WC (Put et al., 2002), and WC/Co (Put et al., 2001) system have already been explored by means of EPD. A prerequisite for successful production of FGM materials by means of EPD is a full control of the kinetics of the process. Kinetic models have therefore been developed for processing an FGM in a multi-component system by means of EPD (Put et al., 2003b). As an example, the composition profile (Fig. 22.4) and microstructure (Fig. 22.5) of an Al2O3/ZrO2 FG (Vleugels et al., 2003) disk with a homogeneous composite core (75 vol% Al2O3), a pure Al2O3 surface layer on one side and a homogeneous composite (90 vol% Al2O3) on the other side, and intermediate symmetrically profiled graded layers, is presented. As will be explained below, a convex alumina gradient profile is suggested to give the highest compressive stress in the alumina outer layers and the lowest tensile stresses in the core of the disk. The ZrO2 (white) and Al 2O 3 (grey) phases can be clearly differentiated in the microstructure. The ZrO2 phase is well dispersed in the Al2O3 matrix in the gradient parts and in the core of the FGM.
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
581
100
Content of Al2O3 (vol%)
95 90 85 80 75 70 Measured profile Predicted profile
65 60 0
1
2 3 Sintered distance (mm)
4
5
22.4 Measured and predicted FGM profile of an Al2O3/ZrO2 FGM disk.
A
B
C
D
E
1 mm
A
B
C
D
E
2µm
22.5 General overview and some detailed micrographs of specific locations in the FGM disk.
22.4.2 Densification of FGM powder compacts A major challenge is the densification of the graded powder compacts. The processing of FGM materials by powder metallurgy methods often faced undesirable excessive bending or warping of the component after sintering (Miyamoto et al., 1999). Due to excessive thermal residual stresses, cracks and other defects may often be observed in the final FGM component unless properly manufactured.
© Woodhead Publishing Limited, 2006
582
Ceramic matrix composites Homogeneous
Al2O3
Gradient Gradient Homogeneous Al2O3/ZrO2
Crack
Gradient
Homogeneous (a)
(b)
22.6 (a) FGM schematic; (b) typical crack observed in Al2O3 /TZP FGM.
Figure 22.6 shows typical cracks observed in symmetrically graded Al2O3/ ZrO2 disks, shaped by electrophoretic deposition and densified by pressureless sintering. Transverse cracks were observed in the ZrO2-rich core of the sintered symmetric TZP/Al2O3 disks. The crack propagation, however, stopped in the outer pure Al2O3 layer, indicating that the in-plane tensile stress is located in the centre of the disks, which should be lowered. Hillman et al. (1996) observed similar defects in symmetrical laminates with Al2O3/ZrO2 layers at the surface and a ZrO2 central layer. Cai et al. (1997a, 1997b) discussed the embryonic stage of a sinter crack as observed in Fig. 22.6(b), and found regions of low density and cavitational defects in Al2O3/ZrO2/ Al2O3 laminates. These defects are most susceptible to residual tensile stresses during cooling in the core, due to the higher coefficient of thermal expansion (CTE) of zirconia. These regions of lower density (pores) must have formed as a result of the tensile stress that develops during the differential shrinkage during densification between the Al2O3 and the Al2O3/ZrO2 layers. The pores then act as pre-existing flaws for the generation of thermal expansion mismatch cracks during cooling via linkage of the pores and cavitational defects. Elimination of the transverse cracks can be accomplished by decreasing the overall shrinkage of the composites. This is done either by decreasing the compositional difference between the different layers (Cai et al., 1997a, 1997b) or by adjusting the green density of the different layers (Novak and Beranic 2005). Another possibility is to decrease the cooling and heating rate during sintering (Cai et al., 1997b). The mismatch stresses during the heating cycle are decreased by the viscous nature of the FGM material at the sintering temperature. The sintering mismatch stress is proportional to the mismatch sintering rate. Reduced cracking under a slow cooling rate is probably due to the relaxation of residual stresses during the initial period of cooling. Almost all ceramic/ceramic bulk FGMs are sintered by conventional pressureless sintering (Wu et al., 1996; Marple and Boulanger, 1994; Cichocki and Trumble, 1998) or hot pressing (Kawai and Wakamatsu, 1995; Vanmeensel
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
583
et al., 2004), depending on the sintering properties of the two components. In metal/ceramic FGMs with a continuous metal phase and a discontinuous ceramic phase, the sintering rates are controlled by the densification of the metal phase and such FGMs can be densified by conventional sintering methods (Neubrand and Rödel, 1997). In most FGMs where a high ceramic phase content is envisaged, however, some special approaches have to be considered for full densification. In addition to conventional sintering, reactive powder processing, also called combustion synthesis or self-propagating high-temperature synthesis (SHS), can be used if the target compounds can be synthezised from the starting powder mixture (Stangle and Miyamoto, 1995). This process comprises a rapid and exothermic chemical reaction to simultaneously synthesize some or all of the constituent phases in the FGM and densify the component. A more advanced technique, such as Spark Plasma Sintering (SPS) or Pulsed Electric Current Sintering (PECS) (Tokita, 1999), is also used for FGM fabrication. It is a pressure-assisted sintering method in which a high current is pulsed through a die/punch/sample set-up, which can be compared with that of conventional hot pressing. The large current pulses generate spark plasmas, a spark impact pressure