Dielectric Materials and Devices
Related titles published by The American Ceramic Society: going-going the Bionic Cat and the Jewel Thief By Lorry L. Hench 02001,ISBN 1-57498-129-3 f i e Magic of Ceramics By David W. Richerson 02000, ISBN 1-57498-050-5 going-going the Bionic Cat By Lorry 1. Hench 02000, ISBN 1-57498-109-9 Glazes and Glass Coatings By Richard A. Eppler and Douglas R. Eppler 02000, ISBN 1-57498-054-8 Electronic Ceramic Materials and Devices (Ceramic Transactions, Volume J 06) Edited by K.M. Nair and AS. Bhalla 02009, ISBN 1-57498-098-X Ceramic Innovations in the 20th Century Edited by John B. Wachtman Jr. 0 1999, ISBN 1-57498-093-9 Finishing of Advanced Ceramics and Glasses (Ceramic Transactions, Volume I 021 Edited by Robert Sabia, Victor Greenhut, and Carlo Pantano 0 1999, ISBN 1-57498-082-3 Dielectric Ceramic Materials (Ceramic Transactions, Volume I 00) Edited by K.M. Nair and AS. Bhalla 0 1997, ISBN 0-57498-066-1 Advances in Dielectric Ceramic Materials (Ceramic Transactions, Volume 88) Edited by K.M. Nair and AS. Bhalla 0 1996, ISBN 1-57498-033-5 Hybrid MicroelectronicMaterials (Ceramic Transactions, Volume 68) Edited by K.M. Nair and Y.N. Shukla 0 1995, ISBN 1-57498-013-0 Ferroic Materials: Design, Preparation, and Characteristics(Ceramic Transactions, Volume 43) Edited by AS. Bhalla, K.M. Nair, I.K. Lloyd, H. Yanagida, and D.A. Payne 0 1994, ISBN 0-944904-77-7 Grain Boundaries and Interfacial Phenomena in Electronic Ceramics (Ceramic Transactions, Volume 4 I ) Edited by Lionel M. Levinson and Shin-ichi Hirano 0 1994, ISBN 0-944904-73-4
For information on ordering titles published by The American Ceramic Society, or to request a ublications catalog, please contact our Customer Service De artment at 614-794-5890 (phone), 614-794-5892 (f&,
[email protected]> (e-mail), or write to Pustomer Service Department, 735 Ceramic Place, Westerville, OH 43081, USA. Visit our on-line book catalog at <www.ceramics.org>.
Dielectric Materials and Devices
Edited by
K.M. Nair Amar 5. Bhalla lapan K. Gupto Shin-lchi Hirano Basavarai V. Hiremath lau-Ho Jean Robert Pohanka
Published by The American Ceramic Society 735 Ceramic Place Westerville, Ohio 43081 www.ceramics.or g
Proceedings of the Advances in Dielectric Materials and Multilayer Electronic Devices Symposium at the I 02nd Annual Meeting of The American Ceramic Socieb St. Louis, Missouri, April 30-May 3, 2000.
Copyright 2002, The American Ceramic Society. All rights reserved. Statements of fact and opinion are the responsibility of the authors alone and do not imply an opinion on the part of the officers, staff , or members of The American Ceramic Society. The American Ceramic Society assumes no responsibility for the statements and opinions advanced by the contributors to its publications or by the speakers at its programs. Registered names and trademarks, etc., used in this publication, even without specific indication thereof, are not to be considered unprotected by the law.
No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording, or otherwise, without written permission from the publisher. Authorization to photocopy for internal or personal use beyond the limits of Sections 107 and 108 of the U.S. Copyright taw is granted by the American Ceramic Society, provided that the appropriate fee is paid directly to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923 USA, www.copyright.com. Prior to photocopying items for educational classroom use, please contact Copyright Clearance Center, Inc. This consent does not extend to copying items for general distribution or for advertising or promotional purposes or to republishing items in whole or in part in any work in any format. Please direct republication or special copying permission requests to Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923 USA 978-750-8400; www.copyright.com. Cover photo of U bluetooth radio module uppeurs courtesy of hicsson AB.
Library of Congress Catalogingin-PublicationData A CIP record for this book is available from the Library of Congress.
For information on ordering titles published by The American Ceramic Society, or to request a publications catalog, please call 614-794-
5890.
Printed in the United States of America.
4 3 2 1-05 04 03 02 ISSN 1042-1122 ISBN 1-57498-118-8
DEDICATION This proceedings is dedicated to Professor L.E. Cross, Pen nsy ban ia State University.
This page intentionally left blank
CONTENTS Introduction. ......................................................
xi
Lattice Defects in HydrothermalPerovskite Powders, Used for the Manufacture of MLCCS D. Hennings, C. Metzmacher, and S. Schreinemacher. ................................
1
Dilatometric Studies on BaTiO, Samples Added with Y,O, And ZrO, C. Gomez-Yanez and H. Balmori-Ramirez. .......................................
13
Reaction and Precipitation Mechanisms in the Low-Temperature Aqueous Synthesis of BaTiO, M. Viviani, M.T. Buscaglia, V. Buscaglia, P. Nanni, P. Piaggio, and P. Bowen . . . . . . . . . . . . . . . . . 25 Hydrothermal Synthesis and Characterization of Barium Titanate Powders E. Ciftci, M.N. Rahaman, and M. Shurnsky. .....................................
.35
Preparation of PZT Thin Film with Compositionally Gradient Buffer Layer by Pulsed Mo-Source CVD K. Shinozaki, A. Endo, A. Iwasaki, A. Saiki, N. Wakiya, and N. Mizutani . . . . . . . . . . . . . . . . . . .47 Factors InfluencingTexture Development in Hot Forged Bismuth Titanate I.S. Patwardhan and M.N. Rahaman. ..................................
.57
Effects of PbO Loss on Microstructural Development and Properties of PLZT Ceramics J. Feng and F. Dogan.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .69
A Study on the Effects of Lanthanum Doping on the Microstructure - 0.1 PbTiO, and Dielectric Properties of 0.9 Pb(Mg,,Nb,,)O,
M. Winter, S. Pilgrim, and M. lejeune. .........................................
77
Effect of the Milling Process on Core-Shell Microstructure for BaTi0,-Based Ni-MLCC Y. Mizuno, T. Hagiwara, H. Chazono, and H. Kishi. ................................
.95
vii
Interrelationship Between Self-Heating and Ferroelectric Properties in PZT Ceramics during Polarization Switching M. lente, D. Garcia, and I. Eiras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measurementsof Pyroelectric Response on Barium Strontium Titanate Single Crystal Fibers
D. Garcia, R. Guo, and A. Bhalla ............................................
105
113
Influence of Crystallizationon Structural and Electrical Properties of PZT Thin Films E. Arauio, D. Garcia, and I. Eiras . . . . . . . . . . . . . . . . . . . . . . . . .. .. .. .. .. .. .. .. .. ..
123
Crystallizationof Strontium Barium Niobate Ferroelectric Thin Films Produced by Chemical Method E.B. Arauio, R. G. Mendes, D. Garcia, and J. A. Eiras ...............................
131
Structural and Dielectric Characterizationof Amorphous SrTiO, Thin Film Prepared by Sol-Gel
E.R. Leite, F.M. Pontes, S.M. Zanetti, E. Longo, 1.A. Varela, and V. Mastellaro . . . . . . . . . . . . . . . 141
Characterization of Residual Stress on Pt/Ti Electrode of Infrared Sensor
K. Kim, M. Yoo, D. Kim, S. lee, and M. Lee. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ceramics Dielectric Properties in Relation with Grains Surface Fractal Nature
V. Mitic, 1. Kocic, and I. Mirtovic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Tailoring of Electromechanical Properties of Pb(Mg,,N b2,,)0,-PbTi0,-BaTi0,-Based Relaxors
C.H. Yoon, A. Sehirlioglu, S.M. Pilgrim, and K. Bridger ..............................
155
169
179
VHF Tunability Measurementsof Ferroelectric Materials Using Doubly Reentrant Cavities
R.G. Geyer and W.E. McKinzie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
High-Dielectric-ConstantCeramic-Polymer 0 - 3 Composites
Y. Bai, Z.3. Cheng, V. Gharti, H.S. Xu, and Q.M. Zhang ............................
Phase Constitution and Microwave Dielectric Properties of the ZNNB,O,-TIO, System
D. Kim, H. Youn, S. Cho, and K. Hong.. ............................
Processing and Properties of Strontium Bismuth Vanadate Niobates Ferroelectric Ceramics
187
.205
. . . . . . . . . . . 213
Y. Wu, M. J. Forbess, S. Seraji, S.1. Limmer, C.P. Nguyen, and G.Z. Cao. . . . . . . . . . . . . . . . . . . 221
...
Vlll
Synthesis, Processing, and Dielectric Properties of Compositions in the Strontium Titanate Strontium Zirconate Solid-Solution System
S.J. Lombardo, R.V. Shende, D.S. Viswanath, G.A. Rossetti, Jr., D.S. Krueger, and A.Gordon. ..............................................
Effect of Dy-Doping on Resistance Degradation of BTZ Sintered in Reducing Atmosphere under the Highly Accelerated Life Test W. Lee.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Effects of Barium Dissolution on Dispersing Aqueous Barium Titanate Suspensions
C. Chiang and J. Jean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Dependence of Dielectric Properties on Thickness (25nm-200nm) for Metal-Organic Chemical Vapor Deposited PZT Thin Films
C.H. tin, P.A. Friddle, X. Lu, and H. Chen. ......................................
Oxidation of CVD Diamonds: An Auger Electron Spectroscopy Approach
1.Y. Howe, L.E. lones, D.N. Braski, and W.D. Porter. ...............................
BaTiO, - Ceramics lntergranular Capacitors in Processing Microstructure - Property Relationship
V. Mitic, I. Mitrovic, and B. lordovic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
'
Hydrothermal Synthesis of Heteroepitaxial Barium Titanate Thin Films
E. Ciftci, M. Rahaman, M. Shumsky, and F.D. Blum ................................
Dielectric Properties of Barium Titanate Sintered with ZnO-Based Fluxes
D. Prakash, B.P. Sharma, P. Gopalan, and T.R. Ram0 Mohan ..........................
Characterizationof Ultra-Fine BaTiO, Powder for Multilayer Ceramic Capacitors
Y. Sakabe and N. Wada. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mechanical and Thermal Properties of Power Electronic Ceramic Multilayer Capacitors
A.A. Wereszczak, L. Riester, J.W. Hill, and S.P. Cygan. ..............................
Plasma and H personic Flame-Sprayed Ceramic Coatings for Dielectrica Applications
1
R. Gadow and A. Killinger. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Effect of DC Field on Dielectric Loss of SrTiO, Single Crystals and Thin Films
C. Ang, Z. Yu, R. Guo, and AS. Bhalla . . . . . . . . . . . . . . . . . ......................
Dielectric Properties of Layered Perovskite Sr,A,B,iN Ferroelectrics (A=La, Ca and x=O,O.l )
.227
.239
.247
257 263
269 279 289
.301
31 1
.323 .339
b,O,
M.J. Forbess, S. Seraji, Y. Wu, S.J. Limmer, C.P. Nguyen, and G.Z. Coo . . . . . . . . . . . . . . . . . . .349
ix
Contribution from Ferroelastic Domain Switching Detected by the X-rays to R-Curve Behavior of PZT Ceramics A.E. Glazounov, M.J. Hoffmann, A. Kolleck, and G.A. Schneider ........................
355
Interpreting Piezoceramic Impedance Measurements A. Ballato.. ........................................................
.369
Effect of Rare-Earth Addition on Microstructure and Electrical Properties in BaTi0,-Based Ceramics for Ni-MLCC H. Chazono and H. Kishi ................................................
.411
Conductivity and Modulus Spectra for a Series of LithiumBorate and Sodium Trisilicate Glasses A.E. Burns.. ..... .............................. .. . .. .. .. .. .. .. .. .. ..
.421
Composition and Temperature Dependence of Microwave Conductivity of Potassium Germanate Glasses S. Krishnaswarni, H. Join, and 0. Kanert ......................................
.431
Highly Accelerated Life Testing (HALT) of a K-4500 Low-Fired X7R Dielectric G.H. Maher.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.443
Ceramic Tapes for Wireless Applications
R.L. Wahlers, S.J. Stein, M.A. Stein, A.H. Feingold, and P.W. Bless ......................
457
low-Temperature Sintering MgCuZn-Ferrites for Muhilayer Ferrite Chips A. Nakano, I. Nakahata, T. Murase, and T. Nomura ...............................
.467
Effect of Rare-Earth Doping on the Temperature - Capacitance Characteristics of MLCCS with Ni Electrodes S. Sato, Y. Fujikawa, and T. Nomura. . . . . . . . . . . . . . . . . . .. .. .. .. .. .. .. .. .. .. .. .
.473
Use of Titanates to Achieve a Temperature Stable LTCC Dielectric for Wireless Applications S.X. Dai, R. Huang, and D. Wilcox Sr.. .......................................
-483
low-Temperature Co-Fired Ceramics and Their Applications K. Wakino, H. Mandai, and N. Nakajirna .......................................
593
Temperature-Dependent Polarization and Electric Potential on Ferroelectric BaTiOJ 100) Surfaces
S.V. Kalinin and D.A. Bonnell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
503
Antiferroelectricity-The Invisible Hand Behind Good Ferroelectrics I. Chen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.519
Microwave Dielectric Property Measurements R.G. Geyer and J. Krupka . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.533
Index.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.561
X
INTRODUCTION L.E. Cross, Father of Modern Electroceramics Rustum Roy Founding Director; Materials Research Lab, Pennsylvania State University
There may be some debate about the parentage of modern ceramics in American universities, in which there is no doubt that faculty at Penn State, Ohio State, and Illinois played major roles. However, there is perhaps no contest in identifying (Leslie) Eric Cross as the father of modern electroceramics. Since the birth of electrical machinery, mica, a natural single crystal mineral (not a polycrystalline ceramic), dominated the world of insulating oxide materials in all electrical systems (In 1945 I was sent to the United States to Penn State to save India’s export market in mica!). In that very year the seeds of its replacement were sown simultaneously all over the world by the discovery of BaTiO, the ceramic dielectric with a k one thousand times greater than mica (see Ref. 1). I have shown elsewhere that this was possibly the greatest step-function discovery in the history of materials science. High Tc superconductors pale in significance compared to this. Eric Cross, then at Leeds in England, was to start his career riding this crest of ferroelectricity, which was soon found to be the cause of BaTi0,’s high permittivity. In 1962 when I was appointed director of Penn State’s newly minted Materials Research Lab, I set about carving a unique space in the materials world for this laboratory. It clearly could not have an emphasis in metals or semiconductors (or polymers). But Penn State was already a powerhouse in “traditional” ceramics (I got my degree in it in 1948): whitewaves, refractories, glass, etc., with luminaries such as Nelson W. Taylor, E.D. Henry, Woldemar Weyl, N.J. Kreidl, George Bair, etc. My mentor, E.F. Osborn, officially professor of geochemistry, had already introduced in 1947 as part of high-science ceramics, albeit, always applications driven-the detailed precise thermodynamics of oxide systems via the determination of phase diagrams. By 1962, Penn State dominated that field, including the new high-pressure aspects, and one thrust of MRL-the chemical side-became the scientific study and practice of novel materials synthesis.
xi
In 1962 the semi-conductor revolution preoccupied nearly all the MRLs of the nation as the physics community got involved with materials. Our choice was therefore to avoid the herd and focus instead on semi-insulators. In the 1950s Penn State already had such a major effort under Prof. R. Pepinsky in the Physics Department, and my synthesis work meshed well with theirs on ferroelectrics; it was with that we were able to attract Eric Cross to join his fellow “Luddite,” G.W. Brindley, at our faculty. Cross became associate director of M U and eventually succeeded me as director for the period 1985-1989. And that made it possible for applications-driven research on electroceramics to remain the focus of MRLs interface with the world of electronics for 40 years. We had explicitly and repeatedly avoided the enormous national push toward mechanical properties of high-tech ceramics; events proved that we had made the right choice. In the next decade under the leadership of Eric Cross, joined soon by Robert E. Newnham, Penn State’s MRL assumed a hegemonic position in then the most significant part of modem ceramics. What is most significant about Eric Cross’ work in the field are two of its aspects:
1. It is always application driven. Eric always can see the connection to a real application. Even in the most esoteric thermodynamic calculations such as those by Aizu, he can find a connection to the real world. 2. He does not mine out the same old territory over and over again. The mark of real creativity is versatility. The following list is truly astonishing. It lists the different areas in which Eric Cross has contributed and illustrates how much the ceramics and electronics community owes him. The second major contribution that Eric Cross has made is shown in the right hand column-the students and post docs he has “trained.” It can truly be said that my colleagues Cross and Newnham together have literally given birth to a whole national and international family of researchers in electroceramics. And that, in the long run, is what the ”fatherhood” of electroceramics is about. And last but not least is the superb human dimension that Eric and his wife Lucilla have exemplified within the MRL community; personal concern for students, associates, family and community, and a commitment to excellence in research and elsewhere. References 1 “Memories of the Early Days of BaTiO,” by Cross and Newnham in “Kyoui no Chitabari” Edited by Prof K. Wakino, Murata CO,Japan (1990).
xii
MAJOR AREAS OF CROSS’ RESEARCH (with collaborators noted)
Origin of high permittivity in fine grain barium titanate.
A.K. Goswami W.R. Buessem
A three-dimensional kittel type analysis of antiferroelectric: ferroelectric switching.
Clevite Corporation
Bismuth titanate studies. Symmetry and structure.
R.E. Newnham, J. Dorian, s. Wolf R.C. Pohanka S.E. Cummins, T. Luke Fousek: Janovec theory
Phenomenological analysis. Optical image storage. Detailed domain analysis. Peculiar ferroelectricity in Gd,(MoO,),. Origin of whole new field of improper ferroelectrics.
A. Fouskova
Importance of phase connectivity in piezoelectric composites: underpinning all future electro-composite work.
R.E. Newnham D. Skinner
Lithium thallium tartrate tetrahydrate. First piezoelectric crystal with
E. Sawaguchi K. Seely
b90% Hydrophone piezocomposites of controlled connectivity. 1 : 3 Composites
K. Klicker, T.R. Gururaja, A. Halliyal J. Ginewicz A. Safari
0 : 3 Composites 3 : 0 Composites Electrostrictive ceramics. Lead magnesium niobate. Surface deformable mirror. Hubble tilt mirror correctors. Super paraelectric model. 0rder:disorder in lead schandium tantalite (PST). Kinetics of re1axor:ferroelectric switching.
S. Nomura K. Uchino, S. Jang T.R. Shout, S. Swartz C. Randall N. Setter Z. Chen Yao Xi
...
Xlll
Piezoelectric in ceramics. Full thermodynamic phenomenology for lead zirconate titanate (PZT). Phase switching AF:Fe actuators. Thin film piezoelectrics.
M. Haun, E. Furman S. Jang S. Yoshikawa, S.E. Park K.R. Udayakumar, D. Chen, K. Brooks B. Xu, C. Gaskey B. Xu J. Bernstein V. Kugel R. Kiu C. Fuller
Square loop AF thick films. Transversely poled double layer thick films. Bimorph based high displacement double amplifiers Single crystal piezoelectrics. Morphotropic phase boundary systems in PMN:PT PAN:PT. Intrinsic model predicting induced monoclinic phase. Optical verification in induced monoclinic phase. Evidence for a stable monoclinic phase in PZT.
S.E. Park T.R. Shrout L.E. Cross P. Hana Zuo G. Ye G. Shirane, B. Noheda S.E. Park, R. Guo
New high strain relaxor ferroelectric irradiated P(VDF:TrFe) copolymer transverse electrostrictive strain -4%. Induced piezoelectric K31 0.45. New chemical methods of inducing relaxor behavior.
Q. Zhang V. Bharti Z.Y. Cheng
-
FUTURE DIRECTIONS New Morphotropic Boundary Systems. Measurements of Flexoelectric Coefficients. Ultra High Strain Soft Elastomers. Induced Piezoelectric Effects.
xiv
LATTICE DEFECTS IN HYDROTHERMAL PEROVSKITE POWDERS, USED FOR THE MANUFACTURE OF MLCCS Detlev Hennings, Christoph Metzmacher and Seriyati Schreinemacher PHILIPS Research Labs, P.O.B. 500145,52085 Aachen, Germany ABSTRACT In ceramic multilayer capacitors (MLCCs) prepared from hydrothermal powders, a fine intra-granular porosity leads to cracks and delamination. The small intra-granular pores segregate as large pores at the inner electrodes, thus leading to “bloating” of the MLCCs, perpendicular to the electrodes. Hydrothermal BaTi03 and (Ba,Ca)(Ti,Zr)03 powders contain large numbers of hydroxyl groups (-OH) in the oxygen sub-lattice. The protons (H’) are compensated by vacancies on metal sites. On heating to 500 “C the point defects disappear, forming fine intra-granular pores of nanometer size. INTRODUCTION Multilayer ceramic capacitors (MLCC) can be produced with several hundred dielectric layers of <4pm layer thickness. High capacitance MLCCs of the EIA temperature specification Y5V and size 1206 have been thus manufactured with Ni electrodes’, showing capacitance values up to 100yF. The manufacture of MLCCs with thin dielectric layers requires very thin ceramic tapes showing a smooth surface which can only be prepared from fine ceramic powders. The most effective way to prepare such fine dielectric powders are wet chemical routes like hydr~thermal~-~, alkoxide’ syntheses or sol-gel preparation‘ . A general problem at sintering MLCCs with thin dielectric layers is the elimination of pores. Since pores can actually not penetrate the inner electrodes, they have to migrate along the dielectric layers to the end terminations. Dielectric materials prepared from hydrothermal powder usually contain large numbers of fine intragranular
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this ublication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright &earance Center, is prohibited.
Dielectric Materials and Devices
1
pores, Figure 1. During grain growth in the final stage of sintering these pores usually aggregate to large pores in the electrode region, Figure 2. Pore agglomeration at the electrodes then gives rise to the so-called “Bloating Effect”. At a certain temperature which is cIosely related to accelerated grain growth, the MLCCs stop shrinkage and begin to expand perpendicular to the electrodes. In the other two directions enhanced shrinkage is observed. It needs no further elucidation that bloating leads to severe mechanical deformation and failures in MLCCs which have been prepared from hydrothermal powder. This paper deals with the investigation of pore formation in hydrothermal powders, resulting from point defect in these powders.
Figure 1:
2
Sintered hydrothermal BaTi03, showing intra-granular porosity
Dielectric Materials and Devices
Figure 2:
Y5V MLCC prepared from hydrothermal (Ba,Ca)(Ti,Zr)03, showing pore aggregation at the electrodes.
EXPERIMENTAL Bloating of MLCCs was studied with TMA on MLCCs of the size 1206 (0.12” x 0.06”). The MLCCs contained 240 dielectric layers of 4pm sintered thickness which have been printed with Ni inner electrodes. Stacks of 4 green MLCCs were sintered in the temperature range 1200°C to 1320°C in a dilatometer (DIL 802 Bahr, Hullhorst/Germany), using a protective atmosphere of moist N2/H2, showing p02=10-’3bar, Figure 3. The dielectric materials were prepared either from hydrothermal (Ba,Ca)(Ti,Zr)03. (SAKAI BCTZ4) or solid state powder. To the BCTZ an appropriate mixture of donor and acceptor dopes was added7 in order to suppress the electron conduction caused by firing in reducing atmosphere*. Before sintering the organic binder was carefully burned out in a mixture of N2/02 at temperatures <350”C. The microstructure of the MLCCs was studied in SEM and light microscopy.
Dielectric Materials and Devices
3
U AL
Time - rel. Change in Length /Temperature
150C
T°C
(YQ) -5
1OO( -10 500
-1 5
-20
0 Figure 3:
100 200
300 400
500
'0
600 700 800 9000 1000
TMA (perpendicular to the electrodes) of YSV 1206 MLCCs, 22pF. Bloating Effect observed perpendicular to the electrodesin MLCCs prepared from hydrothermal powder.
As can be seen in Figure 3, MLCCs produced from conventional solid state powders do not show the Bloating Effect. It must be therefore assumed that special features of the hydrothermal BCTZ, as for instance the large number of lattice defects9 are responsible for the Bloating Effect. In temperature-stable MLCCs of the EIA specification X7R the coreshell structure" preserves the intra-granular porosity. In X7R materials the grain cores, consisting of undoped hydrothermal BaTi03 (CABOT-BT4) are surrounded by the grain shells which are obviously impermeable for the pores, Figure 4. For investigation of the crystallographic and defect chemical properties self-prepared4 and commercial hydrothermal BaTi03 (CABOTBT4) powders were heated at various temperatures between room temperature and 1200°C and thereafter quenched to room temperature. The hydrothermal BaTiO3 powders were of nearly stoichiometric composition, showing only slight deviations from the ideal BdTi atomic ratio of 0.9982BdTi11 .OOO. Using IPC, the impurity content of the hydrothermal powders was analyzed which was less than 100 wt. ppm in total. Thermogravimetric analyses (TGA, Netzsch STA 429) revealed
4
Dielectric Materials and Devices
considerable amounts of water in both hydrothermal powders. The morphology of the powders was studied with transmission electron microscopy (TEM) (PHILIPS CM12) and for high resolution in a PHILIPS CM20-FEG. The powder density of BT and BCTZ was measured, using He pycnometry, while the theoretical density was calculated from the pseudocubic lattice") constants.
Figure 4:
Intra-granular micro-pores in the grain cores of X7R ceramics, prepared from hydrothennal BaTi03.
RESULTS HRTEM of thermally untreated hydrothennal BaTi03 powders revealed undisturbed lattice fringes which were running from one to the other end of the grain. Intra-granular pores could not be detected even at the highest resolution. BaTi03 powders heated above >400"C exhibited in contrast large numbers of intra-granular pores, having diameters of 10-100 nm, Figure 5. After annealing hydrothermal BaTi03 and BCTZ powders to 1200°C the intra-granular pores disappeared. Such particles showed typical sintering effects and considerable grain growth.
Dielectric Materials and Devices
5
Figure 5 :
Intra-granular micro-pores in hydrothermal BaTi03, heated for 2h at 500°C.
XRD revealed a pseudo-cubic' perovskite lattice in as-prepared hydrothermal BaTi03. The pycnometric density of such powder was only ~94% of the theoretical value. The density of solid state prepared BaTi03 showed in contrast almost the theoretical value of p=6.05 g/cm3. The density of annealed hydrothermal BaTi03 powder remained unchanged up to 400°C. Above 800°C the density rapidly increased and reached the theoretical value at ca. 1100°C. In heat treated (>500"C) hydrothermal BaTiO? XRD revealed a slight increase of the unit cell volume from 0.064 (nm)3 to ~ 0 . 0 6 6 However, this slight increase of the unit cell volume cannot explain the strong decrease of the powder density as shown in Figure 6. The low density of hydrothermal BaTiO3 powder observed in the temperature range 500"-800°C can be easily explained by the occurrence of intra-granular porosity. The increase of density after firing at >800"C is then explained by the removal of the intra-granular porosity. However, the large discrepancy between the theoretical and pycnometric density, Figure 6, in the temperature range 20"C-4OO0C needs further explanation. Since no intra-granular porosity could be detected in thermally untreated
6
Dielectric Materials and Devices
hydrothermal BT and BCTZ, point defects must be made responsible for the low density of these powders. In the following the low density of thermally untreated hydrothermal BaTi03 is discussed in terms of a defect chemical model, derived for hydrothermal BaTi03.
Figure 6:
Theoretical (calculated from XRD data) and experimental powder density of hydrothermal BaTi03 as function of the annealing temperature
DEFECT CHEMISTRY OF HYDROTHERMAL BaTi03 As has been shown earlier9>12, hydrothermal BaTiO3 contains large amounts of water which evaporates during heating in the temperature range 100"-500"C. IR and Raman spectroscopy revealed that hydroxyl groups (-OH). are incorporated in the perovslute lattice. Correponding to the large number of OH-, a similar large number of rotons exists in the oxygen sublattice which amounts up to 0.4 mol [H'] per mol BaTi03. The enlarged unit cell volume is closely correlated to the high number of proton defects.
1
Dielectric Materials and Devices
7
The protons on the oxygen sites must be compensated anyway by other point defects to keep up charge neutrality in the perovskite lattice. At least two defect chemical models have been disc~ssed”.’~ for hydrothermal BaTi03. In the case of exclusive Ba- or Ti-vacancy compensation large amounts of Ba or Ti rich second phases should be detectable in a product showing NB-1 gross composition4. At a proton content of -40 mol% [H’] about 20 mol% BaC03 must be formed in the case of Ba vacancy compensation and -10 mol% Ti02 at Ti vacancy compensation. However, in hydrothermal BaTi03 powder heated at 800°C neither considerable amounts of BaC03, Ba2Ti04 nor Ti-rich phases, e.g. Ba6Ti17040, were detected with XRD. In a monophase perovskite showing a correctly adjusted atomic ratio of A/B=1.000 (+0.002)4 the number of vacancies on A- and B-sites must be therefore equal to each other. The defect chemical formula for BaTi03, containing x mol [OH’] per mol perovskite is then
Based on the defect chemical model with Ba and Ti vacancies, and using the experimental’ concentrations of [HI’, the theoretical density of hydrothermal BaTi03 was re-calculated as function of the annealing temperature, Figure 7. In the low temperature region (<200”C) the theoretically calculated density agrees well with the experimental density. The low density of asprepared, thermally untreated hydrothermal powder can be thus considered as a result of the large number of vacancies on Ba and Ti sites. These small point defects, however, could not be made visible in HRTEM. On heating, with decreasing number of protons and metal vacancies the calculated density approaches the theoretical value of p=6.05 g/cm3 at about 500°C.
8
Dielectric Materials and Devices
Figure 7:
Theoretical density of hydrothermal BaTi03, calculated on the base of the point defect model. Comparison to experimental density.
FORMATION OF INTRA-GRANULAR PORES
In spite of the decreasing number of point defects the experimental powder density remains low up to temperatures of 800°C. The discrepancy between calculated and measured density can be explained by the formation of intra-granular pores. The intra-granular porosity becomes visible in TEM after annealing the powders in the temperature range 400"C-800"C. At even higher temperature the intra-granular porosity disappears again and the experimental density approaches that of normal defect free BaTi03. We therefore assume that the intra-granular porosity of hydrothermal barium titanate originates from the point defects. It has been suggested that oxygen vacancies may be transiently formed at the evaporation of water in the temperature range 100"-500"C.
Dielectric Materials and Devices
9
The coexistence of high numbers of Ba, Ti and oxygen vacancies gives rise to a high instability of the perovskite lattice. We have therefore to assume that the point defects on the various lattice sites combine and annihilate each other. The different charges of the point defects compensate each other to neutrality : VA” + VB’l’l
+ 3 Vo**= Nil
At 500°C nearly all point defects have disappeared. The unit cell volume and the theoretical density then take on both the values of normal BaTi03.
The small discrete volume of the point defects is now collected in the intragranular pores. After annealing at higher temperature (>400”C) the small pores have grown to larger ones of 20-200 nm size which can be easily detected with conventional TEM. At temperatures < 800°C the mobility of the intra-granular micropores is very low, so that the pores stay inside of the grains. In the course of accelerated grain growth the majority of the intra-granular pores moves out to the grain boundaries at T>800”C. In thin layer MLCCs the pores aggregate to large pores in the electrode region, where they give then rise to the Bloating Effect and mechanical failures. REFERENCES 1 Y.Yoneda, YKawaguchi, FKobayashi, and Y.Takagi, “Electrical Characteristics and Physical Structure of MLCC with Ultra-high Capacitance,” CARTS-EUROPE 99: 13fhEuropean Passive Components Symposium, 18th-22ndOctober 1999 Lisbon, Proceeding p. 183 2 R.Vivekanandan, S.Philip, and T.R.Kutty, “Hydrothermal Preparation of Ba(Ti,Zr)03 fine Powders”, Mat. Res. Bull 22,99-108 (1986) 3S.Hirano, “Hydrothermal Processing of Ceramics”, Ceram. Bull. 66, 1342 (1987) 4 D.Hennings, G.Rosenstein and H.Schreinemacher, “Hydrothermal Preparation of Barium Titanate from Ba-Ti-Acetate Gel Precursors”, J. Europ. Ceram. Soc. S, 107-115 (1991). ’MKosec and B.MaliE, “Chemical Homogeneity and Microstructure of Alkoxy-Derived PZT Powders”, Proceedings of Electroceramics IV, 4th
10
Dielectric Materials and Devices
Internat. Conf. on Electroceramics & Applications Sept. 5-7. 1994 Aachen, Germany, p. 1245-1250 [ed.. R.Waser, Aachen 1994 ISBN 3-86073-287-01 ‘M. Klee, “Spin-on Processing of Perovskite Thin Films for Innovative Microelectronic Devices”, Proceedings of Electroceramics IV, 4thInternat. Conf. on Electroceramics & Applications Sept. 5-7. 1994 Aachen, Germany, p. 1225-1223 [ed.. R.Waser, Aachen 1994 ISBN 3-86073-287-01 7D.Hennings, K.Albertsen, P.Hansen, and O.Steigelmann,“DonorAcceptor Charge Complex Formation in Barium Titanate Ceramics”, Ceramic Transactions, Multilayer Electronic Ceramic Devices 97,(1999) 41-5 1 [The Am. Ceram. Soc] 8J.Daniels,K.H.Hardt1, D.Hennings and R.Wernicke, “Defect chemistry and electrical conductivity of doped barium titanate ceramics”, PHILIPS Research Reports 31,487-560 (1976) 9 D.Hennings and S.Schreinemacher, “Characterisation of Hydrothermal Barium Titanate”, J. Europ. Ceram. Soc., %(1992) 41-46 10 D.Hennings and G.Rosenstein, “Temperature-stable Dielectrics based on chemically inhomogeneous Barium Titanate”, J. Am. Ceram. Soc. 67, 249-54 (1984) 11 I.J.Clark, T.Takeuchi, N.Ohtori, and D.Sinclair, “Hydrothermal Synthesis and Characterization of BaTi03 Fine Powders: Precursors, Polymorphism and Properties”, J. Mater. Chem. 1999,9-83-91 12 Guido Busca, Vincenco Buscaglia, Marcello Leoni and Paolo Nanni, “Solid State and Surface Spectroscopic Charaterisation of BaTi03 Fine Powders”, Chem. Mater. 6-955-61 (1994).
Dielectric Materials and Devices
11
This page intentionally left blank
DILATOMETRIC STUDIES ON BaTi03 SAMPLES ADDED WITH Y203 AND ZrO2 C. Gomez-Yanez and H. Balmori-Ramirez Department of Metallurgical Engineering, ESIQIE-IPN U. P. Adolfo Lopez Mateos, Zacatenco, D. F. , Mexico C.P. 07301, A. P. 75-593, Tel.: (5)729-6000 Ext: 54208, Fax: Ext:55270 ABSTRACT BaTiO3 samples with TiO2 in excess and additions of 0.12-0.30 mol % Y203 or/and 1-3 wt % 2 1 - 0 2 were analyzed in a themomechanical analyzer using temperatures up to 1500 "C. Additions of ZrO2 diminish the amount of the phase BaTi17040 probably by chemical reactions that consume TiO2. With 1 wt % of Zr02 and a sintering at 1300 "C for 1 hour, the grain growth increases, but at higher 2 3 - 0 2 concentrations there is a refinement of the grain size. On the other hand, additions of 0.12-0.24 mol % Y203 enhance the formation of the liquid phase resulting from the melting of BaTi17040, however, with a concentration of 0.3 mol % of Y203 the mentioned liquid phase is not formed. The addition of Y203 does not affect practically the average grain size, but homogenizes the grain size distribution, impeding in this way the abnormal grain growth present in microstructures added with Zr02. INTRODUCTION BaTi03 is widely used in the electronic industry to fabricate capacitors, thermistors, piezoelectric sensors, etc. Commonly, several kinds of doping elements such as Nb, La, Ca, etc. are added to BaTi03 to improve the operational characteristics in accordance with a specific application. Using Zr as dopant, the capacitor characteristics are improved Also, additions of Y and Zr to BaTi03 increase the PTC effect in thermistors 293. The addition of dopants, however, influences the densification process during sintering and, therefore, the final microstructure of the product. Since electrical properties depend on the microstructural characteristics, it is important to have a detailed knowledge about the densification process. According to T. Armstrong et al. the addition of 1 and 2 wt % ZrO2 to BaTiO3 produces a decreasing in density when the ZrO2 concentration increases, and also, they found a desintering process at temperatures higher than
'.
%-the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication,reproduction, or re ublication of h s publication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright ClearanceCenter, is prohibited.
Dielectric Materials and Devices
13
1320 "C, attributed to increased intragranular porosity resulting from discontinuos grain growth. Adding up to 20 wt % 2 3 - 0 2 and sintering at 1300 "C for 1 hour, relative densities between 95 and 97 % have been obtained. 475 Using sintering at 1450 "C for 6 hours and higher ZrOz content (up to 30 at %), relative densities of 97-98 % were achieved. When Y is added to the BaTiO3 the mean grain size and the relative density (90-96 %) increase when the (0-1 at %) Y concentration is increased. When a relatively high content of Y is used (1-20 at %) Y2Ti207 and YBasTi2Os.s have been detected. 7,8 EXPERIMENTAL PROCEDURE In polypropylane barrels (250 ml) with 2 3 - 0 2 milling media, deionized water and mixtures of BaTi03 powder (Merck), 1-3 wt % Zr02 powder (Tosoh- TZ-0) and/or 0.31-0.78 wt YOYC13.6H20 (Aldrich) (equivalent to 0.12-0.3 mol %O Y203) were mixed during 24 hours. In the BaTi03 powder, a Ba/Ti=0.92 ratio was found using Energy Dispersive Spectroscopy (EDS) technique. After mixing, the powders were dried in an stove at 70 "C. The powder mixture was consolidated by cold isostatic pressing at 294 MPa. Whenever YC13.6H20 was added, the mixture powder was calcined at 1000 "C for 1 hour in order to produce Y2O3 in the powder mixture. Green pieces were introduced in a thermomechanical analizer (Setaram TMA-92), using a heating rate of 5 "C/min. Sintering at 1300 or 1400 "C for 1 hour with 10 "C/min as heating and cooling rates, was carried out. On the other hand, whenever 2 1 - 0 2 was included in the mixture the heating and cooling rates during sintering were 10 "C/min except between 800 and 1200 "C where such rates were 5 "C/min to avoid cracking due to the allotropic transfomation from monoclinic to tetragonal structrure in 2 1 - 0 2 which occurs around 950 "C. Relative densities were determined by the Archimedes method. After conventional polishing and thermal etching (1250 "C for 20 min) the sample microstructures were observed in an scanning electron microscope (Jeol 6300) and the average grain and pore sizes were determined by image analysis (Image Pro 3.0). RESULTS Y2O3 additions The relative densities of the fired bodies are shown in table I where it is possible to observe that the density increases when the doping concentration of Yzo3 increases. The shrinkage of the samples added with 0 to 0.30 mol YOY203 as a hntion of the temperature is presented in figure 1. The densification behavior is clearer when the derivative of the displacement, y, with respect to the time, dy/dt, is plotted against the temperature (Figure 2). In the curve corresponding to BaTi03 withouth additions (Figure 2), two peaks are noted; one at -1210 "C and other at -1320 "C. The peak at -1210 "C corresponds to the sintering of pure BaTi03 10
14
Dielectric Materials and Devices
while the peak at -1320 "C corresponds to the fusion of the phase BaTi17040 which is formed by an excess of TiO2 during the powder processing. TiO2 is normally added in order to improve the densification by sintering in presence of a liquid phase. When 0.12 mol ?40 Y203 is added, the two mentioned peaks are not well defined (Figure 2), however when the Y2O3 concentration is incremented to 0.24 mol % the peak at -1320 C disappears but the peak at -1210 "C is as intense as in the case for pure BaTi03 and a new peak is observed at 1264 "C.
[mol %] Relative density Relative density Sint.: 1300 "C Sint.: 1400 "C 0 91.8 90.8 0.12 93.7 96.3 0.24 I 93.7 I 96.8 I
Y203
I
91.8 94.3 95.0 90.4
0 1 2 3
90.8 96.0 96.4 93.O
Heating rate
1000
1100
1200
1300
s" C h i n
1400
1500
Temperature [ C]
Figure 1 Displacement curves of BaTi03 samples as a function of the addition of Y203 [mol TO].
Dielectric Materials and Devices
15
0.0
n
.-c E
-0.5
\
s z
U
c,
)r
-1.0
Heating rate: 5 OC/rrin s
1000
1100
1200
1400
1300
1500
Temperature [ OC] Figure 2 Curves of dy/dt against temperature of BaTi03 samples as a hnction of the addition of Y203 [mol '701. After sintering at 1300 or 1400 "C for one hour, the average grain and pore sizes increase when the Y2O3 concentration is augmented (Figure 3) although the increment is not big. A slight change in the increasing tendency of the average grain size is observed at 0.3 mol '70Y203. 10
Sint.Time 1 Hour n
E A.-N
a --
6
--
m
v1
a, 0
grain 1400°C
-
grain 1300°C 4 --
pore 1400"C 2--
0 1 4.1
I
I
I
I
0
0.1
0.2
0.3
0.4
Y2 0 3[mol%]
Figure 3 Grain and pore growth as a fhction of the addition of Y203 and the sintering temperature.
16
Dielectric Materials and Devices
ZrO2 additions. It is evident from table I that the addition of 1 and 2 wt % ZrO2 improves the density of the sintered pieces. With 3 wt % 2 1 - 0 2 the increment in density with respect to BaTi03 without additions is not significant. In figure 4 the displacement curves when the temperature is raised for several ZrO2 concentrations can be seen. In figure 5 the corresponding dy/dt against temperature curves are presented. When 1 wt % 2 1 - 0 2 is added to BaTi03, two peaks appear; one at 1163 "C and other at 1210 "C. When 2 or 3 wt % 2 r 0 2 are added to BaTi03 three peaks are observed; with 2 wt % 2 3 - 0 2 the peaks appear at 1110, 1216 and 1250 "C and for the sample with 3 wt % ZrO2 the peaks correspond to 984, 1210 and 1267 "C and, in this sample, a small "shoulder" can be seen at 1300 "C which is within the temperature range of liquid phase formation in the case of BaTi03 with no additions. The peaks around 1100 "C correspond with the allotropic transition of 2 1 - 0 2 from monoclinic to tetragonal structure. 9
-E
0.0 -5.0
CI
c
a,
50 -10.0
m a .'1-15.0 n
I 1000
I
1100
I
U
I
1200
1300
1400
1500
Temperature [ %] Figure 4 Displacement curves of BaTiO3 samples as a fhction of the addtion of 2rO2 [wt %I. The addition of 2rO2 has an important influence on the microstructure as can be seen in figures 6 to 8. A remarkable grain growth occurs with 1 wt % Zr02 and a sintering at 1300 "C for one hour but grain refinement is obtained at the same sintering process with higher Zr02 concentrations (Figure 6 and 8). Also, the duplex microstructure at 2 wt % 2 1 - 0 2 is noticeable. When sintering is carried out at 1400 "C for one hour, the average grain size grows when the ZrO2 concentration increases (Figures 7 and 8) and any of the samples did not show duplex microstructure.
Dielectric Materials and Devices
17
--
0.0
-5.0
-Heating rate:
-10.0 t 1000
II
II
1100
1200
5" C/min
I
I
1300
1400
1500
Temperature [ O C ] Figure 5 Curves of dy/dt against temperature for BaTiOs samples as a fbnction of the addition of 2 3 - 0 2 [wt %I. -
50
Sint. Time 1 Hour
40
30 20 10
0 -1
0
1
zro;!
2
3
4
[wt%]
Figure 6 Grain and pore growth as a fbnction of temperature
25-02
18
Dielectric Materials and Devices
concentration and sintering
Figure 7 Microstructures of BaTi03 with a) 0, b) 1, c) 2 and d) 3 wt % 2 1 - 0 2 sintered at 1300 "C.
Figure 8 Microstructures of BaTi03 with a) 0, b) 1, c) 2 and d) 3 wt % 2 1 - 0 2 sintered at 1400 "C. Y203 and 2rO2 additions. Except for the sample with 2 wt YOZrO2 and 0.12 mol YOY203 in which the density seems to be low, the addition of both Y203 and 2 1 - 0 2 dopants together does not seem to have an important effect on the density of the sintered pieces (Table 11) as compared to the addition of single Y203 or ZrOz (Table I).
Dielectric Materials and Devices
19
Y2O3 [mol %] 0 0.12 0.24 0.30 zr02 [wt %]
Relative density Relative density Sint.: 1300 "C Sint.: 1400 "C 95.0 90.3 2 94.5 93.4
0.24
95.2
Theoretical density: 6.01 g/cm
Since the ZrO2 concentrations used in this work are relatively higher than the Y203 concentrations, the sintering seems to be dominated by the Zr02 additions as suggested by figure 9. The notorious abnormal grain growth presented by the sample with 2 wt % ZrO2 and sintered at 1300 "C for one hour is diminished by increasing concentrations of Y203 as showed in figure 10. The changes in the average grain size provoked by additions of ZrO2 are partially neutralized when 0.24 also added to BaTiO3 (Figure 11). I
Y
0.0
mol % Y *O
-0.5 'C/min
-1.o
3 wt % ZrO2+O.24 mol % Y2 0,
1
I
I
I
I
Figure 9 Curves of dy/dt against temperature for BaTi03 with several concentrations of Y203 and/or 2 1 - 0 2 .
20
Dielectric Materials and Devices
Figure 10 Microstructures of BaTi03 + 2 wt % ZrO2 samples with additions of a) 0, b) 0.12, c) 0.24 and d) 0.30 mol % Y203. 20
15 --
10 --
5
-Sint.: 1300 't= 1 Hour
01 1
I
II
I
I
0
1
2
3
zro,
4
[wt%]
Figure 11 Grain growth in BaTi03 samples as a function of the concentration of Y203 and/or ZrO2. ANALYSIS Y203 additions Figure 2 suggests that the addition of 0.12 and 0.24 mol YOY2O3 obstaculizes the densification at low temperatures of BaTiO3 since the peak at -1210 "C becomes difised but as simultaneous effect, the peak corresponding to the liquid phase formation becomes more intense (-1320 "C). With 0.3 mol YOY203 however, the densification up to 1200 "C looks similar to that for BaTi03 without additions
Dielectric Materials and Devices
21
although the peak at 1320 "C does not appear, instead, one peak at 1264 "C is evident. This behavior suggests the probable existance of transient compounds like YBa3TizO8.5 in whose formation some Ba is consumed, increasing in this way the excess of TiO2 and so, the amount of BaTi17040. On the other hand, the curve corresponding to 0.3 mol % Y203 (Figure 2) suggests the formation of a Ticonsuming compounds like Y2Ti05 or Y2Ti207. 798 additions It is apparent from figure 5 that addition of 1-3 wt % 2 1 - 0 2 decreases the amount of BaTi17040 and, since the peak around 1210 "C is present in all concentrations, it is reasonable to suppose that the main interaction occurs between 2 1 - 0 2 and BaTi17040. It is known 3*11 that the difision of Zr into BaTi03 is a slow process by which when the amount is high enough, ZrO2 segregates in the grain boundary impeding in this way, the grain growth, or when the amount of 2 1 - 0 2 is not enough to be homogeneously distributed throughout the sample, the abnormal growth will be favored. When the sintering temperature is raised the difision improves and so the grain growth.
23-02
Y203 and ZrO2 additions Although there is some controversy about the position that the Y3' ion would occupy into the BaTiO3 lattice, 2 let us assume in sake of the present analysis, that Y3+ occupies the Ba2+ sites. In such case the corresponding lattice distortion generated by this sustitution would be counterbalanced by the distortion povoked by the sustitution of Ti4+by the Zr4- ion since the ionic size of Y3+is larger than that of Ba2-, and the ionic size of Zr4+ is smaller than that of Ti4'. l2 Hence, the net distortion of the BaTi03 lattice decreases and the grains growth can be more pronounced when both dopants are added together. This asumption is supported by the microstructures shown in figure 10.
CONCLUSIONS Additions of Y2O3 or 2 3 - 0 2 improve the sintering characteristics of BaTiO3. Dilatometric results suggest that some transient compounds are produced during sintering of BaTiO3 with additions of Y203 (0.12 - 0.3 mol %) and/or ZrO2 (1-3 wt%) . Y203 additions With 0.12 and 0.24 mol % Y203 it is apparent that the amount of B~Ti17040is incremented while with addition of 0.3 mol % Y203 the phase B~Ti17040is depleted. The average grain size slightly increases as a hnction of the Y203 concentration.
22
Dielectric Materials and Devices
Zr02 additions The results suggest that ZrO2 reacts with the BaTi17040 in such a way that the later phase is depleted. When a sintering at 1300 “C for one hour is carried out and 1 wt % ZrO2 is used, a noticeable grain growth is observed while with 2 or 3 wt YO ZrO2 a grain refinement is obtained. A duplex microstructure is produced with 2 wt % ZrO2. When the sintering temperature is raised to 1400 “C the average grain size increases when the 2 1 - 0 2 concentration is augmented. Y203 and ZrO2 additions Since the lattice distortions produced by insertion of Y3+and Zr4’ ions into the BaTi03 lattice are counterbalanced, the abnormal grain growth produced by ZrO2 is diminished by fbrther additions O f YzO3. REFERENCES 1 D. F. K. Hennings, B. Schreinemacher and H. Schreinemacher, “HighPermittivity Dielectric Ceramics with High Endurance”, Journal of the European Ceramic Society, 13 8 1-88 (1994). 2 R. C . Buchanan, T. R. Armstrong and R. D. Roseman, “Influence of Grain Boundary, Defect and Internal Stress States on Properties of Ferroelectric Materials”, Ferroelectrics, 135 343-368 (1992). 3 T. R. Armstrong, L. E. Morgens, A. K. Maurice and R. C. Buchanan, “Effects of Zirconia on Microstructure and Dielectric Properties of Barium Titanate Ceramics”, Journal of the American Ceramic Society, 72[4] 605- 11 ,(1989). 4 Y. Hirata y T. Kawazoe, “Wet Forming, Sintering Behavior and Dielectric Properties of BaTi0.8Zr0.203”,Journal of Materials Reserch, 11 [ 121 307 1-76 (1 996). 5 Y. S. Her, E. Matijevic and M. C. Chon, “Controlled Double-Jet Precipitation of Uniform Colloidal Crystalline Particles of Zr and Sr Doped Barium Titanates”, Journal of Materials Research, 1 1[ 121 3 121-3 127 ( 1996). GD. Hennings, A. Schnell, and G. Simon, “Diffise Ferroelectric Phase Transition in Ba(Til-, Zr,)03 Ceramics”. Journal of the American Ceramic Society, 65[ 111 539-544 (1982). 7P. Blanchard, J. F. Baumard and P. Abelar, “Effect of Yttrium Doping on the Grain and Grain Boundary Resistivities of BaTi03 for Positive Temperature Coefficient Thermistor”, Journal of the American Ceramic Society, 75[5] 1068-72 (1992). 8 J. Zhi, A. Chen, Y. Zhi, P. Vialrinho and J. L. Baptista, “Incorporation of Yttrium in Barium Titanate Ceramis”, Journal of the American Ceramic Society, 82[5] 1345-48 (1999). 9 R. H. J. Hannink, P. M. Kelly and B. C. Muddle, “Transformation Toughening in Zirconia-Containing Ceramics”, Journal of the American Ceramic Society, 83 [3] 461-87 (2000)
Dielectric Materials and Devices
23
Z. Chen and T. A. Ring, “Sintering of BaTiOs” in Ceramic Transactions, Vol. 32, Dielectric Ceramic Processing Properties and Applications, edited by K. M. Nair, J. P. Guha and A. Okamoto, American Ceramic Society, Westerville, 275-284, 1993. 11 T. R Armstrong and R. C. Buchanan, “Influence of Core-Shell Grains on the Internal Stress State and Permittivity Response of Zirconia-Modified Barium Titanate”, Journal of the American Ceramic Sociev, 73[5] 1268-73 (1990). 12 W. D. Kingery, H. K. Bowen and D. R. Uhlmann, “Introduction to Ceramics”, page 58, John Wiley & Sons, New York, 1976. 10
24
Dielectric Materials and Devices
REACTION AND PRECIPITATION MECHANISMS IN THE LOWTEMPERATURE AQUEOUS SYNTHESIS OF BaTi03 M. Viviani*, M.T. Buscaglia and V. Buscaglia Institute for Physical Chemistry of Materials - National Research Council Via De Marini, 6 I- 16149 Genoa, Italy
P. Nanni Chemical and Process Engineering Department - University of Genoa P.le Kennedy, Pad. D I- 16129 Genoa, Italy
P. Piaggio Chemistry Department - University of Genoa Via Dodecaneso, 3 1 I- 16100 Genoa, Italy
P. Bowen Powder Technology Laboratory, Materials Department Swiss Federal Institute of Technology Lausanne (EPFL) CH- 1015 Lausanne, Switzerland
ABSTRACT In past years it has been demonstrated the possibility to obtain nanosized powders of Barium Titanate by precipitation in aqueous medium through the reaction Ba(OH)2 + Tic14 + 4NaOH -+BaTi03 + 4NaCl+ 3H20 carried out at atmospheric pressure and 80 "C. In this work, the precipitation process was studied by means of X-ray diffraction, specific surface area measurements and scanning electron microscopy . Two different mechanisms of precipitation and crystallisation are reported, depending on the concentration of cationic species. In particular, [Ba2'] plays a critical role. For concentrations higher than 0.1 1 M the reaction leads directly to fully crystalline round-shaped particles with size in the range 20-50 nrn. At [Ba2'] = 0.06M the product is an amorphous Ti-rich precipitate (BdTi = 0.4) which slowly transforms into faceted BaTi03 particles by a dissolution-precipitation process. This conclusion is supported by SEM observation of the particle size and morphology, which are incompatible with in-situ transformation. For intermediate concentrations, amorphous and crystalline phases coexist in the precipitate. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright ClearanceCenter, is prohibited.
Dielectric Materials and Devices
25
INTRODUCTION Barium Titanate (BaTi03) is one of the most important electroceramic materials due to the intrinsic dielectric and ferroelectric properties at room temperature (permittivity = 1000, spontaneous polarisation = 0.1 Cim'), the existence of a ferro-paraelectric phase transition near room temperature (1 20 "C) and the possibility to deeply modify the electronic conduction, the temperature coefficient of resistivity and the figure of merit by doping'. Besides the conventional preparation by high-temperature solid state reaction, a number of chemical routes2 that generally allow to obtain finer powders at lower temperatures, in some cases with the utilisation of inexpensive precursors, have been proposed. Among those syntheses, the hydrothermal method represents a low-temperature route which finds application in the commercial preparation of BaTi03. It consists in the transformation of a fine TiO2 crystalline powder in contact with an aqueous solution of Ba(OH)2 into a closed volume at temperatures ranging from 75 to 400 "C. The product is composed by BaTi03 crystallites in the range of tens to hundreds of nm, retaining in the lattice a certain amounts of OH . Such a process can be modified by the introduction of Ti in more reactive forms, like amorphous particles or gels, composed by polimeric TiOz.xH20, which react with Ba2+ions to form crystalline BaTi03. A different pathway (Low-Temperature Aqueous Synthesis, LTAS) for the preparation of BaTi03 was proposed by Nanni et al.9 and consists in the reaction between pure liquid Tic14 and Ba(OH)2 solution carried out at 80t90 "C in the presence of NaOH. Similarly to hydrothermal precipitations, BaTi03 particles prepared by LTAS are single crystallites of 30-50 nm with cubic structure and hydroxil groups retained into the lattice. A major difference is that direct precipitation of BaTi03 takes place into the Ba solution without any preliminar formation of solid precursors. Whilst the thermodynamics of hydrothermal synthesis of BaTi03 have been clarified and assessed by Lencka and Riman3:4,the reaction mechanism, though intensively studied, is still matter of discussion. In the absence of CO2 and at 90 "C, BaTi03 is the only stable phase for pH > 13.5 and [Ba2+]> 1O-'M. Little modifications in the preparation process, like nature of precursors, mixing technique and temperature, can strongly influence the reaction and are probably responsible for the contrasting results present in the literature. Kutty and Padmini', starting from reactive gels, obtained from the condensation of TIC14 in ammonium hydroxide, and Ba(0H)z were able to follow the progressive crystallisation of BaTiO3 through an in situ reaction, kinetically controlled by diffusion of Ba into the Ti-gel. They report the existence of two regimes: at low initial Ba concentrations ([Ba2']0
26
Dielectric Materials and Devices
concentration ([Ba2']0>O. 15 M) the conversion is rapid and completed after 50-60 minutes. A predominant in situ transformation mechanism was also re orted by Hert16 for the reaction between Ti02 anatase and Ba(OH)2 solution ([Ba 10 = 0.26.94M) on the basis of a detailed kinetic study. Pinceloup et al.7 reported the evidence of a dissolution-precipitation mechanism in the hydrothermal synthesis of BaTi03 starting from aqueousalcoholic solutions of Ba(OH)2 and Ti-isopropoxide. Morphology and structure of powders reacted at different temperatures and for various times are consistent with the dissolution of amorphous Ti02.xHzO and precipitation of crystalline BaTi03, while the absence of partially crystallised particles is incompatible with any in-situ mechanism. A comprehensive survey of possible reaction mechanisms who considered in hydrothermal synthesis of BaTi03 was given by Eckert et both in situ transformation of Ti02 particles in BaTi03 and dissolutionprecipitation reactions. Accurate quantitative X-ray diffractometry of partially reacted powders and kinetic analysis clearly indicated the existence of two kinetic stages, the first one being a dissolution-precipitation mechanism, followed by a slower transformation controlled by Barium ions diffusion. Recently, a statistical study of the effects of LTAS synthesis conditions on the size of particles of crystalline BaTi03 pointed out the significant importance of small variations of concentration of the initial precursors and mechanical energy transferred to the system". In the present work those investigations have been extended to precipitation amorphous powders, observed at low [Ba2+]oand partially crystallised by prolonged ageing, in order to obtain information about the reaction mechanisms involved in the LTAS technique.
R
EXPERIMENTAL Lots of 60 g of powder were prepared by the Low-Temperature Aqueous Synthesis, i.e., through the reaction Ba(OH)2 + Tic14 + 4NaOH + BaTi03 + 4NaC1+ 3H20
(1) carried out by dropping the titanium chloride into an alchaline (pH=l4) aqueous solution of Ba(OH)2 at 80 "C, under N2 at atmospheric pressure. Mixing took 15 minutes to complete and was realised into a PTFE vessel with high stirring rate (100 rpm), followed by ageing at the same temperature with reduced stirring (10 rpm). Details of the instrumented plant and overall procedure can be found elsewhere". Various batch experiments were carried out with different values of the starting barium concentration ([Ba2']0) in the range 0.06 M to 0.47 M. In addition, aliquots of thoroughly mixed slurry were extracted from reactor during ageing to study time evolution of powders from early stages (0.25 h) to long times (100 h). Precipitates were separated from slurries by centrifugation: powders were
Dielectric Materials and Devices
27
repeatedly washed with distilled water and investigated by electron microscopy (SEM, Philips 5 15, Eindhoven, The Netherlands), X-ray diffractometry (XRD, model PW 1710, CO K a radiation, secondary graphite monochromator, Philips, Eindhoven, The Netherlands) and surface gas adsorption (BET, model Gemini 2360, Micromeritics, Norcross, GA, USA). A careful XRD investigation was performed after thermal treatment at 950 "C for 4 h in air. In particular quantitative phase analysis was obtained by the Rietveld method, as implemented in the DBWS 9000 software' .
'
RESULTS AND DISCUSSION Under standard conditions, namely when [Ba2']0 > 0.15 M, reaction (1) was very fast and no morphological evolution of precipitate particles could be observed on the time scale considered. Powders were composed of crystalline(cubic simmetry) spherical particles with average size in the range of 30-50 nm. An example of size and morphology of a powder precipitated with [Ba2+]o=0.31 M and aged 100 h is presented in figure 1.
Figure 1. SEM image of a BaTi03 powder precipitated with [Ba2+Io=0.31 M and aged for 100 h.
On the contrary, when [Ba2+]owas reduced, a certain amount of amorphous phase was present as indicated by a broad contribution to XRD profiles between 20 and 50" 28. The intensity of such a contribution decreased with ageing, revealing a progressive crystallisation of the amorphous particles. When [Ba2+]o= 0.06 M, the powder was fully amorphous (at least within XRD detection limit) after precipitation, with BdTi ~ 0 . 4As . shown in figure 2, at that concentration BaTi03 reflections could be individuated only after several hours of ageing and the crystallisation process was still not completed after 100 hours.
28
Dielectric Materials and Devices
Stoichiometry of the precipitated amorphous phase was studied by XRD on samples thermally treated at 950°C for 4 h. Figure 3 reports the XRD patterns of three powders precipitated with [Ba2']0
T
T
210 140 70 0 20.0
25.0
30.0
35.0
40.0
45.0
28
50.0
55.0
60.0
65.0 70.0
Figure 2. X-ray diffraction profile of as-prepared powders precipitated with [Ba2+],,=0.06 M and aged for different times. T=BaTi03, C=BaC03.
From the following figure it is evident that an increasing amount of phases with BdTi ratio <1 (BaTi2Os and BaTi13030) appeared as barium concentration decreased, suggesting a correlation between [Ba2']0 and the amount of Ba effectively contained into precipitates. The weight fraction of the different crystalline phases was calculated by the Rietveld method and used to determine the overall stoichiometry of powders as a function of [Ba2']0 and time. It is important to note that XRD give an indirect measurement of the Ba content of precipitates and rely on the assumptions: the change in overall precipitate composition is only due to formation of BaTi03, thermal treatment at 950 "C for 4 h doesn't affect the quantity of BaTi03 but just let the amorphous portion of precipitates crystallise. First assumption was verified by the observation of constant precipitate composition in samples prepared from [Ba2']0 = 0.06 M and for ageing time 5 10 h, i.e. before spontaneous crystallisation of BaTi03. Second condition can be justified from results obtained on samples with 0.4
Dielectric Materials and Devices
29
1300’
1
1
1170’-
1
1040t
4910
4780 650
520
390 260 130
n 20.0
25.0
35.0
30.0
20
40.0
__ 45 0
Figure 3. X-ray diffraction profile of heat-treated (950 “C for 4 hours) powders precipitated at different [Ba2’Io and aged for 0.25 hours. l=BaTi03, 2=BaTi205,3=Ba4Ti13030.
cationic ratio lower than 1, like BaTi205 and B~Til3030.On the contrary, BaTi03 was detected after thermal treatment when it was already present as a crystalline phase in the as-prepared powder. Therefore, it can be assumed that the treatment applied wasn’t sufficient to reach equilibrium but only to transform the amorphous phase. Results of Rietveld calculations are plotted in Figure 4 together with specific surface area (SBET) data measured on the same samples. The existence of two regimes for the precipitation and crystallisation of BaTi03 is strongly supported by these curves. At the first stages of the process and low initial Ba concentration, the precipitate is mainly composed by an amorphous and Ti-rich phase, whose formation is inhibited (or its transformation to crystalline BaTi03 is accelerated) when Ba concentration is raised. It is therefore possible to identify two kinetic domains characterised by either amorphous or crystalline product (see fig. 4). In the “crystalline” domain, particle size corresponds to crystal size and SBET dependence on [Ba2+]ois consistent with classical homogeneous nucleation from solution, where the size of stable nuclei depends on the inverse of solute concentration’2. In the “amorphous” domain SBETis not related to crystal size but is mainly affected by the high porosity of the amorphous precipitate, as expected for gels of Ti-hydroxo polimers. The mechanism leading to the formation of BaTi03 crystallites is also clearly different in the low-concentration regime: the morphology of the final product, obtained after long ageing times, characterised by large (200-300 nm) squared crystals and completely different from the fine rounded particles obtained in the high-concentration regime, can be ascribed to a
30
Dielectric Materials and Devices
dissolution-precipitation mechanism. An example of that morphology is shown in figure 5 , coresponding to a powder precipitated from [Ba2’]0 = 0.06M and aged for 1OOh. 1.2
7 amorphous
T
crystalline
I6O
1
0
0.8
0.6
0.4 0
0.1
0.2
0.3
0.4
0.5
0.6
[Ba2+1(M)
Figure 4. Specific surface area (A) and cationic ratio ( 0 ) of powders aged for 5 hours as a function of Ba2’ starting concentration. Continuous lines connect experimental points and guide the eye. Dashed line represents the boundary between “crystalline” and “amorphous” domains.
Figure 5. SEM image of a BaTi03 powder precipitated with [Ba2+]o=0.06M and aged for 100 h
Dielectric Materials and Devices
31
Finally, it is important to note that like in all batch type reactors, also in our case the nuclei and precipitated particles didn’t experience exactly the same conditions and each sample separated from the solution at fixed time was actually composed by parts with a different history. In addition, as the mixing of precursors took place in a finite time, a great difference in the Ba2’ concentration between the beginning and the ending of precipitation has to be considered. The validity of the results presented here has to be therefore intended as average values which, however, fit with a consistent interpretation of phenomena. CONCLUSIONS The structure and size of BaTi03 prepared by the Low-Temperature Aqueous Synthesis are greatly influenced by the initial barium concentration ([Ba2’]o). Two regimes of precipitation have been individuated: at high [Ba2+]othe product is fully crystalline and the particle size is in the range of 30-50 nm, with roundshaped morphology, while at low [Ba2+]othe precipitation of an amorphous Tirich phase is favoured. The composition of such amorphous precipitate varies with time and BaTi03 crystallisation takes place after several hours of ageing. Size around 200-300 nm and cubic-shaped morphology suggest that in this case crystallisation occurs by a dissolution-precipitation mechanism. ACKNOWLEDGMENTS Authors wish to thank Mr. A. Testino for his support in the B.E.T. measurements. REFERENCES A.J. Moulson and J.M. Herbert, Electroceramics, pp. 68-79. Chapman&Hall, London 1990. 2 P.Nanni, M.Viviani , V. Buscaglia, “Synthesis of dielectric ceramic materials”, pp.429-455 in Low and High Dielectric Constant Materials and their Applications, vol. 1. Edited by H.S. Nalwa. Academic Press, San Diego, 1999. 3 M.M. Lencka and R.E. Riman, “Thermodynamic modeling of hydrothermal synthesis of ceramic powders”, Chemistry of Materials, 5 6 1-70 (I 993). 4M.M. Lencka and R.E. Riman, “Hydrothermal synthesis of perovskite materials: thermodynamic modeling and experimental verification”, Ferroelectrics, 151 159-164 (1 994). 5 T.R.N. Kutty, P. Padmini, “Mechanism of BaTi03 formation through gel-tocrystallite conversions”, Materials Chemistry and Physics, 39 200-208 (1995). 6 W. Hertl, “Kinetics of bariu titanate synthesis”, Journal of the American Ceramic Society, 71 879-83 (1988). 7P. Pinceloup, C. Courtois, J. Vicens, A. Leriche and B. Thierry, “Evidence
32
Dielectric Materials and Devices
of a dissolution-precipitation mechanism in hydrothermal synthesis of barium titanate powders”, Journal of the European Ceramic Society, 19 973-977 (1999). 8 J.O.Eckert Jr., C.C. Hung-Houston, B.L. Gersten, M.M. Lencka and R. Riman, “Kinetics and mechanism of hydrothermal synthesis of barium titanate”, Journal of the American Ceramic Society, 79 2929-39 (1996). 9 P. Nanni, M. Leoni, V. Buscaglia, G. Aliprandi, “Low-Temperature Aqueous Preparation of Barium Metatitanate Powders”, Journal of the European Ceramic Society, 14 85-90 (1994). * o M. Viviani, J. Lemaitre, M.T. Buscaglia and P. Nanni, “Low-Temperature Aqueous Synthesis: a statistical design of experiment approach”, Journal of the European Ceramic Society, 20 3 15-320 (2000). 11 Wiles, D.B. and Young, R.A., “A new computer program for Rietveld analysis of X-Ray powder diffraction patterns”, Journal of Applied Crystallography, 14 149-151 (1981). l 2 J.A. Dirkensen and T.A. Ring, “Fundamentals of crystallisation: kinetic effects on particle size distributions and morphology”, Chemical Engineering Science, 46 2389-2427 (1991). l 3 H.M. O’Bryan and J. Thomson, “ Phase equilibria in the Ti02-rich region of the system BaO-TiOZ”, Journal of the American Ceramic Society, 57 522 (1974).
Dielectric Materials and Devices
33
This page intentionally left blank
HYDROTHERMAL SYNTHESIS AND CHARACTERIZATION OF BARIUM TITANATE POWDERS E. Ciftci and M. N. Rahaman University of Missouri-Rolla, Department of Ceramic Engineering, Rolla, MO 65409
M. Shumsky University of Missouri-Rolla, Graduate Center for Materials Research, Rolla, MO 65409 ABSTRACT Crystalline barium titanate powders were precipitated by a reaction between fine TiO2 particles and a strongly alkaline solution of Ba(OH)2 under hydrothermal conditions at 80 "C to 240 "C. The effects of the synthesis temperature and time on the characteristics of the BaTi03 powders were studied by X-ray diffraction, transmission electron microscopy , thermal analysis and atomic emission spectroscopy . The average particle size of BaTi03 increased from -50 nm to 100 nm after 24 hours of reaction at 90 "C and 240 "C, respectively. At synthesis temperatures below -150 "C, the BaTi03 particles had a narrow distribution of sizes and were predominantly cubic in structure. At higher synthesis temperatures, the amount of tetragonal phase increased with increasing temperature. At the highest synthesis temperature (240 "C for 96 h), the particles consisted of a mixture of the cubic and tetragonal phases with a bimodal distribution of sizes. The powders showed little weight loss when heated at temperatures between 100 "C and 600 "C. The influence of particle size and processing related hydroxyl defects on the crystal structure of BaTi03 is discussed.
-
INTRODUCTION Barium titanate is an important material in the electronics industry [ 1-31. Its high dielectric constant and low dielectric loss factor over a wide range of temperature and frequency make it desirable as a dielectric material for the manufacture of capacitors while its ferroelectric properties are exploited for applications such as piezolectric transducers. It is also increasingly used for the manufacture of nonlinear resistors with positive temperature coefficients (PTC resistors) for To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
35
measuring and control applications in the electronic and electrical industries. Barium titanate materials are commonly fabricated by the conventional ceramic processing route involving the consolidation and sintering of BaTi03 powders. The common industrial routes for the preparation of BaTi03 powders are calcination of a mixture of BaC03 and TiO2 powders or thermal decomposition of barium-titanyl oxalate [BaTiO(C204)2-4H20] at temperatures in the range of 8001100 "C [4]. Bonding and agglomeration of the product are common features of these solid state reactions and milling is normally required to achieve particle sizes in the range of microns. Barium titanate powders have also been prepared by chemical methods such as precipitation from solution or sol-gel processing [5,6]. In most of these liquid phase approaches, calcination at 700-800 "C is required to form the oxide or to form the crystalline phase, thereby leading to agglomeration of the product. A milling step similar to that in the solid state methods is therefore required to achieve controlled particle size characteristics. Hydrothermal synthesis provides an attractive route for the synthesis of fine crystalline oxide powders with controlled characteristics at relatively low temperatures. The process, involving precipitation from solution at temperatures typically between the boiling point and critical point of water (100 "C and 374.2 "C) in an autoclave, has been used for decades for the synthesis of fine oxide powders [7]. Crystalline BaTi03 powders have been synthesized at temperatures between -100-200 "C by reacting fine Ti02 particles in a strongly alkaline solution (PH > 12) of Ba(OH)2 [8]. Other titanium sources such as TiC14, titanium alkoxide and Ti02 gels have been used at temperatures in the range of 100-400 "C [9]. Hydrothennal BaTi03 powders typically have fine particles sizes in the range of 100400 nm and a narrow distribution of sizes rrfaking these powders highly sinterable as well as attractive for the production of thin dielectric layers. Hydrothermal BaTi03 powders show a number of structural characteristics that are not observed for powders prepared by conventional solid state reaction at higher temperatures. X-ray diffraction of hydrothermal BaTi03 powders, particularly those synthesized at lower temperatures, reveals a cubic structure that is normally observed only at temperatures above the ferroelectric Curie temperature of 125-130 "C. The possible causes for the apparent cubic and non-ferroelectric structure of fine BaTi03 particles are not clear and have been discussed elsewhere [lO]. They include (i) the idea of a critical size for ferroelectricity arising from factors such as depolarization effects and the absence of long-range cooperative interactions and (ii) particularly for chemically prepared BaTi03 powders, the presence of a high concentration of charged point defects that might upset the long range polar ordering that drives the cubic to tetragonal structural transformation on cooling. For hydrothermal BaTi03 powders (particle size -200 nm) prepared from acetate precursors, Hennings and Schreinemacher [ 111 showed that the development of room-temperature tetragonal structure after heat treatment
36
Dielectric Materials and Devices
was closely associated with the elimination of hydroxyl defects in the structure and not with particle growth. However, for hydrothermal BaTi03 prepared from hydrolyzed titanium alkoxide and Ba(0H)Z solution, Begg et a1 [ 121 concluded that the cubic to tetragonal structural transformation was not associated with the removal of hydroxyl groups during heating but was essentially dependent only on the particle size. The investigation of sol-gel processed BaTi03 polycrystals by Frey and Payne [ 101 indicates that a more complex evolution of subtle structural changes takes place. Barium titanate that were cubic in structure according to Xray diffraction (XRD) and free of hydroxyl defects displayed Raman spectra attributed to the orthorhombic phase. Reduction in grain size was found to enhance the stability of the orthorhombic phase at room temperature. Raman activity for XRD-cubic materials appeared not to be associated only with the presence of hydroxyl defects in the structure. Furthermore, the room temperature tetragonal structure appeared not to be tied directly to the removal of the hydroxyl groups. With increasing grain size from 35 to 100 nm, the room temperature XRD patterns and the Raman spectra exhibited the characteristics of the tetragonal phase. In the present work, the synthesis of hydrothermal BaTi03 powders by a reaction between fine Ti02 particles and a strongly alkaline solution of Ba(OH)2 was investigated. The objective of the work is to understand how the synthesis parameters influence the composition and structure of the particles in order to achieve powders with controlled characteristics. Since a most distinctive feature of fine BaTi03 powders is the tendency to exhibit a XRD-cubic structure at room temperature, the relationship between the XRD structure and the size for this processing method is reported. However, the work does not attempt to determine a cause for the size effect on the structure. EXPERIMENTAL Barium titanate powders were synthesized by reacting Ti02 powder (Degussa Corp., South Planefield, NJ) in an aqueous solution of Ba(0H)z (pH > 14). The Ti02 powder, average particle size ~ 2 m, 5 consisted of -30 weight percent (wt%) rutile and -70 wt% anatase. In the experiment, 8 g of Ba(OH)243H20 (Aldrich, Milwaukee, WI) was added to 12 cc of deionized water in a Teflon-lined autoclave (45 ml capacity; Parr Instrument Co., Moline, IL). The system was purged with argon, sealed and heated to 80 "C until the Ba(OH)2.8H20 dissolved. Two grams of Ti02 was then added to the solution and the system was sealed and heated to the required temperature (in the range of 80 "C to 240 "C) for different times (1 to 48 h). After the reaction, the product was washed with formic acid (-0.1 molar) to remove any BaC03 and then washed with deionized water. The powder was dried in an oven for 24 h at 85 "C. The phase composition, structure and crystal size of the powders were determined by X-ray diffraction (XDS 2000; Scintag Inc., Sunnyvale, CA) using Ni
Dielectric Materials and Devices
37
filtered CuK, radiation (A = 0.15405 nm) in a step-scan mode (28 = 0.01" per step). X-ray diffraction patterns were analyzed using computerized software (Shadow and Riqas; Materials Data, Inc., Livermore, CA) to determine the crystal size as well as the concentration of the cubic and tetragonal phases in the powder. The Shadow and Riqas pattern analysis programs use a direct convolution method in profile and whole pattern fitting [13]. Refined patterns were used to determine the unit cell dimension ratios, the line splitting and the crystal size. The latter is used for quantitative analysis of each phase in the powders based on the Rietveld structure refinement and whole pattern fitting of X-ray diffraction patterns. The two diffraction lines at 28 values of 38.9" and 83.3" were employed in the determination of the crystal size by X-ray line broadening because they represent the only lines in the cubic and tetragonal phase pattern which do not undergo broadening due to line splitting. In the absence of lattice strain, the broadening of these two lines can be attributed to the crystal size effect. The morphology and size of the powders were observed in a transmission electron microscope (Philips EM43OT). The average particle size was determined by measuring the maximum diameters of more than 200 particles. Thermal analysis ("etzsch STA 409), involving thermogravimetric analysis (TGA), differential thermal analysis (DTA) and differential scanning calorimetry (DSC), was performed by heating the powders (previously dried for 24 h at 85 "C) in air at 2" C/min to 1200 "C. The elemental composition of the powder was determined by inductively coupled plasma (ICP) atomic emission spectroscopy (Acme Analytical Labs., Vancouver, BC, Canada). A preliminary examination of the sintering characteristics of the powder was performed by heating the compacted powder (green density M 0.60 of the theoretical) in air in a dilatometer (1600C, Theta, Port Washington, N Y ) at 5 "C per minute to 1300 "C. RESULTS X-ray diffraction revealed that considerable BaTi03 powder was formed after only 1 h of reaction at 150 "C (Fig. 1). Rietveld analysis showed that the product consisted of -93 wt% cubic BaTi03 and -7 wt% unreacted TiOz (rutile). After 12 h, the product consisted entirely of cubic BaTi03. Prolonging the reaction time to 48 h produced a mixture of the cubic phase (-95 wt%) and the tetragonal phase (-5 wt%). The synthesis of BaTi03 was also investigated as a function of reaction temperature for a fixed time of 24 h. It was found that the powders synthesized below 100 "C contained a small concentration of unreacted Ti02 but above this temperature, phase pure BaTi03 was obtained. Furthermore, below -160 "C, the powders were cubic while at higher temperatures, the formation of the tetragonal
38
Dielectric Materials and Devices
1000
-$
750
0 3
-
0 h ._
c
U)
2
c -
500
250
0 I
I
.
21.1276, Rutile T102
I
I
I
.
I
I
.
. 31.0174, Ban03
-
Barium Tmanium Oxide
I
2-Theta(y
Figure 1. XRD patterns of the BaTi03 powder synthesized at 150 "C showing the increase in the formation of BaTi03 and a corresponding decrease in Ti02 after 1, 3,6,12 and 48 h. phase started to become significant. However, even at the highest synthesis temperature (240 "C), complete formation of the tetragonal phase was not obtained. The splitting of the X-ray diffraction peak at -45" 28, indicative of the presence of tetragonal BaTi03, is shown in Fig. 2 for powders synthesized for 24 h at several temperatures. The peak splitting becomes apparent at -180 "C. The concentrations of the cubic and tetragonal phases, as determined by Rietveld analysis, is shown in Fig. 3 for powders synthesized for 24 h at temperatures between 80 "C and 240 "C. According to the data in Figs. 2 and 3, the appearance of the peak splitting for the powder synthesized at 180 "C corresponds to a tetragonal content of -10 wt%. Furthermore, at the highest synthesis temperature (240 "C), the concentration of tetragonal BaTi03 is only -30 wt%.
Dielectric Materials and Devices
39
1500
500
160 0 4 5
44.0
44.5
45.0
45.5
46.0
2-Theta(O)
Figure 2. XRD patterns of the BaTi03 powders synthesized for 24 h at temperatures of 160 "C, 180 "C, 220°C and 240 "C, showing peak splitting above 180 "C as a result of increasing tetragonal phase content. 120
100
:
n
80
Y
QI
U
60
ID
v
5
40
A
c 20
0
0
50
100
150
200
250
300
Te mpe ra tu re
Figure 3. Room temperature X-ray diffraction tetragonal and cubic phase content for the BaTi03 powders synthesized for 24 h at various temperatures.
40
Dielectric Materials and Devices
Figure 4 shows TEM micrographs of the powders synthesized for 24 h at 90 "C, 160 "C, and 240 "C as well as the powder synthesized for 96 h at 240 "C. At lower temperatures [Figs. 3(a) and 3 (b)], the particles appear to have a fairly narrow distribution of sizes. Coarser faceted particles are apparent in the powder synthesized at 240 "C and a bimodal distribution of larger faceted particles and smaller particles is clearly observed for the powders synthesized for 96 h at 240 "C. The average particle size data determined by X-ray line broadening and by TEM for powders synthesized for 24 h at temperatures between 80 "C and 240 "C, are shown in Fig. 5. The data obtained by the two methods agree to within +lO%. They show an increase in the average particle size from -50 nm at 80 "C to -100 nm at 240 "C.
Figure 4. TEM micrographs of the BaTiO3 powders synthesized for 24 h at 90 "C (a), 160 "C (b), and 240 "C (c), and for 96 h at 240 "C (d).
Dielectric Materials and Devices
41
Figure 5. Average particle size determined by X-ray line broadening and by TEM for the BaTi03 powders synthesized for 24 h at various temperatures. Figure 6 shows the change in mass as a function of temperature, measured by TGA, for powders synthesized for 24 h at 80 "C, 160 "C and 240 "C. The total
Figure 6. Weight loss as a function of the heating temperature determined by TGA for the BaTi03 powders synthesized for 24 h at 80 "C, 160 "C, and 240 "C.
42
Dielectric Materials and Devices
weight loss between 100 "C and 600 "C is <2 wt% for the powder synthesized at 80 "C and decreases significantly with increasing temperature of synthesis, becoming < O S wt% at 240 "C. The Ba:Ti elemental ratio for the powders synthesized for 24 h at different temperatures, determined by atomic emission spectroscopy, is shown in Table 1. The ratio is relatively low (0.90) at 100 "C which may be due to the presence of unreacted Ti02. However, at temperatures above -200 "C, the ratio (0.99) is very close to the stoichiometric value of 1.O. Table 1. Ba:Ti elemental ratio for the BaTi03 powders synthesized for 24 h at various temperatures. Synthesis Temperature ("C) 100 160 200 240
BaO Content (wt%) 56.27 55.28 57.58 58.17
TiO2 Content (wt%) 35.96 37.03 34.45 33.92
Ba:Ti Ratio 0.90 0.96 0.99 0.99
DISCUSSION The reaction conditions employed in the hydrothermal synthesis experiments are consistent with thermodynamic predictions for the Ti-Ba-H20 system [ 141 which indicate that BaTi03 is the thermodynamically favored phase at pH > 12 and for high Ba2+concentration (2 M). Similar reaction conditions of pH and Ba2+ concentration were employed by Chien et a1 [ 151 for the synthesis of BaTi03 particles at ambient pressure and temperatures less than 100 "C. Based on these experiments, the thermodynamic foundation for the formation of BaTi03 appears reasonably established. The reaction mechanism is not clear but it has been suggested to involve a dissolutiodprecipitation process in which Ti is hydrolyzed to form either Ti(OH)t- [ 161 or Ti(OH)4 [ 171, followed by subsequent reaction with Ba2' ions to precipitate BaTi03. The mechanism of nucleation and growth of the BaTi03 particles is also not clear. In the present system, heterogeneous nucleation can occur on the fine Ti02 reactant particles, as observed by Chien et a1 [15]. Subsequent growth of the BaTi03 particles can occur by aggregation to form clusters, as proposed for amorphous particles prepared by precipitation from solution [ 181 or by Ostwald ripening in which the smaller particles dissolve and precipitate on the larger particles [ 191. For crystalline particles synthesized under hydrothermal conditions, growth by aggregation may be possible in the early stages. However, since the hydrothermal particles are commonly observed to be single crystals [20], aggregation cannot play a significant role after the early stages.
Dielectric Materials and Devices
43
The data indicate that the reaction product is essentially single phase BaTi03 after only -1 h at a reaction temperature of -150 "C. This reaction time to form predominantly single phase BaTi03 will increase at lower temperatures and decrease at higher temperatures. Therefore, for the reaction time of 24 h used in the experiments, the system for the most part can be considered to consist of a dilute suspension of BaTiO3 particles undergoing coarsening by an Ostwald ripening process. Particles smaller than a critical size dissolve and precipitate on the larger particles and the average particle size increases. The WLS theory for Ostwald ripening of diliute suspensions predicts that the maximum particle size is approximately twice the average particle size [21-23]. Assuming a critical size for ferroelectricity, at lower temperatures, coarsening is expected to be slow, so that most of the BaTi03 particles may be expected to be below this critical size and therefore cubic in structure. At higher temperatures, faster coarsening leads to a significant fraction of the BaTi03 particles having sizes above the critical value for ferroelectricity so that the presence of the tetragonal phase becomes apparent. The X-ray crystal structure data of Fig 3 and the TEM micrographs of the synthesized particles (Fig. 4) are consistent with this picture. The presence of hydroxyl defects in the BaTi03 structure, as outlined before, has been associated with the room temperature cubic structure of hydrothermal BaTiO3 particles. As shown in Fig. 6, the powders prepared in the present work show little weight loss. Furthermore, the weight loss when heated in the temperature range of 300 "C to 600 "C (in which hydroxyl groups are expected to be decomposed) is <0.5 wt% for synthesis temperatures above 160 "C and <1 wt% for the lowest synthesis temperature of 80 "C. The presence of hydroxyl defects do not appear to play a significant role in the structure of the hydrothermal particles prepared in this work. CONCLUSIONS Hydrothermal BaTi03 powders synthesized by reacting fine Ti02 particles with a strongly alkaline solution of Ba(0H)z for 24 h have a room temperature cubic structure by X-ray analysis when the reaction temperature is below -150 "C. Powders synthesized above this temperature have a mixture of the cubic and tetragonal phases, with the content of the tetragonal phase increasing with increasing reaction temperature. The average particle size increased from -50 nm to -100 nm when the reaction temperature was increased from 90 "C to 240 "C. The powders showed little weight loss when heated at temperatures between 100 "C and 600 "C indicating that the concentration of processing related hydroxyl defects in the BaTi03 structure is insignificant. The room temperature X-ray diffraction structure of the powders appears to be consistent with a critical size for
44
Dielectric Materials and Devices
ferroelectricity and cannot be tied directly to the presence of hydroxyly defects in the structure. REFERENCES 1
L. E. Cross, “Dielectric, piezoelectric, and ferroelectric components”, Am. Ceram. Soc. Bull., 63 [4] 586 (1984). 2D. Hennings, “Barium Titanate-Based Ceramic Materials for Dielectric Use,” Int. J. High Tech. Ceram., 3 [2] 91 (1987). 3R. E. Newnhan, Rept. Prog. Phys., 52 [2] 123-156 (1989). 4A. Bauger, J.C. Mutin, and J.C. Niepce, J.Mater Sci., 18, 3041 (1983). ’D. Hennings and W. Mayr, J. Solid State Chem., 26,329 (1978). 6F. Chaput and Boilot. J.P, “Alkoxide-Hydoxide Route to Synthesize Ba Ti03-Based Powders”, J. Am. Ceram. Soc., 73 (4) 942-48 (1990). 7 W. J. Dawson, “Hydrothermal Synthesis of Advanced Ceramic Powders”, Ceram. Bull., 67 [lO] 1673-78 (1988) 8 A. T. Chien, J. S. Speck, F. F. Lange, A. C. Daykin, and C. G. Levi, “Low Temperature/Low Pressure Hydrothermal Synthesis of Barium Titanate: Powder and heteroepitaxial Thin Films,” J. Mater. Res., 10 [7] 1784-89 (1995). ’R. Vivekanandan and T. R. N. Kutty, “Characterization of Barium Titanate Fine Powders Formed From Hydrothermal Crystallization,” Powder Technol., 57 181-92 (1989). ‘OM. H. Frey and D. A. Payne, “Grain Size Efeects on Structure and Phase Transformations for Barium Titanate,” Phys. Rev., 54 [ 5 ] 3 158-68 (1996). ‘D. Hennings and S. Schreinemacher, “Characterization of Hydrothermal Barium Titanate,”J. Europ. Ceram. Soc., 9 41-46 (1992). 12B.D. Begg, E. R. Vance, and J. Nowotny, “Effect of Particle Size on the Room-Temperature Crystal Structure of Barium Titanate,” J. Am. Ceram. Soc., 77 [ 121 3 186-92 (1994). 13S.A. Howard and R.L. Snyder, “The Use of Direct Convolution Products in Profile and Pattern Fitting Algorithms I: Development of the Algorithm:”, J. Appl. Cryst., 22, 238-243 (1989). K. Osseo-Asare, F. J. Arriagada, and J. H. Adair, pp. 47- in Ceramic Transactions, Vol. 1. Ceramic Powder Science 11. Edited by G. L. Messing, E.R. Fuller, Jr., and H. Hausner. The American Ceramic Society, Westerville, OH, 1988. 15 A. T. Chien, J. S. Speck, F. F. Lange, A. C. Daykin, and C. G. Levi, “Low Temperature/Low Pressure Hydrothermal Synthesis of Barium Titanate: Powder and Heteroepitaxial Thin Films,” J. Mater. Res., 10 [7] 1784-89 (1995).
Dielectric Materials and Devices
45
16M. Yoshimura, S-E. Yoo, M. Hayashi, and N. Ishizawa, “Preparation of BaTi03 Thin Film by Hdrothermal Electrochemical Method”, Jpn. J. Appl. Phys., Vol. 28, NO. 11,2007-2009 (1989). 17 R. Bacsa, P. Ravindranathan, and, J. P. Dougherty, “Electrochemical, Hydrothermal, and Electrochemical-Hydrothermal Synthesis of Barium Titanate Thin Films on Titanium Substrates,” J. Mater. Res., 7 [2] 423-28 (1992). l 8 E. Matijevic, “Growth of Oxide Particles by Aggregation,” Langmuir, 4 3 1(1988). 19 S. Lakhwani and M. N. Rahaman, “Hydrothermal Coarsening of CeO2 Particles,” J. Mater. Res., 14 [4] 1455-61 (1999). 20M. N. Rahaman and Y. C. Zhou, “Effect of Solid Solution Additives on the Sintering of Ultrafine Particles,” J. Europ. Ceram. Soc., 15 [lO] 939-50 (1995). 21 G. W. Greenwood, “The Growth of Dispersed Precipitates in Solu tion,” Acta Metall., 4 243-48 (1956). 22 C. Wagner, “Theorie der Aterung von Niederschlagen durch Umlosen (Ostwald- Reifung,” Z. Electrochem., 65 581-91 (1961). 231.M. Lifshitz and V. V. Slyozov, “The Kinetic of Precipitation from Supersaturated Solid Solutions,” J. Phys. Chem. Solids, 19 [12] 35-50 (1961).
46
Dielectric Materials and Devices
PREPARATION OF PZT THIN FILM WITH COMPOSITIONALLY GRADIENT BUFFER LAYER BY PULSED MO-SOURCE CVD Kazuo Shinozaki, Ayanori Endo, Akinori Iwasaki, Atsushi Saiki, Naoki Wakiya and Nobuyasu Mizutani Tokyo Institute of Technology Department of Metallurgy and Ceramics Science 2- 12-1 Ookayama, Meguro-ku, Tokyo 152-8550, Japan ABSTRACT Epitaxially-grown Pb(Zro~5Tio,,)0,(PZT) thin-films with a compositionalgradient buffer layer Pb(ZrxTi,,)O, (x=O-0.5 and x=1-0.5) and PbTiO, buffer layer were prepared on 0.Swt%Nb-doped SrTiO,( 100) by the pulsed MO-source CVD. The Zr/Ti composition ratios of the deposited PZT thin films were successively controlled by the valve-opening-time ratio of Zr and Ti source gases. The gradient of composition in the buffer layer was confirmed by X P S depth profile analysis. All PZT films with different buffer layers, i.e., PbTiO,+PZT, PbZrO,+PZT and PbTiO,, were epitaxially grown with c-axis normal to substrate. Residual stresses in different PZT films were measured by the lattice spacing shift in XRD. All PZT films were under in-plane tensile stress because of lower thermal expansion coefficient of the substrate compared with those of PZT film. Introducing the buffers, in-plane tensile stress could be controlled. The PbTi0,PZT buffer and PbTiO, buffer acted for a stress relaxation compared with the PZT film directly deposited on the substrate. On the other hand, PbZr0,-PZT buffer introduced more tensile stress of PZT film. INTODUCTION Lead-based ferroelectric thin films such as PbTiO,(PT), PZT and PLZT have numerous promising properties such as unique dielectric, pyroelectric, electrooptic and piezoelectric properties. These films have been studied as sensors, actuators or FRAMs. Ferroelectric properties of the thin film are susceptible by the stress. In case of thin film, the anisotropic internal stress is caused by the substrate. Residual stress in the film is generally produced by the thermal ~~
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
47
expansion coefficient difference between the film and the substrate. When the tensile stress was induced in the c-axis-oriented PT thin films by applying inplane stress to the substrates after deposition, the remnant polarization (Pr) of PT film decreased?) This means that the Pr value can be controlled by changing the in-plane stress. When PbZrO,(PZ) film was deposited on the Pt( lOO)/MgO( 100) substrate, the PZ film was epitaxially grown and the PZ film with 87.5nm thickness showed the ferroelectric hysteresis loop, although the bulk PZ ceramics exhibit the double hysteresis loop which indicates antiferroelectricity.'2' The reason why the very thin PZ film showed a ferroelectric nature might be the residual stress of the film. Thus the control of the residual stress of the thin film is very important and may enable us to obtain properties superior to those of the bulk material. There are two main factors in the stress generation of the film. One is the thermal expansion coefficient (TEC) difference between the film and the substrate, and the other is the epitaxial strain caused by lattice mismatch between the film and the substrate. Various buffer layers were introduced for controlling the residual stress of the film. In this paper, the effects on the residual stress of various buffer layers in PZT/buffer/Nb-doped-SrTiO, structure were discussed. PZT thin film was used as a model thin film. Nb-doped SrTi0,(100) (Nb-ST) was used as a substrate. Three kinds of buffer layers were introduced, i.e., two compositional-gradient buffer layer, Pb(Zr,Ti,,)O,, changing x=O to 0.5 and X= 1 to 0.5 both from substrate to PZT film and PbTiO, buffer. The Nb-doped ST substrates have a similar lattice parameter and TEC value to those of PZT thin film. The pulsed MO-source CVD apparatus was used to introduce the compositional-gradientbuffer layer. EXPERIMENTAL PROCEDURE Pb-Zr-Ti-0 thin films were fabricated on 0.Swt%Nb-doped SrTiO,( 100) (NbST) and Si(100) substrates by the pulsed MO-source CVD (pulsed MOCVD). Pb(DPM),, Ti(0-i-C,H,), and Zr(O-t-C,H9), were used as source materials. The source gases were obtained by heating at 130"C, 45°C and 30°C in the oil bathes, respectively. Ar and 0, were used as the carrier gas and the oxidation source, respectively. The carrier-gas(Ar) flow rates through the source cylinders were 15sccm. The gas pressures in the reactor chamber and substrate temperature were fixed at O.8Torr and 650°C. RF plasma(5W) was used to enhance the reaction. Figure 1 shows the schematic diagram of pulsed MOCVD apparatus. Pb precursor that was supplied continuously to the reactor is not described in the figure. Zr and Ti precursors are supplied alternatively(Fig.2). The carrier-gas flow rate was usually controlled to obtain the desired composition of Pb(ZrXTi,~,)O,, However, a non-linear relationship between the gas flow rate ratio, ZrJ(Zrg+Tig),
48
Dielectric Materials and Devices
Fig. 1 Schematic diagram of Pulsed MO-source CVD.
Fig. 2 Schematic diagram of supply system of each precursors.
and film composition was rep~rted.‘~) Pulsed MOCVD method enabled the fabrication of the PZT films with the desired compositions and compositional gradient of Zr/Ti ratios by controlling the valve-opening time-ratios of Zr and Ti sources.(4)The mixed gases with the compositions of PbTiO, and PbZrO, were delivered to the substrate surface alternatively. Adjacent layers were mixed to form a PZT solid solution. The valve-opening-time (t,,: Zr valve opening time, tTi: Ti valve opening time) was initially fixed at t,, : tTi= 20sec : 20sec. The vaporization conditions of the source materials were determined to obtain the stoichiometric Pb(Zro,5Tio.5)03 film. The film deposition rate was 0.3nm per 20sec, that is, 5 O d l hour (180pulses). ARer attaining the Pb(Zr,,Ti,,,)O, composition, the valve opening times were varied. There was an excellent linear relationship between the predicted compositions from the valve opening time and the deposited film corn position^.^^) Simple valve control enables to fabricate the complex PZT films, such as the compositional-gradient andor multilayered film. The film composition was analyzed using an energy dispersive X-ray spectroscopy (EDX) apparatus (DX-95T, EDAX Corp.). Since the PZT film composition deposited on Nb-ST substrate is difficult to determine because of the coexistence of Ti, the film deposited on Si(100) was used. The depth profile of the composition of the thin film was investigated by X-ray photoelectron spectroscopy ( X P S ) using Ar sputter equipment (PHI5500, PERKIN ELMER). Phase identification and crystallographic orientation measurement of the films were carried out by 28-0 scan, pole figure measurements and reciprocal-space mapping by X-ray diffractometers (X’pert-MPD(e-e), X’ pert-MPD(Open Eulerian Cradle) and Extended X’pert Material Research Diffractometer, Philips). RESULTS Characterization of the PZT Thin Film with Various Buffer Layer The compositional-gradient buffer layer was fabricated as follows. For fabricating PT-PZT buffer layer, PT thin film with 5 nm thickness was deposited on Nb-ST substrate, then the Zr valve opening time (t,,) was linearly increased
Dielectric Materials and Devices
49
from lsec to 20sec over lhour. While tz, was increased, the Ti valve opening time (hi)was fixed at 20sec. PZ initial layer with 5nm was introduced in case of PZ+PZT buffer layer. Thus three types of films, (a) PZT(50nm)/[PZT+ PT PT(55nm)]/Nb-ST7 (b) [PZT (55nm)lNb-ST and (c) PZT(SOnm)/NbST were fabricated. The Zr/(Zr+Ti) compositions of PZT were all 0.5. Figure 3 shows the composition ratio (ZrPb, TdPb and (Zr+Ti)/Pb) +
Fig. 3 Composition ratio of the film with the structure PZT/[PZT+PT(S k m ) ] / Nb-ST by X P S .
calculated from X P S measurements as a function of sputtering depth in the film(a). This shows that the Pb and Ti composition are linearly changed with 2 keeping constant APB site ratio in .g perovskite not only in PZT layer, but 3 also in the buffer layer. Excess Pb was observed on the film surface. The 42 43 44 45 46 47 48 49 2 0 I deg. Pb On the presence Of Fig.4 X-ray patterns of (a) PZT/[PZT+PT(55 surface was common for every thin nm)]/Nb-ST, (b) [PZT+PT(55nm)]/Nbfilms fabricated in this experiments. STY(c) PZT/Nb-ST The 26-w scanning patterns of X-ray diffraction reveal that these films are highly c-axis oriented. The pole figure measurements also show that the films aligned with the substrate, i.e., these films were all epitaxially grown on the substrates. Figure 4 shows an example of the (b) [PZT+PT(SOnm)]/NbXRD patterns of (a) PZT/[PZT+PT(5Onm>]/Nb-ST7 ST and (c) PZT/Nb-ST. There is a broad peak (pointed by the arrow at Fig.4) at the lower 20 angle of the PZT(002) peak in the PZT film with buffer layer(a), which corresponds to the peak in the film of buffer layer(b). It is also clear that the d-spacing of PZT(002) peaks of all thin films in this experiments were smaller than that of the bulk PZT value. This is due to the in-plane tensile stress in the film. The c-axis lattice parameter was decreased due to tensile stress in order to maintain the unit cell volume. The PZT(002) peak positions are varied with different film configuration under in-plane tensile stress. Since the peak positions of the PZT(002) in film(a) and the broad peak(002) in buffer layer(b) are very close, the real position of the PZT(002) peak in film(a) was determined by subtracting the film(b) profile from the film(a) profile. The residual stresses in the buffer layers in film(a) and film(b) were different, since the extra stress was
.s
50
Dielectric Materials and Devices
induced by the deposited PZT film in peak(a). Two parameters were introduced at the subtraction procedure. One parameter changed ; the 20 value and the other changed '3 the intensity both of buffer peak in 5 film(b) for minimizing the trace of 42 42.5 43 43.5 44 445 45 buffer peak after subtracting in 2 8 I deg. The profile Of Fig.5 Subtraction of film (b) [PZT+PT(55nm)] film(a) is indicated as (a7) in Fig.5. /Nb-ST profile from film (a) PZT/[PZT+ Since the 20 values of both (a7)and PT(55nm)]/Nb-STprofile in Fig.4. (c) are smaller than that of the bulk [(a')=(a)-(b)l PZT(002) peak angle (28=43.663) as mentioned previously, both films (a) and (c) are subjected to in-plane tensile stress. In Fig.5, the 20value of the PZT(002) peak(a') of film(a) is 0.14 degrees lower than that of film(c). This means that the relaxation of the tensile stress between the PZT thin film and the substrate occurred due to the existence of the buffer layer. We proposed the asymmetric X-ray diffraction method to evaluate the residual stress in epitaxial thin film.(5)This method evaluate the residual stress more exactly. Since the interaction between PZT thin film and the buffer layer was complicated as stated above, simple calculation of the residual stress (0)in the film has been done using lattice strain E in the following equations (1) and (2). E = (4hkl-dhkl)/d,,hkl (1) o = E ' E / (1-V) (2) where dhkland d,,hkl are d spacing in a film and a bulk, respectively. E is Young's modulus and v is Poisson7s ratio. The d spacing value measured by 00 0 peak
.s< 4 a
M
Ml1
Reciprocal lattice spacing Reciprocal lattice spacing parallel to substrate parallel to substrate Fig.6 Reciprocal-spacemapping around PZT and Nb-ST (103) peak of the films (1) PZT/Nb-ST and (2)PZT/ [PZT+PT]/Nb-ST
Dielectric Materials and Devices
n QlW
51
was used. Young's modulus for Pb(Zro,5Tio,5)03 was estimated 9.4~10"N/m2 from the elastic compliance s,, of PT@)and PZ(7)by the equation E=l/s,,. Poisson's ratio was assumed to be 0.2. Calculated stress change corresponding A28=0.14 degrees is 0.37Gpa. Introducing [PT+PZT( 5Onm)l buffer layer into PZT/Nb-ST interface produces 0.37GPa of the relaxation of the tensile stress. XRD Reciprocal-Space Mapping of the PZT Film in PZTBuffer /Nb-ST Figure 6 shows the reciprocal-space mapping of (1) PZT/Nb-ST film and (2) PZT/[PZT+PT(5Onm)]/-ST film around Nb-ST( 103). The (103) peaks of the PT, PZT and Nb-ST substrate are marked in the figure. The peak of the compositional-gradient buffer layer extends continuously from PT( 103) to PZT( 103) in (2). The distances along X-axis and Y-axis between Nb-ST( 103) and PZT( 103) indicate the degree of stress along in-plane and normal to the substrate, respectively. Numeric numbers appeared in the figure indicated the differences. Higher the difference between the film and the substrate, the film is more independent with each other. From the figure, introducing the PT-+PZT(SOnm) buffer layer decreases the interaction. Summary of the Stress Calculation Figure 7 shows the summary of the residual stress calculation for the films with various buffer layers. Comparison of the bulk PZT and the PZT in PZT/Nb-
Fig.7 Summary of the residual-stress calculation of the sample with various kind of buffer layers and the thickness.