and Joule heating. The sintering mechanism and mechanical properties of the sintered compacts show characteristics different from conventional pressure-assisted sintering processes and parts. This technique offers significant advantages for various kinds of new materials and consistently produces a dense compact in a shorter sintering time and with finer grain size than conventional methods. Spark plasma sintering of FGMs uses a temperature gradient in the system, which allows a homogeneous densification of FGMs by matching the temperature gradient to the shrinkage rate gradient of the compact. With a spark plasma system, large ceramic/metal bulk FGMs (~100 mm across) can be homogeneously densified in a short time with heating and holding times totalling less than one hour. Amongst the reported spark plasma sintered systems are WCbased materials (WC/Co, WC/Co/steel, WC/Mo), ZrO2-based composites (ZrO2/steel, ZrO2/TiAl, ZrO2/Ni), Al2O3/TiAl, etc. (Tokita, 1999). A systematic introduction to spark plasma sintering can be found in a review by Tokita (1999) or a recent paper by Hennicke and Kessel (2004), where the process, mechanical properties, size and shape effects, and production machine systems are reported. Microwave sintering is another promising technique for ceramic/metal FGMs to eliminate the difficulty of inequality of the shrinkage rate. As a newly developed sintering technique, microwave sintering uses microwave +irradiation to heat the ceramic or ceramic-based composite compact (Gerdes and Willert-Poradu, 1994; Willert-Porada, 1999; Zhao et al., 2000). The mechanism of microwave heating is based on the dielectric loss of the ceramic phases involved, resulting in a volumetric heating technique in which the heat is generated by the compact itself.
© Woodhead Publishing Limited, 2006
584
Ceramic matrix composites 200
120 100
σy (MPa)
σy (MPa)
150
100
80 60 40
50 20 0 –0.300
–0.200 Y (mm)
–0.100
0 –0.300
–0.200 Y (mm)
–0.100
22.7 Stress distribution in (a) a coating–substrate system and (b) a FGM (Munz et al., 1998).
22.5
FGM design for structural applications
FGMs for ceramics offer great promise in applications where the operating conditions are severe, for example wear-resistant linings for handling large heavy abrasive ore particles, high-speed cutting tools, rocket heat shields, heat exchanger tubes, thermoelectric generators, heat-engine components, plasma facings for fusion reactors, and electrically insulating metal/ceramic joints. They are ideal for minimizing the thermo-mechanical mismatch in metal–ceramic bonding. As presented in Fig. 22.7 (Munz et al. 1998), a stress discontinuity is generated in case of a bi-material combination, which may result in a poor adhesion or even delamination, whereas a significantly lower stress concentration can be established for a graded interface, thereby improving the materials adhesion and reducing the probability of delamination. The residual stresses have a large influence on the properties of a structural body, e.g., a correct design of the composition gradient can generate compressive stresses at those locations which are loaded in tension during application. Compressive stresses at the surface can also have a beneficial effect on the tribological properties of the component (Novak et al., 2005).
22.5.1 Thermal stress calculations in FGM material The calculation of thermal stresses in functionally graded materials is already a relatively old topic (Yang et al., 2003). Two methods can be distinguished: analytical methods and finite element methods. However, the complicating effect of the elastic modulus variation with the position severely limits the scope of problems that can be solved analytically. Therefore, the majority of the analytical work has been for FGM films or other simple structures (Becker et al., 2000). Analytical models have been developed for the calculation of thermal stresses for 1-D FGM symmetrical plates (Jung et al., 2003), non-
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
585
symmetrical plates (Ravichandran, 1995), cylinders (Chen and Awaji, 2003) and spheres (Obata and Noda 1994). Finot and Suresh (n.d.) have written a software program, called Multitherm, to calculate thermal stresses in a plane stress, plane strain or a biaxial state. For a more general 2-D or 3-D problem, numerical methods like finite element analysis are required. These are, costly however, since a full analysis for each material pair, geometry and gradient must be performed. In this section, an analytical solution to calculate residual stresses in an FGM disk is discussed, based on simple linear elastic plate theories of classical mechanics, and used for the calculation of residual stresses in a plane stress state. An equi-biaxial stress analysis differs from a plane stress state by simply replacing the Young’s modulus E by the corresponding biaxial modulus E′ = E/(1 – ν). In this way, the residual thermal stress can be calculated in the centre of the FGM disk, far enough away from the free edges where a complex stress state is present. Consider therefore an initially perfectly planar unconstrained, layered or graded plate as presented in Fig. 22.8, which is subjected to a uniform thermal excursion. During cooling from the sintering temperature, the layers with the highest CTE will contract more than the layers with a lower CTE. This causes a variation in thermal stress along the thickness of the plate. The plate can bend or warp due to the through-thickness strain gradient, thereby accommodating the thermal stress. When the geometrical conditions of the plate (isotropic in-plane) are such that the strain is allowed to be a function of z only, the in-plane normal strain can be expressed as:
ε (z) = εxx = ε0 + βz
(22.1)
where ε0 is the normal strain at z = 0, and β is the curvature of the plate in z
h1
E 1, ν1, α1 E 2, ν2, α2 E 3, ν3, α3 E 3, ν3, α3 E 2, ν2, α2
h2
E 1, ν1, α1 y x
22.8 Schematic of a layered disk.
© Woodhead Publishing Limited, 2006
586
Ceramic matrix composites
its plane. The stress σ(z) = σxx = σyy for the biaxial stress state is given by (Giannakopolous et al., 1995; Suresh et al., 1994): E( z ) [ ε ( z ) – α ( z ) ∆ T ( z )] 1 – ν(z)
σ (z) =
E(z) [ ε + β z – α ( z ) ∆T ( z )] 1 – ν (z) 0
=
(22.2)
where α is the coefficient of thermal expansion (CTE), ∆T is the difference between the sintering temperature and room temperature, ν is the Poisson coefficient, and E is the Young’s modulus. The resultant force and the resultant moment of the stress distribution σ(z) along the height z must be equal to the applied axial force, Fap, and the applied bending moment, Map, respectively:
∫
– h1
∫
– h1
σ ( z )d z = F ap
(22.3)
σ ( z ) z d z = M ap
(22.4)
– h2
– h2
These two conditions, i.e. force balance and moment balance, lead to a linear system of equations for ε0 and β. The solutions, in the absence of buckling, are given by (Giannakopolous et al., 1995; Suresh et al., 1994): – I 2 ( M 0 + F ap ) + I1 ( M1 + M ap ) I12 – I 0 I 2
ε0 =
β=
I1 ( M 0 + F ap ) – I 0 ( M1 + M ap ) I12 – I 0 I 2
(22.5)
(22.6)
with: Ii =
∫
h2
z i E ( z )d z
for i = 0, 1, 2
(22.7)
– h1
Mi =
∫
h2
z i α ( z ) ∆ T ( z ) E ( z )d z
for i = 0, 1
(22.8)
– h1
From these equations, the thermal stress profile through the thickness of the multi-layered material with a compositional profile can be calculated. If E and α∆T vary continuously with z, σ is continuous too. Because, the effective properties change as functions of the position in an FGM material, it is necessary to estimate the local composite properties as functions of
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
587
Table 22.1 Material properties of Al2O3 and ZrO2
Al2O3 ZrO2
Young’s modulus, E (GPa)
Poisson’s ratio, ν
CTE, α (10–6 K–1)
360 180
0.237 0.300
7.40 1.01
95 90
n = 1.75 n=1 n = 0.5
85 80 75
0
1
2 3 4 Distance x (mm) (a)
5
Residual stress (MPa)
Al2O3 content (vol%)
100
150 100 50 0 0
2
4
–50
6
n = 1.75 n=1 n = 0.5
–100 –150 Distance (mm) (b)
22.9 Composition and corresponding residual stresses for a FGM disk with a convex (n = 0.5), linear (n =1) and concave (n = 1.75) composition profile.
composition and temperature. Rules of mixtures can be used to estimate these properties (Suresh and Mortensen, 1998). This analytical model is validated by calculating the stress profile in an Al2O3/ZrO2 FGM disk with homogeneous pure Al2O3 surface layers, a homogeneous Al2O3/ZrO2 composite core and intermediate graded layers with a convex, linear and concave profile. The material properties used in these calculations are given in Table 22.1. The composition profiles and corresponding residual stresses for the FGM disks sintered at 1550°C are given in Fig. 22.9. These calculations reveal that the FGM disks with a convex (n = 0.5) alumina gradient profile result in the highest compressive stress in the alumina outer layers and the lowest tensile stresses in their core.
22.5.2 Al2O3 and ZrO2-based ceramic/ceramic FGMs ZrO2-based FGMs have been of high interest (Marple and Boulanger, 1994, Anné et al., 2004, Zhao et al., 2000, Put et al., 2002) because of the high toughness and excellent strength of tetragonal ZrO2 polycrystalline materials (TZP). ZrO2-based FGMs are developed mainly for energy conversion systems, biomedical applications and cutting tools, where high hardness has to be combined with high toughness or where thermal stresses have to be released (Sanchez-Herencia et al., 2000). Al2O3/ZrO2 FGMs were studied intensively for biomedical applications,
© Woodhead Publishing Limited, 2006
588
Ceramic matrix composites
22.10 Overview of Al2O3 /TZP FGM ball-heads for hip prostheses, made by electrophoretic deposition and a schematic cross-section of an FGM ball-head (Anné et al., 2004).