52
Dielectric Materials and Devices
ST film, PZT film receives 0.87GPa of tensile stress from Nb-ST substrate(. in the bottom axis in Fig.7). Introducing the buffers with the composition of PT, PT -PZT weakens the tensile stress compared without buffer layer. PT buffer(5 0 m ) brings the higher relaxation of tensile stress compared with compositionalgradient buffer. On the other hand, PZ-PZT compositional-gradientbuffer layer enhances the tensile stress. DISCUSSION The epitaxial strain and the thermal strain should be considered to discuss the residual stress of the thin films.@)The epitaxial strain is due to the difference in lattice parameters of the thin film and the substrate. The thermal strain is due to the TEC difference between the thin film and the substrate. The effects of phase transformation of PZT and the buffer layers occurred during cooling should be taken account. Since there is less data of the lattice parameter and TEC as a function of temperature in PZT and the compositional-gradientbuffer, the precise and quantitative discussion might be difficult. Table I shows the TEC and lattice parameters of the PZT, PT, PZ and ST collected from the data book, etc.(9-”)
Pb(Zr,.,Ti,,)O, PbTi0, PbZrO, SrTiO,
thermal expansion coefficient / 10-6K-’ lattice parameter at R.T. / nm at 923K I at 380K (T, in MPB) a C 16.7 ( at 733K in Zr/(Zr+Ti)=0.3) 0.402 0.412 0.414 23 (//aaxis) 23 0.3899 0.415 (a=b) 0.41 1 7.7 (bulk) 11.7 10.4 0.3905
I
Taking the lattice mismatch at room temperature into account, the PZT film in the (00 1)PZT/( 1OO)Nb-ST structure should receive 2.9% of in-plane compressive stress at interface. In case of (OOl)PZT/(OOl)PT/( 100)Nb-ST structure, 3.0% of in-plane compressive stress should occur at PZTPT interface. In a same manner, the PZT thin film might not or less receive epitaxial stress in PZT/buffer/Nb-ST structure with a compositional-gradient buffer, i.e., PT+PZT or PZ-PZT. The epitaxial stress by the lattice mismatch seems not to be so large. Within this experiment, only c-axis preferred-orientation films were obtained. In general, the epitaxial stress will rapidly decrease with increasing film thickness by introducing the dislocations. There exists the critical thickness at which the epitaxial stress will disappear. In case of PT thin film deposited on SrTiO, substrate, the edge dislocations were introduced at the position of -6nm from the interface.(12) Despite of compressive stress by lattice mismatch, all PZT thin films witldwithout buffer deposited on Nb-ST substrates were subjected to in-plane tensile stress as shown in Fig.7. The order of TEC of PT, PZT, Nb-ST and PZ
Dielectric Materials and Devices
53
seems to be am > aPZT >~ t > apz ~ appeared ~ - in Table ~ I. ~ This means that the thermal stress should be dominant, not epitaxial stress. Therefore we mainly focus on the residual stress caused by thermal strain in this work. TEC of PT and Pb(Zro.3Tio.,)03(11) below the phase transformation temperature are about 2 3 . 4 ~0-6 1 and 1 6 . 7 ~ 1 0 -respectively, ~, measured by HT-XED. The TEC of Nb-ST is 11.7x10? Since the TEC data for PZT is not available, we can estimate the relative TEC-value order from known TEC value and the lattice parameter at various temperatures. Since the compositional-gradient buffer layers are the solid solution of PT and PZT, and that of PZ and PZT, the TEC value should be interpolated between PT and PZT. Therefore the TEC of the buffer layer is larger than that of PZT but much less than that of Nb-ST. Taking account of these consideration and the result of Fig.7, the order of the TEC for each materials might be as follows. > a[FT-tPZT] > aPZT > aNb-ST > a[PZ+PZT] > aPZ (3) There might be slight negligible thermal stress at each interface among PZT, the buffer layer and Nb-ST at the growth temperature. During cooling, the stress caused by the TEC difference will be generated at each interface. The PZT film on a Nb-ST substrate receives strong tensile stress from the Nb-ST substrate. However, when the PZT/buffer/Nb-ST structure is formed, the buffer layer receives strong tensile stress. The tensile stress is offset somewhat between the buffer layer and Nb-ST. Consequently, the effect of tensile stress from the Nb-ST substrate is reduced if the PT or PT+PZT buffer layer was introduced. The PZ-+ PZT buffer layer acts as enhance the tensile stress. The degree of stress offset between the buffer layer and the substrate, and PZT thin film and the buffer layer will be affected by the thickness and average TEC of the buffer layer. CONCLUSION We prepared PZT thin films with various kinds of buffer layers by the pulsed MO-source CVD method. The compositional-gradient buffer layer was confirmed by depth profile analysis by X P S . The PZT films with/without buffer layer deposited on Nb-ST substrate had in-plane tensile stress. In-plane tensile stress in PZT deposited on Nb-ST was successively controlled by introducing the buffer layers. The buffer layers with larger TEC that that of PZT, i.e., PT or PZT-PT, produced the relaxation of tensile stress. On the other hand, the buffer with smaller TEC, PZT-PZ, enhanced the tensile stress. Using these phenomena, we might be able to control the properties of the thin film. ACKNOWLEDGEMENT This work was supported by the Grant-in-Aid for Scientific Research, Basic Research(B) from the Ministry of Education, Science, Sports and Culture in
54
Dielectric Materials and Devices
Japan. REFERENCES ‘H. Uchida, A. Saiki, N. Wakiya, K. Shmozaki and N. Mizutani, “Effect of the Residual Stress Induced by External Stress Application on Dielectric Properties of Epitaxial Lead Titanate Film,” Journal of the Ceramic Society of Japan, 108 [l] 21-25 (2000) 21.Kanno, S. Hayashi, M. Kitagawa, R. Takayama and T. Hirao, “Antiferroelectric PbZrO, thin films prepared by multi-ion-beam sputtering,” Applied Physics Letter, 66 [2] 145-147 (1995) M. de Keijser, J.F .M. Cillessen, R.B.F. Janssen, A.E.M. de Veirman and D.M. de Leeuw, “Structural and Electrical Characterization of Heteroepitaxial Lead Zirconate Titanate Thin Films,” Journal of the Applied Physics, 79 [ l ] 393-402 ( 1996) 4A. Endo, A. Iwasaki, N. Wakiya, A. Saiki, K. Shinozaki and N. Mizutani, “Preparation and Properties of PbTi0,-PbZrO, Thin Films by Pulsed MO-Source CVD Method,” Key Engineering Materials, 181-182,77-80 (2000) H. Uchida, T. Kiguchi, A. Saiki, N. Wakiya, N. Ishizawa, K. Shinozaki and N. Mizutani, “Measurement Technique for the Evaluation of Residual Stress in Epitaxial Thin Film by Asymmetric X-Ray Diffraction,” Journal of the Ceramic Society of Japan, 107 [7] 606-10 (1999) S. Ikegami, I. Ueda, T. Nagata, Acoustic Society of America, 50,1060 (1971) M. Marutake, T. Ikeda, Journal of Physical Society of Japan, 10,424 (1955) *C.M.Foster, Z.Li, M.Buckett, D. Miller, P. M. Baldo, L.E. Rehn, G.-R. Bai, D. Guo, H. You, K.L. Merkle, “Substrate Effects on the Structure of Epitaxial PbTiO, Thin Films Prepared on MgO, LaAlO,, and SrTiO, by Metalorganic Chemical-Vapor Deposition,” Journal of the Applied Physics, 78 [4] 2607-2622 ( 1995) “Landort-Bornstein,Numerical Data and Functional Relationship in Science and Technology,” New Series, Group 111, Vol. 13,28 (1981) 10 “Thermal expansion: Nonmetallic Solids, Therrnophysical Properties of Matter”, The TPRC Data Series, Vol. 13, Edited by Y.S. Touloukian, R.E. Taylar and T.Y.R. Lee, Plenum Press, 1977 l1 K. Kakegawa, J. Mohri, K. Takahashi, H. Yamamura and S. Shirasaki, Journal of the Chemical Society of Japan, [5] 7 17-21 (1976) l2 S. Stemmer, S.K. Streiffer, F. Ernst and M. Ruhle, “Dislocations in PbTiO, Thin Films,” Physics of the Solid State (a), 147, 135 (1995)
Dielectric Materials and Devices
55
This page intentionally left blank
FACTORS INFLUENCING TEXTURE DEVELOPMENT IN HOT FORGED BISMUTH TITANATE J. S. Patwardhan and M. N. Rahaman University of Missouri-Rolla, Department of Ceramic Engineering, Rolla, Missouri 65409 ABSTRACT The influence of processing and compositional parameters on the development of an oriented grain microstructure (texture) in bismuth titanate, Bi4Ti3012, by hot forging was investigated. Bi4Ti3O 12 powders, prepared by conventional solid state reaction, were consolidated and sintered to produce compacts with different initial densities for hot forging at temperatures in the range of 800 "C to 1100 "C. Characterization of the oriented microstructure was performed by X-ray diffraction and by scanning electron microscopy. Dense, highly oriented microstructures were achieved by hot forging samples with an initial density of 80 to 85% of the theoretical density at -1050 ' C . The linear strain rate during hot forging increased significantly above the temperature for liquid formation in the Bi4Ti3012/Bi203 system (-870 "C). Compositional variation (excess TiOz or Bi2O3) produced marked changes in the densification of Bi4Ti3012 but did not seriously influence the ability to develop a highly oriented microstructure by hot forging. INTRODUCTION Commonly, ceramics are composed of randomly oriented, equiaxial grains that give rise to isotropic engineering properties normally required for most technological applications. However, the development of a textured microstructure consisting of aligned, elongated grains can provide unique anisotropic properties for enhancing the usefulness of certain ceramics such as piezoelectric ceramics for sensing and actuating devices. The basic requirements for texture development during the fabrication of granular materials are anisotropic grain growth and grain orientation. Anisotropic grain growth is frequently observed in materials with the hexagonal crystal structure such as A1203 [1,2], chainlike structures such as mullite [3,4] and layered structures such as Bi4Ti3012 [5]. The mechanism of growth is not clear but several factors have been suggested as being important for the To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
57
process. These include anisotropic surface energies, twinning, segregation of impurities and dopants, and anisotropic wetting by liquid phases [2,6]. Several processing techniques have been employed to develop grain orientation in ferroelectric ceramics, including hot forging [7,8], hot pressing [9,10], tape casting [11,12], and templated grain growth [ 13,141. Bismuth titanate (Bi4Ti3012) is a candidate material for high temperature piezoelectric applications, memory storage and optical displcys because of its high Curie temperature (-675 "C) and electrooptical properties [15-181. Textured Bi4Ti30 12 have been achieved in several studies [8- 141. However, the mechanism of grain orientation during hot forging or hot pressing is not clear. Furthermore, the effect of deviations from the stoichiometric Bi4Ti3012 composition on texture development has received little attention. According to published phase diagram data of Speranskaya et a1 [19], three incongruently-melting compounds exist in the Bi203-Ti02 system: Bi4Ti3012 (peritectic melting temperature 1210 "C), the bismuth-rich phase BigTi014 (peritectic melting temperature =865 "C) and the titanium-rich Bi2Ti401 1 (peritectic melting temperature ~ 1 2 7 5"C). The data of Bruton [20] are similar to those given by Speranskaya et al, except that the bismuth-rich compound is Bi12Ti020, melting incongruently at -873 "C. The bismuth-rich compound has also been identified as Bi12Ti020 by Levin and Roth [21] who suggested that it is congruently melting, and by Morrison [22] who concluded that congruent melting is favored but that the departure from congruent melting is small. Morrison also found a higher peritectic melting temperature (-920 "C) that those obtained by Speranskaya et a1 [19] and Bruton [20]. It is clear from the phase diagram data that for the fabrication temperatures commonly used for Bi4Ti3012 (1000 "C to 1150 "C), small deviations from the stoichiometric Bi4Ti3012 composition will lead to solid state sintering (for Ti02-rich compositions) or sintering in the presence of a liquid phase (for Bi203-rich compositions). The influence of such compositional variations on the texture development of Bi4Ti3012 is not clear. The objective of the present study was to investigate how key processing factors (initial density, applied pressure and the temperature) and compositional variation (small excess of Bi2O3 or Ti02) influence texture development in Bi4Ti3012 during hot forging. EXPERIMENTAL A conventional process involving the calcination of mixed oxides was use to prepare the Bi4Ti3012. Starting powders, Bi2O3 powder (Ferro Corp., Pen Yan, NY; average particle size = 4 pm; purity 99.9%) and Ti02 powder (Ferro Corp., Pen Yan, NY; average particle size 3 pm; purity = 99.9%,) were weighed out in the molar ratio of 2:3 and mixed by ball milling for 24 h in a polyethylene container while dispersed in isopropanol, using high purity zirconia balls as the mill-
58
Dielectric Materials and Devices
ing media. The mixture was stirred in a beaker until it was nearly dry, heated at 100 "C in an oven to evaporate the remaining liquid and ground lightly in an agate mortar and pestle. The powder was calcined in a high purity A1203 crucible for 1 h at 600 "C followed by 4 h at 850 "C to produce Bi4Ti3012. After calcination, the powder was ball-milled under the conditions described earlier to break down agglomerates, dried, lightly ground in an agate mortar and pestle and sieved using a 100-mesh nylon screen. These powders prepared from a Bi2O3: Ti02 molar ratio of 2 3 will be referred to as Bi4Ti3012. Powders with 1 wt% excess Ti02 (referred to as Ti02-rich Bi4Ti3012) or 1 wt% excess Bi203 (Bi203-rich Bi4Ti3012) were also prepared using the same procedure. The powder was pressed in a uniaxial die at -35 MPa and then isostatically pressed at -275 MPa to produce cylindrical compacts with a green density of 6065% of the theoretical density of Bi4Ti3012 (assumed to be 8.04 g/cm3). The powder compacts were sintered in a dilatometer (1600 C; Theta Industries Inc., Port Washington, NY) in air for 1 h at 1100 "C using a heating rate of 5 "C/min. The densities of the sintered samples were determined from the green density and the shrinkage. Smooth curves were drawn through the data using a curve fitting technique. Hot forging of cylindrical powder compacts (12.5 mm in diameter by 9 mm in height) was performed in air under a load of 5 MPa at a temperatures of 950 "C and 1050 "C (heating rate = 10 "C/min) in an Instron testing machine fitted with a programmable furnace. Prior to hot forging, the compacts were fired to fixed temperatures to produce samples with a range of densities in order to investigate the effect of starting density on grain orientation. The densities of the hot forged samples were measured by the Archimedes method. X-ray diffraction (XDS 2000; Scintag Inc., Sunnyvale, CA) was used to identify the crystalline phases in the samples. For the hot forged materials, the degree of texturing was determined from the X-ray diffraction (XRD) patterns by the Lotgering method [23]. The samples were scanned at 0.03 degreedmin from 10" to 60" 20 on the surfaces perpendicular and parallel to the applied pressure. The degree of orientation, referred to as the Lotgering factor, f, was determined from the relation: f = (P - PO)/ (1 - PO)where P = C I(001) / C I (hkl), where C I(OO1) and C I (hk1) are the sums of the intensities of (001) and (hk1) reflections, respectively, between 10" and 60" 20, and POis the value for a randomized powder sample, taken in this study as the calcined powder. Scanning electron microscopy (JEOL T33OA) was used to observe the microstructure of the fabricated materials. Samples for scanning electron microscopy (SEM) were prepared by polishing down to 0.1 pm diamond finish, followed by thermal etching for 1 h at 950 "C.
Dielectric Materials and Devices
59
RESULTS AND DISCUSSION Figure 1 shows the data for the relative density as function of temperature during constant heating rate sintering at 5 "C/min to 1100 "C for powder compacts of Bi4Ti3012, Bi203-rich Bi4Ti3012 (1 wt% excess Bi2O3) and TiO2-rich Bi4Ti3012 (1 wt% Ti02). The curves have approximately the same shape and the final relative density is -97%. However, compositional variation has a significant effect on the temperature range of sintering. When compared to Bi4Ti3012 that starts to show measurable densification at -875 "C, excess Bi2O3 lowers the sintering temperature while excess TiO2 raises the sintering temperature. Bi203-rich Bi4Ti3012 starts to show measurable shrinkage at temperatures as low as -800 "C. As outlined earlier, the phase diagram data indicate the formation of a liquid phase in Bi~O3-richBi4Ti3012 compositions at 865-870 "C. While the reduction of the sintering temperature for Bi203-rich Bi4Ti3012 is consistent with the formation of a liquid phase, the magnitude of the reduction indicates enhancement of densification at temperatures lower than the peritectic melting temperature for the bismuth-rich Bi12Ti020 phase. Figure 2 shows that the microstructure of Bi4Ti3012 sintered to 1100 "C has a random arrangement of highly elongated grains. The Lotgering factor, f, was in the range of 0.2 to 0.3. Commonly, the growth of highly elongated grains during sintering restricts densification so that high final densities are difficult to achieve. Matter transport into the interstices between highly elongated grains is normally difficult by solid state diffusion. The mechanism of densification coupled with highly elongated grain growth is not clear.
E: Iwt% excess TiOz C: 1 wt% excess Bi203
65
t
0
200
i
i
400 600 800 Temperature O C
1000
1200
Figure 1. Relative density versus temperature for a nominal Bi4Ti3012 composition and for Bi4Ti3012 containing 1 wt% excess Bi203 or TiO2.
60
Dielectric Materials and Devices
Figure 2. SEM micrograph of Bi4Ti3012 sintered at 1100 "C for 1 h. Hot forging experiments on Bi4Ti3012 powder compacts indicated that both the fabrication temperature and the initial density of the body are key variables in optimizing the grain orientation. Lower fabrication temperatures produced inadequate densification and reduced texturing. Figure 3 shows data for the axial strain rate of Bi4Ti3012 powder compacts (initial relative density ~ 0 . 8 5 as ) a function of time at four hot forging temperatures. A rapid increase in the strain rate at early times is observed above 870 "C. As outlined earlier, the presence of a liquid phase is expected for Bi203-rich Bi4Ti3012 above -865 "C. While the present composition is nominally Bi4Ti3012, minor compositional deviations cannot be ruled out. The strain rate data are consistent with the formation of a liquid phase above -865 "C caused by the presence of a small excess of Bi203. Low initial density (60-70% of the theoretical) of the hot forged materials required long times to achieve adequate densification and grain orientation while high initial densities (>90%) resulted in some degree of cracking during sinter forging. A temperature of 1050 "C and intermediate initial densities produced dense microstructures with highly oriented grains. For Bi4Ti3012 with an initial density of 0.80 prepared by sintering (Lotgering factor =0.25), Figure 4 shows the data for the density and Lotgering factor as a function of hot forging time at 1050 "C. The increase in the pressure from zero to the required value (5 MPa) was completed as the hot forging temperature was reached. The data show a rapid increase in the Lotgering factor and in the density as the pressure was being applied and during the early stages of hot forging. These data, coupled with the data of Fig. 3, indicate that a rearrangement process, presumably aided by a liquid phase, may be playing a critical role in the early stages of hot forging. As the rearrangement process slows, diffusional mass transport is expected to become important.
-
Dielectric Materials and Devices
61
87OoC
8OO0C
0
-
c
.-c
I
-
E. -0.05
3 CI
-0.1
-
-0.15
l
-0.2
,
,
,
l
,
,
,
l
,
,
,
l
,
,
,
~
,
,
,
-
Figure 3. Axial strain rate versus time for Bi4Ti3012 powder compacts hot forged at different temperatures.
100
1
v
- 0.8 r --)-
Density (% of theoretical)
+Lotgering factor, f
0,
-
- 0.6
M
2.
3 M
3
0.4 9
"g
0.2 80
0
10
20
30
Ti$g cmin!jO
-h
0
60
70
80
Figure 4. Effect of hot forging time on the density and texture development of Bi4Ti3012 powder compacts. Figure 5 shows XRD measured for the Bi4Ti3012 powder compacts sintered for 1 h at 1100 "C and in surfaces perpendicular and parallel to the hot forging direction for the textured Bi4Ti3012 hot forged for 75 min at 1050 "C. Diffraction from the (001) planes in the surface parallel to the hot forging direction indicates that the grains are oriented along the c-axis.
62
Dielectric Materials and Devices
n c v1 ,
Sintend at U00 C for 1h
E
1
0
U, i2
.r(
v1
E
Hot forged at 1050 C for 75 min
c,
E
H
10
28 (degrees)
Figure 5. XRD patterns of Bi4Ti3012 sintered at 1100 "C for 1 h (random) and hot forged for 75 min at 1050 "C showing texturing in surfaces perpendicular and parallel to the hot pressing direction. The influence of compositional variation on the texturing and anisotropic grain growth of hot forged Bi4Ti3012 is shown in Fig. 6. For Bi203-rich Bi4Ti3012 (1 wt% excess Bi2O3), Fig. 6(a) shows a SEM micrograph of the textured material hot forged for 75 min at 950 "C. The microstructure shows significant grain orientation and the presence of nearly equiaxial second phase grains identified to be Bi12Ti020 by XRD. It is believed that the second phase grains developed by the crystallization of a liquid phase when the material was cooled to room temperature. Corresponding micrographs for the nominal Bi4Ti3012 and for TiO2-rich Bi4Ti3012 (1 wt% excess Ti02) hot forged for 75 minutes at 1050 "C are shown in Figs. 6(b) and 6(c). Both materials show considerable grain orientation. In addition, the TiO2-rich Bi4Ti3012 contains a dispersed second phase of fine particles identified to be Bi2Ti4011 by XRD. For Ti02-rich Bi4Ti3012, the presence of the Bi2Ti4011 phase is in agreement with phase diagram data and the formation of a liquid phase is not expected until the temperature reaches 1210 "C [ 191. According to Figs. 6(a) to 6(c), considerable grain alignment is achieved regardless of the presence of a solid or liquid second phase. The data indicate that a liquid phase may facilitate the achievement of significant grain orientation but it is not a requirement. Table 1 shows that the compositional variation does not have a significant influence on the grain characteristics of the hot forged materials.
-
Dielectric Materials and Devices
63
Figure 6. SEM micrographs showing the effect of compositional variation on the hot forged microstructure (a)-(c) and after annealing for 2 h at 1100 "C (d)-(f). The compositions consist of Bi203-rich Bi4Ti3012 (a), (d), nominal Bi4Ti3012 (b), (e), and TiO2-rich Bi4Ti3012 (c), (f).
64
Dielectric Materials and Devices
Table 1. Characteristics of the elongated grain microstructure produced by hot forging. Composition
Lotgering factor
Average Grain Width (Pm) 3
Average Aspect Ratio
0.90
Average Grain Length (Pm) 15
Bi4Ti3012 + 1 wt% Bi2O3 Bi4Ti3012
0.97
16
3
5.5
Bi4Ti3012 + 1
0.95
18
3
6
wt% Ti02
5
The mechanisms of deformation during hot forging or hot pressing include diffusion (Nabarro-Herring creep and Coble creep), plastic flow (by dislocation motion) and grain boundary sliding [24]. For ceramics subjected to relatively moderate stresses, plastic deformation is expected to be insignificant. The mechanisms responsible for deformation and grain orientation are therefore diffusion and grain boundary sliding. These two mechanisms operate sequentially, i.e., the mechanisms operate interdependently and the overall deformation rate is equal to that of the slower process [25]. Further work is required to determine the rate controlling mechanism. Figures 6(d) to 6(f) shows SEM micrographs of the hot forged materials after subsequent annealing in air for 2 h at 1100 "C. For the Bi203-rich Bi4Ti3012, [Fig. 6(d)], the most significant change is the development of large angular pores, presumably caused by the evaporation of the volatile liquid phase. XRD analysis indicated that the annealed material consisted of single phase Bi4Ti3012. The annealed Ti02-rich material [Fig. 6(f)] also retained the grain orientation of the hot forged sample but the presence of very fine pores, presumably caused by the disappearance of the Bi2Ti4011 particles, is observed. The most drastic microstructural change during annealing is found for the nominal Bi4Ti3012 composition [Fig. 6(e)] where the grain orientation has completely disappeared. The development of nearly rounded pores and highly irregular grain boundaries are also observed. CONCLUSIONS Processing conditions consisting of fabrication temperature, pressure and sample density were determined to achieve nearly fully dense Bi4Ti3012 with significant
Dielectric Materials and Devices
65
grain orientation (Lotgering factor m0.97) by the hot forging of powder compacts. Compositional variation from Ti02-rich to Bi203-rich Bi4Ti3012 coupled with phase diagram data indicate that a liquid phase can facilitate the achievement of grain orientation but it is not a requirement. Significant grain orientation can be achieved with or without the presence or a liquid phase during hot forging. Microstructural evolution of grain oriented Bi4Ti3012, studied by annealing at 1100 “C, is highly dependent on composition. The nominal Bi4Ti3012 showed abnormal grain growth and a complete disruption of the grain alignment. Ti02-rich and Bi203-rich compositions retained the grain orientation but developed porosity caused by the evaporation of volatile second phases. REFERENCES J. Rode1 and A. M. Glaeser, “Anisotropy of Grain Growth in Al2O3,” J. Am. Ceram. Soc., 73 [ l l ] 3292-301 (1990). 2 M. Seabaugh, D. Horn, I. Kerscht, S.-H. Hong, and G. L. Messing, “Anisotropic Grain growth in Alumina Ceramics,” pp. 34 1-48 in Sinterina Technology. Edited by R. M. German, G. L. Messing, and R. G. Cornwall. Marcel Dekker, New York, 1996. 3 S.-H. Hong and G. L. Messing, “Anisotropic Grain Growth in Diphasic-GelDerived Titania-Doped Mullite,” J. Am. Ceram. Soc., 81 [5] 1269-77 (1998). 4T. S. Huang, M. N.Rahaman, T.-I. Mah, and T. A. Parthasarathay, “Anisotopic Grain Growth and Microstructural Evolution of Dense Mullite Above 1550°C” J. Am. Ceram. Soc., 83 [ l ] 204-10 (2000). 5H. S. Shulman, M. Testorf, D. Damjanovic, and N.Setter, “Microstructure, Electrical Conductivity, and Piezoelectric Properties of Bismuth Titanate,” J Am. Ceram. Soc., 79 [121 3 124-28 (1996). %. Kunaver and D. Kolar, “Computer Simulation of Anisotropic grain Growth in Ceramics,” Acta Metall. Mater., 41 [8] 2255-63 (1993). 7J. U. Knickerbocker and D. A. Payne, “Orientation of Ceramic Microstructures by Hot-Forming Methods”, Ferroelectrics 37 [1-41 733-736 (198 1). *T. Takenaka and K. Sakata, Grain Orientation and Electrical Properties of Hot-Forged Bi4Ti3012 Ceramics”, Jpn. J. Appl. Phys., 19 [13 3 1-39 (1980). 9T. Kimura, T. Yoshimoto, N.Iida, Y. Fujita, and T. Yamaguchi, “Mechanism of Grain Orientation During Hot-Pressing of Bismuth Titanate”, J. Am. Ceram. SOC.,72 [l] 85-89 (1989). loY. Inoue, T. Kimura, T. Yamaguchi, K. Nagata, and K. Okazaki, “Grain Orientation and Electrical Properties of Hot-Pressed Bismuth Titanate Ceramics”, Jpn. J. Appl. Phys., 20 [I] 95-99 (1983). “S. Swartz, W. A. Schulze, and J. V. Biggers, “Fabrication and Electrical Properties of Grain Oriented Bi4Ti3012 Ceramics,” Ferroelectrics, 38 765-68 (1981). I’
66
Dielectric Materials and Devices
12H. Watanabe, T. Kimura, and T. Yamaguchi, “Particle Orientation During Tape Casting in the Fabrication of Grain-Oriented Bismuth Titanate”, J Am. Ceram. Soc., 72 [2] 289-293 (1989). 13 J. A. Horn, S. C. Zhang, U. Selvaraj, G. L. Messing, and S. TroilerMcKinstry, Templated Grain Growth of Textured Bismuth Titanate”, J. Am. Ceram. Soc., 82 [4] 92 1-926 (1999). 14 S.-H. Hong, S. Trolier-McKinstry, and G. L. Messing, “Dielectric and Electromechanical Properties of textures Niobium-Doped Bismuth Titanate Ceramics,” J. Am. Ceram. Soc., 83 [ l ] 113-18 (2000). ”A. Fouskova and L. E. Cross, “Dielectric Properties of Bismuth Titanate”, J A&. Phy~.,41 [7] 2834-2838 (1970). “S. E. Cummins and L. E. Cross, “Crystal Symmetry, Optical Properties and Ferroelectric Polarization of Bi4Ti3012 Single Crystals,” Appl. Phys. Lett., 10 [I] 14-16 (1967). I7S. E. Cummins and L. E. Cross, “Electrical and Optical Properties of Bi4Ti3012 Single Crystals,” J. Appl. Phys., 39 [5] 2268-74 (1968). ‘*Y. Masuda, H. Masumoto, A. Baba, T. Goto, and T. Hirai, “Crystal Growth, Dielectric and Polarization Reversal Properties of Bi4Ti30 12 Single Crystals,” Jpn. J Ap 1. Phys., 31 3108-12 (1992). I. Speranskaya, I. S. Rez, L. V. Kozlova, V. M. Skorilov, and V. I. Slavov, Izv. Akad. Nauk. SSSR, Neorg. Mater., 1 232 (1965). 20T. M. Bruton, “Study of the Liquidus in the System Bi203-Ti02,” J. Solid State Chem., 9 173-75 (1974). 21E.M.Levin and R. S. Roth, J. Res. Nat. Bur. Stds., A 68 197- (1964). 22A.D. Morrison, “Some properties of Bi12Ti020 and the System Bi203-Ti02,” Ferroelectrics, 2 59-62 (197 1). 23F.K. Lotgering, “Topotactical Reactions with Ferrimagnetic Oxides Having Hexagonal Crystal Structures-I,” J. Inorg. Nucl. Chern., 9 [2] 113-123 (1959). 24 A. G. Evans and T. G. Langdon, “Structural Ceramics,” Prog. Mater. Sci., 21 [3-41 171-444 (1976). 25R.Raj and M. F. Ashby, “Grain Boundary Sliding and Diffusional Creep,” Metall. Tram., 2 [4] 1113-27 (1971). I’
‘BE.
Dielectric Materials and Devices
67
This page intentionally left blank
EFFECTS OF PbO LOSS ON MICROSTRUCTURAL DEVELOPMENT AND PROPERTIES OF PLZT CERAMICS Jian-Huei Feng and Fatih Dogan University of Washington, Department of Materials Science and Engineering, Roberts Hall, Box 352120 Seattle, WA 98195, U.S.A.
ABSTRACT Commercial lead lanthanum zirconate titanate (PLZT) ceramics were annealed in air at the temperatures between 970 to 1080 "C resulting in various amount of lead oxide loss. The fracture surfaces of the ceramics gradually change from intergranular mode to transgranular mode with increasing PbO loss. Both dielectric and piezoelectric properties degrade with development of new phases due to PbO loss. The related mechanisms for these results were discussed. INTRODUCTION The volatility of the PbO component at high temperatures alters the stoichiometry and properties of sintered PZT ceramics. It has been shown that PbO loss affects both electromechanical coupling factors and dielectric constants of PZT [ 13. A common way of controlling PbO loss when sintering these ceramics is to place samples in a closed crucible surrounded with lead containing atmosphere powders [2-41. This method can maintain a high PbO partial pressure in the crucible and reduce PbO loss from the samples. Other efforts have been put on decreasing sintering temperatures by adding liquid phase agents [5-71. The presence of a liquid phase also promotes densification kinetics during sintering, called liquid phase sintering [8, 91. Excess PbO can act as a liquid phase agent and has been reported to increase grain size and density of piezoceramics [8, 10, 111. Grain size has significant effects on both dielectric and piezoelectric properties of piezoelectric To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, re roduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Jciety or fee pailto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
69
ceramics. For PZT with small grains (1-5 pm), the properties decrease when the grain size decreases, which is explained by the limited motion of domain walls [121 and the polarization locked-in by the space-charge layers of small grains ~31. The objective of this work is to provide a complementary study about the effects of PbO loss on microstructures, phase change, and electrical properties of PLZT ceramics. EXPERIMENTAL PROCEDURE Commercial PLZT sheets (53.7 x 13.5 x 0.5 mm, HD3203, Motorola, Albuquerque, NM) were used for this study. The samples, with consistent original PbO content and electrical properties, were annealed in air between 975 to 1080 “C for 1 hour resulting in weight loss in a range of 0.31% to 5.3 1 %. Microstructures of annealed samples were examined by a scanning electron microscope (SEM) (840A, JEOL, Sweden). The phases of the original and annealed samples were analyzed by a X-ray diffraction (XRD) analysis system (PW 1830, Philips, Netherland). Electrical property measurements were conducted using a HP-4 192A impedance analyzer. In the poling process, the samples were immersed in hot vegetable oil (130 “C) and poled by applying an electric field of 2 kV/mm for 2 minutes. The properties of piezoelectric ceramics may change with time, called “aging” [ 14, 151. In order to achieve stabilized properties, the aging was accelerated by alternately immersing the samples into hot and cold oil for 5 cycles. The duration time for each immersion was 15 seconds. The thicknessmode electromechanical coupling coefficient kt was determined by the resonance method [161. The relative dielectric coefficients were calculated from the measured capacitance at 1 kHz. RESULTS AND DISCUSSION Figs. 1 (a) to (d) show the microstructures of an original PLZT sheet and the annealed samples. The fracture surfaces change gradually from intergranular mode to transgranular mode with increasing PbO loss. According to Kim et al. [ 171, there are several microstructural factors which can affect fracture modes of polycrystalline ceramics, such as boundary segregation of impurities, presence of a second phase, grain boundary morphology and grain size. It was shown that the fracture mode of PZT ceramics is almost intergranular at small grain size and changes to predominantly transgranular at large grain size. Annealing of the samples does not give rise to grain growth at the relatively low temperature and short dwell times. Figs. 1 (a) to (d) reveals that the
70
Dielectric Materials and Devices
Figure 1: Fracture surfaces of annealed PLZT sheets: (a) original sample, (b) 1.54 % PbO loss, (c) 3.57 % PbO loss, (d) 5.31 % PbO loss.
Dielectric Materials and Devices
71
Figure 1: Fracture surfaces of annealed PLZT sheets: (a) original sample, (b) 1.54 % PbO loss, (c) 3.57 % PbO loss, (d) 5.31 % PbO loss.
72
Dielectric Materials and Devices
grain size of the samples remains nearly the same after heat treatment. Therefore, the transition of fracture modes from intergranular to transgranular can be attributed to the PbO loss in the samples. Excess PbO at the grain boundaries may lead to weaker bonding between the grains, so that cracks propagate along the grain boundaries. With the loss of excess PbO at the grain boundaries, bonding strength between the grains is increased leading to a transgranular fracture. This can be advantageous during dicing of piezoceramics which require smooth surface for transducer applications. Fig. 2 shows the X R D patterns of the annealed samples. Development of new phases is observed after a weight loss more than 1.54%. The diffraction peaks belonging to the new phases match most closely to the standard patterns of ZrOz and PbTi03. However, the latter phase may be Pb(Zr,Ti)03 with a lower Zr/Ti ratio considering the phase diagram of PZT [181. 7000
6ooa
8 E
A h, 5.31 % wt. loss
cd
8
i
500C
I
9 400C
h
3
2 % .-
v)
- 300C 200c
n
L
L
3.57 % wt. loss
1.54 % wt. loss
L
R,
L-H
1ooc
t
L
Original Sample
L
30
40
50
2-Theta(deg)
60
Figure 2: XRD patterns of annealed PLZT sheets.
Dielectric Materials and Devices
73
Figs. 3 and 4 show the effects of PbO loss on dielectric constant and kt values of the PLZT sheets, respectively. It is not surprising that the electrical properties were negatively affected by the formation of new phases. However, a slight deficiency of PbO in piezoceramics may not be necessarily detrimental to piezoelectric properties [11. It has been shown [19] that the coupling coefficient and permittivity of PZT decrease when the composition shifts away from the boundary of tetragonal and rhombohedral phases. The PLZT sheets used in this study have a Zr/Ti ratio close to 56/44 and a PbLa ratio close to 94/6. This composition is within the tetragonal region based on the PLZT phase diagram from Haertling and Land [20]. According to Kingon and Clark [4], when tetragonal PZT contains a small amount of excess PbO, the equilibrium composition of the PZT phase shifts closer to the tetragona-rhombohedral phase boundary. Therefore, with the loss of excess PbO in the annealed PLZT sheets, one may expect that the equilibrium Zr/Ti ratio in the PLZT phase might move away from the tetragonal-rhombohedral boundary. As a result, both dielectric and piezoelecric coefficients can decrease with a small amount of PbO loss before forming new phases in the ceramics. The results suggest that the fracture mode of PLZT ceramics with superior properties can be tailored by adjusting PbO stoichiometry. This can be implemented to control the surface roughness of diced piezo-ceramics. 4500 4000
3500 c.
c 9 3000 tn 5 2500 0 0 'E 2000 c.
%
0
6
1500 1000
500 0
0
1
2
3
Wt. Loss (%)
4
5
6
Figure 3: Effect of PbO loss on dielectric constant (@lkHz) of PLZT sheets.
74
Dielectric Materials and Devices
0.55
,
0.5
0.45
*
c
0.4
0
1
2
3
4
5
6
Wt. Loss (%)
Figure 4: Effect of PbO loss on kt of PLZT sheets. CONCLUSIONS Loss of PbO caused a change of fracture mode in PLZT ceramics, where the fracture surface gradually transformed from intergranular to transgranular. Electrical properties of the annealed PLZT sheets also degraded with increasing PbO loss, which mainly resulted from the formation of new phases. The new phases were identified to be mostly close to Zr02 and Pb(Ti,Zr)03 with low Zr/Ti ratio. ACKNOWLEDGMENT The authors would like to acknowledge the support from Washington Technology Center, Praxair Specialty Ceramics, Inc., and ATL Ultrasound, Inc. REFERENCES [l]
H. Webster, T. B. Weston and N. F. H. Bright, J. Am. Ceram. Soc., 50 [9], 490 (1967).
Dielectric Materials and Devices
75
[71
76
G. S. Snow, J Am. Ceram. Soc., 56 [2]91 (1973). G. S. Snow, J Am. Ceram. Soc., 56 [9], 479 (1973). I. Kingon and J. B. Clark, J. Am. Ceram. Soc., 66 [4], 253 (1983). D. E. Wittmer and R. C. Buchanan, J. Am. Ceram. Soc., 64 [8], 485 (198 1). S. Y. Cheng, S. L. Fu, C. C. Wei and G. M. Ke, J. Mater. Sci., 21,571 (1986). G. Zhilun, L. Longtu, G. Suhua and Z. Xiaowen, J. Am. Ceram. Soc., 72 [3], 486 (1989). D. James and P. F. Messer, Trans. J. Br. Ceram. Soc., 77 [5], 152 (1978). R. M. German, Liquid Phase Sintering. Plenum Press, New York, NY, 1985. S.-S. Chiang, M. Nishioka, R. M. Fulrath and J. A. Pask, Am. Ceram. Soc. Bull., 60 [4], 484 (198 1). W. K. Lin and Y. H. Chang, Mater. Sci. & Eng A , 186, 177 (1994). H. T. Martirena and J. C. Burfoot, J. Phys. C: Solid State Phys., 7 , 3 182 (1974). K. Okazaki and K. Nagata, J. Am. Ceram. Soc., 56,82 (1973). A. J. Moulson and J. M. Herbert, Electroceramics: Materials, Properties, Applications. Chapman & Hall, New York, NY, 1990. J. Mendiola, C. Alemany, B. Jimenez and E. Maurer, “Poling Strategy of PLZT Ceramics,” Ferroelectrics, 54, 195 (1984). “IRE Standards on Piezoelectric Crystals: Measurements of Piezoelectric Ceramics,” Proc. IRE, 49, 1161 (1961). S.-B. Kim, D.-Y. Kim, J.-J. Kim, and S.-H. Cho, J. Am. Ceram. Soc., 73 [l], 161 (1985). A. H. Webster, R. C. MacDonald and W. S. Bowman, J. Can. Ceram. Soc., 34,99 (1965). B. Jaffe, W. R. Cook, Jr. and H. Jaffe, p. 142 in Piezoelectric Ceramics, Academic Press, New York, NY, 1971. G. H. Haertling and C. E. Land, J. Am. Ceram. Soc., 54 [I], 1 (1971).
Dielectric Materials and Devices
A STUDY ON THE EFFECTS OF LANTHANUM DOPING ON THE MICROSTRUCTURE AND DIELECTRIC PROPERTIES OF 0.9 Pb(Mg1/3Nb2/3)03-0.1 PbTi03 Michael R. Winter 1688 Paseo Laguna Sec0 Apt. 106 Livermore; CA 94550 (until 8/00>
22 Crystal Drive Dryden, NY 13053
S. M. Pilgrim 2 Pine Street NYSCC at Alfred University Alfred, NY 14802 M. Lejeune Ecole Nationale Supkrieure de Ckramique Industrielle 123 Avenue Albert Thomas 87060 Limoges cedex FRANCE ABSTRACT Lead magnesium niobate, Pb(Mg1/3Nb2/3)03 or PMN, is a relaxor ferroelectric with many potential applications. As a relaxor ferroelectric, PMN displays a diffuse phase transition and the temperature of maximum relative permittivity, Tm, shifts to higher temperatures as frequency increases. When 10 mole % lead titanate, PT, is added, a solid solution is formed with a Tm near ambient temperature. These properties can be manipulated to serve many functions such as transducers, actuators, and active vibration control systems. The purpose of this experiment was to examine the effect of lanthanum doping on the microstructure and dielectric properties of a 0.9 Pb(Mg1/3Nb2/3)03- 0.1 PbTi03 solid solution. Lanthanum was added in concentrations of 0.0, 0.5, and 1.0 mole %. The samples were fired under different lead atmosphere conditions to examine the effect of lead depletion. Lanthanum was found to have no effect on the grain structure of the fired samples, but the grain size did increase when a mass of lead zirconate was added to the system to maintain a high partial pressure To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
77
of lead. The addition of lanthanum also created a weak, grain boundary phase, leading to intergranular fracture. Lanthanum doping decreased the maximum permittivity by approximately 17% and decreased Tm by 10°C for every 0.5 mole 'Yo lanthanum added. The diffuseness of the permittivity versus temperature curve was described using both y and 6. Both parameters increased as lanthanum was added to the system, indicating a broader transition. Lanthanum addition also decreased high-field polarization, but did not have a conclusive effect on induced microstrain. Aging was not observed under any of the examined conditions. INTRODUCTION As a relaxor ferroelectric, PMN displays a diffuse phase transition and the temperature of maximum relative permittivity, Tm, shifts to higher temperatures as frequency increases. Ordered ferroelectrics show a sharp transition peak due to uniform domains of a micrometer scale; however, relaxor ferroelectrics exhibit small domains of a nanometer scale. The composition of these nanodomains varies slightly from the composition Pb(Mg1/3Nb2/3)03and therefore the transition temperature of each domain varies as well. Due to this variation, the transition temperature, Tm, moves from a sharp peak to a diffuse curve. The nanodomains are also responsible for the shift of T, to higher temperatures as frequency increases. In relaxors, an increase in frequency causes some nanodomains to become inactive. The nanodomains that remain active reach maximum relative permittivity, km, at a higher temperature. This results in an increase in Tm as frequency increases. The interest in PMN is due in part to the fact that it exhibits a large dielectric constant, making this material a very attractive candidate for use in capacitors'. Additionally, PMN has a substantial electrostrictive coefficient which is a very important characteristic in applications such as actuat01-s~'~ transducers, and active vibration control4. Because the Tm for PMN is approximately -15"C, it is necessary to add a material with a similar perovskite structure and a higher Tm. Lead titanate, PbTi03 or PT, (Tm = -490°C) is commonly added to PMN to form a solid solution material with a Tm near ambient temperature. The effect of lanthanum on the PMN-PT solid solution has been studied by Gupta et al., but only near the morphotropic phase boundary, corresponding to a composition of 0.65 PMN-0.35 PT5. The effect of lanthanum addition to several PMN-based materials has been well documented and this dopant is believed to increase the ordering of the disordered structure that gives relaxor materials their characteristic diffuse phase transition6. Lanthanum substitutes for lead on the A-site of the PMN perovslute. Because lanthanum is a donor dopant, the excess charge must be compensated. If a high partial pressure of lead is not maintained during sintering, lead depletion of the sample results in the formation of an
78
Dielectric Materials and Devices
undesirable pyrochlore phase. However, if sufficient atmospheric lead exists, the ratio of magnesium to niobium will shift from 1:2 closer to 1:1, or the concentration of magnesium will increase and that of niobium will decrease, compensating for the added lanthanum. The shift in this ratio is accomplished by either precipitating Mg from the lattice as a second phase or dissolving Mg deposits into the lattice. The change in Mg:Nb ratio increases the ordering of the perovskite, changing electrical characteristics. The solid solution of PMN-PT with a T, near ambient temperature (i.e. 10 mole % PT) optimizes relative permittivity over the broadest temperature range. The effect of lanthanum is known qualitatively, however, a quantitative analysis is necessary to determine the most useful composition for practical applications. The purpose of this experiment was to observe the effect of the partial pressure of lead during firing and lanthanum doping on the microstructure and dielectric properties of 0.9 Pb(Mg1/3Nb2/3)03- 0.1 PbTi03 solid solution. Lanthanum was added as a superaddition in concentrations of 0.0,0.5, and 1.O mole percent. PROCEDURE The powder used to create test samples was prepared from a preformed base powder of 0.9 PMN- 0.1 PT with lanthanum, PVA, and PEG additions. The 0.9 PMN- 0.1 PT powder was formed using a solid state reaction process according to Lattard with an excess of 12 mole % Mg07. The additional 12 mole % MgO was added to assure the formation of a high-percentage perovskite powder. A quantity of PEG equal to 1.2 wt. % of the PMN-PT powder was dissolved in warm distilled water while stirring. Once the PEG was dissolved, 1.4 wt. % PVA was added to the solution and stirred for 2 additional minutes. The appropriate amount of lanthanum nitrate, La(N03)3-6H20 was dissolved in a separate beaker containing approximately 60mL of distilled water, then the 0.9 PMN- 0.1 PT powder was added to the solution and stirred for 5 minutes. A separate beaker was used to prevent preferential adsorption of the polymers and improve the distribution of the lanthanum. The solution of PEG/PVA was added dropwise to the lanthanum doped suspension of 0.9 PMN- 0.1 PT powder and then the entire mixture was stirred for 2 hours. A solids loading of 20 wt % was maintained in all batches to allow proper mixing of the constituents. Water was removed by heating the mixture in an oil bath. To prevent segregation of the constituents, the mixture was agitated while drying. Once the visible water had been removed, the dried powder was placed in a drying oven to remove any residual liquid. The dried powder was then crushed and passed through a 250 pm mesh screen. Disk shaped pellets were dry pressed from approximately one gram of the screened powder in a 13mm diameter steel die at 150 MPa for one minute. The geometric density of the green pellets was then calculated.
Dielectric Materials and Devices
79
Pellets were sintered using a sealed double crucible system to control lead volatilization. Excess lead zirconate was added to the crucible systems to create a lead atmosphere and prevent lead volatilization from the samples. The samples were sintered according to the schedule shown in Table I. Table I. Firing schedule used to sinter samples. Batch Fraction PbZr03 Ramp per sample (OC/min) 1 0.25 1 0.2 5
Level 200 450 1200
Dwell (hour) 0.2 2 1
ec>
2
1
1 0.3 5
200 450 1200
0.2 2 6
3
3
1 0.3
200 450 20 1200
0.2 2
N
5
N
6
Samples from batch 3 were cooled to room temperature after the 450°C soak to determine weight loss. After sintering, samples were characterized for geometric density and lead loss. X-ray diffraction was also performed to determine the phase purity of the powder. Conductive surfaces were placed on the samples used in dielectric experiments by sputtering with 20/80 Au/Pd metal. The edges of the samples were covered with nonconductive tape to prevent deposition of metal and then the faces of the samples were metalized. To obtain the relative permittivity as a function of temperature, each sample was placed in a temperature controlled chamber under vacuum. Liquid nitrogen was used to cool the samples to -50°C and measurements were taken while heating at 5°C increments to a maximum temperature of 90°C at frequencies of 0.1, 1.O, 10, and 100 kHz. Diffuseness was calculated using the method described by Uchino and Nomurag as well as the method described by Pilgrim et aL9. Scanning Electron Microscopy was used to observe both the fracture surface and polished surface of samples from each batch. Samples were polished using the standard lapping and polishing technique with a final polish at 1 pm and were thermally etched at 1100°C. Samples were observed using secondary and backscatter electron detectors and electron micrographs were analyzed for grain
80
Dielectric Materials and Devices
size and homogeneity. Grain size was determined according to the linear intercept method described by Wurst and Nelson". RESULTS AND DISCUSSION The most dominant factor in limiting the amount of pyrochlore phase formed is the added excess lead zirconate. As stated before, high levels of atmospheric lead prevent lead volatilization from the samples and therefore limit the formation of lead deficient pyrochlore. The exact amount of excess lead zirconate necessary was not investigated, but a mass greater than the mass of the pellets had no significant effect on the degree of lead volatilization and formation of pyrochlore. The addition of La did not have a significant effect on the formation of pyrochlore, as shown in Table 11.
Dielectric Materials and Devices
81
Table 11. Results of measured properties and characteristics. ~
Fraction Batch LaAdded Excess (mole %) Lead
6
Tm ("C)
Knl
Fired Total Pb Pyrochlore Average Deviation in Corrected A-site Density Loss Concentration Grain Size Grain Size Km Charge (€!/cc) (wt Yo) (YO) (Pm) (Pm)
1
0.0
0.25
1.80
51.7
40
14 846
21 779
1.870
7.30
1.20
8.2
2.02
0.08
1
0.5
0.25
1.83
56.1
30
13 122
20 110
1.894
7.22
1.53
9.3
2.13
0.15
1
1.0
0.25
1.62
63.1
20
12415
17 630
1.884
7.26
2.79
7.8
2.15
0.06
2
0.0
1
1.47
46.5
45
18 352
23 420
1.876
7.45
0.37
5.1
3.43
0.54
2
0.5
1.o
1
1.76
55.8
30
15 667
19 068
1.909
7.44
0.36
4.3
3.21
0.4 1
2
1
1.56
61.2
20
15273
18969
1.897
7.42
1.44
4.8
3.61
0.63
3
0.0
3
1.48
47.1
45
18 184
23 090
1.854
7.31
0.34
5.O
3.78
0.15
3
0.5
3
1.63
50.7
30
16441
19 871
1.893
7.39
0.36
4.1
3.89
0.23
18251
1.892
7.40
1.18
3.9
3.26
0.26
3
82
y
1.o
3
1.83
65.6
15
15 340
Dielectric Materials and Devices
The particle size of the 0.9 PMN- 0.1 PT precursor powder was determined by sedigraph particle size analysis and the average size was found to be -0.2 pm. During sintering, the grain size increased significantly and became more heterogeneous. Shaw et al. reported that the inclusion of lanthanum in a lead zirconate-lead titanate-lanthanum magnesium niobate system created lattice microstrain that limited grain growth, according to the formula xLMN-yPZ-zPT where x+y+z=l and x ranges from 3 to 15 in steps of 311. These microstrains were attributed to the smaller size of the lanthanum ion compared to the lead ion. Gupta et al. reported a decrease in grain size fiom -7 pm to -4 pm with the addition of 10 atom % La to the 0.65 PMN- 0.35 PT system5. In this work, however, the addition of lanthanum did not appear to have a significant effect on the grain size. Figure 1 shows the grain structure of polished samples from three levels of dopant concentration.
Figure 1. Interior surface micrographs of polished and thermally etched samples doped with 0.0,0.5, and 1.O mole % La from batch 2. In Figure 1, the dark regions are MgO inclusions and the dark regions ringed with white are pores. There is very little difference in the microstructure between the three compositions. Because dopant levels were relatively small, it is possible that lanthanum was not present in sufficient quantity to have the effect observed by the two previous works. In all experiments, lead zirconate was added to increase the partial pressure of lead in the crucible atmosphere. Similar to the dopant, this variable also did not appear to have a large effect on the grain size or distribution, as seen in Figure 2.
Dielectric Materials and Devices
83
Figure 2. Interior surface micrographs of polished and thermally etched samples doped with 0.5 mole % La from batch 1, batch 2, and batch 3. When comparing micrographs from the first and second batches, it appears that the addition of a mass of lead zirconate equal to the mass of the pellets increased the grain size. However, there is little noticeable difference between the second and third batches, despite the fact that the mass of lead zirconate added was tripled. Sintering time appeared to have the greatest effect on the size of the grains. While there were a few large grains in the first batch, the average size is quite small. When firing time was increased to six hours, the size of the grains increased considerably. The firing time was held constant for batch 2 and batch 3, resulting in a microstructure with very similar grain size. Homogeneity varies slightly between the three compositions as there are more exaggerated differences in grain size between the undoped and doped samples. The undoped sample has a slightly more uniform grain size, but the difference between the homogeneity of the different doping concentrations is not significant enough to conclude an influence by the dopant. Because the location of dopant is an important factor in grain size and homogeneity, the distribution of dopant in the fired samples was considered. It is possible, but unlikely, that the dopant segregated during drying or firing. X-ray Dispersive Spectroscopy, XDS, was used to examine the polished samples from each batch. However, due to the fact that dopant levels were relatively low compared to levels of titanium and that the XDS peaks for the two elements are very close, the lanthanum peak was obscured and no conclusions could be drawn. The amount of excess lead added to the
84
Dielectric Materials and Devices
system also did not appear to have a significant effect on the homogeneity of the grains. The sintering time was the important factor affecting the homogeneity of the grains, as shown in the deviation in grain size in Table 11. When samples are held at high temperature for longer periods of time, diffusion mechanisms have enough time to create a wide variation in grain size. Small grains are gradually absorbed by large grains or simply coalesce into large grains. This phenomenon was not expressly investigated in this work and a more detailed study is needed to determine the time dependent growth of the grains. The fracture surface of the pellets was also examined both near the edge of the sample and near the middle of the sample. At the surface of the samples, the probability of lead diffusing out of the grains and into the grain boundaries was expected to be higher. Close to the center of the sample, the microstructure was expected to be relatively uniform and homogenous. Gupta et al. reported that the addition of lanthanum caused a weaker, lead-rich grain boundary phase to form between the grains5. As lanthanum is added and substitutes for lead in the lattice, more lead is liberated and can volatilize via migration through the grain boundaries. The migration of lead to the grain boundaries increases the likelihood of intergranular fracture. Figure 3 shows the fracture surfaces from the inner regions of the samples from batch 2. The effect of lanthanum was apparent in the lead deficient atmosphere of the first batch and in the slightly enriched lead atmosphere of the second batch (Figure 3). The addition of lanthanum caused the samples to change from predominantly transgranular fracture to predominantly intergranular fracture. Under lead deficient conditions, lanthanum served to increase the disparity between the compositions of the grain boundaries and the grains and increased the amount of weak grain boundary phase, causing intergranular fracture.
Dielectric Materials and Devices
85
Figure 3. Interior fracture surface micrographs of samples doped with 0.0, 0.5, and 1.O mole % La from batch 2. The effect of lanthanum, however, became less significant as the partial pressure of lead in the firing crucibles was increased, as seen in Figure 4.
Figure 4. Interior fracture surface micrographs of samples doped with 1.O mole % La from batch 1, batch 2, and batch 3. A lead deficient firing atmosphere would likely result in predominantly intergranular fracture as the lead tends to exsol and precipitate at the grain boundaries to form the weak phase. When a sufficient partial pressure of lead vapor was present, the degree of lead migration leading to volatilization and
86
Dielectric Materials and Devices
intergranular fracture was decreased. This decrease in migration increases the homogeneity between the grains and the grain boundaries, resulting in primarily transgranular fracture with very little intergranular fracture. Deposits of excess MgO were found throughout the samples, as expected. The frequency of deposits increased closer to the surfaces of the samples in contact with atmosphere. The weak grain boundary surrounding these deposits resulted in intergranular fracture and the formation of small craters. Micrographs taken from the edges of the samples exhibit this cratered surface due to the pullout of precipitated grains of MgO. Samples could not be broken by hand and had to be fractured using a pestle, indicating that the samples were resistant to mechanical stress and that the addition of excess MgO and lanthanum dopant did not compromise the mechanical strength of the samples. The micrographs show a considerable degree of fracture debris, supporting this qualitative conclusion. DIELECTRIC PROPERTIES The addition of lanthanum significantly affected the dielectric data gathered and analyzed in Table 11. The T, decreased by about 10°C for every 0.5 mole % of lanthanum added, as seen in Figure 5.
Figure 5. Relative permittivity and dielectric loss as a function of temperature at 1 kHz for samples from batch 2 doped with 0.0 and 1.0 mole % La.
Dielectric Materials and Devices
87
An increase in firing time had little or no effect. The addition of excess lead zirconate seemed to augment the effect of the dopant; pushing Tm of the undoped sample slightly higher and the Tm of the heavily doped sample slightly lower. The maximum relative permittivity, km, decreased as the concentration of lanthanum increased, as seen in Figure 5. This trend was also reported previously by Chen et al. and is attributed to the increase of the magnesium to niobium ratio12. The addition of lanthanum decreases k, at 1 kHz in all cases, but the severity of the decrease is not dependent on the concentration alone. In the first batch, the addition of lanthanum did not have a significant effect on the relative permittivity until a doping level of 1.0 mole %. In the other two batches, the addition of 0.5 mole % lanthanum was sufficient to cause a significant decrease in relative permittivity. Other factors must be considered in this analysis, factors that were not measured such as the distribution and location of the pyrochlore phase and the appearance of a lead rich phase at the grain boundaries. The effect of the pyrochlore phase on relative permittivity was also removed using Lichtenecker’s Rule as described by Lattard7:
Where kmeaured is the measured relative permittivity, Xi is the volume fiaction of phase i, and ki is the relative permittivity of phase i. The relative permittivity of the pyrochlore phase was reported by Chen to be about 200 and the volume fraction of the pyrochlore phase was determined via peak area analysis from X-ray diffraction pattern^'^. The pyrochlore concentration is shown in Table 11. Figure 5 illustrates the corrected data and shows a significant increase in relative permittivity due to the removal of the low relative permittivity pyrochlore. The largest increase in corrected relative permittivity was in the undoped samples as these samples contained the highest concentrations of pyrochlore phase. The data were corrected to allow a more accurate comparison of the results of the dielectric tests. The addition of lead zirconate had an evident influence on the maximum relative permittivity. After compensating for the effect of pyrochlore, the relative permittivity of the second batch improved substantially. The effect between the second and third batches was less pronounced, but the addition of lanthanum augments the formation of lead vacancies and these vacancies have an effect on the relative permittivity. Once the effect of the pyrochlore phase was removed, the effect of the lead zirconate was less significant, as shown in Table 11. The lead deficient batch 1 exhibited a slightly lower corrected relative permittivity than the
88
Dielectric Materials and Devices
other two batches with no addition of lanthanum and a 1.0 mole % addition of lanthanum. The most important effect of a high lead atmosphere was to reduce the amount of pyrochlore and excess lead did not have an effect on the corrected relative permittivity. Aging is the formation of a secondary minimum in the curve of relative permittivity as a function of temperat~re’~. This phenomenon was not observed during the measurement of relative permittivity. All samples tested-showed a frequency dependence of Tm and km, as seen in Figure 6.
Figure 6. Uncorrected relative permittivity and dielectric loss as a function of temperature at 0.1, 1.O, 10, and 100 kHz for a sample from batch 2 doped with 1.0 mole % La. As the frequency of the measurements increased, k, decreased and T, increased, as expected. The dielectric loss data, included in the figure, showed an increase in loss and an increase in the temperature at which loss began to drop as frequency increased. As the frequency increased, the loss increased, as expected. The addition of lanthanum also caused a noticeable increase in dielectric loss. Diffuseness is a measure of the breadth of the curve describing the change in relative permittivity as a function of temperature. This parameter was calculated using two separate models using a statistical program called Multiple Correlation Analysis”. The results were examined through a multivariate analysis and are shown in Table 111.
Dielectric Materials and Devices
89
Table 111. Multivariate analysis of diffuseness and A-site charge. T Value R~ Value
Variable
Principle Factor
Y
indeterminant
-
6
lanthanum
6.82
A-site Charge
lanthanum
3.84
S,
Equation
0.869
2.670
6=47.878+14.867(La)
0.768
0.009
A-site=l.867+0. 104(La)-0.0793(La)2
First, diffuseness was calculated using the method described by Uchino and Nomura*, where y, the parameter related to diffuseness, is calculated according to the following equation:
where k, is the relative permittivity measured at temperature T, k, is the relative permittivity measure at T m , and C is the Curie type constant so that y corresponds to the slope of the curve with 1 5 y 5 2 . This parameter describes the deviation of the material from Curie Weiss behavior. Ordered ferroelectrics have a y value near 1 and as a material becomes more disordered, the value of y moves closer to 2. The calculation of this parameter did not yield a conclusive effect on the diffuseness of the curve indicating that the departure from Curie-Weiss behavior was not consistent. When diffuseness was calculated using the method originally proposed by Smolenskii and hrther described by Pilgrim et al., the concentration of lanthanum was the dominant variableg. This parameter describes the Gaussian diffuseness of the curve of relative permittivity as a function of temperature for temperatures greater than T m and is not bounded. Ordered ferroelectrics show a small value for the parameter of diffuseness, 6, and disordered ferroelectrics show a higher value for 6. The resulting equation that describes the effect of lanthanum doping on the diffuseness of the curve indicates that at low concentrations, the diffuseness varies linearly and becomes a squared relationship as the concentration becomes higher. As the dopant level increased, the parameter of difhseness increased in the same fashion regardless of firing conditions, indicating that the excess lead was not a factor. The following equation was manipulated such that the slope of the curve could be used to calculate 6.
90
Dielectric Materials and Devices
In this equation, the variables have the same significance as the previous equation. Although the changes in dielectric behavior are clear and can be described by their macroscopic behavior, the fundamental origin of the changes with La addition should lie with the internal charge distribution. There are many possible ways to compensate for charge imbalances due to the addition of a dopant to PMN-PT. For example, the ratio of magnesium to niobium can change or vacancies can be created to compensate for the dopant addition. The charge on the A-site cation was calculated for the various batches and compositions and is shown in Table 11. The effect of lanthanum and firing conditions was difficult to determine; therefore, a multivariate analysis was conducted. According to the statistical analysis, as shown in Table 111, the A-site charge is dependent on the lanthanum concentration. At lower concentrations, the A-site charge increased slightly, but at higher concentrations, the A-site charge begins to drop. Since the lanthanum increases the probability of lead vacancies, sufficient lanthanum was most likely not present to fill the empty lattice sites of the volatilized lead ions, leading to vacancies and a decrease in A-site charge. Confirmation of this effect was sought through the measurement of the induced polarization and microstrain, however, these did not yield conclusive results. Further experimentation is necessary to quantifj these properties. CONCLUSIONS The effects of lanthanum as a dopant and firing conditions are very important in the formation of high quality 0.9 PMN- 0.1 PT. Lanthanum had a nearly negligible effect on the microstructure. Lead volatilization was slightly increased with progressive La additions, but there was little effect on homogeneity and no effect on the grain size. There does not appear to be a significant effect on the amount of lead deficient pyrochlore phase formed. Lanthanum did promote intergranular fracture in samples fired under lead deficient conditions. The quantities of lanthanum used in this experiment were quite small (1 mole % maximum) and it is possible that a higher dopant concentration would show significant effects on these characteristics. The effect of lanthanum on the dielectric properties is more pronounced and often linear in nature. The increase in diffuseness, according to the diffuseness parameter 6, varies linearly with the addition of lanthanum, as does the decrease in T,. Maximum relative permittivity is also reduced with increasing lanthanum concentrations, but the decrease is not consistent and is most likely dependent on
Dielectric Materials and Devices
91
other factors such as grain size, homogeneity, and second phase location. The charge on the A-site is also varied with the addition of lanthanum, increasing at low dopant levels and decreasing slightly at higher levels. All samples exhibit a characteristic decrease in km and increase in T, as frequency increases. The influence of firing conditions is obvious and control of these conditions is important. Excess lead significantly decreases pyrochlore formation once the mass of lead zirconate added is equal to or greater than the mass of the samples. Excess lead slightly increases the grain size and slightly decreases inhomogeneity of the grains. The mode of fracture changes from intergranular to transgranular fracture as the amount of excess lead increases. The most important effect on the size and homogeneity of the grains is sintering time. Longer sintering times promote mass transport and grain size increases considerably. This also leads to a larger distribution in the size of the grains. The diffuseness parameter 6 is not significantly affected as excess lead is added. Excess lead, in quantities equal or greater than the mass of the samples, increases the relative permittivity of the samples, but only by reducing the amount of low relative permittivity pyrochlore formed. The effects of excess lead are only seen in the uncorrected relative permittivity data. Additional lead seems to augment the effect on Tm at higher concentrations of lanthanum. The charge of the A-site is not statistically affected by the excess lead. By fine tuning the properties of PMN based systems with lead titanate and lanthanum, useful materials for a variety of applications can be formed. These materials exhibit excellent dielectric characteristics such as high relative permittivity and broad temperature stability. Further experimentation will yield more information pertaining to control of grain growth during sintering and grain boundary phases. REFERENCES ‘T.R. Shrout, A. Halliyal, “Preparation of Lead-Based Ferroelectric Relaxors for Capacitors,” American Ceramic Society Bulletin,, 66 [4] 704-71 1 (1987). 2S.J. Jang, K. Uchino, S Nomura, and L. E. Cross, “Electrostrictive Behavior of Lead Magnesium Niobate Based Ceramics,” Ferroelectrics, 27 3 1-34 (1980). 3K. Uchino, “ElectrostrictiveActuators: Materials and Applications,” American Ceramic Society Bulletin, 65 [4] 647-52 (1986). 4 S. Kurutcharry, M. Lejeune, M. Oudjedi, S. Cousty, P. Abelard, “Potentialities of 0.9 PMN- 0.1 PT Ceramics for Active Vibration Control,” ISAF XI IEEE Proceedings (1998). %.M. Gupta and D. Viehland, “Compositional Studies of LanthanumModified Morphotropic Phase Boundary Pb(Mg1/3Nb2/3)03-PbTiO:,,” Journal of the American Ceramic Society, 80 [2] 477-485 (1997).
92
Dielectric Materials and Devices
6L.-J. Lin and T.-B. Wu, “Ordering Behavior of Lead Magnesium Niobate Ceramics with A-Site Substitution,” Journal of the American Ceramic Society, 73 [5] 1253-1256 (1990). 7E. Lattard, “Propriktks Electrostrictives de Ckramiques Massives du Type PbMg1/3Nb2/303”Ph.D. Thesis, E.N.S.C.I., Limoges, France (1996). *K. Uchino and S. Nomura, “Critical Exponents of the Dielectric Constants in Diffused-Phase-Transition Crystals,” Ferroelectrics Letters, 44 55-6 1 (1982). 9S.M. Pilgrim, A.E. Sutherland, S. R. Winzer, “Diffuseness as a Useful Parameter for Relaxor Ceramics,” Journal of the American Ceramic Society, 73 [lO] 3122-3125 (1990). “J.C. Wurst, J.A. Nelson, “Lineal Intercept Technique for Measuring Grain Size in Two-Phase Polycrystalline Ceramics,” Journal of the American Ceramic Society, 55 [2] 109 (1972). “J.C. Shaw, K.S. Liu, I.N. Lin, “Phase Boundary of an LMN-PZ-PT ThreeCom onent System,” Journal of Material Science, 28 [20] 5534-5539 (1993). Chen, H.M. Chan, M.P. Harmer, “Ordering Structure and Dielectric Properties of Undoped and La/Na-Doped Pb(Mg1/3Nb2/3)03,”Journal of the American Ceramic Society, 72 [4] 593-598 (1989). I3J. Chen, “Effect of Powder Purity and Second Phases on the Dielectric Properties of Lead Magnesium Niobate Ceramics,” Journal of the American Ceramic Society, 69 [ 121 C-303-5 (1986). 14W.A.Schulze, K. Ogino, “Review of Literature on Aging of Dielectrics,” Ferroelectrics, 87 361-377 (1988). ’Multiple Correlation Analysis Version, Next Bridge Software, 1999.
‘4.
Dielectric Materials and Devices
93
This page intentionally left blank
EFFECT OF MILLING PROCESS ON CORE-SHELL MICROSTRUCTURE FOR BaTi0,-BASED Ni-MLCC Youichi Mizuno, Tomoya Hagiwara, Hirokazu Chazono, and Hiroshi Kishi Material Development Department, Taiyo Yuden Co., Ltd. 5607-2 Nakamuroda Haruna-machi Gunma-gun gunnma 370-3347, Japan
ABSTRACT The effect of the process parameter in the milling process on the core-shell microstructure was investigated in BaTiO, (BT)
- Ho,O, - MgO system. The degree of damage given by milling
process for BT increased, the crystallinity of BT decreased, the mean grain size decreased, and the number of the chipped particle increased as the amount of the media increased. It was found that the milling damage had a fatal influence on the microstructure. The mean grain size (D,,) determined from the chemically etched samples fired at 1320'C decreased as the damage increased. D,, was almost equal to that of the initial particle. Therefore, little grain growth was occurred for all samples. 100 grains were observed by the transmission electron microscopy (TEM) and analyzed statistically. TEM observation revealed that there were the grains without 90' domain pattern and the grains showing only 90' domain pattern, as well as the grains showing core-shell microstructure. The rate of frequency for grains showing the core-shell microstructure increased as the damage increased, judging from the statistical analysis by TEM.
INTRODUCTION In the past decade, the technology for fabricating multilayer ceramic capacitors (MLCCs) has made great strides by the need for higher volumetric effciencies, and cost reduction. In order to meet these stringent requirements, the number of the dielectric active layers was increased, and the
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
95
thickness was decreased, without compromising product reliability. Furthermore, the noble metal (Pd) is replaced by the base metal (Ni) as the internal electrode material to lower the production cost. In recent years, MLCCs with Ni internal electrodes composed of more than 400 layers of below 3pm thickness have been developed. A deliberate design of the microstructure for individual grains must be essential for such MLCCs with considerably thin dielectric layers since the microstructure has a decisive influence on the electrical properties and reliability of the MLCCs. It is well known that the microstructure has a large influence on the electrical properties and that the material with flat temperature characteristics has so-called grain core-grain shell structure.'*2It was reported that the incorporation of the rare earth ion into BT lattice was dependent on the ionic radius of the rare earth
and that the site occupancy of the rare earth ion affected
the microstructural e v o l ~ t i o nOn . ~ ~the ~ other hand, the microstructure and the electrical properties are also dependent on the fabricating process. Therefore, the influence of the degree of damage given by the milling process on the microstructure was studied for materials in BT - Ho,O, - MgO system. EXPERIMENTAL PROCEDURE
Sample Preparation Samples were prepared by the ball-milling BT with reagent grade oxide powders of l.Omol% Ho,O,, 0.5mol% MgO, O.lmol% MnO, and 1.5mol% BaSiO, for 15h. BT was synthesized hydrothermally with a mean particle diameter of about 0.5pm (SAKAI Chemical Industry Co., Ltd., Osaka, Japan). BaSiO, was used as a sintering aid. The degree of damage for BT powder was controlled by the amount of the milling media. The weight of media used in a milling process was equal to that of BT for the sample- 1, and was twice and four times as heavy as BT for the sample-2 and sample-3, respectively. The powder mixtures were subsequently dried and sieved. The obtained powders with an appropriate organic binder system were uniaxially pressed into disks (10mm4 X 0.6mm') and fired at 132OoCfor 2h in a reducing atmosphere controlled by H,, N,, O,, and H,O, then cooled to room temperature in a weakly oxidizing atmosphere (P(0,)
= 30
Pa at 1000°C).
Characterization Specific surface areas were measured using the conventional nitrogen adsorption (BET)
96
Dielectric Materials and Devices
technique and the samples were observed by the field emission scanning electron microscopy (FESEM) to study the particle diameter changes by milling process. The X-ray difiaction analysis (XRD; Cu-ka, 50kV-150mA; RINT2500, Rigaku Co., Tokyo, Japan) was carried out for milled powders in order to identify the secondary phases and to evaluate the degree of damage given by milling process for BT powder. The XRD measurement was carried out in the 28 range from 20' to 60'. The degree of damage, which will be abbreviated as Ed, hereafter, was defrned as the ratio of the Full-Width at Half Maximum (FWHM) of the peak for the BT powder, indicated follows Ed = P a 1 P O where
p, and PO are FWHM of milled powders and FWHM of as-received raw BT powders,
respectively. FWHM was calculated with the WPPD method accurately,' using the diffraction peaks corresponding to the (200) and (002) planes of the perovskite BT. The shrinkage behavior was measured by dilatometry (DILATO; Macscience, Japan) in a reducing atmosphere, and the phase transition of the sintered samples was characterized by differential scanning calorimetry (DSC; Macscience, Japan). The samples were observed by FE-SEM to study the microstructural changes, such as mean grain size and grain size distribution. The mean grain size was determined by the intercept method with a micrometer from FE-SEM images of the chemically etched surface with HNO, + HF solution. The mean grain size is the grain diameter giving 50% of the accumulated volume. In order to classify the microstructure of the grain, 100 grains with not less than 0.3pm were observed by TEM (200kV; JEM-200CX, Nihondenshi, Japan). The samples were observed with a sufficient tilt in TEM analysis to treat statistically. RESULTS AND DISCUSSION Powder Characterization Powder characteristics were investigated in order to compare the influence of milling strength. Figures 1 and 2 show SEM micrographs and the particle diameter distribution determined by SEM images of milled and raw BT powders, respectively. Small particles of less than O.lpm were not measured. The particle features were summarized in Table I. It was found that the crystallinity of BT decreased as the amount of the media increased, which was determined by FWHM calculation. Furthermore, SEM images revealed that the frequency of chipped particles increased as the damage
Dielectric Materials and Devices
97
0.1
grain diameter (ccm)
1
Fig. 1 The accumulated volume indicating the particle size distribution for raw BT powders and powders milled at various degrees of damage.
Fig.2 SEM micrographs for the raw BT powders and the powders milled at various degrees of damage (bar=l pm).
98
Dielectric Materials and Devices
became heavier, which were not used for the determination of the mean particle size. The increase in chipped particles resulted in the increase in BET. However, the increase in BET between BT and light damage was caused by the specific area of the additives as well as that of the chipped particles, although the contribution of the additives was not estimated in this study. Therefore, it was reasonable that the degree of damage for BT powder by milling process was expressed by FWHM in XRD analysis but not by BET. It was found that the increase in the amount of milling media brought about the chipped particles of BT powders and the decrease of the BT crystallinity.
raw BT sample-1 sample-2 samvle-3
D,, (Pm) 0.500 0.508 0.489 0.479
BET (m'/g> (2.40) 3.53 3.71 4.40
FWHM
Ed
0.103607 0.107956 0.11 1303 0.122402
1.ooo 1.074 1.181 1.773
( "1
(-1
Sintering behavior Figure 3 shows the shrinkage behavior for all samples, indicating that Ed affected the rate of shrinkage. The temperature of the shrinkage onset decreased from 1258' to 1226'C, and that of the shrinkage cease decreased from 1353' to 1316'C, when Ed increased from 1.07 to 1.77 as shown in Table 11. The disk samples fired at 1320'C for 2hr were densed sufficiently for all samples. Microstructual Characterization The secondary phase of all samples fired at 1320 "C was identified by XRD powder method. The XRD profiles were shown in Fig.4. There were three kinds of peaks; large peaks for the modified BT phase (0),small peaks for secondary phase, and the peaks of unreacted raw material. The thud ones were identified to be the BaCO, (#), and the secondary phase was unknown (I). The peak height of the secondary phase decreased as Ed increased, and there were not the secondary phases in the highest Ed sample. Figure 5 shows the grain size distribution determined from the chemically etched surface of the sample fired at 1320'C. The profiles of the curve were similar for all samples, but mean grain size (D5J decreased as Ed increased as shown Table 111. D,, was almost equal to that of the initial particle as shown in Table I. Therefore, little grain growth was occurred for all samples.
Dielectric Materials and Devices
99
5 0 -5 h
8
5
-1° -15 -20 -25 1
temperature (‘c) Fig.3 The shrinkage behavior for milled samples.
1.181
1257.9
1344.2
1000
750
500
250
Ed= 1.773 Ed=l .181 Ed= 1 .074
0 20
30
40
50
&O
20 0
Fig.4 The powder XRD profiles of the samples fired at 1320°C; 0 indicates matrix BT modified phase, # is the peak of BaCO,, and Iis the peak of unknown phase.
100
Dielectric Materials and Devices
100 r
I
I
I
I
75
50 25
0 1.o
0.1
grain size (pm) Fig.5 The grain size distribution determined from the chemically etched surface of the sample fired at 1320°C. Table 111. The mean main size. Ed (-1
1.181
0.495
Furthermore, careful TEM observation revealed that the microstructure of the grain was classified into three categories; the grain showing only 90" domain pattern (named C-grain), the grain showing core-shell structure (named CS-grain), and the grain without 90' domain pattern (named S-grain) as shown in Fig.6. Figure 6 shows the typical TEM micrographs. It was reported that the grain core-grain shell microstructure in CS-grains was formed by the reaction between the additives and BT.*s9It was noteworthy that there were C-grains and S-grains, when the careful and statistical observation by TEM was done. The rate of frequency for three kinds of grains was summarized in Table IV.The rate of frequency for CS-grain increased, and that for C-grain decreased as the damage increased. The rate of frequency for S-grain was almost constant for all samples. In order to obtain the information from the core region, DSC measurement was carried oui since the core is composed of pure BT and the result was shown in Fig.7. The results of DSC measurement are summarized in Table V. The profiles of the endothermic peak at around 125OC
Dielectric Materials and Devices
101
were dependent on Ed. The peak was broadened and the peak area decreased as the damage become heavier, whereas the peak temperature was almost independent on Ed. The change of the peak profile must be conjunction with the volume of the core region and the presence of the internal stress in samples composed of the core-shell microstructure.
Fig.6 The typical TEM micrographs; (a) is the C-grain, (b) is the CS-grain, and (c) is the S-grain (ba~0.5pm).
Ed 1.074 1.181 1.773
C-grain 50 48 23
CS-grain 46 45 70
S-grain 4 7 7
The given damage on BT by milling process has a fatal influence on not only the mean grain size but also the microstructure in a grain. It was suggested that the C-grains were mainly composed of the pure BT, judging from the fact that the grain showed only 90' domain, and that the S-grains were formed by the diffusion of the additives into a whole grain. The damage on BT surface accelerated the incorporation of additives like Ho, Mg, and Mn into BT. The active chipped
102
Dielectric Materials and Devices
particles reacted with the additives and formed S-grains. Therefore, the rate of frequency for CSgrains and S-grains increased as Ed increased. The real rate of frequency for the S-grains must be larger than that shown in Table IV in the sample with the highest Ed, because the grains with less than 0.3pm were not analyzed in TEM observation. Further investigation of the statistical analysis of the grains with less than 0.3pm and the physical and chemical properties such as the compositional distribution in the shell region is necessary to elucidate the relationship between the microstructure and milling damage.
: / -
Ed= 1.773
Ed= 1.074
I
80
90
100
110
120
130
140
150
temperature (“c) Fig. 7 DSC profiles of the fired samples. Table V. DSC peak temperature and the peak area. peak temperature peak area Ed (-1 (“C) (mcal/g) 1.074 124.0 100.00 1.181 124.2 86.66 1.773 123.6 75.08 REFERENCE ‘H. Saito, H. Chazono, H. Kishi, and N. Yamaoka, “X7R Multilayer Ceramic Capacitor with Nickel Electrodes,” Japanese Journal ofApplied Physics, 30 2307-23 10 (1991) 2T. R. Armstrong and R. C. Buchanan, “Influence of Core-Shell Grains on the Internal Stress State and Permittivity Response of Xirconia-Modified Barium Titanate,” Journal of the American Ceramic Society, 73[5] 1268-1273(1990)
Dielectric Materials and Devices
103
,H. Kishi, Y. Okino, M. Honda, Y. Iguchi, M. Imaeda, Y. Takahashi, H. Ohsato and T. Okuda,
‘‘ The Effect of MgO and Rare-Earth Oxide on Formation Behavior of Core-Shell Structure in BaTiO,,” Japanese Journal of Applied Physics, 36 5954-5957 (1997)
4H. Kishi, N. Kohzu, Y. Mizuno, Y. Iguchi, J. Sugino, H. Ohsato and T. Okuda, “Effect of Occupational Site of Rare-earth Elements on the Microstructure in BaTiO,,” Japanese Journal of Applied Physics, 38 5452-5456 (1 999)
5Y. Mizuno, Y. Okino, N. Kozu, H. Chazono, and H. Kishi, “Influence of the Microstructure Evolution on Electrical Properties of Multilayer Capacitors with Ni Electrode,” Japanese Journal
of Applied Physics, 37 5227-523 1 (1988) 6H. Chazono, Y. Okino, N. Kohzu and H. Kishi, “Effect Of Sm and Ho Addition on the Microstructure and Electrical Properties in MLCC with Ni Internal Electrode,”; pp.53-64 in Ceramic Transaction, Vol. 97, Edited by Jau-Ho Jean, T. K. Gupta, K. M. Nair, and K. Niwa, The American Ceramic Society, 1999 ’H. Toraya, “Whole-Powder-Pattern Fitting Without Reference to a Structural Model: Application to X-ray Powder Diffractorneter Data”, Journal Applied Crystaffography. 19, (1986) 440-447 ‘H. Chazono and H. Kishi, ‘‘ Sintering Characteristics in BaTi0,-Nb205-Co,O, Ternary System: 1, Electrical Properties and Microstructure,” Journal of the American Ceramic Society, 82 [ 101 2689-97 (1999)
’H. Chazono and H. Kishi, ‘‘ Sintering Characteristics in BaTi0,-Nb205-Co,04 Ternary System: 2, Stability of So-called “Core-Shell” Structure,” Journal of the American Ceramic Society, 83 [ l ] 101-106 (2000)
104
Dielectric Materials and Devices
INTERRELATIONSHIP BETWEEN SELF-HEATING AND FERROELECTRIC PROPERTIES IN PZT CERAMICS DURING POLARISATION SWITCHING M. H. Lente, D. Garcia* and J. A. Eiras Universidade Federal de S8o Carlos - Departamento de Fisica Gmpo de Cerimicas Ferroeletricas Caixa Postal 676 - CEP 13565-670 - S8o Carlos - SP - Brazil
ABSTRACT This work dealt with the interrelationship between self-heating and both fatigue and depinning process (extracted from P-E hysteresis loop measurements) of doped PZT ceramics. During electric field cycling, niobiumdoped PZT presented fatigue effect on polarisation and coercive field, while for iron-doped PZT, constricted hysteresis changed to normal hysteresis (depinning process). Temperature trend followed the fatigue or depinning behaviours of the ceramics during polarisation switching and was strongly compositional and frequency dependent. The self-heating process could be visualised as domain rotation in a viscous medium. INTRODUCTION Ferroelectric ceramics used to convert mechanical energy into electrical energy and vice versa are very important in the industrial applications such as igniters, transducers and actuators [ 11. Therefore, to improve the performance of these materials we need to understand the process of domain rotation under high electric field and its consequences. An obvious source of energy loss and consequent heat generation in ferroelectric materials lies in their characteristic hysteresis loops. This effect termed macrohysteresis effect is manifested in the familiar (P,E) hysteresis loops and begins when the applied field is sufficient to make a boundary switching from one position to another [2]. Several parameters can influence the energy loss as sample size, electric field strength and frequency [3].
*Actualtemporary address: Material Research Laboratory, The Pennsylvania State University, University Park, PA, 16801, USA. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this ublication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright 8learance Center, is prohibited.
Dielectric Materials and Devices
105
Sample heating rate can be also greatly modified by adulteration of cation site with ions of different valence [4]. It has been also observed that when a bipolar electric field, larger than the coercive field, continuously drives a ferroelectric ceramic three effects may occur: domain depinning [51, ferroelectric fatigue [6, 7, 81 and self-heating of the sample [3, 41. However, it has been not reported a direct relationship between both ferroelectric properties and selfheating trend under high bipolar electric field. Thus, an important question regarding the existence of the correlation between these properties remain open. In this work, we investigated the interrelationship between self-heating and both fatigue and depinning process in PZT ceramics extracted from the P-E hysteresis loop at 60 and 0.1 Hz. In order to investigate the influence of the doping, PZT ceramics were doped with Nb205 (donor) or Fe203 (acceptor). EXPERIMENTAL Pb(Zr0.53Ti0.47)03 ceramics, doped with 1 wt.% of Nb205 or Fe203 (hereafler, PZTN and PZTF, respectively), were prepared by conventional solid state reaction. The sintering temperature was 1250°C in a PbO atmosphere. Scanning electron micrographs showed that the average grain size of the ceramic lies between 3.0 - 3.5 pm. Sintered discs, with 17 mm in diameter, were polished to a thickness of 0.35 mm. After annealing, silver electrodes were painted on both sides of the discs. High amplitude sinusoidal bipolar electric field (0.1Hz and 60 Hz) was applied at room temperature to carry out the polarisation switching measurement, using a system similar to the Sawyer-Tower bridge. The sample self-heat generation during continuous electric field driving was observed monitoring its temperature through a thermocouple attached on one of its face. The specimens were kept in a silicon oil bath, which its temperature was also monitored. RESULTS Figure 1 shows the ferroelectric hysteresis loops for the PZTN measured at 60 Hz for several fatigue cycles. We can see that during the fatigue cycles both the saturation (Ps) and remanent (PR) polarisations gradually decreased whereas the coercive field (Ec) increased. As consequence, the hysteresis loop becomes rounded. Figure 2(a) shows Ps, PR and Ec versus the number of cycles for the PZTN obtained from the hysteresis loop measurements (figure 1). In the initial cycles Ps, PR and Ec change rapidly. While Ec increases quickly, PS and PR decrease accordingly. The specimen temperature as finction of the number of cycles is shown in figure 2(b).
106
Dielectric Materials and Devices
-30
-10
-20
10
0
20
30
Electric Field (kV/cm)
Figure 1. Hysteresis loops for the PZTN ceramics at 60 Hz. The results show that the temperature increased quickly just after the electric field cycling begin. Then, it passes through a maximum (near 90.C) and 1 decreases continuously until reach its stabilisation at 6OoC, after about 7 . 5 ~O5 cycles. In that point, the oil bath temperature was purposely increased from room temperature up to 9OoC,as shown in figure 2(b). Slight increase of PSand PR and decrease of E, occurred. Then, the oil bath was cooled until the room temperature again and the PZTN polarisation kept the same trend, which it had before the oil heating. 180
.
160 -
01
0
"
250
'
I
.
'
"
500 750 1000 Switching Cycles x 103
"
1250
'
14 1500
.
.
,
.
,
PZTN - 60 Hz
0
"
"
250
"
.
,
.
,
.
@
.........................
20
,
"
"
.i..' '
'............_____._..
500 750 1000 Switching Cycles x 1 O3
1250
1500
Figure 2. Dependence with the number of electric field cycles of (a) saturation and remanent polarisations and coercive field; (b) surface specimen temperature for PZTN at 60 Hz. Figure 3(a) show the ferroelectric hysteresis loops for the PZTN measured at 0.1Hz in the initial state and after 4 . 4 ~ 1 cycles 0 ~ and figure 3(b) the
Dielectric Materials and Devices
107
polarisation and coercive field dependence with number of cycles. As observed for the 60 Hz case (figure 2(a)), both PS and PR decreased whereas the coercive field (Ec) increased. However, no temperature change was detected on the sample surface in this case. Figure 4 shows the ferroelectric hysteresis loops for the PZTF measured at 60 Hz for several fatigue cycles. In the initial state, the PZTF was characterised by a constricted hysteresis loop, but after continuous electric field switching, a normal ferroelectric loop was observed. 40
30 20
Y
0 10
3
,E c
0
.I -10 m
2 -20 -30 -40
-30
-20
-10
0
20
10
Electric Field (kVlcm)
0
30
10
20 30 Switching Cicles x 103
40
50
Figure 3. (a) Hysteresis loops at 0.1 Hz, and (b) saturation and remanent polarisation and coercive field dependence with the number of cycles for the PZTN ceramics.
.g
4
m
g
-10
-1s
-50
-20
-10
0
10
Electric Field (kV/cm)
20
30
Figure 4: Hysteresis loops for the PZTF ceramics at 60 Hz. Figure 5(a) shows a remarkable continuos increase of the polarisation and coercive field for the PZTF with the number of cycles. Similar behaviour has been observed in PZT films [ 5 ] .The temperature dependence with the number of
108
Dielectric Materials and Devices
cycles at 60 Hz is shown in figure 6(b). It can be seen that the temperature increased continuously during all electric field cycling.
//..i!
- 16 -0-CoercIve
Fleld
-
T
-
~ 1 4 0
12 3
.
P 100 (
1
'
a h PZTF - 60HZ
r 03 0
"
250
500
'
750
"
1000
"
1250
Switching Cycles x 1O3
'
14 1500
20
0
250
500 750 1000 Switching Cycles x 1 O3
1250
1500
Figure 5: Dependence with the number of cycles of (a) saturation and remanent polarisations and coercive field; (b) surface specimen temperature for the PZTF ceramics at 60 Hz. After the thermal annealing above Tc, the PZTF and PZTN samples recovered again their initial properties without fatigue memory, which indicates fatigue or depinning processes generated by space charges, as discussed in reference [81.
DISCUSSIONS The remarkable observation reported in this work is that for all measurements made at 60 Hz both the sample temperature and the polarisations curves showed the same trend during all cycling time, implying a correlation between them. However, different behaviours generated by the doping action were found. While for the PZTN the polarisation and specimen temperature reached a maximum value and then decreased until stabilise, for the PZTF they increased continuously. This results might be explained supposing the domains moving into a viscous medium, where the viscosity is modified by the defects generated fiom the different kind of additives. The importance of viscosity has been already considered for explaining the switching properties of the ferroelectrics [9, 101. Ferroelectric properties can be strongly modified by impurity doping effects. In the initial state, Pb(Zr, Ti)Os ceramics doped with acceptor atoms (Fe3+as example) have their domains clamped by defects and its hysteresis loop is characterised by a constricted loop [ l 11, as observed in this work. It has been known that oxygen vacancies are introduced by valence compensation when ceramics are doped with acceptor atoms and these vacancies are trapped on the domain walls [11, 121. Acceptor atoms and oxygen vacancies form electric
Dielectric Materials and Devices
109
dipoles called complex defects, which act as pinning points for the domain motion, and consequently the domain rotation is reduced [ 131. A bipolar electric field or dc poling, at high temperature, can induce ferroelectric domain depinning process [5]. In this case, removal of 90" domain pinning occurs and the ceramic shows a remarkable increase of the switching polarisation [5]. Otherwise, when donor atoms are added we have the absence of complex defects and the domains become more mobile [ 11, 131. Considering the assumption that Nb5+acts as donor ion in PZT, its doping effect is making the domains more mobiles. Consequently, due to the high and fast domain amplitude rotation the sample temperature quickly increases. Then, fatigue processes, generated by the arrangement of space charges, continuously clamp the domains, reducing the domain reorientation and consequently decreasing the temperature. When the amplitude of the domain rotation maintains constant, the sample temperature reaches also a constant value. The PZTF in the virgin state presents a constricted hysteresis loop due to the pinning of the ferroelectric domains by the complex defects. When the external electric field is switched ferroelectric domains tend to rotate and the depinning process starts. This increase of the polarisation, by a continuo increase of the degree of the domain reorientation, induced a continuous heating of the sample. In our experiment, a constant polarisation value could not be reached during the cycling time investigated probably because a saturation of the depinning process was not reached. The viscosity of the medium could be modified by different kinds of additives increasing or decreasing the loss. The point that indicates the influence of the doping on the self-heating comes from of the direct comparison between the polarisation value and the respective sample temperature. For the same polarisation, the temperature for the PZTF is always higher than PZTN. This explanation corroborates that different doping produces modifications in viscous medium. The frequency of the electric field cycling is another parameter to consider on the self-heating process. The electric field driving made at 0.1Hz did not produce detectable heating on the PZTN sample surface. Since PZTN domains have high mobility, the slower frequency rotation produces heating at lower rate. Consequently, the possible dissipation of the heat to the oil bath did not permit the sample heating. Although the heating absence in the PZTN measured at 0.1 Hz, the fatigue process was present. Nevertheless, the fatigue rate observed was lower than that measured at 60 Hz (figures 2 and 3). This fact can be explained assuming that space charges are more mobiles at high temperatures [ 141.
110
Dielectric Materials and Devices
CONCLUSION In this work, the interrelationship between self-heating process and both ferroelectric polarisation fatigue and domain depinning processes in doped PZT ceramics was investigated. Fatigue process implied decrease of the sample temperature. For niobium doped PZT, domain depinning process occurred firstly, followed immediately by fatigue process due domain pinning by space charges. For iron-doped PZT, the domain depinning process occurred during all studied polarisation cycling range because of continuous increase of the amplitude of the domain rotation and consequent increase of the self-heating. The temperature arising can be visualised as result of the heat generated by fiiction of the domains in a viscous medium. Higher the switching frequency, higher was the temperature. Acknowledgements The authors thank to FAPESP, CNPq and PADCT/CNPq by financial support.
REFERENCES 1
B Jaffe, W. R. Cook and H. Jaffe, “Piezoelectric Ceramics,” Academic, New York, 1971. * B. Lewis, ”Energy Loss Process in Ferroelectric Ceramics,” Proc. Phys. SOC.73 [ 171 17-24 (1959). 3 J. Zheng, S. Takahashi, S. Yoshikawa and K. Uchino, “Heat Generation in Multilayer Piezoelectric Actuators,” J. Am. Ceram. Soc. 79 [ 121 3 193-3198 (1996). 4 R. A. Gdula, “High field Losses of Adulterated Lead Zirconate titanate Piezoelectric Ceramics,” J. Am. Ceram. Soc. 51 [121 683-687 (1968). 5 M. Kohli, P. Muralt and N. Setter, ”Removal of 90” domain Pinning in (100) Pb(Zro.15 Tio.85)03 Thin Films by Pulsed Operation,” Appl. Phys. Lett. 72 [24] 3217-3219 (1998). 6A. Levstik, V. Bobnar, 2. Kutnjak and M. Kosec, “Fatigue and Piezoelectric Properties of Lead Lanthanum Zirconate Titanate Ceramics,” J. Phys. D: Appl. Phys. 31 2894-2897 (1998). 7 W. Pan, S. Sun and P. Fuierer, ‘%Effects of Ferroelectric Switching on the Dielectric and Ferroelectric Properties in Lead Zirconate Titanate Ceramics and Their Modeling,” J. Appl. Phys. 74 [2] 1256-1264 (1993).
Dielectric Materials and Devices
111
8
Q. Y. Jiang, E. C. Subbarao and L. E. Cross,”Effect of Composition and Temperature on Electric Fatigue of La-doped Lead Zirconate Titanate Ceramics,” J. Appl. Phys. 75 [l 117433-7443 (1994). 9 M. Omura, H. Adachi and Y. Ishibashi, ”Simulations of Ferroelectric Characteristics Using a One-Dimensional Lattice Model,” Jpn. J. Appl. Phys. 30 2384-2387 (1991). ‘OL. Mitoseriu, D. Ricinschi, C. Harnagea, M. Okuyama, T. Tsukamoto and V. Tura, “Grain Size Dependence O S Switching Properties of Ferroelectric BaTi03 Ceramics,” Jpn. J. Appl. Phys. 35 5210-5216 (1996). 11 S. Takahashi, ‘Zffects of Impurity Doping in Lead Zirconate Titanate Ceramics,” Ferroelectrics 41 143-156 (1982). 12 X. Zhang, T. Hashimoto and D. C. Joy, “Electron Holographic study of ferroelctric domain walls,” Appl. Phys. Lett. 60 [6] 784-786 (1992). 13 L. Wu, C. Wei, T. Wu and C. Teng, “Dielectric properties of modified PZT ceramics,” J. Phys. C: Solid State Phys. 16 2803-2812 (1983). 14 K. Okazaki and K. Sakata, “Space Charge Polarization and Aging of Barium Titanate Ceramics,” Electrotech. J. Jap. 7 [131 13-18 (1962).
112
Dielectric Materials and Devices
MEASUREMENTS OF PYROELECTRIC RESPONSE ON BARIUM STRONTIUM TITANATE SINGLE CRYSTAL FIBERS D. Garci;,
R. Guo and A. S. Bhalla
Materials Research Laboratory, The Pennsylvania State University, University Park, PA, 16802 ABSTRACT In this work, temperature dependence of the pyroelectric response and polarization behavior are investigated as a function of Sr content (x=O. 10-0.40) in (Bal-,Sr,)TiOs single crystal fibers grown by the Laser Heated Pedestal Growth technique. Broadening of pyroelectric current peak at Tc, which is shifted down, and lowering of polarization magnitudes are observed gradually with the increasing content of strontium. Hysteresis loops and dielectric measurements were also performed to aid the discussion. INTRODUCTION Barium strontium titanate (BST) solid solutions have been recognized as suitable materials for various types of sensors, especially as infrared pyroelectric detectors (1,2). The Curie temperature of ferroelectric BST solid solutions can be easily adjusted by the BdSr ratio and brought close to room temperature. This fact, added with the achievement of high dielectric constants at paraelectricferroelectric phase transition peak, allows its application in DC-bias field induced pyroelectric devices or dielectric-bolometers (2-5). Nevertheless BST is a well known material, applications as well as studies on properties and phenomenology of doped and undoped BST are based on the * Permanent address: Department of Physics, Federal University of Sao Carlos, Sao Carlos, SP. E-
mail address: ducinei @psu.edu or ducinei @power.ufscar.br
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of h s publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
113
data taken on ceramic or thin film form, due the difficulties in growing good single crystals of this system. High melting points and no congruently melting compositions (6) are the factors that disable the growth of high quality barium strontium titanate single crystals by the regular methods. In this case, crucible-free and fast cooling/sharp temperature gradient are the basic requisites to attempt the growth of BST in the crystalline form. Earlier reports have shown that the Laser Heated Pedestal Growth is a suitable technique for growing BST single crystals (6,7), under suitable growth conditions. This paper reports the pyroelectric response of (Bal-,Sr,)Ti03 (x=O. 100.40) ferroelectric single crystal fibers, grown by the Laser Heated Pedestal Growth (LHPG) technique. Temperature dependence of the dielectric properties and polarization-electric field (P-E) hysteresis loop were investigated to aid the discussions.
EXPERIMENTAL PROCEDURE Ceramic rods of (Bal-,Sr,)TiO3, with x=O.lO-0.40, were prepared to be used as seed and feed during growth. BaTiO3 (Ferro Corp., reagent purity) and SrCO3 (Fisher Chemicals, 99.5 %) were used as starting materials. Mixed raw materials were calcined at 1 100°C/3h in covered A1203 crucibles and ball milled for 24h, with zirconia cylinders in alcohol medium. Ceramic discs, placed in alumina crucibles, were sintered at 1270°C-1330°C (with the increase in Sr content) for 2h, in air. After polishing, the discs were cut into rods of - 1mmx1mmx17mm. Laser Heated Pedestal Growth technique was employed to grow BST single crystal fibers. Basically, the assembled ceramic feedrods were placed in a chamber, where a tunable CO2 laser beam is center focused and creates the molten zone, and then pulled through the hot zone after connected with the seed rod. The growth conditions were: ceramic as seed (considered “free growth”), molten zone temperatures of 1750°C-1850°C (higher concentrations of strontium, higher temperatures) and 12-15 mm/hr for pulling rates. Transparency and light brown color single crystal fibers of -15mm length and 0.9mm in diameter could be grown easily. More details on the equipment and BST growth conditions can be found in the references 8 and 9. X-ray diffraction patterns of crushed fibers showed only BaTi03-like perovskite single phase with a decreasing of tetragonality factor c/a as Sr content increased. At x>0.30, a pseudo-cubic symmetry is observed at room temperature. XRD studies on the circular cross sections also showed that growth direction changed with strontium content and it is <110> for x<0.30 and
for x>0.30. After this semi-quantitative structure analysis, the fibers were cut, polished,
114
Dielectric Materials and Devices
annealed at low temperature (-700°C) to release the mechanical stresses and then gold-electroded on the circular cross-section areas for the pyroelectric and complementary dielectric and P-E loop measurements. The static technique developed by Byer and Roundy was used for the pyroelectric measurements (10). After poling, tbe crystal was cooled and the thermal depolarization current (pyroelectric current) was measured through an HP4140B pA meter under zero-field condition during the heating run (up to the ferroelectric-paraelectric phase transition Tc). After the poling process and during further cooling the samples were short-circuited to avoid built-in charges. The pyroelectric coefficient, p , was calculated by using an expression, p =i/[A(dT/dt)
where i is the pyroelectric current, A , the electroded area and dT/dt is the temperature rate (4’C/min in this case). The remnant polarization P, is calculated from:
Pre-poling conditions were: 3kV/cm, for 15 min., at temperatures -3O’C below the phase transition, i.e., T=TC-30’C. High built-in charge currents were observed in the (Ba09oSro,lo)Ti03 single crystals, especially close to Tc, which mixes up calculation of the remnant polarization (data for these samples is not discussed here). Dielectric measurements during cooling and heating runs with the same temperature rate as used in pyroelectric measurements, i.e. 4’C/min, were carried out using a computer assisted system with the HP4284A LCR-meter. Room temperature hysteresis loops were obtained through a Radiant Technologies RT66A standardized ferroelectric system. RESULTS AND DISCUSSION Figure 1 displays the temperature and frequency dependence of the dielectric properties of (Bal-,Sr,)TiO3 single crystal fibers. Like other BaTi03 based systems (1 l), during cooling, BST undergoes three ferroelectric phase transitions: cubic-tetragonal, tetragonal-orthorhombic and orthorhombic-rhombohedral. In this particular system, as anticipated, the lowering of the three phase transition temperatures with Sr content is observed. Thus, “pinching” effect of phase transition is occurring with the substitution of Ba2+ cation by the smaller radius Sr2+ cation. Relatively sharper transitions at the ferro-ferroelectric and ferro-
Dielectric Materials and Devices
115
paraelectric temperatures appear with high dielectric constants and small dispersion with frequency. Compositions with x=0.30 and 0.40 showed T, 15OC over and below room temperature, respectively. Those results are in good agreement with the values obtained for BST ceramics with the same BdSr ratios (12). Figure 2 gives the temperature dependence of the remnant polarization and pyroelectric coefficient. Gradual lowering of Pr with Sr content occurs, following the trend of tetragonality factor as commented before. Different temperature dependence of P, and p are observed when these are compared for the crystal x=0.20 and the other two, i.e. x=0.30 and 0.40. This result is expected from the fact that the former one has crystal (growth) orientation (at room temperature) in <110> direction and the later two have . The final domain configuration (in
-
20000,
,
I
.
.
I
.
I
. ,
I
.
,
0.8
20000, c 16000
,
I
.
I
.
I
.
.
,
>
Ba, ,Sr,TO i,
(b)
1kHz-1MHz
: 8
.-a,
0.R
I
.
I
-100
.
I
-50
.
I
.
I
0
.
50
8
100
ac
\
o
I 150
-150
-100
-50
0
50
100
150
Temperature ("C)
Temperature ("C)
-
.
- 0.2
4000
D -150
- 0.6 - 0.4
8000
c
0.8
.
16000
a
g 12000
c
: ._ 0
8000
c L
8 .-a, n
4000
o -150
-100
-50
0
100
50
150
Temperature ("C)
Figure 1. Temperature dependence of the dielectric constant and tan& for BalxSrxTi03single crystals, with (a) x=0.20, (b) x=0.30 and (c) x=0.40, measured during cooling runs.
116
Dielectric Materials and Devices
, .
,
. ,
.
,
.
,
.
I
r20.0
.
.
0.002
.
I
I
.
I
.
I
.
I
.
12.5
Y h
- 15.0
E
2
. Y
- 10.0
v
-
0 .-C .-N
.
- 5.0 2 0 .
I
-150
.
I
-100
.
1
-50
.
I
0
.
I
\
I
0.0021
.
-150
150
100
50
Temperature ("C)
P
Io.o
.
.
.-
I
.
I
-100
.
.
-50
0
50
100
150
Temperature ("C) 1
10.0
Y
E
'
-150
-100
-50
0
50
100
150
Tern perat ure ("C)
Figure 2. Temperature dependence of the polarization and pyroelectric coefficient for Bal,Sr,TiOs single crystals, with (a) x=0.20, (b) x=0.30 and (c) x=0.40. the unpoled sample) is different. Even being poled under same poling conditions and along same (as grown) direction and phase symmetry, i.e. tetragonal phase (T=T,-30°C), the effective degree of poling is different. It can be seen that the polarization in the tetragonal phase is much more favorable in the case of (B%.80Sr0.20)Ti03. Comparison between temperature dependence of dielectric constant and remnant polarization is shown in Figure 3. Steps in the polarization curves match the temperature of transformations of BST crystals. The residual remnant polarization, above the temperature of maximum dielectric constant, can be an indicative of phase distribution around nominal phase or due to the result of builtin charges generated at cracks and other defects. Hysteresis loop data at room temperature, displayed in figure 4, showed linear dielectric behavior for x=0.40 (zero remnant polarization, zero coercive field), and P,= 5.6pC/cm2 and 2.55.6pC/cm2 for x=0.20 and 0.30, respectively.
Dielectric Materials and Devices
117
12000
.
,
.
I
.
I
.
,
.
.
I
c
$
. 5 2
15.0
0
0 .& moo
a, 5 -150
-100
-50
0
100
50
-150
,
.
,
.
,
.
I
.
I 15.0
c 0 .c L
N .-
-
0
5a
a, ._
n I
,
-150
,
I
-100
I
-50
I
0
8'.
50
.
100
10.0 150
Tern perature ("C)
Temperature ("C) 16000r
.
0
K 0
0.0 150
'
. ,
c
'Z
5.0
,
c
v
10.0
0
.
I
c
20.048000
0 c
.
12000
25.0
c
.
I
-100
,
,
-50
.
,
0
,
,
50
.
,
100
I10.0
150
Temperature ("C)
Figure 3. Comparative temperature dependence of polarization and dielectric constant for Bal,Sr,TiO3 single crystals, with (a) x=0.20, (b) x=0.30 and (c) x=0.40. Heating runs. For x=0.30, the polarization is close to that obtained from pyroelectric measurements at same temperature, P,=3.5pC/cm2 (figure 2b), but for x=0.20 this value is much lower, Pr=12.5yC/cm2 ( figure 2a). This could be related to the features of domain switching in this particular composition that has <110> as growth direction, allowing the major difference between static and dynamic measurements. Also the loop is not fully saturated in figure 4, thus the observed differences are obsvious. For comparison, two material figures of merit for pyroelectric sensors, the voltage responsivity F, and specific detectivity FD, were plotted as a function of temperature in a range closer to the room temperature and the ferroelectricparaelectric phase transition, as shown in figure 5. The two figures of merit are defined as (5): and
118
Dielectric Materials and Devices
Figure 4. Polarization versus electric field curve loops for Bal,Sr,Ti03 single crystals, with (a) x=0.20, (b) x=0.30 and (c) x=0.40, at room temperature. where the value 3.2x106J/m3K, for the specific heat capacity (c’) was taken from reference 13. It can be seen from figure 5 that figures of merit decrease drastically with increase in Sr content. High dielectric constant with low pyroelectric coefficient lead the compositions x=0.30 and 0.40 to lower values of Fv and FD at room temperature. The single crystal with x=0.20 shows figures of merit on the order of those found in undoped BST ceramics of the corresponding composition (5). Further work is desirable to repeat the experiments for different crystal orientations and under different poling conditions. CONCLUSIONS Three peaks associated with the structural phase transitions are observed in the temperature dependence of pyroelectric coefficient of BST single crystals, between - 130°C and 13OoC. Pr and p trends, as a function of temperature, have strong influence of the crystal orientation and poling conditions. The magnitude of the pyroelectric coefficients decreases with Sr concentration.
Dielectric Materials and Devices
119
40
-
P
-
0 2 3.0 E
UP
x
I
2.0 h
-
a,
n 1.0
P
a, (0
0.0
Temperature ("C)
Temperature ("C)
Temperature ("C)
Figure 5 . Temperature dependence of figures of merit for Bal.Sr.TiO3 crystals, with (a) x=0.20, (b) x=0.30 and (c) x=0.40.
single
Figures of merit around room temperature decrease drastically with Sr content (up to x=0.40) due to the higher dielectric constants of the compositions and for Tc -RT.
ACKNOWLEDGEMENTS This work was supported by a grant from DARPA, under contract no. DABT63-98-1-002. D. Garcia acknowledges the support of FAPESP (Brazilian agency). The authors are grateful also to Dr. Petr Hana for helping in the hysteresis loop measurements.
120
Dielectric Materials and Devices
REFERENCES (1) R.W. Whatmore, P.C. Osbond, and N.M. Shorrocks, Ferroelectrics, 76, 351(1987). ( 2 ) C. Hanson and H. Beratan, in Proceedings of the Ninth IEEE International Symposium on Applications of Ferroelectrics, University Park, PA, pp.657-66 1 (1994). (3) R. Watton and M.A. Todd, Ferroelectrics, 118, 279 (1991). (4) Y.H. Jun, T.-Y. Kim and H.M. Jang, Ferroelectrics, 193, 109 (1997). (5) Idem, Ibidem, 125 (1997). (6) K.-H. Hellwegge, A.M. Hellwegge, ed., “Numerical Data and Functional Relationship in Science and Technology New Series”, vo1.3 (Ferro- and AntiFerroelectric Substances), Spring-Verlag, Berlin-Heidelberg,-New York, 1969. (7) D. Garcia, R. Guo, and A. S. Bhalla, Materials Letters, 42, 136 (2000). (8) D. Garcia, R. Guo, and A.S. Bhalla, in Electronic Ceramic Materials and Devices; Ceramic Transactions, vol. 106, 175 (2000). (9) J. Yamamoto and A.S.Bhalla, Mater. Res. Bull., 24,761 (1989). (lO)R.L. Byer and C.B. Roundy, Ferroelectrics, 3, 333 (1972). (1 l)F. Jona and G. Shirane, “Ferroelectric Crystals”, Dover Publications, Inc., New York, 1993. (12)J.-W. Liou and B.S. Chiou, J. Am. Ceram, Soc., 80[12] 3093-99 (1997). (13)B.M. Kulwicki, A. Amin, H. R. Beratan, and C.M. Hanson, Proceedings of the Eighth E E E International Symposium on Applications of Ferroelectrics, Greenville, SC, pp. 1-10 (1992).