specifically for implants like hip prostheses (Anné et al., 2004). The FGM idea was applied to increase the performance of total hip replacement prostheses through design and development of alumina/zirconia functionally graded materials using electrophoretic deposition as a near-net shaping technique (Fig. 22.10). Today, most of the joint prostheses consist of metallic components articulating against polymer counterparts, of which more than 10% need revision after only 10 years. This is expensive and reduces the quality of life of the patients. Also the clinical effects of metallic ion-releasing implants and polymer wear debris are a matter of concern. This demonstrates the need for more wear-resistant and biocompatible materials such as ceramics. Therefore alumina–zirconia graded femoral ball-heads were developed. The potential of this system follows from the properties of alumina (low wear, high hardness) and zirconia (high strength, high toughness). By combining alumina and zirconia, a functional gradient in hardness (high at the alumina surface) and toughness (high in the zirconia-rich core) can be established. Due to the different thermal expansion coefficients of Al2O3 and ZrO2, residual stresses are developed during cooling from the sintering temperature, which strongly influence the mechanical properties like strength and toughness. Compressive surface stresses in the outer alumina layer will also have a beneficial effect on wear resistance (Fig. 22.11) and strength (Fig. 22.12). Additionally, increased toughness can be observed in this kind of graded ceramic composite (Tilbrook et al., 2005).
22.5.3 WC/Co FGM materials Another interesting group of structural gradient materials is that of cemented carbides, commonly used, for example, for cutting tools and mining equipment where high wear resistance and toughness are required. Using graded WC/
© Woodhead Publishing Limited, 2006
Functionally graded ceramics 2
Al2O3 disks
589
FGM disks
In In (1/S)
1 0 –1
m=9
m = 12
288 MPa
513 MPa
–2 –3 –4 5.0
5.5
6.0
6.5
In (strength)
22.11 Weibull plot of the biaxial strength (ISO 6474) of homogeneous Al2O3 and Al2O3 /ZrO2 FGM disks with 80 vol% Al2O3 in the core, pure Al2O3 on the outside and intermediated convex graded profile. The FGM disks were made by electrophoretic deposition and pressureless sintering.
Wear volume
0.02 0.018 0.016 0.014 0.012 0.01 0.008 0.006 0.004 0.002 0
Ball wear Flat wear
10
175 241 Residual compressive stress (MPa)
260
22.12 Wear volume of step-graded Al2O3 /ZrO2 FGM disks with 80 vol% Al2O3 in the core and pure Al2O3 on the outside (Novak, et al. 2005). The step-graded disks were made by sequential slip casting (Novak, et al. 2005).
Co hardmetals, it is possible to improve both properties at the same time (Cherradi et al., 1994) (Fig. 22.13). This can be achieved by a gradation in the binder content from the centre to the surface of the tool. The improved mechanical properties of the FGM-cemented carbide are due to compressive stresses near the surface, which enable a reduction of the Co-phase content from 6 to 3 wt% without any loss in apparent fracture toughness. Two existing industrial technologies have been reported to produce graded WC-Co hardmetals, i.e., liquid phase sintering of homogeneous WC-Co in a controlled atmosphere or pressure assisted sintering of graded WC-Co compacts. The first method is mainly use for nitride-containing WC-Co or
© Woodhead Publishing Limited, 2006
590
Ceramic matrix composites
FGM
25
KlC
20 15 10 Conventional hard metals 5 700 900 1100 1300 HV10
1500
1700
1900
22.13 Hardness–toughness relationship for WC-Co hardmetals (Cherradi et al., 1994).
WC-Ni cermets (Chen et al., 2000a; Narasimhan et al., 1995). The graded Co distribution is formed during liquid-phase sintering through the establishment of a nitrogen partial pressure gradient (Zackrisson et al., 2000; Chen et al., 2000b) or by a high-temperature carburization treatment of a WC–Co composite with sub-stoichiometric carbon content (Fischer et al., 1988). Pressure-assisted sintering techniques such as hot pressing, hot isostatic pressing and SPS can also be used to consolidate graded WC–Co compacts by solid state sintering (Tokita, 2003). Microwave sintering methods of WC/ Co cemented carbides have also been reported (Willert-Porada et al., 1995). Other graded cermet systems reported include Cr3C2/Ni (Seefeld et al., 1999), TiC/Ti (Cline, 1995), and TiC-Ni (Sabatello et al., 2000). The design concept of the graded cermets is similar to that of WC/Co cemented carbide FGMs.
22.6
Future trends
Already a lot of scientific work has been established in processing graded bulk ceramics and in modelling and optimization of the properties of these materials. Despite these technological successes, the number of commercial successes is still limited because the production costs in most cases are still too high compared to homogeneous materials. However, new production processes are emerging on the market that allow economic production of FGM bulk ceramics. Many new target applications have already appeared or are rapidly emerging,’ including the following: • Thermo-mechanical applications (Kawai and Wakamatsu, 1995; Araki et al., 1994; Wakamatsu et al., 1999; Hofinger et al., 1999; Lee et al., 1996; Jiang et al., 1998)
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
591
• Energy conversion (Niino, 1998; Mahan et al., 1997; Schilz et al., 1999; Sugiyama et al., 1998) • Applications in optics and electronics (Wang et al., 1998; Palais, 1980; Komatsu et al., 1998; Le Goues et al., 1992, Miyamoto et al., 2005) • Electrical and magnetic applications (Yamane et al., 1998; Nishida et al., 1999; Wu et al., 1996) • Graded cemented carbide coatings on steel substrates (Put et al., 2003(c)) • Biomedical applications (Kurzweg et al., 1998, Ban et al., 1999; Rogier and Pernot, 1991, Anné et al., 2004).
22.7
Further reading
Since 1990, an international congress on Functionally Graded Materials is organized by the International Advisory Committee of FGM (IACFGM) every two years. The proceedings of these congresses give a lot of information about research on FGM materials. The symposia on FGMs were held in Sendai (1990), San Francisco (1992), Lausanne (1994), Tsukuba (1996), Dresden (1998), Estes Park, Colorado (2000), Beijing (2002) and Leuven (2004). The next one will be held in Chicago in 2006. Due to their comprehensive description of the design, modelling, processing, and evaluation of FGMs as well as the many applications described, the following books are recommended for further reading: • Fundamentals of Functionally Graded Materials, by S. Suresh and A. Mortensen, A, Institute of Materials, Published by Woodhead, Cambridge, UK, 1998. • Functionally Graded Materials: Design, Processing and Applications, edited by Y. Miyamoto, W.A. Kaysser, B.H. Rabin, A. Kawasaki and R.G. Ford, published by Kluwer Academic Publishers, Boston/Dordrecht/London, 1999. Several review articles have been written on FGMs, covering applications (Cherradi et al., 1994), processing (Mortensen and Suresh, 1995; Neubrand and Rödel, 1997), modelling (Markworth et al., 1995; Li et al., 2000) and fracture mechanics (Erdogan, 1995; Tilbrook et al., 2005). In Japan, the FGM database http://fgmdb.nal.go.jp, http://fgmdb.nal.go.jp/ e_whatsfgm.html is established. This database gives a good overview of the FGM concept, some examples and an overview of FGM literature. Additionally an overview of almost all research groups involved in the FGM concept can be found.
22.8
References
Anné, G., Vanmeensel, K., Vleugels, J., (2004), ‘Electrophoretic deposition as a near net
© Woodhead Publishing Limited, 2006
592
Ceramic matrix composites
shaping technique for functionally graded biomaterials’, Mat. Sci. Forum, 492–493, 213–218. Araki, M., Sasaki, M., Kim, S., Suzuki, S., Nakamura, K., Akiba, M., (1994), ‘Thermal response experiments of SiC/ and TiC/C functionally gradient materials as plasma facing materials for fusion applications’, J. Nucl. Mat., 212–215, 1329–1334. Ban, S., Hasegawa, J., Maruno, S., (1999), ‘Fabrication and properties of functionally graded bioactive composites comprising hydroxyapatite containing glass coated titanium’, Mat. Sci. Forum., 308–311, 350–355. Bao, G., Cai, H., (1997) ‘Delamination cracking in functionally graded coating/metal substrate system’, Acta Mater, 45(3) 1055–1066. Becker, T.L., Cannon, R.M., Ritchie, R.O., (2000), ‘An approximate method for residual stress calculation in functionally graded materials’, Mech. Mat., 85–97. Bernhardt, R., Meyer-Olbersleben, F., Kieback, B., (1999), ‘The influence of hydrodynamic effects on the adjustment of gradient patterns through gravity sedimentation of polydisperse particle systems in newtonian and viscoelastic fluid’, Mat. Sci. Forum, 308–311: 31–35. Cai, P.Z., Green, D.J., Messing, G.L., (1997a), ‘Constrained densification of alumina/ zirconia hybrid laminates. 