Dielectric Materials and Devices
121
This page intentionally left blank
INFLUENCE OF CRYSTALLIZATION ON STRUCTURAL AND ELECTRICAL PROPERTIES OF PZT THIN FILMS E.B. Araujo, D. Garcia and J.A. Eiras Universidade Federal de Sgo Carlos, Departamento de Fisica, Grupo de Cerimicas Ferroeletricas, Caixa Postal 676, 13565-670 Sgo Carlos SP - Brazil ABSTRACT This work reports studies on crystallinity, dielectric and ferroelectric properties of the PZT thin films prepared by oxide precursor method and crystallized by conventional and RTA method. For film deposited on Si substrate the measured degree of tetragonality (CThT) were 1.04 and 1.01 for film crystallized by conventional and RTA method, respectively. Dielectric constant of film crystallized by RTA at all frequency range was greater than values of dielectric constant of films crystallized by conventional furnace. Ferroelectric properties of the film crystallized by RTA presented a higher P, and lower E, than film crystallized by conventional electric furnace. . INTRODUCTION Lead zirconate titanate, PbZrxTi1-x03(PZT), is probably the most studied perovskite-type ferroelectric material as bulk ceramics as well thin films. PZT solid solution ceramics are well known by excellent piezoelectric, dielectric and pyroelectric properties 111. There are many different deposition methods of PZT thin films such as DC planar magnetron sputtering 121, RF sputtering 131, sol gel [4] and others. Among various techniques to prepare ferroelectric thin films, chemical based processes are promising routes for integration of thin layer devices. Solution deposition enable better stoichiometric control of complex compositions than other physical techniques such as RF sputtering, laser ablation [S] or chemical vapour deposition (CVD) [6]. Recently, we proposed a hybrid chemical method for preparation of ferroelectric thin films based on oxide precursors [7]. The oxide precursor method is an alternative chemical method for preparation of ferroelectric thin films starting on a pre-calcination of oxides or carbonates. Described method was applied initially to prepare PZT thin films of good quality. Next, dielectric and To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
123
ferroelectric properties of PZT films were studied showing that some optimizations are still necessary to obtain better properties for fbrther devices applications [8]. In the present work, was studied the effect of rapid thermal annealing on structural, dielectric and ferroelectric properties of PZT thin films prepared by oxide precursor method. Films were crystallized using conventional electric firnace and rapid thermal annealing (RTA) and were studied by X-Ray diffraction (XRD) analysis, dielectric and ferroelectric measurements. EXPERIMENTAL PROCEDURE PZT thin films used in this work were prepared with a Zr/Ti molar ratio of 53/47. At this phase boundary most of the properties, such as dielectric constant, piezoelectric coefficients, electromechanical coupling coefficient and others, show their maximum values [l]. Using the method that was described in our recent paper [7], films of polymeric resin were deposited at room temperature on Si and Pt/Si substrates by spin coating. Boiling in organic solvents cleaned substrates. Organic contaminants were removed by ultrasound. Finally, substrates were rinsed in distilled water and blown dry with nitrogen gas. Polymeric resin was spun on Si and Pt/Si substrates at a rotation speed of 5000 rpm for 50 s and 7000 rpm for 40 s, respectively. The resin films were then fired at 350°C on a hot plate to remove residual solvents and organic. The deposition and heat-treatment procedure was repeated to obtain a thicker coating before heat treatment for crystallization. Films were crystallized using conventional furnace and RTA process. Using conventional furnace, films were crystallized at 700°C for 1-2 hours and RTA process at 700°C for 60 s. Crystallized films were crack-free, uniform and adhered well on both substrates. The average thickness of films was estimated to be around 550 nm. The crystallographic structure of the films was examined by XRD analysis, using CuK, radiation at room temperature. The dielectric and ferroelectric measurements were conducted in Metal-Ferroelectric-Metal (MFM) configuration. For this purpose, several electrodes of gold (0.3 mm in diameter) were deposited over an area of 1 cm2 on the films through a mask to form MFM capacitors. The dielectric constant and dissipation factor were measured using a H p 4 194A impedance analyzer. To measure dielectric constant and dissipation factor a small ac signal of 10 mV amplitude was applied across the sample while the frequency was swept from 100 Hz to 10 MHz. The small electric field (- 0.18 kV/cm) used for dielectric measurements was considerable less than the coercive field of the PZT such that polarization state remain unchanged and effects of the domain wall contribution are minimized. The ferroelectric properties include measurements of P-E hysteresis loops obtained at a frequency of 100 Hz. These properties were measured using a Sawyer-Tower circuit attached to a Tektronix
124
Dielectric Materials and Devices
2232 digital oscilloscope. All measurements were conducted at room temperature.
RESULTS AND DISCUSSION Figure l-A shows the XRD pattern of PZT film on silicon substrate, fired at 700°C for 2 hours using a conventional furnace. In this figure, we can identify the co-existence of the tetragonal and rhombohedral phases and the presence of undesired pyrochlore phase. The lattice constants a T and CT to PZT were calculated for tetragonal phase using the (100)~,(101)~, (1 1 0 ) and ~ (210)~peaks and were found to be 4.02 and 4.18 4 respectively. The axial c T / a T indicates the degree of tetragonality. With basis on previous lattice parameters the obtained degree of tetragonality was c T / a T E 1.04. This c T / a T ratio is very close to those reported for bulk PZT ceramics with the same composition [l] and PZT thin films produced by DC magnetron sputtering [2]. For the rhombohedral phase were used the ( 1 0 1 ) ~(1 , 1 1 ) and ~ (200)R peaks and obtained aR = 4.08 A and 90-a = 0.024'.
20
30
40
50
60
28 (Degree)
Figure I : X-ray diffraction of the PZT thin film crystallized at 700°C for 2 hours (A) and at 700°C for 60 seconds (B). Films were deposited on Si substrate. Figure l-B shows the XRD pattern of PZT film annealed at 700°C for 60 seconds using RTA. Tetragonal and rhombohedral phases were also crystallized such as observed for film annealed in conventional furnace. The lattice constants for PZT tetragonal phase were a T = 4.04 and CT = 4.06 A. For this film the measured degree of tetragonality was cT/aT z 1.01. For the rhombohedral phase were obtained aR = 4.05 A and 90-a = 0.016'.
Dielectric Materials and Devices
125
Based on previous results, we can see that for film crystallized using conventional furnace the degree of tetragonality was greater than obtained for films crystallized by RTA. In addition, the volume of tetragonal unit cells (67.55 A3) for film crystallized using conventional hrnace was slightly greater than of the film crystallized using RTA (66.27 A3). With respect to rhombohedral phase, the distortion was smaller for films crystallized by RTA. Therefore, films crystallized by RTA method leading to a compact tetragonal cell and a small distortion on rhombohedral phase. The PZT crystallization appeared to be more effective with RTA process that may be attributed to the fast rise in temperature and reaching the equilibrium. Advantages of RTA over conventional furnace annealing are that its rise time for heating to the desired temperature is very short as well as the briefness of the overall annealing period. This leads to reduction in surface damage and minimization of the film-substrate interaction. In specific case of this work, using RTA process was possible the crystallization of PZT without pyrochlore phase. Advantages of RTA process were also observed on PZT thin films processed by metallo-organic decomposition [9].
Figure 2: X-ray diffraction of the PZT thin film crystallized at 700°C for 1 hour (A) and at 700°C for 60 seconds (B). Films were deposited on Si/Pt substrate. The structure of the films deposited on Pt/Si substrate was also studied using XRD. Figure 2 shows XRD patters of the PZT films crystallized at 700°C for 1 hour (Figure 2-A) and 700°C for 60 seconds (Figure 2-B), respectively by conventional and RTA method. The lattice constants aT and CT of PZT were calculated for tetragonal phase considering the (100)~,(101)~and (200)~peaks in
126
Dielectric Materials and Devices
Figure 2-A. Obtained parameters were aT = 4.05 and CT = 4.08 A (cT/aT E 1.01). This tetragonality c T / a T was slightly small than calculated for films crystallized on Si by conventional furnace. Based on Figure 2-B the lattice constants were calculated. Obtained values were aT = 4.07 and CT = 4.09 A ( C T / ~ T G 1.01). Films deposited on Pt/Si, crystallized by conventional hrnace and RTA, presented essentially same degree of tetragonality. Peaks of rombohedral are unclear in Figure 2. For this reason, peaks were attributed only to tetragonal phase of PZT. The effect of frequency on the dielectric constant (E) and dissipation factor (tan&) is shown in Figure 3 for PZT films crystallized by different ways, crystallized at 700°C for 1 hour (Figure 3-A) and at 700°C for 60 s (Figure 3-B). From Figure 3-A, E and tan6 values at a frequency of 100 kHz were 358 and 0.039, respectively. Considering same frequency in Figure 3-B, these values were 611 and 0.026, respectively. Mainly difference observed in both graphics is associated with dielectric constant and dissipation factor at higher frequencies. Considering film crystallized by conventional hrnace (700°C for 1 hour) the dielectric constant was relatively unchanged up to fiequency about 1 M H z after which it dropped to a small value (145 at 10 MHz). At around the same frequency, the dissipation factor increased substantially up to 0.743. This frequency dispersion is often characterized by a Maxwell-Wagner type and was also reported in other ferroelectric films [101. 1.o C/1 h --t RTA 700C160s
& Conv 700
0.8 600
0 0.6
i!.
5 0
t 0.4
$
h
G
0.2 200 100
100
Ik
10k
100k
1M
10M
0.0
Frequency (Hz)
Figure 3: Dielectric constant and dissipation factor as a hnction of frequency for PZT film deposited on Au/PZT/Pt/Si configuration. Films were annealed by conventional method at 700°C for 1 h (A) and by RTA at 700°C for 60 s (B).
Dielectric Materials and Devices
127
Dielectric constant of the film processed by RTA was relatively unchanged over all measured frequency range (Figure 3-B). This film did not showed dispersion at high frequency, such as observed for film crystallized by conventional furnace. This fact may be associated with better crystallization in films prepared by RTA. Similar result was reported in other PZT films [9]. Ferroelectric properties of PZT films were obtained from hysteresis loops for films crystallized on Pt/Si substrates using conventional and RTA method. Figure 4-A and 4-B shows hysteresis loops for film crystallized by conventional method at 700°C for 1 hour and using RTA at 700°C for 60 seconds, respectively.
Figure 4: Hysteresis loops of the PZT thin films measured at 100 Hz. Film crystallized by conventional method at 700°C for 1 hour (A) and crystallized by RTA method at 700°C for 60 seconds (B).Films were deposited on Pt/Si substrate For film crystallized by conventional method (Figure 4-A) the remanent polarization (Pr) was about 7.8 pC/cm2 and the coercive field (Ec) of 99 kV/cm. On the other hand, for PZT film crystallized by RTA method (Figure 4-B),the remanent polarization and coercive field were about 15.7 pC/cm2 and 73 kV/cm, respectively. For PZT films the values of Pr and E, ranged from 1-7 pC/cm2 and 26-80 kV/cm, respectively, obtained by sol-gel [4] or 3-30 pC/cm2 and 25-64 kV/cm, for films obtained by dc magnetron sputtering [2]. As we can see, the remanent polarization presented by film crystallized by RTA method is almost two times greater than obtained for film crystallized by conventional method while the coercive field was reduced to about 2/3 in magnitude. There are many works about ferroelectric properties as a function of grain size for different
128
Dielectric Materials and Devices
ceramic materials [ 11,121. In general, the observed slightly lower P, and higher E, may be associated with smaller grain size in comparison with bulk ceramics [l]. In specific case of this work, ferroelectric values presented for film crystallized by RTA method, in comparison with film crystallized by conventional method, were higher P, and lower E,. Such improved ferroelectric behavior may be attributed to improved density, crystallization and smoothness of the PZT films prepared by RTA method. CONCLUSIONS PZT films were deposited by oxide precursor method and crystallized by conventional and RTA method. Studies on crystallinity, dielectric and ferroelectric properties were carried out to understanding the effect of crystallization method on these properties. Films crystallized by RTA method presented better dielectric and ferroelectric properties (higher P, and lower E,) than films crystallized by conventional heating in electric furnace. These data will be fundamental to a future optimization of ferroelectric properties of these films. ACKNOWLEDGMENTS The authors are grateful to Dr. Y.P. Mascarenhas (IFSC-USP), by XRD facilities, to Mr. Francisco J. Picon (DF-UFSCar) by technical support, and to CNPq and FAPESP (process 99/02485-2) for financial support. REFERENCES 1. B. Jaffe, W.R. Cook and H. Jaffe; Piezoelectric Ceramics (Academic, New York, 1971). 2. K. Sreenivas and M. Sayer, J. Appl. Phys. 64, 1484 (1988). 3 . K. Sreenivas, M. Sayer and P. Garret, Thin Solid Films, 172,251 (1989). 4 . G. Yi, Z. Wu and M. Sayer, J. Appl. Phys. 64,2717 (1988). 5 . 0. Auciello, L. Mantese, J. Duarte, X. Chen, S.H. Rou, A.I. Kingon, A.F. Schreiner and A.R. Krauss, J. Appl. Phys. 73, 5197 (1993). 6 . M. de Keijser and G. J. M. Dormans; MRS Bulletin, p. 37, June 1996. 7 . E.B. Araujo and J.A. Eiras, J. Mat. Sci. Letters 17, 833 (1998). 8. E.B. Araujo and J.A. Eiras, J. Phys.: Cond. Matter 1 1, 1975 (1999). 9. H. Hu, L. Shi, V. Kumar and S.B. Krupanidhi, Ceramic Transactions, Ferroelectric films, Edited by A.S. Bhalla and K.M. Nair, volume 25, 113 (1992). 10. P.C. Joshi and S.B. Krupanidhi, J. Appl. Phys. 72, 5827 (1992). 11. K. Okazaki and K. Nagata, J. Am. Ceram. Soc. 56,82 (1973). 12. L. Mitoseriu, D. Ricinschi, C. Hamagea, M. Okuyama, T. Tsukamoto and V. Tura, Jpn. J. Appl. Phys. 35, Part 1 9B, 5210 (1996).
Dielectric Materials and Devices
129
This page intentionally left blank
CRYSTALLIZATION OF STRONTIUM BARIUM NIOBATE FERROELECTRIC THIN FILMS PRODUCED BY CHEMICAL METHOD E.B. Araujo, R.G. Mendes, D. Garcia*and J.A. Eiras Universidade Federal de Siio Carlos, Departamento de Fisica, Grupo de Ceriimicas Ferroeletricas, Caixa Postal 676, 13565-670 Siio Carlos SP - Brazil ABSTRACT Strontium barium niobate (SBN) thin films of good quality were deposited on Pt/Si substrates using a polymeric resin containing metallic ions. Films were crystallized by conventional electric furnace and by rapid thermal annealing (RTA) at different temperatures. Using X-ray diffraction, was identified the presence of polycrystalline SBN phase for films annealed from 500°C to 700°C in both cases. Undesired phases such as SrNb2O6 and BaNbz06 were predominantly crystallized in films annealed at 500°C. These phases disappear at higher temperatures. Dielectric and ferroelectric parameters obtained from films crystallized by conventional furnace and RTA presented essentially same values. INTRODUCTION The tungsten bronze (TB) family is one of several ferroelectric materials. Since 1949, when the tungsten bronze structure was deduced by Magneli [l], numerous tungsten bronzes have been synthesized. The tungsten bronze family includes niobates such as (Sr,Ba)Nb206 (SBN), (Pb,Ba)Nb& (PBN) and (Pb,K)Nb206 (PKN). The PBN and PKN are orthorhombic tungsten bronze structure with a point group mm2. On the other hand, SBN (Sr,Bal-,Nb206), with 0.25<x<0.75, presents a tetragonal (4mm) phase at room temperature. A solid solution of SBN exists in the binary SrNb206-BaNb206 system. Investigations using X-ray diffraction suggested a morphotropic phase boundary (MPB) around x = 0.25, which is characterized by the coexistence of the tetragonal and orthorhombic phases [ 2 ] .
* Actual temporary address: Materials Research Laboratory, The Pennsylvania State University, University Park, PA, 16802.
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
131
The tetragonal SBN structure consists of a framework of NbO6 octahedra sharing corners in such a way that three types of interstitial site result. Figure 1 shows a view of SBN structure along the polar axis presented by Jamieson et al. [3]. Niobium atoms occupy the B1 and B2 sites of the nominal (A1)2(AZ)4C4(B1)2(B2)8030 TB structure. The strontium atom in the A1 site has 12 nearest oxygen atom in distorted cubo-octahedral coordination while bariudstrontium atom in the A2 site is surrounding by nine nearest-neighbor oxygen atoms and this arrangement may be described as in distorted tricapped trigonal prismatic coordination [3]. The structure devised by Jamieson et al. is still accepted for SBN, although there has always been some questions as to whether or not barium is found only on the A2 site. Recent investigations show that barium was found only at the A2 site while strontium occupies the A1 and A2 sites [4].
Figure 1: View of strontium barium niobate structure along the polar c axis. After Jamieson et al. [3]. The excellent ferroelectric and electro-optic properties exhibited by SBN make this material promising for a variety of applications. SBN has received great attention as a ferroelectric material due to its large pyroelectric coefficient [5], piezoelectric [6] and electro-optic properties [7]. In recent years, the development of integrated optical devices has stimulated the demand for thin films using attractive materials such as SBN. SBN thin films have been prepared by several techniques like sol-gel processing [S],
132
Dielectric Materials and Devices
pulsed laser deposition [91 and metalorganic chemical vapor deposition (MOCVD) [ 101. Large-scale processing of high-quality thin films requires lowtemperature synthesis, high reproducibility and simplicity in all processing steps at low cost. Due to this fact, the search for new routes for film preparation remains as an interesting and open subject in order to improve the stability of complex solutions, the control of the stoichiometry of the film composition or to reduce the cost of the process. Very recently, was proposed the preparation of SBN thin films by a chemical method based on a polymeric resin containing metallic ions [l 11. The method was successfblly applied to prepare SBN thin films of good quality and homogeneity. In this work, SBN thin films obtained by early cited method were studied under its structural, dielectric and ferroelectric properties. The effect of the different heat treatment on SBN thin films, using conventional fbrnace and rapid thermal annealing (RTA), was studied. Dielectric and ferroelectric properties were presented for films crystallized by different ways. EXPERIMENTAL PROCEDURE In this work SBN films were prepared by a hybrid chemical method, used to produce a polymeric resin [ 113. The general idea is to distribute the metallic ions homogeneously throughout the polymeric resin, prepared according to the Pechini method [121. The process calls for forming a chelate between dissolved ions with a hydroxycarboxylic acid (citric acid). Heating of the resin in air causes a breakdown of the polymer. Subsequently, the ions are oxidized to form the desired crystalline phases. Barium carbonate (BaC03), strontium carbonate (SrCO3) and ammoniac complex (NH4H2NbO(c204).3H20) were selected as starting materials. The molar ratio of starting materials was calculated to obtain a final Sr0.75Ba0.25Nb206(SBN 75/25) phase. For preparation of the resin, BaC03, SrC03 and NH&t2NbO(C204).3H20 were initially dissolved in water to form a transparent solution. Next, each solution was mixed separately with citric acid and heated to 40°C for 30 minutes. This stage is important to form chelate between mixed cations with a hydroxycarboxylic acid (citric acid). Then, each solution was mixed with ethylene glycol (citric acid/ethylene glycol = 50/50) and polymerized by heating up to 100°C for 30 minutes. Finally, the three solutions were mixed at room temperature and heated again to 50°C and stirred during 20 minutes to homogenization, when a transparent resin was obtained. The final transparent resin indicates that all metallic ions were distributed throughout the polymeric resin. The viscosity of the final resin was controlled with ethylic alcohol. Films of the resins were deposited at room temperature on Pt/Si substrates by spin coating at 4000 rpm for 40 seconds. Films were obtained by depositing multiple layers of this resin. Each layer was annealed at 400°C for one hour, to ~
Dielectric Materials and Devices
133
remove residual solvents. The process was repeated for eight deposited layers to obtain a film with 0.32 pm thickness in average. Deposited films were crack-free, uniform and well adhered on substrates. For crystallization films deposited on Pt/Si substrates were annealed in conventional electric hrnace at 5OO0C, 600°C and 700°C during one hour. The rapid thermal crystallization was carried out in a Research INC power controller Model 664F rapid thermal annealing (RTA) fbmace at 500"C, 600°C and 700°C for 60 seconds. The energy source was comprised of several tungsten lamps positioned on the lower side of a quartz isolation-heating chamber. The structure of the crystallized films at different temperatures was analyzed by X-ray diffraction (XRD) using CuK, radiation at room temperature. The electrical properties include dielectric and ferroelectric measurements, which were conducted in metal-ferroelectric-metal (MFM) configuration. To measure the electric properties, several electrodes of gold (0.3 mm in diameter) were deposited over an area of 1 cm2 on the films through a mask to form MFM capacitors. The dielectric constant (E) and dissipation factor (tan6) were measured from 500 Hz to 500 kHz frequency using a HP 4194A impedance analyzer. A Sawyer-Tower circuit at 100 Hz was used to measure the ferroelectric properties. RESULTS AND DISCUSSION The effect of temperature crystallization in SBN films was investigated under different conditions. Figure 2 shows XRD patterns of SBN films on Pt/Si substrate for different temperatures. Peaks identified in this figure were attributed to tetragonal SBN phase. As show Figure 2, polycrystalline SBN phase better crystallizes on films annealed at 600" and 700°C for 1 hour. Some structural fluctuation can be seen in Figure 2 for films annealed at 5OO0C, when observed the well defined (211) and (400) peaks. These peaks disappear at films crystallized at 600°C and 700°C. Based on XRD patterns, changes of f i l l width at half-maximum (FWHM) for some (hkl) peak represents the degree of crystallization as a function of annealing temperature. The FWHMs of (3 11) peak in Figure 2 were 0.72O, 0.63" and 0.57" for films respectively annealed at 500"C, 600°C and 700°C for 1 hour. These facts show that when annealing temperature increases the crystallization of SBN films on Pt/Si is improved. In Figure 2 we can also identify the .presence of the SrNb206 (SN) and BaNb2O6 (BN) phases for films annealed at 500°C for 1 hour. When temperature increases, BN disappear but SN remains at film annealed at 600°C and 700°C for 1 hour. Neither SN nor BN is a ferroelectric material with the tungsten bronze structure. They are components of the solid solution that yields ferroelectric SBN. BN crystallizes in both hexagonal and orthorhombic forms while SN appears to be isostructural with CaTa206 [3].
134
Dielectric Materials and Devices
Figure 2: X-ray diffraction patterns of SBN thin films deposited on R/Si substrate. Films annealed at different temperatures in electric furnace. Unidentified phase (x), SrNb2O6 (0)and BaNb206 (A).
Figure 3: X-ray diffraction patterns of SBN thin films deposited on R/Si substrate. Films annealed at different temperatures, using rapid thermal annealing (RTA), for 60 seconds. Unidentified phase (x), SrNb2O6 (0)and BaNb206 (A). Figure 3 shows XRD patterns of the SBN films crystallized by rapid thermal annealing (RTA), at different annealing temperatures for a constant annealing time of 60 seconds. The FWHMs of the (3 11) peak were 0.60", 0.57" and 0.57", indicating similar behavior of SBN film crystallized by conventional
Dielectric Materials and Devices
135
method. As observed in Figure 2, BN and SN phases are also present at films obtained by RTA process, but only for films annealed at 500°C and 600°C for 60 seconds, SN and BN phases disappear completely when temperature increases up to 700°C. Based on XRD patterns of SBN films crystallized at 700°C by conventional and RTA method, the lattice constants a and c were calculated for tetragonal phase using the (OOl), (311), (002) and (322) peaks. For the film annealed by conventional method, the calculated lattice constants were a = 11.57 A and c = 4.00 while a = 11.62 A and c = 4.03 A for film crystallized by RTA. The lattice parameters a and c obtained here agreed relatively well with results of the literature for SBN 75/25 thin film obtained by metal-organic chemical vapor deposition (MOCVD), whose values are around a = 12.430 A and c = 3.932 A [lO]. For bulk crystals with same composition the lattice parameters are a = 12.458 A and c = 3.928 A [13]. The slightly smaller a and greater c parameters obtained here for SBN films may be attributed to elastic strain generated along the a and c-axis in SBN films. Films annealed at 700°C by conventional method (for 1 hour) and RTA (for 60 seconds) were used for dielectric and ferroelectric measurements. The dielectric behavior of the SBN films, examined in terms of the dielectric constant and dissipation factor as hnctions of measuring frequency, were summarized in Figure 4.
-(E)
(tan 6) - RTA
0.8
* 200 h
c
150
Figure 4: Dielectric constant and dissipation factor for SBN thin films crystallized by conventional and RTA method. It may be seen that dielectric constant exhibits a slight frequency dependency, which is consistent with the expected normal behavior. The dielectric constant and dissipation factor at a frequency of 100 kHz were 119 and 0.164,
136
Dielectric Materials and Devices
respectively for film crystallized by conventional method at 700°C for 1 hour. For film crystallized using RTA method these values were respectively 107 and 0.088. The value of the dielectric constant obtained here is consistent with other reports for SBN thin films obtained thought metal alkoxide [14]. The frequency dependence for lower frequencies of the dielectric constant and dissipation factor is probably related to the presence of an interfacial surface, at the interface electrodes/film, which results in an undesirable Maxwell-Wagner type dispersion in the dielectric data. In addition, the dielectric constant value on different electrodes varied less than 3%, indicating a good degree of uniformity in the thickness of the films. Hysteresis loops were observed at room temperature, at 100 Hz frequency. Figure 5 shows P-E hysteresis loop of SBN films crystallized by conventional furnace at 700°C for 1 hour and by RTA at 700°C for 60 seconds. Rounded curves were found due to the low frequency high loss characteristics of the films.
Figure 5: Hysteresis loop, measured at 100 Hz, of SBN film annealed by conventional furnace at 700°C for 1 hour and rapid thermal annealing (RTA) at 700°C for 60 seconds. The remanent polarization (Pr) and the coercive field (Ec) were 17.8 pC/cm2 and 180 kV/cm, respectively, for the SBN film annealed in conventional hrnace at 700"C/lhour. Similar results were obtained for films crystallized by RTA method. This film presented Pr = 16 pC/cm2 and E, = 155 kV/cm. However, the coercive field (Ec) and remanent polarization (Pr) values obtained here for SBN thin films may be not correspond to absolute values because conductivity effect was not discounted in loop hysteresis of the Figure 5. Thus, the distortion observed on hysteresis is probably associated with effect of conductivity in SBN
Dielectric Materials and Devices
137
film. In literature, reported values of P, and E, for SBN 75/25 films ranged from 1.9-34 pC/cm2 and 5 1- 180 kV/cm, respectively, obtained by sol-gel [ 15,141. In this work, the observed P, is slightly lower if compared with values observed for crystals, whose values ranged from 27 to 32 pC/cm2 [ 131. CONCLUSION Polycrystalline SBN thin films of good quality were crystallized on Pt/Si substrates using a polymeric resin. Results showed good crystallization of SBN tetragonal phase when films were crystallized using conventional fbrnace or RTA. Undesired phases such as SrNb2O6 and BaNb206 appear in films deposited at 500°C but these phases disappear at higher temperatures. Finally, dielectric and ferroelectric parameters presented essentially same values for films crystallized by conventional fbrnace and RTA. ACKNOWLEDGEMENTS The authors are gratefbl to CNPq and FAPESP for financial support, to Dr. Yvonne P. Mascarenhas (IFSC-USP) for XRD facilities and to Mr. Francisco J. Picon (DF-UFSCar) by technical support. REFERENCES 1. A. Magneli, Ark. Kemi. 1, 213 (1949). 2. S. Nishiwaki, J. Takahashi, K. Kodaira and M. Kishi, “Morphotropic phase boundary and dielectric properties of V2Os-containing SrxBal-&b206 (0.21fi0.4) ferroelectrics”, Jpn. J. Appl. Phys. 35, 5 137 (1 996). 3 . P.B. Jamieson, S.C. Abrahams and J.L. Bernstein, “Ferroelectric tungsten bronze type crystal structures. I. Barium Strontium Niobate Ba0.27Sr0.751\Jb205.7~”,J. Chem. Phys. 48, 4352 (1968). 4. M.P. Trubelja, E. Ryba and D.K. Smith, “A study of positional disorder in strontium barium niobate”, J. Mat. Science 3 1, 1435 (1996). 5 A.M. Glass, “Investigation of the electrical properties of Srl-,BaXNbzO6with special reference of pyroelectric detection”, J. Appl. Phys. 40,4699 (1969). 6 . .J.D. Zook and S.T. Liu, “Tyroelectric effects in thin film”, J. Appl. Phys. 49, 4604 (1978). 7 M: Horowitz, A. Bekker and B. Fischer, “Broadband second-harmonic generation in SrxBal -xNb206 by spread spectrum phase matching with controllable domain gratings”, Appl. Phys. Lett. 62, 26 19 (1 993). 8 Y. Xu, C.J. Chen, R. Xu and J.D. Mackenzie, “Ferroelectric Sro.6oBa0.&b206 thin films by the sol-gel process: Electrical and optical properties”, Phys. Review B, 44(1), 35 (1991). 9 S.S. Thony, K.E. Youden, J.S. Harris Jr. And L. Hesselink, “Growth of
138
Dielectric Materials and Devices
epitaxial strontium barium noibate thin films by pulsed laser deposition”, Appl. Phys. Lett. 65( 16), 20 18 (1994). 10. M. Lee and R.S. Feigelson, “Growth of epitaxial strontium barium niobate thin films by solid source metal-organic chemical vapor deposition”, J. Crystal Growth, 180,220 (1997). 11. R.G. Mendes, E.B. Araujo, H. Klein and J.A. Eiras, “Synthesis of strontium barium niobate ferroelectric thin films by na alternative chemical method”, J. Mat. Science Letters 18, 1941 (1999). 12. P.A. Lessing, “Mixed-cation oxide powders via polymeric precursors”, Ceramic Bulletin, 68(5), 1002 (1989). 13. R.R. Neurgaonkar, W.F. Hall, J.R. Oliver W.W. Ho and W.K. Cory, “Tungsten bronze Srl-,Ba,NbzOb: a case history of versatility”, Ferroelectrics 87, 167 (1988). 14. W. Sakamoto, T. Yogo, K. Kikuta, K. Ogiso, A. Kawase and S. Hirano, “Synthesis of strontium barium niobate thin films through metal alkoxide”, J. Am. Ceram. Soc. 79(9), 2283 (1996). 15. C.J. Chen, Y. Xu, R. Xu and J.D. Mackenzie, “Ferroelectric and pyroelectric properties of strontim barium niobate films prepared by the sol-gel method”, J. Appl. Phys. 69(3), 1763 (1991).
Dielectric Materials and Devices
139
This page intentionally left blank
Structural and Dielectric Characterizationof Amorphous SrTiOs Thin Film Prepared by sol-gel
E.R. Leite(*), F. M. Pontes, S.M. Zanetti, E.Longo LlEC - Department of Chemistry,UFSCar - Federal University of Siio Carlos, CEP 13565-905, SBo Carlos SPYBrazil
J. A Varela IQ - UNESP - Araraquara, SP - Brazil
V. Mastellaro IFSC-USP - SBo Carlos, SP - Brazil
Abstract
Amorphous SrTiOs thin films were processed by chemical solution method. Dielectric and ferroelectric characterization showed that these films present P-E hysteresis loop typical of ferroelectric materials. The P-E hysteresis loop properties are moMied by post annealing treatment due to the modlfication of the dielectric constant of the amorphous matrix. The ferroelectric-likeproperties observed are ascribed to the presence of a small concentrationof Ti06 octahedra (< 20%), identified by XANES (X-
ray Absorption Near Edge Structure)measurements.
(*) e-mail derl@,power.ufscar.br _-
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
141
1. Introduction
The extensive research and development in ferroelectric thin films for nonvolatile random access memory (NVRAM) and high density dynamic random access memory (DRAM) have attracted great attention in the electronic industry. Among the ferroelectric phase, materials such as (Pb,La)TiO, (PLT)[ 11, Pb(Zr,Ti)03 (PZT)[2], PbTi03[3], BaTi03 [4] are well known and have been extensively studied. However, ferroelectricity has been known for these materials only in crystalline phase. In the last few years, perovskite oxides such as strontiumtitanate (SrTi03)[5]have been intensively studied due to their structural and dielectric features. SrTiO, is a well-known perovskitestructured material with a cubic paraelectric phase above 105 K and a bulk dielectric constant of about 300 for sintered ceramics. On the other hand, thin dielectric amorphous films have recently attracted a good deal of attention due to their attractive ferroelectric[6-81, and electro-optical properties[9]. In theory, ferroelectricity should be possible in a noncrystalline material, although the dipole-dipole interaction must be weak[ 101. Several studies have reported ferroelectric-likeproperties in amorphous thin films prepared by physical or chemical methods[6-8]. Recently, however, Xu and Mackenzie[113 described a theoretical explanationfor the ferroelectric-likeproperties of amorphous dielectric films. Their model is based on the presence of ordered clusters formed by octahedral BOGin an amorphous matrix. These ordered regions are separated from one another by disordered regions, although dipole-&pole interaction is possible. This paper describes the structural , dielectric and ferroelectric characterization of amorphous SrTiO, tlun film prepared by a chemical route[ 12,131.
142
Dielectric Materials and Devices
2. Experimental Procedure
Thin films were prepared by the sol-gel technique, using the polymeric
precursor method[ 11,121. A polymeric aqueous solution with a 7 nPa.s viscosity was used for spin coating deposition on a Pt/Ti/SiQ/Si substrate. After deposition, the substrates were dried on a hot plate (-150°C) and heat-treated at 400°C for 8 h in an oxygen flow and 600 "C for 2h. The SrTiQ films were characterized in terms of structure using X-ray diffraction (XRD) (Cu K, radiation). The diffraction patterns were recorded on a Siemens D5000 machine in a 8-28 configuration, using a graphite monochromator. Microstructural characterization was performed by atomic force microscopy AFM to obtain a 3D image reconstruction of the sample surface. These images allow for an accurate analysis of the sample surface and the quantification of very important parameters such as roughness and grain size. A Digital Instruments Multi-Mode Nanoscope IIIa was used. To carry out the electrical measurements, 0.3 mm diameter gold electrodes were deposited by sputtering through a designed mask onto the film surfaces to form The dielectric properties were measured as a metal-insulator-metal capacitors 0. h c t i o n of frequency by using a Hewlett-Packard (4194A) impedance/gain phase analyzer, while the polarization-electricfield (P-E) hysteresis loop was measured using a virtual ground mode test device (Radiant Technology RT 6000HVS). The capacitancevoltage (C-V) characteristics were measured for MIM configuration using a small AC signal of lOmV at 100 kHz.
Dielectric Materials and Devices
143
The signal was applied across the sample, while the DC electric field was swept from positive bias to negative bias and back to positive bias (C-V curves). These electrical characterizations were done on the as prepared film,as well as on post-annealed film in an N2 atmosphere at 200°C for 4 hours. Ti K-edge XANES spectra were collected at the facilities of the LNLS (National Laboratory of Synchrotron Light) in Campinas, SP, Brazil. The LNLS storage ring was operated at 1.36 GeV and 60-100 mA. Data were collected at the Ti K-edge (4966 ev) in the total electron yield mode for thin film samples. The beam was monochromatized using an Si(111) channel cut monochromator with an energy step equal to 0.5 eV. Crystals of J3-Ba2Ti04,Ba2TiSi208(fiesnoite) and rTi02 (rutile) powder were used as a structure reference. The Ti atoms in these structures are coordnated by 4, 5 and 6 oxygen atoms, respectively.
3. Results and Discussions
Figure 1 shows the XRD (X-ray diffraction)patterns of the SrTiO3 thin film on pt/Ti/SiO2/Si substrate annealed at 400 "C. It is.clear from Figure 1 that the XRD patterns did not show any peak corresponding to the crystalline SrTiO, phase. Only the peaks of the substrate were observed, which suggests an amorphous nature of the film. Figure 2 shows the surface morphology using atomic force microscopic (AF'M) of the amorphous film on platinum coated silicon substrate. The AFM pictures revealed a smooth, featureless and crack-free surface for an amorphous thin film with a surface roughness close to 0.25 nm. No evidence of grains or crystallites was observed, suggesting an amorphous structure.
144
Dielectric Materials and Devices
20
28
60
Figure 1: X-ray diffraction patterns of SrTi03 amorphous thin films on Pt/Ti/SiOz/Si substrate.
Figure 2: Atomic force microscopy images of the SrTiO3 thin film, amorphous surface.
The analysis realized by XANES study of the SrTi03 thin film crystallized at 6OO0C for 2h compared to SrTi03 amorphous thin film, showed that in the amorphous
film, a pre-edge feature of a Ti K-edge XANES spectra located at approximately 4970 eV can be observed.
Dielectric Materials and Devices
145
XANES was used to compare the SrTi03 thin film crystallizedat 600°C for 2h with an amorphous SrTiQ thin film (heat treated at 400 "C). In the amorphous film, a pre-edge feature of a Ti K-edge XANES spectra located at approximately 4970 eV could be observed. This peak is related to the ls+3d and it is usually attributed to the energy level transition from Ti 1s to the Ti3d02p molecular orbital[l4]. Farges[l4] made an experimental study of the ls-+3d transition as a function of Ti coordmation. Based on the ls+3d energy position and normalized intensity, Forges obtained structural information of the Ti atoms in crystalline and glassy fresnoites[14]. In the present work, information on the Ti atom coordination was obtained by using a similar approach. The results of this analysis are listed in Table I. Most of the Ti atoms in the amorphous thin film present coordination five, with a small number of Ti atoms in coordination six(octahedron).
Samples
ls+3d energy
14'Ti
15'Ti
16'Ti
transition (ev)
146
*Ba2Ti04
4969.2
*Ba2TiSi208
4970.5
100 100
*r-TiQ
4971.3
SrTi03 -amorphous film
4971.0
80-90
10-20
SrTi03-crystallized film
4970.5
<20
>80
100
Dielectric Materials and Devices
Figure 3 shows the dielectric constant and dielectric loss as a function of frequency for the as prepared films and for the films after annealing in an N2 atmosphere. A drop in the dielectric constant at the low frequency range was observed for the as prepared films.
180
160
j-
10’
Amorphous ( i n 0 2 ) -UAmorphous (in N,) -A-
10’
10‘
10’
-m-
-A-
io6
10’
Frequency (Hz)
Figure 3 : Dielectric constant and &sipation factor as a function of applied frequency for SrTiOs amorphousthin films. No frequency dependence was observed for the &electric constant property after the annealing treatment in N2, (Fig.3). However, a decrease in the dielectric constant after this annealing was observed in all the frequency ranges measured Table I1 shows the values of dielectric constant and &electric loss for the amorphous and crystalline thin films (thin films heat treated at 600°C during 2 h). Table 11: Values for amorphous and crystalline SrTiO, thin films measured at 1OOkHz. Sample Crystalline (in ar) Amorphous (in Amorphous (in N2)
a)
Dielectric Constant 250 78 50
Dielectric Materials and Devices
Dissipation Factor 0.01 0.08 0.01
147
Figure 4 shows the P-E hysteresis loop. A typical ferroelectric hysteresis was observed for the as prepared film, with a remanent polarization (Pr) of 1.8 pLc/cm2and a coercive field (Ec) of 77 kV/cm Similar Pr and Ec values were reported for BaTiO, amorphous thin film[6]. A slim hysteresis was observed after the annealing treatment, coupled with an abrupt decrease in the Pr and Ec values.
4
-SrTiO, (amorphous)- 0, SrTiO, (amorphous)- N,
-4
-300
-200
-100
0
100
200
300
Electric Field (kV/cm)
Figure 4: P-E hysteresis loop for the amorphous SrTiO, thm film in different atmospheres Considering the Xu and Mackenzie model[ll], the decrease in the Ec and Pr values after the annealing treatment can be attributed to the decrease in the dielectric constant values. It can be stated that the remanent polarization (Pr) of amorphous ferroelectric film is given by: Pr=Nv,
148
(1)
Dielectric Materials and Devices
where is the average value of the dipole moments in the z direction, N, is the number of dipoles per volume unit and v is the volume fraction of polar clusters. One can, therefore, express as :
where pc is the dipole of the polar cluster, Ec represents the coercive field, k is Boltzmann’s constant, T represents the absolute temperature and E,.(a) indicates the dielectric constant of the amorphous matrix. Equation 1 and 2 demonstrate that a variation in the dielectric constant of the amorphous matrix promotes a change in the value, leading to a variation in Pr. An increase in the Er(a) value would, therefore, result in an increase of the remanent polarization. Figure 5 displays the typical C-V curve for the MIM capacitors. No voltage dependence and no hysteresis was observed for the capacitance property before and after the annealing treatment in N2. This kind of behavior is typical of paraelectric materials[ 16,171. For ferroelectric materials a strong non-linear capacitance dependence on the voltage is expected and this behavior is related to the domain mobility[ 181. The hysteresis loop suggests a ferroelectric like property. However the C-V curve indicates a paraelectricmaterial. Consideringthese experimentalresults, the origin of the hysteresis loop can be also associatedto electrostrictivemechanical coupling with the dielectric constant[l8]. In this case the change in the dimension due to the electric
Dielectric Materials and Devices
149
field (electrostrictive effect ) can be associated to the polarization of the TiO6 octahedra identified by the XANES experiment. After annealing in N2, the decrease of the dielectric constant will result in a decrease in the dunension variation (AMo).
Figure 5 : Capacitance-Voltage ( C-V curve ) characteristics of an amorphous SrTi03
thin film (room temperature) in Merent atmospheres.
Electrical characterization showed that the amorphous SrTi03 thin films presented P-E hysteresis loop. Structural characterization showed that, in the amorphous
thin films, most of the Ti atoms presented coordination five ((Ti0)04)[101, with a small number of Ti atoms in coordination six (octahedron). Combining the results of structural and electrical characterization, it can be postulated that the presented P-E hysteresis loop observed for the amorphous thin film is related to the presence of a small concentration of Ti06 octahedra in an amorphous matrix formed by (TiO)O4[10]. The TiO6 octahedra form ordered clusters (polar regions inside the amorphous matrix). When an electric
150
Dielectric Materials and Devices
field E is applied, the dipole moments of these ordered regions are polarized and preferentially orientated in the field hection.
5. Conclusions
In summary, amorphous SrTiO, thin films prepared by the sol-gel method showed P-E hysteresis loop. This P-E hysteresis loop is modified by post annealing treatment as a result of the modification of the &electric properties of the amorphous matrix. The origin of the P-E hysteresis loop is not yet clear but can be related to electrostrictive effect or even ferroelectric effect, as postulated by Xu and Mackenzie[113. Anyway, the observed P-E hysteresis loop is attributed to the presence of the small concentration of Ti06 octahedra (< 20%) identified by XANES measurements.
ACKNOWLEDGEMENTS
The authors wish to express their appreciation for the financial support of the Brazilian funding agencies FAPESP, CNPq FINEP and PRONEX.
Dielectric Materials and Devices
151
References
1. A. Li, C. Ge, P. Lii, D. Wu, S. Xiong, and N. Ming, “Growth and ferroelectric
properties of sol-gel derived (PbLa)Ti03 films on metallic LaNi03-coated s~bstrates’~, Mater. Lett. 31 (1997) 15. 2. M. Brazier, S. Mansour, and M. McElfresh, “Ferroelectric fatigue of Pb(Zr,Ti)03
thin films measured in atmospheres of varying oxygen concentration”,Awl. Phys. Lett. 74 (26) (1999) 4032. 3. J. H. Kim and F. F. Lange, “Seeded epitaxial growth of PbTi03 thin films on (001)
LaA103 using the chemical solution deposition method”, J. Mater. Res., 14 (4), 1626, (1999). 4.
M-B. Lee, M. Kawasaki, M. Yoshimoto, and H. Koinuma, “Heteroepitaxial growth of BaTi03 films on Si by pulsed laser deposition”, Appl. Phys. Lett., 66 (1l), 1331, (1995).
K. Fukushima, and S. Shibagala, “Nb doped SrTi03 thin films deposited by pulsed laser ablation”,Thin Solid Films, 315, (1998), 238.
Y. X y C.H. Cheng and J.D. Mackenzie, “Electrical characterization of plycrystalline and amorphous thin films of Pb(ZrxTil,)03and BaTi03 prepared by sol-gel technique”,J. Non-Cryst. Solids, 176, 1, (1994). 7.
J.D. Mackenzie and Y. Xu, “FerroelectricMaterials by the Sol-gel Method”, J. SolGel Sci. and Technology, 8, 673, (1997).
8.
B-S. Chiou and M-C Lin, “Electrical properties of amorphous barium titanate films prepared by low power r.f sputtering’,Thin Solid Films, 248,247, (1994).
152
Dielectric Materials and Devices
9. D.R. Uhlmann, J.M. Boulton and G. Teowee, “New optical materials by wet chamical processing” , J. Non-Cryst. Solids, 196,26, (1996). 10. M.E. Lines, “Microscopic model for a ferroelectric glass”, Phy. Rev. B, 15, 388, (1977). 11. Y. Xu
and J.D. Mackenzie, “A theoretical explanation for ferroelectric-like
properties of amorphous Pb(ZrxTil-do3and BaTiO;’,
J. Non-Cryst. Solids, 246,
136, (1999). 12. S.M. Zanetti, E.Longo, J.A Varela and E.R. kite, “Microstructure and phase evolution of SrTiO, thin films on Si prepared by the use of polymeric precursorsy7, Materials Letters, 31,173, (1997). 13. E.R. Leite, C.M.G. Souza, E. Long0 and J. A Varela, “Influence of polymerization on the synthesis of SrTiG. 1. Characteristics of the polymeric precursor and their
thermal-decompositi~n~~, Ceram. Inter., 21, 143, (1995). 14. F. Farges, “Coordenation of Ti in crystalline and glassy Frenoites: A high-resolution
XANES spectroscopy study at the Ti K-edge”, J. Non-Cryst. Solids, 204, 53, (1996). 15. W. Zhu, O.K. Tan and X. Yao, “Amorphous ferroelectric (B~.61Sro.33)Til.0203 thin films with enhanced H2 induced interfacial polarization potential”, J. Appl. Phys., 84, 5134, (1998). 16. G. M. Rao, and S. B. Krupanidhi, “Study of electrical-propertiesof pulsed excimer laser deposited strontiumtitanate films”, J. Appl. Phys., 75 (S), 2604, (1994).
Dielectric Materials and Devices
153
17. G. M. Rao, and S. B. Krupanidhi, “Pulsed excimer laser ablation of (F‘b, La)Ti03
thin films for dynamic random access memory devices”, Appl. Phys. Lett., 64 (12): 1591, (1994). 18. H. B. Shanna, and A. Mansingh, “Phase transition in sol-gel derived barium titanate
thin films”, J. Phys. D: Appl. Phys., 31, 1527, (1998).
154
Dielectric Materials and Devices
CHARACTERIZATION OF RESIDUAL STRESS ON Pt/Ti ELECTRODE OF INFRARED SENSOR
Ki-beom Kim, Mi-na Yoo, Do-hoon Kim, Sang-il Lee and Moon-Ho Lee Department of Applied Electronics, The Graduate School, Yeungnam University, Kyungsan 7 12-749, KOREA
ABSTRACT Residual stress and surface roughening phenomena of Pt/Ti thin film electrode were investigated by using XRD, SEM and AFM. Pt( 100nm)/Ti(20nm) thin films were sequentially deposited on SiOdSi substrates by DC magnetron sputter without breaking vacuum in the temperature range from room temperature to 600°C. XRD and AFM study were performed to prove the relationship between residual stress and substrate temperature. Stresses on film increased with increasing the substrate temperature. Residual stress on electrode film was originated by the difference of thermal expansion coefficient between substrate and thin film.
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property
of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
155
INTRODUCTION PZT(Pb(Zr,Ti1,)) thin film has been widely studied for the application as IR sensor, NVRAM and microactuator.1) This is mainly due to the desirable properties such as high remnant polarization, fast switching speed and high Curie point. Various deposition techniques have been used to fabricate PZT as thin film such as RF sputtering, sol-gel method, MOCVD and PLD(Pu1sed Laser Ablation). However, all of these deposition techniques commonly need high temperature for crystallization and 02(oxygen) atmosphere due to high vapor pressure of PZT. Because of these two factor, high temperature and 0 2 , electrode material for PZT must have good thermal and chemical stability. Because deposition of PZT thin film is performed under 0 2 atmosphere in general, electrode material must be resistant to oxidation. But, noble metals such as Ag and Au can form solid solution in the Perovskite structure. So, Pt is the only viable candidate, But, there are some drawbacks associated with using Pt. These include hillock formation due to stress relief and poor adhesion to SiOz/Si substrate.W Therefore, Ti has been widely used to have better adhesion between Pt and Si02. So called as a "buffer1' or "glue" layer. Unfortunately the Ti layer has been found to diffuse into Pt during high temperature process and lead to form Pt-Ti alloy and TiO,.4) This degrades sheet resistance and enhances hillock formation in Pt/Ti stack. Many researchers already have been studied about surface roughening and hillock formation. However their study was mainly focused on post annealing as a function of annealing time and temperature.
156
Dielectric Materials and Devices
Table 1. Physical properties of used materials lattice parameter( A)
thermal expansion coefficient( 1o - ~ / 4.03 oc
PZT
4.036
Pt
3.923
14.2
Ti
2.950
6.5
Si02
N/A
0.4
Si
5.430
2.33
Table 1. shows lattice parameter( A) and thermal expansion coefficient( IY ) of materials which used for the FMIS(ferroelectrics/ metal/ insulator/ semiconductor) structure. As can be seen from this table, there are quite large difference on thermal expansion coefficient between Pt and Ti. Assuming that temperature change from high temperature to low temperature when cooled down, tensile stress will be applied to electrode as depicted in Fig. 1. Because thermal expansion coefficient of Pt is more than 6 times larger than that of Si substrate and compressive stress otherwise. Many researchers have tried to prove that hillock is formed to relief the compressive stress and, as a result, surface is roughened. In this study, we first tried to find relationship between residual stress and substrate temperature by XRD study. As the residual stresses cause change of the spacing of crystal planes, reflected as the shift of the diffraction peak to higher or lower angle depending on the nature of the stress(compressive or tensile).5) Also Valdova et al.61, Perry et a1.W and Rickerbys) have reported that measuring the peak shift or the lattice parameter change enables measurements of residual stress. Residual stress characterization on Pt/Ti thin film will be discussed in detail later.
Dielectric Materials and Devices
157
EXPERIMENT Pt and Ti electrodes were sequentially deposited on SiOz/Si substrate by DC magnetron sputter without breaking vacuum. This method is enabled by multi target sputter that can prevent Ti from oxide formation during opening chamber and changing target. The thickness of films were fixed to 100/20nm which known as optimal for electrode for ferroelectric thin film by many other researchers. The two metals are deposited within temperature range from room temperature to 500°C. XRD spectra were taken by using of X-ray diffractormeter(Rigaku, D/MAX-2200H7 Japan) with Cu target at 40kV and 100mA. Scan speed was 5" per step. As mentioned before, stress measurement was achieved by measurement of the change in d-spacing which is revealed as changes in diffraction peak position. The residual stress can be expressed like this in Bragg-Brentano method, which operates in the 8 -2 8 scan mode;lo)
where E,
U,
d , and
do are the Young's modulus, Poisson's ratio,
d
-spacing of the diffraction plane under stress and d-spacing of the diffraction plane free from stress. And surface roughness was measured by , m. AFM(Park Science Instrument, M5, USA) over area of 10 ,Y m X 10 U RESULTS AND DISCUSSION Fig. 2 shows the SEM photos of Pt/Ti electrode deposited at 400"C, 50 0°C and 600"C, respectively. Samples prepared below 400°C have almost flawless surface and do not have difference that can be recognized by SEM image. From 500°C hillocks are started to form, and surface is very roughened at 600°C.
158
Dielectric Materials and Devices
Fig. 3. and Fig. 4. show X-ray diffraction patterns of Pt/Ti thin films on SiOdSi substrate. As can be seen in Fig. 3., spectral widths are narrowing and peaks are shifting as the substrate temperature increase from room temperature to 600°C during deposition. The 4 2 9 reflects the change of the d-spacing of hkl plane at 2 0 . In Fig. 4, spectrum of sample prepared at 200°C has shifted to the left, to the lower 2 9 angle, than spectrum of sample prepared at room temperature. As the shift of the diffraction peak to higher or lower angle depending on the nature of the stress. But, as there are complex information in XRD spectra, it is difficult to decide peak shift is due to residual stress. Table 2. Properties of electrode up to substrate temperature Substrate Temperature( "C) Room temp. 100 200 300 400 500
d-value at Pt(ll1) peak(A) 2.2494 2.254 1 2.2549 2.2656 2.252 1 2.2548
Average roughness( A) 7.30 6.66 6.65 7.40 9.77 10.7
Fig. 5 . shows AFM images of Pt/Ti electrode after deposition. AFM measurement was done on area of 10 ,u mX 10 ,u m and average roughness was measured. Among them, sample prepared at 200°C has the lowest value of average roughness. Table 2. is the list of properties of electrode up to substrate temperature. Assuming that the lowest surface roughness means the lowest residual stress, The d-value at 200°C will be as close as the d-value at Pt(ll1) peak from JCPDS card which is 2.265 A.
Dielectric Materials and Devices
159
Table 3. Comparison of lattice parameter Substrate temperature( E ) Room temperature 100 200 300 400 500
theoretical value d-value 28 2.265 39.80 2.268 39.72 2.271 39.66 2.274 39.61 2.277 39.55 2.280 39.50
experimental value d-value 2.249 2.254 2.254 2.265 2.252 2.254
28 40.062 39.975 39.960 39.746 40.012 39.962
Table 3. is the comparison of lattice parameter up to substrate temperature change. Theoretical values are calculated from d-value of JCPDS card and known thermal expansion coefficient. The d-value increases linearly in theoretical condition, but experimental value shows different result. It is considered that due to interference effect of substrate which is bounding Pt film. CONCLUSION As the substrate temperature increase, Pt(ll1) peak of XRD spectrum has first shifted to lower 2 8 angle and then to higher 2 8 angle. And by the hillock which is considered to be formed to compensate residual stress on Pt/Ti thin film, surface roughness is increased as the substrate temperature do so. Large difference in thermal expansion coefficient is believed to be responsible for the this phenomena, stress induced hillock formation and surface roughness. ACKNOWLEDGEMENT Ki-beom Kim wishs to thank Dong-soo Kim at RIST(Research Institute of Science and Technology) for his help on AFM.
160
Dielectric Materials and Devices
REFERENCES
1) S. S. Dana, IS. F. Etzold and J. Ctabes, Journal of Applied Physics, 69,4398 (1991).
2) P. D. Hren, S. H. Rou, H. N. Al-Shareef, M. S. Ameen, 0. Auciello and A. T. Kingon, Integrated Ferroelectrics,
2, 31 1 (1992).
3) J. Olowolafe, R. E. Jones Jr., A. Campbell, C. J. Mogab and R. B. Gregory, Journal of Applied Physics, 73, 1764 (1993).
4) H. N. Al-Shareef, D. Dimos, B. A. Tuttle and M. V. Raymond, Journal of Material Research, l2, 2 (1997).
5) S. Zhang, H. Xie, X. Zeng and P. Hing, Surface and Coatings Technology, 122, 2-3 (1999). 6) V. Valdova, R. Kuzel, R. Cerny et al., Thin Solid Films, 193/194, 401-408 (1990). 7) A. J. Perry, V. Valvoda, D. Rafaja, Thin Solid Films, 214, 167-174 (1990). 8) A. J. Perry, M. Jagner, W. D. Sproul, P. J. Rudnik, Surface and Coatings
Technology, 42,49-68 (1990).
9) D. S. Rickerby, S. J. Bull, A. M. Jones, F. L. Cullen, B. A. Bellamy, Surface and Coatings Technology, 39/40,4397-408 (1989).
10) B. D. Cullity, "Elements of X-ray Diffraction," p454, Addison-Wesley Inc., Reading, MA., (1978).
Dielectric Materials and Devices
161
At high temperature
As cooled down
Fig. 1. Schematic diagram of residual stress generation.
162
Dielectric Materials and Devices
Fig. 2. SEM photos of Pt/Ti electrode deposited at various temperature: (a) 400’C. (b) 500°C and (c) 600°C.
Dielectric Materials and Devices
substrate
163
20
I
30
40
50
60
70
80
T w o theta
Fig. 3. XRD diffraction peaks of samples deposited at various substrate temperature. (a) room temperature, (b) 100°C, (C) 2OO0C, (d) 3OO0C, (e) 400°C, (f) 500°C and (g)6OO0C
164
Dielectric Materials and Devices
380
385
3.0
38.5
4QO
45
41.0
41.5
420
T W M Fig. 4. XRD diffraction peaks of Pt (111) deposited at various substrate temperature.
Dielectric Materials and Devices
165
Fig. 5. AFM images of Pt/Ti thin film deposited at various substrate temperature. Surface roughness was measured on the area of 1Opm by 10prn. (a) at room temperature, (b) lOO"C, (c) 200°C, (d) 300°C, (e) 400°C and (f) 500°C.
166
Dielectric Materials and Devices
Dielectric Materials and Devices
167
This page intentionally left blank
CERAMICS DIELECTRIC PROPERTIES SURFACE FRACTAL NATURE
IN RELATION WITH GRAlNS
Vojislav V. Mitic, Ljubiia M. KociC and Ivona 2. Mitrovic University of NiS, Faculty of Electronic Engineering, Beogradska 14, 18000 NiS, Serbia, Yugoslavia
ABSTRACT Ceramics grains contacts are essential for understanding complex dielectric properties of electronic ceramics materials. Since the real intergrain contact surface is an irregular object, the theory of fiactal sets can be introduced. BaTiOsceramics, studied in this paper, has fiactal form in, at least, two levels: shapes and distributions of grains and intergrains contacts. Using method of fractal modeling a reconstruction of microstructure configurations, like shapes of grains or intergranular contacts can be successhlly done. Furthermore, the area of grains surface is calculated using fiactal correction that expresses the irregularity of grains surface through fiactal dimension. This leads to a more exact calculus of ceramics dielectric properties as well as more realistic understanding of electrical behaviour of barium-titanate ceramics. INTRODUCTION The complex grainy structure is difficult to describe by using traditional analytic methods. Here, an attempt is made on establishing ceramics grains shapes by fractal modeling. The fiactal method has been systematically employed in describing and studying hierarchical rules and patterns in different fields of science, by many authors. In addition to the Euclidean concept of symmetry, fractal object has a special kind of inner, deeper symmetry, characterized by the Hutchinson's operator: the union of contractions of a metric space. Being a fixed point of such an operator (which is a contraction by itself), a fractal set possesses a fairly complicated structure and the unique feature of self-similarity. The next characteristic of fractal sets (fiactals) is its fiactal or Hausdorf-Besicovitch dimension (a generalization of the classical concept of dimension), which is, as a rule, a fractional number. This number determines, in a way, complexity of the
To the extent authorized under the laws of the United States of America, all copyright interests in t h ~ publication s are the property of The American Ceramic Society. Any duplication, reproduction,or republicationof this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
169
particular fiactal. These features make the fiactal theory an inevitable tool in the study of physical reality. The latest ceramics structure analysis has shown that grains formations can be described by terms of fiactal geometry’. This is confirmed by observing the microphotographs obtained by scanning electronic microscopy (SEM) method or some other similar technique. The micrographs undoubtedly make an impression about grains shapes as irregular structures which can not be well enough defined by Euclidean geometric forms. In turn, ceramics grains can rather be seen as fiactal objects. Therefore, grains’ geometric parameters such as grains’ shapes and surfaces can be estimated if the fiactal dimension of grains surface is previously determined2. Then, the fi-actal model, developed in this paper, may be applied in order to obtain exact graphical representation of grains surface structure presented on SEM microphotographs.
EXPERIMENTAL The modeling of two BaTiO3-ceramic grains in contact is done following the developed procedure of fiactal modeling. The BaTiO3 samples were prepared from “Murata” barium-titanate powder of the composition: 65.24% BaO and 34.70% TiO2. The specimens were pressed under the pressures 86-150 MPa and sintered in a tunnel krnice type CT-10 MCTRATA from 1240-1370°C for two and three hours. Microstructure morphology of BaTiO3-ceramics was observed by the method of scanning electron microscopy (electronic microscope type JEOL-JSMT20, magnification of 3 5000 times, resolution of 4.5nm). Obtained micrographs are the base for applying fractal and hrther stochastic modeling of grains structure (Figure 1).
Figure 1. Microphotographs of BaTi03-ceramics samples with addition of 0, 10% CeOz and 0,14% MnC03obtained by SEM. Sintering parameters are p= 86 MPa, T,ht=129O0Cand ~,i,-,~:a) 2h; b) 3h ( ~ 5 0 0 0 ) . A FRACTAL MODEL
A wide class of fractal sets can be defined as invariant under a contractive
170
Dielectric Materials and Devices
mapping. This overall definition leads to the theory cf Iterated Function Systems (IFS) intoduced by Barnsley3. be a set of contractive Lipschitz mappings with factors ISily=, ..., n, of the metric space (R",d). Let H(Rv) be the set of all nonempty compact subsets of R", and let h denote the HausdorfY metric induced by d,so that (H(R"), h) is a complete metric space. Let p1, ..., pn be positive real numbers such that Cpi'l. The system w
i
= R V ; w ,..., , w,;
I
p , ,. . . , p ,
(1)
is called (hyperbolic)Iterated Function System, (or IFS) with probabilities. The associated Hutchinson operator W : H(R")+ H(R"), defined by is a contraction of (H(R"), h) with the Lipschitz factor s=maxi(lsil}O}, is the (m-1)-simplex with vertices P = [PI...P,JT. Denote V=affi{P1...P,), the affine hull of P. Note that the strict inclusion PAcVcR" takes place. Then, for any point XEV, the components of its position vector x = [XI, ..., &IT, namely its Descartes coordinates, coincide with the barycentric coordinates of X w.r.t. PA, that is X = [XI,..., xm]P, &=I. Let S=[~ij]Tj=l , C Sij = 1 be a real nonsingular row-stochastic matrix, and let a unique linear transformation V-+V be defined by L(P) = SP. (4) This linear mapping is naturally associated with an affine transformation (projection of L )
Dielectric Materials and Devices
171
A(y) = Ay + b, y ~ V = o span{el, ..., em-i}, (5) where A is an (m-l)x(m-1) real, regular matrix, and b an (m-1)-dimensional real vector. The matrix S is uniquely determined by A and b and vice versa. Moreover, S, A and b are connected with
STQ=[l.
.:.
.b. Smm
1,
Q= -1
i
11
where c is an (m-1)-dimensional vector. Let o(M) denote the spectrum of a square matrix M. If o(A)={ h2, h3, ..., L}, then o(S)={l, h2, h3, ..., h,},which is strict consequence of relation (6). It is interesting to note that the (right) eigenvectors of ST, vz, ..., vmassociated with the eigenvalues h2, ..., h, are orthogonal to e, i.e. they form a base in V. Namely, if (u,v) = u*v denotes the usual scalar product of complex-valued vectors U and v, then, hi(vi, e) = (hivi, e) = (hivi)*e = (STvi)*e = (vi*S)e = vi*(Se) = vi*e = (vi, e). Therefore, hi(W, e) = (Vi, e) and since S is a regular matrix, hi +O (i = 2, ..., m), so it must be (vi, e) = 0. Being linearly independent, they form a base in V, i.e. V = span{vz, ..., vm}. Theorem 1. Let L andA be the mappings of the metric spaces (v; d)and (VO, do) respectively, grven by (4) and (5). Den, if one is a contraction, the other is also a contraction. Now, we can state the definition of an AIFS. Definition. Let T = [TI ... T,lT define a nondegenerate simplex TAin Rm-land let pl, ..., pn be the set of probabilities. The system a ( T ) = {T; S1, ..., S,; pl, ..., pn}, where S k are real, nonsingular row-stochastic matrices, is called M i n e Invariant Iterated Function System (AIFS) with probabilities. Theorem 2. Let w be a hyperbolic IFS given by (1) with v=m-I and let F=att(w). Than,for any nondegenerate simplex TA cR"-',the attractor att(LI(T,)) is afSinely equivalent to F. The obvious corollary of previous theorem is that att{a(A(T)))=A { att(Q(T))}. The following theorem offers an algorithm for constructing att(Q(T)). , p n (p, > 0 and = I), Theorem 3. Let the hyperbolic AIFS be gwen. Let p ~..., be a set of probabilities. Let, for some I <j I n , ro E R" satis& Sy ro = roand eTro
a,
=
I . Let the elements of the sequence
8 = [rr ...r",'
E
{xfi))ka
R", prob(8 = Sf
be generated according to:
8-')= pi
(1 I i I n), ) ' x
m
=
c r ; Tj. J
=1
Then, the points x*), k = 1, 2, ... lie on att(Q(T)).
172
Dielectric Materials and Devices
In the next section we will use Theorems 1-3 to construct 3D image of a ceramics grain. Modeling In order to model a ceramic grain we need some input data. One of the methods they can be provided is the SEM method. What we need is an approximative value of the fiactal dimension of a grains' surface and some polyhedral approximation of this surface. Once we have those data we proceed as follows. Denote the grain by G. Let G* be its polyhedral approximation, and let GI, be the set of sides of G*. Obviously G*, is a 3D graph. Select a path in G', that connects vertices T1, ..., T,. Then, approximate the given polygonal line by T=[T1...TmIT. Let the approximation has the form T*=UwI(T), where wi (i = 1, ..., n) is a I-1
contractive afine mapping in R3. Choose Wi such that h(T, T*) < E, where E is a given positive number and h is the H a u s d o e distance. Calculate matrices S1, ..., Sn so that Si contains barycentric coordinates of Ti with respect to T. Select probabilities p1, ..., pn. In this way we have the AIFS Q(T) = {T; S1, ..., Sn; p1, ..., Pn} and we can use Theorem 3 in order to construct the attractor A which, according to the theory satisfies h(T, A) < d(1-S) where s = maxi(si}, Si being a contractive factor of the linear mapping given by the row-stochastic matrix Si. The set of probabilities pl, ..., pn can be used as a modeling tool, as well. Different choice of this set can result in different fiactal images. We usually start by a beginning setting pi = (det A ~/(C ) det A ~, )where Ai is the matrix defining by the i=l
row-stochastic matrix Si through (6). The next important issue is to match dimension. Here we have a partial answer. Namely, in the case when the AIFS is just-touching3 we can use formula that connects fiactal dimension with contractivity factors of matrices S1, ..., S,. Let Si be contractivity factor of Si and D be fractal dimension of the attractor of a justtouching AIFS. Then, i=l
Beginning fiom the known formula s = sup(l1 Sz 11 / )I x 11) for the contractivity IIxII=l
factor s of the matrix S, bearing in mind that the unit sphere in our case is standard simplex P,we can deduce that
Dielectric Materials and Devices
173
where Si is the i-th row of the matrix S and II-II is a suitable vector norm induced by the metric d. Thus, by using formulas (7) and (8) we can prescribe matrices S1, ..., Sn, so that fiactal dimension of the attractor will be D. FRACTAL MODELING OF BaTi03-CERAMlCS GRAINS Considering the fiactal model given, we will represent a typical task of modeling two ceramic grains in contact. First parameter we need is fractal dimension of the surface of ceramic grains. In the case of BaTi03-ceramicsYsome values of fiactal dimensions have been estimated2 e.g. D1=1.071890, D2=1.071993, D3=1.0119546 etc. The next step is constructing an AIFS which attractor is closely resembling the microphotography of a grain contact (see Fig. 1). This construction demands to specify a simplex as well as the set of rowstochastic matrices. To form a simplex we use approximation of the grain by a polyhedron. Then, we choose a polygonal line connecting vertices of this polyhedron and having between three and ten sides. Now, we proceed, by interactively modeling, using contractive images of this initial polygon. In this point, we should take care about the matter of dimensions as it was explained above.
Figure 2. Modeled BaTi03-ceramics grains in contact for different values of the parameter a: 0.1 (left) and 0.12 (right).
1,
In this example we will consider simplex P, with vertices P1=(0, 0, 0.6), P2=( 1, 1, l), P3=(0, 0, l), P4=(0, 1, 0). Subdivision matrices are given by 1-U
-U
174
0
0.5
0
U
0.5
-0.2
0.5
0.8
-0.1
2u
0.5-U
0.5-U
0
0.5
0
0
0.5
0
1-U
0
U
s,
: : :::
0.3 0 . 3 + ~ 0 . 4 - U
0.2
0.7 - U
0
of] 0.1
Dielectric Materials and Devices
0.2 2:
0
0.7
-",8 0.3
0.6 0 0
0.1
0.61' 0.1
1-a
1-a s5=[:a
a
a
-a
1-a
-a
l:j*
-a
1-U
;
The set of pro )abilities is given by pi=0.2 (i=l, ..., 5). The parameter a which appears in all of matrices plays the role o f a shape parameter. Putting ~ 0 . gives 1 attractor which is shown in Figure 2 (left). It is visible that two ceramic grains have a contact which is highly irregular. The new situation is obtained for ~ 0 . 1 2 (Figure 2, right). Obtained results undoubtedly show that BaTi03-ceramics grains are irregular structures which can be successfully described and reproduced by terms of fi-actal geometry. CONTACT SURFACE AREA AND DIELECTRIC PROPERTIES Let us discuss the influence of fractal dimension on increasing the size of contact surfaces. Due to diffisional forces that appear in sintering process we are ready to believe that an approximate form of a contact surface is the shape of a minimal surface - the surface with minimal area size. But, the microstructure of the material makes this surface to be fractal locally (Fig. 3).
Figure 3. An intergrain contact surface has fractal form.
Figure 4. Two grains in contact form a microcapacitor.
Dielectric Materials and Devices
175
Considering the fiactal approach to the intergrain geometry, the formula for the microcapacity of intergranular condensor seen as a planar condensor (Fig. 4) is given into the following form
where EO, EB are dielectric constants in vacuum and in BaTiO3-ceramics material respectively; S- the area of the ‘plates‘and x - distance between condensor ‘plates’, i.e. the condensor thickness and a = ( N { 2 ) k, is a correction factor obtained by a constructive approach to the fiactal surface. This approach uses an iterative algorithm that iterates N self-afine mappings with a constant contractive (Lipschitz) factor I I< 1 k-times3. The underlying theory and techniques for choosing the appropriate mappings are given in previous work.2 Typically, a = D - DT , where D = 2.08744 is the fractal (Hausdorff) dimension of intergrain contact surface3 and DT = 2 is toplogical dimension of the surface. As it is found’y275,BaTiO3-ceramics contact surfaces are of low-irregularity which is caracterized by the small difference D - DT = 0.08744. Derived formula (9) indicate the increase of the value of microcapacity when fiactal approach is applied. Thus, more accurate calculation of microcapacitance generated in grains contact can be carried out leading to a more exact estimation of dielectric properties of the whole sample. CONCLUSION The explanation of sintering phenomena is largely based on empirically established laws being only an approximation of real process. For better analysis new mathematical tools are required. The impact of rapidly developing computeraided design and computer-graphic techniques promises to revolutionize the entire problem of sintering description. The results presented in this paper undoubtedly show that a new modeling tool in ceramic structure analysis is fractal modeling. The fiactal model developed in this study enables the reconstruction of grains surface as well as intergrain contacts. Starting with the fiactal dimension of BaTi03-ceramics grains surface, an AIFS is constructed with the attractor closely resembling the SEM microphotography of a BaTiO3-ceramics grain. Furthermore, the formula for intergranular capacity is improved considering the fractal nature of intergrain contact. In that sense, ceramics dielectric properties can be correlated with grains surface fiactal nature. ACKNOWLEDGEMENT This research has been supported by the project: “Prognosis of material characteristics jnanced by the Ministry of Science and Technology Republic of Serbia. ”
176
Dielectric Materials and Devices
REFERENCES 1. V. Mitic, Lj. Kocic, M. Miljkovid, I. Petkovid, "Fractals and BaTi03-Ceramic Microstructure Analysis", Mikrochim. Acta [Suppl.] 15, pp. 365-369, 1998. 2. V. MitiC, Lj. Kocic, I. Mitrovic, M. kstic, "Models of BaTi03-Ceramics Grains Contact Surfaces'' The 4th IUMRS International Conference in Asia OVTA, Makuhari, Chiba, Japan, September 16-18, 1997.
3. M. Barnsley, "Fractals Everywhere", Academic Press, 1988. 4. Lj. Kocic, A. C. Simoncelli, "Towards Free-Form Fractal Modelling", Mathematical Methods for Curves and Surfaces 11, M. Daehlen, T.Lyche and L.Schumaker (eds.), pp. 287-294, Vanderbilt University Press, Nashville (TN.), 1998.
5 . V. Mitic, 2.Nikolic, Lj. Kocic, I. Petkovid, M. fistic, "Fractal Surfaces and BaTi03Ceramics Dielectric Properties", 100th Annual Meeting of the American Ceramic Society, Cincinnati, Ohio, 1998.
Dielectric Materials and Devices
177
This page intentionally left blank
TALORING OF ELECTROMECHANICALPROPERTIES OF
Pb(Mgl,3Nb,~3)03-PbTi03-BaTi03-BASED RELAXORS C.H. Yoon, A. Sehirlioglu*,and S.M. Pilgrim N Y S College of Ceramics at Alfred University Alfred, NY 14802 K. Bridger Active Signal Technologies Cockeysville, M D 2 1030
ABSTRACT A particulate coating process was used to achieve tailored properties to meet disparate applications by modifying a commercially available Pb(Mg113Nb2/3)03PbTi03-BaTi03 (PMN-PT-BT) based ceramics. By incorporating small amounts (<2 mol %) of chemically derived additives, such as Ti and Zn, into PMN-PT-BT based ceramics, it was possible to synthesize ceramics with improved and tailored properties. The addition of Ti with Zn raised dielectric constant, induced strain, and polarization well above that expected for conventional additions and nearly doubled the induced strain across most of the reduced temperature (T-T,,) range -- a span of 40°C. Diffraction and microscopic work showed that improved properties were correlated with the appearance of discrete second phases. INTRODUCTION Electrostrictive ceramics (specifically those based on Pb(Mgl,3Nb2/3)03 (PMN)) are of interest for a variety of electromechanical systems including actuators and transducers.1-6 PMN-based materials, with their large (>O. 1%) fieldinduced strain at practical fields (-1 MV/m), low electromechanical hysteresis (<5%), high (>GOOpC/N) effective d33,and high energy density, are promising for a wide variety of military and commercial applications. One of the barriers to broader use of PMN in device applications has been the variation of performance with temperature. Although excellent performance can be achieved for a given application, the limited performance window seems to Now at University of Illinois at Urbana Champagne To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright ClearanceCenter, is prohibited.
Dielectric Materials and Devices
179
require a producer to maintain a wide variety of compositions to meet disparate applications. In this study, a novel chemical processing route was incorporated to achieve tailored properties by modifying a commercially available relaxor composition, 0.96(0.91Pb(Mg,,,Nb2/3)03 - 0.09PbTi0,) - 0.04 BaTiO, (PMN-PT-BT). Previous work has shown that the particulate coating processes7-*utilizing sol-gel coating route had an effect on controlling microstructure during densification for diverse purposes and applications. Consequently, the chemical additive process utilizing sol-gel reactions was investigated to modify the base composition and assess the resultant properties of PMN-PT-BT. Several additives such as Ti and Zn were selected as promising candidates for improving the dielectric and electromechanical properties of the given PMN-based composition. These were added at levels between 0 and 2 mol% to the starting composition. EXPERIMENTAL PROCEDURE A commercial PMN-based relaxor with the composition 0.96(0.91Pb(Mg,,,Nb2,,)O,-O.O9PbTiO,) - 0.04BaTi0, (B400090, supplied by Martin Marietta Laboratories) was used as a base material. The base composition was modified by addition of various cationic additives. These were added by coating the base particulate via a sol-gel route using the following reagent grade raw materials: titanium isopropoxide (Ti[OCH(CH,),],) and zinc acetate dihydrate (Zn(CH,CO,),~H,O>. The compositions of the corresponding additives are shown in Table I. Table I. Compositions of corresponding additives Sample ID Compositions Precursor for Additives PTO PMN-PT-BT, base materials PTZ2 PMN-PT-BT with Titanium Isopropoxide, Zinc Acetate Dihvdrate 2 mol% ZnO & 2 mol%TiO., For particulate coating, titanium isopropoxide was diluted with isopropanol maintaining a clear solution. The appropriate amount of Zn acetate (corresponding to the given content in Table I) was then dissolved in the solution. After stirring for -2 hrs for complete dissolution, small amount of nitric acid was added to precipitate the sol-gel reaction. This aqueous solution was then slowly mixed with the Ti isopropoxide-isopropanol solution. The base PMN-PT-BT powders were added into the solution containing the additives while the solution was stirred with a magnetic stirrer. The stirring process of the resultant slurry continued until complete evaporation of the solvent was achieved (at 560°C). The resultant dried powders containing the gel additives were then crushed using
180
Dielectric Materials and Devices
a zirconia mortar/pestle. Each coated powder was mixed with 2wt% PVA solution and then pelletized under 80MPa uniaxial pressure. The specimens were fired at 600°C for 2 hours to burn out the binders and then sintered at 1200°C for 4 hrs in a Pb atmosphere (buried in sacrificial base powder with inverted alumina crucibles). The sintered pellets were electroded with a silver paste after sputtering both sides of the disk-shaped samples with Ag-Pd. The paste electrodes were then fired at 550°C for 1 hr. Dielectric properties of the sintered samples were measured during cooling at 2"C/min using an LCR meter (HP4284A) with variations in frequency (100 Hz to 100 kHi) and temperature (from 150°C to 50°C). An integrating capacitor circuit was used to observe polarization behavior with applied electric field (nominally_+lMV/m and 1 Hz). The ac field was supplied by a Hewlett Packard 8904A waveform synthesizer and amplified by a Trek 610C high voltage amplifier. Longitudinal strains were determined using a MTI 2000 FotonicTMSensor. Analog-to-digital conversion was accomplished using an Iotech ADC488/8SA. A computer, in conjunction with an IEEE 488 bus, was used for control and data acquisition. A program written in LabViewTM was used to automate the process. The data were processed and filtered using subroutines written in MatLabTM. For microstructural analysis of the surface of the sintered specimens, a scanning electron microscope (1810 SEM, Amray Co.) was used. Based on the SEM microstructures, the average grain size was calculated by the linear intercept method. Grain boundary of PMN was studied using transmission electron microscopy (Jem 2000 FX, Joel) with EDS detector (PGT-Prism Light Element X-ray detector). RESULTS AND DISCUSSION Figure 1 shows the variation in dielectric constant (k) measured at 1 kHz with temperature (T). The overall summary of the dielectric properties for modified compositions is shown in Table 11. Table 11. Weak-field dielectric DroDerties of PMN-PT-BT comPositions k at 25°C D at 25°C k at T,, D at T,, Sample ID at1kHz at1kHz atlkHi at1kHz *PTO 17398 0.0180 18477 0.0320 0.0410 **PTZ2 23778 0.0080 25064
*. All numbers are average of 6 samples. **. All numbers are average of 2 samples.
Tm, ("C) 22.5 30.5
The relative permittivity plots show conventional relaxor behavior with moderately high maximum permittivities. Unexpectedly large variations of kmax
Dielectric Materials and Devices
181
and T, for the various compositions were obtained in spite of a small amount of additives. The Zn/Ti compositional modifications show substantial increases in permittivity with an accompanying shift in Tma. Within these compositional modifications, permittivity increases with increasing grain size. 0.25 0.20
- 8 e
0.15 0.10
&
0.05 I4
0
50
100
Temperature ("C)
0.00 150
25000
0.20
3 20000
0.15
; .s 2 9
0
-50
0.25
30000
15000
.s n
0.10;
10000
0
0.05 I4
5000 0
3$
-50
0 50 100 Temperature ("C)
158"
Figure 1. Weak-field dielectric behavior of PTO (base composition) and PTZ2 at O.l,l, 10, and 100 kHzfor a cooling rate of 2"C/min. High-field and low-frequency strain and polarization data for the modified compositions are summarized in Table 111. The Q33values are the coefficient of the x2 component of the second order polynomial curve fitted to the strain vs. polarization plot. Table 111. Electromechanical properties at 25°C of PMN-PT-BT compositions (measured at 1 Hz at f 1 MV/m) Sample # Microstrain Polarization(mC/m2) Q33(m4/C2) ID meas. mean *20 mean *20 mean h20 PTO 3 370 70 165 10 0.013 0,004 PTZ2 6 645 72 22 1 6 0.013 0.001 ~
The strain curves (not shown) are fairly typical for the PMN-based electrostrictive materials. Compared to the base material, the Zn/Ti compositional modifications exhibited substantial increases in strain. This unusual result is consistent with the information presented in a previous work'. Peak strain and polarization are plotted versus reduced temperature (Tma is the peak in 1 kHz permittivity) and this is shown in Figure 2. The addition of Ti with Zn raised dielectric constant, induced strain, and polarization well above that expected for conventional additions and nearly doubled the induced strain across most of the reduced temperature (T - TmJ range -- a span of 40°C.
182
Dielectric Materials and Devices
-
1400
2 * a
2
300,. ,
PTO (Base Canp)
1200 1000
800
g
600
,
I , . , I . , , I , , , I .
,,
20
- 15
200
60
I , ,
- 10 5
.s 400 M
.
n
-40
-20
T - T,,
0
20
40
("C)
0
-60
40
-20
0
20
T - Tmax("C)
40
Figure 2. Longitudinal strain and polarization commonality for PMN-PT-BT compositions (measured at 1 Hz at f 1 MV/m). Figure 3 shows selected SEM micrographs of compositions containing the chemical additives sintered at 1200°C for 4 hrs. In the case of the base composition, abnormal grain growth, consistent with the presence of agglomerates in the dried powders, was found. The overall microstructural characteristics, e.g., grain size, were found to depend on additive composition. The addition of Zn with Ti aided grain growth. The summary of average grain size is shown in Table IV. The relatively uniform grain size distribution in compositions with chemical additives is thought to be related to the homogeneous distribution and the retarding effects of the additives. The observed uniform distribution of gel additives at the particle surface is believed to significantly contribute to the homogeneous distribution of the small additive amounts throughout the base powder. Chemical additive processing has an advantage over the convention batch mixing process since chemical additive processing leads to a more homogeneous dopant distribution and a more uniform microstructure.
Figure 3. SEM micrographs of PMN-PT-BT (base composition) with chemical additives (sintered at 1200°C for 4 hrs.).
Dielectric Materials and Devices
183
Table IV.Average grain size of base and modified compositions. Average Sample ID Maximum Minimum Grain Size (pm) Grain Size (pm) Grain Size (pm) PTO 2.7 0.6 8.7 PTZ2 3.6 8.1 1.5 Figure 4 shows TEM micrographs of modified composition (PTZ2).
(a) bright field image of second phase (b)TEM/EDS of second phase Figure 4. TEM micrograph shows that second phase at triple point is Zn-doped Periclase ((Mg0.68zn0.32)0). In the case of Zn/Ti addition, Zn-doped periclase, (Mg,,8Zno,32)0,forms as a second phase in the triple points. All Ti may go into intrinsic B-site vacancies in the crystal and the system is charge-balanced by either Pb loss or Nb reduction. Ti most probably enters the system as Ti02+. Based on the microscopy data, one may think that the enhanced properties may result from the elimination of a deleterious phase or condition at the grain boundaries of the base materials although the reasons are not clear. CONCLUSIONS A chemical additive process using sol-gel reactions was used to modify a base composition of a PMN-PT-BT material with the intent to improve the resultant performance for electromechanical uses. The method led to intimate mixing of the additives at the nanoscale without an additional ball-mixing process and had the effect of improving homogeneity and the uniformity of grain size. Despite of additives, all properties retained typical temperature dependence. The addition of Ti with Zn raised dielectric constant, strain, and polarization well above that expected for conventional additions. Strain and polarization are nearly double for a 40°C interval with a minimal increase in hysteresis. Electromechanical
184
Dielectric Materials and Devices
properties, coupled with the microscopy data suggest that the enhanced properties result from the elimination of a deleterious phase or condition at the grain boundaries of the base materials. In the case of Ti/Zn addition, Zn-doped periclase was formed as a second phase at triple points. In summary, a 'poor' base material (PTO) can be vastly improved. Thus it appears that a 'poor' material can be coated so as to provide more optimal performance. ACKNOWLEDGEMENTS This work was supported by the Office of Naval Research under No. N00014-98C-0296. REFERENCES L. E. Cross, "Relaxor Ferroelectrics," Ferroelectrics, 76, 241-67 (1987). K. Uchino, "ElectrostrictiveActuators: Materials and Applications," Am. Ceram. SOC.Bull., 65 [4] 647-56 (1986). S. M. Pilgrim, M. Massuda, J. D. Prodey, and A. P. Rotter, "Electrostrictive Sonar Drivers for Flextensional Transducers," in Transducersfor Sonics and Ultrasonics.Edited by M. McCollum, B. F. Harmonics, and 0. B. Wilson. Technormic, Lancaster, PA, 1993. S. M. Pilgrim, M. Massuda, and A. E. Sutherland, "Electromechanical Determination of the High-Field Phase Transition of Pb(Mg1.Nb2,,)0,-PbTi0,(Ba,Sr)TiO, Relaxor Ferroelectrics," J. Am. Ceram. Soc., 75 [7] 1970-74 (1992). M. Massuda, K. Bridger, J. D. Prodey, and S. M. Pilgrim, "High-Field Electromechanical Properties of Some PMN-PT-based Relaxors," Ferroelectrics, 158, 337-42 (1994). S. M. Pilgrim, M. Massuda, and J. D. Prodey, "ElectromechanicalProperties of Some Pb(Mgl,,Nb2,)0,-PbTi03-(Ba,Sr)Ti0, Ceramics: Part Two," J. Am. Ceram. Soc., 78 [6] 1501-506 (1995). 7 F. A. Selmi and V. R. W. Amarakoon, "Sol-Gel Coating of Powders for Processing Electronic Ceramics," J. Am. Ceram. Soc., 71 [ l 11 934-37 (1988). 8 Y.S. Cho, S.M. Pilgrim, H. Giesche, and K. Bridger, "Dielectric and Electromechanical Properties of Chemically Modified PMN-PT-BT Ceramics," accepted J. Am. Ceram. Soc., October (1999).
Dielectric Materials and Devices
185
This page intentionally left blank
VHF TUNABILITY MEASUREMENTS OF FERROELECTRIC MATERIALS USING DOUBLY REENTRANT CAVITIES RICHARD G. GEYER' AND WILLIAM E. JlCKINZIE** National Institute of Standards and Technology RF Technology Division, M.S. 813.01 Boulder, CO 80303 "Atlantic Aerospace Electronics Corp. 6404 Ivy Lane, Suite 300 Greenbelt, MD 20770
ABSTRACT
A transfer matrix transmission line method is applied to two modified doublyreentrant cavities for evaluating the tunable properties of bulk ferroelectric materials at microwave frequencies. The transverse resonator impedances of the equivalent circuit models of the modified doubly-reentrant cavities provide a convenient approach for resonant fixture design and ferroelectric specimen permittivity evaluation under application of a tuning bias. The tuning bias in these measurement systems is parallel to the rf electric field. Tunability measurements on ceramic composite specimens of Ba,Srl-,TiOs (BST) with up to 40 weight percent of added MgO exhibited tunabilities as high as 45% a t tuning biases of 4v/micrometer. Tunabilities compared well with low frequency measurements a t 297 K and did not show evidence of saturation for a tuning bias up to 4v/micrometer. Correlation with microstructural analyses shows that tunability increases with BST-BST grain connectivity. INTRODUCTION Tunability is needed in many areas of microwave electronics such as radars, communication systems, and measurement test systems. Ferroelectric materials are nonlinear dielectrics which possess a n electric field-dependent permittivity. With the incorporation of ferroelectric materials in device configurations such as capacitors and transmission lines: the high frequency reactance can be tuned with a dc electric field. Hence tunable capacitors. filters, resonators, phase shifters, and voltage-controlled oscillators can be developed for microelectronic applications. Advantages of ferroelectric materials compared with magnetically tunable materials, such as ferrimagnetic oxides, are that they are not limited by the frequency range of application, their reciprocal behavior is not dependent on the rf field polarization To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paicfto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
187
state, and their high real permittivities permit large weight and size reductions in device component design. Switching speeds for many ferroelectric materials have been measured to be much less than a microsecond, which is sufficient for most microwave applications. They are also radiation hardened and can be manufactured in bulk, thick- and thin-film forms. T h e tunability of a ferroelectric material is defined as the fractional change in the permittivity with an applied dc bias voltage,
where is the relative permittivity when no bias voltage is applied, and c : , ~ , ~is the relative permittivity when maximum dc bias (or standard bias field) is applied. T h e relative permittivity of a ferroelectric decreases as the bias voltage is increased. Although larger tunability is a desirable feature for most microwave applications, a ferroelectric material with larger tunability usually has greater dielectric loss. Tunability is a function of whether the ferroelectric material is in the ferroelectric as opposed t o the paraelectric state. At the transition temperature between these states (or Curie temperature), maximum tunability is expected, but with accompanying maximum dielectric loss. Optimizing tunability and loss to meet the needs of a particular microwave application remains a challenging technical task. In the development of new ferroelectric materials for use in rf devices, it is desirable t o measure tunability at the operational rf frequency and temperature. One measurement technique that permits accurate evaluation of frequency- and temperature-dependent real relative permittivity, E : , and dielectric loss tangent, t a n 6 , is the coaxial doubly reentrant cavity illustrated in Fig. 1. The normal coaxial reentrant cavity consists of a cylindrical cavity with a center conductor that allows insertion of a specimen for dielectric testing into an adjustable gap. It may be used for accurate measurements of relatively small cylindrical specimens at nominal frequencies that range from 40 MHz to 1 GHz, depending on specimen and cavity dimensions and specimen permittivity. In addition, if the specimen loss characteristics are not known and are relatively high ( t a n 6 > O.OOl), this measurement system often permits an accurate determination of loss. The reason is that the unloaded Q-factor of the cavity with sample insertion may be measurable when that using the specimen as a dielectric resonator is not. When the gap for sample insertion in the center conductor is at either end of the cavity, the resonant system is known as a s i n g l y reentrant cavity; otherwise, it is known as a doubly reentrant cavity. T h e rf electric field in the sample under test is in t h e axial direction, normal to the planar surfaces of the specimen. -A doubly reentrant coaxial cavity has stronger rf-electric fields in the gap than the singly reentrant coaxial cavity, and, consequently, is most often used in dielectric testing.
188
Dielectric Materials and Devices
Coupling
Adjustable Center ,Conductor
L
Cavity
Figure 1: Normal coaxial doubly-reentrant cavity for rf dielectric measurements. L=60 mm, L1=10 mm, q = 7 . 2 5 mm, r2=25 mm. Two adaptations of the coaxial reentrant cavity for rf tunability measurements were designed and fabricated. They are shown in Figs. 2 and 3. In one modification, the external cavity was physically split and a thin, high permittivity dielectric insert placed in the gap to allow application of a dc tuning voltage bias across the sample. An alternative reentrant cavity modification consisted of dielectrically insulating the lower center conductor at positive potential from the rest of the coaxial reentrant cavity at ground potential. Equivalent circuit transfer matrix models of both reentrant cavity modifications were then used t o derive the relations between measured resonant frequency of the lowest order dominant TEM cavity mode as a function of specimen real permittivity, which varies as a function of tuning bias. By using a dielectric insert in a split external cavity wall or by dielectrically insulating the lower center conductor, a dc tuning voltage may be applied to a specimen placed in the internal gap between the coaxial center conductors. A resonant frequency shift is then measured as a function of the dc bias voltage to define the tunability of the ferroelectric specimen. Because the electric field structure in the cylindrical specimen is normal to the flat axial faces, relative permittivities can be underestimated by depolarization effects resulting from any air gaps between the specimen and the coaxial inner conductor of the reentrant cavity. The specimen must be metallized on top and bottom planar surfaces t o avoid these effects. Un-
Dielectric Materials and Devices
189
+
1
Adjustable Center Conductor
Coupling
Cavity
-
r3
Figure 2: Doubly-reentrant coaxial cavity with dielectric insert for microwave tunability measurements of ferroelectric materials. L1=30 mm, L= 60 mm, r1=7.25 mm, r2=25 mm. 7.25 mm
Metalized Ceramic Bypass Capacitor
Attachment Bolt Figure 3: Doubly-reentrant coaxial cavity with lower center conductor dielectrically insulated with a ceramic rf bypass capacitor for bulk ferroelectric tunability measurements.
190
Dielectric Materials and Devices
(Input Reference Plane)
(Output
Re fe re nce
Plane)
Figure 4: Transfer matrix [ A B C D ]representation of a two-port microwave network. biased specimen dielectric losses may be measured in the normal coaxial reentrant cavity by examining the difference in Q-factor observed with the specimen placed in the cavity and that without the specimen for the same gap in the center conductor as for sample insertion.
THEORY Transfer Matrix Met hod
A convenient method for the analysis of microwave circuits is that which utilizes efficient and effective circuit analysis. Slicrowave circuits may generally be classified as N-ports, with the most common being the two-port. In addition, a complex twoport can usually be described as a cascade of simpler two-ports [l],rather than by an N-mesh network. In this work a transfer matrix method is employed for rapid modeling of the reentrant cavity structures shown in Figs. 2 and 3. The transfer matrix method is also known as the chain matrix [ABCD]representation for a microwave two-port network. In this representation the microwave device is generally described as a “black box’’ shown in Fig. 4 with indicated voltage and current sense. The linearized equations that completely describe this two-port network are
Dielectric Materials and Devices
191
I1
=
cv2
i D(-I2),
or in matrix notation,
It is conventional to show currents going into the black box as shown in Fig. 4 If, on the other hand, this black box were cascaded with another, 1 2 from the first two-port is taken as flowing out of the first two-port (polarity of I 2 same as that of Iland -I2 is replaced by +I2 in eq (2)) and becomes I1 for the second two-port both in magnitude and phase. Hence for cascaded two-ports, the convention for the [ABCD] resultant matrix is that the input and output currents have the same polarity (direction). The matrix that resuits from cascaded two-port matrices is called a chain matrix and is simply the product of the two-port matrices. In other words, if two microwave networks are cascaded and characterized individually by their [ABCD] parameters. say, [ABCD], and [ABCD],, the overall chain matrix parameters characterizing the microwave system is the product of the two matrices,
Most microwave networks or devices may then be decomposed into a chain of twoport network elements. For a series element with impedance 2 , the [ABCD]matrix is
For a shunt element with admittance Y the [ABCD]matrix is
The voltages and currents for a transmission line of length d and characteristic impedance Zo at the input and output ports (reference planes) are given in most texts [1,2,3] (Fig. 5) as
Vl
=
V2 cosh($)
II220 sinh(yd)
11 = - sinh(yd) -+ I2 cosh(yd), v 2
20
192
(7)
Dielectric Materials and Devices
Figure 5: Single length of transmission line having characteristic impedance length d.
20
and
where 7 is the propagation constant given by 7 = a+jP with (Y the loss/unit length (Nepers/m), p = 2x/X = 2nf/c is the phase constant (radians/m) and c is the speed of light under experimental conditions. The [ A B C D ] matrix for a series connected transmission line of length d and characteristic impedance 20is then given by
[ t :Id
=
[ -&
cosh(yd) sinh(yd)
20sinh(yd) cosh(yd)
I
’
Equivalent Circuit Models for Modified Coaxial Reentrant Cavities Split Doubly-Reentrant Coaxial Cavity with Dielectric Insert The equivalent circuit model used for the modified doubly-reentrant cavity with dielectric insert illustrated in Fig. 2 is given in Fig. 6. The dielectric insert is modeled as a capacitive impedance ZinseTt = 1/(2~fCi,s,,t) in series with the capacitive impedance 2, = 1/(2nfCs) of the specimen under test (neglecting the dielectric loss of the specimen). The capacitance of the ring dielectric insert is approximated as Cinsert = - r ~ ) € O t : , i n s e r t / t i n s ewhere rt, €0 = 8.854 x 10-l2 Farads/m is the permittivity of free space, cL,inseTt is the relative permittivity of the insert, and tinsert is the thickness of the insert. The capacitance of the dielectric specimen is
~(~32
Dielectric Materials and Devices
193
Figure 6: Equivalent circuit of modified doubly-reentrant cavity shown in Fig. 2 for rf tunability measurements of ferroelectric materials.
C, = K T , ~ C O ~ ; ,f ~, (T ) / t , , where
C : , ~ ( J , T ) is the relative frequency- and temperaturedependent permittivity of the specimen and t , is the specimen thickness. The chain transfer matrix of the split doubly-reentrant cavity is then given by two cascaded matrices, one for the transmission line model including the series impedance elements that represent the capacitive reactances of the insert and specimen with a series connected distributed transmission line of length L2, and one which represents the connected distributed transmission line of length L1. The first transfer matrix is given by
[: ;:] [:f][ =
cosh(yl1) sinh(yl1)
+6
sinh(yl1) cosh(yLl)
20
+
]
(9)
so that A1 = cosh(yl1) sinh(yLl), B1 = 20sinh(yL1) Z c o s h ( y l l ) , C1 = sinh(yll), and D1 = cosh(yl1). The reactive series impedance of the insert and ferroelectric specimen under test is
-&
z(f,4 , i n s e r t
7
4,s)=
1
1 +j2rf Cinsert j 2 r f Cs '
(10)
and Zo is the TEM mode characteristic impedance (ohms) given by [4]
zo = 377 In(--). 7.2
194
Dielectric Materials and Devices
T h e second chain matrix is
Equation (10) is valid when losses in the dielectric insert and specimen are negligible. When we wish to predict resonant frequencies given the specimen permittivity and dimensions or to evaluate relative permittivity from measured resonant frequency of the dominant mode and the losses are sufficiently low so that they do not influence the resonant frequency, this representation of the series impedance is justified. As Schelkunoff [5] notes, the condition for resonance for a resonator circuit is that the sum of the transverse resonator impedances of the equivalent circuit model must vanish. For the transfer chain matrix [ABCD] the driving point impedance at the input port may be derived from eq (3) and is
For Zload = 0 (short circuit), Zi, = B / D . Hence, the transverse resonator impedances from the reference plane separating the two chain matrices are
and 22
= Zotanh(yl2)
The resonant condition for this modified reentrant cavity with sample insertion is expressed by 21 2 2 = 0, or
+
In the limit of negligible dissipative losses (or where the contribution of a series resistance/unit length resulting from any dielectric losses in either the specimen or insert is small), the equivalent -transmission line model may be considered to be lossless so that Q = 0. Then eq (16) may be rewritten
or sin [ P ( h
+ L2)] - j-2 cos(PL1) cos(PL2) = 0. 20
(18)
For predicting resonant frequencies we must find P given 2. Conversely, for evaluating E : , ~ from measured dominant resonant frequency f , we determine the root
Dielectric Materials and Devices
195
of eq (18). T h e rms relative uncertainty in evaluating the ferroelectric specimen real permittivity, A c ; , ~ / c :under , ~ , any bias condition is given in terms of relative uncertainties in r s , t,, f, A;nse.rt = T ( T ~- r:), Lz, L2, Ci,insert, and t i n s e r t by
where
is the sensitivity of the specimen real permittivity relative to the parameter p . Equation (19) is valid assuming there are no gaps between the center conductor and sample surfaces as a result of sample metallization, and there is no uncertainty in the characteristic impedance 20. Analytic expressions for the relative uncertainty in specimen permittivity determination have been worked out. Of importance to note is that in this adaptation of the doubly-reentrant coaxial cavity (Fig. 2), we want to keep the capacitive reactance of the dielectric insert small relative to that of the specimen. This minimizes radiative losses from the cavity and maximizes the change in cavity resonant frequency relative to change in specimen permittivity under bias. Minimization of the capacitive reactance of the dielectric insert can be accomplished by using a thin ( t i n s e r t < 1 mm), high-permittivity > lSO), commercially available ceramic material for the insert.
Doubly-Reentrant Coaxial Cavity Having Lower Center Conductor Dielectrically Insulated with Bypass Capacitor The equivalent circuit model used for the modified doubly-reentrant coaxial cavity with bypass capacitor is given in Fig. 7. This adaptation of the doubly-reentrant coaxial cavity for ferroelectric tunability measurements has the advantage that the cavity remains a closed structure so there are no radiated fields. In this case, the first chain matrix of the equivalent circuit is
196
Dielectric Materials and Devices
-
Reference Plane 21-
'
I I
22
-
I
Figure 7: Equivalent circuit of modified doubly-reentrant cavity shown in Fig. 3 for rf tunability measurements of ferroelectric materials.
+
+
so that A1 = cosh(yl1) % s i n h ( y l l ) , B1 = Zosinh(yl1) Z,cosh(yL1), C1 = s i n h ( y l l ) , and D1 = cosh(yL1). The specimen series impedance is given in terms of the capacitive reactance,
-&
The second chain matrix is
where the series impedance of the bypass capacitor is
and t b y p a s s , and Abypass are the thickness and area of the bypass capacitor. The eigenvalue equation is now derived by setting the sum of the 2 transverse resonant
Dielectric Materials and Devices
197
system impedances equal to 0,
If we consider the limit of negligible dissipative losses and simplify the second term in eq (26), there results
Given the measured cavity resonant frequency (or P ) for an applied bias voltage, we solve eq (27) for 2, or E : , ~ . An uncertainty analysis similar t o that described by eqs (19) and (20) has been performed. For this measurement system, we also want the reactance of the ferroelectric specimen t o dominate the reactance of the bypass capacitor. This was accomplished for 0.25 mm nominal thickness specimens having relative permittivities as low as 100 by using a 1 mm thick ceramic bypass capacitor having a relative permittivity of 160. T h e bypass capacitor has an estimated capacitance of 2200 pF, while that of the samples tested varied between 50 and 150 pF. Resonant Frequency Power Law Dependence on Specimen Permittivity
Either eq (18) or eq (27) may be used t o evaluate resonant frequency as a powerlaw function of specimen permittivity, where li' is a constant and L describes the power law dependence of resonant frequency for a given specimen permittivity. In this case, (29)
Since A f /A€:,, w d f
yielding the following simple equation for tunability as a function of the power law fit parameter x and cavity resonant frequency shift A f,
Af tunability = xf '
198
(31)
Dielectric Materials and Devices
Table 1: VHF and 10 kHz Tunability (%) Comparison at 2 v/pm. Material
~
Ba.50Sr.50TiO3 Ba.55Sr .45 Ti 0 3 Ba.55Sr.45Ti03 Ba.55Sr.45Ti03 Ba GnSr 40Ti03
Weight Percent Added MgO 20 20 30 40 20
Tunability 10 kHz 4.6 14 12.8 14.3 23.8
Tunability VHF 4.7
15 15 12.5 18
Equation (31) was used in the evaluations of tunability as a function of observed normalized frequency shift.
RESULTS Because the modified doubly-reentrant cavity with an insulated lower center conductor post is non-radiative, this cavity was used for ferroelectric tunability at rf frequencies. The results are summarized in Fig. 8 for various ceramic composites of Ba,Srl-,TiO3 and added weight percent SlgO. A comparison between measured tunabilities a t VHF frequencies with those measured a t 10 kHz and 297 K is shown in Table 1. Total estimated relative uncertainties in measured tunabilities are less than 5%. For the 24 "C measurements shown, tunability increases as Ba:Sr stoichiometry increases, whereas increased weight percents of the added non-ferroelectric oxide MgO, decreases dielectric loss and tunability. Generally, tunability decreases with increasing temperature. This is expected as the measurement temperature goes further from the ferroelectric. as opposed to the paraelectric state, of the tested specimens. The Curie temperatures for the specimens and dielectric properties at 10 GHz in the unbiased state are given in Table 2. Tunability for all tested specimens has not saturated, that is, it is still increasing for applied bias voltages as high as 4v/micrometer. Microstructural analysis using scanning electron microscopy shows that connectivity between BST-BST grains versus BST-MgO grains increases tunability (Figs. 9, 10, 11). Ba,Srl-,TiO3 grain segregation occurred for all composites having 20 percent or greater by weight of added MgO [6]. Tunabilities a t measured VHF frequencies compared well with those measured a t low frequencies and 297 K [6,7], although caution must be taken in inferring that the tunabilities would be the same at much higher frequencies.
Dielectric Materials and Devices
199
45
+
40
-E
35
T = 24.5
30
a r-'
/ / /
"C
324MHz
25
F
/
VHF Frequency Range:
) I
c. .-
-
e E a Sr TiO 140 waant % MgO Ba zSr,::TiO>M % MgO +Ba,Sr,TiO,/ZQ weqm % MgO -Oa Sr TiO /30wagnt % MgO *Ba~Ssr,~TiO~/20 wagnt % MgO / /
d
1
05'
, 0
i.
(0
2
20 15
10 5 0
0
0.5
1
1.5
2
2.5
3
3.5
4
Tuning Bias (v/pm)
Figure 8: Radio frequency tunabilities as function of applied tuning bias for various barium strontium titanatelMg0 ceramic composites.
Table 2: Microwave Dielectric Property Measurements on Unbiased Ba,Srl-,TiO3 and MgO Ceramic Composites at 10 GHz 2nd 23OC. [6] Material Ba.50Sr.50TiOs Ba.50Sr.50Ti03 Ba.50Sr.50TiOs B a.50SrsoTi03 B a.55 Sr.45TiO 3 B a.55 Sr.45TiO 3 B a.soSr.4oTiOs
200
Weight Percent Added MgO 0 20 30 60 30 60 60
Curie Temperature "C -25 - 55 - 50 -15 -45 -50 -55
Relative Permittivity 1099 6 16 463 84.5 527 99.8 118
I
Loss Tangent 1 . 8 0 10-2 ~ 8 . 6 7 ~10-3 8 . m 10-3 6 . 5 5 10-3 ~ 1.21x 10-2 7 . 9 5 ~10-3 1 . 2 9 ~10-2
Dielectric Materials and Devices
Figure 9: Bw.ssSro.45Ti03 with 20 weight percent MgO, 2000x. Dark phase contains primarily Mg and 0 [6].
Figure 10: Ba~.5&0.45Ti03 with 40 weight percent MgO, 2000x [S]. weight percent of MgO decreases BST-BST connectivity.
Dielectric Materials and Devices
Increased
20 1
Figure 11: B%.ssSr~..r~TiOs with 60 weight percent MgO, 2000x [ 6 ] .
SUMMARY Two new measurement techniques for performing variable-temperature ferroelectric tunability measurements at I'HF and UHF frequencies have been developed. Both involve adaptation of doubly-reentrant coaxial cavities, and both measure tunability of a bulk ferroelectric specimen in a direction parallel to applied bias field. In order to analyze the relation between the measured dominant resonant frequencies of these cavities and permittivity of the ferroelectric specimen under an arbitrary tuning voltage, transfer (chain) matrix equivalent circuit models were employed. These computational efficient transmission line models are able to predict resonant frequencies with 5% uncertainty for specimen relative permittivities as high as 300. Tunabilities for studied Ba,Srl-,TiO3 and MgO composites appear limited by BSTBST grain connectivity.
ACKNOWLEDGMENTS We wish t o acknowledge Paratek, Inc. and the Army Research Laboratory for providing test specimens. Tl'e also thank Dr. Stuart Wolf for many useful discussions on rf characterization of ferroelectric materials.
202
Dielectric Materials and Devices
REFERENCES R.N. Ghose, Microwave Circuit Theory a n d Analysis, McGraw-Hill, New York, 1963.