1. Experimental observations of processing defects’, J. Am. Ceram Soc., 80(8), 1929–1939. Cai, P.Z., Green, D.J., Messing, G.L., (1997b) Constrained densification of alumina/ zirconia hybrid laminates. 2. Viscoelastic stress computation’, J. Am. Ceram. Soc., 80(8), 1940–1948. Chen, C., Awaji, H., (2003), ‘Transient and residual stresses in a hollow cylinder of functionally graded materials’, Mat. Sci. Forum., 423–425, 665–670. Chen, L.M., Lengauer, W., Ettmayer, P., Dreyer, K., Daub, H.W., Kassel, D., (2000a), ‘Fundamentals of liquid phase sintering for modern cermets and functionally graded cemented carbonitrides (FGCC)’, Int. J. Refract. Met. Hard Mat., 18(6), 307–322. Chen, L.M., Lengauer, W., Dreyer, K., (2000b), ‘Advances in modern nitrogen-containing hard-metals and cermets’, Int. J. Refract. Met. Hard Mat., 18(2–3), 153–161. Cherradi, N., Kawasaki, A., Gasik, M., (1994), ‘Worldwide trends in functional gradient materials research and development’, Compos. Eng., 4(8), 883–894. Chu, J., Ishibashi, H., Hayashi, K., Takebe, H., Morinaga, K., (1993), ‘Slip casting of continuous functionally gradient material’, J. Ceram. Soc. Japan, 101, 818–820. Cichocki, F.R., Jr, Trumble, K.P., (1998), ‘Tailored porosity gradients via colloidal infiltration of compression-molded sponges’, J. Am. Ceram. Soc., 81(6), 1661–1664. Cline, C.F., (1995), ‘Preparation of gradient TiC cermet cutting tools’, in Ilschner, B, Cherradi, N., (eds), Proc. 3rd Int. Symp. on Structural and Functional Gradient Materials, Presses Polytechniques et Universitaires Romandes, Lausanne, 595. Erdogan, F., (1995), ‘Fracture mechanics of functionally graded materials’, Mater. Res. Soc. Bull, 20(1), 43–44. Finot, M., Sureshs (n.d.), http://ninas.mit.edu/lexcom/www/multitherm.html. Fischer, U.K.R., Hartzell, E.T., Akerman J.G.H., (1988), ‘Cemented carbide body used preferably for rock drilling and mineral cutting’, US Patent No. 4,743,515, 10 May 1988. Fukui, Y., Takashima, K., Ponton, C.B., (1994), ‘Measurement of Young’s modulus and internal friction of an in situ Al–Al3Ni functionally gradient material’, J. Mat. Sci., 29, 2281–2286. Gasik, M., (1995), ‘Principles of functional graded materials and their processing by powder metallurgy’, Acta Polytechnica Scandinavica, Chemical Technology Series, No. 226, Helsinki.
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
593
Gerdes, T., Willert-Porada, M., (1994), ‘Microwave sintering of metal-ceramic and ceramicceramic composites’, Mat. Res. Sco. Symp. Proc. 347, 531. Giannakopolous, A.E., Suresh, S., Finot, M., Olsson, M., (1995), ‘Elastoplastic analysis of thermal cycling – layered materials with compositional gradients’, Acta. Metall. Mater., 43, 1335–1354. Hennicke J., Kessel, H.U. (2004), ‘Field assisted sintering technology (FAST) for the consolidation of innovative materials’, Ceramic Forum Int./Ber.DKG, 81(11) (2004) E14–E16. Hillman, C., Suo, Z.G., Lange, F.F., (1996), ‘Cracking of laminates subjected to biaxial tensile stresses’, J. Am. Ceram. Soc., 79(8), 2127–2133. Hofinger, I., Bahr, H.A., Balke, H., Kirchhoff, G., (1999), ‘Fracture mechanical modelling and damage characterization of functionally graded thermal barrier coatings by means of laser irradiation’, Mat. Sci. Forum., 308–311, 450–455. Jiang, W., Watanabe, R., Kawasaki, A, (1998), ‘Compositional dependence of thermal conductivity in sintered Mo/ZrO2 composites’, J. Japan Inst. Met, 62(11), 1018–1024. Jung, Z., Xing, A., Chuanzhen, H., (2003), ‘An analysis of unsteady thermal stresses in a functionally gradient ceramic plate with symmetrical structure’, Ceram. Int., 29, 279–285. Kawai, C., Wakamatsu S., (1995) ‘Synthesis of a functionally gradient material based on C/C composites using an electro-deposition method’, J. Mat. Sci. Lett., 14, 467. Komatsu, T., Benino, Y., Sakai, R., (1998), ‘Fabrication of transparent tellurite-based glass-ceramics with graded optical nonlinearity’, J. Japan Inst. Met., 62(11), 1055– 1102. Kurzweg, H., Heimann, R.B., Troczynski, T., (1998), ‘Adhesion of thermally sprayed hydroxyapatite-bond-coat systems measured by a novel peel test’, J. Mat. Sci. Med., 9(1), 9–16. Lee, W.Y., Stinton, D.P., Berndt, C.C., Erdogan, F., Lee, Y.D., Mutasim, Z., (1996) ‘Concept of functionally graded materials for advanced thermal barrier coating applications’, J. Am. Ceram. Soc., 79(12), 3003–3012. Le Goues, F.K., Meyerson, B.S., Morar, J.F., Kirchner, O.D., (1992), ‘Mechanism and conditions for anomalous strain relaxation in graded thin films and superlattices’, J. Appl. Phys., 71(9), 