J .C. Slater, Microwave Transmission. SlcGraw-Hill, New York, 1942. 3C.G. Montgomery, R.H. Dicke, and E.M. Purcell, Principles of Microwave Circuits. IEE Electromagnetic Wave Series 25. Peter Peregrinus, London, U.K., 1987.
4K.F. Sander, Microwave Components and Systems. Addison-Wesley, New York, 1987. 5 S . i l . Schelkunoff and H.T. Friis: Electromagnetic Waves, D. Van Nostrand. Yew York, 1943.
6 J . Synowczynski, G. Dewing, and R.G. Geyer, “Acceptor Doping of Barium Strontium Titanate and Magnesium Oxide Composites,” Proc. Am Cer. Soc.: Dielectric Materials and Devices, 2000. Private communication, Paratek Microwave. Inc.
Dielectric Materials and Devices
203
This page intentionally left blank
HIGH-DIELECTRIC-CONSTANT CERAMIC-POLYMER 0-3 COMPOSITES Y. Bai, Z.-Y. Cheng, V. Bharti, H. S. Xu, and Q. M. Zhang Materials Research Laboratory Pennsylvania State University University Park, PA 16802
ABSTRACT: A ceramic-powder polymer composite, making use of a relaxor ferroelectric polymer that has a high room-temperature dielectric constant as the matrix, is developed. The experimental data show that the dielectric constant of the composites with Pb(Mgl/3Nb2/3)03-PbTi03powders can reach more than 250 with weak temperature dependence. In addition, the composites under a proper preparation procedure exhibit a high breakdown field strength (>120MV/m), leading to a maximum energy storage density of more than 15 J/cm3. Experimental results also indicate that the high electron irradiation does not have much effect on the dielectric behavior of Pb(Mg1/3Nb2/3)03-PbTi03 powders, possibly due to the relaxor nature of the ceramic. INTRODUCTION By integrating two or more materials with complementary properties, composite materials offer the potential to have performance far beyond those of the constituent materials. For instance; ferroelectric ceramics possess very high dielectric constant but are brittle and have low dielectric strength. On the other hand, polymers are flexible, easy to process with low processing temperature, and possess high dielectric breakdown field. By combining these two, one may be able to develop a new material with high dielectric constant and high breakdown field to achieve high volume efficiency and energy storage density for applications of capacitors and electric energy storage devices. In the past twenty years, a great deal of effort has been devoted to the development of ceramic powder polymer composites (0-3 composite^).^.^ However, due to the low dielectric constant of polymer matrix (usually below lO), the dielectric constant of 0-3 composites developed to date is at the level of about 60 at room temperat~re.~?~.~
'
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this ublication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright {Iearance Center, is prohibited.
Dielectric Materials and Devices
205
Recent research on P(VDF-TrFE) copolymer found that high energy electron irradiation with proper dosage can increase its room temperature dielectric constant to more than 50 over a relative broad temperature range? In addition, the high energy irradiation converts the copolymer from a normal ferroelectric into a relaxor ferroelectric, which removes the large polarization-field hysteresis under high field, typical in the normal ferroelectric copolymers. These features provide a great opportunity for the development of high dielectric constant 0-3 composites. In this paper, we report the results of the development and characterization of 0-3 composites based on the irradiated copolymer as the polymer matrix. We will show that the composite thus developed has a muchimproved dielectric constant. In addition, under a proper processing condition, the breakdown field of the composite can reach more than 120 MV/m.
EXPERIMENT The P(VDF-TrFE) 50150 mol% copolymer (purchased from Solvay and Cie, Belgium) was chosen for the polymer matrix since it can be easily converted into a relaxor under relatively low irradiation dose.7 Pb(Mg 1/3Nb2/3)03-PbTi03(PMNPT) ceramic powder (PMN-85, TRS Ceramics, Inc.) which is also a relaxor ferroelectric with high room temperature dielectric constant was used as filler.' The composite was prepared using solution cast method. P(VDF/TrFE) copolymer was dissolved in methyl-ethyl ketone, and a proper amount of PMNPT ceramic powder (average particle diameter is 0.5 pm) was added into the solution, which was thoroughly mixed with the solvent. The suspension was then poured on to a glass plate to remove the solvent (at room temperature for 1 hour), resulting in a composite film of about 15 pm thick. All these were carried out in a clean bench (class 1000). The film was then heated in a vacuum oven at 70°C for 12 hours to further remove any remaining traces of the solvent. Then it was folded to an average area of 3x2 inches and melt pressed at 170°C (above the melting point of the copolymer) under 15,000 lbs of force, which was used to remove any possible pores in the composite. It was found that only by following this procedure, the breakdown field of the composite can reach more than 120 MV/m. SEM micrograph shows that the composites thus prepared have a uniform ceramic powder distribution in the polymer matrix. The typical thickness of the pressed composite was about 20 microns. Finally, the composite was annealed at 140°C in vacuum for 12 hours and slowly cooled down to room temperature. Composites with volume percentage of the ceramic from 10-60% were prepared and irradiated at 120°C with the energy of the electron source of 2.55 MeV and various dosages (40, 60 and 80 MRad). It should be noted that for such a high energy, the penetration depth of the electron is more than 0.2 mm into a lead plate (more than 90% of the electron energy is still remained).' Hence, for the
206
Dielectric Materials and Devices
composites with 20 microns thickness, the electron will pass through the material without much absorption by the material. For the electric characterization, the films were cut into small pieces of 5x5mm and circular gold electrodes with 3 mm radius were sputtered in the center on both sides of each sample. The dielectric properties as a function of temperature were measured at frequencies from 100 to 100 kHz using a dielectric analyzer (DEA 2970, TA Instrument). The frequency dependence of the dielectric constant and loss at a constant temperature was measured by means of an impedance analyzer (HP 4194A, HP) from 1 kHz to 100 MHz. In order to evaluate the breakdown strength, the sample was immersed in silicon oil and a DC voltage was applied using a high voltage supply (Trek, 610D).
RESULTS The composites prepared before the irradiation were characterized and figure 1 shows the dielectric constant measured at 100 Hz as a function of the volume percentage of ceramic powder (un-irradiated). As expected, the dielectric constant increased with the volume fraction of ceramic filler. In the past, there have been many efforts in developing models for 0-3 composites.'Oil It was found that the expression developed by Yamada et al. can fit the data well (solid curve in the figure),
where K is the dielectric constant of the composite, K, and & are the dielectric constants of the polymer matrix and the ceramic, respectively, q is the volume fraction of ceramic and n is a parameter related to the geometry of ceramic particles." For the copolymer, K, at room temperature is 17, which is directly measured at 100 Hz. The values of & and n obtained from the fitting to equation (1) are 1400 and 10.6, respectively (n value is also in agreement with that found in reference 10). After irradiated with 40 Mrad, K, is increased to 46, while n remains the same since it is only related to the geometry of ceramic powders. & calculated using equation (1) from the dielectric constant of the irradiated composites is 1360. This result indicates that the irradiation does not have a significant effect on the dielectric properties of the ceramic filler. In addition, PMN-PT bulk ceramics with the composition similar to that used in the composite were also irradiated and the dielectric constant before and after the irradiation does not show much change. This could be due to the fact that PMN-PT is a relaxor ferroelectric that already possesses frozen in defects, and therefore, the additional defects induced by the irradiation do not have as much effect in the dielectric properties as that on normal ferroelectric ceramics, where the irradiation effect seems more significant.l 3
Dielectric Materials and Devices
207
Figure 1 Variation of E (f=100Hz) with the volume percentage of the ceramic at 25°C
Figure 1 also reveals that when volume percent of the ceramic powder increases to 60%, the measured dielectric constant of composite becomes much lower than that predicted from eq. (1). This is probably caused by two factors. First, the high volume fraction of ceramic filler in composites may result in an increase in porosity. In addition, the high volume percent of ceramic may lead to agglomeration of powders, which results in a non-uniform distribution of the ceramic powder. Therefore, in this study, only the composites with 50% volume fraction of ceramic powder were chosen for further investigation. 350-
0.08-
300 -
0.06U v)
)
.
5 0.04-
0.02
50 I
0
.
,
20
.
,
40
,
I
60
.
,
80
Ternpe rature ("C)
T
.
100
.
0
,
20
.
,
40
.
,
.
60
,
.
80
100
Ternpe rature ("C) (a) Relative dielectric constant (b) Loss Figure 2 Effect of irradiation on the dielectric properties (1kHz) of PMNPT/P(VDF-TrFE)composite
Presented in figure 2 is the dielectric constant (1 kHz) as a function of temperature for irradiated composites with 50% ceramic volume content. The data shows that over a relatively broad temperature range, the dielectric constant is quite high and exhibits a weak temperature dependence. Furthermore, by adjusting the dosage, the level of the dielectric constant and the flatness of
208
Dielectric Materials and Devices
temperature dependence can also be varied, as can be seen in figure 2, where the dielectric constant for composites irradiated at different dosages is shown. In comparison, the dielectric constant of unirradiated composite is also presented, which shows a stronger temperature dependence. In addition, the copolymer matrix is still a normal ferroelectric and under high fields the polarization-field curve of the composites exhibits large hysteresis. .
250
lk
.......I
. ......., . ......., . ......., . . . .
0.35
40
........ ........ ................. .......Jo.00 I
I
10k
lOOk
I
1M
10M
lOOM
Frequency (Hz)
k'
(a) Frequency dependences of dielectric constant and tan6
(b) Cole-Cole representation of complex permittivity
Figure 3 Dielectric behavior of PMN-PT/P(VDF-TrFE)composite at high frequency
A& 1+ (iot>* Frequency dependence of the room temperature dielectric constant and tan6 of the composite irradiated with 4OMrad at 120°C are shown in figure 3 (a). It is evident that the dielectric absorption of the composite with a maximum near 1 MHz is a simple relaxation process, as shown in Figure 3 (b), which can be fitted quite well with the modified Cole-Cole e q ~ a t i o n , ' ~ &=E,-
K=K,+
AK
(2) l+(iwz)" ' which yields Ks50.832, AK=173.2, a=0.484, ~=61.69ps, indicating that the composite has a dielectric constant of 50 at 100 MHz and this value is comparable to those of current materials used in microwave application^.'^ Moreover, with increased temperatures, the relaxation frequency l h moves progressively to higher frequencies as shown in figure 4, resulting in a higher dielectric constant at high frequencies. For instance, the dielectric constant of the composite can be higher than 150 at 10 MHz when measured at 90" C and above.
Dielectric Materials and Devices
209
- 0.5
0.5
Y
100
lk
10k
lOOk
1M
10M
Frequency (Hz) (a) Frequency dependence of relative dielectric constant at various temperatures
lOOM
0.4
- 0.4
0.3
-
0.2
- 0.2
0.1
- 0.1
100
lk
10k
lOOk
1M
10M
0.3
100M
Frequency (Hz) (b) Frequency dependence of loss at various temperatures
Figure 4 Dielectric behavior of PMN-PT/P(VDF-TrFE)composite at various temperatures (1kHz)
We also characterized the irradiated composites for possible applications of electric energy storage devices. In that application, the maximum stored energy density is an important parameter and for a linear dielectric as the composites studied here, the maximum stored energy per unit volume is 1 K&()E,, 2 U= (3) 2 where K is the relative dielectric constant and E, is the maximum field, which can be ap lied to the material (proportional to the breakdown field of the material)." For dielectric materials, it is well known that the breakdown field will depend on sample thickness and in general, it will increase as the thickness is reduced (due to the avalanche phen~menon).'~"~ For the thickness studied here (-20 pm), the breakdown field for the irradiated copolymer is about 350 MV/m while for the ceramic, it is below 10 MV/m.'*'19 In the 0-3 composites, the complicated geometry makes it difficult to predict exactly the level of the breakdown field and no systematic thickness dependence of the breakdown field was observed. Instead, in this study it was found that the main causes for the breakdown are due to the extrinsic effects such as inclusion of the air bulbs, dust, and residual solvent in the composites. For instance, by preparing the composites in a class 1000 clean bench rather than in a normal environment raised the breakdown field from 80 MV/m to 120 MV/m. Using K of 250 and E,,,=l20 MV/m, equation (1) yields that the maximum stored energy of the corn osite is more than 15 J/cm3 higher than those reported from the current literatures.To
CONCLUSIONS
In the present study, a relaxor-relaxor composite (both PMN-PT ceramic and irradiated copolymer are relaxor ferroelectric materials) was studied. Making use of the high dielectric constant polymer matrix in a recently developed irradiated
210
Dielectric Materials and Devices
P(VDF-TrFE) copolymer, a high dielectric constant 0-3 composite has been developed. It was also observed that the properties of the composite can be changed by varying the irradiation conditions.' The dielectric constant of the composite can be varied from 120 to 350 and the transition temperature can be shifted from 65 to 35" C. A relatively flat dielectric response from 0 to 100" C can be achieved. The composite also has high dielectric constant (>120) at lOMHz and 90°C, which gives the material an opportunity to be used in high frequency applications. The high dielectric strength of the material prepared in clean environment leads to a high-energy storage density of 14J/cm3.
REFERENCES 'R.E. Newnham, Ann. Rev. Mater. Sci. 16,47 (1986). 2C. J. Dias and D. K. Das-Gupta, in Ferroelectric Polymer and CeramicPolymer Composites, edited by D. K. Das-Gupta (Trans Tech Publications Ltd., Switzerland, 1994), p. 217. 3C.J. Dias and D. K. Das-Gupta, IEEE Trans. Electr. hsul. 3,706 (1996). 4K. A. Hanner, A. Safari, R. E. Newnham and J. Runt, Ferroelectrics, 100,255 (1989). 5R. Gregorio Jr., M. Cestari and F. E. Bernardino, J. Mater. Sci. 31, 2925 (1996). 6H. L. W. Chan, W. K. Chan, Y. Zhang and C. L. Choy, IEEE Trans. Electr. Insul. 5, 505 (1998). 7Q.M. Zhang, V. Bharti and X. Zhao, Science, 280,2101 (1998). 8L.E. Cross, Ferroelectrics, 76,241 (1986). 9J. G. Trump, R. J. Van de Graaff and R. W. Cloud, Am. J. Roentgenol. & Rad. Therapy, 40,728 (1940). 'OT. Yamada, T. Ueda and T. Kitayama, J. Appl. Phys. 53,4328 (1982). "D. K. Das-Gupta, Ferroelectrics, 118, 165 (1991). 12T.Furukawa, K. Ishida and E. Fukada, J. Appl. Phys. 50,4904 (1979). 13J.Gao, L. Zheng, J. Zeng and C. Lin, Jpn. J. Appl. Phys. 37, 5 126 (1998). 14B.K. P. Acaife, Principal of Dielectrics (Clarendon Press, Oxford, 1989). 15T. Laverghetta, Microwave Materials and Fabrication Techniques (Artech House, Boston, 1991). 16H. Frohlich, Theory of Dielectrics, Ch. 1 (Oxford University Press, London, 1958). 17R.Gerson and T. C. Marshall, J. Appl. Phys. 30, 1550 (1959). 18J.F.Scott, Ferroelectric Review, 1, 1 (1998). 19E.Furman, Ph.D Thesis, Pennsylvania State University (1987). 20E. Aulagner, J. Guillet, G. Seytre, C. Hantouche, P. Le Gonidec and G. Terzulli, Proc. of IEEE 5th Intl. Con$ on Conduction and Breakdown in Solid Dielectrics, Leicester, UK (IEEE, Piscataway, 1995) p.423.
Dielectric Materials and Devices
21 1
This page intentionally left blank
PHASE CONSTITUTION AND MICROWAVE DIELECTRIC PROPERTIES OF THE ZNNB206-T102 SYSTEM Dong-Wan Kim, Hyuk-Joon Youn, Seo-Yong Cho, and Kug Sun Hong Seoul National University School of Materials Science and Engineering Seoul, 151-742, Korea ABSTRACT Phase constitution and microwave dielectric properties of ( I -x)ZnNb,O,xTiO, were investigated using X-ray powder diffraction and a network analyzer. Four regions were observed: columbite, ixiolite (ZnTiNb,O,), two-phase region of ixiolite and rutile, and rutile solid-solution region. zf could be controlled to around 0 ppm/”C in the two-phase region. The microwave dielectric properties of samples with the same composition showed considerable variations by sintering temperature, time and atmosphere. The change in relative amount of ixiolite and rutile was found to be responsible for that variation. Therefore, microwave dielectric properties of the ( I -x)ZnNb,O,-xTiO, depend mainly on their phase constitution rather than chemical compositions. INTRODUCTION Several synthetic compounds are known to crystallize with a columbite structure, e.g. ANb,O, (where A = Mg, Mn, CO,Ni, Zn).’.2The structure of columbite can be interpreted as an ordered super-structure of a-Pb0,.3-5TiO, has a rutile structure which transforms into the distorted hexagonal close packed aPbO, structure, with the application of external pressure., The close relation between columbite and rutile is attributed to various phase transitions of their solid solution. In the ZnNb,O,-TiO, system, the Ti4’ cation can substitute into the columbite structure, which may cause cation d i ~ o r d e r . ~ , ’ ~ ~ Interest in microwave dielectric materials of high dielectric constant and low dielectric losses continues to grow due to their applications in the telecommunications industry. One of the most important characteristics of microwave dielectric materials is the temperature coefficient of resonant frequency (23. Recently, AB,O, compounds (A = Mg, CO,Ni, Zn and B = Nb, Ta), which are a subcomponent of the complex perovskite, A(B’,”B’’2,3)03were To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
213
found to be very low loss materials, and were investigated by Lee et Among these compounds, ZnNb,O, had a very high quality factor, i.e. 83700. However, E, is small, 25 and zf has a negative value, -56.1 ppm/"C. In contrast, the E, and zf of TiO, are very large and positive, i.e. 100 and 400 ppm/"C, respectively." Our previous work was to compositionally tune E, and zf of ZnNb,O, by mixing it with Ti02.12In mixture region of ixiolite and rutile phase, z, was modified to around 0 ppm/"C. The structural evolution of the ZnNb,O,-TiO, system and the structure-property relationships in the mixture region were investigated by controlling sintering time, temperature, and atmosphere, in this study. EXPERIMENTAL The starting raw materials were ZnO (Seido, Japan), Nb,O, (High Purity Chemical Lab., Japan), and TiO, (Merck, Germany) powders with 99.9% purity. ZnNb,O, powders were prepared using conventional mixed oxide method and calcined at 1000°C for 2 h. Mixtures Of ZnNb,O, and TiO, powders of varying composition were ball-milled in a polyethylene bottle with ZrO, balls for 24 hours using distilled water as a medium. The milled powders were then dried, granulated, and pressed at 1,000 kg/cm2to form pellets 8 mm in diameter and 3 mm thick. The pellets were sintered at 1200-1425°C for 0-28 h in air, O,, and N,. The crystal structure of sintered samples was investigated using X-ray powder diffraction (Model M 18XHF, Macscience Instruments, Japan) in the 28 range of 20 to 60". Step scanning with step size 0.006" and counting time 1 s/step was used for analyzing rutile (1 10) and ixiolite (1 11) reflections. Measured data was deconvoluted into two single waveforms. Integral intensities of each reflection were measured by counting the total area under the deconvoluted peak and then subtracting the background. Microwave dielectric properties of sintered samples were measured using a network analyzer (Model HP8720C, Hewlett Packard, USA) in the frequency range of 5-1 1 GHz. RESULTS AND DISCUSSION In the (1-x)ZnNb,O,-xTiO, samples sintered at 1250°C for 2 h for various values of x, four phase regions were observed with increasing amount of TiO,." The upper limit of a first broad region, the solid solution based on the structure of columbite, was observed up to approximately 40 mole% TiO,. At higher TiO, content, only the ixiolite phase, ZnTiNb,O, appeared. The unit cell volume of the ixiolite structure is one third that of the columbite structure, which means that ixiolite is a disordered modification of ZnNb,O,.* In our system, the ixiolite structure has a statistical distribution of the three metal atoms, Zn, Nb and Ti on
214
Dielectric Materials and Devices
the cation sites of the columbite structure. Between 57 and 60 mole% TiO,, a narrow mixture region of ixiolite and rutile was observed. It is of interest that the ( 1-x)ZnNb,O,-xTiO, system, for this composition region, exhibits a morphotropic phase transition. The fourth solid solution region between 70 and 100 mole% TiO, exhibited a rutile structure. In the (1 -x)ZnNb,O,-xTiO, system, the interrelations of the columbite and rutile structures lead to various structural transitions and broad ranges of solid solutions with composition*, as listed in Table I. Table I. Structural transition of (1-x)ZnNb,O,-xTiO, system. Structure ComDosition (mole fraction of TiO,, x) Columbite 0 5 x < 0.5 0.5 5 x 50.54 Ixiolite Ixiolite and Rutile 0.54 < x 5 0.6 Rutile 0.6 < x 5 1 40000 35000 30000 25000 20000
0
0.42ZnNb,06-0.58Ti0,
6 ;;g;
L
5000 60 0
A
55 50 W.
45 148
L
90 50 70 30 -10 10 -30 -70 -50
O
1200
1250
1300
I
1350
Sintering temperature ("C)
Figure 1. Microwave dielectric properties of (l-x)ZnNb,O,-xTiO, (0.56 I x I 0.6) as a function of sintering temperature. Figure 1 shows microwave dielectric properties of ( 1-x)ZnNb,O,-xTiO, specimens as a function of sintering temperature. In the narrow mixture region of both ixiolite and rutile structures (0.54 < x 50.6), the quality factor sharply decreased with increasing TiO,. However, the relative dielectric constant (E,) and the temperature coefficient of resonant frequency (q)increase linearly with increasing TiO, content in the mixture region. It is noteworthy that zf value is
Dielectric Materials and Devices
215
around 0 ppm/"C. It is postulated that the microwave dielectric properties depend on crystal structure rather than chemical composition in the mixture region.
0.42ZnNb,O6-0.S8Ti0, sintered a1 1200°C for 2 h (e) in air. (0 in NI. ( a ) in 0,
0.42ZnNb20,-0.58Ti02 quenched at 1250°C (a) 0 h. (b) 2 h (a) 18 h. (b) 28 h
lxiolite ( I I I )
Rutilc(l10)
L 1: I
26
27
28
30
29
31
26
27
.
I
.
I
28
28 CuKa
.
l
30
29
.
1
31
28 CuKa
Figure 2. XRD patterns of the 0.42ZnNb,06-0.58Ti0, sintered at 1250°C quenched at preset time: (a) 0 h, (b) 2 h, (c) 18 h, (d) 28 h in air, and sintered at 1250°C for 2 h in (e) air, (f) in O,, (g) in N,.
Ial I2OO"C.Ib) I250"C Ial 1300°C.Ib) 1350°C
Rulilc I 110)
lxiolilc I I I I1
(c)
(b)
A A
(a) A
I
>
27
.
I
.
28
I
29
28
.
1
30
.
I
31
CuKa
Figure 3. XRD patterns of the 0.42ZnNb20,-0.58Ti0, as a function of sintering temperature. (a) 1200"C, (b) 1250"C, (c) 1300"C, (d) 1350"C, (e) 1400 "C, and (f) 1425°C
216
Dielectric Materials and Devices
Figure 2 shows the XRD profiles of the 0.42ZnNb,0,-0.58Ti02 samples for various sintering conditions. The change in relative amount of ixiolite and rutile was found. The relative integrated intensities of the rutile (1 10) and ixiolite (1 1 1) peaks were increased with increasing sintering time at 1250°C [Fig. 2(a)-(d)]. / ( Imtile~llO) + Iixiolite(lll) ) values of the quenched samples at 1250°C for 0 h Irutile(,lO) and 28 h were 0.16 and 0.2 1, respectively. It was also observed that the relative integrated intensity of the sample sintered in N, was increased and that of the sample sintered at 0, was not. From these results, the phase constitution of sintered samples in the mixture region was changed with sintering time and sintering atmosphere.
7 -
5
-+-
0.7
-
0.6
-
0.5
-
0.4
-
2
v
$
.Ili
.
0.3
w-
U
1 . .
--
-
.
1200
1250
1300
1350
1400
1450
Sinterine. temperature (“C)
Figure 4. Relative integrated intensity of the rutile (1 10) and ixiolite (1 11) reflection of the 0.42ZnNb206-0.58Ti0, samples as a function of sintering temperature. Figure 3 shows the change in phase constitution as a function of sintering temperature. The structural transition was observed with increasing sintering temperature in same chemical composition. The relative integrated intensities of the rutile (1 10) and ixiolite (1 11) peaks were plotted in Figure 4. Phase constitution was showed more clearly. The amount of rutile phase increases gradually to 1300°C. The relative integrated intensities of the sample sintered at 1200°C and 1300°C were 0.09 and 0.2, respectively. It is of interest that the relative integrated intensities of the sample sintered at the temperature above 1300°C increased sharply. The relative integrated intensities of the melt sample heat-treated at 1425°C increased to 0.68. These results demonstrated that the considerable change in the amount of rutile and ixiolite phase was observed with variation of sintering temperature.
Dielectric Materials and Devices
217
20000
' cu
I
15000 10000 5000
58
52 50 4a
46 44 42
28 30
2
10 -10
A A
Figure 5. Microwave dielectric properties of 0.42ZnNb20,-0.58Ti0, samples as a function of the relative integrated intensities. Figure 5 shows the microwave dielectric properties of 0.42ZnNb20,-0.58Ti0, sintered at various temperatures as a function of the relative integrated intensities. The Qxf of the specimens was decreased, but and zf was increased sharply as the relative integrated intensity increased. The considerable variation of microwave dielectric properties of (1 -x)ZnNb,O,-xTiO, in the mixture region was shown in Fig. 1 as a function of sintering temperature. Therefore, the microwave dielectric properties of ZnNb,O,-TiO, system in the mixture region depend on the phase constitution of ixiolite and rutile. It is noteworthy that q value is adjusted to around 0 ppm/"C in the mixture region of ixiolite and rutile. In Fig. 1, zf value of each compositions sintered at different temperature was modified to around 0 ppm/"C. In these conditions, E, of specimens have value of 44-46. From these results, phase design can be available in the mixture region, and microwave dielectric properties were controlled suitably. CONCLUSION The phase relationships and microwave dielectric properties of the (1x)ZnNb,O,-xTiO, system were investigated. Four distinct phase regions were observed with mole fraction (x) of TiO,: columbite solid solution, ixiolite (ZnTiNb,O,) solid solution, mixture of ixiolite and rutile, and rutile solid solution
218
Dielectric Materials and Devices
region. In the mixture region of ixiolite and rutile, z, was successfully modified to 0 ppm/OC. The amount of ixiolite and rutile in the mixture region was changed by sintering conditions, mainly sintering temperature. The considerable variations of microwave dielectric properties were attributed to the phase constitutions of ixiolite and rutile.
REFERENCES ‘H. J. Goldschmidt, D.Sc., F. Inst. P., “ An X-Ray Investigation of Systems Between Niobium Pentoxide and Certain Additional Oxides”, Metallurgia, 62, 241-50 (1960). ,H.Weitzel, “Crystal structure refinement of Wolframite and Columbite”, 2. Kristalogr, 144,238-58 (1976). 3F. Laves, G. Bayer, and A. Panagos, “Structural Relations to a-PbO,, FeWO, (Wolframite) and FeNb,O, (Columbite) types, and Polymorphs of FeNbO,”, Schweiz. Mineral. Petrogr. Mitt., 43,2 17-34 (1963). 4 Wenger and T. Armbruster, “Phase Relations of Columbite and Rutile-type Compounds in the System NiNb,O,-TiO,”, N. Jb. Miner. Mh., H. 5,224-32 (1993). ’A. Baumgarte and R. Blachnik, “New M2+M4’Nb,0, Phases”, J. Alloys and compounds, 215, 1 17-20 (1994). ,P. Y. Simons and F. Dachille, “The Structure of TiO2(I1),a High-pressure Phase of Ti0,”Acta. Crystallogr., 23,334-36 (1967). 7E.H. Nickel, J. F. Rowland and R. C. McAdam, “Ixiolite-A Columbite Substructure”, Am. Mineral., 48,96 1-79 (1963). 8 A. Baumgarte and R. Blachnik, “Phase Relations in the System Titaniumoxide-Diniobium-Zinc-Hexoxide”, Mat. Res. Bull., 27, 1287-94 ( 1992). ’M. Maeda, T. Yamamura, and T. Ikeda, “Dielectric Characteristics of Several Complex Oxide Ceramics at Microwave Frequencies”, Jpn. J. Appl. Phys. Supp., 26-2, 76-9 (1987). 10 H. J. Lee, K. S. Hong, S. J. Kim, and I. T. Kim, “Dielectric Properties of MNb206 Compounds (Where M = Ca, Mn, CO,Ni, or Zn)”, Mater. Res. Bull., 32 [7], 847-855 (1997). ” S. B. Cohn, “Microwave Bandpass Filters Containing High-Q Dielectric Resonators”, IEEE Trans. Microwave Theory & Tech., 16 [4], 2 18-26 (1968). I2D. W. Kim, D. Y. Kirn, and K. S. Hong, “Phase Relations and Microwave Dielectric Properties of ZnNb206-Ti02”, accepted in J. Mater. Res.
Dielectric Materials and Devices
219
This page intentionally left blank
PROCESSING AND PROPERTIES OF STRONTIUM BISMUTH VANADATE NIOBATES FERROELECTRIC CERAMICS
Y . Wu, M.J. Forbess, S. Seraji, S.J. Limmer, C.P. Nguyen, and G.Z. Cao University of Washington Materials Science and Engineering 302 Roberts Hall, Box 352120 Seattle, WA 98 195 ABSTRACT Strontium bismuth niobate vanadates, SrBi2(VxNb1-x)209 (with 0 1 ~ 1 0 . 3 )were ~ prepared by reaction sintering of powder mixtures of constituent oxides. With partial substitution of niobium by vanadium cations (up to 30 at%), the single-phase layered perovskite structure was preserved and the sintering temperature of the system was significantly lowered (-200-300°C). The incorporation of vanadium into the layered perovskite structure resulted in a shift of the Curie point to higher temperatures from 418 "C to 459 "C with 30 at% vanadium doping. Dielectric constants at room temperature and their respective Curie points were found to peak at a composition with 10-15 at% vanadium. In addition, a high concentration of vanadium (>15 at%) resulted in a significant increase in tangent loss at low frequencies (4000 Hz). The relationships between the chemical composition, processing condition, microstructure, and dielectric properties of SBVN ferroelectric ceramics have been discussed. INTRODUCTION Recently, bismuth oxide layered perovskite materials, such as SrBi2Nb209 (SBN), SrBi2Ta209 (SBT), and SrBi2(Nb,Ta)209 (SBTN) for FeRAM applications have attracted an increasing attention in the research community, because they are fatigue-free and lead-free [ 1-31. There are many efforts reported recently in the open literature to enhance the properties of layered perovskite ferroelectrics by the addition or substitution of alternative cations. For example, Sr2+,in the perovskite unit cells, substituted by Bi3+or Ba2+[4-51. In general, such substitution resulted in higher polarization; however, no thorough explanation was given. Recently, Forbess etc. [6] have also studied the influences of La3+ and Ca2+ doping on the dielectric properties of SBN ferroelectric ceramics. It was found that the doping of La3+and Ca2+resulted in an appreciable increase in the Curie points and a noticeable decrease in the dc conductivity. There are a lot of research reported in open literature [7-101 on solid solutions of SBTN system. However, few work can be found on improvement of ferroelectric properties of the layered perovskite ferroelectrics through substitution of the B site ions (Nb5+or Ta5+)with other alternate cations. In our previous work [ 11- 121, we have proposed and studied the significant enhancement of ferroelectric properties of SBN ferroelectrics through partial substitution of niobium by pentavalent vanadium cations. The enhanced ferroelectric properties of the layered perovskite ferroelectric properties were explained as follows. The partial substitution of Nb5+(ionic radius = 69 pm with CN = 6) with a smaller cations, V5+(58 pm with CN = 6) resulted in an enlarged "rattling space" which lead to both increased spontaneous polarization and reduced coercive field. Further, it was found that the incorporation of vanadium oxide has lowered the sintering temperature by 200 300 "C. In this paper we will present our systematic results on the processing and properties of the SBVN ceramics.
-
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
221
RESULTS AND DISCUSSION The polycrystalline strontium bismuth vanadium niobate ceramic samples with a composition of SrBi2(VXNbl-.J2O9 (SBVN) with x ranging from 0 to 0.3 (30 at%) were prepared by two-step solid state reaction sintering. Details of the sample preparation were described in our previous work [ 1I]. Prior to characterization and property measurements, all the samples were annealed in oxygen at 800 "C for 3 hrs. X-ray diffraction (XRD, Philips 1830) was used to determine the formation of layered perovskite phase after both first-step and second - step firing. After the first-step firing, the powders were analyzed, whereas the sintered pellets were characterized. The XRD spectrum of NaCl crystal was used as a standard to calibrate the scanning angles. The step size of scan is 0.04 /"28 with a scanning speed of 0.004 "28/second. The platinum was sputtered on both sides of the pellets with a thickness of 300 8, and then connecting Pt wires to the surfaces with silver paste and heat-treated on hot plate around 550 "C for 10 minutes. The dielectric constant and loss tangent as functions of temperature up to 600 "C and frequency ranging from 20 Hz to 1 MHz were measured by a HP Precision LCR Meter 4284A. For all the samples, the overall weight loss was found to be of approximately 3 wt%, less than the extra amount of bismuth oxide added into the systems during powder admixing, and is presumably due to the high vapor pressure of bismuth oxide. No evident relationship between the weight loss and the vanadium content was found; however, the presence of vanadium oxide was seen to promote the densification of the SBVN samples appreciably by lowering the sintering temperatures. However all samples, which consist of various amounts of vanadium ranging from 5 to 30 at%, can be sintered to a relative density of 90% or above, at temperatures as low as 900 "C, which is of approximately 300 "C lower than that of SBN. It is noted that in the current study, two step sintering process was applied. During the first step sintering or pre-firing, vanadium oxide reacted with other constituent oxides and formed layered perovskite (as will be discussed in the following section). There would be no or very little reacted vanadium oxide left for the second step sintering. Further, the increase of vanadium concentration from 5 at% to 30 at% was found to have no noticeable influence on the SBVN densification. Furthermore, it was found that a high vanadium concentration does not necessarily lead to a higher density.
ferroelectric ceramics. (a): x=0.3, sintered at 900 Fig. 1. XRD spectra of the SrBi2(VxNbl-x)209 "C, 2 hrs; (b): x=0.25, at 900 "C, 2 hrs; (c): x=0.2, at 900 "C, 2 hrs; (d): x=0.15, at 950 "C, 2 hrs; (e): x=O.l, at 950 "C, 2hrs; (0:x=0.05,at 950 "C, 2hrs; (g): x=O, at 1200 "C, 2hrs.
222
Dielectric Materials and Devices
Fig. 2.
Curie points of the SBVN ferroelectric ceramics as a function of vanadium concentration.
X-ray diffraction analyses (see Figure 1) indicated that single phase layered perovskites were formed within the composition range studied in this work; no secondary phase was detectable. The lattice constants and single unit cell volume were calculated from the XRD spectra with NaCl reference peaks (not shown here) and are plotted as a function of vanadium doping (see Figure 3). The calculated results indicate that the lattice constant, a, is kept unchanged with an increasing amount of vanadium content at less than 15 at%; however, a further increase in vanadium content results in a gradual reduction in value. The above dependence of lattice constant, a, on the vanadium content could be explained by the limited structure constraint induced by the [Bi2O2I2-interlayer between the perovskite-like units. With a low concentration of vanadium, although the V5+(58pm with CN = 6 [13]) is significantly smaller than Nb5+(69pm, CN = 6 [13]), the [Bi2O2I2interlayers prevent the shrinkage of crystal lattice. However, at a high concentration of vanadium, the shrinking tendency of the crystal lattice constants overcome the limited structural constraint from the [Bi20212-interlayers, therefore, the lattice constant, a, decreases with an increasing vanadium content. It is noticed that the lattice constant, c, showing no noticeable change, although the variation along c-axis would be less constrained by the [Bi2O2I2interlayers. As a result, the unit cell volume remain unchanged till vanadium concentration reached 15 at% and then decreases with an increasing amount of vanadium incorporated into the crystal structure. Further, the current study found that a stable single phase layered perovskite was formed with up to 30 at% vanadium doping although the ionic size difference between Nb5' and v5+is approximately 19%. Figure 2 shows the Curie temperature (T,) as a function of vanadium content. The Curie point gradually increases with an increasing amount of vanadium concentration, which could be another indication that a single phase layered perovskite was formed with up to 30 at% vanadium substitution. In general, the increase of Curie points in the system corresponds to reduced unit cell volume. The increase in the Curie points show a slight discontinuity between 15 and 20 at%, however, no explanation is available. It must be noticed that the T, values are about 20 "C lower than that we previously reported [ 111, which is due to the different measurement setup. Figure 3 shows the dielectric constants as a function of temperature for SBVN ceramics consisting of 0, 10, 20 and 30 at% vanadium, respectively, determined at a frequency of lOOMIz with a oscillating amplitude (50 mV). For all the samples, there was a sharp transition in dielectric constant at their respective Curie points. It is interesting to notice that the dielectric constants of SBVN with 10 at% vanadium were significantly larger than that of all other three samples. For a better comparison, dielectric constants at room temperature were plotted as a function of vanadium
Dielectric Materials and Devices
223
concentration, as shown in Figure 4.The dielectric constants were enhanced with the increasing vanadium content up to 15 at% and then decreased with further increasing vanadium content. At high vanadium concentrations, the dielectric constants were lower than that of SBN. Under the measurement conditions (100 kHz and 50 mV) used in the current study, the dielectric constant consists of ionic and atomic polarization only. Since the ionic radius of V5+is smaller than that of Nb5+,and increasing amount of V5+would lead to a reduced atomic polarization. Ionic polarization would be strongly dependent on the lattice constant or unit cell volume. When the concentration of V5+ is less than 15 at%, the lattice constants or unit cell volume remain almost unchanged according to the XRD analysis. As a result, there would be an increased ionic polarization with an increased V5' concentration, due to a combination of an unchanged unit cell volume and a reduced ionic radius. An increase in dielectric constants indicates that the increase in ionic polarization is predominant over the decrease in atomic polarization. However, a higher concentration of vanadium caused a reduction in the lattice constants and unit cell volume. As a result, both atomic and ionic polarization would decrease with an increasing amount of vanadium introduce into the system and, therefore, lead to reduced dielectric constants. Further, Figure 4 also shows that the loss tangent at a frequency of 100 kHz remains unchanged with an increasing concentration of vanadium. 1400
I
0.1
1200
-
1000
5
800
m
0
0
-
'
600 400 200
100
0
200
500
400
300
600
Temperature ("C)
Fig. 3. Dielectric constant as a function of temperature: : x=O, 0: x=O.l, x: x=0.2, +: x=0.3. 200
I
I 0.02
. I I: , , , , , I 100
m-
--c---r
,,=-. .
0.01
5
0.008 5 0.006 I-
'
0.004 0.002 0
0
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
x values
Fig. 4. Dielectric constants and tangent losses as a function of vanadium concentration measured at 100 kHz.
224
Dielectric Materials and Devices
Figure 5 is the dielectric constants as a function of frequency ranging from 20 Hz to 1 MHz at room temperature. The results from four samples were included and the samples consisted of 0, 10, 20 and 30 at% vanadium, respectively. This figure shows that the dielectric constant for samples consisting of 0 and 10 at% vanadium have a small variation throughout the frequency range studied. However the other two samples consisting of 20 and 30 at% vanadium show a drastic reduction in dielectric constant as the frequency increases from 20 Hz to 1 kHz. Under the current experimental conditions (50 mV and 20 -1 MHz), there are three possible contributions to dielectric constant: atomic, ionic, and space charge polarization, since the electric field (0.5 -2 Vkm) is too small to alter spontaneous polarization. Response frequencies for atomic and ionic polarization are 1015and lOI3 Hz, respectively, whereas space charge has a response frequency of approximately 100 Hz. Such a drastic reduction in dielectric constants as frequency increases from 20 Hz to 1 kHz may be attributed to the space charge polarization. At frequencies higher than 1 kHz, the space charge will no longer exist, and thus dielectric constants remain constant as the frequency increases further. From the above discussion, it is reasonable to assume that the space charge is the main reason caused a drastic decrease in tangent loss as the frequency increases from 20 to 1000 Hz. The exact mechanism of increased space charge polarization is not clear; however, it is clear that the high concentration of vanadium is the key factor. More detailed analyses, such as valence state and distribution of vanadium cations, are in progress.
Fig. 5. The dielectric constants and tangent loss of the four typical samples in the SBVN system over frequency at room temperature. V: x=O; v : x=O.l; 0:x=0.2; 0 : x=0.3. CONCLUSIONS The incorporation of vanadium oxide was found to greatly promote the densification of SBN ferroelectric ceramics by lowering the sintering temperature of approximately 300 "C and a single phase with the layered perovskite structure of SBN was obtained with a vanadium concentration up to 30 at%. Although the ionic radius of V5' was significantly smaller than that of Nb5+,there was no noticeable change in the lattice constants or unit cell volume with a vanadium concentration below 15 at%. However, a gradual decrease in the unit cell volume was observed with a further increase in vanadium concentration above 15 at%. The Curie points increase
Dielectric Materials and Devices
225
gradually from -418 “C for SBN to -459 “C for SBVN with 30 at% vanadium. Dielectric constant was found to gradually increase with an increasing concentration of vanadium, reach a maximum at approximately 10-15 at%, and then reduce with a hrther increase in vanadium concentration. This phenomenon was explained mainly by the variation of ionic polarization caused by the change of unit cell volume as the vanadium concentration increases. In addition, it was found that a high concentration of vanadium was likely to introduce space charge polarization into the system, resulting in a drastic decrease in both dielectric constant and tangent loss as frequencies reduced from 20 Hz to 1 kHz. Although an increase in vanadium concentration resulted in an appreciable change in microstructure, no conclusion on the influence of the microstructure on the dielectric properties could be drawn. REFERENCES ‘C.A.P. de Araujo, J.D. Cuchiaro, L.D. McMillan, M.C. Scott, and J.F. Scott, “Fatigue-free ferroelectric capacitors with platinum electrodes”, Nature 374,627 - 629 (1995). 2J.F. Scott and C.A.P. de Araujo, “Ferroelectric Memories”, Science 246, 1400 -1405 (1989). 3G.Z. Cao, “Ferroelectrics and applications (Chapter 3)”, in Advances in Materials Science and Applications, ed. D.L. Shi, TUP and Springer-Verlag, Beijing, in press, (2000) 4T. Atsuki, N. Soyama, T. Yonezawa and K. Ogi, “Preparation of Bi-Based Ferroelectric Thin Films by Sol-Gel Method”, Jpn. J. Appl. Phys. 34,5096-5099 (1995). 5C. Lu and C. Wen, “Preparation and properties of barium incorporated strontium bismuth tantalate ferroelectric thin films”, Mater. Res. Soc. Symp. Proc. 541,229 - 234 (1999). 6M.J. Forbess, S. Seraji, Y. Wu, C.P. Nguyen, and G.Z. Cao, “Dielectric properties of layered perovskite SrBi2Nb209ferroelectrics doped with CaO and La2O3,” Appl. Phys. Lett. (in press) ’S. B. Desu, D. P. Vijay, “c-Axis oriented ferroelectric SrBi2(Ta,Nb2-,)09thin films”, Mater. Sci. & Engr. B32, 83-88 (1995). *S. B. Desu, T. Li, “Fatigue-free SrBi2(Ta,Nb,-,)209ferroelectric thin films”, Mater. Sci. & Engr. B34, L4-L8 (1995). ’K. Kato, C. Zheng, J. M. Finder, and S. K. Dey, Y. Torii, “Sol-Gel Route to Ferroelectric Layer-Structured Perovskite SrBi2Ta2Ogand SrBi2Nb209Thin Films”, J. Amer. Ceram. Soc. 81 [7], 1869-1875 (1998). 10 S. B. Desu, D. P. Vijay, X. Zhang, and B. P. He, “Oriented growth of SrBi2Ta20g ferroelectric films”, Appl. Phys. Lett. 69, 17 19 - 2 1 (1 996). “Y. Wu and G.Z. Cao, “Enhanced Ferroelectric Properties and Lowered Processing Temperature of Layered Perovskite by Vanadium Doping,” Appl. Phys. Lett. 75,2650-2652 (1999). 12 Y. Wu and G.Z. Cao, “Influences of Vanadium Doping on Ferroelectric Properties of Strontium Bismuth Niobates,” J. Mater. Sci. Lett. 15, 267-269 (2000). 13 CRC Handbook of Chemistry and Physics, 61“ edi., edited by R.C. Weast and M. J. Astle (CRC, Boca Raton, FLY1974).
226
Dielectric Materials and Devices
Synthesis, Processing and Dielectric Properties of Compositions in the Strontium Titanate:Strontium Zirconate Solid Solution System S.J. Lombardo;, R.V. Shende, D.S.Viswanath, Department of Chemical Engineering, University of Missouri, Columbia, Missouri 652 11 G.A. Rossetti, Jr., CAMMP, Northeastern University, Boston, Massachusetts 021 15
D.S. h e g e r , A.Gordon, Honeywell, Federal Manufacturing and Technologies, Kansas City, Missouri 64 141
ABSTRACT The processing and dielectric breakdown behavior of compositions in the strontium titanate:strontium zirconate solid solution system have been examined. Because neither orthorhombic strontium zirconate nor cubic strontium titanate is ferroelectric or piezoelectric, these materials are not subject to strong electromechanical deformations that may contribute to dielectric breakdown. However, due to its incipient ferroelectric behavior, strontium titanate exhibits a relatively high dielectric constant. Secondarily, the cations (S?, Zr4+)in strontium zirconate exist in highly stable valence states. These attributes may be desirable in producing perovskite-based dielectrics having moderate dielectric constants and high breakdown strengths. SrZrO3, SrTiO3, and Sr(ZrXTil-,)03powders have been synthesized using the Pechini method, and these powders were used to fabricate capacitors. The dielectric constant and average breakdown strength for SrZrO3 were 40 and 650 V/mil whereas the corresponding values for SrTiO3 were 378 and 550 V/mil. INTRODUCTION Strontium titanate-zirconate (STZ) perovskite materials have a wide variety of applications as ICs, electroluminescentdisplay devices [I], hydrogen sensors [2], oxygen sensors for liquid metal [3], high-temperature proton conductors [4], and as capacitors [5]. Whereas strontium titanate (SrTiO3) is widely used as a substrate for superconductors, as photoelectrodes [6] and for conventional [101 and internal boundary layer capacitors [7], strontium zirconate (SrZrO3) finds application as superlattices [8], and as a high temperature proton conductor for solid oxide fuel cells [9]. As noted above, strontium titanate has found use as a high voltage capacitor material [ 101. Strontium zirconate, however, has not been referred to so far in the literature as a potential candidate for similar applications. Because neither strontium zirconate nor strontium titanate is ferroelectric or piezoelectric, these materials are not subject to strong electromechanical deformations that may contribute to dielectric breakdown. in addition, the cations (S?, Zr4+)in strontium zirconate exist in highly stable valence states. Such attributes may be beneficial for fabricating perovskite-based capacitors having moderate dielectric constants and high breakdown strengths. This investigation reports on the powder synthesis using the Pechini method [11,12,13], tape cast film preparation, and electrical properties of SrZrO3-SrTiO3 compositions. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
227
EXPERIMENTAL AND CHARACTERIZATIONTECHNIQUES Reagent grade chemicals used for the synthesis were citric acid, ethylene glycol, zirconium acetate (A), and strontium nitrate (B), and the chemicals used for powder processing were 2propanol, and poly(vinyl)butyral, which were all obtained fiom Aldrich. TyzoPGBNGBOtitanium acetyl acetonate (C) was obtained from Dupont. The surfactant DuramaxTMD3005 was obtained from R o b and Haas. The analysis of the precursors (A, B and C) was performed by thermal gravimetry on TGA 7 Perkin Elmer 1020 series analyzer. For the synthesis of SrZr03,a stoichiometric amount of zirconium acetate based on 12.75 g of strontium nitrate was dissolved in 250 ml of de-ionized water in a 3-liter pyrex glass beaker. A strontium nitrate solution (12.75 g/ 100 ml of de-ionized water) was prepared separately, and added in 5 parts to the previously diluted zirconium acetate solution at 3540°C over 15 min. A citric acid solution in water (20 g/250 ml) was prepared and added to the mixture of zirconium acetate and strontium nitrate over 15 min. The mixture was stirred continuously at 80-90°C for about 1 hr. To this solution, ethylene glycol (18.6 ml) was added over 30 min and heated till a glassy transparent mass was formed. The solid mass was gently powdered in a mortar and then charred at 400°C and calcined at various temperatures ranging fi-om 600-1200°C. SrTiO3 and Sr(ZrxTilJ03 compositions were prepared in a similar manner with use of the precursor titanium acetyl acetonate. Slurries in water (40-55% v/v solid loading) and in propanol (40-65% v/v solid loading) were prepared by milling for 48 hrs. Poly(viny1)butyral was used in the concentration of 8-10 % (w/w) for the non-aqueous slurries. Formulations were tape cast on Mylar film and dried initially in propanol atmosphere and subsequently in an oven at 105°C for 16 hrs. The tape cast films made using aqueous formulations were initially dried at 25°C and then at 85°C for 12 hrs. The films were sintered at temperatures of 1300-1560°C in a Thermolyne 46100 furnace. The fired density was determined using the sample weight and dimensions and the Archimedes principle. whereas the green density was established by the former method only. The synthesized powders were characterized by x-ray diffraction using a Bruker D5005 8/28 Bragg-Brentano difhctometer using CuKa radiation. The difhctometer was equipped with a curved, graphite-crystal diffracted beam monochromator and a NaI scintillation detector. The instrument was calibrated using NIST standard reference materials SRM 660 ( L a 6 powder), SRM 1976 (a-A1203plates). The diffraction data was collected in the step scanning mode over an interval of 20-60" 28, using a step increment of 0.02" 28 and a count time of 1 sechtep. Specimens were mounted on the plate using a fugitive liquid (2-propanol) for dispersion to produce a reasonably flat specimen with sufficient integrity to enable analysis at any diffraction angle. Analysis for phase composition was carried out by comparing the data for experimental specimens to that of reference spectra complied by the International Centre for Diffiction Data (ICDD), to that taken on material commercially available fiom Aldrich, and also to x-ray patterns published by other invzstigators [2,14,15,16]. The dielectric constant (K), dissipation factor (tan 6), and conductance fiom 102-10' Hz were measured using an HP 4263B LCR meter. For breakdown testing, the monolithic films (4-6 mil thick) were electroded by gold sputtering (-10 nm on each side) and tested in Fluorinert. RESULTS AND DISCUSSION The initial synthesis of strontium zirconate was conducted by using the nominal two moles of water of hydration present with ST(NO~)~, and on this basis the stoichiometric amount of zirconium was calculated. The reaction mixture was processed as described in the Experimental section and the material was calcined at 900°C for 12 hr (sample B-004 in Table 1). From the xray diffiction results in Figure 1, nearly all of the peaks present could be assigned to the orthorhombic Pnma phase of strontium zirconate. The presence of an impurity phase was
228
Dielectric Materials and Devices
evidenced by peaks centered near 25.5' 28 (with d-spacing 3.5 1 and 3.43 A ) which could not be indexed to the orthorhombic Pnma phase. We speculate that these reflections may arise from complex phases as Sr3Zr207-2H20or Sr2Zr20&03. Table 1Sample codes, starting composition and calcination schedule for SrZrO3, SrTi03 and Sr(ZrxTil-x)03 materials.
Sample Code
Compound
Starting composition
Calcination schedule
Aldrich B-002 B-003 B-004
SrZrO3 SrZrO3 SrZrO3 SrZrO3
Aldrich Stoichiometric 5% excess Zr Sr(N03)2.2H20 stoichiometric with Zr Stoichiometric Aldrich Stoichiometric stoichiometric without acetic acid and stoichiometric with 10% acetic acid and stoichiometric
--------
B-005 c-00 1 c-002 cx-1 cx-2 cx-3
6OO0C/12hr 6OO0C/12hr 9OO0C/12hr
4OO0C/3hr; 9OO0C/8hr
---------
400°C/3hr; 9OO0C/8hr 4OO0C/3hr; 9OO0C/8hr 4OO0C/3hq 900°C/4 hr, 1200°C/8hr 4OO0C/3hr, 9OO0C/4hr. 12OO0C/8hr
EDS analysis (see Figure 2) was performed to verify the compositional homogeneity of the grains. We found that the Sr:Zr ratio in samples containing mixture of phases (such as B-004, B003) was lower as compared with the stoichiometric samples (sample B-005 and Aldrich SrZrO3) which suggests the presence of a zirconium-rich impurity. In order to avoid the formation of mixed phases, the precursors A, B and C, were analyzed by thermal gravimetry. The assay as oxide (w/w %) was 48.41% and 18.38% for strontium nitrate and zirconium acetate, respectively, and 14.89% for titanium acetyl acetonate. Preliminary calculations did not prove the presence of any water of hydration in the Sr(NO3)2 used for the synthesis. Based on ths result, the correct stoichiometry was determined for zirconium. As seen in Figure 1, the x-ray pattern for this sample (labeled as B-005) calcined at 4OO0C/3 hr-9OO0C/8 hr exhibits superior phase purity as compared to sample B-004 and contains no peaks near 25.5' 28. The XRD pattern we obtained is consistent with the XRD patterns reported by others [2,14,15]. Finally, from the data in Table 1 and Figure 1, we can also observe that the samples calcined at lower temperatures (samples B-002, B-003) or containing excess of one reagent (sample B-003) resulted in a mixture of phases. The x-ray pattern for SrTiO3 is shown in Figure 3. Although the difhction peaks were shifted to somewhat higher 20 angles as compared to the ICDD-XRD patterns, the pattern is superior with respect to the intensity and peak height-to-width ratio as compared with the Aldrich material and consistent with the results of other investigators [10,13,16]. There was no evidence of peaks corresponding to impure phases such as T407 at 28 of 28-32' and Sr3Ti207 at 31-32' 28 as reported by other investigators [ 16,171. In contrast to the high degree of phase purity obtained for the pure end members, the synthesis of phase-pure Sr(Zro.2Ti,-,8)03 was more problematic. As seen in Figure 4 for sample CX-1, the peak located at 25.5' 28 was again present. On increasing the temperature of calcination from 900 to 12OO0C,however, we observed a slight decrease in this peak height, (see sample CX-2 in Figure
Dielectric Materials and Devices
229
4). The addition of excess of a 10% solution of acetic acid at 7-8% v/v of the reaction mixture lead to an XRD pattern with almost no peak intensity at 25.5" 28. The reason could be as follows: titanium acetyl acetonate was supplied in a solution of isopropanol, 1-butanol and methanol. The titanium acetyl acetonate may precipitate when zirconium acetate and water are added due to change in the solvent to precursor ratio. This interpretation is consistent with the observations made regarding titanyl acylate precursor by others [161. After synthesis, the powders were pressed into pellets of 12.7 mm diameter and 3.2-4.8 mm height and then sintered for one hour. The maximum in density of 5.24 g/cc (96.4% of the theoretical) for SrZrO3 is attained at 1520°C. For SrTiO3, the maximum in density is achieved at a fired density 5.01 g/cc (98.0% of the theoretical) at 1400°C. A summary of the processing conditions is given in Table 2. In aqueous phase, the maximum solids loading was 52% for both strontium zirconate and strontium titanate. The green density values for slip cast specimen of SrTiO3 and SrZrO3 were 3.23 and 3.48 g/cc whereas the corresponding fired density values at the respective optimum temperatures were 5.05 (98.4% of theoretical) and 5.33 g/cc (98.1% of theoretical) respectively.
Table 2 Effect of powder processing in water and propanol carrier fluid: solid loading, green density, fired density for SrTiO3 and SrZrO3 materials. Material
Fluid carrier
Solid Loading,%
g/cc
g/cc
SrTi03 SrTi03 SrTiO3 SrTiO3 SrTiO3 SrTiO3
water water water propanol propanol propanol
40.2 45.1 51.5 45 .O 55.1 60.2
3.21 3.22 3.23 3.26 3.28 3.31
5.00 5.05 5.05 5.06 5.07 5.08
SrZr03 SrZrO3 SrZr03 SrZrO3 SrZrO3 SrZr03
water water water propanol propanol propanol
40.3 45.3 52.4 45.7 55.4 61.2
3.40 3.43 3.48 3.58 3.62 3.63
5.29 5.31 5.33 5.36 5.39 5.42
Pgreen,
Pfired,
Table 2 also shows that it was possible to achieve 60% (v/v) solid loadings using propanol as against 52% (v/v) in aqueous phase. Using propanol based slurry formulations, both slip cast and tape cast samples were fabricated. The green density values for slip cast and tape cast bodies were near 65.5% of the theoretical density values for both the substrates as against -63% for specimens prepared fiom aqueous phase formulations. The hghest fired densities were obtained using propanol as the carrier fluid at high solids loading and correspond to fired densities of 99.4 and 99.6% of the theoretical density for SrTiO3 and SrZrO3, respectively. The electrical properties of these high density thin films were also measured. As seen in Figure 5 and Table 3 for SrZrO3, the dielectric constant at 25°C decreases fiom 60 to 40.6 as the frequency increases from 0.1 to 100 kHz. The variation in K with respect to test temperature from 25-1 10°C was 17% at 0.1 kHz, whereas at 100 kHz the change in K was less than 6%. The values of tan 6 decrease from 0.03 1 at 0.1 kHz to 0.0002 at 100 kHz and are also not strongly dependent
230
Dielectric Materials and Devices
upon temperature. SrZrO3 capacitors thus possess qualities advantageous in high frequency applications, as does SrTi03. The change in dielectric constant with applied voltage of STZ films (not shown here) was relatively low and there was no CV hysteresis observed with respect to DC bias of 0-2 volts. This demonstrates the insensitivity of the materials to electromechanical deformation and this is consistent with what has been observed by other investigators for STZ capacitors [ 5 ] . Table 3 Dielectric constant (K), dissipation factor (tan 6 ) and conductance for SrZr03 and SrTi03 substrates measured at 25°C.
+
Substrate Frequency
K
Tan6
63.3 56.6 46.5 43.1 40.0
0.127 0.107 0.031 0.007
KHZ
0.1 0.12 1 10 100
SrZrO3 Conductance,
us
0.032 0.877 2.632 8.375
SrTi03 K Tan 6 379 359 190 159 103
0.050 0.003 0.004 0.028
Conductance,
us
0.163 0.192 1.127 3.146 8.304
To determine the breakdown strength, the tape cast films of SrZrO3 and SrTi03 were electroded by gold sputtering (-10 nm thick) on each side. The breakdown strength values obtained were 650 V/mil for SrZrO3 and 550 V/mil for SrTi03 and are very encouraging for thin films fabricated via a powder processing route. For comparison, a value of 635 V/mil was reported [ 181 for SrTi03 substrates fabricated by pulsed excimer laser. CONCLUSIONS Propanol was found to be a better dispersion medium as 60% solid loading was achieved as compared to 52% in aqueous phase. This modification results in a significant gain in fired density; for instance, fired densities of 99.4 and 99.6% of the theoretical density for SrTiO3 and SrZr03, respectively, were obtained. The change in dielectric constant of SrZr03-SrTi03films with respect to frequency was moderate; the change in capacitance, however, with applied voltages (100-1000 mV) was almost insignificant and there was no CV hysteresis observed at low DC bias voltages. The &electric constant of 60 and 378 at 100 Hz applied fiequency and breakdown strength of 650 and 550 V/mil for SrZr03and SrTi03 suggests that these materials can be used for high voltage capacitor applications. ACKNOWLEDGEMENT This project was funded by the Honeywell Corporation, FM&T, which is operated for the United States Department of Energy under contract No. DE-AC04-76-DP00613. REFERENCES 1 . T. Matsuoka, JKuwata, Y. Fujita, and A. Abe, “Concentration Profiles of Composing Ions in Radio Frequency Sputtered Sr(Zro.2Tb.8)03 Films,” Journal of Applied Physics, 64[7] 35 12-15 (1988). 2. W.Zheng, W. Pang and G. Meng, “Hydrothermal Synthesis of SrZrO3_,(M=Al,GayIn,x<0.20) Series Oxides,” Solid State lonics, 108 37-41 (1998).
Dielectric Materials and Devices
23 1
3.
K.T. Jacob and Y. Waseda, “Potentiometric Determination of the Gibbs Energies of Formation of SrZrO3 and BaZrO3,” Metallurgical and Materials TransactionsB, 26B 775-8 1
4.
F. Krug and T. Schober, “The High-Temperature Proton Conductor Stontium Zirconate: Thermogravimetry of Water Uptake,” Journal of American Ceramic Society, 80 [3] 794-96
(1995).
(1997).
K.A. Vorotilov, M.I. Yanovskaya, L.I. Solovjeva, A.S. Valeev, V.I.Petrovsky, V.A.Vasiljev and I.E.Obvinzeva, “Ferroelectric Capacitors for Integrated Circuits,” Microelectronics Engineering,29 41-4( 1995). 6. G. Koster, B.L. Kropman, G.J.H.M. Rijinders, D.H.A.Blank and H. Rogolla, “fl Influence of the Surface Treatment on the Homoepitaxial Growth of SrTi03,” Materials Science and Engineering B56 209-12 (1998). 7. K.D. Budd and D.A. Payne, “Preparation of Strontium Titanate Ceramics and Internal Boundary Layer Capacitors By the Pechini Method,” MRS Symposia Proceedings, 32 239-45 5.
(1984). 8. H.M. Christen, L.A. Knauss, and K.S. Harshavardhan, “Field-dependent Dielectric Permittivity of Paraelectric Superlattice Structures,” Materials Science and Engineering, B56 200-03 (1998). 9. B. Gharbage, F.M.B. Marques and J.R. Frade, “Electrochemical Behavior of S r Z r l J l y x 0 3in~ Atmospheres Containing H2 and H20,” Electrochimica Acta, 43 [18] 2687-92 (1998). 10. H.K. Varma, P.K. Pillai, M.M. Sreekumar, K.G.K. Warrier and A.D. Damodaran, “Strontium Titanate Prepared by Spray Drying of Redispersed Metal Alkoxide Gel,” British Ceramic TransactionJournal,” 90 189-91 (199 1). 11. M.P. Pechini, “Methods of Preparing Lead and Alkaline Earth Titanates and Niobates and Coating Method Using the Same to Form a Capacitor,” U.S. Pat. No. 3,330,697, July 11, 1967. 12. N.G. Eror and H.U. Anderson, “Polymeric Precursor Synthesis of Ceramic Materials,” MRS Symposia Proceedings, 73 571-77 (1986). 13. P.A. Lessing, “Mixed-Cation Oxide Powders via Polymeric Precursors,” Ceramic Bulletin, 68[5] 1002-7 (1989). 14. J.S. Smith, R.T. Dolloff and K.S. Mazdiyasni, “Preparation and Characterization of AlkoxyDerived SrZrO3 and SrTiO3,” Journal ofAmerican Ceramic Society, 53 [2] 91-95 (1970). 15. T. Kasai, Y. Ozaki and S. Yamamoto, “Preparation of BaTi03 and SrTi03 from Metal Alkoxides,” Journal of The Ceramic Society Japan, 95 [ 101 1000-6 (1987). 16. C.F. Kao and W.D. Yang, “Preparation and Electrical Properties of Fine Strontium Titanate Powder From Titanium Alkoxide in a Strong Alkaline Solution,” Materials Science and Engineering B38 127-37 (1996). 17. M. Fujimoto and M. Watanabe, “Tin02n-lMagneliPhase Formation in SrTi03 Dielectrics,” Journal of Material Science, 20 [ 101 3683-90 (1985). 18. M.G. Rao and S.B. Krupanidhi, “Electrical Properties of SrTi03 Films by Pulsed Excimer Laser,” Journal ofApplied Physics, 75 [5] 2604-11 (1994).
232
Dielectric Materials and Devices
sm
h
c 3
0
a
2-Theta - Scale
Figure 1 Powder x-ray diffraction patterns for strontium zirconate.
Dielectric Materials and Devices
233
1400 1200
1000
f 800 3
8
* Q
z
600
400
200
0 0.00
1.oo
2.00
3.00 KeV
4.00
5.00
6.00
Figure 2 Comparison of EDS patterns of SrZxQ materials.
234
Dielectric Materials and Devices
21
30
4)
2-Theta - Scale
Figure 3 Powder x-ray diffraction patterns for strontium titanate.
Dielectric Materials and Devices
235
2-Theta - Scale
Figure 4 Powder x-ray difiaction patterns for Sr(Zr0.2Ti0.8)03.
236
Dielectric Materials and Devices
"k 60
U
C
8 v)
+298
K
+323
K
+343
K
-363
K
-X
55
-383 K
0
. I
U L
% 0
50
E
45
2
3
4
5
6
Log F
Figure 5 Dielectric constant and dissipation factor for SrZrO3 at T=298-383 K as a fbnction of applied fiequency.
Dielectric Materials and Devices
237
This page intentionally left blank
Effect of Dy-Doping on Resistance Degradation of BTZ Sintered in Reducing Atmosphere Under the Highly Accelerated Life Test Wen-Hsi Lee Philips Passive Components Kaohsiung 16, West 3rdStreet, N.E.P.Z. Kaohsiung, Taiwan ,R.O.C. Abstract Substitution of Dy rare earth ions were studied in Ba(Ti,Zr)03 dielectric materials , using X-ray diffraction and thermogravimetry. D g + ions enter both the A- and the Benters Ba sites as sites of the peroviskite structure, depending on Ba- or Ti-excess. donor and maximum solubility is 2 mol% for Ti-excess. However, the total solubility on A-and B-site are up to 8mol% for Ba-excess. There are two discrepant mechanisms of occupancy. 1) Dy completely enters Ti-site as acceptor when Dy concentration is below 2 mol% ; 2). Dyf3 both enters Ba- and Ti-site when Dy concentration is above 2 mol%. The donor of Dys;' is highly effective in improving the life stability by orders of magnitude. Impedance measurements gave an indications that insulating barrier layer were formed by Dy+3 incorporated into Ba-sites-donors on the grain boundary which could help to suppress electromigration of oxygen vacancies, thus improving the life stability.
Dg'
1. Introduction Dielectric formulation and process technologies for multilayer ceramic capacitor have been extensively investigated in the electronics industry[ 11, Recently multilayer capacitor with base-metal internal electrodes such as nickel have been developed to reduce the process cost[2]. In this case, dielectric should be co-fired in a reducing atmosphere to prevent the oxidation of internal electrode. However, BaTi03 dielectric material for capacitor with Ni-base metal electrode are usually fired in protecting atmosphere of N*/H2. Under such reducing conditions the dielectric perovskite material (Ba)(Ti,Zr)03 forms large numbers of ionized oxygen vacancies Vo. At the ionoization of Vo a high number of electrons are formed which give rise to a high electric conductivity. The electrons have to be therefore trapped by acceptor dopes. As far as the number of oxygen vacancies are below that of the acceptors, the BME material remain in a highly insulating state. The activation energy of acceptors on Ti-site is generally rather high(Ea>l SeV). Even at room temperature the ionized oxygen vacancies are rather mobile in the electric field. BME materials thus exhibit a high ionic contribution to the conductivity. The high ionic conduction often lead to a low insulation resistance(1.R.) and poor life stability of BME MLCC. Recently, rare earth elements were suggested by Randall and Shrout[3] for improvement on life stability of BME MLCC sintered in reducing atmosphere. In this paper, we therefore present some preliminary results about the defect chemistry of Dy with Ba- and Ti-site occupancy as well as the effect of D$+ on resistance degradation for (Ba)(Ti,Zr)03. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
239
2.Experiments The standard experimental material was BDTZ of the composition (BalxDyx)(Tio.78Zro.2i)03 The raw materials of Ti02 (Fuji Titanium), Zr02(Z-Tech), BaC03(Merck) and Dy203(Merck) were mixed in isopropanole, using agate ball mill and then calcined to perovskite at 1200C for 6 h in air. The atomic ratio A/B of the BDTZ was systematically varied by addition of appropriated amounts of Ti02 and BaC03 in the range 0.005 mol Ti02 to 0.15 mol BaC03. Ti02 and BaC03 additives were finely distributed over prereacted BDTZ by wet milling in isopropanole, using 2mm+ Zr02 balls. The powders were uniaxially pressured at 3kbar to discs of 6mm diameter and 600um thickness. The predensified samples were fired for 2 h at 1450°C in a reducing atmosphere of moist (H20-20C) N2/H2-9515. The incorporation of Dy in BTZ was studied by means of X-ray diffraction (XRD). The lattice parameters were determined , using a Si standard at room temperature. The incorporation of Dy on the B-sites of BDTZ was thermogravimetrically studied on the powders, containing various addition of BaCO3. The TGA experiments were carried out with prereacted mixtures of BDTZ and BaC03 in pure CO2 at 1250°C. For electrical measurements, 500um thick ceramic discs were evaporated with CrNiAu electrodes. The life stability of dielectric materials are commonly evaluated from long time measurements of the insulation resistance(1.R.) under electric filed and temperature stress. The long time measurements were abbreviated bu employing so-called highly accelerated life tests(HALT) under specially high temperature stress. The I.R. of such tested MLCC then breaks down after rather a short time. HALT experiments on MLCC were performed at 300°C and 300V. Impedance measurements were carried out over the range 100 Hz - 1MHz, using Hewlett Packard 41 92A instrumentation. The samples were allowed to equilibrate for 1-2 h at each temperature prior to measurements. The temperature range was 25 -300°C; temperature were accurate to f 1°C. Admittance analysis techniques in the frequency(f) domains were carried out For the temperature range in which the ceramics show a relatively low resistivity(500-900k). A frequencydomain analyzer(Mode1 HP 41 94A, Hewlett Packard) was used, covering the frequency range from lOOHz to lOMHz 3. Results and Discussion (I).XRD All materials prepared showed a cubic perovskite lattice. The lattice constant determined as hnction of the Dy concentration at 25"C, are shown in Fig.(l).
A/B ratio B-excess B-excess A-excess A-excess A-excess
240
Dy mole% 0-2mol% >2mol% 0-2mol% 2-8mol% >8mol%
Incorporation of Dy A-site Max. Solubility B-site A- and B-site Max.Solubility
Conductivity Semiconductor Semiconductor Insulator Insulator Insulator
Dielectric Materials and Devices
Fig.(l) 4.08 -
4.075 -
3 m A
4.07
/
-
Dy incorporated into Ti-site
,/
-
Ba-excess
4.065 4.06 -
4.055 -
..,K Ti-excess
4.045' 0
'
I
2
'
I
4
'
I
6
'
I
'
I
8 10 Dy(mol%)
'
I
12
'
I
14
'
16
The lattice constants of BDTZ with Dy completely incorporated on B- or A-sites follow simple Vegard lines. Dy+3obviously exhibits a preference for the B-sites, According to the larger atomic radius[4] of D$+(Rr~1=0.09lnm) compared to Ti4+( Ri~1=0.068nm), samples containing Dy on B-sites show a continuous increase of the lattice constants. The maximum solubility of Dy on B-sites is up to ca. 8 mol%. The maximum solubility of D$+ on A-site is in contrast only 2 mol%. Corresponding to the smaller ionic radius on A-sites show a slight decrease. The lattice parameters of BDTZ containing D P both on A- and B-sites are between those of the Vegard lines. At too high Ba or Ti excess in the material the following second phases were often observed. For too high A-excess(BaC03): Ba12Dy4.67Ti8035 For too high B-excess(TiO2):
Dy2Ti207
There phases were identified, using the ICDD File[S]. Dy2Ti207 is isomophous to Gd2Ti207, thus showing a very similar XRD diagram. Dy-containing second phases were also formed in the case of not carehlly mixing the raw materials. Such " nonequilibrium" phases were very stable and would not disappear even after long term annealing.
(a)TGA Measurements At the reaction of BaC03 with BDTZ certain amount of Dy are shifted from A- to B-sites and CO2 gas is evaporated which can be determined with TGA. (Bal-,Dy,)TiO3 + 2x BaCO3 (Bal+,)(TiDy,)Ti03
+
+ 2x CO2 --------(1)
Dielectric Materials and Devices
24 1
BaO + CO2
2BaCO3 <+
--------(2)
Reaction (2) is in contrast to Reaction(1) reversible in CO2 atmosphere at T<126OoC. Reaction(2) was therefore used in combination with Reaction(1) to determine exactly the amount of BaC03 required for the shift of Dy from A to B-sites. TGA diagram, Figure 2. Shows the reversible decomposition of BaC03 which was not consumed by Reaction(1). Similar experiments were carried by Hennings[6] at the shift of Ca+2-ionsfrom A to Bsite in BTZ. The TAG confirmed, that Dy+3can be completely shifted from A to B site and exactly two Ba were needed to shift one Dy from A to B sites. Fig.2
(ID).Life Time
HALT measurements show that insulation resistance(1.R.) and life stability of BTZ strongly depend on the Dy concentration, see Fig.(3). With increasing Dy from 0 to 2 mol% ,the lifetime are lowered. The effect
2 3 4
5
242
1 .o 1.5 2.0 2.5
10 3 1.5 32
Dielectric Materials and Devices
Fig
I.R. M-ohm
can be explained by the increased ionic conductivity due to the high number of mobile Vo at B/A<1[7]. However, the donor of DYB;' is highly effective in improving the life stability by orders of magnitude when occupancy of DY'~from B-site to A-site was happened. A lot oxygen vacancies which are contributed to poor stability of life for BaTi03 was generated during sintering in reducing atmosphere. The addition of donor to the dielectric was expected to compensate the oxygen vacancies generated by acceptor doping, thus, stability of lifetime is effectively improved. Therefore, the improvement of stability of life due to the shift of Dy incorporation from B- to A-site is plausible because of reduction of oxygen vacancies
According to previous study[8], the various types of donor was given different influence of stability of life. In the present investigation, reduction of oxygen vacancies by incorporation of Dy on Ba acting as donor is still insufficient to explain on the prolongation of life stability by orders of magnitude due to the shift of Dy from B- to Asite. (IV). Electrical measurement Admittance measurement of BTZ with 2.5mol% Dy addition revealed MaxwellWagner relation, see Fig.(4). Grain boundary layers ,seems to exist only at incorporation of Dy on Ba acting as donor.
Dielectric Materials and Devices
243
Fig. (4) -2 -3
c
4 L-,
-5
A
-6 - 573K -7 Q
.523K 473K 423K 373K
1
2
3
4
5
6
log(flHzl
The electrical measurements of BTZ with Dy doped reveal that the conductivity is dominated by ionic effects, The electronic contribution to conductivity seems to be negligibly small. Barrier layer effects occur only at Dy incorporated on Ba. Small shift of Dy from B-site to A-site in BTZ seems to give rise to reduced bulk resistance and life stability. Impedance measurements clearly indicated a pronounced Maxwell-Wagner relaxation, see Fig.(S). The increases of the barrier layer height, however, seems to be corrected with the reduction of the bulk resistance [9]. The increase of the barrier height must be seen relative to the reduction of the bulk resistance Fig 5
I*DY
100 WO
02sxoy
mo.000
, IW.000
00
m.000
400.000
6w.m
L'laml
ao0.m
, '
,
,
I
I.0W.m
1m
0
The study of Dy-doped BTZ clearly showed that Dy is a bulk donor dopant which strongly improves the life stability of the bulk material. Since the electrical conductivity of BTZ is widely determined by the ionic contribution of Vo, we have to conclude that Dy -donor obviously retard the mobility of Vo. There were indications of insulating barrier layers formed by D Y Bdonor ~ on the grain boundaries which could help to suppress electromigration of oxygen vacancies. However, this is only a plausible assumption. The actual mechanism of the Dy effect has still to be cleared by further experiments. 4. Conclusion
1. D++ ions enter both the A- and the B-sites of the peroviskite structure, depending on Ba- or Ti-excess. DJ?' enters Ba sites as donor and maximum solubility is 2 mol% for Ti-excess. However, the total solubility on A-and B-site are up to 81~101%for Ba-excess
244
Dielectric Materials and Devices
2. The experimental results have demonstrated that D ~ B acting , as donor can improve drastically the stability of life due to the reduction of oxygen vacancies and formation of barrier of grain boundary to suppress electromigration of oxygen vacancies.
5.Reference 1. Y. Sakabe, K. Minai and K. Wakino, “High-Dielectric Constant Ceramic for Base Metal Monlithic Capacitor”, Janpan. J. Applied physics, Vo1.20( 1981) Supplement 20-4, ~~147-150. 2. Y. Sakabe, K. Minai and K. Wakino, “High-Dielectric Constant Ceramic for Base Metal Monlithic Capacitor”, Janpan. J. Applied physics, Vo1.20( 1981) Supplement 20-4, pp 147-150. 3.A.Hitomi, T.R.Shrout and C.A. Randall ”Hypothesis on the rare Earth Doping of BaTi03 Capacitor Ceramics” Proc. 7th U.S.-Japan Seminar on Dielectric and Piezoelectric Ceramics,pp255( 1995) 4. R.D.Shannon,” Revised Effective Ionic Radius and Systematic Studies of Interatomic Distance in Halids and Chalcogenides”, Acta Cryst. A32( 1976) 75 1-767 5.International Center for Diffraction Data (ICDD), Newton Square, PA 190733273,USA( 1998) 6. D.F.K.Hennings & H.Hcsreinemacher,”Ca-acceptor in Dielectric Ceramic Sintered in Reducing Atmosphere”, J. Eur. Ceram. Soc. 15(1995) 795-800 7. H.L.Tuller and K.K.Baek,” Electrical Activity at Individual Grain Boundaries and Interfaces in Semiconducting Oxides”, Grain Boundaries and Interfacial Phenomenea in Electronic Ceramics pp 19-34 Edited by L.M.Levinson and S.I.Hirano. 8.R.Waser”Bulk Conductivity and Defect Chemistry of Acceptor-Doped Strontium Titanate in the Quenched State”, J.Am.Ceram.Soc.,74[8] 1934-40(1991). 9.H. Neumann and G. Arlt, “Maxwell-Wagner Relaxation and Degradation of SrTiO3 and BaTi03 Ceramics,” Ferroelectrics,69,179-86( 1986)
Dielectric Materials and Devices
245
This page intentionally left blank
EFFECTS OF BARIUM DISSOLUTION ON DISPERSING AQUEOUS BARIUM TITANATE SUSPENSIONS
Chia-Wen Chiang and Jau-Ho Jean Department of Materials Science and Engineering National Tsing Hau University Hsinchu , Taiwan, ROC
ABSTRACT
Dissolution of barium ion and its effect on dispersion behavior of aqueous barium titanate suspensions at various pH values have been investigated. The amount of leached barium ion decreases with increasing pH value. The dissolution of barium ion also causes an increase in pH value of suspension, but the change decreases with increasing initial pH value. The iso-electric point (IEP) of leached barium titanate powder increases with increasing leaching pH value and solid loading as well. The dissolution of barium ion enhances the colloidal stability of aqueous barium titanate suspension, in agreement with zeta potential measurement.
INTRODUCTION
Aqueous-based tape casting systems have been increasingly received attention in the electronic ceramic industry due to their low cost and reduced environmental impact. To prepare green sheets with high packing density and uniform microstructure, a well dispersed aqueous slurry is essential. To do that, the surface chemistry of oxide powders in aqueous suspensions has to be fully understood before the aqueous casting vehicle is formulated. In this study, barium titanate is chosen because it is the most widely used dielectric materials in the multilayer ceramic capacitor industry. It has been reported [ 1-31 that barium titanate is not thermodynamically stable in the acidic aqueous solutions. Ba+2ions are leached out of the powder
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
247
which results in a titanium rich surface layer. It was also found that the amount of Ba+2leached increases with decreasing pH [2-31. This increases the pH of resulting suspension and changes the surface chemistry of BaTi03. The reported iso-electric points (IEP) of commercial BaTi03 in aqueous suspensions vary considerably from pH=4 to pH=lO [4-121, depending upon the Ba/Ti ratio, processing history and measurement equipment. This makes the effects of pH on the dissolution and dispersion behavior of BaTi03 powders in aqueous solution still unclear. In this investigation, effects of barium dissolution on dispersion behavior of aqueous barium titanate suspensions at various pH values and solid loading are investigated. BaTi03 powders with different Ba/Ti ratios are used. Dissolution behavior of BaTi03 is assessed by inductively couple plasma atomic emission spectrometer (ICP-AES) and electron spectrometer for chemical analysis (ESCA). Surface chemistry of leached BaTi03 powders at different pH values are characterized by zeta potential measurements. Two techniques including acousphoretic titration (ACT) and electrokinetic mobility (EM) for concentrated and dilute suspensions are used to identify the effect of solid loading of BaTi03 on its surface chemistry at different pH values. Rheological measurements are conducted to characterize suspension stability and its relation to the dissolution of BaTi03.
EXPERIMENTAL PROCEDURE (1) Raw Materials The ceramic powders used in this study were high-purity BaTi03 (TICONHBP, TAM Ceramics, USA) with Ba/Ti ratios in the range of 0.992-1.004. The powder had a median size of 1.1- 1.5 pm and a specific surface area of 1.2-3.0 m2/gm, measured by light scattering (LS-230, Coulter Counter, USA) and N2 adsorption (BET-2300, Micromeritics, USA) methos, respectively. Deionized and distilled water was used, and the pH was adjusted by HCl and NaOH. (2) Experimental methods (a) Dissolution Studies Aqueous suspensions of various amounts of BaTi03 were prepared at pH ranging from 1.5 to 12. The slurries were deagglomerated by a high-energy ultrasonic horn, and mixing was continued by milling with Y203-stabilized Zr02 media for different periods of time. After mixing, the suspensions were centrifuged at a speed of 3000 rpm for 10 min to obtain supernatants. The amounts of barium ion left in the supernatants were determined by inductively couple plasma atomic emission spectrometer (ICP-AES) (SCIEX ELAN-5000, Perkin Elmer, USA). (b) Zeta Potential Measurements Two different techniques including electrokinetic mobility (EM) (ZetaPlus,
248
Dielectric Materials and Devices
Brookhaven Instruments, USA) and acousphoretic titration (ACT) (DT 1200, Dispersion Technology, USA) were used to measure the zeta potential of aqueous barium titanate suspensions. Suspensions were prepared by mixing various amounts of BaTiO3 powder in deionized water at different pH values. The slurries were prepared by the procedure described earlier. After mixing, a small amount of supernatant was removed by centrifuging, and then the zeta potential of the remaining powders in the supernatant was measured by the EM technique. For the ACT technique, however, a highly concentrated suspension without dilution was used to measure the zeta potential of aqueous barium titanate suspensions. (c) Rheological Behaviors Rheological behaviors of 60 wt% BaTiO3 powder suspensions at various pH values and amounts of electrolytes were determined using a concentric cylinder viscometer (RV-lOO/CV-200, Haake, Germany). Viscosity at a shear rate of 100 s-' was used for comparison.
RESULTS (1) Dissolution The amounts of barium ion leached from 60 wt% BaTi03 suspensions at pH=2.97-10.48 as a function of milling time are shown in Fig. 1 using BaTi03 powder with a Ba/Ti ratio of 0.992 as an example. It is found that the dissolution of barium ion from barium titanate powder takes place and completes almost immediately at all pH values investigated. Moreover, the total amount of barium ion leached decreases with increasing pH value, and becomes insignificant when the initial pH value is greater than 9.2. Similar phenomena are also observed at other Ba/Ti ratios ranging from 0.992- 1.004, and the results are summarized in Fig. 2. A higher concentration of leached barium ion is found for the BaTi03 powder with a greater Ba/Ti ratio. Note that little titanium ion is detected in the supernatants for all conditions investigated in Figs. 1-2.
.-+-
+-.ia
% #
1
Q
I
I
.
I
.
25
I
5Q
.
I
75
7.08
-7.85
.
I
.
1QQ
I
125
.
I
.
150
I
175
Time [hours)
Figure 1 The amount of Ba2+dissolved as a function of milling time for 60 wt%
Dielectric Materials and Devices
249
BaTi03 (Ba/Ti=0.992).
-
2oooo 180QQ
6afT i
-
1 I)0a
-
in00
-0.99d
0.992
aooo
I
o
1
2
1
4
5
a
7
6
P
io
.
11
Figure 2 The amount of Ba2' dissolved as a function of leaching pH for 60 wt% BaTi03 with different Ba/Ti ratios. The effect of solid loading on dissolution behavior of barium in aqueous barium titanate suspensions is shown in Fig. 3. A linear relation between the amount of leached barium ion and solid loading of barium titanate is noted, and the slope, which indicates the amount of barium ion leached from per gram of barium titanate powder, is about 355 ppm. The above results indicate that the amount of barium ion leached from barium titanate powder is independent of solid loading of barium titanate in the suspensions.
IQ
20
40
IQ
BaTiO, [gm 120gm H$)
250
Dielectric Materials and Devices
Figure 3 The amount of Ba2+dissolved as a function of solid loading of BaTi03 with a Ba/Ti=0.992, leached at pH=2.6. Fig. 4 shows that the pH value of aqueous barium titanate suspensions increases with milling time, and the change after milling for 150 hours decreases with increasing initial pH value. No significant change in pH value is observed at pH>9.2. Note that the specific surface area of BaTi03 powder remains relatively unchanged even after milling for 150 hours.
-
12
11 I-
10.2s
-9.20 -7.85
10
+7 .oa
9
a
-6.18
-A-s.ss
7
d .76
6
-2.97
5 4
1 2 Q
K
50
75
IQ0
125
150
175
Time (hours)
Figure 4 The pH of suspension as a function of milling time for the suspension with a solid loading of 60 wt% BaTi03 (Ba/T.i=0.992), dispersed at different pH values. (2) Zeta Potential Measurements Results of zeta potential of concentrated aqueous barium titanate suspensions at various pH values, measured by the ACT technique, are shown in Fig. 5 as a function of solid loading. The results are close, with an IEP in the range of 9.2 for all solid loading investigated. Note that not enough signal data is collected by the ACT technique when the concentration of barium titanate suspension is lower than 10 wt%.
Dielectric Materials and Devices
25 1
Figure 5 Zeta potential, measured by the ACT technique, as a function of pH for the suspensions with 10-30 wt% BaTi03 (Ba/Ti=0.992). Fig. 6 summarizes the zeta potential data for the suspensions with BaTi03 contents ranging from 0.2-30 wt%, measured by the EM technique. Note that the dilute supernatants were used for measurements. It is interesting to note that the IEP increases with increasing BaTi03 content, from p H ~ ~ = 3for . 5 0.2 wt% to pH~P=9.2for 10-30 wt%.
-30 I
.
,
1
.
,
.
I
1
2
.
l
.
I
5
4
.
I
.
I
.
I
7 . a
6
.
1
Q
.
1
.
1
.
L
1 0 1 1 1 2
(PHI
Figure 6 Zeta potential, measured by the EM technique, as a function of pH for the suspensions with 0.2-30 wt% BaTi03 (Ba/Ti=0.992). For the leached BaTi03 powders, the zeta potential is also affected by the leaching pH value, as shown in Fig. 7. The IEP decreases with decreasing leaching pH, from pH~pd3-9at PHleaching>6 to pH~p=3-4at PHleaching=l.5-5. 40
-
-
I~dm
p
15
3D +SD
+BD
-4- 9D
-a-w
,
~
.
i
,
.
~
,
.
~
I
.
I
4
.
I
.
5
I
.
6
I
.
7
I
.
I
a
.
I
~
.
I
.
i
~
i
i
PH
Figure 7 Zeta potential, measured by the EM technique, as a function of pH for
252
Dielectric Materials and Devices
i
z
i
the suspensions with 10 wt% BaTi03 (Ba/Ti=0.992), leached at different pH values. (3) Rheological behavior The colloidal stability of aqueous barium titanate suspensions prepared at various pH values is also assessed by measuring their rheological behavior, as shown in Fig. 8. For the suspensions with pH>l.97, the viscosity decreases with increasing shear rate indicative of shear-thinning behavior, suggesting a flocculated powder suspension. However, when the pH continues to decrease down to pH=l.97, a relatively constant viscosity with increasing shear rate is found, indicative of a Newtonian response corresponding to a well-dispersed powder suspension. Similar phenomena are also observed at other Ba/Ti ratios investigated. For comparison, the viscosity at an arbitrary shear rate, e.g., 100 secis used and the data are summarized in Fig. 9. It is found that the viscosity curves show a max
',
1QQ 0
25
*
I
75
I
,
IQQ 125
I
I*
r 175
I
-
r
7
Il)r)Esm275
Shear Rate [sec-)
Figure 8 Viscosity as a function of shear rate for the suspensions with 60 wt% BaTi03 (Ba/Ti=0.992) at different pH values.
Dielectric Materials and Devices
253
Figure 9 Viscosity at 100 sec-’ as a function of initial pH value for the suspensions with 60 wt% BaTi03 (Ba/Ti=0.992). CONCLUSIONS Dissolution of Ba2’ takes place almost immediately, and the amount of leached Ba2+decreases with increasing pH value. Moreover, the dissolution of Ba2+also causes an increase in pH value of suspension, but the change decreases with increasing initial pH value and becomes insignificant at pH>9.2. The isoelectric point (IEP) of leached barium titanate powder increases with increasing leaching pH value, from p H ~ ~ = 3for - 4 PHleaching=l.5-5.0 to p H ~ ~ = 8for -9 pHleaching>G. It is further found that the IEP of barium titanate suspension increases with increasing solid loading, from pH~p=3.5for 0.2 wt% to p H ~ ~ = 9for . 2 10-30 wt%. The above IEP results are explained by the dissolution of Ba2+and specific adsorption of Ba2+onto the leached BaTi03 powder surface. The dissolution of Ba2+enhances the colloidal stability of aqueous barium titanate suspension, in agreement with zeta potential measurement.
ACKNOLWLEDGEMENTS Samples provided by TAM Ceramics, Inc. is greatly appreciated. We are also grateful to Dr. Mike Chu of TAM for his encouragement and helpful comments. This study was supported by the National Science Council of Republic of China under Grant No: NSC 88-2216-E-007-036. REFERENCES [l]H.W. Nesbitt, G.M. Bancroft, W.S. Fyfe, S.N. Karkhanis and A. Nishijima, ”Thermodynamic Stability and Kinetics of Perovskite Dissolution”, Nature, 289 [29] 358-362 (1981). [2]D.A. Anderson, J.H. Adair, D. Miller, J.V. Biggers, and T.R. Shrout, “Surface Chemistry Effects on Ceramic Processing of BaTiO3 Powder”; pp 485-92 in Ceramic Transactions, Vol. I , Ceramic Powder Science IIA, Edited by G.L. Messing, E.R. Fuller Jr., and H. Hausner, American Ceramic Society, Westerville, Ohio, 1988. [3]M.C. Blanco-Lopez, B. Rand, and F. L. Riley, “The Property of Aqueous Phase Suspension of Barium Titanate”, J. Eur. Ceram. Soc., 17, 281-287 (1997). [4] J.-H. Jean, and H.-R. Wang, “Dispersant of Aqueous Barium Titanate Suspensions with Ammonium Salt of Poly(Methacry1ic Acid)”, J. Am. Ceram. SOC.,81 [6] 1589-1599 (1998).
[5] P. Gherardi, and R. Matijevic, “Homogeneous Precipitation of Spherical Colloidal Barium Titante Particles”, Colloids Surf., 32, 257-274 (1988).
254
Dielectric Materials and Devices
[6] M.C. Blanco-Lopez, B. Rand, and F. L. Riley, “The Property of Aqueous Phase Suspension of Barium Titanate”, J. Eur. Ceram. Soc., 17,281-287 (1997).
[7] T.J. Eade, I.A. Rahman, M.C. Blanco-Lopez, L.S. Tovey, and F.L. Riley, “Aqueous Processing of Barium Titanate Powders”, pp. 233-242 in British Ceramic Proceeding, vol. 52, Edited by W.E. Lee and A. Bell, The Institute of Materials, London, 1994. [8] A.W.M. Laat, and G.L.T. Hewerel, “Competitive and Displacement of Polyvinyl Alcohol and the Ammonium Salt of a Polyacrylic-Acid on BaTi03,” Colloids Surf. A, 70, [2] 179-187 (1993).
[9] A.W.M. Laat, and W.P.T. Derks, “Colloids Stabilization of BaTi03 with Poly(viny1 alcohol) in Water”, Colloids Surf. A, 71, 147-153 (1993).
[lO] A.W.M. Laat, and G.L.T. Hewerel, “Molecular Weight Fraction in the Adsorption of Polyacrylic Acid Salts onto BaTi03,” Colloids Surf. A, 98, 53-59 (1995). [ll]Z.-C. Chen, T.A. Ring, and J. Lemaitre, “Stabilization of Aqueous BaTi03 Suspension with Polyacrylic Acid”, Ceram. Trans., 22,257-263 (1991).
[12] Z.-C. Chen, T.A. Ring, and J. Lemaitre, “Stabilization and Processing of ~, Aqueous BaTi03 Suspension with Polyacrylic Acid”, J. Am. Ceram. S O C . , [~121 3201-3208 (1992).
Dielectric Materials and Devices
255
This page intentionally left blank
DEPENDENCE OF DIELECTRIC PROPERTIES ON THICKNESS (25 nm - 200 nm) FOR METAL-ORGANIC CHEMICAL VAPOR DEPOSITED PZT THIN FILMS C. H. Lin, P. A. Friddle, X. Lu and Haydn Chen Department of Materials Science and Engineering, Frederick Seitz Materials Research Laboratory University of Illinois at Urbana-Champaign Urbana, IL 61801 ABSTRACT Thin films of Pb(Zro,5Tio,5)03 (PZT), with thickness in the range 25 nm to 200 nm, were grown by metal-organic chemical vapor deposition (MOCVD) on LaNiO,/Pt/Ti buffered Si substrates. P-E studies showed a remanent polarization value of 17 - 20 pC/cm2, when tested at 5 V AC, for all but the 25 nm thick film. This film showed a remanent polarization value of 10 pC/cm2 when tested at 4 V AC. The coercive field increased from 50 to 350 kV/cm as thickness decreased. Fatigue testing showed a lower rate of fatigue in the 25 nm film. C-V measurements showed a shoulder near 0 V for the thinner films indicating a significant space charge contribution to the capacitance. INTRODUCTION Recently, the fabrication of integrated ferroelectric non-volatile dynamic random access memories (NV-DRAMS) has drawn much attention.[l] Several device structures have been proposed, such as the 1 transistor (1T) /1 capacitor (1C) structure used in conventional DRAM or the metal-ferroelectric-semiconductor field effect transistor (MFSFET).[2] [3] One of the major concerns in the development of ferroelectric based NV-DRAMSis the scaling limit of ferroelectric switching, since this limitation would affect the size of the device fabricated. Currently, most studies of PZT thin films are focused on the thickness range of 100-300 nm. There are few reports concerning the ferroelectric properties of PZT thin films below 5 0 m , which is an objective of the current investigation. EXPERIMENTAL PROCEDURE (100)-textured LaNi03 (0.2 pm) thin films were deposited on Pt (0.15 pm)/ Ti (0.05 pm)/ Si02 (0.15 pm)/ Si substrates by radio frequency magnetron sputtering at 300°C. The subsequent PZT thin film growth was carried out in a low pressure, horizontal, cold wall MOCVD equipped with a quartz reactor and a resistive substrate heater. The metal-organic sources used were Pb(TMHD)2, Z T ( O C ~ H ~and ) ~ ,Ti(OC3H7)+ Ultra high purity nitrogen and oxygen were used as the carrier gas and oxidant respectively. X-ray diffraction (XRD) scans were carried out using a Philips diffractometer with CuKa radiation. A Hitachi S-4700 field emission scanning electron microscope (SEM) was used to measure the film thickness in cross sectional view. To measure the electrical properties, Au (200 nm thick) square patterns (200 pm x 200 pm) were evaporated onto the PZT films as the top electrode of the metal ferroelectric metal (MFM) structure. A Sawyer Tower circuit was employed for measuring the hysteresis behavior (P-E loop). The C-V behavior was measured by To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paicfto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
257
employing a HP 4276 LCR meter with an external DC bias and an AC oscillating voltage of 50 mV (1OkHz). The DC voltage sweep rate was 0.01 V/sec. RESULTS AND DISCUSSION Microstructure X-ray diffraction of the LNO and PZT thin films was employed to determine the crystalline orientation. Shown in Fig. 1 is the 8-28 scan profile of a 25 nm-thick PZT thin film exhibiting a highly (1 00)-textured orientation. Fig. 2 contains the cross-sectional SEM micrographs, showing grain morphologies of 60 nm and 25 nm-thick PZT layers on the LNO layer. It was found that the lateral grain size of the PZT layer was around 0.05-0.1 Pm and the variation in thickness did not greatly change the lateral grain size. It was also found that grains in the PZT layer were mostly aligned with correspondiflg grains in the LNO layer, which confirmed the grain-to-grain epitaxial relation between the PZT and LNO. Since the lateral grain size is not a major factor affecting the electrical properties, the variation in electrical properties should arise from the interfacial or surface layer.
20
25
30
35
45
40
50
Two Theta (degree)
Figure 1. XRD pattern of a 25 nm-thick PZT grown on LNO buffered Si.
Figure 2. SEM micrographs of (a) 60 nm and (b) 25 nm thick PZT grown on LNO buffered Si.
258
Dielectric Materials and Devices
Hysteresis Behavior The P-E curves of 25-200 nm-thick PZT thin films with applied voltage ranging from 1V-5V are shown in Fig. 3. It was found that the remanent polarization (P,) values of 50-200 nm-thick films are all in the range between 17 pC/cm2 to 20 PC/cm2 for an applied voltage of 5 V. Also, the coercive voltages (V,) are all close to 0.8-1 V. However, the 25 nm thick-films broke down at 5 V. For 4 V applied voltage, the P, value of the 25 nm-thick film is around 10 pC/cm2 with the V, value being 0.6V. These P, values are lower than those recently reported by Wouter et a1.,[4] where it was reported that the P, value of sol-gel derived 4 1 1 > oriented 75 nm-thick PZT thin films on PVSi is around 30 pC/cm2 with a 2 V applied voltage. However, it was also reported by Bjormander et a1.[5] that <001>-oriented 60 nm-thick PZT thin films epitaxially grown on YBCO/LaA103 have similar P, values (2;0 pC/cm2) to those reported here. Moreover, the P, values of 100-600 nm-thick -oriented PZT deposited on LSCO/Si, as reported by Cillessen et al., [6] are around 20 PC/cm2. Thus, the difference in P, values might be due to differences in the film orientation. It is known that <11 I>-oriented PZT films generally have higher P, values than <001>-oriented PZT films, since it is easier to re-orient the permanent dipole moment using an electric field along <111> orientation due to the multiplicity, especially for compositions near the morphotropic boundary. However, it should be noted that the uncompensated P-E loop acquired using a Sawyer-Tower circuit might yield an inaccurate remanent polarization value if the sample is lossy. The coercive field increased substantially as thickness decreased from 50 kV/cm for the 200 nm film to 350 kV/cm for the 25 nm thick film. The most important effects on the feiroelectric/dielectric properties resulting from a reduction of film thickness are (1) an appreciable increase of the coercive field and (2) a noticeable decrease of the remanent polarization.
40 R
E 32 20
Y
0 -20
40 -2
0 Electric Fleld (kVlcm)
Figure 3. P-E behavior of PZT(50:50) thin films with different thickness under 1-5V.
Dielectric Materials and Devices
259
Fatigue tests of samples of different thickness were conducted. Fig. 4. shows the P-E behavior of a 25 nm-thick and 50 nm-thick PZT thin film after 4 x 108cycles, tested using a 3V triangular wave (1OOkHz). The 25 nm-thick film showed a slight degradation in remanent polarization up to 4 x 108cycles before its breakdown. On the contrary, it was found that a 50 nmthick PZT capacitor with same MFM structure ( i.e. Au/PZT/LNO) showed much pronounced fatigue after 4 x 108cycles and the P, value decreased from 15 &Xm2 to about 7.5 PC/cm2. The lower fatigue rate of the 25 nm-thick film may be attributed to single-domain grains. It is possible that the lack of domain boundaries prevent domain ginning from occurring.
40,
30
0 'C
n"
-20 -30 -40
I
4
-3
-2
,
I
-1
.
I
0
.
,
1
.
I
2
.
I
3
.
: ,
-4
.
,
.
2
,
.
0
,
.
2
I
4
Voltage (V)
Figure 4. Fatigue test of (a) 25 nm and (b) 50 nm thick PZT. Fatigue tests of samples of different thickness were conducted. Fig. 4. shows the P-E behavior of a 25 nm-thick and 50 nm-thick PZT thin film after 4 x 10' cycles, tested using a 3V triangular wave (1OOkHz). The 25 nm-thick film showed a slight degradation in remanent polarization up to 4 x 108 cycles before its breakdown. On the contrary, it was found that a 50 nmthick PZT capacitor with the same MFM structure ( i.e. Au/PZT/LNO) showed much pronounced fatigue after 4 x 108cycles and the P, value decreased from 15 W c m ' to about 7.5 Wcm'. The lower fatigue rate of the 25 nm-thick film may be attributed to single-domain grains. It is possible that the lack of domain boundaries prevents domain pinning from occurring. Capacitance-Voltage(C-V) Behavior In Fig. 5, the temperature-dependent C-V behaviors, with voltage sweeping from negative to positive values (upward sweeping), of 25-200 nm-thick PZT thin films are illustrated. The C-V curves are not symmetric about zero applied voltage. For films thicker than 75 nm, the capacitance values gradually increase with increasing temperature, because the dielectric constant of PZT increases with temperature (TcT,). In addition, all curves have their peak values located around 0.5-lV, which corresponds to the coercive voltages for an upward voltage sweep (-V to +V). As the film thickness is reduced to 50 nm, the capacitance values do not increase with the measuring temperature. Moreover, in addition to a peak in capacitance located at the coercive voltage, a shoulder located near zero voltage appears with increasing temperature (50- 140 "C). For the C-V curves of 25 nm-thick PZT films, the peak in capacitance occurs at 0.8 V and a
260
Dielectric Materials and Devices
shoulder near zero voltage appears in the upward voltage sweep (i.e. from -V to +V) at all measuring temperatures (20- 140 "C).
Figure 5. Temperature dependent C-V behavior of PZT(5050) thin films. C-V behavior of PZT thin films was discussed by Chai et a1.[7] They showed that two factors control the C-V behavior: (1) domain switching, and (2) distribution of space charge in the film. The variation of capacitance in the C-V curve of a ferroeletric material is traditionally considered to be due to domain switching from one poling state to the other when the DC voltage is sweeping from negative to positive. For very thin films (25nm-50nm),the variation of space charge distribution and therefore the built-in electric field near the PZT/electrode interface becomes more significant than thicker films. The C-V results indicate that the space charge variation becomes a dominant factor and the polarization behavior is greatly affected by the electric field distribution in these PZT thin layers. The shoulder at V=O and the invariance of measured dielectric constant with temperature indicate the effect of space charge variation on the C-V behavior.[8] CONCLUSIONS P-E studies showed a remanent polarization value of 17 - 20 pC/cm2, when tested at 5 V AC, for all but the 25 nm thick film. This film broke down at 5 V. The 25 nm film showed a remanent polarization value of 10 pC/cm2 when tested at 4 V AC. The coercive field increased from 50 to 350 kV/cm as thickness decreased. It is suggested that this is due to the increased significance of the space charge in the film. Fatigue testing showed a lower rate of fatigue in the 25 nm film. This may be attributed to decreased domain pinning due to the single domain nature of the grains. C-V measurements showed a shoulder near 0 V for the thinner films indicating a significant space charge contribution to the capacitance.
Dielectric Materials and Devices
26 1
ACKNOWLEDGEMENTS This work is supported by the U.S Department of Energy under the contract DEFG0296ER45439 through the Frederick Seitz Materials Research Laboratory, University of Illinois at Urbana-Champaign (UIUC). Fruitful discussions with Professor David Payne of UIUC are deeply appreciated. REFERENCES 1. J. F. Scott and C. A. P. de Araujo, Science 246, 1400-1402 (1989). 2. T. Yamazaki, K. Inoue, H. Miyazawa, M. Nakamura, N. Sashida, R. Satomi, A. Kerry, Y. Katoh, H. Noshiro, K. Takai, R. Shinohara, C. Ohno, T. Nakajima, Y. Furumura and S. Kawamura, IEEE IEDM Tech. Dig. pp.613-616 (1997). 3. S. Mathews, R. Ramesh, T. Venkatesan and J. Benedetto, Science 276,238-240 (1996). 4. D. J. Wouters, G. J. Norga and H. E. Maes, Mat. Res. Soc. Symp. 541, 381-391 (1999). 5. C. Bjormander, K. Sreenivas, M. Duan, A. M. Grishin and K. V. Rao, Appl. Phys. Lett. 66, 2493-2495 (1995). 6. J. F. M. Cillessen, M. W. J. Prins and R. M. Wolf, J. Appl. Phys. 81,2777-2783 (1997). 7. F. K. Chai, J. R. Brews, R. D. Schrimpf and D. P. Birnie 111, J. Appl. Phys. 82, 2517-2527 (1997). 8. C. H. Lin, Microstructure and Electrical Properties of MOCVD Derived Perozskite PbZr,Til-,03 and Pb(ScTa),_,Ti,O3Thin Films on LaNi03 Electrode Buffered Si, (Ph.D Thesis), University of Illinois at Urbana-Champaign, 2000
262
Dielectric Materials and Devices
OXIDATION OF CVD DIAMONDS: AN AUGER ELECTRON SPECTROSCOPY APPROACH J.Y. Howe and L.E. Jones joneslQalfred.edu School of Ceramic Engineering and Materials Science Alfred University Alfred, NY 14802 D.N. Braski and W.D. Porter High Temperature Materials Laboratory Oak Ridge National Laboratory Oak Ridge, TN 3783 1 ABSTRACT CVD diamond films, ET 100, were isothermally treated in oxygen at two conditions: 1) 550 'Cl95 kPa 0 2 for 170 min; and 2) 1478 "C/ 10-9Pa 0 2 for 600 min. The films were then investigated using Auger electron spectroscopy. Auger spectra of oxidized diamonds were compared with that of glassy carbon. Evidence is given for diamond minimizes its surface energy by: 1) oxygenchemisorption and retaining sp3 character at high pressure (95 k Pa 0 2 ) ; and 2) surface reconstruction to an sp2 character at low pressure (< 10-9Pa 02). INTRODUCTION There is an interest in the oxidation behavior of diamond surfaces because of the development and application of diamond-based electronic devices. It has long been argued that diamond oxidation and phase transformation to carbons are connected phenomena yet, they are not well understood. Identifying surface active sites, activity of specific crystallographic facets, and the extent to which surface conversion is involved in the oxidation process is key to the fabrication and design of diamond materials. The carbon atom in its ground state has the 2s' 2p2 electronic configuration. In diamond, the four electrons hybridize and adapt an sp3 configuration in which all four valence electrons have the same energy. Carbon and graphite materials have an sp2 hybridization. A phase transition in diamond may occur when the electronic configuration changes from sp3 to sp2, To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
263
and it is argued that this is the phenomenon closely related to the oxidation of diamond. Auger electron spectroscopy (AES) has been successfully applied to discriminate between diamond and carbon materials. Change in the detailed Auger spectra from diamond as compared to those from graphite and amorphous carbon, have been reported by Lurie and Wilson [l], Pate [2], Ramaker [3], and Howe, Jones, and Braski [4]. The major distinction is the peak shape of the first satellite Auger peak region from 255 to 265 eV. Diamond is a material with extremely high surface energy. According to Harkins' calculation, the surface energy is 5500 mJ/m2 on { 11 1}, as compared to 1200 mJ/m2 on silicon { 1 1 1} [5]. In contrast, the surface energy of graphite {OOOl} is as low as 70 mJ/m2 by wetting angle measurements [6]. In a recent paper, Howe and Jones proposed that diamond minimizes its surface energy via two possible mechanisms; either oxygen-chemisorption or relaxation to an sp2 configuration and each is governed by the partial pressure of oxygen [4]. Our previous Auger work supports in part this concept. At atmospheric oxygen pressure, only chemisorption occurs without surface reconstruction to the sp2 configuration. Cheniisorption and surface reconstruction to sp2 occurred simultaneously in oxygen partial pressures from 0.5 to 15 Pa at elevated temperatures [7]. At issue is what happens on the diamond surface in very low oxygen pressures at less than 10m9Pa at temperatures from 1000 to 1500 "C. In current study, AES has been used to investigate diamond surfaces after heat treatment. The bond nature of diamonds on the surface have been determined from the line shape of Auger spectra at the first satellite region around 260 eV. EXPERIMENTAL CVD diamond film ET100 manufactured by Norton Diamond Film, was the subject of the research effort. Highly oriented pyrolytic graphite (HOPG), grade ZYH, manufactured by Union Carbide, and the glassy carbon, manufactured by SGL, were the reference. HOPG and glassy carbon were chosen simply because they are the crystalline and amorphous carbon materials with sp2 bond character. Isothermal oxidation at low oxygen partial pressures was conducted using a Stanton-Redcroft (now Rheometric Scientific Inc.) Simultaneous Thermal Analyzer (STA 1500s) at 1018, 1318, and 1478 "C for 170 to 600 min. The experimental runs were conducted using flowing titanium-gettered helium. A flow rate of 50 ml/min was used for all exposures. Prior to heating, the instrument was purged until the oxygen partial pressure in the exhaust stream was below 10-9Pa, as indicated by a Centorr oxygen analyzer. Alumina crucibles were used and specimen size was approximately 50 mg in all cases. Isothermal oxidation at 550°C/95 kPa 0 2 was carried out in a Setaram TAG24 Thermoanalyzer for 170 min with a flow rate of 80 sccm.