4230–4243. Li, H., Lambros, L., Cheeseman, B.A., Santare, M.H., (2000), ‘Experimental investigation of the quasi-static fracture of functionally graded materials’, Int. J. Solid. Struct., 37, 3715–3732. Mahan, G., Sales, B., Sharp, J. (1997), ‘Thermoelectric materials: new approaches to an old problem’, Physics Today, 3, 42–47. Markworth, A.J., Ramesh, K.S., Parks Jr. W.P., (1995), ‘Modelling studies applied to functionally graded materials’, J. Mat. Sci., 30, 2183–2193.. Marple, B.R., Boulanger, J., (1994), ‘Graded casting and materials with continuous gradients’, J. Am. Ceram. Soc., 77(10), 2747–2750. Miyamoto, Y., Kaysser, W.A., Rabin, B.H., Kawasaki, A., Ford, R.G., (eds) (1999), Functionally Graded Materials: Design, Processing and Applications, (Kluwer Academic Boston/Dordrecht/London. Miyamoto, Y., Kirihara, S., Takeda, M.W., Honda, K., Sakoda K., (2005), ‘A new functional material: photonic fractal’, Mat. Sci. Forum, 492–493, 77–84. Mortensen, A., Suresh, S., (1995), ‘Functionally graded metals and metal-ceramic composites: Part 1 Processing’, Int. Mater. Rev., 40(6), 239–265. Munz, D., Schaller, W., Yang, Y., (1998), ‘Where is the benefit of a graded material?’,
© Woodhead Publishing Limited, 2006
594
Ceramic matrix composites
presented at FGM’98 Workshop Computer Aided Design of Functionally Graded Materials, 26 October 1998, Dresden. Narasimhan, K., Boppana, S.P., Bhat, D.G., (1995), ‘Development of a graded TiCN coating for cemented carbide cutting tools – a design approach’, Wear, 188(1–2), 123– 129. Neubrand, A., Rödel, J., (1997), ‘Gradient materials: an overview of a novel concept’, Z. Metallk, 88(5), 358–371. Niino, M., (1998), ‘Working towards the goal of creating new energy conversion materials’, FGM News (FGM Forum), 37, 25–28. Nishida, T., Prezzotti, G., Shiono, T., (1999), ‘Preparation and characterization of substrates with functionally graded dielectric constant’, Mat. Sci. Forum, 308–311, 539–545. Novak, S., Beranič, S., (2005), ‘Densification of step graded Al2O3–ZrO2 composites’, Mat. Sci. Forum, 492–493, 207–212. Novak, S., Kalin, M., Beranič S., Lukas, P., Anné, G., Van Der Biest, O., (2005), ‘Properties and wear behaviour of step-graded alumina-ZTA composites’, IX Conf. & Exh. Eur. Cer. Soc., Portorož, Slovenia. Obata, Y., Noda, N., (1994), ‘Steady thermal stress in a hollow circular cylinder and a hollow sphere of a functionally gradient material’, J. Therm. Stress., 17, 471–487. Palais, J.C., (1980), ‘Fiber coupling using graded-index rod lenses’, Appl. Optics., 19(12), 2011–2018. Put, S., Vleugels, J., Van der Biest, O., (2001), ‘Functionally graded WC–Co materials produced by electrophoretic deposition’, Scripta. Mater, 45(10), 1139–1145. Put, S., Anné G., Vleugels, J., Van der Biest, O., (2002), ‘Functionally graded ZrO2-WC composites proceessed by electrophoretic deposition’, Key. Eng. Mat., 206(2), 189– 192. Put, S., Vleugels, J., Anné, G., Van der Biest, O., (2003a), ‘Functionally graded ceramic and ceramic–metal composites shaped by electrophoretic deposition’, Colloids and Surfaces A: Physicochem Eng. Aspects, 222, 223–232. Put, S., Vleugels, J., Van der Biest, O., (2003b), ‘Gradient profile prediction in functionally graded materials processed by electrophoretic deposition’, Acta. Mater., 51(20), 6303– 6317. Put, S., Vleugels, J., Anné, G., Van Der Biest, O., (2003c), ‘Processing of hardmetal coatings on steel substrates’, Scripta Mater., 48(9), 1361–1366. Ravichandran, K.S., (1995), ‘Thermal residual stress in a functionally graded material system’, Mat. Sci. Eng., A201, 269–276. Requenna, J., Moya, J.S., Pena, P., (1993), ‘Al2TiO3–Al2O3 functionally gradient materials obtained by sequential slip casting’, in Holt, J.B., Koisumi, M., Hirai, T., Munir, Z.A., Functionally Gradient Materials, American Ceramic Society, Westerville, O.H., 203– 210. Rogier, R., Pernot, F., (1991), ‘Glass-ceramic metal composites for making graded seals in prosthetic devices’, J. Mat. Sci. Med., 2, 153–161. Sabatello, S., Frage, N., Dariel, M.P., (2000), ‘Graded TiC-based cermets’, Materials Science and Engineering A288, 12–18. Sanchez-Herencia, A.J., Moreno, R., Jurado, J., (2000), ‘Electrical transport properties in zirconia/alumina functionally graded materials’, J. Eur. Ceram. Soc., 20, 1611– 1620. Schilz, J., Muller, E., Helmer, L., Kang, Y.S., Noda, Y., Niino, M., (1999) ‘On optimizing the composition function of graded thermoelectric materials’, Mat. Sci. Forum., 308– 311, 647–652.