264
Dielectric Materials and Devices
Auger spectra were obtained using PHI 680 Auger Nanoprobe. The accelerating potential was 2 kV and the beam current was 1 nA. Prior to transfer into the Auger Nanoprobe, as-received (AR) diamond and glassy carbon were ultrasonicate-rinsed in methanol or ethanol for 10 min. RESULTS AND DISCUSSIONS Figure 1 is the Auger spectra of CVD diamonds and glassy carbon. Spectra of CVD diamonds were collected from as-received (AR) surface as well as heat-treated specimens.
Fig. 1. Auger spectra of CVD diamonds and glassy carbon. Difference of bond character is shown in the first satellite peak Al. Heat treatment in low oxygen partial pressure leads to a surface reconstruction to sp2 bond character.
Dielectric Materials and Devices
265
Carbon has two core electrons and four valence electrons. Therefore carbon KLL transition is specified as KVV through the context. Because of the complexity of Auger transition in solids, a full and unambiguous interpretation is still pending, yet, a comparison between the Auger fine structure and predominant features in the band structure is possible. Ramaker showed that theoretically, carbon KVV (KLL transition is specified as KVV in context) Auger lines shapes are the convolution of the main peak generated by the normal KVV process as well as up to four satellite peaks due to the resonant electron excitations [3]. Peak shifts observed in the diamond spectra were mainly due to charging. Charging masked the subtle change of the main peak position (Ao) caused by the change of valence band structure [1,8]. However, the peak shape of the first satellite peak (AI) still truthfully reflects the bonding information. We thus limited our discussion to line shape of the major satellite peak, AI, at 262 eV. The diamond spectra, collected from diamonds treated at 550 "U95 kPa was quite similar to that of the AR specimen. The high shoulder of A I is the signature of the sp3 bond character [4]. As we have known that diamond has been oxidized after isothermally treated at 550 "C for 170 min[9], the similarity of line shapes between AR and oxidized diamonds implied that a direct oxidation from sp3-bond carbon occurred. In contrast, spectra of diamond treated at low oxygen pressure (< 10-9Pa 02) was similar to that of glassy carbon reference. The lower shoulder at A I is the significant feature of sp2-bound carbons. This clearly showed that after heat treatment at 1478°C/10-9 Pa 0 2 for 600 min, the surface of diamond was reconstructed from sp3 to sp2. Evidence of reconstruction from sp3 to sp2 was also observed from the change of appearance; a layer of black carbon, up to 3 p m thick, was formed on the diamond surface. Previously, we have proposed that diamond minimizes its surface energy by either cheinisorption or through surface reconstruction to an energy-favorable configuration, sp2. Surface reconstruction and direct oxidation can be a competing process depending upon oxygen partial pressure, temperature, and energy on specific crystal facets. The driving force of these transitions is thought to be the high surface energy of diamond. The Auger study on the oxidized diamond proved the fact that reconstruction does occur. An investigation on the quantity of surface energy as a function of crystal facets is under way. Computer simulation work is being carried on to address this issue.
-
SUMMARY CVD diamond films, ET 100, were isothermally treated in oxygen at two conditions: 1) 550 "U95 kPa 0 2 for 170 niin; and 2) 1478 'CC/ 10-9Pa 0 2 for 600 min. The diamonds were then investigated using Auger electron spectroscopy.
266
Dielectric Materials and Devices
Auger spectra of oxidized diamonds were compared with that of glassy carbon. It is suggested that diamond minimizes its surface energy by: 1) oxygenchemisorption and retaining sp3 character at high pressure (95 k Pa 02); and/or 2) surface reconstruction to an sp2 character at low pressure (< 10-9Pa 0 2 ) . ACKNOWLEDGEMENT This work was sponsored by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Transportation Technologies, as part of the High Temperature Materials Laboratory User Program, Oak Ridge National Laboratory, managed by Lockheed Martin Energy Corp. for the U.S. Department of Energy under contract DE-AC05-960R22464. REFERENCES I P.G. Lurie and J.M. Wilson, “The diamond surface 11. Secondary electron emission”, Surf. Sci. 65,476-98 (1977). 2B.B. Pate, “The diamond surface: atomic and electronic structure”, Surface Sci., 165, 83-142 (1986). 3D.E. Ramaker, “Chemical effects in the carbon KVV Auger line shapes”, J.Vac. Sci. Technol. A 7(3), 1614-1622 (1989). 4 J.Y. Howe, L.E. Jones, D.N. Braski, ”An Auger and XPS Study on CVD and Natural Diamonds”, accepted, MRS symposium paper, 2000 5 W.D. Harkins, “Energy relations of the surface of solids: I. Surface energy of the diamond,” J. Chem.Phy., 10 268-272 (1942). 6R.J. Good, L.A. Girlfalco, G. Kraus, “A theory for estimation of interfacial energies. 11. Application to surface thermodynamics of teflon and graphite,” J. Phy. Chem. 62, 14 18-2 1 (1 958). 7 J.Y. Howe and L.E. Jones, Unpublished data, 2000. 8 F.R. McFeely, S.P. Kowakczyk, L. Ley, R.G. Cavell, R.A. Pollak, and D.A. Shirley, “X-ray photoemission studies of diamond, graphite, and glassy carbon valence bands”, Phy. Rev. B, 9(12), 5268-78 (1974) 9 J.Y. Howe, L.E. Jones and D. W. Coffey, “The evolution of microstructure of CVD diamond by oxidation”, accepted, Carbon (2000).
Dielectric Materials and Devices
267
This page intentionally left blank
BaTi03-CERAMICS INTERGRANULAR CAPACITORS IN PROCESSINGMICROSTRUCTURE-PROPERTY RELATIONSHIP Vojislav V. MitiC, Ivona 2. MitroviC University of Nii Faculty of Electronic Engineering Beogradska 14, 18000 Nii Serbia, Yugoslavia;
Branka JordoviC Technical Faculty-CaEak Sv. Save 65, Serbia Yugoslavia
ABSTRACT In the process of BaTi03-ceramics consolidation, technological parameters like pressure, temperature and duration of the process essentially determine final electrical properties of the ceramics. A slight change of particular sintering parameter can change significantly the microstructure as well as the electrical parameter being observed. Since both intergranular structure and electrical properties of BaTi03-ceramics strongly depend on consolidation parameters, it is of extreme importance to correlate them. BaTi03-ceramics samples used have been sintered using sintering temperatures up to 1370°C. The microstructures of BaTiO3-ceramics observed by SEM method have shown that recognizable structural complex grain-contact-grain can be seen as a microcapacitor. Thus, the intergranular impedance model can be established. This model is developed for five-grain cluster. Obtained theoretical results are compared with frequency characteristics measured on pure as well as BaTiO3-ceramics samples with additives. This study presents the step firther to BaTi03-ceramics dielectric properties prognosis according to correlation synthesis-structure-property. INTRODUCTION The main goal of this paper is to define the corresponding frame for modeling a complex structure of different electrical parameters in a BaTiO3-ceramics sample bulk. In final, the idea is in controlling electrical and especially dielectrical properties through processing and microstructure parameters. Thus, through the controlling of microstructure from the point of view of processing parameters we are making relation between technological and final electrical properties. This triad
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paiBto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
269
is very important due to the research task of making the high level of electrical parameters integration in a BaTi03-ceramics bulk. The relation considering the influence of consolidation parameters and structure properties on dielectrical properties is expressed. In microstructure analysis we used previous results based on fi-actal structure nature and fractal method, too. EXPERIMENTAL The specimens for this study were prepared from “Murata” barium-titanate powder as well as powders of additives CeO2 and MnC03. To investigate the influence of the grain growth on the dielectric characteristic the following pressures have been used: 86, 105, 130 and 15OMPa. The specimens were sintered in a tunnel furnace type “CT-10 MURATA” at 1180°C to 1370°C for sintering times of 2 and 3 hours. The capacitance was measured on the specially prepared components, based on sintered samples, using “Hewlett-Packard” equipment (4276A LCZ Meter) [l]. The samples were tested in Heraeus D-6450 Hanau chamber. The microstructure investigations have been done by scanning electronic microscopy (SEW method (electronic microscope type JEOL-JSM-T20, magnification of 35000 times, resolution of 4.5nm), with a special attention to the metallographic analysis. The measurements were carried out on free surfaces using a semiautomatic device for quantitative analysis “MOP - Videoplan - Kontron” with automatic data processing. By this method the size and the shape of grains, as well as the number of grains can be obtained. This analysis provides the research of a great number of samples and view fields so that reliable statistical data have been obtained. Micrographs of BaTiO3-ceramic samples with additives 0.10% CeO2 and 0.14% MnCO3 obtained by SEhf method are shown in Fig. 1.
Figure 1. SEM micrographs of sintered BaTi03-ceramics (.r,ht=2h): a. Tsint=137O0C, p=3.6x103 kg/m3(x.5000); b. Tsint=l 18OoC, p=3.2x103 kg/m3 (~10000).
270
Dielectric Materials and Devices
RESULTS
AND DISCUSSION
The Intergranular Impedance Model The latest microstructure research of BaTi03-ceramics shows that recognizable structural complex grain-contact-grain can be seen as an impedance with dominant capacitance property [2]. As it is shown [3], the electrical model of BaTi03-ceramics sample consists of parallel connection of condensor C1 and three additional elements: condensor C, inductor L and resistor R (Fig. 2)
Figure 2. The electrical model of BaTi03-ceramics sample. Since ceramics sample consists of numerous sintered grains organized in clusters of different sizes, it is reasonable to suppose that each cluster of grains within the sample and even each intergranular contact within the cluster show similar behavior. Thus, in s-domain the expression for an elementary intergranular impedance can be written in symbolic [4] form as Z(S) =
1 + CR . s +CL.s2
(C, + C ) . s + C l C R . s 2+ClCL.s3
'
The dominant electrical parameter of this impedance model in wide frequency range is capacitance C1 [3,5]. The connection between C1 and geometrical parameters of two grains in contact can be established by assumption that the contact area between two grains can be viewed as a planar condensor (Fig. 3). Actually, when particles of barium-titanate powder which should be sintered, form a contact, in that area interatomic forces start an action forming a 'neck' of particle. In hrther process, a neck begins to grow and this process is controlled by different diffusion mechanisms (lattice diffusion, grain boundary difksion etc.) with the rates determined by total flux of atoms coming to the neck. The contact area looks like a new-emerged crystal structure. The boundaries of this area can be considered as defects in crystal structure and hence they can be treated as plates of a capacitor. Between the "plates" is a defect-free crystal structure that behaves as a dielectric (gray zone in Fig, 3). Let us assume that the surfaces of condensor plates correspond to intersection
Dielectric Materials and Devices
27 1
Figure 3. Two-sphere contact model presented in the plane section. surface (Sc) of two grains. Therefore, the formula for the capacitance of the planar condensor formed in the contact is given by
c = Eo&,-,S C X
where EO, E~ are, dielectric constants in vacuum and in BaTi03-ceramic material respectively; S, - the area of the plates and x - distance between condensor plates, i.e. the condensor thickness. The parameters of the model shown in Fig. 3 are: r l , r2-radii of the spherical grains, rc-radius of the grains intersection, d-distance between centers of grains, x-grains’ penetration thickness. Since the spherical shape of the grain is assumed, the contact surface is the circle of the area SC=xr:. It is obvious that rc depends on d. If condensor plates are on distance x, where x is a fhction of d according to the relation x=rl+r2-dy the capacitance is given by
Furthermore, let p and T are pressing pressure and sintering temperature respectively. If C = C ( p , T ) is the capacitance distribution, then E , = E, ( p , T ) can be computed from Eq. (2) as
where r, = rc ( t , T ) is the neck radius time dependent function. In the absence of pressing pressure, sintering is driven by the difference in surface curvature between sources and sinks. The curvature difference must go zero when the pore becomes spherical, that is when [6]
(*3hPijV3
r c . -rc = - . a ,
272
(5)
Dielectric Materials and Devices
where rCf is the final value of rc when 100 % density is reached, A, and A~ are the theoretical and initial densities of the compact, and a is the particle radius. In that sense, the limit value of Eq. (4) can be defined as
It can be shown that [6] [0.55.u, for two spheres
I
rcf = 0.74 - c1, for the compactof spheres.
(7)
If G S = G S ( p , T ) is the grain size distribution, then expression (7) should be modified for the point ( p , ~ as ) follows
w7 for two
10.jj.
“f I(p.T)
=
I
2
spheres
0 . 7 4 - F y for thecompact of spheres.
The dielectric width x is very difficult to estimate. However, it can be concluded that x is a fkction of grain size distribution Gs(p,~). We therefore introduce the dielectric width as where f = f ( p , T ) is the proportionality hnction. If we assume that C ( p , T ) and GS(p,T) are obtained as a result of experimental observation, then we can introduce the average proportionality fimction (f),and then Eq. (6) can be modified as
Fractal Correction: According to the conjecture that BaTi03-ceramics have the fiactal structure [7], the contact area is a fiactal-like surface. This means that the geometric structure of this surface is closer to fiactal then to smooth, Euclidean object (Fig. 4). If we, in addition accept the simplified type of the self-similar fiactal structure, this will allow us to construct the fiactal contact surface by successive applying the iterative procedure that contains of N contractions, each
Dielectric Materials and Devices
273
having contractive factor E. So, if we start with a flat surface having the area &, after k steps we will end up with the surface having & = (NE')~&as the area.
Figure 4. The intergranular capacitance model: a) two ellipsoihl grains in contact; b) fractal structure of contact surface; c) model of planar condensor with fiactal surfaces.
Suppose that D is the fiactal dimension of a self-similar fiactal being a model of our contact surface. It is known that E - = ~ N is the relationship that connects E, N and D. Thus, & can be expressed as Ak
- E ( ~ - ~- A) o~,
k 2 1.
Since, 0 < 6 1 and 2
Now, the formula for intergranular impedance (1) can be given as
274
Dielectric Materials and Devices
The Cluster Intergranular Impedance Model The relation (1) is valid only for one intergranular contact, and should be inserted for every contact between any pair of grains within the cluster. Let us consider an equivalent impedance of a cluster formed by five grains. This gives a hexahedron formed by two pyramids with a common triangular basis where the grains are placed as nodes. A five-gain cluster and its equivalent impedance scheme is given in Fig. 5. 1
Figure 5 . a) A cluster of five grains and b) its equivalent impedance model.
Grains within a cluster make nine intergranular contacts. Thus, an equivalent impedance model consists of nine intergranular impedances connected as shown in Fig. 5 b). For randomly chosen parameters for nine different elementary impedances [ 5 ] we obtained results presented in the Fig. 6. The Fig. 6 a) presents the magnitude of the equivalent impedance when opposite peaks of pyramids were chosen for ports, while the another (Fig. 6 b)) shows IZ,J for the same cluster when the equivalent impedance is looked between two nodes from the basis.
f (HL}
a)
b)
Figure 5 . The magnitudes of the equivalent impedances as a function of frequency: a) Iz12/(&> = f(Hz); b) IZM~(~B) = fWS.
Dielectric Materials and Devices
275
Experimental Results On the base of experimental results, the following 3D diagrams are obtained and shown in Figs. 7 and 8. Fig. 7 presents the capacitance as a function of pressing pressure and sintering temperature for pure BaTi03-ceramics samples while Fig. 8 presents the capacitance as a hnction of additives concentration (CeOz and MnC03) and exploitation temperature.
Figure 7. Capacitance as a function of pressing pressure and sintering temperature for pure BaTi03-ceramics samples.
Figure 8. The capacitance as a function of exploitation temperature and additives concentration: a) CeOz and b) MnC03 (T,,,=1370°C, tSbt=3h,p=86 MPa).
As it can be noticed, distribution of the capacitance (Fig. 7) is relatively uniform. For the lower temperature c ( ~ , T ) mainly depends on sintering temperature and very little on pressing pressure; for the higher temperature the influence of sintering temperature is lower than the influence of pressing pressure.
276
Dielectric Materials and Devices
It can be seen important decreasing of c ( p , ~ from ) 600 to 300 pF inside the temperature interval of 100°C. The minimal value is of about 250 pF under the pressure of about 130 MPa. The frequency characteristics for pure BaTi03-ceramics as well as BaTi03ceramics with addition of CeO;! and MnC03 are shown in Figs. 9 and 10.
Figure 9. Frequency characteristics of BaTi03-ceramics sintered on 124OOC and pressed under the pressure: a) 86 MPa; b) 105 MPa; c) 130 MPa; d) 150 MPa. In Fig. 9 the minimum of frequency characteristics is shifted towards higher frequencies for the increase of pressure from 86 MPa to 150 MPa. This expands the frequency range for the condensor's application. In broad frequency region (Fig. 9 right) more resonant peaks are observed. For the case of BaTi03-ceramics doped with CeOz and MnC03, the higher percentage of both additives improve the frequency characteristics (Fig. 10).
Figure 10. Frequency characteristics of BaTi03-ceramics samples with additives: a) 0.30% Ce02; b) 0.45% Ce02; c) 0.30% &CO3; 0.45% MnC03. Experimental results shown in Figs. 9-10 are just particular examples that confirm the accuracy of the intergranular impedance modeling given above.
Dielectric Materials and Devices
277
CONCLUSION Electrical properties of BaTi03-ceramics directly depend on the structural parameters. From the point of optimization of BaTi03-ceramics it is very important to establish the interrelation synthesis-microstructure-dielectric characteristics. In this paper, the model of intergranular impedance is developed taking into consideration fiactal nature of contact surface. The five-grain cluster is studied. For randomly chosen parameters for elementary impedances the results for magnitude of equivalent impedance of the cluster in frequency domain are plotted. Obtained diagrams are in agreement with experimentally obtained frequency characteristics of BaTi03-ceramics samples. Actually, taking into consideration more interconnections between clusters within the sample from the electrical point of view would outcome in better matching of theoretical and experimental diagrams. This will help in hrther dielectric properties design of electronic ceramics based on barium-titanate leading to a higher level of integration of B aTi03-ceramics components. ACKNOWLEDGEMENT
This research has been supported by the project: "Prognosis of materials characteristics financed by the Ministry of Science and Technology Republic of Serbia. "
REFERENCES V.V. Mitic, "The optimization of interrelations of microstructure and electrical properties of BaTiO3-cerarnics",Ph. D. Thesis, University of NiS, NiS, 1995 J. Daniels, K.H.Hardt1 and Wernicke, "The PTC effect of barium titanate", Philips Technical Review, 38, (3), 73-82, 1978f79. V.V. Mitic, P. Petkovic and M. Radmanovic, "Design of BaTi03-Ceramics Reactive Properties - Symbolic Approach", Proc. o f M E L '97, pp. 87-90. G.Gielen, W. Sansen, "Symbolic Analysis for Automated Design of Analog Integrated Circuits", Kluwer Academic Publishers, 1991. V.V. Mitic, P. Petkovic, I. Mitrovic, "BaTi03-ceramics Cluster Intergranular Impedance Model", Abstracts of the 3'd International Conference MicroMat2000, April 17-19, 2000, Berlin, Germany, pp. 259-260. M.F. Ashby," A First Report on Sintering Diagrams", Acta Metall., Vol. 22, 275289 (1974). V. Mitic, Lj. Kocic, M. Miljkovic, I. Petkovic, "Fractals and BaTi03-Ceramic Microstructure Analysis", Mikrochim. Acta [Suppl.] 15, pp. 365-369, 1998.
278
Dielectric Materials and Devices
HYDROTHERMAL SYNTHESIS OF HETEROEPITAXIAL BARIUM TITANATE THIN FILMS
E. Ciftci and M. N. Rahaman University of Missouri-Rolla, Department of Ceramic Engineering, Rolla, MO 65409 M. G. Shumsky University of Missouri-Rolla, Graduate Center for Materials Research, Rolla, MO 65409
F. D. Blum University of Missouri-Rolla, Department of Chemistry, Rolla, MO 65409
ABSTRACT
Heteroepitaxial BaTi03 thin films were deposited in an aqueous solution under hydrothermal conditions on single crystal substrates (100) of SrTi03 and (012) LaA103. The reactants consisted of fine Ti02 particles in a strongly alkaline solution of Ba(0H)z at a temperature of 150 "C. The growth of the films was studied by atomic force microscopy, high resolution scanning electron microscopy and X-ray diffraction. The formation of the films occurred by nucleation of {loo} faceted islands followed by growth to produce nearly complete coverage of the substrate. Repeated hydrothermal treatment improved the film thickness and surface coverage at the expense of slightly increased surface roughness. X-ray diffraction coupled with pole figure analysis showed that the films had the same in-plane and out-of-plane orientation as the substrate. INTRODUCTION Thin films of oxides with high dielectric constant are of significant interest for a variety of technological applications, such as dynamic random access memory (DRAM), sensors, thermistors and electroluminescent elements [ 11. Furthermore, epitaxial growth of the films can be used to optimize the anisotropic properties of electronically important oxides such as barium titanate, BaTi03, and lead zirconate titanate, Pb(ZrxTil-,)03, abbreviated as PZT. Conventional routes to the synthesis of ceramic thin films, such as sol-gel processing [2] and metal-organic ~~~~
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
279
chemical vapor deposition [3], require relatively high temperatures at which interdiffusion, interfacial reaction and evaporation of volatile constituents can lead to a severe deterioration of the electronic properties. There is interest in the electrochemical route [4,5], the hydrothennal route [6,7] or a combination of hydrothennal and electrochemical techniques [8,9] for the synthesis of ceramic films from aqueous solutions because they can provide low temperature, alternative routes to the high temperature methods. The growth of epitaxial films from aqueous solutions has been reported for simple oxides such as Ti02 and ZnO [ 10,111. The hydrothermal route has been used to deposit epitaxial BaTi03 and PZT films on SrTiO3 single crystal substrates [ 12-141. The purpose of the present work is to describe the influence of key parameters on the nucleation and growth of BaTi03 films on single crystal substrates by deposition fiom aqueous solution under hydrothermal conditions. Barium titanate films were deposited on (100) SrTi03 and (012) LaA103 single crystals by reacting fine Ti02 powder and a strongly alkaline solution of Ba(OH), solution at 150 "C in a Teflon-lined autoclave. The growth of the films was characterized by atomic force microscopy (AFM), X-ray diffraction (XRD) and scanning electron microscopy (SEM). EXPERIMENTAL The chemicals for the deposition of the BaTi03 films consisted of barium hydroxide octahydrate [Ba(OH)2.8H20; Aldrich, Milwaukee, WI] and Ti02 (average particle size =25 nm; Degussa Corp., South Planefield, NJ) consisting of ; -30 weight percent (wt%) rutile and -70 wt% anatase. Four grams of Ba(OH)2.8H20 was added to 24 cc of deionized water in a Teflon-lined autoclave (45 ml capacity, Parr Instrument Co., Moline, IL). The system was purged with argon, sealed and heated for 3 h at 90 "C to dissolve the Ba(OH)2-8H20 completely. One gram ) a polished single crystal of Ti02 powder was added to the solution @H ~ 1 3 . 3and of SrTi03 or LaA103 (6.25 mm by 6.25 mm by 0.5 mm thick; Superconductive Components, Inc., Columbus, OH) was suspended in the solution using a Teflon thread. The autoclave was sealed and the system was heated for fixed times at 150°C. Two types of runs were performed. In one case, the deposition process was performed for fixed times between 3 h and 24 h using a different crystal for each run. In the other case, to investigate the nucleation and growth process progressively, the same crystal was used for repeated deposition steps. After each step, the reactant solution in the autoclave was replaced with a fresh solution. Prior to characterization, the single crystals were washed with deionized water in an ultrasonic bath to remove loose particles fiom the surface of the film and dried at -80 "C.
280
Dielectric Materials and Devices
The crystalline structure of the deposited films was determined by XRD (XDS 2000; Scintag Inc., Sunnyvale, CA) using nickel filtered Cu & radiation (h = 0.15405 nm) at a step-scan mode with 28 = 0.01" per step. Texture analysis by XRD was performed using an azimuthal-scan mode at fixed tilt angle of 45". The pole figure measurements were made by employing a texture goniometer in which the crystal was tilted at a fixed angle (45") corresponding to the planes of interest (e.g., 110) and rotated azimuthally to investigate preferred orientation phenomena. X-ray rocking curves were also used to interpret the quality of synthesized layers by examining the curve width in comparison to its theoretical width. Broadening in the X-ray rocking curves can be resulted mainly by defects such as the mosaic spread-wise formation [ 151. The morphology and microstructure of the films was examined with a field emission SEM (Edax Phoenix System, Edax Inc., Mahwah, NJ). Atomic force microscopy (Nanoscope 111, Digital Instruments Inc., Santa Barbara, CA) was used to determine the crystal; size, thickness and surface roughness of the films. RESULTS AND DISCUSSION In the present experiments, the reaction system consists of fine Ti02 particles in a solution of Ba(OH)2 into which a single crystal substrate (SrTi03 or LaA103) is suspended. The reaction mechanism is not clear but it has been suggested to involve a dissolutiordprecipitation process in which Ti is hydrolyzed to form either Ti(0H):- [9] or Ti(OH)4 [16], followed by subsequent reaction with Ba2' ions to precipitate BaTi03. Several interactions are possible. For example, the BaTi03 can nucleate homogeneously in solution and grow to form individual particles or agglomerates, eventually leading to an uncoated substrate and BaTiO3 particles. Nucleation can also occur heterogeneously on the fine TiO2 particles [ 131, resulting again in an uncoated substrate and BaTi03 particles. In the experiments, it was found that the reactant concentrations used earlier [ 171 to synthesize fine BaTiO3 particles did not lead to film deposition. By reducing the concentration of the reactants at a fixed synthesis temperature (150 "C), film deposition was achieved. Thus, the concentration of the reactants is a key variable in the deposition process. Figure 1 shows AFM images of the nucleation and growth process at 150 "C of BaTi03 on a single crystal SrTiO3 surface. The formation of BaTi03 nuclei is observed after -20 min [Fig. l(a)] and contiues rapidly so that a significant fraction of the surface is covered with fine faceted crystals after 30 min [Fig. l(b)]. By refreshing the reactant solution after 30 min, almost complete coverage of the single crystal surface is observed after 1.5 h [Fig. 1(c)].
Dielectric Materials and Devices
281
Figure 1. AFM images of BaTi03 film on SrTiO3 single crystal for deposition times of (a) 20 min, (b) 30 min and (c) 1.5 h. Figure 2 shows AFM images of BaTi03 deposited on a SrTiO3 single crystal surface after 0.5 h, 3 h and 9h. The reactant solution was refreshed after each run. Nucleation and growth of the BaTi03 crystals occur simultaneously, eventually leading to almost complete coverage of the substrate surface. Additional observations indicate that after coverage of the substrate has been achieved, the film thickens by nucleation and growth on the BaTi03 film surface.
Figure 2. AFM images of BaTi03 film on SrTi03 single crystal surface after deposition times of (a) 0.5 h, (b) 3 h and (c) 9 h.
282
Dielectric Materials and Devices
SEM observation of the films deposited on SrTiO3 shows complete coverage of the substrate by a fairly smooth film [Fig. 3(a)]. High resolution SEM indicates the presence of some fine porosity [Fig. 3(b)]. However, other observations revealed that deposition is hindered by the presence of flaws on the substrate surface (e.g., scratches and other defects produced by insufficient polishing of the substrate), thereby reducing the coverage and uniformity of the film.
Figure 3. SEM images of BaTi03 films deposited on SrTi03 substrates. While AFM revealed almost complete coverage of the SrTiO3 substrate surface after -1.5 h (Fig. l), the deposited film became thick enough to be detected by XRD after -5 h of deposition. Figure 4 shows XRD patterns of the BaTi03 films deposited on SrTiO3 and LaA103 substrates. Only the reflections fiom the (100) and (200) planes of the film are observed, indicating that the films have the same out-of-plane orientation as the substrate. Pole figure measurements of the (1 10) reflection of the films confirm that they also have the same in-plane orientation as the substrate (Fig. 5). Rocking curves were used to investigate the crystalline quality of the films by comparing their full width at half maximum (FWHM) with that for the substrate. Commonly, the FWHM is -10" for polycrystalline materials and 4" for single crystals. The data given in Table 1 indicate that the FWHM for the films are larger that those for the single crystal substrates but significantly smaller than the values commonly observed for polycrystalline materials.
Dielectric Materials and Devices
283
Figure 4. XRD pattern of BaTi03 films on (a) SrTiO3 substrate and (b) LaA103 substrate, showing that the films have the same out-of-plane orientation as the substrate. Table 1. Data for the full width at half maximum (FWHM) of the BaTi03 films and for the SrTiO3 and LaA103 single crystals on which they were deposited. Material BaTi03 film on SrTi03 SrTi03 single crystal BaTi03 film on LaA103 LaA103 single crystal
284
FWHM (degrees) 0.38 0.26 0.93 0.65
Dielectric Materials and Devices
Figure 5. Azimuthal scans for the (1 10) reflections for (a) BaTi03 film on SrTiO3 substrate and (b) the uncoated SrTi03 substrate, showing that the film has the same in-plane orientation as the substrate.
Dielectric Materials and Devices
285
The thickness of the deposited films was determined using SEM fkom crosssections mounted in epoxy resin. After 5 runs, with each run lasting for 24 h and with the reactant solution refreshed after each run, the thickness of the BaTi03 film on the SrTiO3 substrate was measured as -0.6 pm (Fig. 6). As outlined earlier, AFM images revealed nearly complete coverage of the substrate after only 1.5 h, followed by thickening of the film. After 1.5 h, the film thickness is estimated to be -0.1 pm. Figure 7 shows AFM images of the surfaces of an uncoated SrTi03 substrate and BaTi03 films after various deposition times. The film appears relatively smooth at shorter deposition times but the roughness appears to increase slightly at longer deposition times.
-
-
Figure 6. SEM image of the cross section of a relatively thick BaTi03 film deposited on a SrTiO3 substrate. The film was deposited in 5 runs, with each run lasting 24 h.
CONCLUSIONS Heteroepitaxial BaTi03 thin films were deposited on single crystal substrates of (100) SrTi03 and (012) LaA103 by reacting fine Ti02 particles in a strongly alkaline solution of Ba(OH)2 under hydrothennal conditions 150 "C. The formation of the films occurred by nucleation of {loo} faceted islands followed by growth to produce nearly complete coverage of the substrate after only -1.5 h. Repeated hydrothermal treatment improved the film thickness at the expense of slightly increased surface roughness. Dense films with thicknesses in the range of 0.1 to 0.6 pm were deposited. X-ray diffraction coupled with pole figure analysis and rocking curves showed that the films had a high quality of crystallinity and the same in-plane and out-of-plane orientation as the substrate.
286
Dielectric Materials and Devices
Figure 7. AFM images of the surfaces of SrTiO3 substrate (a) and BaTi03 films deposited in 1 run lasting 1.5 h (b), 1 run lasting 9 h (c) and 5 runs, with each run lasting 24 h. REFERENCES ‘R. E. Newnham, Rept. Prog. Phys., 52 [2] 123-156 (1989). *R. G. Dosch, “Preparation of Barium Titanate Films Using Sol-Gel Techniqes,” Mater. Res. Soc. Symp. Proc., 32 157-61 (1984). 3B. W. Wessels, “Metal-Organic Chemical Vapor Deposition of Ferroelectric Oxide Thin Films for Electronic and Optical Applications,” Ann. Rev. Mater. Sci., 25 525-46 (1995). 4 E. Bohannan, M. G. Shumsky, and J.A. Switzer, “Epitaxial Electrodeposition of Copper (I) Oxide on Single-Crystal Gold (loo)”, Chem. Mater., 11 2289-92 (1999). 5W.-S. Cho, M. Yashima, M. Kakihana, A. Kudo, T. Sakata, and M. Yoshimura, “Room-Temperature Preparation of Highly Crystallized Luminescent SrW04 Film by Electrochemical Method,” J. Am. Ceram. Soc., 78 [ l l ] 31 10-12 (1995). 6W. J. Dawson, ‘‘Hydrothermal Synthesis of Advanced Ceramic Powders”, Ceram. Bull., Vol. 67[10], 1673-78 (1988)
Dielectric Materials and Devices
287
7K. Kajijoshi, N. Ishizawa, and M. Yoshimura, “Preparation of Tetragonal Barium Titanate Thin Film on Titanium Metal Substrate by Hydrothermal Method,”J. Am. Ceram. Soc., 74 [2] 369-74 (1991). ‘M. Yoshimura and W. Suchanek, “In Situ Fabrication of MorphologyControlled Advanced Ceramic Materials by Soft Solution Processing,” Solid State Ionics, 98 197-208 (1997). 9M. Yoshimura, S. E. Yoo, M. Hayashi, and N. Ishizawa, “Preparation of BaTiO3 Thin Film by Hydrothermal Electrochemical Method,” Jpn. J. Appl. Phys., 28 [2] L2007-L2009 (1989). 10 Q. Chen, Y. Qian, Z. Chen, W. Wu, Z. Chen, G. Zhou, and Y. Zhang, Appl. Phys. Lett., 66 1608- (1995). 11 Q. Chen, Y. Qian, Z. Chen, and Y. Zhang, Mater. Lett., 2293- (1995). 12 K. Kajiyoshi, N. Ishizawa, and M. Yoshimura, “Heteroepitaxial Growth of BaTi03 Thin Films on SrTi03 Substrates under Hydrothermal Conditions,” Jpn. J. Appl. Phys., 30 [lB] L120-L123 (1991). 13 A. T. Chien, J. S. Speck, F. F. Lange, A. C. Daykin, and C. G. Levi, “Low Temperature/Low Pressure Hydrothermal Synthesis of Barium Titanate: Powder and Heteroepitaxial Thin Films,” J. Mater. Res., 10 [7] 1784-89 (1995). 14 A. T. Chien, J. S. Speck, and F. F. Lange, “Hydrothermal Synthesis of Heteroepitaxial Pb(ZrxTi1-,)03 Thin Films at 90-150°C,” J. Mater. Res., 12 [5] 117678 (1997). ”H. Klug and L. Alexander, &Ray Dlffvaction Procedures, 2nd Edition. Wiley, New York, 1974. 16 R. Bacsa, P. Ravindranathan, and, J. P. Dougherty, “Electrochemical, Hydrothennal, and Electrochemical-Hydrothermal Synthesis of Barium Titanate Thin Films on Titanium Substrates,” J. Mater. Res., 7 [2] 423-28 (1992). 17 E. Ciftci and M. N. Rahaman, “Hydrothermal Synthesis and Characterization of Barium Titanate Powders,” This Symposium.
288
Dielectric Materials and Devices
DIELECTRIC PROPERTIES OF BARIUM TITANATE SINTERED WITH ZnO-BA4SEDT;r,UXE3 Deep Prakash and B.P. Shanna Powder Metallurgy Division, Bhabha Atomic Research Centre Vashi Complex, Turbhe, Navi Mumbai - 400705 INDIA
P. Gopalan and T.R. Rama Mohan Department of Metallurgical Engineering and Materials Science, I.I.T. Bombay, Powai, Mumbai-400076 INDIA
ABSTRACT Samples of barium titanate containing pre-reacted ZnO-B203 flux compositions were liquid phase sintered in air at different temperatures. No shift in the Curie - temperature (Tc) was noticed while a hghly resistive p i n boundary phase was observed, which was responsible for improvement of tan delta. In addition, grain refinement took place which led to increase in dielectric constant. The role of ZnO-based fluxes on the sintering behavior, development of microstructure and dielectric properties is elaborated. INTRODUCTION Ever growing trend of miniaturization in electronic industry has pressed the demand for high volumetric eficiency capacitors. This has resulted in emergence of multi-layered ceramic capacitor (MLCC) conft-mations. The MLCC manufacturing consists of stacking alternative layers of dielectric ceramic tape and an internal electrode ink, which are co-fired. Barium titanate (BaTiO3) is used extensively for these applications due to its high dielectric constant (K) [l]. The co-firing is done at the sintering temperature of barium titanate, under oxidizing atmosphere. Sintering of titanales under reducing atmosphere may give rise to reduction of titanium ion, thereby deteriorating dielectric properties [2]. A constraint of maintaining high partial pressure of oxygen during co-firing necessitates use of noble metal as a choice for internal electrode. Pd-Ag alloy is generally used for this purpose. High cost of this alloy is a major concern for commercial usage, which is mainly contributed by Pd, the expensive constituent of the alloy. However, tremendous cost-savings may be realized by reducing the
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
289
amount of palladium in the alloy. Composition of Pd-Ag alloy to be used as electrode ink, depends on the temperature of application. At a temperature of -1350" C, at which BaTi03 based ceramics are generally sintered, an alloy of 70%Pd is required, whereas alloy containing only 30% Pd would suffice, if the sintering temperature is brought to 1150" C [11. In order to decrease the sintering temperature, adtlltion of various types of sintering aids (fluxes) have been attempted [3-71. Though good densification was achieved by using low-melting glasses and glass-compositions, electrical properties were compromised due to dilution of dielectric constant. Therefore, need of a flux was felt, whch may improve dielectric properties. Studies of Burn [6] and Wang [7] were attempted in this direction, but these works showed graingrowth in flux-sintered ceramics, which may not be beneficial for achievement of temperature-stable characteristics. This paper reports investigations on a binary flux based on ZnO-B203 system. Role of flux on the densification and microstructural development is presented and correlation with dielectric properties is discussed.
-
EXPERIMENTAL WORK Barium titanate powder was prepared by conventional mixing and grinding process. Stoichiometric amounts of high purity titanium dioxide and barium carbonate powders were mixed in a ball-mill using alumina balls as grinding medium. Mixed powders were calcined in air at 1100"C. Repeated grrnding and calcination steps were carried out to ensure formation of single phase powder. Characterization of barium titanate powder was done by x-ray diffraction (XRD) and particle size analysis. The XRD wa..done with Cu-K, radiation, using Philips Analytical diffractometer. Particle size analysis was carried out by Horiba LA-500 laser diffraction particle size analyzer. Binary flux composition was prepared by mixing weighed amounts of zinc oxide and boric acid powders in an agate mortar and pestle. The mixed powders were pre-reacted at -300" C, ground and added in amounts of 4 wt.% to barium titanate powders. Green pellets were compacted using 1 wt.% PVA as binder. Pure barium titanate pellets were sintered at 1350" C, whereas flux-added pellets were sintered at 1OOQ", 1 100' and 1 150" C Sintered pellets were lapped, coated with silver paste and electrical contacts were soldered. Dielectric measurements were carried out between room tempeEture and 150" C using LCR meter, model SR-720, manufactured by Standard Research Systems, U.S.A.,from 0.1 to 100 kHz. Microstiuctural examinations were cai-ried out by scanning electron microscopy (SEM) and transmission electron microscopy ( E M ) , using Cambridge Stereoscan 240 and Jeol FX-2000 respectiveiy.
290
Dielectric Materials and Devices
RESULTS X-ray diffraction pattern of the barium titanate powder is shown in fig. 1. All the peaks were assigned to barium titanate and no impurity phase was detected. Particle size analysis of the powder indicated the medmn of the particles to be at 0.5 pm with all the particles below 1.2 pm. Sintered density of the pure barium titanate was found to be 5.61 gcc. Densities of flux-sintered ceramics, sintered at different temperatures varied linearly with increase in sintering temperature ( Fig.2). Fig.3 shows the variation of shrinkage in diameter ( A 4 /4 ) against sintering temperature for flux-sintered pellets. The shnkage was found to increase linearly with sintering temperature. SEM photomicrograph of pure barium titanate is shown in fig.4. The microstructure shows an average grain size of 5 ym,whereas some o f the grains grew up to lOpm also. On the other hand, flux-sintered ceramics were found to hzvc averagc graiil sizcs below 1.00 pm. Fig.5 shows bright field TEM photograph of pellet sintered at 1150" C. Sub-micrometer size grains of barium titanate can be ciearly noticed in the micrograph. Dielectric constants of pure and flux-sintered pellets, measured at roomtemperature are shown in fig6 as a function of frequency and sintering temperature. The dielectric constants of flux-sintered ceramics were found to be higher than those obtained for pure barium titanate. Also, the magnitude of the dielectric constant (K) increased with increase in the sintering temperature at a given frequency for flux-sintered samples. For all samples, K decreased with increase in the frequency. The drop in K with frequency was almost linear on logarithmic scale, except for the pellet sintered at 1000" C, for whch the drop in K was sharp till 1 &, beyond which it became linear. Fig.7 shows room temperature dielectric loss (tan 6) values of flux-added ceramics as a hnction of frequency, Tt can be seen that the tan ci values are higher for pellet sintered at 1000" C, but nearly independent of sintering temperature for pellets sintered at 1100" a d 1 150" c. Fig.8 shows dielectric constant of samples as a function of temperature, at a frequency of I kHz. All t'he samples exhibited Curie-point at 130" C. Pure barium titanate exhibited a room-temperature dielectric constant of 700 which increased to 1600 at the Curie temperature. The flux-sintered ceramics showed K values to slightly increase with increase in the sintering temperature. However, K values at the Curie-point are significantly less than that exhibited by pure BaTi03. The resistivity - temperature behavior of pure and flux-sintered pellets at 1 kHz have been shown in fig.9. The resistivity values for pure barium titanate was between 1 05-1O6 ohm m. However, flux-sintered samples exhibited higher resistivities ( > 106ohm m). Also, the resistivity was found to be independent of ternpmme.
Dielectric Materials and Devices
29 1
Measurements at different frequencies showed similar qualitative behavior for dielectric constant as well as resistivity, whch are reported elsewhere[8].
DISCUSSION Density of pure barium titanate pellets, sintered at 1350" C waq found to be 5.61 g/cc, which is -94% of the theoretical density. On the other hand, densities of flux-sintered samples were found to be 5.16, 5.46 and 5.63 g/cc which correspond to 87%, 92% and 95% weighted theoretical densities for pellets sintered at 1000"; 1100" and 1 150" C ; respectively. Lowering of sintering temperature in the case of flux-sintered pellets could be attributed to the liquidphase sintering. In the ZnO-BzO? phase-diagram, a eutectic reaction occurs at 961" C [9]. The composition of flux used in this study was 46 mol% ZnO. This composition correspond to a liquid-phase formation tempratwe of 982" C. Emergence of a compatible liquid phase is known to enhance sintering rate [101. When barium titanate is heated with ZnO-B*Ot fluxes, there could be initially interdiffusion of zinc, boron and barium ions among the phases. Once the temperature reaches above 9 8 2 O C , a liquid-phase containing 46 molO/nZn0 forms and is likely to wet the grains and fill the pores. It is also possible that some of the BaTi03 can dissohe in the liquid-phase with increase in the sintering temperature, thereby rounding of the sharp corners of the grains. During cooling, because of the decrease in solubility a secondary barium titanate p i n is likely to precipitate from the liquid-phase. In fact, TEM micrograph of flux-sintered samples showed a grain-size distribution and rounded grains, ail below 1 pm. On further cooling, eutectic phase of ZnO-B203 is likely to separate out, possibly preventing graingrowth. In the current experiments, the sintering temperature of 1100" and 1150" C did not indicate any difference in tan 6 values, suggesting that these temperatures are enough for the formation of stable microstructures. Dielectric constant of barium titanate is strongly influenced by its grain-size [ll]. Ceramics with finer grain-size (-lpm) show higher values of K at roomtemperature, whereas peak in permittivity at Curie-temperature is lower, as compared to ceramics with grain-size larger than 2pm. This phenomenon could be explained on the basis of ferroelectric-domains. A large grain may contain multiple domains, whereas in fine-grain ceramics, it is energetically favorable for every grain to contain single domain only. In large-grain ferroelectic ceramics, motion of domain wall provides large contribution to the permittivity, especially at the Curie-temperature. On the contrary, motion of domain walls is clamped by the grain-boundaries in fine-grain ceramics. The absence of domain wall motion results in a lower permittivity at the Curie-temperature. Below the Curietemperature, stresses generated by the clamping of the, domains result in contribution to the permittivity [2]. In these experiments, addition of fluxes
292
Dielectric Materials and Devices
resulted in smaller grain-size, which is responsible for the observed dielectric constant-temperature behavior. In addition, dielectric constant of flux-sintered pellets was found to increase with sintering temperature, though no appreciable change in the grain-size of was observed. This may be attributed to gradual decrease in porosity and hence lesser dilution of dielectric constant. it is important to note that tan 6 values of pellets sintered at 1100” and 1150” C are much less than those obtained for pure barium titanate and pellet sintered at 1000°C. Low values oftan 6 were attributed to the formation of highly resistive boundary phase. XRD patterns of flux-sintered pellets showed appearance of second phase. Moreover, bright field TEM micrograph in fig. 10 confirmed such a homdary phase. This is attributed to the increase in resistivity by almost an order of magnitude as well as its temperature-independent behavior. CONCLUSIONS 1. Addition of ZnU-B203 flwes helped in achieving higher sintering densities at lower temperature due to the liquid-phase sintering. The rounding up of the grains and presence of distribution in grain sizes suggest that solution re-precipitation is responsible for liquid-phase sintering. The particles of the eutectic phase of the ZnO-B203present at the grain-boundaries, have helped in retarding the grain-growth. 2. ‘i’hepresence of resistive grain-boundary phase has minimized the tan 6 values. The fine-grain microstructure has most likely contributed to a gradual change in dielectric constant and coupled with resistive grainboundary to a suppression of Curie-peak.
REFERENCES 1. L.E.Cross, ‘Terroelectric Ceramics: Tailoring properties for specific applications”, pp. 1-41, in Ferroelectric Ceramics, Edited by N.Setter and E.L.ColTa, Birkhausen, Verlag Basel, (1993) 2. S.L.Swartz, “Topics in elecronic ceramics”, IEEE Transactions on Electrical Insulation, 25(5), 935-1000, (1990) 3. GDesgardin, I. Mey, B. Reveau and J.M. Haussonne, “BaLiF3- A New Sintering Agent for BaTi03 based Capasitors”, American Ceramic Society Bulletin, 64(4), 564-70, (1985) 4. J. M. Haussonne, G. Desgardin, P. H. Bejoiet and B. Raeau, “Barium titanate Perovskite Sintered with Lithium Flouride”, Journul of the Amerrcan Ceramic Society, 66,801-807, ( ‘1983j 5. D. A. Tolino and J. B. Blum, “Effect of Ba:Ti Ratio on Densification of LiF-Fluxed BaTiO3”, Journal of the American Ceramic Society, 68, c-292, ( 1985)
Dielectric Materials and Devices
293
6. I.Burn, “Flux sintered BaTiO:, ceramics”, Journal of Materials Science, 17, 1398-1408, (1982) 7. C.-F. Yang, “The Influence of CuO-BaO Mixture on the Grain Growth and Dielectric Characteristics of BaTiQ Ceramics“, Ceramics International, 24,34 1, ( 1998) 8. D.Prakash, B.F.Sharma, P.Gopalan and T.R. Rama Mohan, “Dielectric Characteristics of Liquid-phase Sintered Barium Titanate Ceramics”, Proc. Con? DAE-BRNS Symposium on ”Recent Trends in Elecro-magneto Ceramics”, 18-20 February, (1 999), Kolhapur, India. 9. ZnO-BzQ Phase Diagram, page 1 19, in “Phase Diagram for Ceramrsts vol. I ” , Edited by E.M. Levin, C.R. Robbins and H.F. McMurdie, The American Ceramic Society, Oho, (1964j 10. R. M. German (ed.), Liquid Phase Sintering, Plenum Press, New York, (l989j 1 1. G.Arlt, D.Hennings and G.deWith, “Dielectric properties of fine grained barium titanium ceramics”, Journal of Applied Physics, 58(4 j, 16 19-1625, (1 985)
294
Dielectric Materials and Devices
Fig.1 : XRD pattern of barium titanate powder
Sintering Temperature ( C)
-
Fig.2 : Densities of flux-sintered ceramics
Dielectric Materials and Devices
295
42.0 J
-12.5
-13.0 n
$ f 2.
-
-
-13.5-13.5
-14.0
-
-14.5
1
980
loo0
.
I
1020
'
I
1040
'
I
.
1060
I
.
1080
l
'
1100
I
.
1120
I
.
I140
l
1160
Sintering temperature ( C)
Fig.3 : Shrinkage in diameter for flux-sintered ceramics
Fig.4 : SEM photomicrograph of pure barium titanate sintered at 1350" C
296
Dielectric Materials and Devices
Fig.5 : Bright field TEM photomicrographof sample flux-sintered at 1150" C
--
sintefed at
l ~ a c
-+- Sitered at 1 IOO'C --c smtered at 1 1W'C
860650?, 100
. . . . ....,
Ik
,
. . . ,...,
1Ok
,
. . . ....,
look
Frequency(Hz)-
Fig.6 : Dielectric constants of as a h c t i o n of frequency
Dielectric Materials and Devices
297
0.08
1..
I
U;
B
Fig.7 : tan 6 values of flux-sintered pellets
*-F lml I l4 I 18w
'"3
~OOO
'
20
--ePure,i3500C +- flux-sintered,100o0C -A-
flux-sintered,l1OO°C
-v-
flux-sintered,l150°C
io
ix
60
iix
i20
iJG
is2
Temperature ( 'C)
Fig.8 : Dielectric constant-temperature behaviot of sintered pellets
298
Dielectric Materials and Devices
.-aA
10'-
pure BaTiO, flux-sintered, ~ O O O ~ C
flux-sintered, ~IOO"C
I
.
I
.
,
.
Temperature ( 'C)
Fig.9 : Resistivity-temperature behavior of sintered pellets
Fig.10 : Bright-field TEM micrograph showing grain--
Dielectric Materials and Devices
phase
299
This page intentionally left blank
CHARACTERIZATION OF ULTRA-FINE BaTiO, POWDER FOR MULTILAYER CERAMIC CAPACITORS
Y. Sakabe andN. Wada, Murata Mfg. Co. LTD., Kyoto Japan ABSTRACT Hydrolysis BaTi0, powder included rectangular shaped inclusion. During heat treatment, dissociationof OH ions had taken place and completed at around SOO'C, the rectangular shaped lattice pore about lOnm in diameter had been left. XRD data and TEM analysis showed existence of pore reduced the tetragonality of the fine grain BaTiO, powder. The powder was structured of two regions having low and high tetragonality(c/a). The higher c/a regions had grown epitaxially on the surface of the host particle absorbing the neighboring particles. BaTiO, powder of higher tetragonality was obtained by calcining the hydrolysis powder of smaller particle size. Pore-fiee BaTiO, powder was prepared by calcining the nano-sized BaCO, and Ti02powder in a low air pressure. Tetragonality of the powder was higher than that prepared by the chemical wet process. It is expected that BaTiO, powder of about l O O n m in diameter can have higher c/a than 1.009. Multilayer ceramic capacitor with thinner dielectric layer of 1.3ym was demonstrated using the ultra-fineBaTiO,. INTRODUCTION BaTiO, is a most popular and useful dielectric for multilayer ceramic capacitors (MLCs) of today. Reduction of the dielectric thickness of the MLCs has been accelerated for designing the capacitorsof higher volumetric efficiency which allow high density circuits. Fine grain BaTiO, ceramics of high dielectric constant has been required for high performance MLCs. It well known that dielectric property of ferroelectric BaTiO, change seriously with voltage and mechanical stress. Therefore, it is important to investigate the dielectricchange with grain size and dielectric thickness of the capacitor with high layer count. Generally, fine-grained BaTiO, ceramicsrequire the more fine BaTiO, powder of high crystallinity. Extensiveresearch work has been done with reference to the size effect of BaTiO,'. Arlt et al. reported that there was an optimal value of grain size at about 0.8ym for the maximum dielectric constant2.Uchino et al. reported that ferroelectric properties, which were represented by tetragonality, were reduced with decreasing particle size in BaTiOJ and PbTiO,. The size effect was explained on the basis of the surface tension stress for small particle^,^. Hennings et al. examined BaTi0, powders of 0.2 ym in diameter prepared by hydrothermal method. They To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
301
showed that the cubic phase transferred to the tetragonal phase with heat treatment, due to the release of hydroxyl groups in the expanded cubic structure5. Wada et al. also examined ultrafine BaTiO, powder of 9Onm in diameter by a hydrothermal method. They suggested that tetragonality was influenced by the correlational size of dipoles due to the amount of the hydroxyl lattice defects6. Several models were also proposed fiom viewpoint of the inner structure of particles. Hsiang et al. reported that single domain structure appeared in particles smaller than 80nm in diameter, resulting in the change of tetrag~nality~. Takeuchi et al. proposed the existence of cubic phase on particle surface'. Frey et al. observed that a locally orthorhombic phase in the particle smaller than lpm in diamete?. Viswanath et al. reported that the limit size of the tetragonal phase was 65nm at BaTi0,prepared by sol-gel method". Li et al. suggested that cubic phases coexisted in the particles smaller than 80nm in diameter". Liu et al. experimentally showed that depolarization field was an important factor of the size effect 12. Zhang et al. recently explained the size effects on the basis of the Landau phenomenological theory 13. They showed that the cubic phase was stable in particles smaller than 40nm in diameter. It is certain that the size effect exists for fine BaTiO, particles, but explanations about it are insufficient. This is because of the difference in methods of preparing fine BaTiO, powder. In this work, we analyzed the particle structure more detail to reveal the dependence of dielectric properties on crystal size. The solid phase reaction process for preparing BaTiO, powder was also employed to understand the effect of the included pore defect. MLCs with thin dielectriclayer were prepared using the ultra-fine BaTiO, powder and Ni inner electrode. EXPERIMENTAL Hydrolysis BaTi0, powders were synthesized using Ba(OH), and Ti alkoxide and calcined at various temperatures ranging fiom 500°C to 1100°C. BaTiO, powders were also prepared using nano-sized BaCO, (120nm) and Ti0, (35nm) and calcined at various temperatures ranging fiom 650°C to 1000°C in a low air pressure(4 X 102Pa).Purity of both BaTi0, was hgher than 99.98%. Molecular ratio of Ba and Ti was 1.OOO 0.002. The obtained powders were examined by a differential thermal analysis with thermo-gravimetry(TG-DTA) and a Fourie transform infiared spectrometer(FT-IR). Scanning electron microscope (SEM) observations were performed to measure the average particle size. Inner structures of some particles were observed by Transmission electron microscope (TEM). Particles larger than lOOnm in diameter were thinned by ion milling in order to observe the inner structure clearly. Hot stage TEM was applied to observe particles growing in-situ during heat treatment. The crystal structure was analyzed by X-ray diffiaction (XRD) and lattice constants were fitted by Rietveld analyzingprogram using the RIETAM 910814. The calcined hydrolysis BaTi03 powder with the particle size of 15Onm was blended with metallic soaps of Ba, Mg, Mn, Dy, and Si alkoxide in ethylalcohol. Chemical composition was designed to obtain the core-shell ceramics. The mixture was evaporated and heated at 400°C. The obtained powder was mixed with an organic binder, ethlalcohol and toluene to prepare the slurry. Ceramic green sheets were prepared using the doctor-blade method. MLC samples were prepared with 1.6pm thick green sheets and Ni inner electrodescreen-printed on the sheets. The
*
302
Dielectric Materials and Devices
active layer of MLC was 100 layer and the size of the chip was 1.6 X 0.8 X 0.5mm. This sample was fired in a reducing atmosphere. Capacitance and dissipation factors were measured at lkHz, 1Vrms using LCR meter. RESULTS and DISCUSSION Characterizationof the Hydrolysis BaTi03Powder
Fig.1 TEM & SEM images of BaTiO, particles (a)as-prepared, (b) calcined at 950°C. TEM images of the as-prepared hydrolysis BaTiOJ are shown in Fig.l(a). XRD data showed that as-prepared particles had uniform structure of the expanded cubic symmetry. From TGA data, remarkable weight loss was observed fi-om 200°C to 800°C. Fig2 shows the curves of FT-IR spectroscopy for the powder calcined at various temperatures. An absorption band near 35OOcm-' was assigned to the stretching mode of internal OH- ions. The absorption
0.30 Q)
0
E ep
0.20
5: c
0.10
.$I p
0.00 4000
3800
3600
3400
Wave number
3200
3000
(cm-')
Fig2 Infiared spectra for hydrolysis BaTiO, powder calcined at various temperatures.
Dielectric Materials and Devices
303
strength decreased heating temperature. It was confirmed that the weight loss during heat treatment was caused by the release of the hydroxyl groups, and it completed at 800°C. With increasing of calcine temperature, rectangular shaped pore was found at inside the particles. The pore was clarified to be empty in mass by TEM analysis, as shown in Fig.3. Particle growth was not recognized while dissociation of hydroxyl groups. Particle growth started after ending of this dissociation. The high ferroelectric tetragonal phase was detected in the powder calcined at temperature higher than 800°C.
Fig.3 TEM image of hydrolysis BaTi03particle calcined at 1000°C.
In our previous work15,dependence of tetragonality on the BaTi03particle size was discussed based on E M observation and XRD data analysis. The schematic diagram of the inner structurechanges is illustrated in Fig.4. Fig.5 shows micrographs of the hot-stage TEM observation during heat treatment. The particle growth occurred instantaneously absorbing the neighboring particles. The newly developed crystals surrounded the original region that included the pores. The Rietveld analysis of XRD data confirmed that the grown particle was composed of binary phase. Fig.6 shows c/a values obtained by single and binary phase models. The newly developed region had larger c/a value than the original crystal region. According to this structure, the finer as-prepared powder could prepare fine BaTi03powders with high c/a value.
304
Dielectric Materials and Devices
5 sec.
Keeping time Osec.
12 sec.
Fig.5 In-situ images of hot-stage TEM of hydrolysis BaTiO, at 8 10°C. 20
Reciprocal of particle size / 15
10
p m-'
0
1.009 1.008
-
1.002 0.05
I
0.07
I
0.1
Particlesize/
I
I
0.2
0.9
p m
Fig.6 Tetragonality change of the hydrolysis BaTiO, powder with particle size by binary phase model.
In order to clarifL above-mentioned consideration, two kinds of as-prepared BaTiO, powders with a different particle size were prepared and calcined at various temperatures(Fig. 1). Initial particle size was changed with contents of Ba and Ti ions in the solution. Fig.7 shows c/a changes(sing1e model) with the calcining temperature. Larger c/a value was obtained when the size of as-prepared BaTiO, particle was small. Fig.8 shows the relationship between the particle size of as-prepared BaTiO, powder and c/a value (single model) of the calcined powder that had the same particle size of 0.2pm in diameter. It was found that the change of c/a value has a linear relation with the particle size of as-prepared BaTiO, particle. This result confirmed that the c/a value of the BaTiO, powder depended on the volume of the newly developed region while particle growing. BaTiO, Powder Synthesized by Solid Phase Reaction Ultra-fine BaTiO, powders were also prepared by the conventional powder process. SEM micrographs the obtained powder was shown in Fig.9. Tetgagonality (c/a) change with the
Dielectric Materials and Devices
305
1.01 1 1.01
1.009
2
1.008
x .Z 1.007
a
s
&
Particle size of as-prepared BaTiO, powder
1.006
-
1.005 1.004
55nm
1.003 600
700
800 900 1000 Calcined Temp.,T/ "c
1100
Fig.7 Tetragonality change with calcining temperature for the two grade of hydrolysis BaTiO, powder. 1.011
I
1.01
g
z
a
1.009
*1 :Hennings (1992) 1.008
5 1.007 0
1 1.006 0
?
b 1.005
G
0
0.05 0.1 0.15 0.2 0.25 Particle Size of as-prepredBaTiO3 powler , DBETI fi m
Fig.8 Relationship between particle size of as-prepared hydrolysis BaTiO, powder and tetragonality of calcined BaTi03powder (particle size: 0.2pm ). DB,;diameter calculated fiom specificsurface area of powder. calcining temperature is shown in Fig.10 in comparison with the hydrolysis powder. High c/a value was obtained in the powder that was calcined under low air pressure. The powder of high c/a value had perfect crystal structure as shown in a TEM micrograph of Fig.11. This atmosphere dependence of the calcination reaction was found when we had observed the reaction of BaC0, with Ti02 on the hot-stage of the TEM. Decomposition of Ba-carbonate was caused at 650°C,then reaction between BaO and TiOz started to compose the pore fiee BaTiO,. This reaction was reproduced in a tube fimace. The reaction temperature was lower than by the normal calcination in air. The reaction between nano-sized BaO and Ti02 at lower temperature resulted in ultra-fine BaC0, with high tetragonality. The obtained powder did not
306
Dielectric Materials and Devices
800"C, 4 X 10'7Pa
850"c, 4X 10'Pa
950"c, 4 X 10ZPa
9OOC, 4 X 1O2Pa
Fig.9 SEM images of BaTi03powder by solid phase reaction method under low air pressure. 1.01 1 1.01 1.009 m
1.008
2 1.007
$ 1.006 9 1.005
; 0
1.004 -m-
1.003
1.002
-o-
solid phase reaction under normal air pressure solid phase reaction under
1.001 1 20
40
I
8
60
80
I
100
I
I
120
140
160
0
180
I
200
Particle size I nm
Fig.10 Tetragonality vs. particle size of BaTi03.
Fig. 11 TEM image BaTi03powder by Solid phase reaction under low air pressure.
Dielectric Materials and Devices
include the specific pores as seen in the hydrolysis BaTi03powders. The volume of the pore free region was estimated for hydrolysis powder by subtracting the volume of the host particles fiom the grown particle, and plotted in Fig.12 with the particle volume by the solid phase reaction. From this figure, we could confirm that the crystal volume of the pore free region determined the c/a value. BaTi03 powder of about loom in diameter is expected to have higher c/a than 1.009.
307
Particle size / nm
, 150 175 I
1.012 50 75 100
. 6
1.008
8
1.006
prepared particle) Hydrolysis BaTi03 (large asprepared particle) A Solid phese reaction BaTi03
1.002
0
0.5
1 1.5 2 2.5 High c/a phase volume /106nm3
3
Fig. 12 Volume of high c/a phase region vs. tetragonality of BaTi03. Application of the Ultra Fine BaTiO, Powders to MLCs A fine grained core-shell ceramics with the grain size of 0.15pm was prepared using BaTiO, powder calcined at 900°C to examine the electric properties of MLC with thin ceramic layers. MLC samples were fred at 1150°C in a reducing atmosphere. The core region had a domain structure caused by a tetragonal crystal structure, and trace of the doped components was not found in this region. On the other hand, Dy ion was detected in the shell region where the domain structure was disappeared. Dielectric layer of 1.3pm thick with Ni inner electrode is shown in Fig.13. Electrical properties of the demonstrated MLC are shown in Table 1. The dielectric constant was stable in the wide temperature range. High dielectric constant ceramics with stable characteristicsunder high AC and DC voltage stress are indispensable in the design of MLC with yet thinner layer. Dielectric properties of the ultra-fine ceramics under high voltage stress were more stable than conventional ceramics. This is big advantage for high volumetric capacitors with thin layers. It
Fig. 13 SEM image of the thin BaTi0, layer MLC.
308
Dielectric Materials and Devices
Table 1. Electrical properties of MLC with ultra-fine core-shell grains. (average grain size 150nm) Chip size (m) Dielectric thickness ( pm) Active layer’s number Internal electrode Firing temperature (“C) Capacitance (nF) at 1Wl Vrms Dielectric constant at 1 W l V r m s Dissipation factor (%) at 1 W l V r m s Resistivity ( 61 ) log R at 4v Break down voltage (v)
1.6 X 0.8 X 0.5 1.3 100 Ni 1150 770 1560 1.7 9.37 130
was supposed that the small voltage dependence was caused by the high surface tension of ultra-fine grains. By incorporating 1.3pm dielectric layers and Ni electrode, replacement of Ta and A1 electrolytic capacitors with large capacitance MLCs will be realized in the capacitance ranging fiom 10 to 100pF.
SUMMARY (1) Binary phase model could explain the particle size effect on tetragonality(c/a) of hydrolysis BaTiOJ powder. The c/a value was determined by the volume fiaction of a newly developed region while particle growing. (2) Smaller as-prepared particle size of the hydrolysis BaTi03 powder provided formulated dielectricpowder with higher tetragonality c/a value. (3) Ultra-fine powder was also prepared by solid phase reaction under low air pressure using nano-sized BaC03and Ti02.The c/a value was higher than the powder by wet process. (4) MLC with 1.3pm thick dielectrics and Ni electrode was demonstrated with ultra-fine and high crystallinity BaTi03 powder, which was 15Onm in diameter and had c/a value of over 1.009. REFERENCES A. Yamaji, E. Enomoto, K. Kinoshita and T. Murakami, “Preparation, Characterization and Properties of Dy doped Small-Grained BaTi03Ceramics ”, J. Am. Ceram. Soc., 60, E341 ,pp.97-101,1977. G. Arlt, D. Hennings and G. de With, “Dielectric Propertiesof Fine-grain Barium Titanate Ceramics”, J. App. Phys., 58, [4] ,pp.1619-25, 1985. K. Uchino, E. Sadanaga and T. Hirose, “Dependence of the Crystal Structure on Particle Size in Barium Titanate”,J. Am. Ceram. Soc., 72, [81 ,pp.1555-58,1989. K. Uchino, E. Sadanaga, K. Oohashi, T. Morohashi and H. Yamamura, “Particle/Grain Size Dependence of Ferroelectricity”, Ceramic Transaction, 8, pp. 107-115,1989.
Dielectric Materials and Devices
309
D. Hennings and S. Schreinemacher, “Characterization of Hydrothermal Barium Titanate”, J. EWO.Ceram. Soc., 9, pp.41-46, 1992. S. Wada, T. Suzuki and T. Noma, “Role of lattice Defects in the Size Effect of Barium Titanate Fine Particle”, J. Ceram. Soc., 104, [51 ,pp.383-392, 1996. H. Hsiang and F. Yen, “Effect of Crystallite Size on the Ferroelectric Domain Growth of Ultra-fine BaTi03Powders”, J. Am. Ceram. Soc., 79, [41 ,pp.1053- 60,1996. T. Takeuchi, K. Ado, T. Asai, H. Kageyama and Y. Saito, “Thickness of Cubic Surface Phase on Barium Titanate Single-Crystal Grains”, J. Am. Ceram. Soc., 77, [61 ,pp.1665-68, 1994. M. H. Frey and D. A. Payne, “Grain-size Effect on Structure and Phase Transformations for Barium Titanate”, Phys. Rev.B, 54, pp.3 158-67, 1996. R. N. Viswanath and S. Ramasamy, “Preparation and Ferroelectric phase Transition studies ofNanocrystallineBaTiO;’, Nano Structued Materials, 8, [2] ,pp.155-62, 1997. I ’ X.Li and W. Shih, “Size Effect in Barium Titanate Particles and Clusters”, J. Am. Ceram SOC.,80, [ll] ,pp.2844-52, 1997. X.Liu, W. Y Shih and W. H. Shih, “Copper Coating Effect on the Crystal Structure of Fine BaTi03 Particles”, 99th Annual Meeting Abstract for Am. Ceram. Soc., p. 110, 1997. W. L. Zhong, Y. G. Wang, P. L. Zhang and B. D. Qu, “Phenomenological Study of the Size Effect on Phase Transitions in Ferroelectric Particles”,Phys. Rev.B, 50, pp.698-703, 1994. l4 F.Izumi, “The Rietveld Method”,Oxford University Press,ch. 13,1993. l5 Y. Sakabe, N. Wada and Y. Hamaji, ” Grain size Effects on Dielectric Properties and Crystal structure of Fine-Grained BaTiO, Ceramics”, J. Korean Phy. Soc., 32, pp.260-264, Feburuary 1998.
’*
310
Dielectric Materials and Devices
MECHANICAL AND THERMAL PROPERTIES OF ELECTRONIC CERAMIC MULTILAYER CAPACITORS1
POWER
A. A. Wereszczak2 and L. Riester Mechanical Characterization and Analysis Group Oak Ridge National Laboratory Oak Ridge, TN 37831
J. W. Hill and S. P. Cygan Olean Advanced Products Division of AVX Olean, NY 14760 ABSTRACT Elastic modulus, hardness, and fracture toughness of the barium titanate (BaTiO,) dielectric were in-situ measured in a class of high-voltage (500-6OOV) power electronic multilayer capacitors (MLCs) using a mechanical properties microprobe. Additionally, the apparent thermal expansion and directionallydependent thermal conductivity (measured in all three orthogonal directions) of the MLCs were examined between -50 and 200°C. The modulus, hardness, and fracture toughness were equivalent to those values that the authors have previously measured with BaTiO, in surface-mount MLCs. The apparent coefficient of thermal expansion (CTE) of the MLC was a function of temperature and was also equivalent to that of monolithic BaTiO,. The apparent intraplanar thermal conductivity in the MLC was higher-valued than that of monolithic BaTiO, (because of the metal electrodes having a higher thermal conductivity) and was transversely isotropic, while the apparent interplanar thermal conductivity in the MLC was equivalent or slightly higher that that for the monolithic BaTiO,. The relatively low fracture toughness, high thermal expansion, and low thermal conductivity of the BaTiO, or the MLC combine to illustrate why these power electronic MLCs can be so susceptible to thermal shock if caution is not exercised to minimize their exposure to abrupt thermal excursions during manufacturing or while in service.
Research sponsored by the U. S. Department of Energy, Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Transportation Technologies, as part of the Advanced Automotive Materials Program, under Contract DE-AC05-000R22725, managed by UT-Battelle, LLC. Now with the Army Research Laboratory, Aberdeen Proving Ground, MD. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
311
I. INTRODUCTION Over the last decade, the market for switch mode power supply (SMPS) stacked multilayer ceramic capacitors (MLCs) has grown annually by over 50% [11. Such ceramic MLCs (typically barium titanate-, BaTi0,-, containing) are very attractive alternatives to aluminum electrolytic or polycarbonate film capacitors because they have lower equivalent series resistance (ESR) and inductance (ESL), occupy much less space and are lighter for equal capacitance, and can operate over a wider temperature range [l-31. Consequently, they are ideal candidates for high-power and high-current converters and power supplies where space is limited (i.e., automotive power electronics). As an indication of their attractiveness, government agencies such as the U. S. Department of Energy [4] are striving to fund their continued improvement and development for the improved viability of hybrid vehicles. Experience and field-service have shown that these MLCs can mechanically fail if caution is not exercised during their manufacture or service. For example, the importance of proper soldering during MLC manufacture is strongly emphasized in the literature [l-3, 51. Thermal shock failure can occur during soldering if the MLC is heated too quickly (e.g., > 4”C/s) to the soldering temperature or if the MLC is not pre-heated to a temperature approximately near the soldering temperature (within 30-50°C). During service, SMPS can fail in a shorted condition. Such failures can cause extensive or catastrophic damage to printed circuit boards because there is negligible or no current limiting in power supplies. Clearly, it is advantageous for both the MLC manufacturer and end-user to recognize and address these issues and work toward avoiding them. To understand or predict thermal shock behavior, availability of both mechanical and thermal property data is required. The present study set out to address this by mechanically and thermophysically characterizing candidate MLCs for automotive power electronic devices. The approach was to perform insitu, micromechanical testing of the BaTiO, dielectric in such a MLC and also examine apparent thermal expansion and thermal conductivity behaviors of the MLC. The combination and availability of all the data would then allow manufacturers and end-users to better predict and avoid potential thermal shock scenarios. 11. EXPERIMENTAL PROCEDURES 1I.A. Description of the MLCs The MLCs shown in Fig. 1 are examples of those that were characterized in the present study. The operating capacitances and voltages for the shown MLCs are 0.1 pF - 600V (2225), 0.33 pF - 600V (3640), and 4-4.5 pF - 500V (SM02). All three MLCs can be used in “stacked” arrangements, and are applied in high power, high frequency converters. The level of required filtering, as well as voltage and current ratings, will determine the capacitance/size needed (i.e., which of the MLCs in Fig. 1) for a particular application. Regarding their preparation for testing, a 2225 MLC was sectioned for the micromechanical testing, a SM02 MLC was sectioned for the apparent thermal expansion testing, while one 3640 MLC and several of the 2225 MLCs were sectioned for the apparent thermal conductivity testings. Directionality was
312
Dielectric Materials and Devices
examined owing to the inherent transversely isotropic symmetry of the MLC (i.e., isstropy in the XY plane shown in Fig. 1). The measured responses of the MLC thermal expansion and thermal conductivity testings were compared to those of monolithic BaTiO,.
1I.B. Elastic Modulus, Hardness, and Fracture Toughness Testings A fine surface finish was a prerequisite for the planned micromechanical testing. To achieve this, a 2225 MLC was sectioned, then set in metallographer’s mounting epoxy, and diamond-paste polished to a 0.25 pm finish. The mechanical properties can be affected by residual stresses; consequently, the region for in-situ testing was chosen to be in the margin at a distance from the termination metal where their effects were deemed a minimum.
Figure 1. SM02 (top), 3640 (bottom left), and 2225 (bottom right) MLCs. The layers comprising the capacitors are oriented in the shown Cartesian XY-plane. A mechanical properties microprobe (MPM)3 was used in the measurement of the Young’s modulus (E), hardness (H), and for the determination of fracture toughness (KJ of the dielectric ceramic in the margin of a 2225 MLC. The MPM is an automated instrument that consists of four primary components shown in Fig. 2.: an indenter whose vertical displacement (nanometer resolution) and Nanoindenter 11, Nano Instruments, Oak Ridge, TN.
Dielectric Materials and Devices
313
applied load (microgram resolution) are controllable, an optical microscope (OM) with several objective lenses, a precision X and Y stage that translates the metallographically prepared specimen between the OM and the indenter, and a computer that controls the OM’s lens turret (50- 1500x magnification), stage movement, and indenter load and displacement. A diamond Berkovich indenter was used for the E and H measurements, while a diamond Vickers indenter was used for the K,, measurements. The computer also collects data on the indenter’s displacement and load, and is also interfaced with a TV-monitor and camera that show the microscope’s field of view and allow inspection of each indent. The MPM (not including the computer) is housed inside an insulated cabinet that minimizes its susceptibility to laboratory room temperature fluctuations and vibrations (problematic when controlling displacements at the nanometer level). The computer is outside the insulated cabinet to further minimize the introduction of thermal instabilities, and remotely operates the MPM. Relatively low loads applied with the Berkovich indenter were used for the E and H determinations (no cracks produced). The load/displacement history during indentation provided data which were interpreted with the aid of an appropriate model to calculate E and H of the material [6]. Forty indents were applied in the dielectric ceramic in each MLC’s cover layer, and the corresponding average E and H were calculated for each of them. Higher loads were applied with a Vickers indenter to produce cracks which emanated from the corners of the indent, and these cracks were used to determine KIc. The indent-generated cracks were imaged at a magnification of 7000x with a scanning electron microscope (SEM) to measure their lengths. The crack length, its corresponding applied indent load, and the measured E and H for each MLC dielectric ceramic were needed to calculate K,, using [7]
K,c =
<
where is a material independent constant for the Vickers indenter, E is the elastic modulus, H is the hardness, P is the applied load, and c is crack length. The use of Eq. 1 is valid for use with the small indents produced by the MPM [8]. The values of E and H used in Eq. 1 were those measured with the MPMgenerated with crack-free indents. Ten indents were applied at high loads to produce cracks, and up to four cracks were produced per indent. However, many indents did not have four cracks, so the total number of fracture toughness values comprising the average was less than forty.
314
Dielectric Materials and Devices
1I.C. Apparent Thermal Expansion and Conductivity Testings Apparent thermal expansion and thermal conductivity testing were performed by a commercial laboratory.4 The adjective “apparent” is used in this study because the thermal measurements sampled the net or combined thermal effects of all the dielectric and electrode layers in the MLC. Apparent thermal expansion was continuously measured between -75 and 200°C while apparent thermal conductivity was determined at discrete temperatures of -50, -25, 0, 50, 100, 150, and 200°C. For the thermal expansion analysis, a specimen was sectioned out of a SM02 MLC with its major axis parallel to the SM02’s X-direction.5 For reference, the dielectric and electrode layers comprising the MLC are oriented in the X-Y plane in Fig. 3. The thermal expansion of a monolithic BaTiO, specimen of equivalent dimensions was also examined and compared to that of the MLC.
Figure 2. A mechanical properties microprobe was used for the testing.
Holometrix Micromet, Bedford, MA. Y- and Z-direction measurements of the SM02 are presently pending and were not available at the time of this article’s writing.
Dielectric Materials and Devices
315
3 mm
Y
Figure 3. Dimensions of specimen sectioned from a SM02 MLC for X-direction measurement of apparent coefficient of thermal expansion. In order to determine thermal conductivity, the thermal difhsivity and heat capacities were first measured using the laser flash method [9]. The thermal conductivity (k) was then calculated according to k = D*C;p , where D is thermal diffusivity, C, is heat capacity, and p is density. A specimen with a relatively large aspect ratio is a prerequisite for thermal conductivity testing, so special preparation efforts were performed in order to obtain valid results in all three orthogonal directions for at least one of the MLCs shown in Fig. 1 (2225 MLC). A Z-direction measurement was performed too with a 3640 MLC to compare apparent thermal conductivity with the Z-direction in the 2225 MLC. To avoid gaps and the introduction of erroneous testing for the X- and Y-direction measurements, the sides of all pieces were carefully ground flat, parallel, and perpendicular. The width of the 2225 MLC.was approximately 2 mm, so several pieces were mechanically held in parallel during testing (using a small wire “tourniquet” about the pieces) in order to determine the X- and Ydirection measurements (Fig. 4). Likewise, such machining procedures were performed for the Z-direction measurement of the 2225 MLC as well as for the measurement of the monolithic BaTiO, specimen. 111. RESULTS & DISCUSSION 1II.A. Elastic Modulus, Hardness, and Fracture Toughness The average Young’s modulus of the BaTiO, dielectric ceramic in the 2225 MLC was 212 GPa (No. = 40, Stan. Dev. = 11.6 GPa). This average value for E is approximately 5% higher than what the authors measured in BaTiO, in surface-mount 0805 MLCs [lO-111. What is significant about the value of E? Principally, if the dielectric ceramic in a MLC has a relatively large-valued Young’s modulus, then the imposed stresses in it will be higher than if its E were lower-valued (for the same applied strain). The average hardness of the BaTiO, in the 2225 MLC was 12.6 GPa (No. = 40, Stan. Dev. = 1.3 GPa), and is approximately 10% higher than the
316
Dielectric Materials and Devices
hardness of BaTiO, in those same surface-mount MLCs that the authors previously examined [10-111. The average fracture toughness in the 2225 MLC was 1.28 MPaOm (No. = 17, Stan. Dev. = 0.21 MPaOm). The authors have measured fracture toughnesses of the BaTiO, in surface-mount MLCs of 1.11, 1.36, and 1.53 MPaOm [lO-111, so the toughness of the BaTiO, in this 2225 MLC is equivalent. What is significant about the value of K,,? Structurally speaking, ceramics with higher K,, are preferable for load-bearing applications because K,, K strength. Dielectric ceramics in MLCs may be subjected to tensile stresses during manufacturing or service, so using a dielectric ceramic with higher K,, will lessen the likelihood of mechanical failure (all other things being equal of course).
2 rnrn
2 rnm 4 pieces
4 pieces
X-direction 4
5 mm
2 rnrn 1 piece
Z-direction
Y-direction 8 rnrn m
//
2 rnrn
1 piece
Monolithic BaTiO,
Figure 4. Dimensions and layouts of specimens for thermal conductivity measurements. Indicated directions are those perpendicular to the largest planar surface. 2225 MLCs were sectioned for the X-, Y-, and Z-direction measurements. One 3640 MLC was prepared for a Z-direction measurement.
Dielectric Materials and Devices
317
1II.B. Thermophysical Properties 1II.B.1 Apparent Thermal Expansion The apparent coefficient of thermal expansion (CTE) in the X-direction of a SM02 MLC and in monolithic BaTiO, were functions of temperature and were essentially equivalent through the temperature range of -75 to 200°C as shown in Fig. 5. Their CTE's were approximately 6 ppm/"C at -75°C and increased in a transient fashion to approximately 8 ppm/"C at 0°C. The rate of increase in the CTE was constant between 0 and = 110°C where it had increased to M 8.9 ppm/"C. The phase transformation in the BaTiO, around 125°C caused an increase in the CTE and it eventually went through a maxima (3 11.5 ppm/"C) at around 180°C. The higher thermal expansion of the electrode metal (a = 16.8 ppm/"C at 27°C [12]) compared to that for the BaTiO, (=a= 8 ppm/"C at room temperature) does not produce an apparent thermal expansion in the X-direction higher than that of monolithic BaTiO,. This CTE information is quite useful because it (1) shows that the CTE of BaTiO, is far from being a constant value and (2) provides continuous data that may be used in finite element analysis to model (residual) stress development and changes in this temperature range. III.B.2. Apparent Thermal Conductivity The apparent thermal conductivities in the X-, Y-, and Z-directions in the 2225 MLC were higher-valued than that of monolithic BaTiO, (see Fig. 6); however, the apparent thermal conductivity in the Z-direction of the 3640 MLC was equivalent to that of the monolithic BaTiO,. The apparent thermal conductivities in the X- and Y-directions were equivalent (and higher-valued than the Z-direction) which is consistent with the transversely isotropic symmetry in MLCs; they were approximately 50% more than for monolithic BaTiO,. The apparent thermal conductivity (k,) parallel and perpendicular to a series of dual-material lamina is respectively represented by [ 131 k,
= v,k, +v,k,
and
where v, and v2 and k, and k, are the fractions and thermal conductivities of the two material constituents. Measurements of v, (considered here as the metal electrode constituent) and v, (BaTiO,) were made (using an optical comparator) in the thermal conductivity specimens after their cross-sections were polished. For the 2225 and 3640 MLCs, v, = 0.04 and v, = 0.06, respectively (or v, = 0.96 and v, = 0.94, respectively). Using a thermal conductivity of k, = 322 W/mK for the 30/70 Pd/Ag metal electrode [121 and k, = 1.8 W/mK for the barium titanate results in k, = 15 W/mK for the 2225 X- and Y-directions, and k, = 1.9 W/mK for both the 2225 and the 3640 2-direction.
318
Dielectric Materials and Devices
The measured apparent thermal conductivities for the X- and Y-directions at room temperature were = 2.6 W/mK which does not correlate well with the calculated k, = 15 W/mK from Eq. 2. The use of Eqs. 2 assumes continuity of the lamina; however, this likely was violated in the measurement because the electrodes in these thermal conductivity samples were not continuous through their depth (and are not in a MLC as well). Consequently, the contribution of the high thermal conductivity from the metal electrodes to the overall thermal conductivity of the MLC in the X- and Y-directions is much smaller than predicted. This same effect likely accounts for lack of CTE difference between monolithic BaTiO, and the 2225 X-direction, as shown in Fig. 5. The measured apparent thermal conductivities in the Z-direction correlate much better with the values calculated from Eq. 3. The apparent thermal conductivity of Z-direction in the 2225 and 3640 were = 2 and 1.8 W/mK, respectively, at room temperature while the value calculated using Eq. 3 was 1.9 W/mK. For this direction, the use of laminate theory to calculate thermal conductivity worked quite well. This thermal conductivity information is quite useful because it shows (1) its dependencies on orientation and temperature and (2) that even with the metal electrodes that the absolute values of thermal conductivity are still relatively low.
1II.C. Ramifications on Mechanical Robustness The relatively low fracture toughness, high thermal expansion, and low thermal conductivity of the BaTiO, or the MLC combine to illustrate why these power electronic MLCs can be so susceptible to thermal shock or thermalgradient-induced failure. The low fracture toughness of BaTiO, indicates that it cannot withstand relatively high mechanical loadings. The high thermal expansion shows that the MLC can exhibit relatively large dimensional changes over a relatively small temperature change. Lastly, the relatively low thermal conductivity shows that both BaTiO, and the MLC cannot dissipate heat very efficiently. These three characteristics combine to result in a worst-case scenario for mechanical integrity. Clearly, if caution is not exercised to avoid abrupt thermal excursions, then thermal stresses in the MLC can be significantly high so to exceed the strength of the (weak or poor toughness) BaTiO, and cracking and mechanical failure can be consequences. This effect is more severe if a large AT gradient is abruptly set up in the Z-direction than for the same AT gradient in the X- or Y-directions because the apparent thermal conductivity is lower-valued in the Z-direction than in the Xor Y-directions. The results from this study quantify effects that have been empirically recognized in the MLC industry for many years.
Dielectric Materials and Devices
319
121. L
6 4
.................................
(
1
K
=
F
Monolithic BaTi03
SM02 - X Direction
; 1
J
U
C
E
2 -
0
0
.-Q,
0
9th Order Polynomial Fit . of Expansion-Temp Data '
1
'
' 1 '
' 8 ' 1 s '
8
'
1
"
I
' 1
'
1
'
' 1 '
' " 1 ' '
:
"
Figure 5. Apparent coefficient of thermal expansion as a function of temperature for monolithic BaTiO, and the X-direction in a SM02 MLC.
4
h
7
'E
k
X-Direction (2225)
4
Y-Direction (2225)
3 ,
-
2 -
b
1 -
E
c
Monolithic BaTiO & Z-Direction (3640)
Z-Direction (2225) . ,
' " ' ~ ~ " ' ~ . . " ~ . ~ " ~ ' ' ' ' ~ ' ' ' ' ~ ' . ' '
-100
-50
0
50
100
150
200
250
Temperature ("C)
Figure 6. Apparent thermal conductivity as a function of temperature and directionality in a 2225 MLC, a 3640 MLC, and monolithic BaTiO,.
320
Dielectric Materials and Devices
The data that were generated in this study may be combined with probabilistic strength data, finite element analysis, and an appropriate multiaxial failure criterion to predict the survival probability of a MLC under any loading scenario. Such analysis is common to the structural ceramic community and the authors have demonstrated this approach to understand what effect the stress-freetemperature has on mechanical reliability in surface-mount MLCs [141. Such analysis with automotive power electronic MLCs would therefore be worthwhile.
IV. CONCLUSIONS Elastic modulus, hardness, and fracture toughness of the BaTiO, dielectric in high-voltage power electronic MLCs were measured in-situ using a mechanical properties microprobe. The apparent thermal expansion and directionallydependent thermal conductivity of the MLCs were also examined between -50 and 200°C. The modulus, hardness, and fracture toughness were equivalent to those previously measured with BaTiO, in surface-mount MLCs. The apparent CTE of the MLC was a function of temperature and was also equivalent to that of monolithic BaTiO, - The apparent intraplanar thermal conductivity in the MLC was higher-valued than that of monolithic BaTiO, and was transversely isotropic, while the apparent interplanar thermal conductivity in two of the MLCs was equivalent or slightly higher that that for the monolithic BaTiO,. The relatively low fracture toughness, high thermal expansion, and low thermal conductivity of the BaTiO, (or the MLC) combine to illustrate why they can be so susceptible to thermal shock if caution is not exercised to minimize their exposure to abrupt thermal excursions during their manufacturing or service. ACKNOWLEDGMENTS The authors wish to thank L. O’Rourke for the specimen preparation for the thermal expansion and conductivity testings, T. Kirkland for the dimensional measurements of the MLCs, M. Ferber and G. Hsueh for reviewing the manuscript and for their helpful comments, and D. Stinton for the support of this work. REFERENCES J. Hill and S. Cygan, “Stacked SMPS Capacitors - One Vendor’s Experience,” Capacitor and Resistor Technology Symposium (CARTS), pp. 283-291, 1999. G. J. Ewe11 and J. Siplon, “Stacked Ceramic Capacitors: A User’s Guide,” Capacitor and Resistor Technology Symposium (CARTS), pp. 137-143, 1998. R. W. Dobson and R. E. Wagoner, “ A Military Contractor’s Experience with DSCC 87 106 Switch Mode Capacitors,” Capacitor and Resistor Technology Symposium (CARTS), pp. 203-212, 1998. Progress Report for Propulsion Materials-FY 1999, U. S. Department of Energy, Office of Transportation Technologies, Office of Advanced Automotive Technologies, October 1999.
Dielectric Materials and Devices
321
R. M. Suarez, “Experiences in the Development of High Capacity MLC Stacks,” Capacitor and Resistor Technology Symposium (CARTS), pp. 45-52, 1993. W. C. Oliver and G. M. Pharr, “An Improved Technique for Determining Hardness and Elastic Modulus Using Load and Displacement Sensing Indentation Experiments,” Journal of Materials Research, 7 1564-83 (1992). [71 G. R. Anstis, P. Chantikul, B. R. Lawn, and D. B. Marshall, “ A Critical Evaluation of Indentation Techniques for Measuring Fracture Toughness: I, Direct Crack Measurements,” Journal of the American Ceramic Society, 64 533-538 (1981). K. Breder, A. A. Wereszczak, L. Riester, and T. P. Kirkland “Determination of Strength for Reliability Analysis of Multilayer Ceramic Capacitors” Ceramic Engineering and Science Proceedings, 20 565-572 (1999). W. J. Parker, R. J. Jenkins, C. P. Butler, and G. L. Abbott, “ A Flash Method of Determining Thermal Diffusivity, Heat Capacity, and Thermal Conductivity,” Journal of Applied Physics, 32 1679 (196 1). ORNL/TM- 1999/202 Report: “ Toward the Assessment of Mechanical Robustness of Ceramic Multilayer Capacitors (MLCs),” A. A. Wereszczak, K. Breder, E. Riester, T. P. Kirkland, and R. J. Bridge, ‘ICJ.S.DOE Office of Transportation Technologies, October 1999. A. A. Wereszczak, K. Breder, E. Riester, T. P. Kirkland, and R. J. Bridge, “ In-Situ Mechanical Property Evaluation of Dielectric Ceramics in Multilayer Capacitors,” SAE Paper No. OOFCC-116, SAE 2000 World Congress, Arlington, VA, Apr. 2000. S. P. Cygan, “ High Temperature Ceramic Capacitors,” Report WL-TR-942049, Aeropropulsion and Power Directorate, Wright Laboratory Air Force Material Command, Wright Patterson AFB, OH, March 1994. W. D. Kingery, H. K. Bowen, and D. R. Uhlmann, Introduction of Ceramics, Second Edition, John Wiley & Sons, New York, 1976. A. A. Wereszczak, K. Breder, M. K. Ferber, R. J. Bridge, L. Riester, and T. P. Kirkland, “ Failure Probability Prediction of Dielectric Ceramics in Multilayer Capacitors,” Multilayer Electronic Ceramic Devices, Ceramic Transactions, Vol. 97. DD. 73-83. 1999.
322
Dielectric Materials and Devices
PLASMA AND HYPERSONIC FLAME SPRAYED CERAMIC COATINGS FOR DIELECTRTCAL APPLICATIONS Rainer Gadow* and Andreas Killinger Institute for Manufacturing Technologies of Ceramic Components and Composites Allmandring 5b D-70569 Stuttgart, Germany ABSTRACT Oxide ceramics play a major role as dielectric materials for the use in barrier discharge applications. Refractory oxides with high melting points can be processed to functional coatings in a performing and cost effective way using thermal spray technologies. Incompatibility problems of the thermophysical properties of such coatings on metal, glas and polymer substrates can be solved by selective thermal spray parameter variation and simultaneous cooling to obtain appropriate residual stresses in the layer composite. Thennophysical and electrophysical properties of such coatings, e.g. the dielectric behaviour in barrier discharge of coated tubular ozone generators are shown and results for various oxides (alumina, zirkonia, titania) are discussed. The influence of different thermal spray techniques, i. e. plasma spraying ( A P S ) and hypersonic flame spraying (HVOF) onto the physical properties of the thermally sprayed coatings will be presented. INTRODUCTION Oxide ceramics play a major role as dielectric materials for multiple purposes in industrial applications. Insulators, dielectric discharge barriers, corona treatment devices, printer rollers etc. are manufactured as coating systems using surface technologies and have replaced expensive bulk ceramic composites in many cases. The processing of refractory oxides with high melting points using thermal spray technologies is a performing and cost effective way for manufacturing coating sytems in the range of 30 pm up to several 1000 pm. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, wilhout the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
323
Problems that arise fkom the thermophysical incompatibilities of the ceramic coating and the substrate material (for instance metal, glass or polymer) can lead to failure by crack formation or complete delamination of the coating composite under thermal andor mechanical load. The problem can be solved by selective thermal spray parameter variation and simultaneous cooling to obtain an appropriate residual stress distribution in the layer composite. The present study gives an overview of the thermophysical and electrophysical material properties of thermally sprayed oxide coatings and describes the influence of thermal spray parameters for atmospheric plasma spraying ( A P S ) and hypersonic flame spraying (HVOF) on the physical properties. Some examples of industrial applications will be introduced. These are thermally sprayed dielectrics for use as dielectric barrier discharge material in tubular ozonizers and multilayer composites with different electric properties for applications in the printing industry. THERMAL SPRAY TECHNIQUE Functional coatings can be applied by various techniques. For the deposition of thick coatings consisting of high melting metals and ceramics thermal spraying techniques are a cost effective and promising technology. One of the advantages of this technology, compared to other coating processes is the possibility to use a wide range of materials ranging from low melting polymers, metals and metal alloys to high melting oxide ceramics ” ’. A key feature of the thermal spray technique is the substrate’s low thermal load during the coating process compared to many other methods. Using simultaneous air or liquid CO2 cooling techniques, the substrate temperature can be kept relatively low e. g. between 80 - 150°C. There exist various thermal spraying techniques e.g. arc-, plasma-, or flame-spraying. Within all these variations two thermal spray processes are commonly used for manufacturing functional or protective coatings, the Atmospheric Plasma Spraying ( A P S ) and the High yelocity Oxygen Fuel process (HVOF). During A P S and HVOF the spray powder, suspended in a carrier gas, is injected into the heat source of the torch. After being totally or partially molten and accelerated, the powder particles impact the substrate, are quenched and build up the coating. The substrate surface is not fused during the process. In A P S a plasma serves as the energy source, an electric arc discharge between a water cooled copper anode and a tungsten cathode dissociates and ionises the working gas and builds up a plasma, that expands into the atmosphere forming a plasma gas jet (see figure I a) ’. The HVOF process uses liquid fuels or fuel gases for high energetic combustion with oxygen in a combustion chamber. A subsequent relaxation and acceleration of the combustion gas in an expansion nozzle leads to hypersonic high gas and particle velocities at the nozzle exit (see figure I b and 324
Dielectric Materials and Devices
c). The powder is inserted either into the combustion chamber or into the nozzle entrance
’.
Figure I. (a) Schematic sketch of an Atmospheric Plasma Spray torch with axial powder injection. Schematic sketches of HVOF torches used for thermal spraying: (b) Second generation type operated by diffusionflame, short combustion chamber (c) Third generation operated by spray flame, long combustion chamber, 6 - 10 bar, lava1 shaped nozzle. For details refer to text.
A P S is characterised by extremely high plasma temperatures and medium range particle velocities. The HVOF performs highest particle velocities but reduced process temperatures and is commonly used for materials with melting points below 2000°C. Therefore the A P S is preferentially used for the deposition of refractory oxides, the HVOF system for metal and cermetic materials. Typical particle temperatures and velocities are summarized in table I. Precedent to the coating process a washing, degreasing and grit blasting of the substrate surface is performed. The roughening of the surface with corundum of defined size improves the micromechanical adhesion of the coating and induces compressive stresses into the substrate material. Following to the coating deposition a mechanical (grinding, polishing) or thermal post-treatment of the coating surface takes place. The quality of the thermally sprayed coating with respect to the residual stress situation, deformation, microstructure, porosity and coating properties can be varied to a certain extend by tuning the spray parameters, like specific energy supply, spraying distance, torch geometry, simultaneous cooling, or additional selective heating, spraying powder selection etc.
Dielectric Materials and Devices
325
Table I. Comparison of particle temperatures and velocities for various thermal spray processes Spray method and torch type
Arc spraying Flame spraying Atmospheric Plasma Spraying ( A P S ) Vacuum Plasma Spraying (VPS) HVOF System TopGunTM HVOF System JetKoteTM HVOF System DiamondJetTM JP5000 TopGun-K
Fuel / plasma gas
Particle temp. ("C)
2700 2500 3500 3700 2600 22002500 propane 22002500 Liq. fuel (Isoparafine) 1800 Liq. fuel (Isoparafine) 1700
propane A r / H2plasma A r / H2plasma acetylene propane
Particle velocities ( d S )
50 100 210 250 400-450 400-450 400-450 600-650 750-900
POWDER TECHNOLOGY FOR THERMAL SPRAYING Depending on what material is processed thermal spray powders have distinct morphologies that strongly affect their fluidability (supportability) through the powder feeder, their melting behaviour and thereby the coating efficiency. The microstructure of the deposited film, i. e. its porosity and phase composition, is strongly influenced by the powder morphology and thermal history. Figure I1 summarises the common morphology types for ceramic powders.
Figure 11. Types of powder morphologies used for thermal spraying of oxide ceramics: According to
Fused and crushed qualities are the standard powders and they are quite cost effective. These powders are electrothennally fused and milled to an appropriate grain size. More expensive are sintered qualities based on submicron primary grains because they include an additional sintering step after the agglomeration process. Granulate size distributions typically range 326
Dielectric Materials and Devices
from 5 - 20pm, 25 - 40pm and 45 - 90pm respectively. Narrower size ranges can be achieved by additional sieving or separating techniques.
(a) (b) Figure 111. (a) Fused and crushed alumina powder; (b) Chemically synthesized alumina powder Mixed oxides are mechanically mixed powders (consisting of two separate phases) or alloyed ones (mixed crystals). High end quality oxide ceramic powders are obtained by redox reactions and sol gel techniques based on organometallic compounds or hydrothermal synthesis. They exhibit a very narrow size distribution and can be scaled down to a few 100 nm particle size. Electron micrographs of alumina particles, fused and milled and chemically synthesized, are shown for comparison in figure 111.
Figure IV. Porosity and structural fineness of particles as a function of powder production method '
Figure IV illustrates the correlation of the production route with porosity and grain size respectively. In general it can be stated, that very dense and fine powders are most expensive to manufacture and most sensitive in handling and control of the feed stream. Very crucial in thermal spraying is the grain size distribution. Often powder lots have a bimodal distribution with a fine fraction band. An example Dielectric Materials and Devices
327
is given in figure V. During HVOF spraying, a fine fraction of the oxide powder may partially evaporate and condense at the nozzle exit leading to a built up of disturbing residuals. This will influence the continuous operation of the HVOF torch, lower the process stability and consequently the coating quality. A coarse fraction will not contribute in a promising way to the coating build up because the particles cannot be fully molten and densified in a homogenous coating. With increasing coating thickness and total hardness the particles will escape as overspray and lead to a decrease of the deposition efficiency of the coating process and to an undesired increase of costs.
Figure V. Grain size distribution spectra of oxide ceramic powders.(a) Spinell powder with bimodal distribution; (b) A1203 powder with narrow monomodal distribution. THERMAL SPRAYING OF OXIDE CERAMICS For many applications A P S sprayed coatings will meet all requirements especially when the coating thickness is not restricted to a certain limit. Compared to other coating methods (see figure VI) thermal spraying and spray laquer techniques are most cost effective deposition methods for thick films in the range of 50- 1000 pm.
Figure VI. Appropriate coating processes for depositing insulating coatings
328
Dielectric Materials and Devices
However, most applications have additional requirements specifications like low porosity, extended surface quality with extremely low surface roughness or a limited coating thickness. In this case HVOF spraying or a combination of the A P S and HVOF method can solve the problem (see table I). It should be stressed that both techniques lead to different phase compositions in the coating that can differ from the initial phase composition of the powder material. Dependent from the temperature / time history of the particle and the heat conductivity of the substrate different phases corresponding to the thermodynamic equilibria in multiphase diagrams are formed. Due to the rapid cooling process metastable high temperature phases can be observed. Within the coating there may be also a spatial distribution of different phases. Many oxides like titanium oxide exhibit a distinct oxygen loss behaviour during the quenching process in thermal spraying, this phenomenon leads to an increase in electron conductivity due to the oxygen defects in the crystal lattice. Table I. Comparison of thermal spray methods concerning dielectric manufacturing HVOF Higher particle velocity Lower gas temperature Higher heat impact Oxidising flame possible Running costs
APS Lower particle velocity Higher gas temperature and higher gas enthalpy Reduced heat impact on workpiece Running costs
Benefits and disadvantages dense coatings, higher breakthrough voltage but reduced elasticity of the coating less efficient in oxide powder processing extensive cooling and reduced coating thickness reduced oxygen loss for titania and zirconia more expensive Benefits and disadvantages higher porosity, lower breakthrough voltage but higher elasticity of the coating Very efficient in oxide powder processing Moderate cooling and coating thickness up to several mm Less expensive
THERMOPHYSICAL PROPERTIES AND RESIDUAL STRESS ADJUSTMENT A remarkable aspect of thermal spraying is the possibility to adjust the internal stress distribution of the layer composite and the ability to partial compensation of the thermophysical mismatches that can arise when substrate and coating materials are different. Table I1 shows values for typical substrates and thermally sprayed coatings respectively. Most crucial is the ratio of heat expansion values of the substrate and the coating material respectively. For substrate materials it can vary in a wide range from 3 - 24x10-' K'. For thermally sprayed oxide ceramics, it ranges from 3 - 1I x ~ O K' - ~ (refer to table 11). Dielectric Materials and Devices
329
Table 11. Comparison of thermophysical properties of substrate materials and thermally sprayed oxides Material
2+
$,b2 2
A1 alloy (AIMg,) Mg alloy (AM50) Cr-steel (X5CrNil1810) Borosilicate glass
6.+g
A1203 (APS) TiO,(APS) ZrO, N20392/8 (APS)
'' &gij
CTE a (W6 1K)
Thermal conductivity h (W/mK)
Youngs modulus E (GW
23,8 26 16 3,3
134 65 15 1,12
70 45 200 63
3-5 7-9 (20"-400") 9-1 1
3-5 (2O0-4OO0)' 5 (20"-400") 225
40-50' 97'
Depending on the material combination of the substrate and the coating very opposed situations can emerge where either the CTE a of the substrate or the coating exhibits the lower value. This will contribute to an overall residual stress of the composite in form of tensile coating stresses for acoating > asubstrate and in COmpreSSiVe stresses for acoating < asubstrate. Substrate pre-processing, cooling down after coating and mechanical surface post-treatment cause additional stresses that superimpose and form the residual stress of the whole composite after cooling. For a more detailed discussion see
'.
(4 (b) Figure VII. (a) Circular microhole drilling and milling process in thermally sprayed coatings. (b) Flow chart of the measurement procedure.
The appropriate method to determine residual stresses in thermal spray coatings is an advanced incremental microdrilling and milling method (see figure VII). The advantage of the method is the possibility to aquire a depth resolved set of data that starts at the coating surface and reaches through the interface into the substrate. 330
Dielectric Materials and Devices
In several drilling and milling processes a circular, cylindrical shaped microhole is brought step by step (- 5 - 10 pm) into the component surface. The released stress leads to a relaxation of the area of the bore hole and is detected by the strain gauge When spraying ceramics for industrial applications, thermal and mechanical loads of the coated component under operation have to be taken into account. If the component is operated at elevated temperatures and the substrate material has a high a (e. g. aluminum) the adjustment of compressive stresses may be advantageous. With the help of experimental stress measurements, an optimization process can be developed to spray coat a sample under appropriate conditions. Online infrared thermography, simultaneous cooling and intelligent torch handling are tools to optimize the coating process iteratevly '. Table I11 summarizes measured thermomechanical properties of plasma sprayed ceramics.
'.
Table 111. Measured thermomechanical properties of thermally sprayed oxide ceramic coatings NO
A
Coating - material (supplier / type of torch) AI203 (Hochrhein / F4)
universal hardness [N/niin']
E/(I-P)
PPal
1
Rz
[PI
CTE a 10" [ I /K]
phase
composition
crystallographic shucture
6543
152f3
3,9 / 26,4
6,63
y-AI203 a-A1203
cubic rhomboedric
B
A1203/ TiOz 9713 (Hochrhein / F4)
7424
179f3
3,6 / 26,l
7,16
y-A I ?03 TiOz anatase
cubic tetragolial
C
AI2O3/TiO2 92,5/7,5 (Hochrhein / F4)
8620
221 + 4
5,5 1 36,O
7,35
y-AlaOs AITi3 Ti02 anatase
D
AI2O3/TiO287113 (Hochrhein / F4)
7544
184f3
5,B I 36,6
7,32
y-Al203 Ti02 anatase TiO
E
AI2O3/TiO267/33 (H.C. Starck / PTG)
7571
187 f 3
4,O / 27,3
7,68
A1203 TiO? rutile
cubic hexagonal tetragonal (bee) cubic tetragonal (bcc) cub1c (fcc) oi-thorliombic tetragonal
F
TiO? (H.C. Starck 1 F4)
6043
165 f 3
2,7 / 19,O
8,73
G
ZrO2 /CaO 70130 (Hochrhein / F4)
3050
97+2
2,4 I 17,8
10,30
H
zl'O2/Y203 9317 (H.C.Starck / PTG)
63 12
158f3
3,4 I 24,4
10,80
I
ZrO2 /MgO 76/24 (Hochrhein / F4)
6180
169 f 3
3,8 I 26,3
10,50
zro2 Zr
orthorhombic hexagonal
J
Hf02/Y203 6092 155f3 95,5/4,5 (Ceram 1 F4) I supplier info ' a i n temperature range 200-400°C
4,2 / 29,l
7,87
y2~f701, Hf02
monoclinic
Dielectric Materials and Devices
331
Figure VIII. Influence of the chemical composition on electrophysical properties of A1203/Ti02 mixed oxides ELECTROPHYSICAL PROPERTIES OF OXIDE COATINGS The electrophysical properties of a dielectric system can be determined by complex impedance spectroscopy. Most important electrophysical properties are permittivity, loss factor and volume resistance. It should be noted that for most materials the values are frequency dependant. Thermally sprayed oxides data vary from bulk materials ones found in the literature in most cases. That has two reasons: (1) Thermal spray coatings exhibit a certain porosity, (2) The phase composition in the coating differs from the spray powder material and is not homogeneous in many cases. For insulator materials the breakthrough voltage is another crucial factor and has to be measured via high voltage breakthrough tests. It is especially this value that is strongly influenced by the microstructure and the surface morphology of the coating, i. e. its porosity and the presence of cracks. Table IV summarizes some of the electrophysical data measured on thermally sprayed samples. Mixing and spraying oxides with varying electrophysical properties allows to adjust the values of permittivity and volume resistance of the coatings in a certain range. An example is given in figure VIII. Alumina / titania oxides have been plasma sprayed in varying compositions starting from pure alumina up to pure titania. With the increase of titania content the volume resistivity can be lowered from 10" Rcm to 107 ncm. Approximately at a alumindtitania weight percent or mass ratio of 60/40 there is a gap and the volume resistivity drops to a value of 105Rcm. The permittivity behaves in a respective manner. Pure titania finally shows very low volume resistivity values in the region of 1O2 Rcm. This behavior offers interesting possibilities of tailoring dielectric properties of thermally sprayed oxides for various new industrial applications. 332
Dielectric Materials and Devices
Tabelle IV. Measured dielectric properties of thermally sprayed oxide ceramic coatings.Measurements carried out at room temperature. ~
No. A
B C
D
E F
G
H
I J
Coating material (supplier / type of torch)
A1203 (Hochrhein / F4) A1203 / T i 0 2 97/3 (Hochrhein / F4) A1203/Ti02 92,5/7,5 (Hochrhein / F4) A1203/Ti02 87/13 (Hochrhein / F4) A1203/Ti02 67/33 (H.C. Starck / PTG) Ti02 (H.C. Starck / F4) Zr02/CaO 70/30 (Hochrhein / F4) Zr02/Y203 9317 (H.C. Starck / PTG) Zr02/MgO 76/24 (Hochrhein / F4) Hf02N203 95,5/4,5 (Cerem / F4)
'
_
.
E,.
(at 50 Hz)
tan6 ( 10-4)
Volume resistance
13,Of 1,2
6
22 -t 6
(Szcm) 1,2*109
9,5 - 11,l
31
1
2,6* 107
<1
47
3
2,3*107
<1
867
1
2,5 *106
4
>1000
0
i,8 *104
<1
>1000
0
4,9 * 1o2
9,9 f l,o
21
170
2,6 *10"
4s
24
453
6,3 *10"
337
16
71
1,4 *10"
9,O - 10,3
23
62
8,8 *lO'o
Ed
(kV/mm)
INDUSTRIAL APPLICATIONS In the following sections some recent studies of the IFKB will be presented where APS and HVOF sprayed dielectrics could be inserted successfully into industrial applications. These studies have been carried out using a TopGun-GTMand a TopGun-KTMinstallation for HVOF spraying and a 85 kW M e t c 0 - F 4 ~plasma ~ torch for A P S spraying. All spray processes are controlled by a fully programmable GTV-MF-P-1200 controller unit. The torches can be mounted either on a Staubli industry robot system or on a two axis linear handling system that is controlled by a Sinumeric control unit. APPLICATION 1: OZONIZER TUBES Thermally sprayed oxides are promising candidates for several barrier discharge applications. Tubular ozonizers are one example where the application of plasma spraycoated borosilicate glass tubes revealed increased ozone production efficiency '. Ozone is formed with the help of a dielectrical barrier discharge ("silent" discharge) driven by a controlled high voltage alternating field (figure IX a) '. The discharge burns in a small gap formed by an inner and an outer electrode (two concentric tubes). One electrode is covered by a dielectric (figure IX b). The processed gas (pure oxygen or air) flows through the gap and the discharges lead to dissoziation and forming of activated oxygen species that react and enrich the gas with ozone as a product. Dielectric Materials and Devices
333
(4 (b) (4 Figure IX. (a) Schematic diagram of a capacitor arrangement for ozone production. (b) Schematic crossections of conventional and improved tube designs. (c) Commercial ozonizer with tube array The dielectric has an important influence on the ozone forming efficiency of the ozonizer. In conventional tube designs the borosilicate glass tube itself serves as the dielectric. In an improved design, developed by the IFKB and partners Io, electrode and dielectric exist as thermally sprayed composite layers on the glass tube. A key feature of the thermally sprayed coatings is their composite structure as the coating consists of a dielectric and a metallic electrode. The thickness of the dielectric can reach 1000 pm, the metal interlayer measures about 50pm. Figure X shows micro cross sections of plasma sprayed composite layers with an A1203 and a ZrO2 dielectric respectively.
(4 (b) Figure X. Thermally sprayed composite layers for ozonizer tubes on a borosilicate glass substrate. (a) A1~0~-A1/Si-composite.(b) ZrOz-Al/Sicomposite. One important influencing factor is the dielectric's permittivity but experiments have shown that more determining factors like surface structure, work function and surface conductivity of the dielectric are also involved. In Contrast to the common procedure the glass substrate needs no gritblasting before coating. The glass surface is activated by the plasma flame itself and subsequently coated with the Al/Si alloy that exhibits extraordinary high adhesion to the glass surface. 334
Dielectric Materials and Devices
Most important for a high electrical breakdown strength is a proper adjustment of the roughness at the AlWceramic interface. However, this has to be done thoroughly because a too low surface roughness leads to minor adhesion and delamination of the ceramic coating.
(a)
(b)
Figure XI. (a) Coated glass tube destroyed by the intrinsic stresses of the coating. (b) Coated glass tube after process optimisation Most crucial when applying thick ceramic coatings on a glass tube is the optimisation of the spray process regarding stress adjustment. The temperature field of the tube must not exceed 150°C. Online cooling, torch handling and powder feedstream quantity have to be optimised empirically to avoid destruction of the glass tube during or after coating.
Figure XII. Glass tube prototype with applied metal dielectric coating Figures XI11 and IX show results of ozone production measurements in a three tube test ozonizer. In figure XI11 the ozone yield vs. energy density is plotted. The curves indicate a better ozone yield at lower energy densities for the Zr02 and A1203 coated tubes respectively. However, the energy consumption is limited for the novel tube types due to electrical breakdown in the coating materials or reaching the power limit for the electrical supply of the ozone generator. This fact limits the absolute ozone concentrations of coated tubes and is outlined in figure XIV. To increase this value the electrical power supply of the ozonizer has to be adapted to the electrical properties (capacity, loss factor etc.) of the novel tube design.
Dielectric Materials and Devices
335
Energy density [kWhNm3]
Figure XIII. Comparison of measured ozone yields vs. energy density in a three tube test ozonizer for standard duran tubes and coated tubes.
Dielectric material
Figure XIV. Comparison of measured ozone production efficiency in a three tube test ozonizer APPLICATION 2: DIELECTRIC COMPOSITE COATINGS Figure XV shows an example for an APS / HVOF sprayed multilayer designed for an application in the professional printing industry. The APS F4 sprayed bottom layer consists of alumina and serves as a dielectric bulk material. The HVOF (TopGun-G) sprayed top layer is an alumina - titania mixed oxide with two functions: (1) Due to its low porosity, the surface can be grinded and finished to extremely low roughness values and thus a higher surface quality can be reached than with an APS sprayed coating. (2) The 336
Dielectric Materials and Devices
mixed oxide exhibits different electric properties that are especially needed in this application.
Figure XV. (a) Thermally sprayed composite layer consisting of an APS sprayed A1203 (330 pm) and a dense HVOF sprayed A1203/Ti02 top layer (70 ym); (b) Detail of the HVOF sprayed A1203 / TiO2 As the HVOF coating can be sprayed with high accuracy and a low as sprayed surface roughness ( A P S : r, = 30 - 40 pm, ra = 3 - 4 pm; HVOF: r, = 2- 3 ym, r, = 20 - 30 pm) the mechanical post-processing of the coating, i. e. cost intensive precision grinding and finishing can be minimised. Combining the two spray methods can save cost especially where a high surface quality is required.
Table V. Surface roughness, porosity and hardness of thermally sprayed coatings spray R, Rz Porosity HU HV0.05 method
[Pml
[Pml
[%I
[N/nm2]
[l]
A1203
APS as sprayed
3.74
27.2
631
4267,4
706,4
Top coat A1203/Ti02 67/33
HVOF as sprayed
5,5
29
1,9
45473
765,6
0,07
1,02
179
45473
765,6
Top coat HVOF A1203/Ti02 finished
CONCLUSIONS Thermal spraying is a powerful and cost effective tool in manufacturing fimctional coatings are required in industrial applications. General distinguishing marks of thermal spraying are (1) low substrate temperatures and thermal load during the coating process, (2) principally applicable under atmospheric conditions, (3) film thickness in the range of 30 pm to several 1000 ym scale (4) high coating efficiency and fast production cycle times. A P S and HVOF sprayed coatings exhibit distinguished properties like designed porosity, phase composition and intrinsic stress distribution that lead Dielectric Materials and Devices
337
to different interesting thermophysical and electrophysical properties of the thermal spray coating system. Therefore the coating method has to be chosen carefully to accomplish the special needs of the application. Oxide ceramic coatings and multilayer structures for applications as dielectric in barrier discharge and in electrically controlled color / ink transfer rollers in professional printers have been fabricated and proved the thermal spray technique to be suitable and performing. Process control becomes more and more important. Residual stress adjustment of sprayed coatings is possible by optimised process sequences including torch and workpiece movement, simultaneous cooling techniques and heat flow control during the coating process. REFERENCES ‘R. B. Heimann; “Plasma Spray Coating: Principles and Applications”, VCH, 1996, ISBN 3-527-29430-9 2L. Pawlowski; “The Science and Engineering of Thermal Spray Coatings”, Wiley, 1995, ISBN 0-471-95253-2 3P. Heinrich, Ch. Penszior, H. MeinaD; “Gases for High Velocity Oxy-Fuel Flame Spraying”, in 4. HVOF Spraying Colloquium, Erding, Germany, 1997 4Berreth K., Buchmann M., Gadow R., Tabellion J.; “Evaluation of Residual Stresses in Thermal Sprayed Coatings”; Conference Proceedings of the UTSC’99, Dusseldorf , ISBN 3-87155-653-X, pp. 670 - 675 ’Buchmann, M.; Gadow, R. High speed circular micro milling method for the determination of residual stresses in coatings and composites 24th Annual Cocoa Beach Conference&Exposition, 23. - 28.01.2000, Cocoa Beach, USA
C. Friedrich, A. Killinger; “Thermographic imaging - a new tool for on line processing control in coating manufacturing by thermal spraying” CIEC 7, 13-1 5 September 2000, Genoa, to be published 7Gadow R., Riege G., Deutsche Patentschrift, Nr. 195 11 001.3, 1995 ‘Lemmerich J., “Die Entdeckung des Ozons und Die ersten 100 Jahre Ozonforschung“; Sigma, Berlin, 1990 9 Kogelschatz U. , Eliasson B ., “Die Renaissance der stillen elektrischen Entladung“; Physikalische Blattev, Weinheim, 1996 6
“Gadow R.; Killinger A.; Friedrich C.; “Thermally Sprayed Multilayer Coatings as Electrodes and Dielectrics in High Efficiency Ozonizer Tubes“, UTSC‘ 99, United Thermal Spray Conference and Exposition, Diisseldorf, Germany, 17. - 19. March 1999, ISBN 3-87155-656-X, pp. 676 - 682
338
Dielectric Materials and Devices
Effect of dc field on dielectric loss of SrTi03 single crystals and thin films CHEN Ang, ZHI Yu, Ruyan Guo, and A.S. Bhalla Materials Research Laboratory, The Pennsylvania Stute University, University Park, PA 16802
ABSTRACT The dielectric behavior, especially the dielectric loss of SrTi03 single crystals and thin films under high dc bias is reported in this paper. A rounded dielectric constant peak, so-called “induced ferroelectric mode” is induced by dc bias in both SrTiO3 single crystals and thin films. The dielectric loss shows more complicated behavior. The results show that dielectric loss under dc bias consists of several components coming from “defect modes” and “induced ferroelectric mode”. For the so-called “defect modes”, the T,,, is independent of dc bias, however, whose intensity is suppressed with increasing high dc fields. At high dc bias, only the ”induced ferroelectric mode” remains. The physical nature of the “defect mode” is briefly discussed.
I. INTRODUCTION The dielectric behavior, especially the dielectric loss of the ferroelectric and paraelectric materials under dc bias is less reported in the literature. Recently, in order to meet the need of continuously increasing development in data processing and microwave communication, a series of microwave devices, such as frequency agile filters and tunable high Q resonators, are being developed. This demands a dielectric material with both high electric-field tunability (K = [E(O) - E (E)] / ~ ( 0 ) ) and very low dielectric loss at microwave frequency range.[ 11 This demands the revisit on the dielectric behavior of the materials under dc bias. To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
339
The paraelectric SrTi03 has been recognized as a promising candidate material, which shows high electric-field tunability at low temperatures along with reasonably low dielectric loss in the form of single crystal.[2-51 However, it is found that the dielectric quality factor Q is deteriorated for more than an order of magnitude in SrTi03 thin films [6,7]. Several fxtors, including nonstoichiometry, defects, high strain, and geometry effect in the thin film are recognized as the probable causes. However, in fact, the understanding of the dielectric loss under dc bias for both of the SrTiO3 single crystals and thin films is not clear so far. In this paper, we report the effect of dc electric-fields on dielectric properties, especially the dielectric loss of SrTiO3 single crystals and thin films in an effort to obtain more experimental data to understand the physical mechanism of the dielectric loss in SrTi03 under dc bias.
11. EXPERIMENTAL PROCEDURE The SrTi03 single crystal samples with polished surfaces (100) were obtained from commercial source. The SrTiO3 thin film samples with thickness of 1 pm were prepared by the pulsed laser deposition technique on the SrTi03 single crystal substrates. The preparation details were reported in Ref. 8. Complex dielectric permittivity was measured using HP 4284A LCR Meters. The temperature dependence of dielectric properties was measured in a cryostat system, while the specimen was being cooled or heated up at a cooling/heating rate of 1 K per minute and readings were taken at every 1 K or 2 K intervals. The dc voltage is applied to the samples and a blocking circuit was adopted to separate the high dc voltage and LCR meters.
111. RESULTS A. SrTi03 single crystals Temperature dependence of the dielectric constant (E) and loss (tan&)with and without dc bias (0 - 50 kV/cm), is shown in Fig. 1. Without dc bias, from 300 K to 12 K, E increases continuously with decreasing temperature and attained the value E = 10300 at 12 K. However, for tan&,in the temperature range 50 - 100 K, a notable peak with frequency dispersion is observed around 70 K (1 kHz), the peak is denoted as mode I (see below). With further decreasing temperature, as T < 50 K, the tan6 increases sharply and E increases correspondingly.
340
Dielectric Materials and Devices
w
Fig. 1 (a) Temperature dependence of the E and tan6 for the SrTiOS single crystal for the different electric-fields from 0 to 20 kV/cm at 10 kHz. Solid curves: peak A, dash curves: mode
m.
Dielectric Materials and Devices
34 1
1500
0.002 0.001 0
1000 1200 ' d
900 1000
r
i
40 kVlcmI
750 1000
0
126
80
40
0.002 0.001 0 0.002 0.001 0
T(K) Fig. 1 (b) Temperature dependence of the E and tan6 for the SrTiO3 single crystal for the different electric-fields from 25 to 50 kV/cm at 3 kHz.
80
s. 60 h
r
40
fl Made 111
2o0 0
10
20
30
40
50
E (kV/cm)
Fig. 2 Phase diagram of the dielectric relaxation modes for different electric-fields 0 - 50 kV/cm.
342
Dielectric Materials and Devices
The relaxation rates of the mode I were derived from the temperature dependence of the imaginary part of the permittivity. The data can be well fitted to the Arrhenius law, v = vo exp [ U/KBT] , (1) where vo is the relaxation rate at infinite temperature, U the activation energy for relaxation, and T the temperature. The fit parameters are U = 0.1 1 eV, and vo = 10” Hz. The activation energy U = 0. I I eV is in good agreement with the parameter reported by Mizaras and Loidl. [9]
Fig. 1 (a) shows that the temperature dependence of the E and the tan6in the field range of 2 - 20 kV/cm. It can be seen that in E, by applying the dc bias, a peak is induced. As suggested in the earlier literature [ 10-131, the peak induced by electric-field in the dielectric constant indicates an occurrence of a “induced ferroelectric peak”. Generally, for a “ferroelectric peak”, the corresponding loss peak should occur at the same temperature. Therefore, it is reasonable to assign a loss peak which occurs at the temperature of the “induced ferroelectric peak” in dielectric constant, as peak A (shown as solid curves in Fig. 1). However, tan6 vs. T under the dc bias shows more complicate behavior, in addition the loss peak corresponding to the peak A, there are three sets of peaks, the mode I and other two peaks, under dc bias. They are ( 1 ) near peak A, a peak is noticeable, denoted as mode I11 (shown as dash-line); (2) a mode I1 around 50 K. One question is raised, why under dc bias, the two modes I11 and I1 appear? This might be explained as that, these two modes originally exist. However, because of the quickly increase in the tan6 as decreasing temperatures in the case of the zero dc bias, the modes 111 and I1 are overlapped; by an application of a dc bias, the tan6 is overall suppressed, and the modes I11 and I1 show up. With further increasing field, from 25 to 50 kV/cm, as shown in Fig. 1 (b), it is observed that mode I, I1 and I11 almost disappear, and only peak A remains, whose temperature exactly corresponds to that of the E peak.
Fig. 2 shows a summary of the dielectric relaxation modesand peak A in the single crystal under the different electric-fields 0 - 50 kV/cm i n the form of a phase diagram. The T,,, for three modes, I, 11, and I11 do not change with electricfield. However, modes I, 11, and I11 vanish at high electric-fields. The T,,, of peak A shifts to higher temperatures with increasing electric-fields, which is the only one that stays with increase in electric-field E 2 25 kV/cm.
Dielectric Materials and Devices
343
B. SrTi03 Thin films The temperature dependence of the E and the tan6of the SrTiO3 thin film sample under dc bias at 10 kHz, is shown in Fig. 3. For the E, by applying dc electric-fields, a broad dielectric constant peak is induced; obviously this is the same peak as the "induced ferroelectric peak", i.e.. peak A, observed in the SrTiO? single crystal, so here also denoted as peak A. For the tan6, different from those observed in the SrTi03 single crystal, the thin films show a notable dielectric loss peak with frequency dispers.ion around 150 K (at 1 kHz), which is denoted as mode IV.In addition, a peak around 70 K is also observed, especially under dc bias. This is the same temperature range in which the mode I locates in the single crystal. 2500
0 kVlcm
100
150
200
240
350 kVlcm
Fig. 3 Temperature dependence of E and tan6 for the SrTiO3 thin film sample under different dc electric-fields from 0 to 350 kV/cm at 10 W z .
For the mode IV, the relaxation rates also follow the Arrhenius law ( l ) , the fit parameters are U = 0.25 eV and vg = 9x 10" Hz. It should be also stressed that although the intensity of the modes IV and I in the thin film decreases with the increasing dc bias. However, the temperature (Tm)of the modes IV and I remains almost the same, the relaxation rates follow the Arrhenius law with the almost same activation energy U and the attempt frequency vg under different dc biases.
-
344
Dielectric Materials and Devices
IV. DISCUSSION As shown above, in both SrTiO3 single crystals and thin films, there are mainly two different kinds of dielectric modes, which behave in different ways under dc bias. One is the peak A that induced by dc bias, whose T,,, increases with increasing dc bias, this peak is recognized as “induced ferroelectric peak” by some authors.[ 10-131 Another kind of the peaks are those we denoted as modes I, 11, I11 and IV in the present paper, whose T,,, does not change with variation of dc biases, however, the intensity decreases as increasing dc bias. The peak A has been extensively studied in the literature,[ 10- 131 so in the present work, we only briefly discuss the possible physical nature of the modes I, 11, 11, and IV. Viana et a1.[16] attributed mode I11 (-28 K) to a possible coherent state.[l7] However, it is recognized that the possibility of the existence of the quantum coherent state around 37 K is still an open question. Mueller et a1.[17] suggested that it is a static response rather than a dynamic response, but mode I11 obviously exhibits a dynamic behavior, mode 111 thus can not be explained as a quantum coherent state. The relaxation modes I (- 70 K) observed in both the single crystal and the thin film. have also been observed in single crystals by Mizaras and Loidl [9], Viana et al. [16]; in polycrystalline samples by Chen et al.[ 14,151, and in thin films by Li et al. [8]. Mizaras and Loidl attributed mode I to dynamic response of the domain walls that occur at the cubic-to-tetragonal phase transition.[9,18] Mode IV only appears in the thin film samples, which was previously reported by other authors,[8] and was explained to be caused by stresses in the thin films, for example, caused by the mismatch of the lattice between the thin films and the substrates.[8] Mode I1 has not been discussed by others. The important and common characteristic of modes I, 11, 111, and IV is their field-independent T,,,. Similar dielectric modes were also reported in pure and Bi doped SrTi03 ceramics,[ 14,151 in which, some “impurity modes” and “defects modes” were observed, whose T,,, is independent of the concentration of the doping impurity, and of the applied dc bias. [ 14,151 This indicates that besides the contribution of the defects/impurities, an intrinsic mechanism should be involved. Generally speaking, even in nominally pure quantum paraelectric SrTiO3, low levels of unavoidable defects and/or impurities (for example, oxygen vacancies) may be present. These defects and/or impurities may behave like the dipoles, whose reorientation under weak ac field will contribute to the dielectric response. Indeed, in the high quality single crystal quantum paraelectric KTa03, the dielectric relaxation at low temperature (-40 K) was found to be related to the defects/impurities [ 19,201.
Dielectric Materials and Devices
345
It is well known that SrTi03 is a typical soft mode paraelectric, specially, at cryogenic temperatures, SrTi03 displays varies of phonon modes. In Bi doped SrTi03, we found that the activation energies of the “impurities modes” derived from the dielectric relaxation are in good agreement with the phonons energies observed from the Raman spectra.[21] This strongly indicates that there exists the coupling between the impurities/defects dipoles and the phonon modes. That is the effect of the impurity and/or defects in the SrTiO3 is amplified via the interaction between the impuritieddefects dipoles and the phonon modes. Therefore. in the present work, it is highly possible that such defects and/or impurities could interact with the soft modes at low temperatures, and contribute to dielectric relaxation behavior. We suggest that the modes I, 11, 111 and IV observed could be attributed to the re-orientation of the dipoles (formed by the impurities and defects) that interact with the soft modes. However, at this moment, a clear picture of the polarization of the “defect modes” is lacking, further work is needed.
V. CONCLUSION The dielectric behavior of SrTiO3 single crystals and thin films under high dc bias is reported. For both SrTiO3 single crystals and thin films, a rounded dielectric constant peak is induced by dc bias, which was attributed to the “induced ferroelectric peak”. The dielectric loss shows more complicated behavior, the results show that dielectric loss under dc bias consists of several components coming from “defect modes” and “induced ferroelectric mode”. For the SrTiO3 single crystals, three “defect modes” occur below 100 K. For the SrTi03 thin films, a notable dielectric loss peak, “defect mode” around 150 K (at 1 kHz), and a small mode around -70 K are observed. All of these “defect modes” whose T,,, is independent of dc bias, however. whose intensity is suppressed under high dc bias, and at high dc biases, only the “induced ferroelectric mode” remains. The physical nature of the “defect modes” is explained as the reorientation of the “defect dipoles” and/or “impurity dipoles” that formed by the unavoidable defectdimpurities in the samples. Theses “defect dipoles” and/or “impurity dipoles” can interact with the phonon modes of the host lattice of the SrTi03, and hence the dielectric responses are amplified by the phonon modes.
346
Dielectric Materials and Devices
Acknowledgments : This work was supported by the grant from DARPA under contract No DABT6398- 1-002.
REFERENCES
[ 101 [ 1 11 [ 121 [ 131
[14] [ 151 [ 161 [ 171 [ 181
O.G. Vendik, Ferroelectrics, 12, 85 ( 1 976); O.G. Vendik, I.G. Mironenko, and L.T. Ter-Martirosyan, Microw. RF, 33, 67 (1994); O.G. Vendik, E. Kollberg, S.S. Gevorgian, A.B. Kozyrev, and 0.1. Soldatenkov, Electronics Lett. 31, 654b( 1995) A.B. Kozyrev, T.B. Samoilova, A.A. Golovkov, et al., J. Appl. Phys. 84, 3326 (1998). F.W. Van Keuls, R.R. Romanofsky, D.Y. Bohman, et al., Appl. Phys. Lett., 71,3075( 1997). R. E. Treece, J. B. Thompson, C.H. Mueller, T. Rivkin, and M. W. Cromar, IEEE Trans. Appl. Supercond., 7,2363 (1997). CHEN Ang, A.S. Bhalla, Ruyan Guo, and L.E. Cross, Appl. Phys. Lett., 76, 1929(2000). CHEN Ang, Ruyan Guo, A.S. Bhalla, and L.E. Cross, J. Appl. Phys., 87, 3937(2000). D. Galt, J. Price, J.A. Beall, and R.H. Ono, Appl. Phys. Lett., 63, 3078 ( 1993). D. Galt, J. Price, J.A. Beall, and T.E. Harvey, IEEE Trans. Appl. Supercond., 5. 2575 (1995). Hong-cheng Li, Weidong Si, Alexander D. West, and X. X. Xi, Appl. Phys. Lett., 73,464 (1998) R. Mizaras and A. Loidl, Phys. Rev. B 56, 10726 ( 1997). C. Frenzel and E. Hegenbarth, Phys. Stat. Sol., 23,5 17( 1974). H. Unoki, and T. Sakudo, J. Phys. Sco. Japan, 23, 546 ( 1 967). P. A. Fleury, and J. M. Worlock, Phys. Rev. 174, 6 13 ( 1968). J. Hemberger, P. Lunkhemer, R. Viana, R. Bohmer, and A. Loidl, Phys. Rev. B52, 13159 (1995). CHEN Ang, J. F. Scott, ZHI Yu, H. Ledbetter, and J.L. Baptista, Phys. Rev. B59, 666 1 ( I 999). CHEN Ang, ZHI Yu, J. Hemberger, P. Lunkhemer, and A. Loidl, Phys. Rev. B59, 6665 (1999). R. Viana, P. Lunkenheimer, J. Hemberger, R. Bohmer, and A. Loidl, Phys. Rev,. B 50,60 1 ( 1994). K.A. Miiller, W. Berlinger and E. Tosatti, Z. Phys. B84,277( 1991). M. Liu, T. R. Finlayson, and T. F. Smith, Phys. Rev. B55, 3480 (1997).
Dielectric Materials and Devices
347
[19] B. Sake, J.L. Gravi, and L.A. Boatner, J. Phys. CM, 6,4077 (1994). [20] V. Trepakov, F. Smutny, V. Vikhnin, V. Bursian, L. Sochava, L. Jastrabik, and P. Syrnikov, J. Phys. CM, 7, 3765 (1995). [21] CHEN Ang and ZHI Yu, Phys. Rev. B61, I1363 (2000).
348
Dielectric Materials and Devices
DIELECTRIC PROPERTIES OF LAYERED PEROVSKITE S ~ , - ~ A ~ B i 2 N b ~ 0 9 FERROELCTRICS (A = La, Ca and x = 0,O. 1) M.J. Forbess, S. Seraji, Y . Wu, S.J. Limmer, C.P. Nguyen, and G.Z. Cao University of Washington Materials Science and Engineering 302 Roberts Hall, Box 352120 Seattle, WA 98 195 ABSTRACT In this paper, we report an experimental study on the influences of 10 at % Ca2' and La3+ doping on dielectric properties and DC conductivity of SrBi2Nb209ferroelectric ceramics. All the samples were made by two-step solid-state reaction sintering at temperatures up to 1150 "C for 0.5- 1 hour in air. X-ray diffkaction analysis indicated that single-phase layered perovskite ferroelectrics were obtained and no appreciable secondary phase was found. The Curie point was found to increase from 418 "C without doping to 475 "C with Ca2' doping and to 480 "C with La3+ doping. Dielectric constants, loss tangent, and DC conductivity of SrBi2Nb209ferroelectrics doped with Ca2+ and La3+ were studied and the relationships among doping, crystal structure, and dielectric properties were discussed. INTRODUCTION Ferroelectrics have reversible spontaneous polarization making them ideally suited for use in non-volatile random access memories (NvRAMs). Polarization is due to dipoles that can switch directions spontaneously under the influence of an electric field and the dipoles are a result of the noncentrosymmetric crystal structure [ 1-21. Previous research has focused on Pb(Ti,Zr)03 (PZT), but one of the current problems with PZT is the fatigue resistance of the material. PZT thin films tend to degrade most of the initial amount of switching charge ("fatigue") after 106 108cycles of full polarization switching [3]. Bismuth layered perovskite materials have high fatigue resistance and are able to withstand 10l2 erasehewrite operations [4] and therefore have attracted an increasing attention for NvRAM application [5]. Among the layered perovskites, SrBi2Nb209 (SBN), SrBi2Ta209(SBT) and their solid solutions (SBNT) are the most promising candidates, because they possess a reasonable spontaneous polarization which is one of the key parameters for the information storage applications. However, they suffer from relatively high processing temperature, relatively low spontaneous polarization, and relatively high dielectric loss [6-71. A lot of research has been reported in the open literature aimed at improving the dielectric and ferroelectric properties of the SBNT materials [e.g., 8-91. In particular, doping with various metal oxides has been demonstrated one of the effective approaches to improve the properties [ 10-111. For example, Bi3+doping has resulted in an appreciable enhancement of dielectric properties [ 101. Wu and Cao [ 12-131 have substituted Nb5' with vanadium and found a significant enhancement in dielectric and ferroelectric properties. In addition, it was found that the sintering temperature was significantly reduced by approximately 200 "C. In this letter, we report our recently study on the influences of Ca2+and La3+doping on the electrical properties of SBN ceramics. One of the objectives of this research is to enhance the
-
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
349
dielectric and ferroelectric properties of SBN by substituting Sr2+(1.44A) with smaller ions, Ca2' (1.34A) and La3' (1.36A) (all with coordination number = 12) [14]. It is anticipated that the incorporation of smaller Ca2' or La3+could lead to a reduced lattice constant of perovskite unit cells; however, in layered perovskites, the Bi202interlayer would impose a structural constraint so that the perovskite unit cell would not be free to shrink. Consequently, a combination of incorporation of small cations and unchanged unit cells would lead to both large spontaneous and ionic polarization. Another objective is to study the possibility of reducing the DC conductivity and loss tangent of SBN ferroelectrics, a similar approach as that widely used in isotropic perovskite ferroelectrics. The chemicals used were: CaC03, La2(C03)3.8H20, SrC03, Bi203,and N I 3 2 0 5 (all from Aldrich Chem. Co., with a purity of 99%). Three compositions were studied in the current research: SBN, Cao.ISr0.9Bi2Nb209 (hereinafter referred to as CSBN), and Lao.lSr0.9Bi2Nb209 (LSBN). Details of the sample preparation and characterization were the same as reported previously [9]. RESULTS AND DISCUSSION Table I summarizes the sintering conditions and the relative density (percentage of the theoretical density calculated from lattice constants) of the final products. Table I. Sintering conditions and relative density of SBN, CSBN, and LSBN. Samples First Sintering Final Sintering Relative Density Conditions Conditions 1150°C; 0.5hr 94% SBN 900°C; 2hrs 1150°C; lhr 95% CSBN 900°C; 2hrs LSBN 900°C: 2hrs 1150°C: lhr 96% This table shows that the incorporation of calcium oxide and lanthanide has improved the sinterability. This result is similar to that with the addition of vanadium oxide [12], albeit the improvement is less significant. The weight loss of all three samples is comparable and is less than the amount of extra bismuth oxide added. Figure 1 shows the X-ray diffiaction spectra of the three samples studied and indicates that a single phase layered perovskite in all samples was formed after pre-firing and no secondary phase detectable. The XRD results indicate that both Ca2' and La3+were incorporated into the layered perovskite structure and presumably occupied Sr2+sites. Although it is not experimentally shown whether Ca2' and La3+replaced Sr2'or Bi'+, comparison of the ionic radii reveals that Bi3+(0.96 A [14]) is more than 30% smaller than the other ions and the substitutions in the Bi3+sites are not very likely. Figure 2 shows the dielectric constants of samples SBN, LSBN, and CSBN as a function of temperature, determined at a frequency of 100 kHz and an alternating voltage of 50 mV. It is seen that with 10 at% lanthanum and calcium doping, the Curie points increased from 4 18 "C for SBN, to 475 "C and to 480 "C for CSBN and LSBN, respectively. However, the dielectric constants at both room temperature and the Curie point were found to decrease with Ca2' and La3+doping. Decrease in dielectric constants can be explained by a decrease in atomic polarization. Dielectric constant is determined by four polarizations: atomic, ionic, dipolar and space charge polarization. Under the experimental conditions used in the current study, i.e., a high frequency of 100 kHz and a low electric field (- 0.4 V/cm), only atomic and ionic polarization would contribute to the dielectric constants. The incorporation of smaller cations (7% smaller in ionic radius) would result in a reduced atomic polarization. Figure 3 is the DC conductivity of samples SBN, CSBN, and LSBN, as a function of temperature. This DC conductivity was estimated from the bulk resistance, R,, which was derived from the equivalent circuit model in complex impedance planes of a resistor and a capacitor in
350
Dielectric Materials and Devices
Figure 1.
XRD spectra of (A) SBN, (B) LSBN, and (C) CSBN.
Dielectric constant of SBN ( O ) , CSBN(A), and LSBN(0) as a function of Figure 2. temperature, measured at a frequency of 100 kHz.
Dielectric Materials and Devices
35 1
parallel [15]. It was found that throughout the temperature range studied in the current research, the DC conductivity of both LSBN and CSBN samples are appreciably lower than that of SBN. There are two regions predominant with different conduction mechanisms. At the low temperature region, the conduction is presumably predominant by the extrinsic impurity conduction with a very similar activation energy, whereas the conduction at the high temperature range is likely predominant by intrinsic defects. The activation energy, AEj, of LSBN and CSBN, at the high temperature range was found higher than that of SBN and AEj is assumed to be the energy necessary to create and move vacancies. The transition temperature between the intrinsic and extrinsic conduction increased from 205 "C for SBN to 300 "C for LSBN. Higher transition temperature and higher activation energy for LSBN could be explained by the fact that the substitution of La3+into the Sr2' sites resulted in an enhanced stability of the perovskite structure due to the larger chemical bond strength of La-0 bonds (799 f 4 kJ/mol at R.T.[16]), as compared with that of Sr-0 bonds (425.5 k 16.7 kJ/mol at R.T. [16]). In addition, stronger chemical bonds would certainly suppress the formation of intrinsic defects and, thus, result in an increased transition temperature. At the low temperature range, a reduced DC conduction may also be attributed to the substitution of trivalent La3+into the crystal structure. The simplified defect chemical reaction could be written as follows: La203------------3 2 LaSr' -t3 Oox + VSr" Where LaSr*represents a lanthanum ion in the strontium site with one effective positive charge, Oox represents an oxygen ion in the oxygen site without effect charge, and VS," represents a strontium vacancies with two negative charges. For ionic conduction in oxide systems, oxygen vacancies are commonly charge carriers for both intrinsic and extrinsic conduction. The partial substitution of divalent Sr2+by trivalent La3+ would introduce one extra oxygen anion which would occupy possible oxygen vacancies available or create Sr2+vacancies. Consequently the amount of charge carriers, i.e., oxygen vacancies would be decreased and thus the DC conductivity reduced. However, the bond strength could not explain the high activation energy observed in CSBN and no other explanation is found at the moment. Figure 4 shows the room temperature dielectric constant and loss tangent of samples SBN, CSBN, and LSBN, as a function of frequency, which ranged from 20 Hz to 1 MHz. Doping of Ca2+ and La3+ resulted in reduced dielectric constants throughout the frequency range. This reduction could be attributed to the reduced atomic polarization due to the smaller ionic radii of both Ca2+and La3+. However, the loss tangent of the LSBN sample was found higher than that of SBN ferroelectrics; whereas the Ca2+doped sample exhibited a reduced loss tangent except at low frequencies. It is known that the loss tangent in ferroelectrics is due to a combination of the space charge polarization and domain wall relaxation. The exact mechanism for the variation in loss tangent with La3+or Ca2' doping is not clear; however, it is possible that the La3+or Ca2+doping may change the mobility of domain walls. This phenomenon was widely observed in isotropic perovskite ferroelectrics [17]. More study is required to achieve a better understanding of the relationship between the compositions, microstructure and physical properties of layered perovskite ferroelectrics. CONCLUSIONS In summary, a single phase layered perovskite SBN ferroelectric doped with 10 at% of La3+or Ca2' has been synthesized through solid state reaction sintering. It was found that the incorporation of both Ca2+ and La3+ resulted in a somewhat reduced dielectric constant, but increased Curie points, which could be attributed to the reduced ionic radii of Ca2+and La3+and the slightly reduced lattice constants. The incorporation of La3+has led to an appreciable reduction in DC conductivity throughout the temperature range studied in the current research. The reduced DC conductivity is explained by the reduction in the concentration of oxygen vacancies, and by
352
Dielectric Materials and Devices
enhanced structure stability due to the stronger chemical bond strength of La-0 bonds. The study of ferroelectric properties, such as P-E hysteresis, is under progress.
Figure3. DC conductivity of SBN (a), CSBN (A),and LSBN (0) as a function of temperature.
Figure4. Loss tangent and dielectric constant of SBN (a), CSBN (A),and LSBN (0)as a function of frequency at room temperature.
Dielectric Materials and Devices
353
REFERENCES: 1. F. Jona and G. Shirane, “Introduction”; pp. 1-27 in Ferroelectric Crystals, Pergamon Press, New York, 1962. 2. B. Jaffe, W.R. Cook, Jr. and H. Jaffe, “The Ppiezoelectric effect in ceramics”; pp. 7-21 in Piezoelectric Ceramics, Academic Press, New York, 1971. 3. Hitoshi Watanabe and Takashi Mihara, “Preparation of ferroelectric thin films of bismuth layer structured compounds,” Jpn. J. Appl. Phys., Part Z34,5240 (1995). 4. C. A-Paz de Araujo, L.D. McMillan, J.D. Cuchiaro, M.C. Scott, and J.F. Scott , “Fatigue-fiee ferroelectric capacitors with platinum electrodes,” Nature 374,6523 (1995). 5. J.F. Scot and C.A.P. de Araujo, “Ferroelectric memories,” Science 246, 1400 (1989). 6. J.F. Scott, “Thin Film Ferroelectric Materials and Devices”, pp.115 in Layered Perovskite Thin Films and Memory Devices, ed. R. Ramesh, Kluwer, Nonvell, MA, 1997. 7. J.F. Scott, “High-dielectric constant thin films for dynamic random access memories (DRAM),” Annual Review of Materials Science, 28, 79 (1998). 8. S.B. Desu and D.P. Vijay, “C-axis oriented ferroelectric SrBi/sub 2/(Ta/sub x/Nb/sub 2x/)O/sub 9/ thin films,” Mater. Sci. Eng. B32, 83 (1995). 9. K. Kato, C. Zheng, J.M. Finder, and S.K. Dey, “Sol-gel route to ferroelectric layer-structured perovskite SrBi2Ta209and SrBi2Nb209thin films,” J. Am. Cerarn. Soc. 81, 1869 (1998). 10. P. Duran-Martin, A. Castro, P. Millan, and B. Jimenez, “Influence of Bi-site substitution on the ferroelectricity of the Aurivillius compound Bi2SrNb209,”J. Mater. Res. 13, 2565 (1998). 1 1. C. Lu and C. Wen, Mater. “Preparation Properties of Barium Incorporated Strontium Bismuth Tantalate Ferroelectric Thin Film,” Res. Soc. Symp. Proc. 541,229 (1999). 12. Y. Wu and G.Z. Cao, “Enhanced ferroelectric properties and lowered processing temperatures of strontium bismuth noibates with vanadium doping,” Appl. Phys. Lett. 75, 2650 (1999). 13. Y. Wu and G.Z. Cao, “Influences of vanadium doping on ferroelectric properties of strontium bismuth niobates,” J. Mater. Sci. Lett., 19, 4 (2000). 14. F. Scordan, “Ionic Crystals”; pp. 420-421 in Fundamentals of Crystallography, Edited by C . Giacovazzo, Oxford, New York, 1998. 15. T. Chen, C. Thio, and S. B. Desu, “Impedance spectroscopy of SrBi2Ta209and SrBi2Nb209 ceramics correlation with fatigue behavoir.” J. Mater. Res. 12,2628 (1997). 16. J.A. Kerr, “Strenghts of Chemical Bonds”; pp. F166 - F118 in CRC Handbook of Chemistry and Physics, lStStudent ed., edited by R.C. Weast and M. J. Astle, CRC Press, Inc., 1988. 17. A.J. Moulson and J.M. Herbert, “Elementary Solid State Science”; pp. 39 in Electroceramics, Chapman & Hall, London, 1990.
354
Dielectric Materials and Devices
CONTRIBUTION FROM FERROELASTIC DOMATN SWITCHING DETECTED BY THE X-RAYS TO R-CURVE BEHAVIOR OF PZT CERAMICS A. E. Glazounov and M. J. Hoffmann Institute for Ceramics in Mechanical Engineering, University of Karlsruhe, D-76 131 Karlsruhe, Germany A. Kolleck and G. A. Schneider Advanced Ceramics Group, Technical University Hamburg-Harburg, D-2 1073 Hamburg, Germany ABSTRACT The ferroelastic domain switching induced during crack growth and the Rcurve behavior were investigated as a function of composition for ferroelectric ceramics Pb(Zr,Ti)03. The compositions had different values of critical tensile stress necessary to switch the domains, which depends upon the Zr/Ti ratio and the type of additive. It is demonstrated that the amount of domain switching is correlated with the increment of fracture toughness in the R-curves and with the magnitude of critical tensile stress. The material with smaller critical stress has a larger amount of the domain switching and a higher toughness increment. INTRODUCTION It is known that the fracture toughness of ferroelectric ceramic materials, such as barium titanate, BaTi03 (BT), and lead zirconate titanate, Pb(ZrxTil+ ) 0 3 (PZT x/( 1-x)), is higher in the ferroelectric phase than in the paraelectric phase [ l , 21. The higher values of fracture toughness are attributed to the so-called Rcurve behavior [3], i.e., an increasing resistance to the crack growth, which was observed both in BT [4, 51 and PZT [6]. It is believed that the major mechanism responsible for the increase in fracture toughness is a crack-domain interaction [2, 5-71. High tensile stresses developed around the crack tip causes ferroelastic domain switching, which consumes energy. In addition, compressive stresses To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the propertv of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, withoit the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
355
developed in the crack wake after switching shield the tensile stress around the crack tip. Both effects act together and slow further crack growth [3, 51. Another possible mechanism leading to a R-curve behavior is microcracking [3]. In ferroelectric ceramics, microcracks typically develop along the grain boundaries as a result of residual stresses which are not compensated by the formation of the domain structure. The appearance of microcracks depends upon the grain size and the magnitude of spontaneous strain due to the phase transition [ 1, 81. In general, both mechanisms, domain switching and microcracking, may coexist, and some efforts to separate their contribution to fracture toughness were undertaken both for BT [ l ] and PZT [6]. The objective of the present work is to investigate systematically ferroelastic domain switching induced during crack growth and the R-curve behavior as a function of composition for PZT ceramics. The compositions selected differ in the magnitude of critical stress required to switch the domains. This approach is justified because it is known that the critical stress plays an important role in materials where R-curve behavior is related to stress-induced phase transformation [3, 91. In PZT, the critical stress depends upon the Zr/Ti ratio [lO] and the type of additive [ 10, 111, which is used to produce "soft" or "hard" PZT [8]. Therefore, several compositions are prepared, which were doped either with La ("soft") or Ag ("hard"). The amount of domain switching near the fracture is characterized using X-ray diffraction (XRD), which gives direct quantitative information about the distribution in the population of domains with different orientation. The results of XRD are compared with R-curves measured from the same materials, to evaluate the contribution from domain switching to the toughness increment. EXPERIMENTAL The compositions investigated are summarized in Table I, along with the type of crystallographic modification of the ferroelectric phase, and the reflections in diffraction patterns. which were selected to characterize domain switching. The type of the crystal lattice was determined from the XRD measurements from sintered ceramic samples. From Table I, one can notice that two materials with the same Zr/Ti ratio, 54/46, have different crystallographic modifications of the ferroelectric phase. Composition doped with 2%La is at the morphotropic phase boundary [81, and contains the mixture of rhombohedral and tetragonal phases, whereas material doped with 0.7%Ag contains only rhombohedral phase. This difference can be attributed to the shift in the morphotropic phase boundary caused by doping with different additives.
356
Dielectric Materials and Devices
Table I. Investigated compositions of PZT ceramics, their crystallographic modification, and reflections in diffraction patterns, which were used to study the domain switching. Zr/Ti
Additive [mol%]
54/46
2%La
Crystallographic modification
Measured reflections
at morphotropic phase boundary - mixture of rhombohedral and tetragonal
(002)/(200) and (1 1I ) / ( ]11)
71)
60/40
2%La
rhombohedral
( I I I)/(I
45/55
2%La
tetragonal
(002)/(200)
54/46
0.7%Ag
rhombohedral
( I II)/(lII)
The samples were prepared by using the conventional mixed-oxide route, as was described in details elsewhere [12, 131. The sintering was done at 1225°C for 2 hours, and the density of the sintered bodies was always higher than 99% of theoretical density. In Table I, the compositions are ranked in order of the value of critical stress for the domain switching, so that PZT 54/46 + 2%La and PZT 45/55 + 0.7%Ag have the lowest and the highest values of stress, respectively. For La-doped PZT, the stress was measured using a 4-point bending test, as described in Refs. [ 14, 151. For Ag-doped materials, the stress was estimated from the data reported in Ref. [ 101 for "hard" PZT ceramics with similar compositions. The detailed description of the experiments is given below along with corresponding results. In addition, an extended version of this work will be published elsewhere [ 161. EXPERIMENTS AND RESULTS Amount of domain switching To determine the amount of domain switching, XRD was measured from the fracture surface, Fig. 1(a), using a standard laboratory diffractometer (D5000, Siemens, Germany) with Cu K a , radiation, A = 1.5406 A . The samples were
Dielectric Materials and Devices
357
first fractured using conventional 4-point bending technique. A special attention was paid to the following detail. In materials which exhibit R-curve behavior, there are two stages of crack growth, which are characterized by a different rate of crack extension: stable crack growth (which is slow, and its rate can be controlled by the applied force), and unstable, where the crack grows with a speed of sound. It is expected that the two regions should be characterized by the different amount of domain switching, because domain switching in ferroelectrics is a dynamic process, which depends upon the rate at which the applied force changes with time. Since in the unstable region the crack propagates with the speed of sound, the elastic force around the crack tip changes very fast with time and therefore, here the amount of the domain switching is expected to be smaller than in the stable region [2, 161.
(b) region with instantaneous crack growth measured region notch sample
Figure 1. X-ray diffraction was measured from the fracture surface, to determine the amount of the domain switching (a). Here the sample was cut into two parts along the dashed line in (b), to study the domain switching from the region with stable crack growth. At the same time, only the region of stable crack growth determines the Rcurve behavior of the material. Therefore. in order to measure XRD only from the stable region, the fractured samples were first cut into two parts, as shown in Fig. l(b). It should be noted, though, that the measured part of fracture surface still contained a mixture of regions with stable and unstable crack growth. When the sample is fractured using 4-point bending technique, the width of the stable region can be as large as 200 - 300 p n , whereas the smallest part of the fracture surface, which could be measured using standard laboratory diffractometer, was approximately equal to I rnrn , Fig. l(b). This smallest width was determined by the necessity to determine the amount of domain switching with an accuracy of about -t I % , in order to make the comparison of the data meaningful.
358
Dielectric Materials and Devices
Another important detail of this experiment is that samples were unpoled before fracture, in order to have a well-defined initial state with isotropic domain orientation. The amount of domain switching was determined from the comparison of two XRD patterns. The first pattern was measured immediately after fracture. Then, the sample was annealed for 4 hours at 5OO0C, which is above the temperature of ferroelectric phase transition, and the second diffraction pattern was measured after annealing from the same sample. The purpose of the annealing was to restore the isotropic domain orientation, and therefore, the difference between the diffraction patterns measured after fracture and after annealing should be directly related to the change in domain structure induced during fracture.
300
200
I
I
300
after fracture (1-11)
1
2
.a
37,6
37.8
.*. ...*..
38,O
38,2
c
I
38,4 38,6
,
[
after annealing
(1-11)
(b)
200
OL
Figure 2. X-ray diffraction patterns measured from the fracture surface of PZT 60/40+2%La after fracture (a), and after annealing (b). Open circles show experimental data, and the solid line corresponds to the fit of the peaks using Gaussian functions. Fit of individual peaks is shown with closed circles. Figure 2 shows an example of the diffraction patterns measured from the fracture surface of PZT 60/40 + 2%La after fracture, Fig. 2(a), and after annealing, Fig. 2(b). One can clearly see the difference in the ratio of peak heights I ( ] / I ( l ~ l iwhich , reflects the domain switching. After annealing, the intensity ratio is I ( ] ] ] /, I(l71) = 0.33, which is characteristic of the isotropic domain orientation for a rhombohedral modification [81. By denoting the ratio of peak heights before annealing as R , and after annealing as R ' , the fraction of the domain switching during fracture was calculated according to [ 171:
Dielectric Materials and Devices
359
s=
R' - R ( I + R')(I + R )
For tetragonal PZT, R and R' correspond to the ratio I(oo2)/ I(2oo), and in the . I1 case of rhombohedral PZT, they correspond to the ratio I ( l l l ) / I ( , T ~ )Table summarizes the values of 6 obtained for all the studied compositions. It gives 6 expressed as a fraction of the domains that can be switched by the external uniaxial stress. In unpoled ceramics, the amount of "switchable" domains is equal to 66.7% for the materials with tetragonal structure, and 75% for rhombohedral compositions [8, 17, 181. Composition at the morphotropic phase boundary, PZT 54/46 + 2%La, represents a special case. Here both the tetragonal and rhombohedral phases coexist, and hence the spontaneous polarization and spontaneous elastic strain have in total 14 orientation states (8 states from rhombohedral phase and 6 states from tetragonal phase) [8]. Therefore. two kinds of switching are possible under the high tensile stress around the crack tip: (i) within each of the phases, and (ii) from one phase to another. To investigate the possibility of stress-induced phase transformation, we calculated the volume fraction of rhombohedral phase, y , both after fracture and after annealing. y was calculated using:
where IFo2,, and I/:&,) denote the heights of the peaks from the rhombohedral, 'lrh", and tetragonal, "tet",phases, which were determined from the fit of measured (002)/(200) reflection. In addition, domain switching within each of the two phases was evaluated using Eq. (1): for the tetragonal phase from the ratio of peak heights I;;;*, / I/;;*, , and for the rhombohedral phase - from the ratio I I I I I , / I ( l T l j . No evidence for the stress-induced phase transformation was detected in PZT 54/46 + 2%La, in the sense that the tensile stress developed during fracture did not favor any of the two states on the macroscopic scale. The volume fraction of rhombohedral phase had close values after fracture, y = (32 f I ) % , and after the annealing, y = (30 k I ) % , compared to the experimental uncertainty. However, this result does not rule out completely such a phase transformation. It is possible that in some parts of the fracture surface transformation occurred from
360
Dielectric Materials and Devices
rhombohedral to tetragonal phase, whereas in the other parts tetragonal phase transformed into rhombohedral, so that the overall amount of both phases remained nearly constant. The amount of domain switching within tetragonal and rhombohedral phases in PZT 54/46 + 2%La. is also given in the Table 11. From this Table, one should notice the important result of this study, that the amount of domain switching induced during fracture is noticeably larger for La-doped PZT, than for Ag-doped ceramics. To our knowledge, this difference in the values of S is demonstrated for the first time for fractured PZT, and agrees with a known fact that in "soft" PZT domain switching is enhanced compared to ''hard'' PZT [8, 10, 113. Table 11. Volume fraction of domain switching, 6, expressed as a fraction of domains that can be switched by external stress, increment of fracture toughness, A K R , of the R-curve, and critical tensile stress, a,, required to switch the domains. Composition
S in terms of "switchable" domains
[%I
[ Mpa&] MR
0,
[n/c~a]
PZT 54/46 + 2%La
22 (tetragonal) 20 (rhombohedral)
0.70
45
PZT 60140 + 2%La
19
0.56
55
PZT 45/55 + 2%La
17
0.20
120
PZT 54/46 + 0.7%Ag
5
0.16
150
PZT 45/55 + 0.7%Ag
1.5
0.1
200
In order to check whether there was any reverse domain switching after fracture, the diffraction patterns were monitored as a function of time for all studied samples. The possibility of reverse switching was suggested by the recent data for BT [S], where R-curves were studied under repeated loading and unloading of the specimen. It was observed that after the load was released, the fracture toughness decreased to its initial value in the R-curve. The result was attributed to the reverse domain switching at the crack wake, which reduced the shielding of the tensile stress at the crack tip [ 5 ] .
Dielectric Materials and Devices
36 1
To detect the reverse domain switching, in the present work the diffraction patterns were monitored as a function of time for all the studied samples. Figure 3 shows the time dependence of the ratio of peak heights, for two compositions: Ill I ) / I( ]T], for PZT 60/40 + 2%La, and I(oo2)/ Ilzoo) for PZT 45/55 + 2%La. No change in the intensity ratio has been observed for both materials within the scattering of the data, which indicates that no domain switching occurred in the samples within the studied time period after fracture.
1
45155 + 2%La
60140 + 2%La 0,o 0.2 O
10'
t
10'
-
4
IO*
103
t
104
105
Time after fracture [Min.]
Figure 3. The ratio of peak heights for two La-doped materials measured from the fracture surface as a function of time elapsed after breaking the sample. Note that the offset in the scale for PZT 60/40 + 2%La was made to avoid overlapping of both plots, since both materials by pure chance have close values of the ratio of peak heights. Although Fig. 3 does not suggest reverse domain switching discussed in Ref. [ 5 ] , we believe that the two results do not contradict each other. The results can differ due to the difference in time scale of both experiments. In Ref. [ 5 ] , the reduction of the fracture toughness occurred within less than 1 minute after the load was released. On the other hand, the shortest measurement time in the present experiments was approximately equal to 5 minutes, Fig. 3, and hence we were unable to probe the "fast" domain reversal [ 5 ] . Therefore, in order to detect reverse domain switching after opening of the crack, more elaborate experimental methods are required.
362
Dielectric Materials and Devices
(002)1"1 (002)* (2,),,
350 F
1
(a)
300 1
23,O
43,5
44,O
i: 44-5
45,O
45,s
20 [Desl
300
i
E'
+
t
50 0 43,O
(c)
(002)'" (002)* (,(XI),,
0
250
'Jj
150
F -
100
50 43,O 0
43-5
44,o
44.5
45.0
20 [Wl
45.5
43,5
44,O
44,5
20 [Wl
45,O
455
Figure 4. X-ray diffraction patterns measured from the fracture surface of PZT 54/46 + 2%La. Plot in (a) was measured after annealing the sample. Plot in (b) corresponds to the region with stable crack growth, and plot from (c) shows the data measured from the region with instantaneous crack growth. Open circles show experimental data and the solid line corresponds to the fit of the peaks using Gaussian functions. Fit of individual peaks is shown with closed circles. The two regions are characterized by the different amount of domain switching. By using from Eq. (l), one obtains 6 = 22% for the ratio of peak heights / (b), and S = 18% for (c).
Figure 4 compares X-ray diffraction patterns measured in PZT 54/46 + 2%La from two regions of the fracture surface, with stable and instantaneous crack growth. One can see that compared to the annealed sample, the change in the ratio
Dielectric Materials and Devices
363
of peak heights
/
after fracture is stronger in the region with stable
crack growth, Fig. 4(b). By using Eq. (I), one can calculate the corresponding amount of domain switching, 6 , within the tetragonal phase. It is equal to 22% for the stable region, Table 11, and to 18% in the region where the crack propagates instantaneously. This data thus supports the argument discussed above that slow crack growth should result in a larger amount of domain switching, due to the dynamic nature of switching process in ferroelectrics. R-curve measurements The R-curves of the same materials were measured using compact tension specimens, as described in detail in Refs. [14, 161. Unpoled samples were studied, in order to compare the R-curve data with the amount of domain switching determined using XRD. Upon loading, the average crack growth velocity was kept constant at about 0.5 - 2 p / s e c , and crack length was measured in-situ with an optical microscope. The fracture toughness, K , , was calculated from the crack length and the applied load, which was measured directly using a force sensor [5]. Figure 5 shows K , measured as a function of crack extension, for compositions doped with 2%La, Fig. 5(a). and with 0.7%Ag, Fig. 5(b). La-doped PZT exhibit raising R-curves, where the fracture toughness starts from 0.6 MPa&, and reaches the plateau value, which is equal to 1.21 MPa& for PZT 54/46 and PZT 60/40, and to 0.87 M P a h for the PZT 45/55 composition. On the other hand, in Ag-doped PZT the R-curve behavior is less pronounced. For example, PZT 45/55 + 0.7%Ag, the initial and plateau values of K R are almost the same, and equal to 1.0 MPa& and 1.1 MPa&, respectively. From Fig. 5 , the difference, AK,, between the plateau value and initial value of the fracture toughness was calculated, and it is listed in Table I1 for all investigated compositions.
DISCUSSION In order to check how well domain switching accounts for the R-curves of the studied "soft" and "hard" PZT, we will now compare the obtained values of the amount of switching, 6 , and toughness increment, dK, . The discussion is based on the argument that in all the studied compositions, domain switching is the only mechanism responsible for the R-curve behavior. A possible contribution from ~
364
Dielectric Materials and Devices
'0:O'
'
'012'
'
'014'
'
'016'
'
'018'
'
'1:O'
'
Crack Extension [mm]
0,4
0,O
0,2
0,4
0,6
0,8
1,0
Crack Extension [mm]
Figure 5 . R-curves (fracture toughness, K R , plotted against crack extension) for the studied PZT compositions, doped with La (a), and Ag (b). The legends give the Zr/Ti ratio. microcracking to the increase in fracture toughness [ l , 61 can be excluded from the discussion because a careful examination of polished and etched surfaces of studied ceramics using SEM did not reveal the presence of microcracks, both in Ag and La-doped PZT [ 191. From Table 11, one can clearly see that the materials with a larger amount of switching exhibit stronger resistance to crack growth. For example, the highest toughness increment, A K , = 0.70 MPa&, is observed in PZT 54/46 + 2%La, which also exhibits the largest amount of the domain switching, 6 = 22%. At the
Dielectric Materials and Devices
365
same time, PZT 45/55 + 0.7%Ag has the smallest increment of fracture toughness, A K , = 0.1 MPa&, and the smallest amount of switching, 6 5 2 % . Therefore, this result clearly demonstrates the role of domain switching in the R-curve bkhavior of PZT ceramics: the material where domain switching is enhanced, i.e., "soft" PZT, will demonstrate stronger resistance to fracture. In addition. one can also correlate the amount of domain switching with the critical tensile stress, crc, required to switch domains. The values of a, are listed in Table 11, and they were obtained as follows. For La-doped compositions, the critical stress was determined from stresdstrain curves measured using 4-point bending test [ 14, 151. At the same time, the tensile stress developed in this test was not sufficient to cause domain switching in Ag-doped compositions, before fracture of the samples occurred. Therefore, the values given in Table I1 were estimated from the stresdstrain curves measured under compression for "hard" PZT ceramics with similar compositions [lO]. The tensile stress, oC,was calculated assuming that the critical stress is about 20% smaller for tension than that for compression, which has been demonstrated recently for PZT ceramics The comparison of data for the amount of domain switching with critical stress, Table 11, also demonstrates a clear trend: the smaller the critical stress, the higher is the fraction of switching, and the difference is especially strong between La-doped and Ag-doped compositions. This result agrees with a known fact that in "soft" PZT domain switching is enhanced compared to "hard" PZT [S, 10, 113. Although the qualitative agreement between the data for 6 and A K , is established, at present we cannot provide their accurate quantitative comparison. For example, from Table I1 one can see that in the three studied compositions of La-doped PZT, A K , exhibits substantial difference, whereas the amount of domain switching has close values. First of all, in order to explain this difference in A K , , one should also take into account other parameters related with fracture process. According to fracture mechanics, the toughness increment in the R-curve can be written as [20]:
where a = 0.22 is a constant, E is Young's modulus, v is Poisson's ratio, AS, is the strain related with the domain switching, and w is the width of the process zone where switching occurs. Most of these parameters are not known yet for the studied compositions of PZT ceramics, and they are the subject of ongoing experiments.
366
Dielectric Materials and Devices
There is another reason which may explain why the amount of domain switching does not match exactly the fracture increment of the R-curve. Strictly speaking, in Eq. (3), A K , must be compared with S measured from the region with a stable crack growth only, which we could not achieve at the present stage. As was described above, the measured part of fracture surface was a mixture of both regions: of stable and unstable crack growth. Therefore, S given in Table I1 most likely represents the average value from both of them. Since unstable crack growth is characterized by smaller amounts of domain switching, as shown for example in Fig. 4, the values given in Table 11 are smaller than 6 corresponding to stable crack growth. However, at present it is not clear how large this underestimate is, and to which extent it changes from one material to another. SUMMARY Ferroelastic domain switching induced during crack growth and the R-curve behavior were investigated as a function of composition for PZT ceramics. The compositions selected differed in the magnitude of critical tensile stress necessary to switch the domains, which depends upon the Zr/Ti ratio and the type of the additive, La or Ag. The following results were presented. The correlation was demonstrated between the volume fraction of domain switching, 6 , toughness increment of the R-curves, AK, , and the critical tensile stress, 0, . The highest values of the amount of the domain switching, S = 22%, and toughness increment, A K , = 0.56 M P a h , were measured for composition PZT 54/46 + 2%La, which has the smallest value of the critical stress, CT, = 30 MPa . At the same time, the composition PZT 45/55 + 0.7%Ag with the highest value of critical stress was characterized by a very small amount of domain switching and toughness increment. Thus, this result demonstrates that the material with enhanced domain switching, such as "sofill PZT, exhibits stronger resistance to the fracture. No evidence for a reverse domain switching after fracture has been detected, during a long time period starting 5 minutes after fracture. It was pointed out that this result does not rule out the possibility of the reverse domain switching discussed in Ref. [ 5 ] , because most likely switching may occur over a very short time after crack opening.
Dielectric Materials and Devices
367
ACKNOWLEDGEMENTS The authors would like to thank Mr. J. Lupaca-Schomber (IKM) for his help in samples preparation. This work was supported by the German Research Foundation (DFG) under the contract #Ho 1165/3-5. REFERENCES 1. R. Pohanka, P. L. Smith, and S. W. Freiman, in "Electronic Ceramics" edited by L. Levinson (Marcel Dekker Inc., New York, 1987), 51-138. 2. K. Mehta and A. Virkar, J. Am. Ceram. Soc. 73, 567 (1 990). 3. D. Munz and T. Fett, "Ceramics: mechanical properties, failure behavior, material selection" (Springer-Verlag, Berlin, 1999). 4. R. F. Cook, B. R. Lawn, and C. J. Fairbanks, J. Am. Ceram. Soc. 68, 604 (1 985). 5 . F. Meschke, A. Kolleck, and G. A. Schneider, J. Europ. Ceram. Soc. 17, 1143 (1 997). 6. S. Baik and S. M. Lee, J Mater. Sci. 29, 61 15 (1994). 7. Zh. Zhang and R. Raj, J. Am. Ceram. Soc. 78, 3363 (1995). 8. B. Jaffe, W. Cook, and H. Jaffe, "Piezoelectric Ceramics" (Academic Press, London, 197 1). 9. D. J. Green, R. H. J. Hannink, and M. V. Swain, "Transformation Toughening of Ceramics" (CRC Press, Boca Raton, 1989). 10. A. B. Schaufele and K. H. Hardtl, J. Am. Ceram. Soc. 79,2637 (1996). 11. H. Cao and A. G. Evans, J. Am. Ceram. Soc. 76, 890 (1993). 12. M. Hammer and M. J. Hoffmann, J. Am. Ceram. Soc. 81,3277 (1998). 13. M. Hammer. C. Monty, A. Endriss. and M. J. Hoffmann. J. Am. Ceram. Soc. 81, 721 (1998). 14. A. Kolleck, J. Saldana, and G. Schneider, Annual report for DFG project #Ho 1 165/3-4 (1 999). 15. T. Fett, D. Munz, and G. Thun, J. Mater. Sci. Lett. 18, 1641 (1999). 16. A. E. Glazounov, H. Kungl, J.-Th. Reszat, M. J. Hoffmann, A. Kolleck, G. Schneider, and T. Wroblewski, subm. to J. Am. Ceram. Soc. 17. E. C. Subbarao, M. C. McQuarre, and W. R. Buessem, J. Appl. Phys. 28, 1194 (1 957). 18. D. Berlincourt and H. Krueger, J. Appl. Phys. 30, 1804 (1959). 19. M. Hammer, A. Endriss, and M. J. Hoffmann, Annual report for DFG project #Ho 1 1650-3 (1 998). 20. R. M. McMeeking and A. G. Evans, J. Am. Ceram. Soc. 65,242 (1982).
368
Dielectric Materials and Devices
INTERPRETING PIEZOCERAMIC IMPEDANCE MEASUREMENTS Arthur Ballato US Army Communications-Electronics Command Fort Monmouth, NJ 07703-5201 ABSTRACT Impedance measurements of piezoceramics require careful interpretation in order to yield consistent results. Addressed are some of the potential pitfalls associated with this type of measurement. Topics include: practical ranges of material parameters encountered, behavior and suitability of various equivalent networks, proper incorporation of loss mechanisms into circuit descriptions, and definitions of characteristic frequencies available for measurement. INTRODUCTION Electrical impedance measurements provide simple and cost-effective means of inferring pertinent material properties of electroceramics. Material values of interest include dielectric permittivity, piezoelectric coupling, elastic stiffness, and coefficients relating to various loss mechanisms. These are required to be known with ever-increasing accuracy to meet the demands of wireless communications, and other modern technology applications. Interpreting measurement results requires an appropriate network model, and proper concordance of the material quantities with each circuit element. This paper reviews simple network models, and discusses how measurement results interpreted uncritically can lead to inaccuracies. It is instructive to begin by recalling the measurement situation that obtained when piezoelectric vibrators were a novelty. Quartz resonators were first used in the 1920s for the purpose of controlling oscillator frequencies. These components had mechanical quality factor (Q) values in the range 103 to 106. The presence of such high Q values led, in the early years, to difficulties in measuring the equivalent circuit parameters. These difficulties were dealt with, for example, by adding resistors or capacitors in series or parallel with the crystal to worsen the crystal characteristics artificially. Knowing the values of these added elements allowed the crystal electrical parameters to be inferred, so that the crystal could be matched to the using oscillator, and overall performance optimized. Eventually, the procedure was turned around, and the resonator method for determining material parameters was developed. In this method, the electrical circuit parameters of a resonator are measured. Since the equivalent circuit To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the propertv of The American Ceramic Society. Any duplication, reproduction, or re ublication of this ublication or any part thereof, without the express written consent of The American Ceramic Society or fee paicfto the Copyright &earance Center, is prohibited.
Dielectric Materials and Devices
369
parameters are presumed to be known functions of the material constants and dimensions, knowing the circuit values allows the material constants to be determined. If the substance under investigation is not piezoelectric, it is bonded to a piezoelectric transducer of known properties to form a composite resonator, which is then measured. Knowing the characteristics of the piezo-resonator and the electrical behavior of the bonded composite, allows unknown material coefficients, such as elastic stiffnesses, to be found. With regard to loss, the situation with piezoceramics is just the opposite what it was in the early days of the quartz resonator. Instead of artificially ‘blunting’ the resonance in order to be able to measure it, one now has to deal with the relatively flat response that results when the resonator mechanical Q becomes lower than 102or so. It might be asked why, if high Q materials are readily available, and are of relatively low cost, low Q substances are even considered for use. The chief answer is cost, another is coupling. It is true that some quartz resonators, such as those in wristwatches, are quite inexpensive. This is largely because of the uniqueness of the single-frequency requirement (32 kHz; actually 2’’ Hz), coupled with exceedingly large demand. When the parameter specifications (e.g., frequency), are variable, and the demand not very large, the cost of singlecrystal piezo-resonators and transducers can be appreciable. Cost drives most applications. In the modern telecommunications business, fractions of a cent per component in a hand-held unit can mean the difference between success and failure. Applications such as these look very appealing for piezoceramic resonators, which can be manufactured at very low cost, compared to single-crystals such as quartz and lithium niobate. Piezoceramics are polycrystalline, ferroelectric substances that can be readily fashioned in required geometrical forms, and subsequently polarized (‘poled’) in an external electric field to induce piezoelectric properties. The piezoelectric coupling can be very large, in some instances exceeding 75%. These attributes make piezoceramic materials suitable for wideband filter applications as well as resonator and transducer usage at affordable costs. While in the past the process of making electrical measurements on resonators was difficult, and the results prone to errors of various sorts, at present instrumentation employing, e.g., digital, crystal-controlled frequency synthesizers and network analyzers has dramatically improved, and made immittance measurements over wide frequency ranges a relatively routine task today. (Immittance refers generically to either impedance or admittance, when it is not necessary to make a distinction.) While network analyzers have accuracy/precision specifications, these are today not limiting factors for our purposes, and it is presumed that instrumentation errors are negligible.
370
Dielectric Materials and Devices
Interpretation of immittance measurement results, however, is often problematical, particularly when dealing with lower Q and higher coupling materials. One temptation is to let the automatic measuring equipment do the thinking as well as the measuring. Apart from this failing, the main interpretative sources of error appear now to be: - failure to distinguish between the circuit differences arising from the type of excitation (applied electric field either parallel to or perpendicular to acoustic wave propagation direction) - lack of appreciation of the fact that the ‘width’ of a piezoceramic resonance curve is a function conjointly of both piezocoupling and losses - confusion over where various loss mechanisms are to be incorporated into the equivalent circuits use of inappropriate formulas found in outmoded standards documents Each of these points will be addressed in the sequel.
-
BUTTERWORTH-VAN DYKE CIRCUIT The simplest, and most often used model is the Butterworth-Van Dyke (BVD) circuit, consisting of a series RlLlCl arm with shunting capacitor CO [1]-[4]. This four-element model has been used extensively since it was devised, and has a rich literature, e.g., [5]-[23]. It is often adequate in the vicinity of a single mechanical resonance. The traditional BVD circuit does not distinguish between mechanical and electrical Qs and can therefore lead to confusing results for certain ranges of material values. Distinction between various losses will be considered further subsequently. The circuit may be modified slightly by the addition of a shunt conductance Go in parallel with COto accommodate various types of dissipation in both the high and low loss regimes. When necessary to make a distinction, the four-element network will be called BVD4, and the augmented version, to be discussed in greater detail below, will be referred to as BVDS. To illustrate the correspondence between material coefficients and circuit element values, we give expressions for the simple thickness modes of plates, where lateral variations in motion are absent: CO= E (Nt), static capacitance C1 = (8/n2) (e2/c) (Nt), motional (dynamic) capacitance RI= (n2/8) (q/e2) (t/A), motional (dynamic) resistance L1 = (1/8) (p/e2) (t3/A), motional (dynamic) inductance 9
Material coefficients are: p, mass density; q, acoustic viscosity; e, piezoelectric constant; permittivity; and cE, elastic stiffness at constant electric field
Dielectric Materials and Devices
E,
dielectric
37 1
Auxiliary definitions are: A, electrode area; t, plate thickness c = (cE+ e2/&),piezoelectrically stiffened elastic constant 21 = q/c = R1C1, motional time constant k2 = e2/(&c), piezoelectric coupling coefficient Q = Q1= d(Ll/Cl)/Rl= (l/n)-[d(c*p)/q]*(t),mechanical quality factor r = (C&) = (7r2/8)*(& c/e2)= capacitance ratio 01 = n*d(c/p)-(l/t)= 1 / d ( ~ 1cl), series resonance circular frequency M = Q/r = figure of merit E = M=Q= Q2/r = M2.r = figure of excellence The capacitance ratio, r does not contain loss, only piezocoupling, and the mechanical quality factor, Q does not contain piezoelectric coupling, only loss. M and E contain both piezoelectricity and loss. Notice that the presence of piezoelectricity appears in two distinct aspects: In the volumetric piezostiffening, (c - cE) = et/&,which for ceramics, can be quite large, and In the series RlLlCl arm circuit values that contain geometrical factors; the piezo contribution is due to the wave interactions at the resonator surfaces. In many practical situations involving high frequency plate modes, one is forced to consider the true lateral distribution of motion. More complicated expressions for the circuit values then appear, but often one can use the above simple expressions for uniform plate motion, but modified by multiplicative factors associated with the lateral distribution. For other geometries and modal types, e.g., extensional modes of beams, the same BVD circuit is applicable, but with different formulas relating the circuit parameters and material coefficients. Having a correspondence between physics and circuit elements can be used in either direction: knowing material coefficients allows the circuit elements to be optimized for a given application; alternatively, knowing the circuit elements permits inferring the material coefficients. When the BVD4 circuit is adequate, then there exist a variety of means of extracting the circuit values from immittance measurements; a few are sketched below; see also Ref. 24. IMMITTANCE FUNCTIONS VERSUS FREQUENCY Figure l a shows the BVD4 circuit; the four element values are constants for a given resonator. In Figs. l b and l c are the respective equivalent series impedance and shunt admittance representations. Series resistance, Rs and
372
Dielectric Materials and Devices
reactance, X, and shunt conductance, G, and susceptance, B, are functions of frequency. In order to show the frequency behavior of these and other functions at a convenient scale, values of r = 2, Q = 6 (M = 3), and R I = 1 ohm will be assumed [9, 111. In Fig. 2 is plotted complex admittance Y = G, + jB, as function of frequency. The space curve assumes a helical aspect in the vicinity of resonance; also shown is the projection of the ‘admittance circle’ on the G, - B, plane. Impedance functions lZl, R,, X,, and X1 (reactance of the RlLlCl arm), are plotted in Fig. 3, and Fig. 4 shows the frequency behavior of the real and imaginary parts of the admittance, G, and B,. Figures 5 and 6 depict, respectively, the impedance and admittance circles. The curves in Figs. 2 to 6 are obtained from the relations [9,11]:
Z = l/Y = jX{(A - j q)/[(l - A) + j q]} = R, + jX, R, = X q/[(l - A)2+ q2] X, = X [A (1 - A) - q2]/[(1- A)2+ q2] G, = (q/X)/(A2+ q2) B, = ( m u - “(A2 + q2)1),
= (G,
+ j B,)-’
where the following definitions have been used: o = 2nf; X = l/oC, = Rl/q; R ; = a12= 1/d(L1 C1); q = R/M; and A = r (R2 - 1). Shown in Figs. 2 to 6, and defined in Table I are various critical frequencies that may be used in determining the BVD4 circuit element values. Expressions for these frequencies in terms of the generalized frequency variables A and q, plus the circuit parameters are found in Ref. 11. At the unique equiimmittance frequencies f, and fh, complex immittance has the same value. At these frequencies, and with the abbreviation N = d[l + 1/2r - 1/4Q2], we have:
= o/o, = f/f,
(R,/R1) = (2r + l)-’ (Xsmd = “-24 N)/(2r + I)] (G, RI) = r/2Q2 (BP RI) = N/M (Y RI) = (R1/Z) = d(l + 1/2r)/M tan(@ = (Xs/Rs)= (-B,/G,) = (- 2Q N) (fit, - a$= d[(2Q2/r) - 1]/Q (Rh Q$= (2 N) a h oRg=l.
Dielectric Materials and Devices
373
Table I. Critical frequencies of the BVD4 circuit Symbol Condition fg ,fh Equi-immittance points f, ,fy Parallel capacitance extrema fw9,f,, Parallel susceptance extrema f,,,,f, Immittance extrema fs Series resonance (XI = 0)
Symbol f, (fq) f, ,fa f, ,f, fP’ fP
Condition Maximum phase Resonance/Antiresonance Series reactance extrema Maximum series resistance Parallel resonance (RI= 0)
THICKNESS AND LATERAL EXCITATIONS There are two ordinary electrode configurations used for exciting plate modes; these are designated thickness excitation (TE) and lateral excitation (LE). Sketches of the two cases are shown in Fig. 7. In the pertinent theory [16,17,22,23], a distinction is made between the BVD circuits appropriate to each case. For LE, the BVD of Fig. l a yields the proper relationship between the capacitance ratio and piezocoupling, viz., r = (Co/C1) = (n2/S)/(k2).In the TE situation, however, the BVD must be augmented with a negative capacitor of magnitude COin series with C1. This has the effect of increasing the effective capacitance ratio for TE by unity. It is often the case that, in measuring TE resonators, the (- CO)is omitted from the representation, and the BVD4 of Fig. l a is used. This is a perfectly acceptable procedure if the resultant value for r is then increased by unity when determining the piezocoupling factor; that is, kTE = (eTE)/d(&TE c) = (n/h)/(r + 1). Figure 8 is a sketch of the regions in (Q, r) - space occupied by four classes of piezoelectrics; see also Ref. 25. It is seen that for quartz resonators, omission of the unity factor is usually of little consequence, whereas for ceramics, the omission can result in serious error in determination of the true coupling factor. The difference between TE and LE is shown in Fig. 9, as function of r, where r refers to the value of (Co/Cl) reckoned in the BVD circuit appropriate for each excitation, without (- CO)for LE, and with (- CO)for TE. In Fig. 10 is given the relation between the true and apparent coupling values when the TE coupling is incorrectly evaluated using the LE version of the BVD4 circuit. CIRCUIT VALUE EXTRACTION BY IMMITTANCE MEASUREMENTS Several methods, from among many, whereby the element values of a BVD4 circuit may be determined are outlined briefly below. These are the parallel capacitance method, the nfh-powerpoints method, and the maximum phase method.
374
Dielectric Materials and Devices
The Parallel Capacitance Method One simple method by which to extract the BVD values is to use the input (parallel) capacitance curve of the resonator. Parallel capacitance is defined as C, = B,/o. Figure 11 shows the frequency variation of normalized capacitance (Cp/Co), in the range of resonance, for both a lossy resonator, and for a resonator with negligible loss. For the BVD circuit, C, is obtained from the relation [(C,/Co) - 11 = r/[A2+ q2] = [(l - R2)/r]/[(l - Q2)' + (Q/Q)21. Then Q/r = M is found from the extrema of the capacitance curve using = f (M/2)/[1 T
[(Cp/Co) - llextrema
-
1/(2Q)] M/2, SO that
The maxima of C, occur at frequencies R = d(1 f UQ), whereas the reactance zeros occur at R 1 and d(1 + l/r). The capacitance ratio, r, follows from
-
-
In this last relation, R is confined to the nearly hyperbolic region of normal dispersion, viz., somewhat away from the region between the extrema of C,. Knowing the static capacitance, COand the nominal resonance frequency, f1 yields C1, L1, and RI. As a check, [a(C,/C,)/aR]n = 1 = - 2 Q2/r = - 2 E. The nth-Power Points Method of Determining Q A method often used for determining circuit Q is based on the presumption that the width of a 'resonance' curve (which may be any of a number of the network functions described above, e.g., lyl, or R,) is inversely proportional to Q. One measures the frequency span between points on both sides of the maximum, that are a fixed fraction thereof. The success of the method depends on the network function used, and on the value of piezocoupling. The method was originated many years ago for resonant systems representable by a series RlLlCl circuit, where the COwas absent. In this case, the method is exact when applied to the admittance magnitude. With CO included, use of method without taking the influence of k into account can lead to misinterpretating the data. The case where COis absent can be thought of as the (impossible) limit of the BVD4 where k + CO. Then IZi"I2= R12+ X12, and normalized impedance is
Dielectric Materials and Devices
375
lzll = (ZinI2 (01C1)~.Admittance is lyll = 1z11-l = + (0- 1/Q)2]-1’2, which has a maximum value of Q2 at Q = 1. It is an even, but not a symmetric function of Q. Setting Iy1I2 = n Q2, and solving for Q; yields the general relation for the nth- power points [26]: (AQ Q) = d(n-’ -1). Often used are the relations: Q = (f,,/Af) = l/AQ, for n = % (3-dB points), and Q = 3/AQ, for n = 0.1 (10-dB points). When COis present, both k and Q influence the shapes of the ‘resonance’ curves. Figure 12 shows the admittance versus frequency for r = 2, Q = 6, and R1 = 1 ohm. The curve assumes an asymmetric aspect with the addition of CO,which influences the result. There is cooperation between coupling and loss in the sense that when coupling increases, the minimum Q for which the method is accurate, decreases. In this respect, it is fortunate that the relatively low values of Q encountered in piezoceramics are usually accompanied by high coupling. Figures 13 and 14 plot normalized admittance of a BVD4 resonator versus frequency for, respectively, r = 2, with Q as parameter, and Q = 100, with r as parameter. With low coupling (high r), the asymmetry leads to loss of accuracy; as r diminishes, the curves become more symmetric, and the method yields increasingly better results. Tables I1 through V contrast the results obtained using the 3 dB points of lyl for various combinations of r and Q to demonstrate the ranges of applicability of the method 1271. At low k values, the situation degrades rapidly for low Q. Table 11.3-dB bandwidth method; k = 8.8%; r = 159 Q a(-) a(+) AQ (Ydmax Q AQ 1000 0.99932 1.0004 0.00108 1024.8 1.08 100 0.9221 1.0019 0.0798 216.42 7.98 10 0.7015 X X 159.96 X 1 X Table 111.3-dB bandwidth method; k = 30%; r = 13.7 Q a(-) Q(+) AQ (Ydmax Q *AQ 1000 0.99949 1.0005 0.00101 1000.2 1.01 100 0.99347 1.0038 0.01033 101.75 1.033 10 0.8183 1.0205 0.2022 19.18 2.022 3 0.648 X X 14.138 X 1 X
376
Dielectric Materials and Devices
Table IV. 3-dB bandwidth method: k = 60%: r * 3.43 Q Q(-, a(+) ASZ (J’drnax Q *ASZ 1000 0.9995 1.0005 0.001 1000 1.00 100 0.99466 1.0047 0.01004 100.12 1.004 10 0.9154 1.0271 0.1117 11.033 1.117 3 0.6921 1.0837 0.3916 4.9734 1.175 1 X Table V. 3-dB bandwidth method; k = 90%; r = 1.52 Q a(+) ASZ (y1)max Q*AQ 1000 1.oo 1000 0.9995 1.0005 0.001 100 0.99486 1.0049 0.01004 100.02 1.004 10 0.93622 1.0375 0.10128 10.224 1.0128 3 0.7471 1.0796 0.3325 3.5736 0.9975 1 0.5173 X X 1.8418 X If the function g, = (G, RI)is used for measurement of the resonance width, it will be shown below that the result is exact because it does not depend on coupling. Figure 15 shows the frequency dependence of g, for various Q values. For the BVD4 circuit, the admittance is
Y = joC, (1 + (1 - SZ2)/rD}+ oC, SZ/QD, where D = ((1 - SZ2)2 + (SZ/Q)2}. Hence, G, = oC, SZlQD, and g, = (G, RI) = (SZ/Q)2/D. When Q = 1, g, = 1, which is the maximum. At the half-power points, g, = %; the frequency relation is Q2 k (SZ/Q) - 1 = 0, and is a function of Q only. Roots are: = (+d[1 + (1/2Q)2]- (1/2Q)}, and
a(+) = (+d[l + (1/2Q)2]+ (1/2Q)}. Therefore, ASZ = (a(+) - SZ,) = 1/Q, and ASZ Q = 1. Summary of the nth- Power Points Method of Determining Q 0 Without CO:using lyll, relationship Q ASZ = d(n-’ -1) is exact. With CO:using (RJRI), relationship Q AQ = d(n-’ -1) is not exact; it - (1 + l/r)I2+ depends on r and Q. (R,/R1) = (l/r)’/Dl; D1 = (a2 f D. 0 With CO: using lyl= [IYl/(olCt)] = jSZ Ir + K’I, where H = [(l - Q2) + j(SZ/Q)]. Relationship Q AQ = -1) is not exact. 9
Dielectric Materials and Devices
377
0
Same remarks hold when using lzl, B,, C,, and X, as for lyl and (R,/Rl). With CO:using (G, RI), relationship Q AQ = d(n-' -1) is exact.
-
DISAPPEARANCE OF INDUCTIVE REGION Between the resonance (fR) and antiresonance (fA) frequencies, the BVD4 resonator appears inductive. It is instructive to consider the condition where loss increases to the point where fR = fA, and the inductive region shrinks to a point. When mechanical Q = Q1 is lowered to the point where
resonance ceases to exist. This condition may also be stated as k = (7d2) (l/dQ) d[l + 1/(2Q)] = (x/2) (l/dQ), for sufficiently high Q.
So, provided that k2 Q is greater than approximately ( ~ / 2 )resonance ~, will occur. The size of k2 Q is a measure of how robust is the resonance; this is closely related to the results in Tables I1 to V. Table VI provides equivalent formulas expressing the condition. Table VI. Alternative statements of the fR = fA condition. F(r) F(Q) F(M) r= r Q'/(2Q + 1) 1/[M(M - 2)] l/(M - 2) Q Q = r + d[r(r+l)] M 2 + 1/Q M= 1 + d(l+l/r) M/(M-2) E= 1+2{r+d[r(r+l)]} 1 + 2 4
F(E) (E - 1)L/(4E) (E - 1)/2 2E/(E - 1) E
A SECOND LOSS PARAMETER: ELECTRICAL Qo In ceramic resonators, there is a second loss element that often cannot be neglected [15, pp. 95-96]. This is the dc conductive path that exists in parallel with the static capacitance CO[28-341. When it is necessary to include this feature, it is modeled by adding a conductance Go = G (A/t) in shunt with a lossless CO;0 is the effective conductivity, comprised of ohmic conductivity, along with dielectric loss; this is permissible as long as &"(a)= constant in the frequency range considered. Some circuit models incorporate shunt loss by allowing COto assume a complex value; this is an equivalent picture. One not infrequently finds in the literature a confusion concerning the attribution of the various loss mechanisms to elements in an equivalent circuit. Sometimes conductivity effects are lumped with RI, (which ordinarily
378
Dielectric Materials and Devices
only accommodates the acoustic viscosity [35] and ambient loading); this lumping into a single loss element obscures the distinction of the effects, and leads not only to incorrect assessments of circuit values, but also to inaccurate predictions of circuit frequency behavior. In the usual BVD4 circuit, mechanical Q is defined as Q1 = (01 q)-' = d(Ll/Cl)/Rl, and electrical Q is undefined. In the five-element refinement electrical Q is defined as containing the shunt Go =
lm,
Qo = (01
2 ),
= &Co/d(LIC1) = (z) [ E ~ ( c / ~ ) / c J ](l/t).
With the addition of Go to BVD4, the composite is called BVD5. When Q is not further qualified, or bears no subscript, Q1 is meant. The two resistances are characterized by two material time constants, z1 = q/c = RI C1 and zo = E/G = & Co. One desires 71 << 1, and zo >> 1. Conductivity dominates over static capacitance when 01 zo = Qo << 1. The relation
is a useful measure of total loss. Loss tangent is related to Qo as follows: Qo = 01 & CO= 0 1 E/G, whereas loss tangent is defined as tan(6) = 1/Qs at a low frequency, coo, often 1 kHz. If we take QS = coo z, where zo is the same time constant used in defining Qo, then zo = Qdo, = Qo/ml ,and
Table VII provides typical values for room-temperature specific resistivities p = l / ~ of , metals, semiconductors, and insulators [36, 371. Although quite variable, 'average' specific conductivity values for piezoceramics range between those of polyethylene and silicon; a typical value being 0 = 10-9 S/m. Cermets formed by thermal processing of ceramics in a reducing atmosphere [38] range in specific conductivity values between those of lead and nichrome; a typical value might be 0 = 10+6 S/m. Additional data are contained in Ref. 10. Addition of Go to the traditional BVD4 affects some quantities, and leaves others invariant. The effect on the admittance circle is shown in Ref. 10 (Fig. 30, p. 244). Since Go appears in parallel with G,, the entire circle is merely shifted a constant amount to higher conductance values; the effect on the impedance circle is somewhat more involved. In order to apply the nfh-power method to the normalized conductance g,, one has first to subtract the constant value (Go RI) value. The quantity (Go RI) = (.rr2/8)-(oq/e2);
Dielectric Materials and Devices
379
contains no geometry, but is an admixture of both loss mechanisms. The C, curve is unchanged, because Go adds solely to the real part; therefore, results for C, apply to both the four- and five-element BVD citcuits. The critical frequencies of BVD4 that are changed in BVD5 are: fg, fh, f,, f,, f, (= fq), f,, f,, f,', and f,. Those unaffected are: f,, fy, fwl, fvq, f,, f,, and fa. Since resonance and antiresonance frequencies are unchanged, the formula relating r and Q subject to the constraint f, = fa is unchanged. Instrumentation may be used to balance out COand/or Go;the effect on IYI of altering COin the BVD4 circuit is shown in Fig. 16. Table VII. Specific resistivities, p of various materials [36] Metal (20°C) Ag cu At Brass Ni Fe Steel Constan tin Nich rome
p (Q-m), 1.6 E-8 1.7 'E-8 2.8 E-8 -7 E-8 7.8 E-8 10 E-8 -11 E-8 49 E-8 100 E-8
Insulator Polyethylene Glass Porcelain, unglazed Rubber, hard EPOXY Pure Semiconductor Si Ge C (graphite)
P (Q-m) 2 E+11 E+12 E+12 E+13 E+15 p (a-m) 2.6 E+3 4.2 E-1 3.5 E-5
-
PHASE ANGLE MEASUREMENTS Phase angle cp (= e) is defined from the relations: @ = tan cp = XJR, = - B,/G,.
For the BVD4 circuit with numerical values of Fig. 2, the variation in the region of resonance is shown in Fig. 17. Even in the case where the inductive region is not reached (cp < 0), the frequency of maximum phase is always available for measurement. Obtaining the peak of the phase versus frequency curve experimentally for piezoceramics is analogous to that pertaining to Xray 'rocking curves' [39], because the observed phase curves are quite similar to that in Fig. 17. One rocks to both sides of the maximum and estimates the maximum by interpolation; many network analyzers, however, are able to find the maximum electronically. When the Q is high, however, the phase is nearly 90" over most of the region between the resonance and antiresonance frequencies, and finding the phase maximum then is akin to making Brewster angle measurements about a shallow extremum. A typical specification of phase accuracy with network analyzers (e.g., HP 4194
380
Dielectric Materials and Devices
Impedance/Gain-Phase Analyzer), is f %O, so the shape of the phase curve is of considerable importance. Figure 18 depicts the variation in angle of phase maximum versus Q, with capacitance ratio r as parameter. At high Q values, the maximum phase,
@(R) = [r Q2- 2 r Q2 R 2 + r Q2 R4 + r R2 + Q2- Q2 0 2 ]/ [Q R] d@(R)/dR = [- r Q 2 - 2 r Q2 R2 + 3 r Q2 R4+ r R 2 - Q 2 - Q2 R2]/ d2CD(R)/dR2= 2 Q [r + 3 R4r + 11 /
[a2Q]
[a3].
These are to be evaluated at frequencies Ru, Rs,and a,,. When the BVDS circuit is required, the values Q1 and Qo, conjointly with the piezoelectric coupling factor k, determine completely the shape of the phase versus frequency curve. Whereas the phase curve for BVD4 is monotonic with frequency on both sides of a, that for BVDS exhibits a local minimum below R,. This is seen in Fig. 21, where, for graphical convenience, Qo is taken to equal Q1. The phase maximum continues to occur between Rs= 1 and R,,= d(1 + l/r), but R, for BVDS is less than R, for BVD4. For the BVDS circuit, as R increases, @(a) approaches (- Qo) O; slope is (- Qo). As R approaches zero, O(R)approaches (- Qo (r + l)/r) 0; so slope is (- Qo (1 + l/r)). For fixed r and Q1, as l/Qo becomes smaller, R m i n (frequency of phase minimum) goes to zero. At R = 0, (Qo finite), CD(0) = 0; this combination of zero value at R = 0, and R m i n approaching zero makes the inverse slope approach zero. On the other hand, the point of maximum phase
Dielectric Materials and Devices
38 1
is very little influenced as l/Qo goes to zero. One finds that Ru increases with increasing Qo, approaching a limit point as l/Qo + 0. The change in Ru with Qo is not large (for reasonable values of Qo), whereas the change in R m i n is considerable, going from R = (but less than) 1 to 0, with increasing Qo. The phase behavior at Rs = 1 is a useful condition to examine, because at this frequency G, is a maximum, and is relatively easy to measure. For BVDS,
so cp < 0 at the series resonance frequency, Rs.As l/Qo + 0, (vanishing o), tan(@ = ( r/Q1) = -l/M; as 1/Q1+ 0, (vanishing q), tan(@ = 0, independently of Qo. The phase behavior at R, = d(l + l/r) is also of interest, as it is very close to the frequency R,, where Rsis a maximum, and is relatively easy to measure. For BVDS, tan(@ = [-
Qo
(r + 1) d(r(r + I))] / [(r + 1) (r + QoQ1) + Ql2] c 0,
so q < 0 at the parallel resonance frequency, R, = d(1 + l/r) also. Phase slope is negative at R,, so R, occurs before R,.
PASSAGE TO LOSSY BVD FORM FROM LOSSY TANGENT FORM The original BVD circuits included resistances to represent losses due to internal friction in the resonator material, as well as mounting losses and damping arising from air-loading, etc. The practical utility of the lossy BVD forms has been established over many years. Exact expressions for input immittance of a single simple thickness plate mode excited by TE and LE are: Yin (TE) = joC, / [1 - k2 T(0/2)], and Yin (LE) = joC, [ l + k2 T(0/2)], where T(0/2) = tan (0/2)/(0/2), and 0/2 = (n/2)*R = (n/2).(Uf0) = oh/v. The appearance of tangent functions represents the acoustic wave propagation aspects of the resonators. These trigonometric functions have poles located at regularly spaced frequency locations. In the absence of loss the immittance poles are located on the real frequency axis. The pole locations are analytically continued to complex frequencies by the inclusion
382
Dielectric Materials and Devices
of loss, and expansions about these complex poles yield lumped networks incorporating resistances in the traditional BVD form [40]. The BVD4 circuits for TE and LE (with and without the - CO)may be derived from partial-fractions expansions of the tangent functions in the exact expressions for Yi" given above; see Fig. 22. CERAMIC STANDARDS The IEEE ceramic standard [20] was published in July 1961, and has provided relations that are still used today for reducing the results of measurements to BVD circuit parameters. The exact solution for the piezoelectric plate problem, however, only became available in 1963 [41]. It is not surprising, therefore, that some of the relations contained in the standard are not entirely correct. This fact, plus an incomplete understanding of the circuit differences between TE and LE, discussed above, lead to difficulties in evaluating piezoceramic coefficients, such as coupling factor. For example, equation (15) in Ref. 20 is:
where Af = (f,, - fs). The quantity lZ,l is the minimum magnitude of impedance; see Fig. 23. This is only good if Qo >> 1, and BVDS becomes BVD4. In addition, f,, is associated with a lossless vibrator; the approximation additionally assumes that lZ,l = RI.Similar remarks apply to equation (13), which is for the TE case, and reads k: = (nfS/2f,,)tan(nAf/2fP). Rather than discuss this standard here in detail, it is simply suggested that the relations given therein be examined carefully in the light of subsequent developments, particularly with respect to the use of LE formulas to describe the TE situation. Because of the high coupling factors of piezoceramics, the difference can lead to large errors. FUTURE MEASUREMENT PROTOCOLS In all the foregoing electrical measuring techniques, the data to be analyzed consist of complex immittance and the real (sinusoidal) frequencies at which they were obtained. From these data, the elements of the equivalent electrical circuit are derived. A novel method has recently been developed where the concept of frequency is generalized to encompass complex (non-sinusoidal) drivingpoint frequencies [42]. Software-programmable network analyzers now are
Dielectric Materials and Devices
383
able to produce temporal waveforms that are non-sinusoidal, so complex excitation functions are available to tailor the source waveform to the natural circuit response, thereby to enhance the accuracy with which low Q materials can be characterized. The method is able to be implemented directly in practical situations. CONCLUSIONS Using the simple thickness modes of electroceramic plates as an illustrative application, various equivalent circuits are considered. These include the traditional BVD, the BVD augmented by a shunt resistor, and the modifications to these necessitated by the distinction between lateral and thickness excitation. Also considered are interpretative sources of error such as where various loss mechanisms are to be incorporated into the equivalent circuits, and use of inappropriate formulas found in outmoded standards documents. REFERENCES AND REMARKS ‘S. Butterworth, “On a null method of testing vibration galvanometers,” Proc. Phys. Soc. (London), vol. 26, pp. 264-273,1914. 2S. Butterworth, “On electrically-maintained vibrations,’’ Proc. Phys. Soc. (London), vol. 27, pp. 410-424,1915. 3K. S. Van Dyke, “The electric network equivalent of a piezo-electric resonator,” (abstract), Phys. Rev., vol. 25, no. 6, p. 895, June 1925. 4K. S. Van Dyke, “The piezo-electric resonator and its equivalent network,” Proc. IRE, vol. 16, pp. 742-764,1928. 5 W. G. Cady, Piezoelectricity, McGraw-Hill, New York, 1946; Dover, New York, 1964. Chapter 14, “The electrical equivalent of the piezo resonator,’’ Table 23, p.355. Table of critical frequencies. 6G. E. Martin, “Determination of equivalent-circuit constants of piezoelectric resonators of moderately low Q by absolute-admittance measurements,” J. Acoust. Soc. Amer., vol. 26, no. 3, pp. 413-420, May 1954. ’A. R. von Hippel, Dielectrics and Waves, (1954), MIT Press, Cambridge, MA, 1966. *R. M. Glaister, “Measurement of coupling coefficient and Q of low-Q piezoelectric ceramics,” British J. Appl. Phys., vol.11, pp. 390-391, August 1960. 9R. Bechmann and A. Ballato, “Parameters of a piezoelectric crystal,” Proc. IRE, vol. 50, no. 12, pp. 2496-2497, December 1962. “D. A. Berlincourt, D. R. Curran, and H. Jaffe, “Piezoelectric and piezomagnetic materials and their function in transducers,” Vol. 1A, Chap.
384
Dielectric Materials and Devices
3, pp. 169-270, in Physical Acoustics: Principles and Methods, W. P. Mason, Ed., Academic Press, New York, 1964. l1A. Ballato, “Resonance in piezoelectric vibrators,” Proc. IEEE, vol. 58, no. 1, pp. 149-151, January 1970. ”G. E. Martin, “Dielectric, elastic, and piezoelectric losses in piezoelectric materials,” IEEE Ultrasonics Symp. Proc., pp. 613-617, Milwaukee, WI, November 1974. 13A. Ballato, “Doubly rotated thickness mode plate vibrators,” in Physical Acousiics: Principles and Methods, W. P. Mason and R. N. Thurston, Eds., vol. 13, chap. 5, pp. 115-181, 1977. Academic Press, New York. ISBN: 0-12477913-1. I4A. Ballato, “Frequency-temperature-load capacitance behavior of resonators for TCXO application,” IEEE Trans. Sonics Ultrason., vol. SU-25, no. 4, pp. 185-191, July 1978. ”A. Ballato, “Piezoelectric resonators,” Chap. 3, pp. 66-122 and pp. 432436, in Design of Crystal and Other Harmonic Oscillators, by B. Parzen, Wiley, New York, 1983. ISBN: 0-471-08819-6. 16A. Ballato and J. Ballato, “Accurate electrical measurements of modern ferroelectrics,” Ferroelectrics, vol. 182, nos. 1-4, pp. 29-59,1996. ”A. Ballato and J. Ballato, “High frequency piezoceramic equivalent circuit,” J. Amer. Ceramic Soc., vol. 79, no. 5, pp. 1413-1415, May 1996. ‘8‘cIRE standards on piezoelectric crystals - the piezoelectric vibrator: definitions and methods of measurements, 1957,” Proc. IRE, vol. 45, no. 3, pp. 353-358, March 1957. 19“IRE standards on piezoelectric crystals: determination of the elastic, piezoelectric, and dielectric constants - the electromechanical coupling factor, 1958,” Proc. IRE, vol. 46, no. 4, pp. 764-778, April 1958. (IEEE Standard no. 178). 20“IRE standards on piezoelectric crystals: measurements of piezoelectric ceramics, 1961,” Proc. IRE, vol. 49, no. 7, pp. 1161-1169, July 1961. (IEEE Standard no. 179). This standards document is reprinted as an appendix in Ref. 25. ”“Standard definitions and methods of measurement for piezoelectric vibrators,” IEEE Standard no. 177, IEEE, New York, May 1966. 22“IEEE Standard on piezoelectricity,” IEEE Standard 176-1978, IEEE, New York. Reprinted in IEEE Trans. Sonics Ultrason., vol. SU-31, no. 2, Part 2, 55pp., March 1984. 23‘CIEEEStandard on piezoelectricity,” IEEE Standard 176-1987, IEEE, New York. 24E. Hafner, “The piezoelectric crystal unit - Definitions and methods of measurement,” Proc. IEEE, vol. 57,110.2,pp. 179-201, February 1969.
Dielectric Materials and Devices
3 85
25B. Jaffe, W. R. Cook, Jr. and H. Jaffe, Piezoelectric Ceramics, Academic Press, New York, 1971. ISBN 0-12-379550-8. p.219 lists QMvalues (= Q1) for Pbl-, Ba, Nb2 0 6 of 2 to 1000, and values of l/(loss tangent) of 25 to 100. 26F. K. Priebe and A. Ballato, “Measurement of mode parameters by sweep frequency methods in the frequency range from 20 to 250 MHz,” Proc. 20fhAnnual Frequency Contol Symp., pp. 465-499, Atlantic City, NJ, April 1966. 27Ref.25, p.31, indicates that the effects of R1 (presumably on the resonant and antiresonant frequencies, etc.) may be disregarded for strong modes (k > 0.3) and reasonably large mechanical Q = QM> 50. 28J.J. Kyame, “Conductivity and viscosity effects on wave propagation in piezoelectric crystals,” J. Acoust. Soc. Amer., vol. 26, No. 6, November 1954, pp. 990-993. 29A. R. Hutson and D. L. White, “Elastic wave propagation in piezoelectric semiconductors,’’ J. Appl. Phys., vol. 33, no. 1, pp. 40-47,1962. 30G. Arlt, “Resonance-antiresonance of conducting piezoelectric resonators,” J. Acoust. Soc. Amer., vol. 37, no. 1, pp. 151-157, January 1965. 31R. Holland, “Representation of dielectric, elastic, and piezoelectric losses by complex coefficients,” IEEE Trans. Sonics and Ultrasonics, vol. SU-14, no. 1, pp. 18-20, January 1967. 32T. Ikeda, Fundamentals of Piezoelectricity, Oxford University Press, Oxford, 1990. ISBN: 0-19-856339-5. 33A. K. Jonscher, Dielectric Relaxation in Solids, Chelsea Dielectrics Press, London, 1983. ISBN: 0-9508711-0-9. 34A. K. Jonscher, Universal Relaxation Law, Chelsea Dielectrics Press, London, 1996. ISBN: 0-9508711-2-5. 35J.Lamb and J. Richter, “Anisotropic acoustic attenuation with new measurements for quartz at room temperatures,” Proc. Roy. Soc. (London), vol. A293, pp. 479-492,1966. 36Adapted from H. Ohanian, Physics, 2”d expanded ed., W. W. Norton, New York, 1989; ISBN: 0-393-95750-0. 37Ref.25, p.192, lists the specific resistivity of K1/2Nal/2Nb03as 10” SZ-m, its electrical Q (= Qo) as f 20 at 1 kHz (that of another sample as 50 at 100 kHz), and its mechanical Q (= Q1) as 130 to 240. 38G. H. Haertling, “Rainbow ceramics - A new type of ultra-highdisplacement actuator,” American Ceramic Society Bulletin, vol. 73, No. 1, January 1994, pp. 93-96. 39B. D. Cullity, Elements of X-Ray Diffraction, 2”d ed., Addison-Wesley, Menlo Park, NJ, 1978, pp. 277-279. 40The Mittag-Leffler Theorem describes the pole expansion of analytic functions having only well-separated poles as singularities (meromorphic
-
386
-
-
Dielectric Materials and Devices
functions). It states that the series in the complex variable z, f(z) = f(0) + Cn bn converges to f(z), where the poles are at finite z = Zn, and ((z - Zn)- + z,,have residues bn. The quantity z is proportional to complex frequency, and f(z) is related to an immittance function. 4’H. F. Tiersten, “Thickness vibrations of piezoelectric plates,” J. Acoust. Soc. Amer., vol. 35, no. 1, pp. 53-58, January 1963. 42R. A. Pastore, Jr., “A new characterization technique for lossy piezoceramic resonators,’’ PhD Dissertation, Stevens Institute of Technology, Hoboken, NJ,May 2000.