© Woodhead Publishing Limited, 2006
Functionally graded ceramics
595
Schindler, J., Meyer-Olbersleben, F., Kirbach, B., (1998), ‘Fabrication of FGM-foils for joining application by wet powder spraying’, presented at 5th International Symposium on Functionally Graded Materials, Dresden, October 1998. Seefeld, T., Theiler, C., Schubert, E., Sepold, G., (1999), ‘Laser generation of graded metal-carbide components’, Mat. Sci. Forum, 308–311, 459–464. Stangle, G.C., Miyamoto, Y., (1995), ‘FGM fabrication by combustion synthesis’, MRS Bull., XX(1), 52–53. Sugiyama, A., Kobayashi, K., Ozaki, K., Nishio, T., Matsumoto, A., (1998), ‘Preparation of functionally graded Mg2Si–FeSi2 thermoelectric material by mechanical alloying– pulsed current sintering process’, J. Japan Inst. Met., 26(11), 1082–1087. Suresh, S., Mortensen, A., (1997) ‘Functionally graded metals and metal-ceramic composites: Part 2 Thermomechanical behaviour’, Int. Mater. Rev., 42(3), 85–116. Suresh, S., Mortensen, A., (1998), Fundamentals of Functionaly Graded Materials, Woodhead, Cambridge, UK. Suresh, S., Giannakopoulos, A.E., Olsson, M., (1994), ‘Elastoplastic analysis of thermal cycling – layered materials with sharp interfaces’, J. Mech. Phys. Solids, 42, 979– 1018. Tilbrook, M., Moon, R., Hoffman, M., (2005), ‘Crack propagation in graded composites’, Comp. Sci. Tech, 65, 201–220. Tokita, M., (1999), ‘Development of large-size ceramic/metal bulk FGM fabricated by spark plasma sintering’, Mat. Sci. Forum, 308–311, 83–88. Tokita, M., (2003), ‘Large-size WC/Co functionally graded materials fabricated by spark plasma sintering (SPS) method’, Mat. Sci. Forum, 39, 423–425. Van der Biest, O., Vandeperre, L., (1999), ‘Electrophoretic deposition of materials’, Ann. Rev. Mat. Sci., 29, 327–352. Vanmeensel, K., Anné, G., Jiang, D., Vleugels, J., Van der Biest, O., (2004), ‘Homogeneous and functionally graded Si3N4–TiCN composites shaped by electrophoretic deposition’, Silicates Industriels, Special Issue, 69(7–8), 233–239. Vleugels, J., Anné, G., Put, S., Van der Biest, O., (2003), ‘Thick plate-shaped Al2O3/ZrO2 composites with a continuous gradient processed by electrophoretic deposition’, Mat. Sci. Forum., 423–425, 171–176. Wakamatsu, Y., Shoji, T., Ogawa, K., Hino, I., (1999), ‘Health diagnosis of functionally graded C/SiC coating on C/C composites’, Mat. Sci. Forum, 308–311, 416–421. Wang, X., Masumoto, H., Someno, Y., Hirai, T., (1998), ‘Optical characterization of SiO2–TiO2 thin films with graded refractive index profiles’, J. Japan Inst. Met., 62(11), 1069–1074. Watanabe, Y., Yamanaka, N., Fukui, Y., (1998), ‘Control of composition gradient in a metal-ceramic functionally graded material manufactured by the centrifugal method’, Composites Part A, 29A, 595–601. Willert-Porada, M., Gerdes, T., Borchert, R., (1995), ‘Application of microwave processing to preparation of ceramic and metal-ceramic FGM, Ilschner and N. Cherradi (ed.), Proceedings of the 3rd International Symposium on Structural and Functional Gradient Materials, Presses Polytechniqus et Universitaires Romandes, Laussanne (1995), 15– 21. Wu, C.C.M., Kahn, M., Moy, W., (1996), ‘Piezoelectric ceramics with functional gradients: a new application in material design’, J. Am. Ceram. Soc., 79(3), 809–813. Yamane, H., Morita, K., Shimada., M., (1998), ‘Structure and dielectric properties of composition graded K(Ta,Nb)O3 crystals’, J. Japan Inst. Met., 62(11), 1051–1054.
© Woodhead Publishing Limited, 2006
596
Ceramic matrix composites
Yang, Z.M., Zhou, Z.G., Zhang, L.M., (2003), ‘Characteristics of residual stress in Mo– Ti functionally graded material with a continuous change of composition’, Mat. Sci. Eng., A358, 214–218. Zackrisson, J., Rolander, U., Jansson, B., Andren, H.O., (2000), ‘Microstructure and performance of a cermet material heat-treated in nitrogen’, Acta Mater, 48(17), 4281– 4291. Zhao, C., Vleugels, J., Vandeperre, L., Van Der Biest, O., (2000), ‘Cylindrical Al2O3/TZP functionally graded materials by EPD’, Brit. Cer. Trans., 99(6), 284–287.
© Woodhead Publishing Limited, 2006