’
’>
Dielectric Materials and Devices
387
CO
0
Q
% b
RS
GP
*S
C
Figure 1
388
Dielectric Materials and Devices
fa
Figure 2
Dielectric Materials and Devices
389
-
FREQUENCY
/
Figure 3
390
Dielectric Materials and Devices
-
1.1
I
.o -
0.9-
0.8 -
1
I
I
0.7 -
I
I
0.6 w
I I
I
-
U
20.4
I
s k0.3 z a
I
I
\
\
0.2
0.1
0
-0.I
-0.2
I
-0.3 0 Figure 4
1
.2
I
.4
I
.6
I
.8
Dielectric Materials and Devices
I
1.0
I
1.2
I
I
I
I
I
I
I
I
1
1.4 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 FREQUENCY SZ
391
5
OHMS
4
r = 2 ; R,=IOHM; Q=6
3 2
I '
T O -I
I1
I 1
I 1
8
I 1
-
10 RS
I
I I
12
I
OHMS
-2 -3 IMPEDANCE CIRCLE
-4 -5
Figure 5
392
Dielectric Materials and Devices
I.o
0.8
t
a
o.6
m
0.4
0.2
Dielectric Materials and Devices
393
for
Figure 7
394
Dielectric Materials and Devices
Figure 8
Dielectric Materials and Devices
395
1.o
0.8
0.4
0.2
0
2
4
6
r
8
10
12
14
Figure 9
396
Dielectric Materials and Devices
0.e
0.6
ktrue
0.4
0.2
0
0
0.2
0.4
0.6
0.8
1
k apparent
Figure 10
Dielectric Materials and Devices
397
7
c 1
0
-I
0.5 PA R A LLEL CA PACI TAN C E R , - 0 : LOSSLESS
Figure 11
398
Dielectric Materials and Devices
*[ SIEMENS = MHOS
A DMITTA NC E
Figure 12
Dielectric Materials and Devices
399
120
I
I
I
I
I
I
I
r=2 80
I yl 40
-
0
0.8
0.9
1
1.I
c2 Figure 13
400
Dielectric Materials and Devices
1.2
300
200
100
0 0.8
0.9
I
I.I
I.2
Q Figure 14
Dielectric Materials and Devices
40 1
1
gP
0.5
0
0
1
3
4
Figure 15
402
Dielectric Materials and Devices
2.0 'balance
CO
+2 0
2c0
-2
0 I .5
I 'I
balance
t 1.0
HYBRID BRIDGE ADMITTANCE
r=-2
Q=6 RI= lohm
0.5
0.0
0.5
I.o
I
I .5
1
2.0
Q Figure 16
Dielectric Materials and Devices
403
goo
70° 50 °
30°
Q
I oo O
- loo -30' -5OO
- 70°
- goo
0
-2
.4
.6
.8
1.0
1.2 1.4 1.6 FREQUENCY
a
1.8
2.0 2.2
2.4
Figure 17
404
Dielectric Materials and Devices
1ooo
50°
(Pu
O0
-5OO
-1ooo
0.1
1 .o
10
100
1000
Q Figure 18
Dielectric Materials and Devices
405
1.5
1
1
1
r=l 1.0
Q"
0.5
0.0
0.1
1.0
10
100
1000
Q
Figure 19
406
Dielectric Materials and Devices
goo
60’
30°
8
0
-30’
-60’
-90’
Figure 20
Dielectric Materials and Devices
407
500
1
I
I
I
I
I
I
I
-
0
-
n
,4 S
rn
-
-500
-1000
-
-1500
-
-2000
I
0
Q~= 103 = Q,
r=10 I
0.4
I
I
0.8
I
I
I.2
I
I
1.6
Figure 21
408
Dielectric Materials and Devices
yin = + j Y, tan
(2)
1 1 1
0
.
1 111) 1 1 1
1 L= 4 Y, ( v / I )
V
Figure 22 Dielectric Materials and Devices
409
8-
OHMS
6-
-
4-
-
t
2-
-
0
II
I I
-
-2
I
I
i
s1
-4-
-
-6-
IMPEDANCE MAGNITUDE
121' = R,
2
+ X,
2
-8-
Figure 23
410
Dielectric Materials and Devices
EFFECT OF RARE EARTH ADDITION ON MICROSTRUCTURE AND ELECTRICAL PROPERTIES IN BaTi0,-BASED CERAMICS FOR NiMLCC Hirokazu Chazono and Hiroshi Kishi Material Development Department, Taiyo Yuden Co., Ltd. 5607-2 Nakamuroda Haruna-machi Gunma-gun Gunma 370-3347, Japan
ABSTRACT The microstructural evolution of the sample containing dysprosium (Dy) and holmium (Ho) was investigated by changing the firing temperature in BaTi03 (BT)-MgO-R,O, (R=Dy or Ho) system. The mean grain diameter and the grain size distribution determined by measuring the grains for the chip surface of the multilayer ceramic capacitor with nickel internal electrode were considerably dependent on the firing temperature for the sample containing Dy, while they were slightly dependent on the firing temperature for the sample containing Ho. The sample containing Dy fired at 1240°C and the sample containing Ho fired at 1300°C showed almost the same mean grain diameter and the grain size distribution. In addition, it was found that they had the same size of core region, indicating that both samples were composed of the similar core-shell microstructure, which was determined by the differential scanning calorimetry. The temperature characteristics (TC) of the dielectric constant and the aging characteristics under direct current (dc) bias field of 2Vipm were measured using these two chip samples. It was found that the electrical properties, such as TC and the aging characteristics under dc-bias field, were almost the same when the samples were composed of the same microstructure, even though the samples contained the different rare earth oxide.
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee pailto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
41 1
INTRODUCTION BaTiO, (BT) is a most convenient and prevalent material for dielectrics. Many additives are known to depress the temperature dependence of the dielectric constant (E) for BT, yielding materials for multilayer ceramic capacitors (MLCCs) which meet the specification of X7R in the Electronic Industries Association standard (EIA). Microstructures observed in these materials with flat temperature characteristics (TC) of
E
can be characterized by small grains and core-shell
microstructure.'-' The core is composed of pure BT and the shell is produced by the reaction of the additives with BT. Kahn' reported that two TC peaks were observed in BT systems modified with niobium with highly suppressed grain growth. The peak at the high temperature was the Curie temperature and remained constant in appropriate firing temperature. The broad peak at low temperature was due to a quasi-stable state of niobium distribution. Pathumarak et al. suggested that the flat TC of E was explained by the superposition of many separate E-T curves for this region Therefore, the TC of with a composition gradient having slightly different Curie temperature (E). ceramics should be dependent on the corehhell volume ratio; larger core volume yields a higher peak at Tc and larger shell yields larger E in the temperature range below Tc, since E is the sum of the contribution of core and shell. The base metal, nickel (Ni), has been used preferably to palladium (Pd) for the internal electrode especially for MLCC with large capacitance. Although Ni is an excellent candidate for the internal electrode of MLCC in view of the reduction of production costs, the dielectrics must be fired in a reducing atmosphere. In general, the dielectric materials suffer from the low resistivity and poor reliability when fired in such a reducing atmosphere. Saito et al.' reported that a high reliable MLCC conforming to X7R specification of EIA was developed by the addition of rare earth oxide. In addition, Okino et
showed that the life at highly accelerated life test
(HALT) was considerably improved by adding the rare earth oxide having small ionic radius, such as dysprosium (Dy), holmium (Ho), and erbium (Er). The incorporation of the rare earth ions into BT lattice depends on their ionic radius as pointed out experimentally by Takada et al.' and using computer simulation by Lewis et al. I0. Kishi et al. reported that Dy and Ho dissolved into both Ba and Ti sites of BT perovskite lattice from the powder X-ray diffraction analysis at 300°C for the system (Bal-2xRzx)(Ti,,Mgx)03, where R indicated Dy and Ho. On the other hand, Okino et all2 reported that the microstructural evolution and electrical properties including TC, aging characteristics under direct current (dc) bias, and life at HALT,
412
Dielectric Materials and Devices
were greatly influenced by the composition in BT-MgO-R203system (R=Sm and Ho). Chazono et
al.l 3 reported that microstructural evolution for the sample containing Sm was more dependent on the firing temperature than that for the sample containing Ho in the similar BT-MgO-R,O, system. Moreover, the electrical properties such as TC and aging characteristics under dc-bias was affected by the microstructural change. In addition, Mizuno et
all4
reported that the core-shell
microstructure of the sample containing Ho was more stable than that of the smaple containing Dy against the firing temperature, and that the electrical properties were sensitive to the microstructural evolution. Many studies indicating that there is a strong relationship between the microstructure and the electrical properties have been reported as mentioned above. However, it is very important to understand the effect of the rare earth addition by comparing the samples having the same microstructure. Therefore, the aims of this work are to prepare the samples with the same microstructure in the system BT-MgO-R,O, system (R=Dy and Ho) first, and to compare the electrical properties between them next.
EXPERIMENTAL PROCEDURE Sample Preparation The main starting material was BaTiO, (BT) with an average particle size of about 0.35pm synthesized hydrothermally (Sakai Chemical Industry Co., Ltd.). Reagent grade MgO, MnO, rare earth oxide (Dy,O, and Ho203),and BaSiO, were weighed and mixed to BT in each composition given in Table I. They were mixed and then dried. These powders with an organic binder system were cast into green sheets. Nickel (Ni) internal electrode was printed on green sheets of about 5pm. 21 green sheets with Ni internal electrodes as well as the protective sheets at the upper and lower sides were stacked and pressed into a bar of about 800pm thickness and then cut into small pieces. Terminal Ni electrodes were formed on both sides of the chips. The chips were fired at various temperatures and cooled to 1000°C in a reducing atmosphere controlled by H,,
N,,O,, and
H,O, then to room temperature in an weakly oxidizing atmosphere (Po2=20Paat 1000°C). Table I. Sample Composition (atomic%b
BT-Sm BT-HO
BT
MgO
Dy03,
100 100
1.o 1.o
1.5
Dielectric Materials and Devices
HoO,, 1.5
MnO 0.1 0.1
BaSiO, 1.5 1.5
413
Characterization The microstructure was observed by the field emission scanning electron microscopy (FESEM; Hitachi, Japan). The grain size of the chip surface was measured by the intercept method with a micrometer. The mean grain size is defined as the grain diameter giving 50% of the accumulated volume, which is determined by the summation of each grain volume on the assumption that the each grain is a sphere. The phase transition of the samples was characterized by differential scanning calorimetry (DSC; Macscience, Japan). Electrical Properties The temperature characteristics (TC) of the dielectric constant was measured at lkHz,
1.OVrms using an LCR meter (HP-4284A; YHP,Japan), covering the temperature range from -55" to 140°C. The dielectric constant of the chips was calculated geometrically by the thickness and the crossing area of the active layers. The aging of the capacitance under dc-bias field was measured as follows. The initial capacitance was measured at room temperature 24h after heat treatment at 150°C-lh. 2V/pm dc-bias field was applied to MLCCs for desired periods. The capacitance was measured after the dc-bias removal. The capacitance change was measured using the same sample in a sequence.
RESULTS AND DISCUSSION Ni-MLCC samples were fired at various temperatures. The surfaces of samples were observed with FE-SEM. Figure 1 shows SEM micrographs of the free surfaces of the sample containing Dy fired at 1220", 1240", 1260", and 1280°C and of the sample containing Ho fired at 1260", 1280", 1300", and 1320°C. It was found that the rate of grain growth for the sample containing Dy was qualitatively larger than that for the sample containing Ho. There observed a secondary phase in the sample containing Dy fired at 1280"C, although it was not identified. Using these photographs, the grain size was measured for more than 300 grains by the intercept method with a micrometer. Figure 3 shows the accumulated volume indicating the grain size distribution. The curve of the accumulated volume was changed as the firing temperature changed for the sample containing Dy. However, it was almost constant for the sample containing Ho, although the firing temperature was higher for the latter than for the former. Figure 4 shows the mean grain size as a parameter of the firing temperature. The mean grain size increased linearly as the firing temperature increased for the sample containing Ho. However, it increased exponentially for the sample containing Dy.
414
Dielectric Materials and Devices
Fig.1 SEM micrographs for the chip free surface fired at various temperatures for the sample containing Dy and Ho (bar = 2pm).
Dielectric Materials and Devices
415
Fig.2 The accumulated volume indicating the grain size distribution.
I
1200
I
,
1
I
0 sample containing Dy 0sample containing Ho , , I , , I
1250 1300 temp. (“C)
1350
Fig.4 Definition of m value.
Fig.3 Mean grain size as a parameter of the firing tem peratu re.
On the other hand, the slope of the curve in Fig.2 gives the grain size distribution. Therefore, the slope, m value, of the accumulation volume curve was defined as illustrated schematically in Fig.4 and the following equation (1). m = l/log(d,/d,)
416
Dielectric Materials and Devices
Figure 5 shows m value as a parameter of the firing temperature. The increase in the firing temperature brought about the decrease in m value. Similarly to the mean grain size, the dependence of m value on the firing temperature was larger for the sample containing Dy than for the sample containing Ho.
m 1
- 0 sample containing Dy - 0sample containing Ho
Kishi et al." reported that Dy and Ho were incorporated into both Ba- and Ti-sites in BT (R=Dy and Ho). perovskite lattice in the study of solid solution system, (Ba,-2xR2x)(Ti,,Mgx)09xMgx)03 Moreover, they reported that the substitution ratio of Dy into Ba-site in BT lattice was larger than that of Ho, indicating that Dy behaved more likely as a donor dopant than Ho. It can be speculated that the rate of diffusion of the ion in Ti-site is slower than that of the ion in Ba-site. Therefore, the larger dependence of the mean grain size and m value, seen in Figs.3 and 5, on the firing temperature can be ascribed to the larger activity of Dy than that of Ho in the shell phase. Judging from the mean grain size in Fig.3 and m value in Fig.5, the sample containing Dy fired at 1240°C and the sample containing Ho fired at 1300°Cwas chosen for further investigation. These two samples were abbreviated as D124 and H130, respectively, hereafter. The mean grain size and m value for the samples D124 and H130 were the same. Therefore, it was supposed that these two samples were composed of the same microstructure. However, it can not be conclude that the samples D 124 and H 130 have the same core-shell microstructure yet. Information only on the core region can be obtained by DSC measurement if the grains are composed of the core-shell microstructure, since a core is composed of pure BT, which has the latent heat at Tc. Figure 6 shows the DSC profiles for the D124 and H130. A peak at around 397K was observed for both
Dielectric Materials and Devices
417
samples, and the broadness of the peak was the same for both samples. The temperatures of these endothermic peaks are very close to that of Tc for pure BT, which indicates that there is a core region composed of pure BT in both D124 and H130. Accordingly, it was found that the samples D124 and H130 are composed of the same microstructure including the core-shell microstructure in a grain. enthLalPY (a.u.)
exo: endo 50
100
150
temp. (“C)
Fig.6 DSC profiles for D124 and H130. Electrical properties were measured for the samples D124 and H130. Figure 7 shows the TC of
E
for the samples D124 and H130. Figure 8 shows the aging characteristics measured at the
room temperature under a dc-bias field of 2V/pm. A slight difference in
E
between the two
samples was in the error range due to the geometrical calculation. It was found that both samples, D124 and H130, showed the same TC of E and the aging characteristics. It was consistent with the
4500
5000 4500
E
2
E 4000
3500
3500
-60
-20
20
60
100
temperature (“C)
Fig.7 TC of E for D124 and H130.
418
140
time (hr)
Fig.8 Aging characteristics under a dcbias field of 2V/pm for D124 and H130.
Dielectric Materials and Devices
result of the aging characteristics obtained by Mizuno et a1.,14 although the composition was different between them and this work. It was noteworthy that the electrical properties, such as TC
of E and the aging characteristics, were the same when the samples were composed of the same microstructure. It is suggested that the more elaborate design and control of the individual grains must be essential for the further decrease of the active layer thickness in Ni-MLCC. The other electrical properties, such as degradation behavior and the leakage current at HALT, must be investigated in a future work. REFERENCES
‘M. Kahn, “Influence of Grain Growth on Dielectric Properties of Nb-Doped BaTiO,,” Journal of the American Ceramic Society, 54 [9] 455-57 (1971).
2B. S. Rawall, M. Kahn, and W. R. Buessem, “Grain Core-Grain Shell Structure in Barium Titanate-Based Dielectrics”; pp. 172-88 in Advances in Ceramics, Vol. 1, Grain Boundary Phenomena in Electronic Ceramics. Edited by L. M. Levinson. American Ceramic Society,
Westerville, OH, 1981. 3T. R. Armstrong, K. A. Young, and R. C. Buchanan, “Dielectric Properties of Fluxed Barium Titanate Ceramics with Zirconia Additions,” Journal of the American Ceramic Society, 73 [3] 700706 (1990). 4D. Hennings and B. S. Schreinmacher, “Temperature-Stable Dielectrics Based on Chemically Inhomogeneous BaTiO,,” Journal of the American Ceramic Society, 73 [ 121 3562-68 (1990). 5H. Saito, H. Chazono, H. Kishi, and N. Yamaoka, “X7R Multilayer Ceramic Capacitors with Nickel Electrodes,” Japanese Journal ofApplied Physics, 30 [9B, Part I] 2307-10 (1991).
%. Pathumarak, M. Al-Khafaji, and W. E. Lee, “Microstructural Development on Firing NbzOsand Bi203Doped BaTiO,,” British Ceramic Transactions,93 [3] 114-18 (1994).
7H.Chazono and H. Kishi, “Sintering Characteristics in the BaTi0,-Nb,O,-Co,O, Ternary System: 11, Stability of So-called “Core-Shell” Structure,” Journal of the American Ceramics Society, 83 [l] 101-106 (2000). ‘Y. Okino, H. Shizuno, S. Kusumi, and H. Kishi, “Dielectric Properties of Rare-Earth-OxideDoped BaTi0, Ceramics Fired in Reducing Atmosphere,” Japanese Journal of Applied Physics, 33 [9B, Part I] 5393-96 (1994).
Dielectric Materials and Devices
419
’K. Takada, E. Chang, and D. M. Smith, “Rare Earth Additions to BaTiO,,”; pp.147-152 in Advances in Ceramics Vol. 1, Edited by J. B. Blulm and W. R. Cannon, The American Ceramic Society, 1985.
‘OG.V. Lewis and C. R. A. Catlow, “Defect Studies of Doped and Undoped Barium Titanate Using Computer Simulation Techniques,” Journal of Physics and Chemistry of Solids, 47 [ 11 8997 (1986).
“H. Kishi, N. Kohzu, Y. Okino, Y. Takahashi, Y. Iguchi, H. Ohsato, K. Watanabe, J. Sugino, and T. Okuda, “Effect of Rare-Earth Oxides on Formation of Core-Shell Structures in BaTiO,,”; pp.33-40 in Ceramic Transactions, Vol. 100, Edited by K. M. Nair and A. S. Bhala, The American Ceramic Society, 1999.
12Y. Okino, N. Kohzu, Y. Mizuno, M. Honda, H. Chazono, and H. Kishi, “Effects of the Microstructure on Dielectric Properties for BaTi0,-Based MLC with Ni Electrode,” ;pp9-15 in CSJ Series - Publications of the Ceramic Society of Japan, Vol.1, Electroceramics in Japan 11, Edited by N.Mizutani, K. Shinozaki, N. Kamehara and T. Kimura, Trans Thech Publications Ltd., Switzerland, 1997. I3H. Chazono, Y. Okino, N. Kohzu, and H. Kishi, “Effect of Sm and Ho Addition on the Microstructure and Electrical Properties in MLCC with Ni Internal Electrode,”; pp.53-64 in Ceramic Transactions, Vol. 97, Edited by Jau-Ho Jean, T. K. Gupta, K. M. Nair, and K. Niwa, The American Ceramic Society, 1999.
I4Y. Mizuno, Y. Okino, N. Kohzu, H. Chazono, and H. Kishi, “Influence of the Microstructure Evolution on Electrical Properties of Multilayer Capacitor with Ni Electrode,” Japanese Journal Applied Physics, 37 [9B, Part I] 5227-31 (1998).
420
Dielectric Materials and Devices
CONDUCTIVITY AND MODULUS SPECTRA FOR A SERIES OF LITHIUM -BORATE AND SODIUM TRISILICATE GLASSES Andrew E. Bums Department of Chemistry Kent State University-Stark Campus 6000 Frank Avenue Canton, OH 44720-7599 ABSTRACT Using recently developed modulus scaling and conductivity scaling analyses, dielectric data up to 2 GHz from a series of lithium-borate and sodium trisilicate ionic conducting glasses have been analyzed. The results indicate that one can plot the dielectric data in a variety of formats to show the transport model in which one believes. INTRODUCTION Ion transport in glasses is a well-studied discipline, and one, which is continuously evolving. Typically, one synthesizes the glasses to be studied, and then performs an experiment to obtain the conductivity or dielectric information. Usually, this is done with a low frequency dielectric bridge, but recently, new high frequency bridges and microwave techniques have allowed for dielectric data to be obtained into the GHz range of frequency [ 1-41. The dielectric data is then analyzed in terms of the dielectric constant, conductivity, or Modulus representation. Depending on the format used to investigate the data, results and interpretations can be different and contradictory. The Modulus representation for dielectrics was developed by Macedo et a1 [ 5 ]to reduce the DC conductivity (ode) effects at low frequency. Further analysis showed that the peaks in the temperature dependence of the imaginary part of the modulus (M") data produced the same activation energy as from conductivity results [ 5 ] . Normalization of the modulus data, done using the frequency and dividing it by the peak frequency, showed that all the temperature M" plots were superimposable. This implies that the ion transport dynamics, regardless of frequency, have the same thermal activation energy [6].
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
42 1
Roling and coworkers [7] recently developed a scaling analysis of conductivity spectra without using any arbitrary parameters. Their model for a series of compositionally related glasses through the mole ratio (x), involves normalizing the conductivity (0) by the DC conductivity (qc) and the frequency (f) by odcT, which takes into account the temperature to produce a scaling law represented by
They then showed how the DC conductivity could be replaced by the carrier mobility showing that the ion dynamics can be described by a universal function, independent of composition and temperature. This new scaling law predicts the same proportion of ions is mobile in a glass independent of composition, in agreement with the strong electrolyte model. Funke and Wilmer [8] recently reported a new model, involving the concept of mismatch and relaxation (CMR). The model explains how the conductivity dispersion is related to the rates of relaxation. The conductivity is given by
Where F is a function such that coo is defined as the angular frequency of onset such that o(o)/odc= 1.303 at co = coo. This model claims to explain the shape of the conductivity dispersion. EXPERIMENTAL All the glasses analyzed in this paper were homogeneous and x-ray amorphous and whose standard melt-quench preparation has been described earlier [ 13. Glasses of composition Na20 + 3sio2 and x Li20 + (1-x) B203, 0.19<x<0.35 were synthesized and studied. Three separate dielectric techniques were used to obtain dielectric data up to 2 GHz. A low frequency bridge technique was used for frequencies up to 2 10 kHz. The other two methods used were a high frequency bridge and time domain reflectance system, with frequency windows from 500 kHz to 100 MHz and 875 MHz to 2 GHz, respectively. Both of these techniques have been reported in detail elsewhere [ 1,2]. RESULTS AND DISCUSSION The results for the modulus and normalized modulus spectra for a lithiumfluoro-borate glass are shown in figures 1-3. These results are typical for any single composition glass with a variation in temperature. Pate1 and Martin [6]
422
Dielectric Materials and Devices
0.14
*M'136
2
I
0.07
0.00
2
4
8
6
10
log f
Fig 1. The real M' part of the electric modulus plotted against frequency for 0.35 Li20 + 0.65 B203.
0.04
-
c
-1
. 0.02
A
0
M" 136C M" 168C M" 21 1C M" 268C
2
6
log f
10
Fig 2. The imaginary M" part of the electric modulus plotted against frequency for 0.35 Li20 + 0.65 B203.
Dielectric Materials and Devices
423
were the first to observe this effect for a series of sodium thioborate glasses, attributing this to the idea that all dynamic processes occurring at different frequencies exhibit the same thermal activation energy. In addition, they showed that for different amounts of alkali ion, the shape and width of the modulus curve changes, due to the relaxation mechanism depends on the total ion concentration and thus these glasses behave as strong electrolytes. The spectral results for the lithium-borate glass and other glasses agree with their results.
*
X
M" 136C M" 168C A M" 21 1C M" 268C * M'l36 M168
M' M"
-5
+c
+? + :
Y
*
*a
5
0
log (fieq/peak fi-eq)
Fig 3. The real (M') and imaginary (M") part of the electric modulus plotted against the normalized frequency for 0.35 Li20 + 0.65 B2O3. Figure 4 shows the lithium borate series normalized modulus spectra for all three glasses at the same temperature, 268°C. In these spectra, the different compositions do not overlap although their shape is nearly identical. Roling and coworkers [7] claim that the width and shape of the normalized M" curves do change with composition as the spectra become narrower with decreasing alkali concentration, although this does not appear in these spectra. Sidebottom et al. [9] believe the modulus representation is misleading in that an incorrect representation of the relaxation mechanism is used. Roling and coworkers.
424
Dielectric Materials and Devices
log(fireq/peak fireq) Fig. 4. The normalized M" spectra for x Li20 + (1 -x) B2O3 at 268OC.
Fig 5. The normalized MI1/M"max spectra for x Li20 + (1-x) B2O3 at 268OC.
Dielectric Materials and Devices
425
believe the reason the modulus formalism is incorrect is that the mobile ion number density is not used at all [7,10,11]. The M" peak used in normalization scales with the DC conductivity but not with the mobility. However, when the modulus peak is normalized to unity by the maximum in peak height, the spectra do overlap, as shown in figure 5. The conductivity scaling spectrum using eqn 1 is shown in Fig 6 for the lithium-borate glasses at the same temperature up to 2 GHz. This spectrum agrees with this "master" scaling within scatter from the experimental techniques. This "master" scaling has been observed previously by Roling and others [7,12]. As with the modulus, it also suggests a temperature independent relaxation mechanism. This scaling, unlike modulus normalization, takes into account the mobile ion number density.
Fig. 6 The conductivity scaling curve for x Li20 + (1-x) B203 at 268OC.
426
Dielectric Materials and Devices
Figure 7 shows the data as plotted using the CMR model. Unlike the conductivity and scaling, the data do not seem to overlap. This contradicts the CMR model [S], which implies that the conductivity dispersion is not equivalent to the rates of relaxation via the single and multiple particle routes.
Fig 7. The CMR modeling of the Na2O + 3sio2 glass at various temperatures using equation 2. SUMMARY The results of the sodium trisilicate and lithium-borate series of ionically conducting glasses have been shown to scale with the modulus and conductivity up to 2 GHz. The reason the conductivity scaling agrees with the data is that it incorporates the number density of mobile ions. The scaling is independent of composition thus suggesting that the strong electrolyte model describe ionic transport better than the weak electrolyte model. The modulus normalization of the frequency was investigated and not found to scale the data due to its lack of
Dielectric Materials and Devices
427
incorporating the number density. However, when the imaginary part of the modulus was normalized to the peak height, the modulus formalism does scale the data. The CMR model did not scale the conductivity data. What this suggests, is that one can use either method of plotting and analyzing dielectric data. The question remains as to which method is best and should be used to interpret and explain ionic transport in glasses. REFERENCES 1. A. Burns, G.D. Chryssikos, E. Tombari, R.H. Cole, and W. M. Risen, "Dielectric Spectra of Ionic conducting Oxide Glasses to 2 GHz, "Physics Chem. Glasses, 30 [6] 264-270 (1989). 2. R.H. Cole, J.G. Berberian, S. Mashimo, g. Chryssikos, A. Bums, and E. Tombari, "Time Domain Reflection Methods for Dielectric Measurements to 10 GHz," J. Appl. Physics, 66 [2] 793-802 (1989). 3. C. H. Hsieh and H. Jain, "Are the low temperature-low frequency and high temperature-high frequency ac conductivity of glasses the same phenomenon?" J. Non-Cryst. Solids, 203 293-299 (1996). 4. K. Funke, C. Cramer, B. Roling, T. Saatkamp, D. Wilmer, and M. D. Ingram, Yonic and polaronic glassy conductors: conductivity spectra and implications for ionic hopping in glass," Solid State lonics, 85 293-303 (1996). 5 . P.B. Macedo, C.T. Moynihan, and R. Bose, "The role of Ionic diffusion in Polarization in Vitreous Ionic Conductors," Physics Chem. Glasses, 13 [61 171179 (1972). 6. H.K. Pate1 and S.W. Martin, "Fast Ionic Conduction in Na2S + B& Glasses," Phys. Rev.B, 45 [ 181 10292-10300 (1992). 7. B. Roling, A. Happe, K. Funke, and M.D. Ingram, Carrier concentrations and Relaxation Spectroscopy: New Information from Scaling Properties of Conductivity Spectra in Ionically Conducting Glasses, " Phys Rev Lett, 78 [I 11 2160-2163 (1997). 8. K. Funke and D. Wilmer, "Concept of mismatch and relaxation derived from conductivity spectra of solid electrolytes," Mat. Res.Soc. Symp. Proc, 548 403-414 (1999). 9. D.L. Sidebottom, P.F. Green, and R.K. Brow, "Comparison of KWW and Power Law Analyses of an Ion-conducting Glass," J. Non-Cryst. Solids, 183 151-160 (1995). 10. B. Roling, "What do Electrical Conductivity and Electrical Modulus Spectra Tell Us about the Mechanisms of Ion Transport Processes in Melts, Glasses, and Crystals?" J. Non-Cryst. Solids, 244 34-43 (1999). 11. B. Roling, "Scaling Properties of the Conductivity Spectra of Glasses and Supercooled Melts," Solid State lonics, 105 185-193 (1998).
428
Dielectric Materials and Devices
12. B. Roling, M.D. Ingram, M. Lange, and K. Funke, "Role of AgI for Ionic Conduction in AgI-AgP03 Glasses," Phys. Rev. B, 56 [21] 13619-13622 (1997).
Dielectric Materials and Devices
429
This page intentionally left blank
Composition and Temperature Dependence of Microwave Conductivity of Potassium Germanate Glasses S. Krishnaswami, H. Jain' Department of Materials Science and Engineering Lehigh University, Bethlehem, PA 18015
0. Kanert Institute of Physics, University of Dortmund D-4422 1 Dortmund, Germany
ABSTRACT The dependence of microwave conductivity, GMW, on temperature (T) and alkali concentration has been studied for potassium germanate glass of composition xK20.(l-x)Ge02, 0.00231 x 10.247. From the empirical analysis, it is established that conduction at high frequency (f) and high T in these glasses is related to the conduction at low f and low T. The temperature dependence of GMW shows a weak thermal activation for all compositions as predicted by the model of j ellyfish-like localized fluctuations of atoms in asymmetric double yell potential configurations (ADWPC). The variation of GMW with alkali oxide concentration is however weaker than that observed at low T. It increases only by a factor of 2 for an increase from 0.23 to 25 mol% of K20 at 1GHz. INTRODUCTION There is considerable interest in using glass and glass ceramics in highspeed computers and communication systems operating at microwave frequencies. The replacement of crystalline microwave integrated chip by glass microwave integrated chip is motivated by low loss and conductivity of glass at these frequencies. On the other hand, highly conducting glasses have found applications in microwave sintering, firing and processing of ceramics [ 1,2]. Corresponding author. Email: [email protected]
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
43 I
Among all these efforts, the use of microwave radiation for sintering of ceramics has drawn most attention. However, there exist unsolved problems like the catastrophical phenomenon of thermal runaway hindering the development of this emerging technology. Recently, microwave processing of glasses is also being explored [3,4], but it is unclear how this phenomenon will affect it. More than fifty years ago, Navias and Green [5], von Hippel [6] and Stevels [7] began investigating the mechanism of dielectric loss at microwave frequencies in oxide glasses. Their measurements were limited to one or two frequencies and room temperature. The results of these studies indicated at least five independent sources for microwave loss in oxide glasses [8]: (1) Migration loss that arises from the transport of mobile alkali cations, (2) Vibration loss that arises due to the localized vibrations of the individual cations or other massive structural units, (3) Deformation loss due to small deformation of the glassy network, (4) Electronic hopping and ( 5 ) Interfacial polarization loss due to presence of electrodes. However, these theories are either inconsistent with experimental data or lack satisfactory microscopic description [9]. Thus, in spite of all plausible sources, the microscopic origin of CTMW has remained unknown for glasses. Recently, theories for explaining the mechanism of microwave loss in glasses are emerging based on the better understanding of the models that have been established for low frequency dielectric loss. For instance, Funke and coworkers [ 10,l l ] have advocated that microwave conduction is simply an extension of low frequency single ion hopping (universal dynamic response region, UDR). Any additional contribution is believed to come from high frequency far infrared vibrational conductivity. Microwave conductivity is then the short time observation of the relaxation associated with the hopping of ion. However, it was later established that CTMW cannot be described by the simple addition of low frequency ion hopping and high frequency vibrational conductivity in a silicate glass system [ 121. Furthermore, Funke’s model predicts a plateau at microwave frequencies and beyond [lO], which has not been experimentally observed in oxide glasses. This suggests the presence of an additional mechanism of microwave loss in oxide glasses. The other current model by Jain et al. [13] proposes the origin of microwave loss to be different from UDR conductivity. In this case CTMW arises from localized movements of a group of atoms similar to that of a ‘jell’yjkh’ in the ‘glassy ocean’ [14]. Mathematically, it can be expressed in terms of thermally activated change in its configuration over an asymmetric double well potential (ADWP) barrier [ 15,161. Preliminary studies on lithium silicate glass suggested that the room temperature CJMW has a common origin with low T - low f conductivity [13]. The low temperature ADWP conductivity has been observed in a wide range of disordered materials [ 15,17,18,19,20]. Recent systematic studies on Gd3+ and Y3+ -doped
432
Dielectric Materials and Devices
crystalline Ce02 by Nowick [21] have shown a transition from Debye type dipolar loss to the ADWP type constant loss with increasing concentration of the charge carrier. Thus, the real part of ionic conductivity described by the ADWP model can be expressed by summing up contributions from all ADWP configurations in a unit volume of the amorphous solid. It is expressed by: G(O,T) =
oNe2R2 12kT
I /sec h 2(A / 2kT)$(A, V) 1+ oz
vIllA~
AdV
where, $(A,V) is the distribution function assumed to be of form commonly used in the literature [22], o (=2nf) being the angular frequency, N is the concentration of ADWP configurations per unit volume which is assumed to be independent of frequency and temperature but dependent on the glass composition, R is the distance between the two wells, typically 4-5& e is the electronic charge, V, and A,,, are the cut-off energies for the barrier height, V, and asymmetry, A, respectively. z is the effective relaxation time given by z = zosech(A/2kT)exp(V/kT) where, zo is a constant prefactor inversely proportional to phonon frequency. Here kT has the usual meaning. Both V and A have distributions in the amorphous solid due to the disorder in the structure and interactions among atoms. Due to their low concentration, the ADWP transitions are assumed to be uncorrelated. Despite its success in explaining the low temperature ion dynamics, the description of ADWP is not fully understood. Lu et al. studied the variation of ADWP configurations (ADWPC) with glass composition for a binary lithium germanate series at low T and low f. It was observed that the ADWP concentration follows a power law, i.e. N a x0.79,where x is the lithium oxide content [ 151. This sublinear dependence on composition suggests that more than one alkali ion effectively contribute to one ADWP configuration, which also includes the network former atoms. Thus the alkali ions are important in determining the low T conductivity in oxide glasses. For a long time the structure of a jellyfish configuration remained a mystery, but systematic studies on several series of glass compositions have revealed significant details about the constitution of these entities. Most recently it has been found that the ADWPCs that are responsible for conduction at high f and high T are significantly different from those at low T and low f in a silicate glass [121. This is not surprising, since different kinds of ADWPCs have been observed at different temperature and frequency regimes. For example, Kanert et al. identified three kinds of ADWPCs in lithium silicophosphate glass [23]. Therefore, the scope of this paper is to first establish if the low ‘r- low f conductivity and high T - high f conductivity have the same common origin in germanate glass system.
Dielectric Materials and Devices
433
This is done by comparing the data in these two regions by an empirical relation. We further examine the mechanism of microwave loss by investigating the composition and temperature dependence of GMW of the potassium germanate glass series. EXPERIMENTAL Potassium germanate glass samples of composition xK20.( 1-x)Ge02, x = 0.0023, 0.02, 0.05, 0.074, 0.10, 0.15, 0.20, and 0.247, were prepared using the conventional melt quench method. The samples were cut and polished into lcm x lcm x lmm flat discs for low frequency electrical measurements. The surfaces were deposited with gold electrodes in a 3-probe configuration. For measurements at microwave frequencies, the glass sample was precision machined to fit into a coaxial airline of inner and outer conductor diameters 3mm and 7mm respectively. Low frequency electrical conductance in the frequency range 10 Hz to 100 kHz were measured using the GR1616 Precision Capacitance Bridge at temperatures 4K to 400K. For cryogenic temperatures, the sample was placed in a cryostat under He flow and temperature was controlled with an accuracy of O.lK. Conductivity was then calculated by multiplying conductance with the geometric factor (=d/A, d is the diameter of the top electrode and A is the area) of the sample. Microwave conductivity was measured by the transmission line method using a network analyzer (Hewlett Packard 8753C) and S-parameter test-set (85047A) in the frequency range from 10 MHz to 3 GHz and temperatures from room temperature (RT) to 400K. The sample holder consisted of a 7mm coaxial airline with its ends terminated with APC-7 connectors. Full two-port calibration was performed before the actual measurement to minimize errors and uncertainties from the instrument and port cables. The reflection and transmission coefficients, in the form of S-parameters, were measured in the specified frequency range. A Newton-Raphson iterative procedure [24] was utilized to calculate complex permittivity from the S-parameters. The conductivity was then calculated from the imaginary part of the permittivity. RESULTS The frequency dependence of G between 10 Hz-100 lcHz for the 24.7 mol% K20 glass is shown in fig. 1 for different temperatures as a log-log plot. The low temperature data are from the work of Lu [25]. At temperatures below 200K, CY increases linearly with frequency, the slope being close to unity. The conductivity in this region has been established to be due to thermally activated
434
Dielectric Materials and Devices
fluctuations of ADWP configurations, which can be approximately written as, El51
where, C is a constant independent of frequency and temperature but dependent on glass composition, y w 0-0.25 and p w 1.0-1.15 at low temperatures for most glasses studied [151. In fact, this linear region is observed at low temperatures for all compositions within this glass series. However, for temperatures 2200K, the increase in (T is no longer linear and contribution from diffusive single ion hopping begins to appear at the low frequency end of the plot. The frequency independent dc conductivity (ode) region appears at T 2 346.3K. It increases with temperature and finally at 394.7K, it spreads over at least two decades of frequency (see fig. 1). At the higher frequency end, 2104 Hz, (Tdc component is small. The transition from dc to ac region is rather smooth, with the slope increasing from 0 to 0.6 corresponding to the dc plateau and the UDR (i.e. the sublinear power law) regions, respectively. In order to establish that (T in the two regions, namely, low T - low f and high T - high f have the same origin, the data in the two regions have been analyzed using the empirical relation of the form similar to eqn. (2). The low T (4K - 100K) and low f (10 Hz - 100 kHz) data are extrapolated to room temperature using eqn. (2) and compared with the high frequency experimental data. The result of this extrapolation is shown in fig. 2 for the highest potassium content glass. One can clearly see that the two regions connect reasonably well, indicating the mechanism of conduction to be related in the two regions, as shown before for a silicate glass [9]. The dashed line joining the symbols is a least square fit with slope -1.1. Empirical analysis performed on other glass compositions yielded results similar to that presented for the 24.7 mol% glass. The variation of microwave conductivity with frequency for the glass composition 0.247K20.0.753Ge02 is shown in fig. 3 for temperatures from RT to 400K. Similar results were obtained for other glass compositions as well. From the figure, one can see that CY increases with a slope of approximately unity in a narrow range between 4x 10' Hz and 1O9 Hz. The temperature dependence of (TMW for 24.7 mol% alkali content glass is shown in fig. 4 for several frequencies. The dashed lines are fit to Ty power law, with y typically ranging between 0.8 - 0.9 at high frequencies. These values are higher than that observed at low T and low f by Lu et al. [IS] for this glass composition. The variation of CYMW with alkali concentration in the glass has also been studied at different temperatures. The results in iig. 5 show the composition dependence at 5 12 MHz and 1 GHz frequencies for three different temperatures.
Dielectric Materials and Devices
435
I
0.247K20-0 .753Ge02 10-7
’
10-1
d
10-12
I
I
I
I I I I I I
1 1 1 1 1 1 1
1 o2
10’
I
I
I 1 1 1 1 1
I
I 1 1 1 1 1
105
104
103
Frequency (Hz) Fig. 1: Frequency dependence of conductivity of 24.7 mol% K20 glass at low frequencies. Lines connecting the symbols are drawn to guide the eye. The dashed line is a reference for slope=l .O. 10-3
10-4
10-1’
102
103
104
105
106
107
108
109
1010
Frequency (Hz)
Fig. 2: Comparison of empirically extrapolated low f data (open circles) and experimental high frequency data (solid circles) of 24.7 mol% K 2 0 glass at 300K. The dashed line is a least square fit with slope x l . 1.
436
Dielectric Materials and Devices
Frequency (Hz) Fig. 3: Conductivity versus frequency in microwave region of 24.7 mol% K20 glass for temperatures from 300K to 400K. The symbols represent experimental data points and the dashed line is reference slope=l .O.
A
I 0-5
350
300
400
Temperature (K) Fig. 4: Temperature dependence of (JMW for 24.7 mol% K20 glass at different frequencies. Frequency values are 5x107, lxlO*, 5x108, lx109, 2x109 and 3x109 Hz from bottom to top respectively. The dashed lines are fit to Ty.
Dielectric Materials and Devices
437
One can see that OMW increases weakly with alkali concentration at these fiequencies. For example, at 1 GHz and 300K, conductivity increases by only a factor of 2 upon increasing the potassium content from 0.23 to 25 mol%. This is in stark contrast to the dependence of low T - low f conductivity, where 0 increases by almost an order of magnitude [25].
3x10-4
zx 10-4
-0- 388K
h
E
5
1 -4 .g 9x18-5
.I8 ~ 1 0 - ~ 5 7~10-~ 4 6~10-~ 5~10-~
6
4~10-~ 3x 10-5
zx 10-5
0
3
6
9
12
15 18 21
24
KzO(rnol%) Fig. 5: Microwave conductivity at 512 MHz (circles) and 1 GHz (squares) as a function of K 2 0 concentration at three different temperatures. The symbols represent experimental data points. Lines are drawn to guide the eye.
DISCUSSION One can clearly see from fig. 2 that the temperature and frequency dependence of 0 at low temperature-low frequency and high temperature-high frequency regions can be well described by a simple empirical relation given by eqn. (2). In other words, we are able to predict the frequency as well as the temperature dependence of microwave conductivity solely using the lowfrequency low-temperature conductivity data. Therefore, the mechanism of conduction should be of the same origin in these two regions. This conclusion is in agreement with what has been observed in the silicate glass system studied earlier [9]. This means, the localized movements 4)f jellyfish-like structures
438
Dielectric Materials and Devices
composed of network former atoms as well as alkali ions can describe microwave conduction in oxide glasses. The temperature dependence of microwave conductivity for all glass compositions follows a Typower law as predicted by the ADWP model (eqn. (2)). The values of the exponent y obtained from the fit of experimental GMW are found to vary between 0.5 and 0.9 for the lowest potassium to highest potassium content glass. These values are consistently higher than those obtained at low f and low T by Lu [25] for the same glass series. This discrepancy in the temperature exponent can possibly arise from the influence of additional mechanism such as low frequency single ion hopping (UDR region) to GMW. To establish this, the magnitude of UDR contribution to conduction at microwave frequencies was calculated using a ‘master’ plot of d o d c against O/TCYdc and then subtracted from the observed oMW. As a result of this subtraction, the temperature exponent y decreases from 0.86 to 0.16 for the 24.7 mol% K20 glass. The latter value agrees well with the y experimentally observed at low T and low f for this glass composition. From the composition dependence of OMW (refer fig. 5), we can see that conductivity increases weakly with potassium concentration. Nevertheless, the greater the number of alkali ions present, higher is the microwave loss, and hence conductivity. However, the variation of OMW below 2 mol% is complex owing to the structural rearrangement that occurs in the glass network upon the initial addition of alkali oxide [26,27]. It is believed that for low K20 content glass (0 I x 5 0.02), six-membered rings rearrange to form smaller ring structures, probably three-membered rings. This process is manifested by the abrupt downward shift of the asymmetric stretching vibration frequency of Ge-@-Ge linkages in the infrared spectra [26]. Unlike in silicates, the change in o with alkali concentration in germanate glasses is non-monotonic. For example, in sodium aluminosilicate glass with varying A1203, Topping and Isard found the loss to monotonically increase with A1 increase until Al/Na ratio was equal to unity [28]. The nonmonotonic behavior is common among germanates leading to the anomaly seen in many physical properties as a function of alkali concentration. The differences observed in the composition dependence between low f low T and high f - high T conductivity suggests that the mechanisms in these two regions, even though are of a common origin, may have some differences in the constitution of the species responsible for conduction. P, more careful and detailed investigation is underway to determine how the constitution of these high frequency ADWPC species differs from their low temperature - low frequency counterparts. It is possible that the composition dependence of UDR mechanism influences the high frequency data, resulting in the discrepancy.
Dielectric Materials and Devices
439
CONCLUSIONS The empirical analysis of the temperature and frequency dependence of microwave conductivity of potassium germanates shows that the mechanism of microwave conduction has a basically common origin to conduction at low f and low T. In other words, the high fiequency, high temperature dielectric loss in oxide glasses can be established to originate from j ellyfish-like localized fluctuations of atoms in asymmetric double well potential configurations. Additional contribution from single ion hopping is found to influence the temperature dependence of dielectric loss at microwave frequencies. Furthermore, the composition dependence of GMW shows non-monotonic increase with alkali concentration. ACKNOWLEDGEMENT The authors wish to thank Drs. C.M. Weil, J. Baker-Jarvis and J.H. Grosvenor of NIST, Colorado for their help with discussions and suggestions in setting up the microwave measurement system. This work was supported by National Science Foundation under grant no. DMR-9225072. REFERENCES ~
[ 13 Y. Haessler and L. Johansen, “Microwave Heating of Fused Quartz to High Temperatures in the Fabrication Process of Optical Fibers”, Matls. Res. Soc. Proc., 124 (1988) pp:273. [2] W.H. Sutton, “Microwave Processing of Ceramics”, Am. Ceram. Soc. Bull., 68 (1989) pp:376. [3] D.J. Duval, M.J.E. Terjak and D.E. Clark, “Microwave melting of ion-conducting glasses ”, MRS Symp. Proc., 430 ( 1996) pp: 125. [4] M.P. b o x and G.J. Copley, “Use of Microwave Radiation for the Processing of Glass”, Glass Technol., 38 (1997) pp:9 1. [ S ] L. Navias and R.L. Green, “Dielectric properties of glasses at ultra-high frequencies and their relation to composition”, J. Am. Ceram. Soc. 29 (1946) pp: 267. [6] A.R. von Hippel, Dielectric Measurements and Applications, Technology Press of MIT, Cambridge, MA, 1954. [7] J.M. Stevels, ‘Dielectrische verluste des glases’, Glasstech. Ber. 29 (1953) pp: 227. [8] J.M. Stevels, “The Electrical Properties of Glass”, Handbuck der Physik XX (1957) pp:35 0. [9] H. Jain, S. Krishnaswami, 0. Kanert, “New Mechanism of Microwave Conductivity of a Silicate Glass”, Ceram. Trans. 106 (2000) pp:35 1. [lO] K. Funke, C. Cramer, B. Roling, T. Saatkamp, D. Wilmer, M.D. Ingram, “Ionic and polaronic glassy conductors: conductivity spectra and implications for ionic hopping in glass”, Solid State Ionics 85 (1 996) pp:293.
440
Dielectric Materials and Devices
[ 113 C. Cramer, K. Funke, “Observation of two relaxation processes in an ion conducting glass yields new structural information”, Ber. Busenges. Phys. Chem. 96, (1992) pp: 1725. [ 121 S. Krishnaswami, H. Jain, K.I. Kamitsos, J. Kapoutsis, “Connection between the microwave and far infrared conductivity of oxide glasses”, J. Non-Cryst. Solids, in press. [ 131 C.H. Hsieh, H. Jain, “Are the low temperature low frequency and high temperature high frequency ac conductivity of glasses the same phenomenon?”, J. Non-Cryst. Solids, 203 (1996) pp:293. [ 141 H. Jain, “’Jellyfish’ atom movement in inorganic glasses”, Met., Matls. Processes, 11 (1999) pp:317. [ 151 X. Lu, H. Jain, “Low temperature ac conductivity of oxide glasses”, J. Phys. Chem. Solids, 55 (1994) pp: 1433. [ 161 X. Lu, H. Jain, 0. Kanert, R. Kuechler and J. Dieckhoefer, “Low temperature dynamics of inorganic glasses: electrical conductivity vs. nuclear spin relaxation”, Phil. Mug. 70 (1994) pp: 1045. [ 171 W.K. Lee, J.F. Liu, A.S. Nowick, “Limiting behavior of ac conductivity in ionically conducting crystals and glasses: A new universality”, Phys. Rev. Lett. 67 (1991) pp:1559. [ 181 C. Cramer, K. Funke, B. Roling, T. Saatkamp, D. Wilmer, M.D. Ingram, A. Pradel, M. Ribes, G. Taillades, “Ionic and polaronic hopping in glass”, Solid St. Ionics 86-88 (1996) pp: 48 1. [ 191 A. Pradel, G. Taillades, C. Cramer, M. Ribes, “Ion dynamics in superionic chalcogenide glasses studied in large frequency and temperature ranges”, Solid St. Ionics 105 (1998) pp:139. [20] D.L. Sidebottom, P.F. Green, R.K. Brow, “Scaling behavior in the conductivity of alkali oxide glasses, polymers and doped crystals”, J: Non-Cryst. Solids 203 (1 996) pp:300. [2 11 A.S. Nowick, “Exploring the low-temperature electrical relaxation of crystalline oxygen-ion and protonic conductors”, Solid St. Ionics, in press. [22] K.S. Gilroy, W.A. Phillips, “An asymmetric double-well potential model for structural relaxation processes in amorphous materials”, Phil. Mug. B, 43 (198 1) pp:73 5. [23] 0. Kanert, R. Kuchler, J. Peters, A. Volmari, H. Jain, H. Eckert and E. Ratai, “Structural basis of low-frequency excitations in silicophosphate glasses”, J Non-Cryst. Solids, 222 (1997) pp:321. [24] J. Baker-Jarvis, M.D. Janezic, J.H. Grosvenor, R.G. Geyer, “Transmission /Reflection and short-circuit line methods for measuring permittivity and permeability”, NIST Technical Note 1355 (revised), 1993. [25] X. Lu, “Study of Ion Movement in Inorganic Glasses at Low Temperatures”, Ph.D. Dissertation, Lehigh University, 1994. [26] Y.D. Yiannopoulos, E.I. Kamitsos, H. Jain, “Structure of potassium germanate glasses by vibrational spectroscopy”, in Physics and Applications of Non-Cwstalline Semiconductors in Optoelectronics, NATO Advanced Research Workshop, Moldova, 1996. [27] H. Jain, W.C. Huang, E.I. Kamitsos, Y.D. Yiannopoulos, “Significance of intermediate range structure for electrical conduction in alkali germanate glasses”, J. Non-Cryst. Solids, 222 (1997) pp: 361.
Dielectric Materials and Devices
44 1
[28] J.A. Topping, J.O. Isard, “The dielectric properties of sodium aluminosilicate glasses at microwave frequencies ”, Phys. Chem. Glasses, 12 (1971) pp:145.
442
Dielectric Materials and Devices
HIGHLY ACCELERATED LIFE TESTING (HALT) OF K-4500 LOW FIRED X7R DIELECTRIC * Galeb H. Maher MRA Laboratories North Adams, MA U.S.A. ABSTRACT Highly accelerated life test was performed on an 0805 - 100 nF X7R multilayer chips with 12 microns active layer thickness. The material is a low fired commercial product (SF-422) produced by MRA Laboratories, with a dielectric constant of about 4500. The test conditions consisted of 125, 140, 155, and 175OC, and voltages ranging between 250 and 600 volts. The preliminary data showed a temperature dependence of an activation energy of 1.21 electron volt, nearly similar to that reported by other researchers on an X7R - BaTi03 system. However, the voltage dependence of acceleration factor (n) was found to be in the range of 5.4 to 7.1, almost twice as large as those reported in the literature. The experimental data showed a good fit with the Weibull statistical distribution. These observations suggest that this dielectric should be usehl for high voltage and high temperature applications. For this type of MLC chip, tested at twice rated voltage (lOOV, 125OC), the mean life time was predicted to be in excess of one million hours. INTRODUCTION We recently reported(')on the physical and electrical properties of a high K (4500), low fired, barium titanate base X7R dielectric. This material is commercialized as product SF-422. The shift in market demands for higher voltage (>200V) and higher temperature (>125"C) X7R applications have prompted us to examine the intrinsic capability of this dielectric for these applications. As a first phase of this study, we performed a highly accelerated life testing (HALT) in the temperature range of 125 to 175OC and voltage range of 250 to 600 volts on 0805-1OOnF chips with fired layer thickness of about 12 microns. Accordingly, the objective of this paper is to report the HALT results on this dielectric. Prokopowicz and Vaskas were the first to use an empirical relationship to predict the mean time to failure of multilayer ceramic capacitor, based on accelerated life testing at higher voltage and temperature.
* Persented in parts at the US. -Japan
Seminar, Nov. 1999, Ohnawa, Japan and at Passive Componentsfor Power Elecronic, Penn State Universiv. April 2000
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
443
Where: t, =the mean time to failure at Vl and T1 tz =the mean time to failure at V2 and T2 V1, Vz = test voltages of the MLC, in Volts T1, T2 = test temperatures of the MLC in OK n = voltage stress exponent E, = activation energy (electron-volt) K = Boltzman constant = 0.86 x 104ev/”K Based on the results observed by many researchers, this equation appeared to fairly predict the life expectancy of the capacitor and is widely used in the industry to assess the properties of the MLC capacitors. For BaTi03 based MLC chips, Prokopowicz(2’ reported a value of 2.7 for n and 0.9 ev for activation energy E,.
Other researchers have reported slightly different values. Very recently, however, Sakabe ()’ reproted a voltage acceleration factor that ranged between 5.5 and 6.9 for Y-doped X7R BaTi03 dielectric. The HALT data were generated on MLC, with very thin active layers around 2 microns and Ni internal electrode. The Table I below summarizes their findings. Table I.
The Values generally used in the industries are: n=3 and E,=l . 1 ev During our research, we were surprised to find that the voltage acceleration exponent n, was greater than 5, however the average activation energy E, was 1.21 ev and it is within the reported range shown in Table I.
EXPERIMENTAL SAMPLE MLC capacitors of 0805 size with 30 layers of about 12 microns fired thickness were manufactured by the “wet” deposition process in our laboratories. The internal electrode was 70 Ad30 Pd composition. The sample used for this study was taken fiom a single firing representing about 20,000 chips. After silver termination, the chips were measured for capacitance and dissipation factor, and all showed 100 nF k 5%. No other testing was done prior to the HALT study. The pertinent physical and electrical properties of this sample are shown in Table 11. The temperature coefficient of capacitance (TCC) and SEM image of a polished cross section are shown in figures 1 and 2, respectively.
444
Dielectric Materials and Devices
Table 11. - Sintering Temperature: 1 130°C/3 hour - Fired Density Ceramic only 5.85 g/cm3 (97.5% of theoretical) - Size: 0805 (2.0 x 1.25mm) - Number of Layers: 30 - Fired Layer Thickness: 12 microns - Average Capacitance at 1 KHz, 1 Vrms, 24 hours: 101 nF - Average Dissipation Factor: 1.8% - Calculated Dielectric Constant "K': 4600 HALT SETUP A test fixture was assembled to accommodate 24 individual chips held in place by a spring loaded strip. The temperature chamber is a Blue-M oven Model OV-490A-1. The test temperature was held to 2°C from the set value. A 1000 l 2 resistor and a relay were connected in series with each capacitor. The leakage current of each capacitor was monitored during the test. The test circuit was interphased with a PC which recorded the exact time of failure on each chip. When the leakage current exceeded 105 ampere, the relay opened up, thus removing the failure from the test and the time was recorded by the PC. The power supply is a KepCo Model 2000M where the voltage was held to +1 volt from the set value.
+
TEST RESULTS Although many researchers have used lognormal and Weilbull statistics to analyze the HALT data, in this study we used the Weilbull to determine the mean time to failure at 63% failure level of the test samples. Table III below shows the different test conditions that were used. 22 chips were tested at each condition until more than 90% of the sample had failed. Table 111. Life Test Conditions
600 I 5 5 0
I450
[ 350
I
Figures 3, 4, 5 and 6 show the Weibull plot curves for 175, 155, 140 and 125OC for three different voltages, respectively. The early failures (below 30%) were considered to be caused by manufacturing defects such as thin spots in the active layers. The mean time to failure at 63% failure level for each test condition was determined from the graphs with the help of a MathCad program and curve fit linear regression statistic. Using equation (l), the voltage acceleration factor n, and the activation energy E, were determined for the various test conditions as shown in Figures 7 and 8, respectively. It is interesting to note that a higher n (>6) was observed for test temperatures at 155 and 175, while for 125 and 14OOC conditions, n was between 5.4 and 5.7. Similarly the activation energy of failures was higher (E, for 300V, with an E, = 1.1 ev.
=
1.35 ev), for higher stress voltage 550V than
PREDICTED VERSUS ACTUAL TIME TO FAILURE Using an average voltage stress, n = 6.17, and an average activation energy, E, = 1.21 ev, the predicted time to failure at 63% failure level was compared to the actual time observed in this study. The time to failure, t2,at T2 = 175"C, and V2 = 350, were used to generate the data shown in Table IV.
Dielectric Materials and Devices
445
As can be seen from the results, the predicted TTF at 125 and 140°C are somewhat lower than the actual values while for 155 and 175OC, the actual TTF values, are slightly lower than the predicted one. These differences can, perhaps, be explained by slight variation in test temperature. Assuming we can further predict the TTF of these chips at lOOV and 125"C, which corresponds to twice rated voltage, the life time will be greater than 106 hours (1 14 years). Table IV. Examples of Time to Fail (TTF) Predictions Based on the data for 175"C, 350 Volt case: n = 6.17 (Average), Activation energy = 1.21 ev. (Average) P = predicted TTF, A=actual TTF (at 63% Failed) Oven temperature has a strong effect on the predictions. For example, the 125OC, 550 volt case would be 133 hours at 123O, and 93 hours at 127", instead of 1 1 1 hours at 125"
INSULATION RESISTANCE (IR)ON TEST The insulation resistance of each sample at each test condition was monitored during the HALT. The IR at the maximum stress voltage for the various temperatures at 63% failure are shown in Figure 9. It should be noted that there were no degradation in IR until the time of failure. ULTIMATE DIELECTRIC BREAKDOWN VOLTAGE To gain hrther insight into the capability of this dielectric, we have also performed a voltage breakdown analysis on a 22 chip size sample at -65,25, 125, 140, 155 and 175°C. The HALT fixture was used for this test. The voltage was raised slowly (about one minute) on each chip, until failure (relay opened). The dielectric breakdown voltage was also examined at liquid nitrogen temperature (-200°C). In this case, the sample was tested to 800V, at 25"C, to remove the low voltage failure. The failures distribution for each temperature are shown in Table V. Table V. Dielectric Breakdown at Different Temperatures
446
Dielectric Materials and Devices
20 15 Q)
m + .0
10
m Q 5 m
0
.-C
O
c
.-0 -r
.-(II -5 >
a,
f3-10 s -15 -20
I
-55
I
-35
I
-15
I
I
I
I
45 65 Temperature, Celsius
5
25
I
85
I
105
1
125
Figure 1 Temperature Coefficient of Capacitance
Dielectric Materials and Devices
447
HBar Marker = 10 ,urn
Figure 2 Polished Cross-section
448
Dielectric Materials and Devices
It is interesting to note that the ultimate dielectric breakdown remained relatively high in the range of -200 to +175”C. Further work will be performed with increase in temperature to 300°C. SUMMARY A HALT analysis was performed on 0805, 100 nF chips, of a low fired K-4500 X7R dielectric, between 125 and 175OC, and voltage stress ranging from 250 to 600 volts, on 12 microns layer thickness. - The voltage acceleration exponent, n, ranged between 5.4 to 7.1, while the activation energy, E,, ranged between 1.1 and 1.35 electron-volts. The insulation resistance on test was monitored and remained relatively high >108 ohms and constant until breakdown (thermal runaway). The ultimate dielectric breakdown voltage remained relatively high, >75V/micron even to 175OC. Using the time to failure value generated at 175°C and 350V, and averages n = 6.17, and E, = 1.21 ev, the predicted life time was in close agreement with actual values at the lower test temperatures and higher voltage. Assuming that the average n and E, generated in this study to be valid, the predicted life time of the 0805, 100 nF chip will be: at lOOV, 125OC : 4 x 106hours and at 50 V, 125OC: 3 x 108hours
-
ACKNOWLEDGEMENT The author wishes to thank Dr. George Shim for the analysis of the HALT data and Messrs. John Martin and Richard Zona for sample preparations and performing the test. REFERENCES: 1. Maher, G.M., et al., “Recent Developments in Low Fired X7R Dielectrics” Ceramic Transactions, Vol. 97, P17, 1999. 2. Prokopowicz, T., et al., “Research and Development Intrinsic Reliability, Subminiature Ceramic Capacitors” Final Report ECOM-9075-F, NTIS AO-864068,69. 3. Munikoti, R., “Highly Accelerated Life Testing (HALT) for Multilayer Ceramic Capacitor Qualification” E E E Transactions on Components, Hybrids, and Manufacturing Technology, Vol. 11, No.4, 1988. 4. Minford, W., “Accelerated Life Testing and Reliability of High K Multilayer Ceramic Capacitors”. IEEE Transaction and Components, Hybrids, and Manufacturing Technology, Vol. CHMT-5, No.3, 1982. 5 . Kurtz, S., et al., “Infant Mortality, Freaks and Wear Out:..” Proceeding of the Center for Dielectric Studies Symposium on Improvement of Multilayer Capacitor Reliability”. Penn State University 1989. 6. Confer, R., et al., “Use of Highly Accelerated Life Test (HALT) to Determine Reliability of Multilayer Ceramic Capacitors”. Proceeding of Electronic Component conference 1991. 7. Sato, S., et al., “Effect of Y-Doping on Resistance Degradation of Multilayer Ceramic Capacitors with Ni Electrode under Highly Accelerated Life Test”. J.J. Appl. Phys., Vol. 36 (1997), pp 6016-6020. 8. Pak, H., et al., “Reliability Prediction of Multilayer Ceramic Capacitors Using an Improved Accelerated Life Testing and Weibull Analysis Technique”. 1997 International Symposium on Microelectronic, pp 3 62-367. 9. Sakabe, Y. “MLC Technologies of Today and Future” U.S.-Japan Seminar Conference, Nov. 1999, Okinawa, Japan.
Dielectric Materials and Devices
449
I..'
II
. < 63 % ---
-I
'
.a
3
J
6
(TTF) Hours Slopes of > 30% Failed Data. 175 C
Ln
Wcibull Plots of 175 Degree C Data
X - 250 Volts 0 - 300 Volts + - 350 Volts
Figure 3
450
Dielectric Materials and Devices
I
e n (TTF) Hours
A
f
i
3.5
J
45
5
55
t n (ITF) Hours
6
65
Slopes of > 30% Failed Data. 155 C Weibull Plots of 155 Degree C Data
X - 300 Vcits 0 - 350 Voits + - 400 V c i k
Figure 4
Dielectric Materials and Devices
45 1
Q
ZSP
+
l
-
t+3
3 E-
p
n Y
0
z-
y
n 3
0 I-
?
$
2 L.
0
0
CZ W W
I
I
3
7
5
4
I
6
&n (TTF) Hours
5
4
45
5
5.5
I
6
& (TTF)Hours Slopes of 125 C Data, > 30% Failed Weibull Plots of 125 Degree C Data
X - 500 Volts 0 - 550 Volts + - 600 Volts
Figure 6
Dielectric Materials and Devices
453
125"C, n = 5.7 a 14OoC,n = 5.4 b 155"C, n = 7.1 c 175"C, n = 6.5 d
1
100
0
11.
C2
VOLTS
Voltage Acceleration Factor, n Figure 7
454
Dielectric Materials and Devices
1 o - 300 Volts, 1.1 ev a x - 350 Volts, 1.17 ev b + - 550 Volts, 1.35 ev c
looo
h
\
U
I
\
n
10 2.L
.d.
2.3
2.2
1/T, per degree K * 10*3 ACTIVATION ENERGY, E, Figure 8
Dielectric Materials and Devices
455
1.4-10’ 1.2-10’ 1-10’
.........................................
125 C. 600 Volts ....................................... ........................................... 140 C. 550 Vofts
X
+
...
8-10’ 6- 1O8 4- 1O8 2-108
0
__^________-_--_---________^________
155 C,400 Volts
___.__^______.___.-_-------.-----------.0 175 C,350 Volts
Figure 9 Insulation Resistance - Time (63% Failed) during HALT at Maximum Test Condition
456
Dielectric Materials and Devices
Ceramic Tapes for Wireless Applications R.L. Wahlers, S.J. Stein, M.A. Stein, A.H. Feingold & P.W. Bless Electro-Science Laboratories 4 16 E Church Rd., King of Prussia, PA 19406 Phone: 610-272-8000, Fax:610-272-6759,email: [email protected]
The rapidly growing wireless industry needs new high performance materials to build low loss, high density, thermally stable integrated packages. This paper describes cofire and transfer ceramic tapes developed to meet these needs. Low loss gold and silver based compositions for screen print and photoimaging technologies are used for top and inner layer conductors. Size reduction using buried passive components and techniques for reducing cross-talk are discussed. Design, processing and material interaction effects on dielectric constant and loss characteristics are presented for frequencies up to 20 GHz. Introduction The wireless telecommunications industry is growing at a rapid pace. Applications include automotive (safety, control and entertainment), global positioning system mapping (GPS), multifunctional portable phones, home entertainment and office voice, video and data transmission through wireless local area networks (WLAN). New materials are required to meet the increased speed and multifunctionality needed in these systems. Their GHz operating frequencies necessitate the use of dielectric matrix materials with low loss and low dielectric constants. Low loss is critical in applications requiring long battery life. Low K is needed for isolating signal traces and for high signal propagation rates. Polymer based circuit boards offer low K but suffer from poor thermal conductivity and, in some cases, lossy behavior at high frequencies (1). Ceramic substrates like alumina meet many of the requirements but have relatively high dielectric constants and suffer from the general requirement that refractory metal conductors must be used in multilayer configurations. In the frequency range of most telecommunications applications, where conductor losses dominate, the higher resistivity of these materials is undesirable. The solution proposed in this paper involves burying passive components in low firing ceramic tape. Burying the surface-occupying passive components frees up surface space for additional functions. It also increases the number of circuits per substrate which lowers system cost. Use of tape having low loss and low K values provides the dielectric properties needed in the matrix. Ceramic Tape System Figure 1 gives an outline of the two approaches used in building up the multilayer tape modules along with the advantages associated with each. In the cofire tape approach green tape sheets each with its appropriate screen printed circuitry are laminated together and cofired. The transfer tape process, on the other hand, involves sequential firing of individual layers laminated to a ceramic substrate such as alumina. The cofire approach To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
457
allows for a single burnout/firing cycle thereby improving cost savings, especially as the number of layers increases. Technological challenges are greater with cofire tape, however, since all materials must be matched in coefficient of thermal expansion and shrinkage during firing to avoid warpage of the final package. The transfer tape approach more closely resembles conventional thick film processing and offers more flexibility in the selection of materials, largely because planar shrinkage is effectively constrained by the ceramic substrate. Processing costs necessarily increase with the number of fired layers.
Transfer Tape
Cofire Tape
Processiny (Seauential) Tape sheets cast Lamination to substrate Metallization Component formation
Processin? (Parallel) Tape sheets cast Metallization Component formation Lamination of layers
Advantages Strength Heat dissipation Zero shrinkage (XY) Multiple fire
AdvantaFes Low labor Single fire Layer capability Sheet inspection
Figure 1. Approaches for building multilayer tape modules
The properties of the tapes, conductors, and buried components discussed in this paper are given in Tables 1-3. Table 1 lists the properties of two tape chemistries in both cofire and transfer formats. The “1 11” composition is distinguished by a very low dielectric constant (K) around 4 to facilitate high signal propagation rates, excellent isolation and low insertion loss. The “101” composition possesses a still low K of about 7 while offering some advantages in materials compatibility.
Table 1. Tape Properties
,
Tape Designation 11 1-TT 101-TT 11 1-CF 101-CF
458
I
Dielectric Constant -4 -7 4.2 7.3
1 Insertion Loss” I -0.004 dB/mm -0.006 0.0038 0.0057 “
“
Peak Temperature 850°C LL
8501875°C &L
I
FurnaceType
1
Belt LL
Box 66
“
Dielectric Materials and Devices
The conductors used in this study are shown in Table 2. They were formulated to provide low loss, ohmic contact, shrinkage match to the tape matrix and minimal interaction with the components. Metallurgies include Ag, PdAg and Au. The Ag and Au conductors are also available in photoimageable versions for high frequency applications requiring fine lines and precise edge definition.
Table 2. Conductors Designation Ag- 1 PdAg- 1 Au- 1
Metallurgy Ag Pd, Ag Au
Processing Screen Printed
Firing Temperature 850-875°C
LL
LL
LL
Lb LL
I
Au-PI
Au
LL
1
LL
Table 3 lists the designations and properties of the capacitors which were embedded in the tape matrix. For the capacitor tape, nominal stand-alone K and DF values along with conductor and firing temperature recommendations are listed. The reported properties/processing of the capacitor pastes are those related to conventional use as thick film inks fired on 96% alumina. These values may be compared with the buried component values reported later in the paper. Deviations from the numbers in Table 3 serve as an indication of the degree of interaction between the capacitors, conductors and tape.
Table 3. Properties of Capacitor Materials Capacitor Product Number 41 13 41 17 4151 4152 4153 41210-70C*
Dielectric Constant (K)
Dissipation Factor (DF, %)
110 300 300
I
1,000 2.400
100
c2.0
1
Electrode Metallurgy
Firing Temperature ("C) 930-980
Pd/Ag
LL
LL
Ag, PVAg
850-930
&a
a6
(a
LL
LL
LL
~2.5
PdAg
I
850
Firing temperatures for the tapes fall into the range of 850-875°C which allows for a greater selection of materials based upon conventional thick film pastes. This temperature range, below the melting point of pure silver, accommodates the use of high conductivity, air firable precious metal conductor systems. Conventional thick film tunnel kilns are appropriate for transfer tape firing while programmable box furnaces are preferred for cofire tapes which require a longer cycle containing adequate burnout time in the 450°C range. Figure 2 outlines these firing schedules.
Dielectric Materials and Devices
459
Firinp Conditions Belt Furnace (Transfer Tape) 580°C peak / 50 minute cycle (burnout) 850°C peak / 45 minute cycle (sintering) Box Furnace (Cofire Tape) 5"C/minute to 450°C Hold at 450" for 60 minutes 5"C/minute from 450" to peak temperature* Parts cool with furnace *Peak Temperatures of 850 & 875°C 12 minute hold time at 850°C 30 minute hold time at 875°C Figure 2. Firing conditions of dielectric tapes A ring resonator pattern was used to measure the insertion loss and dielectric constant versus frequency of the dielectric tapes. The specifics of test part preparation and measurement methods used in this technique are described in another publication (2). The loss versus frequency characteristics of the 111-CF and 101-CF tapes are plotted in Figure 3. Their values are close in the low frequency region with the superiority of the low K 11 1-CF composition becoming evident at higher frequencies. Ag-1 silver was used for the ground plane and top layer ring pattern for both parts. Firing was done in a box furnace at a peak temperature of 850°C. Comparison of the loss vs. frequency curves of 11 1-CF with silver on alumina and copper on FR-4 can be seen in Figure 4. The loss of 11 1-CF is close to that of silver/alumina and clearly better than that of copper/FR-4.
Buried Components Embedding passive components in the tape matrix can provide a significant amount of surface space on which additional functions can be placed. Component counts indicate that 90% or more of the components used in wireless applications are passives (3). The majority of these are de-coupling capacitors with tolerances in the 510% range. Since this performance is obtainable with buried capacitors the potential for saving space with buried components is great. Buried resistors and capacitors, especially when cofired, are subject to physical and chemical interactions with the surrounding tape layers. Values and tolerances are affected, sometimes dramatically so. Significant work is involved in selecting and developing passives which approach their conventional counterparts. High range resistors are particularly problematic because of their sensitivity to glass diffusion from
460
Dielectric Materials and Devices
the dielectric tapes. Buried resistors with tolerances of 20-30% are still the norm. Generally, it can be said that the materials used for buried passives are.still evolving.
0.050 h
2 0.040
-T
m
f 0.030 v1
?
4 5
0.020 0.010
A
0.000 0.0
5.0
10.0 Frequency (GHz)
15.0
20.0
Figure 3. Loss versus frequency of 1 11-CF and 101-CF
0.080
,
I
n
3 \
c) e
.-0
E
Y
0.060 0.040 0.020
c
n
0.000 0.0
5.0
10.0 Frequency (GHz)
15.0
20.0
Figure 4. Loss comparisons to alumina and FR-4
Dielectric Materials and Devices
46 1
Capacitors The simplest method for producing capacitors involves using the dielectric tape itself as the capacitor and printing the appropriately sized electrodes to achieve the capacitance desired. Buried capacitors constructed in this way with 101-CF and both Ag and PdAg conductors can be cofired at 875°C to produce flat packages with TCCs within S . 2 % from -55 to +125"C (EIA classification X7C).
Table 4. Buried Hiph K CaDacitor Tape Matrix Tape 111-CF 101-CF
K 50 79
DF (%) 1.2 1.2
AC (%, -55°C) -6.1 71 1.5
AC (%, 125°C) 3.9 13.5
Figure 5. SEM of K-100 tape buried in 11 1-CF tape
Higher values can be achieved by substituting a ferroelectric based tape composition for the capacitor layer. The K 100 tape was cofired at 850°C with P d A g electrodes in 101-CF and 11 1-CF with the results shown in Table 4. Dielectric constants are reduced by about 20-50% from the nominal value of 100 in these buried cofired composites because of interactions with the contiguous low K tape layers. Dissipation factors are maintained at low values around 1.2%. Temperature characteristics are X7R
462
Dielectric Materials and Devices
or better depending upon the surrounding matrix. The interaction between the capacitor tape and the host tapes is likely that of glass diffusion into the capacitor layer. This interaction does not appear to have any deleterious effects on the bonding between the tapes. Figure 5 is a photomicrograph of the K 100 tape with Pd/Ag-1 electrodes buried in 11 1-CF. No delamination is evident and boundaries are clearly defined. Fired film thickness of the single capacitor tape layer is approximately 95 pm. Maintaining planarity with buried components is often difficult for screen printed capacitors where multiple layers of dielectric and conductor are usually required. One approach which minimizes this problem is the use of an interdigitated electrode pattern followed by a layer of dielectric to fill in the spaces between conductor lines. The result is a nearly planar capacitor in which the capacitance becomes a function of line dimensions and, inversely, of spacing. Figure 6 is a photomicrograph of interdigitated Ag electrodes with 40pm lines and 60pm spaces obtained by photoimaging. These dimensions result in an eightfold increase in capacitance over those obtainable with the 250ym capability of standard thick film techniques. Table 5 shows the effect of line and space dimensions on capacitance. It should be pointed out that these interdigitated capacitors are inherently asymmetric in the z direction and careful selection of materials is required to avoid possible warpage.
Figure 6. SEM of photoimaged electrode pattern (40ym lines, 60pm spaces) As an indication of future developments, Figure 7 shows an SEM of a photoimaged Ag conductor which had been processed on a K 4, screen printed multilayer dielectric (ESL 491 1). Line width is 8pm on a 50pm pitch. At present, line definition is
Dielectric Materials and Devices
463
not optimized at this level of resolution but improvement would be expected to occur with processing on the smoother dielectric tapes.
Figure 7. SEM of photoimaged Ag conductor (8pm lines on 50pm pitch) Capacitors were also prepared in more conventional fashion using a parallel plate arrangement of Pd/Ag electrodes and two dielectric layers buried in both 101-CF and 11 1-CF. All layers were cofired at 850°C. The capacitor dielectrics tested were ESL 41 13 and ESL 41 17. The results are listed in Table 6. Electrode size was 1mm2and the capacitance density ranged up to about 3000 pF per cm2 of electrode area. Dielectric constants are reasonable but below the values obtained in standard thick film processing which has the advantage of separate firing schedules for all layers and lack of intimate contact with low K glass phases.
Table 5. Buried Interdipitated Capacitors Conductor Processing Screen Print
464
Line Width (w-0
Line Height (pm)
Line Spacing
250
15
250
(pm)
Capacitance Increase Over SP-Std SP-Std
Dielectric Materials and Devices
Table 6. Screen Printed Capacitors Buried in Cofire TaDe Capaci tornape 41 13 / 111-CF 41 13 / 101-CF 41 17 / 11 1-CF 41 17 / 101-CF
K 73 75 103 129
DF (%) 0.5 1.1 0.3 1.7
AC (70,-55°C) -1.1 -9.1 2.0 -2.0
AC (%, 125°C) 0.6 5 .O -1.0 1.2
Table 7. Screen Printed Capacitors Buried in Transfer Tape Capacitornape 4113/ 111-TT cc
41 13 / 101-TT
I
'L
41 17 / 11 1-TT C6
41 17 / 101-TT C6
4153 / 11 1-TT LC
4153/101-TT 6'
I
Temp ("C) 850 930 850 930 850 930 850 930 850 930 850 930
I
I
K 46 114 114 142 63 30 1 247 400 53 57 43 862
I
I
DF (%) 1.5 1.o 1.4 1.1 4.6 1.5 1.5 1.2 0.6 1.o 2.0 1.5
AC (%, 125°C) -3.2 I -1.1 13.6 -0.4 3.6 -8.8 -5.9 -8.5 -4.4 -8.2 I -8.9 -35.7
I
I
Separate firing of layers is possible with the transfer tape method and two approaches were tried. In the first, the dielectrics and electrodes were printed and fired on tape coated alumina followed by lamination and firing of a top tape layer. All firings were performed at 850°C. In the second, thick film dielectrics and conductors were fired onto terminated alumina substrates at temperatures higher than that used for subsequent firing of the tape. The ability of transfer tape processing to allow higher firing temperatures of the capacitors prior to their being buried means that the capacitors will achieve higher fired densities and higher K values while minimizing interactions with the glassy matrix of the tape. Table 7 shows the results of these approaches. Terminations were Ag-1 in all cases and the matrix tapes were 101-TT and 111-TT. The data for the 850°C firings represent properties obtained using tape coated alumina (first approach) and the data for the 930°C firings represent capacitors prefired onto alumina (second approach). The general conclusions are that prefiring the capacitors at a higher temperature can dramatically raise the dielectric constant but that interactions with the matrix can still be significant. The nature of the interaction depends upon the chemistries involved. Among the capacitor dielectrics tested, ESLA 153 shows the greatest variability with K values ranging from 43 to 862 depending on processing details. Generally, however, the highest K values are achieved with the transfer tape method.
Dielectric Materials and Devices
465
Resistors Resistors are arguably the most process sensitive of thick film materials and initial attempts to bury resistors in dielectric tapes resulted in sometimes dramatic changes in value and temperature characteristics. Undesirable physical and chemical reactions such as blistering could occur when resistors were buried in tapes. Cofiring the materials often exacerbated the problems. Rational approaches aimed at improving the compatibility of the compositions have yielded some positive results. Table 8 lists the resistor properties of a nominal lKQ, 100ppd"C resistor modified for tape use and buried in cofire and transfer tapes with Ag and PdAg conductors. Buried in 101-CF the resistor gave values similar to those that would be expected on bare alumina. The resistance values were higher and the TCR values more negative when this same resistor was buried in transfer tape of a similar composition. The larger changes in the latter are likely the result of the additional firings associated with the transfer tape process.
Table 8. Resistor Buried in Dielectric Tapes Matrix Tape
Conductor
111-TT
Ag- 1 Pd/Ag- 1 Ag- 1 PdAp- 1 Ag- 1 PdAg- 1
&b
101-TT bb
101-CF (6
Resistance (KQ) 13.4 11.5 8.3 5.3 1.26 0.78
CTCR (PPd"C> -75 1 -747 -680 -548 -183 -25
HTCR (PPd"C> -439 -423 -403 -3 13 -181 -26
Conclusions The rapidly growing wireless communications industry is placing demands on the materials and packaging industries to provide microwave circuits with high performance in small sizes and at low cost. One promising way to accomplish this goal involves the burying of passive components in low K dielectric tapes. The advantages of this approach include space and cost savings, good thermal dissipation, environmental protection of the buried components, reduction in the number of soldered connections, mechanical robustness and good high frequency performance. We have described a set of materials comprising low K tapes, compatible Ag, Pd/Ag and Au conductors and capacitors and resistors that can be buried in the tapes to form planar, dense structures in either the cofire or transfer tape formats. References ( € )Skurski, M.A., Smith, M.A., Draudt, R.R., Amey, D.I., Horowitz, S.J., and Champ, M.J., Microwaves & RF, Feb, 1999, pp. 77-86
(2) Proceedings of IMAPS Europe, Prague, June, 2000 (3) T.G. Reynolds, The Application of LTCC Technology to Improved Integrated Passive Components, LTCC Wireless Workshop, NIST, Nov, 1999
466
Dielectric Materials and Devices
LOW TEMPERATURE SINTERING MgCuZn ERRITES FOR MULTILAYER FERRITE CHIP Atsu yuki Nakano, Isao Nakahata, Taku Murase, Takeshi Nomura Materials Research Center TDK Corporation 570-2, Matsugasita, Minamihatori, Narita city, Chiba, 286-8588, Japan ABSTRACT The low temperature sintering MgCuZn ferrites were developed for the multilayer chip ferrite components. It was reported that electromagnetic properties of the chip are influenced by stresses from the Ag internal electrode and Ag diffiision [2][3]. As the magnetostriction of MgCuZn ferrites is lower than that of NiCuZn ferrites, it is expected that the ferrite materials have high potential in multilayer chip ferrite component applications, not only chip inductors but also LC and EM1 chip filter components. In this paper, we have discussed the preparation of low temperature sintering MgCuZn ferrites, and also the electromagnetic properties of the sintered ferrite's body and the multilayer chip ferrite. INTRODUCTION The incessant demand for higher density circuits in electronics during the recent decades has required the continued miniaturization of all components. In this situation, surface mount devices (SMD) have been also developing toward miniaturization and densification. In 1980, multilayer chip ferrites were developed using thick-film printing and co-firing technology [l]. Figure 1 illustrates a structural model of multilayer chip ferrites. Low temperature sintering NiCuZn ferrites are being used for the multilayer ferrite chip components since the ferrite chip was developed. It is well known that the magnetic characteristics of the ferrite materials depend on composition, additives, microstructure and stress. In the case of co-fired NiCuZn ferrite materials and Ag metal materials, stresses from Ag-diffusion and from Ag-electrode were expected easily.
Figure 1Structural model of a multilayer ferrite chip Previous studies have already reported that the Ag electrode brought about significant stress which had a deleterious effect on the magnetic characteristics [2]. By TEM investigation, it was observed that a great number of interference fringes existed at the grain boundary. At the center of interference, it was confirmed that Cu,.,O and Ag are present in high concentrations. It is supposed that Cu,.,O and Ag coexistence on the grain boundary brings the compressive stresses to the ferrite grains and causes the deleterious effect on the magnetic characteristics of multilayer NiCuZn ferrite chips [3] [4]. According to this equation of permeability, it was thought that even if stress were large, a chip would have high performance if magnetostriction were kept low. This is probably the first time investigation that focuses on the magnetostriction in ferrites in order to achieve high chip performance. It is known that magnetostriction of MgCuZn ferrites is lower than that of NiCuZn ferrites. Therefore we studied low temperature sintereable MgCuZn ferrites for To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
461
multilayer ferrite chip. Usually, a temperature higher than 1250°C is necessary to fire MgCuZn ferrites. However, in order to use these ferrites for the multilayer ferrite chip components, the firing temperature must not be over the melting point of Ag (940°C). This paper presents the effect of specific surface area, calcining temperature and chemical composition of MgCuZn ferrites on the densification characteristics were investigated. Multilayer MgCuZn ferrite chip preparation and the magnetic characteristics will be also discussed. EXPERIMENTAL The com osition of low temperature sintering MgCuZn ferrite was Fe20,=47.5m01%, ZnO=2lmol& Cu0=5.5-13.5mol%, and MgO=31.5-x mol% (Where x is CuO content). These raw materials were weighed and mixed by ZrO, ball-mill for 16hrs. Afterward, the mixture powder were dried and calcined at 700--900°C for 10hrs. The calcined powders were comminuted so as to give a specific surface area of 3-15m2/g. Then the ferrite powders were mixed with 1Owt%PVA binder solution and pressed to form T5-toroidal of size 3 X 3 X 4mm. M e r the samples were fired at 87O-93O0C, measured permeability and density. The densification characteristics of these samples were obtained by measuring the size of the pellets. Multilayer ferrite chips were prepared by printing multilayer method with same composition of low temperature sintering NiCuZn ferrite and MgCuZn ferrite. The green chips were fired at 870-9930°C. Then the magnetic characteristics such as inductance, Q and inductance depending on frequency were measured with LCR meter.
RESULT AND DISCUSSION The effect of cupric oxide content on the densification characteristics of MgCuZn ferrites is shown in figure 2. Iron oxide and zinc oxide contents were 47.5mol% and 21mol% respectively. These samples were heated at 10"C/min. up to 1100°C in air. According to higher the cupric oxide content, the densification characteristics of MgCuZn ferrites shifted towards low-temperature side. When the cupric oxide content was 11.5mol%, a marked densification occurred. From the view point of low temperature sintering, high CuO content material is used for multilayer chip ferrites. This is due to the fact that the magnetic properties of NgCuZn ferrite depend on composition strongly and the final composition can be not selected from this view point alone. Consequently, cupric oxide content should be as high as possible without compromising the magnetic properties. Figure 3 shows the effect of specific surface area on the densification behavior of NgCuZn ferrites. It was shown that the densification temperature of NiCuZn ferrites decreases with increasing specific surface area. Especially, the densification characteristic was clearly increased when powder having a specific surface area of more than 6.5m2/g was used. I a n g time for ball milling is required to get the powder with larger specific surface area. Thus, to achieve over 8 m2/g, techniques were developed to adequately extend the milling time while minimizing contamination from the media. 0 n
-4 8
-1 2 -16
-20
500
6M)
700
800
900
1000
1100
Temperature (T) Figure 2 Effect of cupric oxide content, on the densification characteristics of MgCuZn ferrites.
--
500
m
700
800
900
1000
1100
Temperature ('C) Figure 3 Effect of specific surface area on the densification characteristics of MgCuZn ferrites.
In order to get more densification characteristic of MgCuZn ferrite, calcination condition was investigated as shown in figure 4.It was observed that the densification temperature of MgCuZn
468
Dielectric Materials and Devices
ferrites decreases with calcination temperature. Usually, the calcination temperature of NiCuZn ferrite is about 700°C. However in the case of MgCuZn ferrite, Fe203still existed in the powder with 700°C calcined. XRD investigation showed that the peak of Fe203 decreased with the increasing calcination temperature. It is important to choose the calcination temperature to obtain homogeneous composition. As a result of these investigations, low temperature sintering MgCuZn ferrite required not only higher CuO content and specific surface area but also adequate calcination temperature.
-20
" "
500
'
I
"
600
' I
"
700
"
800
"
'
900
1000
1100
Temperature ('C) Figure 4 Effect of calcination conditions on the densification characteristics of MgCuZn ferrites. Figure 5 shows the effect of permeability (p) of low temperature sintering MgCuZn ferrite and NiCuZn ferrite on sintering temperature. These ferrite materials were prepared by the same composition. At between 870 and 900"C, MgCuZn ferrite sample shows lower p than NiCuZn ferrite sample. However above 910"C, p of MgCuZn ferrite sample is higher than that of NiCuZn ferrite sample, especially p of MgCuZn ferrite sample which fired at 930°C showed about 100 higher than that of NiCuZn ferrite sample. The density of these samples increased with sintering temperature. It is known that densification of MgCuZn ferrites depended on the sintering temperature significantly as compared to the NiCuZn ferrites.
860 870 a80 890 900 910 920 930 940
Temperature ("C) Figure 5 Effect of sintering temperature on p. The effect of compressive stress on change in p of MgCuZn ferrite is shown in figure 6. It is known that compressive stresses cause a decrease in pof both materials. However the change in p
Dielectric Materials and Devices
469
of MgCuZn ferrite is smaller than that of NiCuZn ferrite. Figure 7 shows the effect of co-fired with dielectric material on change in p of MgCuZn ferrite. In this experiment, when the green body was pressed, dielectric material was also pressed with the ferrite. Then the samples were fired. The dielectric material is also the low temperature sintering material which is TiO, added CuO Swt%. It was expected that stress occurred from interface between dielectric material and ferrite material. In the case of co-fired sample of NiCuZn ferrite, the p shows 22.9% decrease. On the other hand, p of the MgCuZn ferrite sample decreases only 3%. These results are consistent with our assumption. Therefore it was thought that MgCuZn ferrite has a high potential for multilayer chip components. 290t
0 n -10
5
3 0
U
,
,
,
,
,
j
270 260
-20
2 2%
.rl
&,
I
280
-30
240 230
-40
220
-50
0
5
10
15
20
310
25
30
Ferrite(0.7g)
Compressive stress (MPa) ~i~~~~ 6 Effect, of compressive stress on change in p.
Ferrite(0.7g)
t Dielectric rnateriaJ(0.5g)
Figure 7 Effect of co-fired with dielectric material on change in p .
We attempted to make the multilayer chip ferrite using printing method with low temperature sintering MgCuZn and NiCuZn ferrites. Figure 8 shows the effect of sintering temperature on inductance of multilayer chip ferrite. As shown in Figure 5, p of NiCuZn ferrite was higher than MgCuZn ferrite between 870 and 910°C sintering temperature. However, MgCuZn ferrite chip shows almost same inductance of NiCuZn ferrite chip between 870 and 890"C, then above 890 "C MgCuZn ferrite chip obtained higher inductance than NiCuZn ferrite chip. In order to confirm the MgCuZn ferrite's potential, moreover we tried to make several compositions of MgCuZn ferrite and NiCuZn ferrite, and measured inductance of those MgCuZn ferrite chips and NiCuZn ferrite chips. From the results, It was confirmed that inductance of MgCuZn ferrite chip was higher than the NiCuZn ferrite chip at all composition.
060 070 880 890 900 910 920 930 940
Temperature ("C) Figure 8 Effect of sintering temperature on the inductance of chip ferrit,es.
470
Dielectric Materials and Devices
Also the inductance and Q vs. frequency was measured as shown in Figure 9. The inductance frequency characteristic of MgCuZn ferrite chip shows lower frequency than that of NiCuZn ferrite chip. And Q peak of MgCuZn ferrite is lower than that of NiCuZn ferrite. To find the reason, we investigated to measure insulation resistance and dielectric constant of these ferrite materials. However these results were almost the same. At present, the cause of this contradiction is unknown, and should be investigated in fitrther detail.
I_.__._.._
MgCuZn ferrite (L) NiCuZn ferrite (L)
I
-MgCuZn ferrite (Q) -.-.---.NiCuZn ferrite (Q)
I
0
1.5 105 n
8
1 105
I 4
5 106 0 1
Frequency (Hz)
Figure 9 Inductance (L), Q VS. Frequency. CONCLUSION (1) Densification characteristics of MgCuZn ferrites depend on specific surface area, amount of CuO and especially calcination temperature. (2) In order to obtain the low temperature sintering MgCuZn ferrite, SSAZ7m2/g, C u O z 7.5mol% and calcination temperature =760-850"C were required. (3) Under compressive stress, change in permeability of MgCuZn ferrite was lower than that of NiCuZn ferrite. (4) The multilayer ferrite chip using MgCuZn ferrite showed higher inductance than the chip using NiCuZn ferrite. (5) Th? inductance frequency characteristic of MgCuZn ferrite chip showed lower frequency than that of NiCuZn ferrite chip (6) It is thought that MgCuZn ferrite has high potential for Multilayer chip applications. References 1 T. Nomura, M. Takaya, HYBRIDS 3, 15(1987). 2 A. Nakano, T. Suzuki, Y. Kanagwa, H. Watanabe and T. Nomura, Proc.of lothTakei Seminar, P1-8 (1990). [3] A. Nikano,' H. Momoi, T. Nomura, Proc.of International Conference on Ferrites, P1225-1228 (1992). [4] A. Nakano, T. Sato, T. Nomura, Proc.of A.Cer.S, Vo1.47 (1995), P241-249.
[I
Dielectric Materials and Devices
47 1
This page intentionally left blank
EFFECT
OF
RARE-EARTH
DOPING
ON
THE
TEMPERATURE-CAPACITANCE
CHARACTERISTICSOF MLCCS WITH Ni ELECTRODES Shigeki Sato, Yoshinori Fujikawa and Takeshi Nomura Materials Research Center, TDK Corporation 570-2 Aza-Matsugashita, Minamihatori,Narita-shi, Chiba-ken, 286-8588 Japan
ABSTRACT Doping effect of rare-earths on the temperature-capacitance characteristics of MLCCs with Ni electrodes has been investigated for BaTi0,-MgO-Rare-earth system. Temperature dependence of capacitance became smaller with smaller radii ions such as Ho, Er, Tm, Yb, and Lu. Especially, Tm, Yb, and Lu were effective dopants to depress the temperature dependence of capacitance. The Curie temperature also shifted toward higher temperatures by the doping of smaller ionic radii rare-earth elements. On the other hand, temperature dependence of capacitance became larger and the insulation resistance became lower by the doping of larger ionic radii rare earth elements such as Tb and Gd. It is clear that ionic radius of rare-earth dopants is an important factor to control the temperature-capacitance characteristics as well as improving the reliability. Ni electrode MLCCs with X8R characteristics specification for automotive use have been developed newly using rare-earth dopant having smaller ionic radii such as Tm, Yb, Lu. INTRODUCTION The requirement for higher capacitance and further miniaturization in multilayer ceramic capacitors (MLCCs) is driven by the downsizing trend in electronics. Structurally, this has meant reduction of layer thickness and an increase in the number of layers. The choice of internal electrode has shifted from Pd to Ni in order to realize low cost production. These requirements also have demanded in automotive uses with the introduction of electronic controls, such as ECU(Engine Control Unit), PGM-FI(Programmed Fuel Injection), ABS(Anti-Lock Brake System), and so on. Recently, since the weight reduction of automobile in order to improve the fuel consumption, elimination of wire harness and the control module to be mounted in engine room is
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication,reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
413
desirable. Hence, electronic parts in the control module for the automotive applications must keep their performance under higher temperatures. The Bi20,-Pb0-Ti02 composition has been commonly used for the X8R-MLCCs. However, the Bi203-Pb0-Ti02type MLCCs could not be realized with thinner layers, and Pd or Ag-Pd as inner electrode has to be used. Therefore, the demand of X8R MLCCs with Ni electrodes for automotive application have become stronger with the trend towards miniaturization, cost reduction and environmental protection (Pb-Free). In order to realize above, the authors have been studying dielectric formulation and microstructure, especially focusing on the temperature characteristics of capacitance of BaTiO, based materials, and successfully developed the X8R-MLCCs with Ni electrodes. In this paper, we report the relationship between temperature-capacitance characteristic and rare earths doping in Ni-X7R material. EXPERIMENTAL PROCEDURE The dielectric compositions were BaTiO, +R203(R=Gd,Dy,Ho,Er,Tm,Yb,Lu)+ MgO + MnO + V,O, + (Ba,,Ca,,)SiOj. Highly pure oxalate or hydrothermal BaTi03 powders were employed here. The additives were reagent grade oxides or carbonates. MLCCs were prepared by the so-called sheet methods. Green sheets were formed by doctor-blade casting and the thickness was controlled between 5 and 10 microns. After casting, and dring, Ni electrodes were printed on the sheets using screen-printing system. Next, 4 to 100 layers of the sheets were stacked, laminated, and cut into green chips. Binder burn-out was followed by sintering between 1260 and 1340°C. During sintering, the oxygen partial pressure was controlled between 1O-' and 1O-I3 MPa by adjusting the amounts of H2and H,O in the N2-H2-
H 2 0 gas mixture. The chips were then annealed between 104 and 10-9MPa in a N2-H20gas mixture in order to re-oxidize the dielectrics. The permittivity, dissipation factor and their temperature dependencies were measured using a HP4284A LCR meter (Hewlett-Packard) at 1 KHz with 1.0 V,.
The microstructure was analyzed by TEM,
and XRD. The Curie temperature was determined using DSC. Phase transition mode was investigated by Raman spectroscopy. The lattice constant of diffused area where rare-earths substitute into BaTiO was measured by XRD. The composition of diffused area replica were BaTi03+MgO(1mol%)+Re203.BaTiO,, MgO and rare-earth raw mixtures were prepared by wet ball-milling forl6hrs. After drying, the mixture was calcined for 2hrs at 1150°C. The diffused aria replica was obtained by wet ball milling for 16hrs and dried off calcined powder.
474
Dielectric Materials and Devices
RESULTS AND DISCUSSION Temperature-CapacitanceCharacteristics Figure 1 shows the temperature-capacitancecharacteristics of dielectric doped with rare earths. The temperature coefficient of capacitance became smaller with the smaller ionic radii such as Y, Dy, Ho, Er, Tm, Yb, and Lu, and satisfied the X7R specification. Especially, it should be noted that Tm, Yb, Lu doped samples satisfy the X8R specification. On the other hand, temperature coefficient of capacitance became larger by the doping of the larger ionic radii rare-earth such as Tb and Gd. In addition the insulation resistance was became lower.
1
2--20 -40 -50 -60
-Tb -Gd -Dy +Y
oH;
-50
0
+Er
,-L;
+Tm -Yb
DY 50
100
150
Temperature ["C]
200
Fig. 1 Temperature dependence of capacitance for various rare-erarth doped BaTiO, DSC-Analysis The Curie temperature of various rare earth doped BaTiO, were measured using DSC and the results are shown in Fig.2. An endothemic peak due to tetragonal-cubic phase transition was observed at a temperature between llOoC to 135oC. The peak shifted toward higher temperatures and the peak width was broadened by the doping of smaller ionic radii rare-earth.
The peak shift toward higher
temperatures indicates that the tetragonal phase has become stable. The broadening of peak width suggests that the tetragonal-cubic phase transition became dispersive. It is considered that increase of Curie temperature and dispersion of phase transition improves the temperature-capacitancecharacteristics at high temperatures.
Dielectric Materials and Devices
475
1
900
-t 400 300
200
Fig.2 DSC profiles of BaTiOJ doped with various rare-earth elements. Phase-Transition-Mode-Analysis The (004), (400) peaks for the Y-doped and Yb-doped dielectrics were measured between 25oC to 2000C using the hot stage XRD, in order to investigate the effect of the rare earth doping on phase
transition of BaTiO,, as shown in Fig.3 XRD peak width of Y-doped X7R-material decreased at a temperature above 125oC, along with the tetragonal-cubicphase transition. On the other hand, for the Yb-
doped dielectric, phase transition was not observed clearly by XRD. This phenomenon corresponds to DSC results. Y-doped
Yb-doped
1-
"--.4.--..-A& cuJ(IL_
A A .Jpk K a l Ka2
2%
-20093
16093 15097
-14%
14097 1 s t 13ot
Fig.3 Phase transition of Y-doped and Yb-doped dielectrics measured by hot stage XRD Raman spectroscopy was done between 25°C to 200°C for Y and Yb doped dielectrics, in order to study the effect of rare-earth doping on phase transition mode for the BaTiO,. Fig.4(a) and (b) show the Raman
476
Dielectric Materials and Devices
spectra for the Y and Yb doped dielectrics. Raman spectra indicates that tetragonal phase (ET0 mode) exists above 140oC in Yb doped dielectric. On the other hand, Y-doped dielectric, tetragonal phase disappears above 125°C. Fig.5 shows the temperature dependence of integral peak intensity of ET0 mode peak. It is clear that both dielectrics indicates the primary phase transition mode. It is thought that dispersive transition phenomena in DSC and hot stage XRD does not relate to the tetragonal-cubic phase transition. Therefore, it is considered that there is no relationship between the increase of Tc and phase transition mode of BaTi03.
Fig 4 Temperaturedependence of Raman spectra for Y-doped and Yb-dope dielectrics.
0.05
-
0.03 0.04
0.02 0.01 0.00
-
-
50
100
150
-
Temperature ["C]
200
Fig.5 Temperaturedependence of relative integral intensity of Raman peak at 310-lcm
Dielectric Materials and Devices
477
Microstructure-Analy sis The effect of rare earth doping on microstructure was investigated. Fig.6 shows the XRD profiles around (002), (200) peaks for the rare-earths doped BaTiO,.
The distance between (002) and (200)
peaks tended to increase by the doping of smaller ionic rare-earth elements. Peak intensity at (002) also increased by the doping of smaller ionic rare-earth. These XRD results mentioned that the tetragonality become stronger by the smaller ionic rare-earth doping. It is assumed that increase in tetragonality shifted toward higher temperatures. Gd and Tb doped dielectrics also indicate the increase in tetragonality, however, microstructure of both dielectric show the grain growth. It is considered that the core-shell structure was broken by the doping of Gd and Tb.
Fig.6 XRD profiled for various rare-earth doped BaTi03 around (002),(200) TEM-Analysis TEM micrograph indicats the grain size decreased with the decrease in the ionic radii of rare-earth as shown in Fig.7. Gd doped dielectric show the grain growth and destruction of core-shell structure. Dy doped dielectric has clearly a core-shell structure. On the other hand, Microstructure of Ho, Y and Yb doped dielectric have a mixture of largely homogeneous (non-diffused grain) and inhomogeneous grains (partially diffused grain). These difbed and non-diffused phases were not like the so-called core-shell structure, which is commonly observed in X7R type dielectric materials. However, a very small number of core-shell like grains were also observed. Therefore, it has been demonstrated that the core-shell structure was not necessary in order to stabilize the temperature-capacitancecharacteristics as X7R and
478
Dielectric Materials and Devices
X8R Gd203
Dy2°3
Ho203
Yb203
.
Fig.7 TEM micrograph of various rare-earth doped BaTiO, Fig.8 shows a TEM micrograph of the Y and Yb doped dielectric. Many interference fringes, probably due to stress, were observed for Yb-doped dielectric. TEM-EDS observation indicated that Yb exists at partial diffised areas and grain boundaries. Therefore, the stress might be originated from substitution of Yb into BaTiO,. It is assumed that increasing of Curie temperature and tetragonality could be due to this stress.
Y doped)
Yb doped
Fig.8 TEM micrograph of Y-doped and Yb-doped dielectric. Fig.9 shows the lattice constant of diffused area compositions for various Y and Yb doped BaT03+MgO system. Both diffised area compositions tends to become the cubic structure by the doping of rare-earth.
Dielectric Materials and Devices
479
Tetragonality of the Yb-difhsed area is smaller than that of Y-diffused area at same content of the rareearths. It is considered that the stress, probably the cause of interference fringes observed in E M ,have originated from miss-fit of lattice constant between diffused and non-diffused area. Thus, it is assumed that stress generated by the miss fit of lattice constant increases the tetragonalityand Curie temperature.
+I
s
' ; 3
8 0)
+:
2 pureBaTi0,
,
4.04r
4*021
''.AYb-dope
4.01
L Y ~ o p ea-axis
. I
3
0.0
...0
0.5
1.0
1.5
2.0
2.5
1 4
I
3.0
Content of rare-earth [mol%]
Fig.9 Rare-earth content dependence of lattice constant for diffusion area replica (BaTiO,+MgO+Rare-earth).
Finally, the electrical properties for newly developed XSR-MLCCs with Ni electrode are shown below. Specifications
20
-40
I
-50
0 50 100 150 Temperature ["C]
3216-0.lpF 16pm -45 Layers W.V=SOV 8 p u -9 1 Layers W.V=SOV 20 12-0.1pF Internal electrode is Ni Temperature-capacitance specification is X8R A C 6 1 5 % (T=-55°C~1500C) Permittivity q=I900 tan6 tan6$l% Insulation resistance I R 2 1.4*1010C2 CR-product CR=l6OOMQpF DC Break down voltage vBD=120V /pm HALT -3Ohr at 2OO0C-15V/p
Fig. 10 Newly developed X8R-MLCCs with Ni electrode. CONCLUSIONS 1. It is found that the temperature-capacitancecharacteristicsfor BaTi03-MgO-Rare-Earthsystem could
480
Dielectric Materials and Devices
be controlled using various ionic radii rare-earths. Especially, smaller ionic radii rare-earths such as Tm, Yb and Lu were effective dopants to realize the increase in Cure temperature. 2. Curie temperature shift towards higher temperature and the dispersal phase transition were observed
by DSC and hot stage XRD in the case of smaller ionic radii rare-earth doping. Raman spectroscopy also indicated the increase in Cure temperature in Yb doped dielectric. The tetragonal-cubic transition shows the primary phase transition mode.
4. Many interference h g e s , probably due to stress were observed for Yb-doped dielectric in TEM. It is considered that the stress originated fiom the miss-fit of lattice constant between diffised and nondiffused area compositions. Thus, It is assumed that the strong stress increases the tetragonality and the Curie temperature.
4. We have newly developed X8R-MLCCs with Ni electrodes for automotive application through the investigation of suitable formation and microstructure for Yb-doped BaTiO, system. REFERENCES IT. Nomura, A. Sat0 and Y. Nakano: J. Soc. Mater. Eng. Resour. Jpn. 5 (1992) 44 [Japanese]. 'T. Nomura, S . Sumita, Y. Nakano and K. Nishiyama: Proc. 5th US-Jpn. Seminar Dielectric Piezoelectric Ceramics, 1990 (Kyoto, 1990) p. 20.
'S. Sumita, M. Ikeda, Y. Nakano, K. Nishiyama and T. Nomura: J. Am. Ceram. Soc. 74 (1991) 2739. 4T.Nomura, A. Sato, A. Hitomi and Y. Nakano: J. Jpn. Soc. Powder & Powder Metall. 39 (1992) 590 [in Japanese].
'T.Nomura and Y. Nakano: Denshi Tokyo 31 (1993) 168. 6Y.Nakano, A. Sato, A. Hitomi and T. Nomura : Ceram. Trans.32 (1993) 119. *T. Nomura, A. Hitomi, A. Sat0 and Y. Nakano: J. Jpn. Soc. Powder & Powder Metall. 40 (1993) 677 [in
Japanese] 19( 1999)1061-1065. 9S.Sato,Y.Nakano,A.Satoand T.Nomura,J.Eur.Ceram.Soc.
Dielectric Materials and Devices
48 1
This page intentionally left blank
USE OF TITANATES TO ACHIEVE A TEMPERATURE STABLE LTCC DIELELCTRIC FOR WIRELESS APPLICATIONS Steve X. Dai, Rong-Fong Huang and David L Wilcox Sr. Ceramics Technologies Research Laboratory, Motorola Labs 7700 South River Parkway Tempe, AZ 85284 ABSTRACT A low loss and near zero temperature coefficient of resonant frequency (T,.) LTCC (low temperature cofired ceramic) host dielectric was developed for portable consumer wireless device applications. The low T, was realized by compensating the Al,O,-filled-glass dielectric with admixtures of TiO, (negative T, - temperature coefficient of dielectric constant) in the starting formulation. XRD data indicates a portion of the TiO, in the starting formulation dissolves into the glass and extensive formation of crystalline titanium compounds was observed via a nucleation and growth mechanism. The dissolution of TiO, in the glass and subsequent formation of titanium compounds was believed to result in the relatively small amount of TiO, required to achieve a near zero T, in the final sintered structure. INTRODUCTION Wireless communication is one of the fastest growing segments in the consumer electronics industry. The consumer wireless applications, especially the portable devices, demand material systems and processes that can achieve low cost, high performance, high reliability and light weighdfunction employing fabrication technologies that enable rapid prototype turnaround times. The radio frequency (RF) multilayer ceramic integrated circuit (MCIC) utilizing the IOW temperature cofired ceramics (LTCC) technology provides a viable solution for these challenges'. However, the unique requirement of RF frequency circuits operated in the 0.5 to 6 GHz range for most consumer wireless applications imposes significant challenges on the traditional LTCC material systems. In addition to the need of a low loss (high Q) substrate dielectric, it is desirable that the dielectric exhibit a near zero temperature coefficient of resonant frequency (T,.).This is because some of the key elements integrated in MCIC such as the filters and resonators need to be very stable under the fluctuation of the temperatures. The temperature coefficient of dielectric constant (T,) is commonly cited in the literature for capacitor dielectrics. However, the temperature coefficient of resonant frequency (T,.), which measures the shift of resonant frequency of a dielectric resonating circuit over a temperature change, is a parameter more directly useable by the device designers. A T, within +10 ppm/"C can in general provide high quality filtering and temperature stability. T, can be correlated to T, by the following equation
where a is the linear thermal expansion coefficient of the dielectric. a is typically in the range 3-15 ppm/"C for most ceramics. T, is clearly the dominant part in T,.. Adjustment of T,.towards 0 ppm/"C may be achieved by modifying the base composition with materials of opposite T,. For To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
483
example, materials with negative TE such as CaTiO, (TE =-1850 ppm/"C), SrTiO, (T, =-3000 ppm/"C) and TiO, (TE =-750 ppm/"C) can be added into LTCC dielectrics with negative T, to approach a T, of 0 ppm/"C. There are currently several types of LTCC material systems in terms of chemistry. The first type is based on a mixture of low melting glass with ceramic powder (i.e. alumina) as a filler for the dielectric tape'. In this system, the glass acts as a bonding agent to hold the ceramic particles together and little reaction between the glass and ceramic filler has occurred. The second type dielectric is made of re-crystallizeable glass'. During firing, the glass re-crystallizes to low loss phases and produces a low dielectric loss ceramic body. The loss factor of < 0.0004 at high frequency is not uncommon. This type of dielectric tape is suitable for 20-30 GHz applications such as in military and aerospace where very low loss is required. The third type of dielectric is also a mixture of low melting glass and ceramic filler. However, in addition to the role as a bonding agent, the glass is also designed to react with the filler ceramic at the sintering temperature to form high Q crystalline phases. In the second and third type LTCC dielectrics the microstructure, phases and final properties are controlled by the sintering conditions such as heating rate, sintering temperature and soaking time. An example of the third type dielectric is the Motorola advanced dielectric (designated as T2000 in this paper) for LTCC system4. T2000 dielectric is a glass-ceramic material system that includes a specially formulated lead-free glass, A1,0, as ceramic filler and Ti02 as T, adjustment agent. Based upon the fact that the residual glass is a dielectric Q limiting phase at high frequency in the sintered dielectric, this dielectric was designed to have a minimum amount of glass after sintering. The approach is to form a glass that first helps densification of the dielectric and later reacts with A1,0, to form high Q crystalline phases in a self-limiting crystallization process. Ingredients in the T2000 glass are B203, KzO, Si02, CaO, SrO and BaO. As depicted schematically in Fig. 1, during sintering the A1203 reacts with glass and forms anorthite type crystalline phases MSi2A1208 (M=Ca, Sr or Ba). The consumption of Si02, CaO, SrO and BaO greatly reduces the volume of glass in the final structure and results in high dielectric Q. Fig. 2 shows a high resolution transmission electron microscopic (HRTEM) image of a AlzO, particle in the sintered T2000. The dissolution of A1,0, into the glass and subsequent formation of crystalline phases via diffusion process are clearly demonstrated'. Details on the chemistry of the system could be found in references6. The sintered T2000 dielectric has a dielectric constant of -9.1 and dielectric Q in a range 1000- 1200 at 0.5 GHz.
Figure 1. Synthesis strategy of the T2000 dielectric.
484
Dielectric Materials and Devices
Figure 2. HRTEM image of an A1203 particle showing the dissolution of A1203 into glass and formation of the crystalline phases on the interface. T, modification by adding TiO, in the first type glass-ceramic dielectric has been reported by Yano et a1' in which the reaction of TiO, with glass was purposely minimized by coarsening the TiO, particles to surface area < 2m2/g via heat treatment. It was demonstrated in their work that about 15 wt% TiO, was necessary to reduce the original T,.of an undoped LTCC dielectric at -50 ppm/"C to near 0 ppm/"C. There is no report on the similar modification in the re-crystallizeable or reactive (type 2 and 3) dielectrics in which TiO, might react with the glass. The focus of this paper will be on T, modification by TiO, in T2000 dielectric. The correlation between the presence of titanium in the sintered structure and the T, will be addressed. EXPERIMENTAL PROCEDURE The dielectric tape of proper formulation was produced by a doctor blade technique. Fine TiO, powder with median particle size less that 0.5 pm and surface area greater that 4 m'/g was used. A stripline resonator was built by standard multilayer green tape processes for T, measurement. The stripline was designed to resonate at 1.3 GHz via a capacitive coupling to the U 0 pads. The green ceramic laminate was sintered at 875 "C for 30 minutes. The resonant frequency was measured via S21 parameter by HP8753D network analyzer. The resonator was placed in a Delta 2300 temperature chamber and cycled between 4 0 and 80 "C. Xray diffraction was performed on a Simens D5000 diffractometer. The scanning electron microscopy (SEM) analysis was primarily performed on a Leo field emission SEM (FESEM) and Princeton GammaTech EDS/F'"S x-ray spectrometer in Rutgers Univer~ity~'~. Samples were examined both high and low voltages. The transmission electron microscopic (TEM) was done on a Philips EM-420 microscope operating at an accelerating voltage of 120 kV. RESULTS AND DISCUSIIONS T, of T2000 Dielectric Figure 3 shows the change of resonant frequency of a stripline over temperature prepared by T2000 dielectric formulations with and without Ti02. The dielectric has a Tf around -80 ppm/"C without the addition of Ti02. The Ti02 modified dielectric shows a Tf close to zero near 25 "C. Tt over the entire temperature range is 4.2 ppm/"C. It is found that the Tf of T2000 can be continuously adjusted over a wide range, including 0 ppm/"C, depending upon the amount of Ti02 in the formulation.
Dielectric Materials and Devices
485
1> -
T. Measurement
1.248 1.246
-+- TIO
added
+No TiO
1.244
v
-40
-20
0 20 40 Temperature (“C)
60
80
Figure 3. Change of resonant frequency with and without Ti02 addition. T2000 Microstructure Observation Fig. 4 is a FESEM image of a well-polished T2000. Three “phase” constituents with different general chemistry (X-ray spectra) were identified. The microstructure is shown best at low voltage (1.63kV) with use of the In-lens detector. The three phases are (1) AI20, with particle size from 0.5-5 pm, (2) “glass” which is general areas of matrix, and (3) wedge shaped aluminosilicate crystalline phase. The identification is made on morphology and local chemistry by X-ray spectra. The A1,0, is straightforward and the glass is only qualified in that the chemistry appears to vary in terms of local K, Ca and Ti. The aluminosilicate crystalline phases were identified based on chemistry (high AVSi XRF count), the angular morphology of these regions, and relatively dark contrast to the glass and A120,.
Figure 4. FESEM image of polished T2000 at increasing magnification.
486
Dielectric Materials and Devices
The microstructure was examined with ET-SE/BSE (secondaryhack scatter electron), in-lens SE-I, and Oxford Tetra BSE detectors to determine if any titania particles were present. Many hundred particles each were observed for both polished and fracture surface samples at various firing conditions. Little direct image evidence of titania (Ti02) particles was found down to a size level of about 0.1 pm. However, small, local (< 1 pm) regions were found with higher Ti levels, suggestive of titania. One particle about 150 nm was seen with an extremely high Ti level and is probably a titania particle with other elemental signals contributing (overlapping) from both beneath the particle and around it. The results would be consistent with a minor level (a few percent) of sub-micrometer titania particles embedded in an aluminosilicate matrix. but this can not be concluded with technical uniqueness. Micro X-ray spectra were taken in these phases to examine the existence of Ti. It was found: (1) A1,0, - no detectable Ti (2) Aluminosilicate crystalline phases - No Ti within the detection limit. (3) Glass - Ti detected but the proportions of Ti, K, and Ca vary very widely over sample and even within a few micrometers. It appears that Ti is dissolved in variable amount in the glass, probably depending on the local chemistry of the original powder on firing. In addition, there would be low long-range chemical mobility of the glass provided that the glass becomes more viscous with dissolution of AI1O,and especially with A1,0, particulate. Quantitative Analysis of TiO, in T2000 A quantitative X-ray diffraction analysis (XRD) using CaF, as internal standard was performed to calculate the weight percentage of TiO, and AI,O, in sintered T2000. It is shown that" in a mixture of phases A, B, C, ..... with a fixed amount of the internal standard phase S IJJS = K-W, where I, is the intensity of a diffraction line i of a phase A; 1,: is the intensity of a diffraction line j of the internal standard S; K is a constant for a given crystalline structure, diffraction lines and a set of test conditions; w, is the weight percentage of phase A in original composition. A calibration curve can be prepared from measurement on a set of reference samples, containing known concentration of A and a constant concentration of the internal standard S. Once the calibration curve is established, the concentration of A in an unknown sample is obtained by simply measuring the ratio I,A/Ijsfor a composite sample containing the unknown and the same proportion of standard S. Fig. 5 is a XRD pattern of sintered T2000 powder with additional 20 wt% CaF,. Phases of MSi2A1208 (M=Ca, Sr, or Ba), Al,O,, TiO, and CaF, are marked in the picture. The overlap of phases in the low angle 28 region is obvious. The arrows in the graph point to the clean peaks of the A1203,TiO, and CaF,phases selected for calculation. Table 1. Compositions of the reference samples and measured weight Percentage of TiO, and AI,O, in both sintered and unfired T2000 Wn,, W,,,,? Wr.m 0 100 20 Ref 1 Ref 2 30 70 20 60 40 20 Ref 3 Ref 4 100 0 20 Sintered T2000 3.5 40.3 20 Unfired T2000 6.2 49.1 20
Dielectric Materials and Devices
487
Four reference samples with a fixed 20 wt% CaF, and Al,O,/TiO, ratio at 100/0, 70130, 60/40 and 0/100 were prepared to construct the calibration curve for both TiO, and A1?0,. Fig. 6 shows the calibration curves for TiO, and A1,0,. The curves showed linear relation between the intensity ratio and the portion of the TiO, and A120, phases. Table 1 lists the compositions of the reference samples and measured weight percentage of TiO, and Al,O, in both sintered and unfired T2000. The unfired T2000 with known formulation is measured to calibrate the measurement. The TiO, in sintered and unfired T2000 is measured about 3.5 and 6.2 wt%, respectively. The amount of TiO, in the unfired T2000 is within &0.3wt% of the actual amount of TiO, in the formulation. The results suggest that a substantial amount of TiO,, about 45-50 %, is dissolved into the glass during sintering. The particle size of the residual TiO, in the T2000 could be so small that TiO, is hard to be identified conclusively by the SEM examination. The fact that about 9 wt% A1,0, dissolved into the glass and subsequently form high Q crystalline phase is consistent with the microstructure observation in Fig. 2 and Fig. 4.
Figure 5. XRD pattern of T2000 with 20 wt% CaF,. The arrows point to the peaks selected for quantitative calculation.
Figure 6. Calibration curves for A1,0, and TiO,.
488
Dielectric Materials and Devices
Existence of TiO, in Sintered T2000 In a morphological examination of T2000 by transmission electron microscope (TEM), A1,0, particles were seen throughout the sample. However, careful search revealed regions where TiO, particles were identified. Fig. 7 is a bright field image of such a region in T2000. A coexistence of TiO, particles with a amorphous matrix and wedge-shaped MSi2A1208 (M=Ca,Sr,Ba) phase was observed. The size of TiO, particles is in 50-300 nm range. The TiO, particles were identified by both selected area diffraction pattern (SAED) and energy dispersive spectrum (EDS). The [OOl] SAED pattern showed typical lattice diffraction of a TiO, crystal. The EDS conclusively confirmed the nature of the particles.
Figure 7. Bright field image of T2000. TiO, particles are identified by morphology, selected area diffraction pattern (SAED) and energy dispersive spectrum (EDS). MASO=(Ca,Sr,Ba)Si2A1208
EDS examination of the amorphous matrix also showed the existence of Ti, as shown in Fig. 8. The relative amount of Ti as well as other elements fluctuates from region to region, reflecting the change of local chemistry due to the viscous glass reaction at the sintering temperature. The existence of Ti in the glass matrix clearly demonstrated the dissolution of Ti02 into the glass, in consistent with the results from the quantitative XRD analysis. The well-defined edges of MSi2A1208(M=Ca,Sr,Ba) crystalline inclusions in the glass matrix (Fig. 7) indicate the growth of crystals along preferred crystallographic directions. The formation of the crystalline phases can significantly reduce the volume of the residual glass, contributing to the overall higher Q of the T2000 dielectric. In addition, the formation of the crystalline phases also reflects the importance of T2000 sintering in order to control the kinetics to maximize the desired phases while having minimum impact on other properties.
Dielectric Materials and Devices
489
Figure 8. EDS of the glass matrix in T2000. Existence of Ti is clearly seen. Ti-rich Crystalline Phases via Nucleation and Growth Mechanism Extensive TEM search of the T2000 also revealed the existence of plate-like, Ti-rich crystalline phases, as marked in Fig. 9. The crystallinity of the phases was reflected in the SAED patterns in the picture. The amorphous ring in the SAED pattern shows the coexistence of the plate-like phases in glass matrix. It is also obvious from the EDS spectrum that there is a significant amount of Ti in the crystalline phases as well as the embedded glass matrix. The Ti peak is found much higher than the average T2000 dielectric.
Figure 9. Bright field image of T2000. Ti-rich plate-like phases were identified by SAED patterns and EDS.
490
Dielectric Materials and Devices
It is suggested that during sintering part of the TiO, dissolves into the glass matrix and acts as nucleation agents for the subsequent growth of the Ti-rich crystalline phases. The process would likely occur in the regions rich in Ti where TiOl particles originally exist. Coexistence of the (Ca,Sr,Ba)SizAl2O8phases is also seen next to the Ti-rich plates in Fig. 9, indicative of a formation of the conventional high Q phases in T2000 in the Ti-lean regions.
Figure 10. EDS of two different Ti-rich plates. The composition fluctuates significantly from plate to plate. Attempts trying to analyze the SAED pattern in Fig. 9 do not results in conclusive crystalline structure of the Ti-rich plates. Fig. 10 shows EDS patterns of two Ti-rich plates. which is representative of all the plates examined. Fluctuation of Ti-content as well as other elements are obvious. It is concluded that the composition of Ti-rich plates depends on very much the local chemistry of the T2000 which is inhomogeneous through out the samples.
T, Compensation in T2000 Dielectric
Due to the multiple forms of the titanium existence in T2000 dielectric, the compensation mechanism of the TiO, to T, could be significantly different from that of a TiO, addition to a nonreactive LTCC dielectric. The authors propose that the T, of T2000 is adjusted by a combination of the residual TiO, which is around 3.5 wt%, the Ti in the glass matrix, and the Ti-rich plate-like crystalline phases. The dissolved Ti and subsequently formed Ti-compounds probably have much higher negative T, value than TiO,, contributing to more adjustment of T,. This is probably the reason that the modification on T, by TiO, in T2000 is so effective, only less than half of the TiO, is necessary to achieve a near 0 ppm/"C T, comparing to the amount required for the type 1 LTCC dielectrics in which TiO, remains unreacted. CONCLUSIONS The T, of T2000 dielectric can be reduced to close to 0 ppm/"C by incorporating a relatively small amount of TiO, into the composition. The TiO, dissolves into the glass during sintering and forms Ti-rich-plate-like crystalline phases via a nucleation and growth mechanism. This process makes the T, modification more efficient than a simple mixing of TiOl into the composition. The T, of T2000 is believed to be compensated by the multiple forms of Ticompounds in the sintered T2000 dielectric. The dissolution of TiO, into the glass and the consequent effect on T,.reduction is a unique feature of the T2000 dielectric, and is first reported in current study.
Dielectric Materials and Devices
49 1
ACKNOWLEDGEMENT The authors would like to thank Dr. Zhengkui Xu at the University of Illinois for part of the TEM analysis. The authors would also like to thank Prof. Victor Greenhut at the Rutgers University for the access of the FESEM. REFERENCES ‘D. L. Wilcox, R.F. Huang, and D. Anderson, Proceedings of 1997 ISHM, Philadelphia (1997) 17. 2A.L. Eustice, S.J. Horowitz, J.J. Stewart, A.R. Travis, and H.T. Sawhill, 36“’ Electronic Components Conference, Seattle, WA, (1986) pp37-67 3J.H. Alexander, S.K. Muralidhar, G.J. Roberts, T.J. Vlach, in Proc. Int’l Sym. on Microelectronics, Orlando. FL, (1991) pp414-417. 4Shelly Bethke, Ross Miesem, Wayne Chiou abd Richey Pastor, “Ceramic Composition”, U.S. Pat. No. 5 821 181, Oct. 13, 1998. ’David Wilcox, Rong-Fong Huang and Steve Dai, “Enabling Materials For Wireless Multilayer Ceramic Integrated Circuit (MCIC) Applications”, Ceramic Transaction, Vol. 97,201-213 (1999). %teve Dai, Rong-Fong Huang and David Wilcox, “Temperature Stable, Low Loss and Low Fire Dielectric for Consumer Wireless Applications”, Proceedings of the lst China International Conf. On High Performance Ceramics, Oct.3 1-Nov.3, 1998, Beijing, China. ’Shintuka Yano, Nagoya Hirofumi, Yamaguchi Komaki and Takami Hirai, “Distributed Constant Circuit Board Using Ceramic Substrate Materials”, U.S. Pat. No. 5 232 765, Aug. 3, 1993. ‘V. A. Greenhut and J.J. Friel, “Application of Advanced Field Emission Scanning Electron Microscope and Energy Dispersive Spectroscopy to Ceramic Materials”, Microscopy and Analysis, March 1997. ’John Friel and Victor Greenhut, “Novel technology for X-ray Mapping of Ceramic Microstructure”. Journal of the American Ceramic Society, 80 [12], 3205-208 (1997). “B.D. Cullity “Elements of X-Ray Diffraction”, 2’IdEdition, Addison-Wesley Publishing Company. 1978.
492
Dielectric Materials and Devices
LOW TEMPERATURE CO-FIRED CERAMICS AND ITS APPLICATIONS Kikuo Wakino Murata ManufacturingCo., Ltd 2-26- 10, Tenjin, Nagaokakyoshi, Kyoto, 617-8555 Japan
Harufimi Mandai Murata Manufacturing Co., Ltd 2-26-10, Tenjin, Nagaokakyoushi, Kyoto, 617-8555 Japan
Norio Nakajima Murata Manufacturing Co., Ltd 2288, Oshinohara, Yasu-cho, Shiga, 520-2393 Japan ABSTRUCT Recent progresses of communication-, information-, and data processing-systems request ultra high speed and high-density signal process technology. Of cause silicon LSI is playing the leading part of this demand, but fast and less distorted signal transmission capability and reliable packaging are also key issue to meet the crucial requirement for the intelligent era. Multilayer MCM that based on the low temperature cefired ceramics is one of the most promising technologies to adapt to these new age. In this review, the recent status and trend of Low Temperature Co-fired Ceramics (LTCC) and its applications will be described. INTRODUCTION High-density multilayer package for the high speed CPU of host computor, using Alumina and Molybdenum-Manganese-Tungsten system, was developed and put into practical use by IBM in 1970s. This package was powerful and reliable for the complex and fast computing system, but too expensive for consumer applications. The advent of less expensive mu1tilayer package with highly conductive inner electrical connection was expected to fulfil1 the demand of fast growing advanced social systems, such as; multimedia network home system, mobile communication, highly integrated social system, ultra high speed computing system in hctory and so on. Many kind of LTCC materials have been developed to use the high conductive electrodes such as copper or silver combining with multilayer technologies. Around 1980, LTCC were practically adopted for main frames of the super-computer. As to contribute to realize more compact and lighter Cellular terminals, LTCC multilayer LC filters were acknowledged as “promising’ solution and designed in for the first time in early 1990s. LTCC have started with low dielectric constant (5 10) materials, thereafter higher dielectric constant (20-60) materials were also developed to reduce the device size for low GHz fiequency range application. Chip monolithic type couplers, b a l m and even antennas using the said LTCC multilayer technologies were developed one after another and increasingly adopted by almost all cellular phones in the world. In addition to those chip monolithic type devices, the same LTCC multilayer technologies have been also applied to develop firther integrated RF devices with
-
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property
of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the Copyright Clearance Center, is prohibited. the express written consent of The American Ceramic Society or fee pa!iot
Dielectric Materials and Devices
493
more sophisticated functions using Transistors and ICs. The main reason why LTCC multilayer technologies have advantage in FW area is that; lower parasitic capacitance or coupling design is achieved using the lower dielectric constant material; less inductive connection is attained with shorter wiring such as via holes; higher temperature stability is attainable using NPO material; tremendous miniaturization is realized by 3D design that embeds small elements in the ceramic multilayer substrates. In this report the recent progress of LTCC materials, multilayer substrate, MCM and their applications especially about cellular phones will be presented. LOW TEMPERATURE COFIRABLE CERAMICS Various kinds of LTCC materials have been developed so fhr. Roughly, they can be classified; (a) basic material wise, glass system and non-glass system; and (b) conductor material wise, copper system and silver system. Table. 1 shows the representative materials and their properties announced fiom several manufactures.
Table. 1 LTCC compositions and characteristics
q
27xld
494
294
132
4 . m d 3Mob
274
193
a06
245
167
5.Md
7.7xl0(
7.Md
3Md
Dielectric Materials and Devices
Table. 1
(continued)
Gass
I
+zhQo.sio,
I 3.1 xlO3 167-196
157
3.0-7.0
11.6
6.8
2.5 6.1
2.0 15.0
7
4
20
20
>10l2
>1O1O
>10l2
OJ
OJ
cu
-I---+7.Wd
>1O1O
5.9~10~
>lolz
k P g ltemalculductor4l
Table. 2 shows specific characteristics of BAS ceramics in more detail; which is mainly consisted of Ba-, Al- and Si-oxides and co-fireable with copper electrodes under nitrogen atmosphere. BAS has relatively low dielectric constant, small TC of dielectric constant, reasonably small in tan 6 and large insulation resistance. It is also reasonably sturdy for multilayer substrate. Table.2 Dielectric Properties of (BAS) Material Basic Formulation Relative Dielectric Constant (at 1MHz) Q (at IMHz) Q (at 5GHz) T. C. of Dielectric Constant Resistivity Dielectric Strength Flexural Strength Firing Temperature Firing Atmosphere
Dielectric Materials and Devices
BaO,Al2O3,Si02 6.1 1400 300 +60 p p d C >1014 Bcm >10 MV/m 160000 N/m2 980 "C Nitrogen
495
Manufacturing process of multi-layer LTCC devices or modules are basically similar with those of multi-layer ceramic capacitors. The major differences between MLC processes are; the patterns of multi-layer substrate or package are complicated, different in design fi-om layer to layer to perform different functions in each layer; use of number of via holes to realize the shortest connections between fitnctional sub circuits; resistive paste andor magnetic materials are applied in some case as well as with dielectric. AVERAGE DESIGN RULES Finer wiring can realize smaller device design but on the other side it gives the disadvantage; higher electrical resistance; higher parasitic capacitance; lower yield and so on. The commonly used average design rules to design the multi-layer LTCC devices or MCM are briefly summarized in Table.2. Several RF and Microwave simulators, such as MDS, HFSS, XFDTD, Momentum, Sonnet and others will be effectively used to shoot the first approximated or trial design. Usually, only a little corrective action is required to reach the final desired design. For the time saving, these tools are quite useful and effective. Table.3 Design Rules Length & Width Total Thickness Layer Thickness Line and Space Conductor Via Hole Diameter Electric Conductivity Printed Resister Resistance Range Tolerance TCR Buried Capacitor Capacitance To1erance TCC Impedance Strip Line Substrate
20mm x 20mm max 2.0mm max. 50 p Nstd.) 100 p d 1 0 0 p m min. 130 m(std.) 4 x 10+7s/m 5Oohm- 1OOKohm t5% t 3 0 0 ppmpc 1pF/mm2/layer -+ 5% 80 2OppdC 5Oohm std. 1OOohm max. _+
DEVICES Although LTCC devices; such as multilayer semi-dntributed LC-filter, coupler and balun are not so highly intelligent, but these simple and some what clever devices are most widely used and playing important roles in sophisticated systems as a knctional module of small size, lightweight and tuning free. Fig.l shows the structure and the equivalent circuit of 3dI3 coupler. The performance of LC-filter for the latest ETC system; 5.8GHz is shown in Fig.2. The geometry of this device is 2.0~ 1 . 5 1~.Omm.
496
Dielectric Materials and Devices
Fig. 1 Structure and equivalent Circuit
Fig.2 LC filter characteristics for ETC
Dielectric Materials and Devices
497
Fig.3 The equivalent circuit of diode switch
MCM In analog system, dielectric ceramic duplexer has been mainly used for the antenna filter, but in digital system, like GSM, diode switch can be used in stead of the duplexer. The diode switch has the advantages; lighter, smaller and cheaper than duplexer. Therefore, almost all GSM cellular phones are using it now, Equivalent circuit of diode switch is shown in Fig.3, As a low pass filter, strip lines and capacitors are integrated in the ceramic substrate, two diodes and one resister are mounted on the ceramics. Regarding active element, Si bare chips are mounted on the ceramic surface with die bond and wire bond technology, or Si flipchip is mounted using flipchip technology, Beside these, filters and other peripheral circuit are integrated in the ceramics. Fig.4 shows a concept of 3D integration. ANTENNA The whip antenna and the inverted F-type antenna are commonly used for the cellular phones so kr. But both of these antennas are not SMD type and little bit large in size. Chip type antenna has been introduced to cellular phones using multi-layer LTCC technology; as a main antenna for PHS and diversity antenna for PDC; both for Japanese cellular systems. Fig.5 shows the typical appearance of chipantenna. Helical coil is embedded in the ceramic block. This miniaturized chip antenna is expected to design in Bluetooth; and as the key component for wireless LAN.
498
Dielectric Materials and Devices
Fig.4 A concept of 3D integration
The radiation patterns of the chip antenna placed in the PHS as a main antenna are shown in Fig.6. The chip antenna shows same sensitivity level with currently used whip antenna, although the dimension of ground of PCB gives a great influence
Dielectric Materials and Devices
499
Fig.6 The Radiation patterns of the chip antenna
CONCLUSION Chip type filters, baluns, couplers, diode switches, antennas and MCMs using multi-layer LTCC technologies were developed and widely adopted by almost all cellular phones. Multi-layer LTCC technologies are going to be expected as a most permissible technologies for the intelligent traffic control system as well as mobile wireless communication equipment. The main reasons why multi-layer LTCC technologies have advantage in RF area are as follows. 1.lower parasitic capacitance or coupling design is achieved using the lower dielectric
constant material 2.Less inductive connection is attained with shorter wiring such as via holes 3.Higher temperature stability is attained using NPO material. 4.Tremendous miniaturization is realized by 3D design that embeds small elements in the ceramic multilayer substrata.
500
Dielectric Materials and Devices
REFERENCES [l] Yuzo Shimada, Yoshinobu Kobayashi, Keiichirou Kata, Masayuki Kurano and Hideo Takamizawa: "Large Scale Multilayer Glass-Ceramic Substrates for Supercomputer", E E E Trans. on Components, Hybrids and Manufacturing Technology, Vol. 13, No. 4, Dec. 1990 [2] William A.VitroI and Lerry L. Steinberg; "Development of a Low Temperature Co-fired Multilayer Ceramic Technology", The second International Journal for Hybrid Micro-Electronics, Vol. 6, No. 1. 1983 [3] Troung Dinh Than, Nobuo Iwase, Hanrtoshi Egami and Eikichi ichimori; "Low Temperature sintered Ceramics for Hybrid Functional Circuit Substrates"; IMC 1984 Proceedings, Tokyo [4] K. Yokouchi, N Kamehara and K. Niwa: "Packaging Technology for High-speed Computers
Multilayer GlasdCeramic Circuit Board"
ISHM '9 1 Proceedings
151 Haruhmi Mandai,
Kimuhide Sugo, Kazuyoshi Tukamoto, Hiroji Tani and Michihiro Muraw "A Low Temperature Cofiered Multilayer Ceramics Substrate Containing Copper Conductors", IMC 1986 Proceedings 161 N. Ichinose; Electronic Ceramics, Vol. .8, 1991
[7] Harufbmi Mandai, Kikuo Wakino, Hisatake Okamura: " A Low Temperature Cofiered Multilayer Ceramic L-C Filter with Copper Conductors", Ceramic & Science Technology, Congress I989 Proceedings [8] Takahiro Watanabe, Koji Furutani. Norio Nakajima and Harufbmi Mandai; "Antenna Switch Duplexer for Dual-band Phone (GSMiDCS) using Multilayer LTCC Technolo&'; 1999 EEE MTT-S INTERNATIONAL MICROWAVE SYMPOSIUM DIGEST
Dielectric Materials and Devices
50 1
This page intentionally left blank
TEMPERATURE-DEPENDENT POLARIZATION AND ELECTRIC POTENTIAL ON FERROELECTRIC BaTi03 (100) SURFACES Sergei V. Kalinin and Dawn A. Bonnell Dept. Mat. Sci. Eng., University of Pennsylvania, 323 1 Walnut St. Philadelphia, PA 19104 ABSTRACT Variable temperature atomic force microscopy (AFM), electrostatic force microscopy (EFM), scanning surface potential microscopy (SSPM) and piezoresponse imaging (PRI) were used to study model BaTi03 (100) surfaces. A systematic comparison of local measurements with analytical and numerical models provides the basis to quantify local materials properties. Domain induced surface corrugations and piezoelectric response were found to disappear above the Curie temperature. Surface potential was found to decay slowly after the transition. Temperature induced changes of topography, surface potential and piezo response below and during the ferroelectric phase transition were used to elucidate contrast formation mechanisms. The experimentally determined temporal response of surface charge and local deformation are discussed within existing models of ferroelectric screening. INTRODUCTION The development of spontaneous polarization and related lattice distortions below the Curie temperature of a ferroelectric material results in regions of uniform polarization, i.e. ferroelectric domains. Depolarization energy limits the size of the domains giving rise to complicated domain structures. The formation and static properties of domain structures in bulk crystalline ferroelectrics have been extensively studied since the discovery of ferroelectricity at the eve on the century and are well-under~tood.'-~ However, polarization and domain behavior in the vicinity of surfaces or interfaces as well as dynamic properties of domain structures are less well understood. In particular, a vast number of experimental and theoretical studies have addressed the issue of the "dead" layer at a
To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
503
ferroelectric surface, usually attributed to the formation of surface layer with a low dielectric constant. It has also been realized that surface polarization charge due to the normal component of the polarization vector must be screened by electronic states, charge transfer to the electroded surface, or band bending and formation of a space-charge layer similar to semiconducting materials. A detailed discussion of screening phenomena in ferroelectric semiconductors is given by Fridkir~.~ A closely related phenomenon is polarization behavior in the vicinity of electroactive interfaces, in which interplay between ferroelectricity and local charge at the interface gives rise to positive temperature coefficient of resistance (PTCR) behavior. Dynamic properties of ferroelectric domain structures which govern the hysteresis, piezoelectric and fatigue behavior of ferroelectric materials are traditionally studied on bulk samples, i.e. average properties are determined. Only a small number of in-situ experiments on domain wall motion under applied lateral bias or ferroelectric phase transition by optical or electron microscopy have been r e p ~ r t e d . ~ - ' ~ Contact and intermittent mode atomic force microscopy along with lateral force microscopy has been widely used to characterize domain-related topographic features.16-24 Direct information about local polarization, charge 'distribution, and electromechanical properties of surfaces can be obtained by such techniques as electrostatic force microscopy (EFM)25-29and piezoresponse imaging (PRI).30-34 The contrast formation mechanism in many variants of SPM is still unclear due to the complex nature of tip-surface interactions. Attempts to quantify non-contact SPM of ferroelectric surfaces utilize the assumption that the polarization charge is unscreened and SPM contrast is attributed to the normal component of electric field above the surface. However, theoretical arguments suggest that polarization is screened on a ferroelectric surface. Hence, in the present research static and dynamic behavior of ferroelectric domain structure (i.e. domain wall motion and phase transition) are studied to determine the nature and properties (e.g. mobility and relaxation times) of surface charges. EFM and SSPM images of ferroelectric domain structures are quantified in terms of screened surface polarization. EXPERIMENTAL DETAILS The AFM and SSPM measurements were performed on commercial instrument (Digital Instruments Dimension 3000 NS-III). Both conventional silicon tips (e = 125 pm, resonant frequency 270 kHz) and metal coated tips (t =: 225 pm, resonant frequency - 60 kHz, k = 1-5 N/m) were used. To perform piezoresponse measurements, our AFM was additionally equipped with a Wavetek function generator and SRS830 lock-in amplifier. W2C coated tips35(e =: 125 pm, resonant frequency 350 kHz) were used for these measurements. These
-
-
504
Dielectric Materials and Devices
tips can also be used for SSPM measurements, however, due to large spring constant ( k =: 40 N/m) EFM imaging is complicated. Variable temperature measurements were performed on a home-built heating stage. During measurements, the temperature was increased in steps of -10°C and the system was kept at the selected temperature for -0.5 h in order to achieve thermal equilibrium. The cantilever was re-tuned at each step in order to stay in the vicinity of the resonance frequency. Thermal drift was corrected by adjusting lateral offsets to position domain-unrelated topographical features. The lateral displacements of the tip with respect to the surface were usually 2-3 pm per 10°C, except in the vicinity of the Curie temperature, where the ferroelectric phase transition was accompanied by significant (- 10 pm) lateral displacements. A barium titanate (100) single crystal ( 5 x 5 ~ 1mm, Tc = 130°C, Superconductive Components, Inc) was used on which the roughness of the (100) face did not exceed 15 A. Prior to analysis the crystal was repeatedly washed in acetone and deionized water. In order to obtain a reproducible well-developed domain structure the crystal was heated above the T,, kept at 140°C for -0.5 h and cooled down on a metallic surface. Static results
Figure 1. Surface topography (a,b), surface potential (c,d) and schematics of domain structure (e,f) in a-domain region with c-domain wedges (a,c,e) and in c-domain region with a-domain wedges (b,d,f).
Dielectric Materials and Devices
505
The surface domain structure of a ferroelectric surface can, in some cases, be unambiguously determined by SPM. Surface topography in ferroelastic materials is directly related to the misorientation angle between domains with different polarization directions, e.g. for tetragonal perovskites the corrugation angle, 8, associated with 90" a-c domain walls is 8= .n/2-2arctan(a/c), where a and c are the parameters of the tetragonal unit cell. Complimentary information on surface potential obtained by non-contact (SSPM, EFM) or contact (PRI) SPM allows the orientation of polarization vector (e.g. c+ - c- domains) to be distinguished, thus providing a reconstruction of surface domain structure. Polarized light microscopy, AFM and SSPM allowed the following major types of domain structures to be characterized. The central part of the crystal is formed by large lamellar domains oriented ,at 45" to the edges of the crystal. The absence of significant topographic and potential variations allows this domain structure to be ascribed to al-a2 domain arrangements. Close to the edge of the crystal regions with a-c orientation are present. If the size of the c-domains is relatively small, then 180" walls perpendicular to 90" domain boundaries between a and c domains (Fig.1 a,c,e) are formed. Similar domain arrangements are reported elsewhere.36This domain pattern can be ascribed to c domain wedges in the crystal with dominating a domain structure. The formation of 180" walls within the wedge minimizes the depolarization energy. If c domain regions are large (Fig. 2 b,d,f), irregular 180" walls separating c+-c- domains exist. These walls are continuous through a domain regions, indicating the presence of a wedge domains in preferentially c domain material (Fig. 1 b,d,f). More complex domain structures can also be observed. Fig. 2 shows the boundary between regions with al-a2 (left side) and c+ - c' (right side) domain arrangements. The optical micrograph clearly indicates the presence of a 1-a2 boundaries (left). Minor lines (right) can be observed only for small focus depths indicating a nearsurface character. Large scale AFM imaging indicates that large surface corrugations (Fig. 2 a) are associated with the presence of 90" domain walls. The measured corrugation angle 8 =: 0.62" is very close to calculated value (8= 0.629"). The surface potential indicates that the left region of the image is not associated with significant potential features, while clear c+-c- domain regions are present on the right side. Noteworthy is that small horizontal potential features are also observed on the SSPM image. Shown in Fig. 2 d,e is the enlarged scan of the right region. Surface corrugations corresponding to the 90" domain walls are now clearly seen (note the difference in vertical scales between Fig. 2 a and d). The surface potential image from the same region (Fig. 2 e) shows both potential features corresponding to surface a-c domain structure and bulk c+-c- domain arrangement. Thus, surface potential measurements on ferroelectric surfaces in certain cases provide information about surface and bulk domain structures. In this particular case domain contrast originates from surface polarization as well as
506
Dielectric Materials and Devices
Figure 2. Surface topography (a,d), surface potential (b,e), domain structure reconstruction (c) and polarized light optical micrograph (0 in the region with complex domain arrangement. Scale is lOOnm (a), 10 nm (d), 0.2 V (b,e).
Figure 3. Surface topography (a), surface potential (b) and EFM images of BaTi03 (100) surface at tip bias of 5 V (c), 0 V (d), -2 V (e) and -5 V (0. Note the inversion of domain contrast with tip bias and abnormal image at large negative bias.
Dielectric Materials and Devices
507
from bound screening charges on charged domain boundaries. This surface domain structure probably relieves the strain in the near-surface layer associated with macroscopic 90" domain wall between with al-a2 and c+-c- domain regions. Surface topography, surface potential (SSPM) and force gradient (EFM) images of a similar region are compared in Fig. 3. Note that for positive tip bias (Fig. 3 c) the EFM image is similar to the SSPM image. For negative tip bias the EFM image is inverted, as expected. Noteworthy is that for zero tip bias the EFM image has the same sign as for a negatively biased tip, indicative of positive surface potential. For large negative biases the EFM image is unstable as seen on Fig. 5 f. It is yet unclear if this effect should be attributed to a feedback loop instability or tip-induced desorption or charge transfer in the surface layer. In order to minimize the influence of this effect on effective surface properties, quantitative measurements were performed well inside the stability region. Ferroelectric phase transition
Figure 4. Surface topography and potential distribution at BaTi03 (100) surface before ferroelectric phase transition at 125°C (a,b), 4 min after transition (c,d) and after 2.5 h annealing at 140°C (e,f). Apparent intensity differs due to the different scale (0.1 V for (b), 0.5 V for (d) and 0.05 V for (0).Note that the sign of surface potential features does not change during the transition.
Screening on ferroelectric surfaces can be studied by variable temperature measurements of surface properties. Above the Curie temperature spontaneous polarization disappears, as evidenced by the absence of surface corrugations. This
508
Dielectric Materials and Devices
is also confirmed by variable temperature piezoresponse imaging.37At the same This effect time, SSPM indicates a spurious increase of potential amplitudes.38139 is ascribed to the release of uncompensated screening charges after spontaneous polarization disappears above Tc. Noteworthy is that the sign of potential features remains the same after the transition (Fig. 4). This implies that the sign of domain related potential features are determined by the screening charges and is opposite to that expected from polarization orientation. Domain wall motion The relationship between polarization orientation and surface potential can also be established from the observation of domain wall motion. Fig. 5 shows SSPM images of c+ - c- domain structures obtained at a 12 h interval. It can be seen that the shrinking of the negative domain results in a dark rim in the direction of domain wall motion. It should be noted here that shrinking occurred spontaneously rather than under applied tip bias or lateral bias, and thus charge transfer from tip to the surface is minimized. The formation of the rim is ascribed to the slow relaxation of screening charges after the displacement of domain wall.
Figure 5 . Surface potential images of c+-c- domain region BaTi03 (100) acquired at 12 h interval (a,d), corresponding average profiles along the boxes (b,e) and the scheme of surface charge distribution (c,f).
Simple considerations (Fig. 5 c,f) imply that formation of a negative rim in the direction of wall motion is possible only if domain related potential features are determined by the screening charges. Model and fields In quantification of electrostatic SPM on ferroelectric materials the majority of authors assumed that a ferroelectric surface is characterized by unscreened
Dielectric Materials and Devices
509
polarization charge density CJ = P - n , where P is the polarization vector and n is the unit normal to the surface. However, it is well known that polarization is usually screened on a ferroelectric surface4. The screening can be due to adsorbates and/or surface states, or free charges and formation of depletion or accumulation layers. In the latter case an additional constraint is that the electric field in the surface layer can not exceed coercive field. Thus polarization and charge distribution in the surface layer is rather complex. For extrinsic screening adsorption results in surface double layer and surface polarity depends on spatial localization of polarization and screening charges and the degree of screening. Electrostatic force miscroscopy We assume (and it is usually true) that domain size is comparable or larger than characteristic tip size, but much smaller then the cantilever size. Thus we assume that tip interacts with single domain, and cantilever "feels" the average surface potential. In the following discussion we assume that the polarization charges are almost completely screened by surface adsorbates and/or free carriers equivalent to formation of double layer characterized by potential V,. The electrostatic force between the tip and the surface in the absence of unscreened charge is
w=
(vtip - v s ) 2 ~ t ( z ) + ( v , i p -Vuv)2Fc(d (1) where F,(z) is the tip contribution and F&) is the cantilever contribution equal to the derivatives of corresponding capacitances. Force gradient can be derived from Eq. (1) and after grouping the terms
F'(z) = vt; {Ft' + F,'}+ vtip{- 2vs Ft' - 2VUVF,'}+V,2Ft'+ v,2,F,' The average force gradient determined as the average over all image points can be obtained as or
F& ( z) = vt; {F; + F,' } - 2Vt@vuv{Ft' + F,'
} + v,2,(Ft' + F,' )
(2)
(3)
(4) provided that the image size is large compared to domain size. From Eq. (2) the force gradient difference between regions with surface potentials Vl and V2 (i.e. between domains with different polarities) can be found as G
or
(2)
=
+W t i p +
A0
~2( z) = -2vtip (v,- v:!)F; + (v?- V; )F; F;
510
V
( z )= Bytip + Bo
(5) (6)
Dielectric Materials and Devices
Provided that the experimentally determined average force gradient and domain force gradient are quadratic and linear in voltage respectively (Fig. 6), the constants A2, A1 and B1, Bo can be extracted. 100-
2
50-
e a
0-
s 0
H -503
f
a ! v
-P
U.
c
*p.
-100d
-150.
.
I
.
1
.
I
.-5) 0
. =\
%
-
i
.
I
. b.
A, (Dull)
.
Bl (Dull) Bl (Sharp)
.
1;
0 0
. .
U
-
.-c E
-
.
3 1°! . I
LH=126nm LH=56nm
A
-
5 0.1 ;
-I
Figure 7. Coefficients sharp tip.
(b)
A2
100 Distance, nm
I
1000
(a) and B 1 (b) as a function of tip-surface separation for blunt and
In order to quantify the distance dependence of EFM data, the coefficients and B1 for tips used were determined as a function of tip-surface separation and are shown of Fig. 7 for sharp and dull tips. These dependencies can be well linearized in log-log coordinates. Noteworthy is that effective slopes are larger than those expected for cone tip model (-1) and smaller than expected for s here tip model (-2) in agreement with previous studies on different systems! As expected, the effective slope is smaller for sharp tip, since the contribution of tip bulk (cone contribution) is larger in this case. In fact, the effective slope for average force gradient for a sharp tip is almost equal to unity, implying that a cone model can be used to describe the capacitive interaction in this case.
A2
Dielectric Materials and Devices
511
Table I. Distance dep
main (D) frequency shifts. Tip
effective slope--
Dull A
-1.17 f 0.04
Sharp A
-1.02 f 0.05
Dull D
-1.41 f 0.02
Our estimates4' suggest that the cantilever contribution,
Fi,
can be
neglected compared to F; for intermediate tip-surface separations. In this case in the absence of a Coulombic contribution from unscreened charges the coefficients in Eqs. (4,6) yield the following universal ratios:
Bo --Vl +v2
. A1 -= -2v,, B1 -2 ' A2 These ratios are independent of the cantilever properties. Provided that the contribution of uncompensated charge on the ferroelectric surface is negligible, these ratios are distance-independent. Conversely, if these ratios are distance independent, then the observed contrast between domains of different polarity can be attributed to the double layer contrast without a free charge contribution. Experimentally it was found that the ratios BI/A~and Bo/Bl for sharp and dull tips are distant independent for small tip surface separations (z < 100 nm). For larger tip-surface separations the measurements are small compared to typical noise and drift in force gradient (-0.1-1Hz) and correspondingly errors in fitting are large. Using Eqs. (7a,b,c), the absolute potentials of domains of different polarity with respect to the tip are calculated as V1 = 668 mV (dull), 628 mV (sharp), V2 = 533 mV (dull), 473 mV (sharp) and shown in Table I1
B1 = -2(v1
A2
-v2),
= ..-Bl/A2 (Vl+V2)/2 = -Bo/Bl .................T !I? ......................................2.........(Vl-V,) ...... ..... ......................... ........................ ........................................ .............,..,..... ..................... Dull 0.27 f 0.03 0.60 f 0.08
....
....
Vav
= -A1/2A2 ....._..
0.53 It 0.05
Sharp c+-c-
0.31 f 0.04
0.55 f 0.09
0.60 f 0.07
Sharp a-c
0.17 k 0.02
0.63 f 0.09
0.60 k 0.07
The potential difference between c+ and c- domains (independent of tip properties) is correspondingly 135 mV (dull) and 155 mV (sharp). Noteworthy is that the average image potential VaVis approximately equal to (Vl+V2)/2, i.e. the
512
Dielectric Materials and Devices
effective surface areas of c+ and c- domain regions are equal, as might be expected from the energy minimization considerations. The potential difference between a and c+ domains was found to be 85 mV, i.e. approximately equal to expected value (V,+V2)/2 = 77 mV. It should be noted here that domain polarities Vl and V2 and average image potential V,, are combinations of four independent parameters A', A2, Bo and B1 and thus are independent. The image formation mechanism in scanning surface potential microscopy on similar surfaces is presented elsewhere4'. The average surface potential and potential difference between the domains determined by SSPM are in excellent agreement with EFM data. Screening on ferroelectric surface Both EFM and surface potential measurements yield the potential difference between c+ and c- domains as AVc-c = 150mV, and the potential difference between a and c domains a AVa-c = AVcJ2. These values are much smaller than expected for the unscreened case. Both EFM and SSPM contrast is found to be uniform within the domains with rapid variation at the domain boundary. The origin of observed signal is attributed either to pure electrostatic field contrast for an unscreened surface or surface potential contrast for completely screened surface. Observations of the ferroelectric phase transition and domain wall motion, as well as the distance dependence of the universal coefficient ratios, suggest that the latter is the case. Thus, the surface condition of ferroelectric BaTiO3 ( 100) under ambient conditions corresponds to complete screening of polarization bound charges. The phase transition, domain wall motion, and piezoresponse imaging suggest that the potential of surface is inverse to that expected from polarization orientation, i.e. it is negative for c+ domains and positive for c- domains. While complete screening and overscreening are expected when domain switching is performed with a charged tip, in our case the pristine equilibrium domain structure is studied. The sign of potential features indicates that screening charges are located closer to the tip than potential bound charges. Calculation of the surface potential suggests that a potential difference of 0.175 V is equivalent to a double layer of 0.25 nm and €1 = 80 (H20) of ferroelectric substrate (external screening) or a layer of 9.5 nm and ~2 = 3000. The former estimate is reasonable for a molecular adsorbate layer or occupation/depletion of surface states, while the latter is unreasonably small for a depletion layer width for semiconductor with low charge carrier concentration (-lpm). Thus, surface adsorption or intrinsic surface states are the dominant mechanisms for screening on ferroelectric surfaces in ambient conditions, though minor contribution from intrinsic screening can not be excluded.
Dielectric Materials and Devices
513
CONCLUSIONS The combination of AFM, EFM and SSPM provides a powerful tool to determine surface and sub-surface domain structures on well-defined BaTi03 (100) surfaces. EFM and SSPM studies of domain wall motion and thermal phase transition indicate that polarization bound charge is completely screened on this surface and surface potential is reverse to that expected from domain polarity. These conclusions are corroborated by piezoresponse imaging technique. Quantification of EFM data allows extraction of absolute domain potentials with respect to the tip. Extracted potential difference between domains of opposite polarities suggest that polarization bound charge is completely screened by adsorbates or charge carriers on intrinsic surface states. We also found that surface potential from SSPM data does not saturate for small tip-surface separations and correspondingly this technique can not be used to unambiguously determine surface potential amplitudes. The latter are also found to be independent on feedback parameters unlike the absolute values of surface potential. ACKNOWLEDGEMENTS We acknowledge the support from MRSEC grant NSF DMR96-32596. Authors are grateful for D.L. Gorbachev for the development of image analysis software and A. Farrow for SEM measurements. REFERENCES 1 F. Jona and G. Shirane, "Ferroelectric Crystals," Dover Publications, New York, 1993. 2G.A. Smolenskii, V.A. Bokov, V.A. Isupov, N.N Krainik, R.E. Pasynkov and A.I. Sokolov, "Ferroelectrics and Related Materials," Cordon and Breach, New York, 1984. 3B. Jaffe, W.R. Cook, Jr. and H. Jaffe, "Piezoelectric Ceramics," Academic Press, London, 1971. 4 V.M. Fridkin, "Ferroelectric Semiconductors," Consultants Bureau, New York, 1980. %. Zhu and W. Cao, "Imaging of 180" Ferroelectric Domains in LiTa03 by Means of Scanning Electron Microscopy," Phys. Stat. Sol. A, 173 (2), 492-502 (1999). 61.M. Reaney, "TEM Observations of Domains in Ferroelectric and Nonferroelectric Perovskites," Ferroelectrics, 172 (1-4), 115-25 (1995). 7 M.L. Mulvihill, K. Uchino, Z. Li and W. Cao, "In-situ Observation of the Domain Configurations during the Phase Transitions in Barium Titanate," Philos. Mag. B, 74 (l), 25-36 (1996).
514
Dielectric Materials and Devices
8V. Gopalan, Q.X. Jia and T.E. Mitchell, "In situ Video Observation of 1800 Domain Kinetics in Congruent LiNb03 Crystals," Appl. Phys. Lett., 75 (16), 2482-84 (1999). 'L. A. Bursill and P.J. Lin, "Electron Microscopic Studies of Ferroelectric Crystals," Ferroelectrics, 70 (3-4), 191-203 (1986). "E. Snoeck, L. Normand, A. Thorel and C. Roucau, "Electron Microscopy Study of Ferroelastic and Ferroelectric Domain Wall Motions Induced by the in situ Application of an Electric Field in BaTi03," Phase Transitions, 46 (2), 77-88 (1994). "Z. Xu, D. Viehland, P. Yang and D.A. Payne, "Hot-stage Transmission Electron Microscopy Studies of Phase Transformations in Tin-modified Lead Zirconate Titanate," J. Appl. Phys., 74 ( 5 ) 3406-13 (1993). l20.O. Popoola and W.M. Kriven, "In-situ Transmission Electron Microscopy Study of Phase Transformations in KNb03 Perovskite," Philos. Mag. Lett., 75 (l), 1-5 (1997). 13 E. Snoeck, C. Roucau, P. Baules, M.J. Casanove, M. Fagot, B. Astie and J. Degauque, "Use of In situ TEM Experiments for Phase Transition Studies,'' Microsc., Microanal., Microstruct., 4 (2-3), 249-64 ( 1993). I4A.A. Sogr, "Domain Structure of Ferroelectrics Observed in the Scanning Electron Microscope," Ferroelectrics, 97,47-57 ( 1989). I5N. Yamamoto, K. Yagi and G.Honjo, "Electron Microscopic Studies of Ferroelectric and Ferroelastic Gadolinium Molybdate (Gd2(M004)3). IV. Polarization Reversal and Field Induced Phase Transformation," Phys. Status Solidi A, 62 (2), 657-64 (1980). 16Y.G.Wang, J. Dec and W. Kleemann, "Study on Surface and Domain Structures of PbTi03 Crystals by Atomic Force Microscopy," J. Appl. Phys., 84 (12), 6795-99 (1998). 17A.L.Gruverman, J. Hatano and H. Tokumoto, "Scanning Force Microscopy Studies of Domain Structure in BaTi03 Single Crystals," Jpn. J. Appl. Phys., 36 (4A), 2207-11 (1997). "M. Takashige, S.-I. Hamazaki, N. Fukurai, F. Shimizu and S. Kojima, "Atomic Force Microscope Observation of Ferroelectrics: Barium Titanate and Rochelle Salt," Jpn. J. Appl. Phys., 35 (9B), 5181-84 (1996). "M. Takashige, S.-I. Hamazaki, F. Shimizu and S. Kojima, "Observation of 900 Domains in BaTi03 by Atomic Force Microscopy," Ferroelectrics, 196 (14), 21 1-14 (1997). 2oS.Balakumar, J.B. Xu, J.X. Ma, S. Ganesamoorthy and I.H. Wilson, "Surface Morphology of Ferroelectric Domains in BaTi03 Single Crystals: an Atomic Force Microscope Study," Jpn. J. Appl. Phys., 36 (9A), 5566-69 (1997).
Dielectric Materials and Devices
515
21G.K.H.Pang and K.Z. Baba-Kishi, "Characterization of Butterfly Single Crystals of BaTi03 by Atomic Force, Optical and Scanning Electron Microscopy Techniques," J. Phys. D, 31 (20), 2846-53 (1998). 22 L.M. Eng, M. Friedrich, J. Fousek and P. Gunter, "Deconvolution of Topographic and Ferroelectric Contrast by Noncontact and Friction Force Microscopy," J. Vac. Sci. Technol. B 14 (2), 1191-96 (1996). 23 H. Bluhm, U.D. Schwartz and R. Wiesendanger, "Origin of the Ferroelectric Domain Contrast Observed in Lateral Force Microscopy," Phys. Rev. B, 57 (1) 161-69 (1998). 24A.Correia, J. Massanell, N. Garcia, A.P. Levanyuk, A. Zlatkin and J. Przeslawski, "Friction Force Microscopy Study of a Cleaved Ferroelectric Surface: Time and Temperature Dependence of the Contrast, Evidence of Domain Structure Branching," Appl. Phys. Lett., 68 (20), 2796-98 (1996). 25R.Luthi, H. Haefke, K.-P. Meyer, E. Meyer, L. Howald and H.-J. Guntherodt, "Surface and Domain Structures of Ferroelectric Crystals Studied with Scanning Force Microscopy," J. Appl. Phys., 74 (12), 7461-71 (1993). 26R.Luthi, H. Haefke, W. Gutmannsbauer, E. Meyer, L. Howald and H.-J. Guntherodt, "Statics and Dynamics of Ferroelectric Domains Studied with Scanning Force Microscopy," J. Vac. Sci. Technol. B, 12 (4), 2451-55 (1996). 27B.D.Terris, J.E. Stern, D. Rugar and H.J. Mamin, "Contact Electrification Using Force Microscopy," Phys. Rev. Lett., 63 (24), 2669-72 (1989). 28J.Ohgami, Y. Sugawara, S. Morita, E. Nakamura and T. Ozaki, "Determination of Sign of Surface Charges of Ferroelectric TGS Using Electrostatic Force Microscope Combined with the Voltage Modulation Technique," Jpn. J. Appl. Phys., 35 (5A), 2734-39 (1996). 29L.M.Eng, J. Fousek and P. Gunter, "Ferroelectric Domains and Domain Boundaries Observed by Scanning Force Microscopy," Ferroelectrics, 191 (1-4), 21 1-18 (1997). 30A.Gruverman, 0. Auciello and H. Tokumoto, "Scanning Force Microscopy for the Study of Domain Structure in Ferroelectric Thin Films," J. Vac. Sci. Technol. B, 14 (2), 602-605 (1996). 310. Kolosov, A. Gruverman, J. Hatano, K. Takahashi and H. Tokumoto, "Nanoscale Visualization and Control of Ferroelectric Domains by Atomic Force Microscopy," Phys. Rev. Lett., 74 (21), 4309-12 (1995). 32A.Gruverman, 0. Auciello and H. Tokumoto, "Imaging and Control of Domain Sructures in Ferroelectric Thin Films via Scanning Force Microscopy,"Annu. Rev. Mat. Sci.,28, 101-123 (1998). 33G.Zavala, J.H. Fendler and S. Trolier-McKinstry, "Characterization of Ferroelectric Lead Zirconate Titanate Films by Scanning Force Microscopy," J. AppZ. Phys., 81 (1 l), 7480-91 (1997).
516
Dielectric Materials and Devices
34L.M.Eng, H.-J. Guntherodt, G.A. Schneider, U. Kopke and J. Munoz Saldana, "Nanoscale Reconstruction of Surface Crystallography from ThreeDimensional Polarization Distribution in Ferroelectric Barium-Titanate Ceramics," Appl. Phys. Lett., 74 (2), 233-35 (1999). 35WzCNCSC-12 from Silicon-MDT 36Y. Cho, S. Kazuta and K. Matsuura, "Scanning Nonlinear Dielectric Microscopy with Nanometer Resolution," Appl. Phys. Lett., 75 (18), 2833-35 (1999). 37S.V. Kalinin and D.A. Bonnell, "Characterization of Ferroelectric BaTi03 ( 100) Surfaces by Variable Temperature Scanning Surface Potential Microscopy and Piezoresponse Imaging," Mat. Res. Soc. Symp. Proc., in press. 38S.V.Kalinin and D.A. Bonnell, "Effect of Phase Transition on the Surface Potential of the BaTi03 (100) Surface by Variable Temperature Scanning Surface Potential Microscopy," J. Appl. Phys. 87 (8), 3950-58 (2000). 39S.V. Kalinin and D.A. Bonnell, "Dynamic Behavior of Domain-Related Topography and Surface Potential on the BaTi03 (100) Surface by Variable Temperature Scanning Surface Potential Microscopy," 2. Metallkd., 90 (12), 983989 (1999). 40Y.Liang, D.A. Bonnell, W.D. Goodhue, D.D. Rathman and C.O. Bozler, "Observation of Electric-Field Gradients near Field-emission Cathode Arrays," Appl. Phys. Lett., 66 (9), 1147-9 (1995). 4'S.V. Kalinin and D.A. Bonnell, "Scanning Probe Microscopy of Ferroelectric Materials. I. Electrostatic Force Microscopy and Scanning Surface Potential Microscopy," Phys. Rev. B, submitted
Dielectric Materials and Devices
517
This page intentionally left blank
ANTIFERROELECTRICITY-THE INVISIBLE HAND BEHIND GOOD FERROELECTRICS I-Wei Chen Department of Materials Science and Engineering University of Pennsylvania, Philadelphia, Pa 19104-6272
ABSTRACT Although antiferroelectric perovskites are not usually considered to have high dielectric and electromechanical coefficients, we suggest that good ferroelectrics always contain incipient antiferroelectric interactions. Such incipient antiferroelectric interactions compete with ferroelectric interactions, causing local structure randomness and eventually frustration. The symmetry change caused by the competition is typically temperature independent, giving rise to morphotropic phase boundary. Very large dielectric and electromechanical responses are associated with compositions near the morphotropic phase boundary. The evidence for the above mechanism is seen in the phase relations and local atomic structures of many perovskites, and a statistical mechanical theory akin to spin glass theory provides a first basis for understanding this mechanism. INTRODUCTION It is well known that among perovskite ferroelectrics (FE), those that have a composition straddling two symmetry phases have outstanding properties. Compositions near a morphotropic phase boundary (MPB) are especially in that the phase boundary is essentially temperature independent, therefore the condition for symmetry straddling is preserved over a broad range of temperature. The above observation is usually explained in terms of the increased multiplicity of polarization directions, since at MPB polarization directions of both symmetries may become simultaneously active. The validity of this explanation, however, is not self-evident. Indeed, it is known that multiplicity is not an important criterion that determines good magnetism, inasmuch as the behaviors of an Heisenberg magnet, which allows spins to assume any direction, and of an Ising magnet, which allows spins to assume only certain crystallographic directions, are rather similar in the mean field theory.[61 Prototypical ferroelectric perovskites that are known to experience an To the extent authorized under the laws of the United States of America, all copyright interests in this publication are the property of The American Ceramic Society. Any duplication, reproduction, or republication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paid to the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
519
enhancement in ferroelectricity, such as large dielectric and electromechanical coefficients, at the MPB include Pb(Zr,Ti)03 (PZT ) solid solution,[’”1 and various solid solutions between relaxors and PbTiO3.[4-51 We have recently proposed a new concept that incipient ferroelectric (FE) /antiferroelectric (AFE competition is a key element in many FE perovskites that have large responses.17jBy quantifying the FE and AFE interactions, respectively, we have identified the MPB to be the condition when the FE/AFE competition is balanced. This condition is usually temperature independent; moreover, the balance can be greatly disturbed by an external field, giving rise to large responses. This concept is somewhat counterintuitive in that most AFE perovskites have relatively low permittivity and are not known as desirable ferroelectrics except for specialized application^.[^-'^] Indeed, compared to FE perovskites, AFE perovskites have been understudied despite the early work of Cross, Shirane, Sawaguchi and Megaw on NaNb03 and PbZrO3 revealing fascinating crystallography and phase transitions.“ 54 71 Nevertheless, as we shall see below, the evidence for the importance of AFE in good ferroelectrics is manifest in the structures and phase relations of a large class of ferroelectrics. In the following, we will summarize such evidence paying special attention to Pbcontaining perovskites. Hopehlly, this concept of FE/AFE tuning will inspire hrther development of new high performance ferroelectrics.
PHASE RELATIONS Pb-containing relaxors are a class of complex perovskites that have attracted current interest because of their outstanding ferroelectric properties.[’8-221 Phase relations observed in many AFE-relaxor-PbTi03 systems provide a first glimpse of the role of AFE in FE responses.[7y231 (See Fig. 1) It is well known that the binary alloys of relaxors and PbTiO3 contain an MPB, and that near the MPB composition, a peak in dielectric and electromechanical coefficients is seen.[4251 It is less appreciated, however, that this is just one part of the overall prototypical phase relations observed in the AFE-FE binary.r23y241In the case of Pb(B71/2B”1/2)03-PbTiO3binary, where Pb(B’ 1/2B7’1/2)03 is an ordered AFE containing mixed cations on the B-site of perovskite, the phase diagram shows a depression of T, and a relaxor regime between (AFE) Pb(B’1/2B’71/2)03and (FE) PbTiO3. There is also an MPB between relaxor and PbTi03 where peak dielectric and electromechanical responses are seen. More generally, even when the relaxor has a composition that is of the (1:2) type, i.e., Pb(B71/3B772/3)03 with a B’:B” ratio of 1:2, and is not capable of complete ordering as in the (1:1) type to become an M E itself, its solid solution with a Pb(B71/2B”)1/2)03 AFE still produces the same feature of a depressed T, in the phase diagram. In addition, the symmetry of all the phase diagrams shown in Fig. 1 follows a systematic trend, from that of orthorhombic or monoclinic distortion for M E , to rhombohedral for relaxor, to
520
Dielectric Materials and Devices
tetragonal for FE PbTiO3. It is thus clear that the relaxor can be regarded as an intermediate phase that experiences competing FE/AFE interactions. The transition from relaxor ferroelectricity (which is anhysteritic) to normal ferroelectricity (which is hysteritic) at the (relaxor/PbTiO3) MPB can likewise be regarded as a transition that coincides with a critical stage in the FE/AFE competition.
PT
AFE
AFE
AFE
b 4
R Relaxor 4 b -
R
PT FE
Fig. 1. Prototypical phase dlagrams of Pb-containing complex perovskites showing
competition between AFE and an FE phase, PbTiO, (PT). Left, between an ordered AFE and PT; middle, between an ordered AFE and relaxor (R); right, between a relaxor and PT.
Lower case symbols for symmetry, cubic (c), tetragonal (t), rhombohedral (r) and orthorhombic/monoclinic (o/m). The rhombohedral phase region always corresponds to a relaxor region.
There are other examples of FE/AFE competition that causes an AFE perovskite, which typically has a relatively low permittivity, to become an FE with a very large permittivity once it is modified to the point of just losin AFE and becoming FE. NaNb03 is a simple AFE perovskite (T, = 64OK).[l5-' The phase diagram of NaNbOs-PbTiO3 has the characteristic depression of T, in the intermediate composition, accompanied by a large increase in permittivity,[241 following the prototypical behavior seen between ordered AFE and PbTiO3. A second example is PbZrO3, a classical AFE. When PbZr03 is alloyed with PbTi03, the PZT family obtained has very large permittivity. Moreover, when Pb is hrther substituted by La, the AFE phase field expands to impinge the PbTi03rich (tetragonal) FE field, and at this boundary composition relaxor characteristics
8
Dielectric Materials and Devices
521
(such as anhysteretic slim loop and smeared transition) emerge.[25-271 (See Fig. 2) It is also at these compositions of impingement where a large electrostrictive response is found. PbZr03 I00
80
60
40
20
PbTi03
0
Fig. 2. Phase Qagram of (Pb,La)(Zr,Ti)03.The Ti-rich region is tetragonal, the Zr-rich region
is orthorhombic, the intermediate composition of Pb(Zr,Ti)03 is rhombohedral, the shaded region is pseudocubic, and the high La composition region contains mixed phases. Their P-E hysteresis lops also shown.
External fields that alter the balance of FE/AFE competition can change the phase relations. Obviously, an electrical field enhances FE and suppresses AFE. As a result, the tetragonal phase field expands at the expense of rhombohedral relaxor. Such field-induced relaxor-to-ferroelectric transitions have been reported many times in the literat~re.[~*-~'] Another more interesting and non-trivial transition is induced by pressure. As it turns out, all paraelectric-to-FE transitions involve an increase in volume, and all paraelectric-to-AFE transitions (with one exception, to my knowledge) involve a decrease in ~ o l u r n e . ' ~Therefore, '~~] a pressure should expand the AFE field at the expense of relaxor and ferroelectric.
522
Dielectric Materials and Devices
Indeed, when an FE is pressurized, an FE-to-relaxor transition is seen as in Lamodified PZT.[311Likewise, when a relaxor is pressurized, a relaxor-to-AFE transition is seen as in Pb(Tn1/2Nb1/2)03.[321 These mechanically induced transitions provide hrther evidence of the presence of FE/AFE competition in relaxor ferroelectrics that are of current interest.
SPIN GLASS MODEL To quantifl the AFE/FE competition, we have used spin glass model and defined two parameters, J, for the average FE interaction favoring dipole alignment, and A for the dispersion of the interaction allowing misalignment [7933y341 The actual dipole-dipole interaction has a strength that varies from Jo-A to Jo+A, depending on the sites of the dipoles. This site-randomness distinguishes a
T kT
kT = A /’Spin Glass I
0
1
JdA
/
0
/
/
Jo
Jo’A
Jo
Fig. 3. (a) PreQcted phase diagram according to spin glass theory, assuming Ising spins interacting via a gaussian distribution of coupling coefficient that has a variance A and mean
Jo. (b) Phase diagram replotted assuming A is linearly correlated with Jo.
spin glass from a simple FE or AFE material, since simple FE corresponds to A = 0 and J, > 0, and simple AFE corresponds to A = 0 and J, < 0. (In contrast, simple spin glass corresponds to A > 0 and J, = 0. But as we shall see, the more interesting case is when J, > 0.) The above model has a direct analogy in magnetism. This is justified by the observation that in magnetism, the antiferromagnetic (AFM) and ferromagnetic (FM) competition, such as the one seen in the (AFM) GdS-(FM) EuS binary, likewise produces, at the intermediate composition, a depression in Curie temperature and a spin glass regime over which the ferromagnetic transition is smeared.[352361
Dielectric Materials and Devices
523
The predicted phase diagram based on the above model is shown in Fig. 3 in two alternative The spin glass/FE transition occurs at J, = A. When J, > A, there is a paraelectric-to-FE transition of the normal type, with a sharp permittivity peak and a P-E hysteresis below the Curie temperature. This transition is smeared when J, < A, where the paraelectric state is reached when thermal agitation exceeds “bond” dispersion. Meanwhile, there is an incipient transition at a lower temperature when kT = J,. We proposed[71that the relaxor state corresponds to the spin glass state, and the high temperature transition may be identified as Tmax, namely the temperature of the maximum permittivity. We also propose that the lower transition temperature may be identified with Td, below which P-E hysteresis is seen and the field-cooled polarization can be maintained. Using this identification, we can quantifjr the magnitude of J, by Td, and the magnitude of A by Tma,. With these parameters, we have also shown that all the experimental data of the 1:2 type of relaxors, specifically Pb(B71/3Nb2/3)03 with B’ = Mg, Ni, and Zn, and their modifications by substituting Ba, Sr and La on the A-site, and Ti on the B-site, fall onto a reduced phase diagram (Fig. 4)[71 that is consistent with Fig. 3b. This implies that these three relaxors of different B’ cations have the same MPB temperature. Indeed, when they are alloyed with PbTi03, the transition temperatures move upward to the same point, at about 440 K, corresponding to the tri-critical point in Fig. 3b. It is also clear that the MPB is determined by the condition J, = A, and is temperature independent. The reduced phase diagram for Pb(B71/2M2,2)O3type of relaxors, including their modifications by substituting A-site cations and B-site cations, have not been constructed since the data of Td are currently lacking. However, reviewing the binary phase diagrams of these relaxors with PbTi03, we notice that their T, at the MPB boundaries all lie around 520 K. Therefore, we believe that the tri-critical point is at 520 K and the reduced hase diagram can be readily delineated once some data of Td become available. [7P Lastly, if the model is hrther extended to span the range from J, > 0 to J, < 0, then it should be able to predict features of the phase diagram of the AFE-FE binaries, including the depression of T, in the intermediate composition, and a relaxor region between the FE and the AFE end.t7923y241 This is expected since the phase diagram of Fig. 3a should have some mirror symmetry going from J, > 0 to J, < 0 in the context of the present model.
ALLOYING EFFECTS ON FE AND AFE Relaxor perovskites are complex oxides containing Pb on the A-site, with typically two types of cations on the B-sites.[18-201 One of the two types of B-site cations may be considered ferroelectrically active, such as Ti, Nb, and Ta, while the other may be considered ferroelectrically inactive, such as Mg, Ni, Zn and Sc.[18,371 As we pointed out previously, ferroelectrically active B’ cations in
524
Dielectric Materials and Devices
relaxors are smaller than the ferroelectrically inactive B” cations, and this is a basic structural requirement for relaxors.r231 Closely related ordered (1 :1) type perovskites may additionally exhibit AFE behavior when the size mismatch between B-site cations is large enough (e.g., Pb(Yb&b1,2)O&[381while other (1: 1) type perovskites with a smaller size mismatch are FE (e.g., Pb(Scl/2Tal/z)Os) when they are ordered and then become relaxors when they are d i s ~ r d e r e d . [ ~ ~ - ~ ~ ] Thus, the size misfit, along with cation order, is a critical tuning parameter that can turn AFE to relaxors and ultimately to FE.
600
400
200
0 0
fig. 4. T,
200
400
600
plotted against Td for some relaxors based on the Pb(B’1/3%/3)03 family. These
relaxors are based on three compounds, Pb(Nil/3mn)03, Pb(Mgl/3Nb2/3)03,and Pb(Zn,1~m/3)03 along with their solid solutions by substituting A and B site cations. This diagram resembles Fig. 3(b) indicating a tricritical point (T*)at about 440 K.
Dielectric Materials and Devices
525
To understand the size effect in terms of the parameters in the spin glass model, we first compare three (1:2) relaxors, Pb(B’1,3Nb2,3)03 with B’ = Ni, Mg, and Zn. These three cations have increasing ionic radii. In Fig 4, we can see that they also have increasing T,, and Td, indicating increasing A and J,. This correlation can be understood as follows. An increasing B’ cation size increases the volume of the unit cell, thereby increasing the FE interaction. This is consistent with the trend that a large unit cell promotes FE in the so-called rattling model and also consistent with the pressure effect.[411On the other hand, the pressure effect implies that AFE is suppressed when the unit cell is enlarged. Thus, the observed trend that an increasing B’ cation size increases A, hence promoting AFE, can not be due to the unit cell expansion. Instead, it must be caused by the increased size misfit between B’ and B” (in this case Nb). This is consistent with the trend that a large size misfit in (1: 1) ordered perovskites causes an FE-to-AFE transition. We have also shown elsewhere that the Curie temperature of (1:l) ordered AFE perovskites with a fixed B” indeed increases with the unit cell volume.[71These results led us to propose that AFE may be regarded as a buckling transition on the B-sites, in contrast to the rattling transition on the B-site for FE .[71 This buckling transition may be triggered by an internal pressure due to the size misfit on the Bsite, or by an external pressure. (Alternatively, if there is only one kind of B-site cation involved, a smaller A-site cation can be used to increase the internal pressure on the B-site, hence promoting AFE, whereas a larger A-site cation should do the opposite. This explains the alloying effect of La, Ba, Ca, and Sr on PbZr03.[241)
LOCAL MODES OF AFE POLARIZATION The local atomic structures of ferroelectrically active cations, especially those that are substantially undersized, e.g., Ti, Nb and Ta, are usually displaced along Such displacement creates the 111 direction in these complex pero~skites.[~~] dipolar interactions that are mostly FE in nature (favoring alignment), even though some AFE character (favoring anti-alignment) always exists for electrical dipoles that stand side by side. More important AFE competition, however, arises when a hi hly polarizable Pb enters the A-site and causes two modes of distortions.[ 723,421 First, when the B site cation is too large, the tendency is always to have a coordinated rotation of the B06 octahedra. The rotation of neighboring octahedra is in opposite directions, which constitutes an AFE mode of polarization. This is the case of AFE PbZr03 and PbHfD3 and also in FE Pb(Zr0.9Ti0.1)03 that has R3c symmetry at room temperature. (See Fig. 5a) Second, when Pb enters a mixed perovskite that has two types of B-site cations, it develops a strong preference to one set of B-site cations over the other, triggered when Pb recognizes a size difference between the two types of B-site cations.
F
526
Dielectric Materials and Devices
fig. 5. Ionic Qsplacements in rhombohedral symmetries of perovskites.
(a) all B cations are equivalent, (b) two distinguishableB positions.
This results in a large 111 displacement of Pb toward one of the B-sites with associated B-site displacements. (See Fig. 5b) Since there are two neighboring Pb along a body-diagonal centered at each B-site, their displacements must be in
Dielectric Materials and Devices
527
opposite directions and again constitute an AFE mode of p o l a r i ~ a t i o n . [(Unlike ~,~~~ rotation, these are in-line dipoles of opposite directions.). This preference of Pb displacement can be satisfied in a coordinated way over a long range if the B-site cations are hlly ordered. A complete AFE thus results as in ordered Pb(ml/2ml/2)03 and Pb(Mg1/2W1/2)03. In alloys with mixed B-site cations that are at least partially disordered, full AFE-type of displacements are not possible and the competition of FE/AFE interactions causes various degrees of frustration.[77231 This frustration is at the core of the very strong dielectric responses seen in PZT, (Pb,La)(Zr,Ti)03 (PLZT), and well-known relaxors such as Pb(Mg1/3Nb2/3)03.In the latter case, the random arrangement of Mg and Nb on at least one €3-site sublattice causes frustration of the Pb and Nb di~placernents,[~~~~~~~] both having strong FE and AFE tendencies. A completely correlated FE or AFE state of these displacements is not possible. Such frustrated FE/AFE competition is seen as a means to tune dielectric and electromechanical responses. Moreover, when both the FE and AFE interactions are strong, as in many Pb-containing complex perovskites, the interactions and competition are temperature independent and can be balanced at a particular composition that corresponds to the condition of “mean bond strength” equal to “bond dispersion.” The underlying random nature of the balance, however, makes the state a precarious one that can be easily affected by an external field creating large electromechanicalresponses.
LARGE-FIELD RESPONSES The above statistical mechanical model is capable of explaining the essential features of relaxors and large response FE regarding phase transitions and zero/weak field responses. Large field responses such as domain switching, however, are more difficult to model using such rigorous methods but can be more conveniently understood using activated state theories based on domain wall dynamics.[441The method here is similar to the dislocation the0 of crystal plasticity and interface theory of martensitic phase transformations,3 4 7 1 since all these phenomena involve large fluctuations of incommensurates over obstacles that have a broad spectrum of size and strength. In practice, the obstacle spectrum evolves, e.g., during switching, and such evolution impacts the observed dynamics. In our previous research we have studied the temperature, frequency and history dependence of the switching resistance in both small and large signal regimes.[447451 These dynamic data have been compared with model predictions to hrther understand the relaxor and FE behavior in active switching applications. ACKNOWLEDGEMENT This work was supported by the U.S. National Science Foundation, Grant No. DMR 99-88853. The use of facilities at the University of Pennsylvania
528
Dielectric Materials and Devices
supported under the NSF MRSEC program, Grant No. DMR 96-32598 is also acknowledged.
REFERENCES 1. B. Jaffe, W.R. Cook and H. Jaffe, Piezoelectric Ceramics, Academic Press, London, 1971. 2. K. Car1 and K.H. Hardtl, “On the Origin of the Maximum in the Electromechanical Activity in Pb(ZrxTil-,)03 Ceramics near the Morphotropic Phase Boundary,” Phys. Stat. Sol. , (a), 8, 87 (197 1). 3. M. Fukuhara, A.S. Bhalla and R.E. Newnham, “Morphotropic Phase Boundary in the Pb(ZrxTi1-,)03 System,” Phys. Stat. Sol., (l), 122, 677 (1990). 4. S. Nomura, T. Takahashi and Y. Yokomizo, ‘Terroeelctric Properties in the System Pb(Zn1/3Nb2,3)03-PbTiO3,” J. Phys. Soc. Jpn., 27, 262 (1969). 5 . L. Hahn, K. Uchino and S. Nomura, “On the Phenomenon of Morphotropic Tetragonal-Rhombohedral Boundary in the Ferroelectric Ceramics,” Jpn. J. Appl. Phys., 17 [4], 637 (1978). 6 . P.M. Chaikin and T.C. Lubensky, Principles of Condensed Matter Physics, Cambridge Univ. Press, Cambridge, UK, 1995. 7. I-W. Chen, “Structural Origin of Relaxor Ferroelectrics-Revisited,” J. Phys. Chem. Soli&,61,197-208 (2000). 8. D. Berlincourt, H.H. Krueger, and B. Jaffe, “Stability of Phase in Modified Lead Zirconate with Variation in Pressure, Electrical Field, Temperature and Composition,” Phys. Chem. Solids, 25, 65974 (1964). 9. D. Berlincourt, H. Jaffe, H.H. A. Krueger and B. Jaffe, ‘Release of Electric Energy in PbNb(Zr,Ti, Sn)03 by Temperature and by Pressure-Enforced Phase Transitions,” Appl. Phys. Lett., 3, 90-98 (1963). 10. L.E. Cross, “Antiferroelectric-Ferroelectric Switching in Simple “Kittel” Antiferroelectrics,” J. Phys. Soc. Jpn., 23, 77-82 (1967). 11. D. Berlincourt, “Transducers Using Forced Transitions between Ferroelectric and Antiferroelectric States,” IEEE Trans. Sonics Ultrason. , 13, 116-24 (1966). 12. W. Pan, Q. Zhang, A. Bhalla and L.E. Cross, ‘Field-Forced Antiferroelectricto-Ferroelectric Switching in Modified Lead Zirconate Titanate Stannate Ceramics, ” J . Am. Ceram. Soc., 72 [4] 571-78 (1989). 13. W.Y. Pan, C.Q. Dam, Q.M. Zhang and L.E. Cross, “Large Displacement Transducers Based on Electric Field Forced Phase Transitions in the Tetragonal (Pbo.97Lao.02)(Ti7Zr, Sn)03 Family of Ceramics,” J. Appl. Phys., 66 [12] 6014-23 (1989).
Dielectric Materials and Devices
529
14. B. Xu, P. Moses, N.G. Pai and L.E. Cross, "Charge Release of Lanthanum-
doped Lead Zirconate Titanate Stannate Antiferroelectric Thin Films," Appl. Phys. Lett., 72 [ 5 ] , 593-95 (1998). 15. L.E. Cross and B.J. Nicholson, Phil. Mag., Ser. 7, 46,453 (1955). 16. H.D. Megaw and W. Wells, Acta Cryst. 11, 858 (1958). 17. F. Jona and G. Shirane, Ferroelectric Crystals, Pergamon Press, New York (1962). 18. G.A. Smolensky, "Physical Phenomena in Ferroelectrics with Diffised Phase Transition," J. Phys. Soc. Japan, 28,26-37 (1970). 19. L.E. Cross, "Relaxor Ferroelectrics,"Ferroelectrics, 76, 241 (1987). 20. L.E. Cross, "Relaxor Ferroelectrics: An Overview," Ferroelectrics, 151, 30520 (1994). 21. S.E. Park and T.R. Shrout, J. Appl. Phys., 82, 1806 (1997). 22. Y. Yamashita and K. Marada, Jpn. J. Appl. Phys., 263 (1998). 23. I-W. Chen, P. Li and Y. Wang, "Structural Origin of Relaxor Perovskites," J. Phys. Chem. Solids, 57 [lO] 1525-36 (1996). 24. See compilation of numerical data and hnctional relations in LandoltBornstein, New Series, Group I11 Vol. 13 and Vol. 28, Springer -Verlag, Berlin (198 1, 1990). 25. C.G. F. Steiger and A.J. Burggraaf, "Study of Phase Transitions and Properties of Tetragonal (Pb,La)(Zr,Ti)03 Ceramics-I," J. Phys. Chem. Solids, 41, 17-23 (1980). 26. C.G. F. Steiger and A.J. Burggraaf, "Study of Phase Transitions and Properties of Tetragonal (Pb,La)(Zr,Ti)Os Ceramics-11," J. Phys. Chem. Solid, 41,25-30 (1980). 27. G. Burns and F.H. Dacol, "Crystalline Ferroelectrics with Glassy Polarization Behavior," Phys. Rev., 28B, 2527-28 (1983). 28. G. Schmidt et al., Cryst. Res. Technol. 15, 1415 (1980). 29. E.V. Colla et al., Ferroelectrics, 151, 337 (1994). 30. R.Sommer, N.K. Yushin, and J.J. van der Klink, Phys. Rev., B 48, 13230 (1993). 3 1. G.A. Samara, Tressure-induced Crossover fi-om Long-to Short-Range Order in Compositionally Disordered Soft Mode Ferroelectrics," Phys. Rev. Lett., 77[2], 314-17 (1996). 32. K. Nomura, T. Shingai, N. Yasuda, H. Ohwa and H. Terauchi, "Pressureinduced Structural Phase Transition from Relaxor Phase to Antiferroelectric Phase in Disordered Pb(Inl/zNbl/2)03, J. Phys. Soc. Japan, 68[3] 866-70 (1999). 33. K. Binder and A.P. Young, "Spin Glasses: Experimental Facts, Theoretical Concepts and Open Questions," Rev. Mod. Phys., 58, 801-976 (1986). 34. K. Moorjani and J.M.D. Coey, Magnetic Glasses, Elsevier, Amsterdam, 1984.
530
Dielectric Materials and Devices
35. J. L. Tholence, F. Holtzberg, T. R. McGuire, S. Von Molnar, and R. Tournier, J. Appl. Phys., 50, 7350 (1979). 36. A. Berton, J. Chaussy, J. Odin, R. Rammal, S. Souletie, J. L. Tholence, R. Tourier, F. Holtzberg, and S. Von Molnar, A. Appl. Phys., 52, 1763 (1981). 37. N.W. Thomas, “A New Framework for Understanding Relaxor Ferroelectrics,” J. Phys. Chem. Solids, 52 [121 1419 (1990). 38. F. Gallasso, Structure, Properties and Preparation of Perovskite-type Compounds,Pergamon Press, New York, 1969. 39. N. Setter and L.E. Cross, “The Role of B-site Cation Disorder in Difise Phase Transition Behavior of Perovskite Ferroelectrics,” J Appl. Phys., 5 1 [8]4356-60 (1980). 40. N. Setter and L.E. Cross, “The Contribution of Structural Disorder to Difise Phase Transitions in Ferroelectrics,” J Mater. Sci., 15, 2478-82 (1980). 41. J.C. Slater, Phys. Rev., 78, 748 (1950). 42. J.B. Goodenough and J.M. Longo, “Magnetic and Other Properties of Oxides and Related Compounds,” Landolt-Bornstein,New Series, Vol. 12, Ed. K.H. Hellwege, Springer-Verlag, Berlin (1978). 43. M.A. Akbas and P.K. Davies, “Domain Growth in Pb(Mg1/3Ta2/3)03 Perovskite Relaxor Ferroelectric Oxides”, J Amer. Ceram. Soc., 80, 293336 (1997). 44. I-Wei Chen and Ying Wang, “A Domain Wall Model for Relaxor Ferroelectrics,”Ferroelectrics, 206/1-4 & 207/1-2,245-263 (1998). 45. J.P. Hirth and J. Lothe, Dislocation Theory, 2nd Ed., J. Wiley & Sons, New York, 1982. 46. U.F. Kocks,A.S. Argon and M.F. Ashby, “Thermodynamics and Kinetics of Slip”, Prog. Mater. SciL,Vol. 19, Pergamon Press, New York, 1975. 47. G.B.Olson and M.Cohen, “Dislocation Theory of Martensitic
Transformations,” in Dilsocations in Solids, Ed. F.R.N. Nabarro, NorthHolland, Vol. 7, 297-407, 1986. 48. Y. Wang, Dielectric and Ferroelectric Properties of Perovskite Relaxors, PhD Dissertation, University of Michigan, 1998.
Dielectric Materials and Devices
This page intentionally left blank
MICROWAVE DIELECTRIC PROPERTY MEASUREMENTS RICHARD G. GEYER* AND JERZY KRUPKA** *National Institute of Standards and Technology R F Technology Division, M.S. 813.01 325 Broadway Boulder, CO 80303
** Politechniki Warszawskiej , Instytut Mikroelektroniki Warszawa, Poland
ABSTRACT Materials, whether in the solid, liquid or gaseous states, may be electrically nonlinear, anisotropic, inhomogeneous and dispersive with respect to frequency. Dispersion results from loss mechanisms that differ in different types of materials and vary with temperature. Dielectric loss tangent measurements reflect the different loss mechanisms occurring in a material placed in an electric field. Because of all these factors, both measurement techniques and accuracies for evaluation of dielectric properties are requisite for physical understanding. Various permittivity and dielectric loss tangent measurement techniques, including low-frequency complex impedance, free space, waveguide transmission and reflection, and resonance methods are reviewed. Measurement uncertainties are also discussed.
ELECTROMAGNETIC CHARACTERISTICS OF MATERIALS Physical Concepts All materials are characterized electromagnetically by permittivity E (F/m): magnetic permeability p (H/m), and electrical conductivity 0 (S/m). Maxwell’s equations, together with the constitutive equations relating field quantities in terms of material properties, completely govern electromagnetic wave propagation and behavior in that medium. The constitutive equations for a linear, homogeneous and isotropic medium may be expressed in the frequency domain as
5 f 6
= pif, = =
02, CZ,
(1)
To the extent authorized under the laws of the United States of America, all copyright interests in h s publication are the property of The American Ceramic Society. Any duplication, reproduction, or re ublication of this publication or any part thereof, without the express written consent of The American Ceramic Society or fee paifto the Copyright Clearance Center, is prohibited.
Dielectric Materials and Devices
533
2
where the magnetic induction (Wb/m2) is related to the magnetic field l? (A/m) by the magnetic permeability; the current density f ( A / m 2 ) is related to the electric field I? (V/m) by the conductivity; and the dielectric displacement field 5 (C/m2) is related to the electric field by the permittivity. Any deviation from linearity is usually included by making E , p , or IY field dependent. For anisotropic media, E , p , or 0 is a second rank tensor rather than a scalar function of frequency or simply a constant. For inhomogeneous media, E , p, or IY is a function of spatial coordinates. Material dielectrics that are linear, isotropic, and homogeneous when placed in an electric field at one frequency may not be homogeneous when placed in an electric field at another frequency or be linear and isotropic under different temperature and pressure conditions. Similarly, dielectrics that behave linearly when placed in weak electric fields may not be linear in strong fields. Dielectric Anisotropy
When a dielectric is placed in an electric field and the dielectric displacement field is written
I?) the material becomes
polarized,
where f ' is defined as the electric polarization of the material (dipole moment per unit volume), or ; ?? - = xo€ (3)
F
the proportionality constant x is called the electric susceptibility, and the factor € 0 (free space permittivity equal to 8.854 x 10-l2 F/m) is included in eq ( 3 ) to make x dimensionless. Then eq (2) becomes
or
+
+
5=EOE~E,
(5)
where E: = 1 x is called the complex permittivity of the medium relative t o a vacuum. (The subscript r denotes relative.) The presence of a dielectric always affects the ratio of 8 to I? by a factor of E: . For linear materials the dipole moment induced in a dielectric by an external field I? is directly proportional to I?. As long as the electric properties of the dielectric are independent of direction, it is isotropic; that is, and 2 are collinear. For an anisotropic material, however, the polarization (or charge separation) obtained when an electric field is applied along one coordinate axis will be different than that produced by the same field applied
534
Dielectric Materials and Devices
along a different coordinate axis. Quantitatively, this is expressed by writing
+
where X = [xzcii x y j j t xzkk] and xz,xy, xz are the principal components of the electric susceptibility tensor expressed in dyadic form. For isotropic materials xz = xy = xz,and eq (6) reduces to eq (3). Equation (6) shows that @ and I? are not collinear when xz # xy = xz or when xz = xy # xz or when xz # xy # xz (for two- or three-dimensional anisotropy), so that the electric susceptibility tensor may, in general, be viewed as an operation which takes a vector I? and converts it into a new vector which is not collinear with 2.Accurate determination of the dielectric properties of anisotropic materials usually involves specimen orientation relative t o both an axis of anisotropy and the applied electric field in the measurement system. -tt
++
++
Polar versus Nonpolar Materials Dielectric materials can be described as polar or nonpolar. A nonpolar material (such as inert or rare gases) is simply one that contains no (equivalent) dipoles (or separation of charge) when the material is not in an electric field. A polar material, on the other hand, possesses permanent polarization, even in the absence of an electric field, due to its molecular structure. Polar materials have permanent dipole moments at the microscopic or molecular level. In general, any distribution of charge may be described in terms of its multipole moments [l]. The relevance of this discussion to dielectric material properties is that the existence of permanent dipole moments on the molecular level gives rise to a type of polarization mechanism when an electric field is applied which is frequency-dependent. Without an applied electric field in polar materials, the individual molecular dipole moments can point in random directions, so that macroscopically, their vector sum vanishes. In the presence of the applied electric field I?, though, there is a pronounced tendency of the dipoles to line up in the direction of I?, creating an orientational polarization whose magnitude can be computed and measured [a]. Ferroelectric materials are those in which there is a spontaneous alignment of electric dipole moments at the molecular level, which occurs at the Curie temperature. The permittivity of a ferroelectric material is dependent on the field strength of an applied electric field bias. This dependence allows ferroelectric materials to be used in a variety of nonlinear devices, such as piezoelectric transducers, voltagecontrolled oscillators, varactors, tunable filters, and phase shifters.
Dielectric Materials and Devices
535
Material Electrical Constituent Properties The solution of Maxwell's equations yields all of the quantities that describe the propagation of electromagnetic waves in terms of the propagation constant y = j k , where k 2 = W ~ ( W E- ja) (7) for exp(+jwt) time dependence for angular frequency w and time t. In general, the constituent electrical properties may be written as complex quantities; that is, for exp(+jwt) time dependence,
where po is the free space permeability equal to 47r x 10-7 H/m. Each component of E , a,or p (which for anisotropic materials are tensor matrices) is, in general, a complex quantity. The imaginary part of the propagation constant contains all necessary information about energy loss in a material medium during wave propagation. If magnetic properties are ignored, we may consider only the complex form of E and a in eq (7): WE - j
a = U(€' - j c " )
- j(o'
+ ja") = (a"+ WE ) - j(a' + W E ' / ) . I
+
.
(9)
+
Here (WE' a") may be considered an effective permittivity and (a' W E " ) as an effective conductivity. The term (a'+ja'') physically represents carrier transport due to ohmic and Faraday diffusion mechanisms, whereas (E' - j ~ " represents ) dielectric relaxation mechanisms. From eq ( l O ) , the loss tangent is simply defined as 7r
t a n s = t a n ( + + -) 2
= a' + WE" ~
a"
+WE',
where $I is the phase between I? and f. If there are no dielectric losses, Similarly, if there are no Faraday losses, o" + 0; hence, tans =
E"
+
0.
a'
-, WE'
which describes losses physically due to ohmic conductivity. Actual dielectric measurements are indifferent to the underlying physical processes. To the extent, however, that physical and chemical processes are understood, distinctions can be made and materials designed t o have certain electromagnetic characteristics.
536
Dielectric Materials and Devices
Polarization Mechanisms in Materials A pulse or “signal” of any specified initial form can be constructed by superposition of harmonic wave trains of infinite length and duration. The velocity with which the constant-phase surfaces of these component waves are propagated (phase velocity) depends on the propagation constant, or on the parameters E , p , and a. If the medium is nonconducting and the quantities E and p are independent of the frequency of the applied field, the phase velocity is constant and the signal is propagated without distortion. The presence of a loss mechanism, however, yields a functional relation between the frequency and phase velocity, as well as between frequency and attenuation. Hence in a lossy or absorptive medium the harmonic components suffer relative displacements in phase in the direction of propagation, and the signal arrives at a distant point in a modified form. The signal is dispersed and attenuated, and a medium in which the phase velocity is a function of frequency f (or in which the complex dielectric constant E* is a function of frequency) is said to be electrically dispersive. The quantity ~ ’ ( z, f ; y, z ) is a measure of the polarization of the material. There can be a number of different polarizing mechanisms, each having a characteristic relaxation frequency and dielectric dispersion centered around this relaxation frequency. At the relaxation frequency there is maximal absorption. Figure 1illustrates the dispersion of E (and a) that may be observed in materials in the frequency range 103 to 1015 Hz. At the highest frequencies, the polarizing species in a material are the electrons. Electronic polarization occurs when an applied electric field causes a net displacement of the electron cloud of an atom with respect to its nucleus. At frequencies below about lOI3 Hz, there is also a contribution from atomic polarization. Atomic polarization occurs in structures (molecules, solutions) in which atoms do not share electrons equally and electric fields displace the electron clouds preferentially towards the stronger binding atoms. It also occurs when charged atoms are displaced with respect to each other. Dipolar polarization, that is, the orientation of polar molecules (molecules with asymmetric charge distributions), occurs at frequencies below about 1010 Hz. At frequencies below about 105 Hz, there are various types of charge polarization which may be collectively referred to as Maxwell-Wagner mechanisms [3,4]. One of these, interfacial (space-charge) polarization, occurs when migrating charge carriers are trapped or impeded in their motion by local chemical or electric potentials, causing local accumulations of charge and a macroscopic field distortion. Another low-frequency mechanism that can occur is due to mixtures of materials having differing electrical properties (such as conducting spheres embedded in a dielectric). Several different equations are available to describe the resultant properties for vari-
Dielectric Materials and Devices
537
* c
c.
n
Maxwell-Wagner (Interfacial)
,"Conductivity Frequency
(Hz)
Figure 1: Dielectric dispersion for various types of polarization. ous geometries of the embedded conductor [5-7]. The common causes of these effects are the distributions of charge that occur at conductor-dielectric boundaries and the resultant action under applied electric fields which can yield very large low-frequency dielectric constants. Dispersion Processes in Materials
Polarization occurring in material media as a result of electromagnetic wave propagation is physically damped by either resonance or relaxation. Resonance is the state of a harmonic oscillator that is driven at its preferred frequency. Relaxation, on the other hand, is the state of a critically damped or overdamped oscillator. The characteristics of E', E" for these two differing types of dispersion and absorption processes are shown in Fig. 2. At microwave frequencies, dipolar or orientation polarization phenomena principally occur. In this case, the frequency is sufficiently low so that the rotation of polar molecules has time t o take place. At a frequency of w = l / r , E' decreases because the individual dipoles can no longer keep in step with the applied field. The relaxation time r represents the time required for the dipoles to revert to a random distribution. This is a diffusion process which is represented by Fig. 2(a). Atomic and electronic polarization processes take place in the infrared and optical portion of the spectrum (1 THz and above) and lead to the
538
Dielectric Materials and Devices
t
tI
E'
I
log f
log f
Figure 2: Differing types of dispersion and absorption processes occuring in dielectrics as a function of frequency: (a) relaxation and (b) resonance. E' is real permittivity and E" is the dielectric loss index. resonance-type dispersion and absorption phenomenon represented by Fig. 2( b). A given medium may display any or all of these characteristic polarization phenomena, depending on its composition and molecular or atomic structure. Relaxation processes are those observed in dielectric materials at microwave frequencies and below. Relaxation models are based on the general equation of charge motion, q t (p.)-'i (p€>-'q = 0, (12)
+
where q is the charge and - represents differentiation with respect to time.
Debye Relaxation Materials having single relaxation time constants are called Debye materials. The complex permittivity in a Debye material is given by [8-101 E'
- jet' = E,
+ 1€s+-w E,2 r 2
- E,)WT
- j
1+
d r 2
'
where E , is the relative dielectric constant at zero frequency ( q c= . € , C O ) , and E, is the optical relative permittivity (at angular frequency w >> l / ~ )In. general, apart
Dielectric Materials and Devices
539
from liquid dielectrics, single relaxations are seldom observed. Multiple relaxations or distributions of relaxations are instead found.
Generalized Relaxation Distributions A generalized expression for material media in which multiple relaxations are found may be written as [ll],
where D ( r ) is the time constant distribution function normalized such that roo
10
D(r)dr =
One of the most commonly observed simple relaxation distributions in lossy media is the Cole-Cole distribution. In the Cole-Cole distribution eq (14) reduces t o
where 0
5 m 5 1. The loss tangent for the Cole-Cole distribution [8] is tan6 =
E" E'
=
1
+e + (2+
e(wr)l-m e)(wr)l-m
sin [(1 - m)]; [(I - m)?]
COS
+
'
( ~ ~ ) 2 ( l - m )
(17)
where 8 = ( E ~ E,)/E,. The m = 0 case corresponds to a Debye material (single relaxation). The m = 1 case corresponds to an infinitely broad continuous distribution (one having no relaxation). In the latter case the imaginary part of the complex permittivity disappears, and the real part becomes frequency independent. The Cole-Cole distribution corresponds to a symmetric distribution of relaxation times of width m. Whereas a Debye material yields a plot of E"(E') that is a semicircle whose center lies on the E" = 0 axis, a Cole-Cole E"(E') plot is a semicircle whose center lies below the horizontal E'' = 0 axis, on a line drawn from (E' = E,, E" = 0) that makes an angle of mn/2 with the horizontal axis. This is shown in Fig. 3. In addition to the Cole-Cole expression, there are other empirical relations commonly used t o describe a non-Debye response. These are the Cole-Davidson [12], the combined Cole-Cole, and the Williams- Wat kins [ 131 expressions. A characteristic feature of all these empirical relations, besides being based on eq (12)) is that at frequencies away from the (dominant) relaxation frequency, they reduce to expressions showing a power-law dependence [14] on frequency for both E' and E".
540
Dielectric Materials and Devices
Figure 3:
E"
versus
E'
for (a) Debye and (b) Cole-Cole materials.
Generalized Relation between Permittivity and Dielectric Loss Index A generalized relation between E' and E" for linear dielectric materials possessing an arbitrary number of relaxation times may be derived by regarding the permittivity as a system function characterizing the electrical properties of a material with the applied electric field as input and the displacement field as output. In the time domain, the material permittivity is simply the transient (causal) system response, which can always be decomposed into the sum of even and odd functions. The Fourier transforms of the even and odd functions yield the (real) permittivity and (imaginary) dielectric loss index. The real permittivity and dielectric loss index are then related by the following Hilbert transforms, also known as the Kramers-Kronig relations, €'/(U)=
and €'(U)
= E,
-P 7r
Jrn 1
-
- a wE'(v) - v dv7
1
- -P 7r
€"(U)
-dv,
-mU-W
where P denotes the Cauchy principal value. The application and limitations of eqs (18) and (19) for band-limited measurement data have been described in [ll]. In addition, the use of an inverse power law of the distribution function for predicting
Dielectric Materials and Devices
541
expected changes in dielectric loss tangent from measured changes in permittivity at two selected frequencies is treated [ll]. Effect of Temperature Changes
A classical statistical thermodynamic model using a double potential well was used [113 t o describe the dispersive dielectric behavior for a bistable dielectric as a function of frequency, the temperature-dependent dipolar polarizability Q D = N p 2 / l c ~ Tthe , activation energy U , and the high frequency (optical) permittivity at temperature T (K). N is the total number of bistable dipoles in the material having dipole moment p and k g is Boltzmann’s constant ( 1 . 3 7 ~ 1 0 -J/K. ~ ~ A bistable model is that of an elementary dipole within a dielectric whose molecular groupings can be characterized by well-defined dipole moments. In this type of model, we assume that a charge q may be in one of two states, depicted by states 1 and 2, that are separated by a distance.d. The states are defined as minima of the potential energy function, and an electric field acting on the dielectric causes movement of charge from the minimum of state 1 to the minimum of state 2. The results are E’(W)
T ) = E,(T)
+ 1 +QDu2r2 ~
and where r = e-U/“BT/2A and A is a constant (that may or may not depend on temperature) describing the number of dipoles jumping per unit time from one potential energy state to a higher state within the dielectric. Equations (20) and (21) are limited t o dielectric materials where interaction between individual dipoles can be neglected and for conditions where pE << kBT (nonsuperconducting states). Langevin considered the electrostatic case of interacting molecules. Langevin treated a Maxwell-Boltzmann statistical ensemble average of the molecular angular alignment with an applied electric field E . The ensemble consisted of point dipoles having equal dipole moments in thermal equilibrium at temperature T . He derived the well-known Langevin function shown in Fig. 4,
where 0 is the angle between field and dipole and y = p E / ( l c ~ T )The . ensemble average < cos0 > increases with increasing y; for values of E / T >> 1, the orienting effect of the electric field dominates over the disorienting action of the temperature. Implicit in the derivation of the Langevin function are the assumptions that the
542
Dielectric Materials and Devices
t
0
1
2
3
4
5
Figure 4: Behavior of Langevin function versus p E l ( k ~ T ) . molecules are point dipoles that have isotropic polarizability, that ergodicity holds, and that the system obeys classical Maxwell-Boltzmann statistics. Additional discussion on relaxation models is to be found in the classical texts of Von Hippel [15] and Bottcher [16]. These physical relaxation models provide insight into what dispersive permittivity and dielectric loss tangents can be expected both as a function of temperature and frequency. However, their applicability and validity must be ascertained by accurate measurements.
PERMITTIVITY AND DIELECTRIC LOSS TANGENT MEASUREMENTS Low-Frequency Complex Impedance Measurements
The use of a plane-parallel capacitor having a vacuum capacitance CO = ~ o S / t , where S and t are the respective surface area of the electrode plates and separation of the plates, is commonly used for low-frequency dielectric measurements. If a low frequency voltage V = Voejwt is applied to this capacitor, a charge Q = CoV appears on the electrodes that is in phase with the applied voltage. The nondissipative displacement current in the external circuit is then given by
I = Q = jwCoV
Dielectric Materials and Devices
(23)
543
which is 90" out of phase with the applied voltage. If the volume between the electrodes is filled with a lossless, nonpolar insulating material, the capacitor has a capacitance C = €:CO. In this case the new displacement current is
The capacitance is larger than the vacuum capacitance, but remains 90" out of phase with respect to the applied voltage. For lossy dielectric materials, the current is not 90" out of phase with the voltage since there is a small conduction GV due t o charge motion in phase with the applied voltage. If the charges are free, the conductance G is independent of frequency. However, if the charges are bound, G is frequency dependent, and the dipole relaxation phenomena previously described become relevant. In general, I = ( j w C t G)V, (25) . where G = a S / t is the conductance due to free charges and C = ~ i S / t Whenever dissipation is not exclusively due to free charges, but is also due to bound charges, the conductivity is itself a complex frequency-dependent quantity and a distinction cannot be made between ohmic conductivity and dielectric loss factor or between Faraday diffusion transport and in-phase polarization. Free Space Measurement
Free space measurements of the complex permittivity and complex permeability usually involve placing a plate specimen orthogonal to the axis between the transmitting and receiving antennas. A plane electromagnetic wave is passed through the specimen. The complex permittivity or permeability can then be evaluated from measurements of the propagation constant ys = j k , = j w d q of the plane electromagnetic wave propagating in the specimen or from the measured impedance 2, = 0pu,,/EOE, ,of the specimen. The accuracy of free space measurements depends on the appropriate choice of a theoretical model representing the experimental measurement system and the accuracy of the measurement system. Accuracy of either dielectric or magnetic loss tangent is largely constrained by radiation losses. Radiation losses occur as a result of specimen misalignment and diffractive edge effects. For a normally incident transverse electromagnetic (TEM) wave on the specimen surrounded by air (see Fig. 5) the transmission and reflection coefficients, To and Ro, are given by
d
544
c
Dielectric Materials and Devices
Figure 5 : Reflection and transmission coefficients for an electromagnetic plane wave incident on a dielectric plate specimen.
$--
where 70 = j27r/Xo and ys = j27r ~;,,p;,,/Xo. Equations (26) and (27) can be solved for the complex permittivity and permeability of a magnetic plane-parallel plate. If the specimen is nonmagnetic, the transmission coefficient is solved for E ; . A common reflection technique for complex permittivity evaluation is to place a conducting plate (short) behind the specimen and measure the reflection coefficient. In this case,
Generally, complex permittivity evaluations are more accurate in reflection (oneport scattering parameter) measurements when the specimen is surrounded by air, whereas permeability evaluations are most accurate from reflection measurements when the specimen is backed by a shorting plane.
Waveguide Transmission Line Met.hods Transmission line techniques, usually made in rectangular or coaxial waveguides, are the simplest of the relatively accurate wavs of measuring permeability
Dielectric Materials and Devices
545
and permittivity. Coaxial lines are broadband in the TEM dominant mode and, therefore, are at tractive for spectral characterization of lossy magnetic materials. However, measurement uncertainty problems exist in complex permittivity determination when air gaps occur between the sample and the coaxial line center conductor. Details of two-port, reference-plane invariant scattering parameter expressions that can be used for determining permittivity and permeability are given elsewhere [17]. One set of equations for dielectric and magnetic measurements of a single sample, in terms of two-port scattering parameters that can be taken with an automatic network analyzer, is
and where
R = PYO - POY PYO t POT' T = exp(-yL), Yo =
\i
W
(,)2
2T
- (-)2,
Club
(33)
c,,, and C l a b are the speed of light in vacuum and laboratory, w is angular frequency, Ac is cutoff transmission-line wavelength, e;, p: are the effective relative complex permittivity and permeability relative to vacuum, and Lair,L are air-line and specimen lengths. Equations (29) and (30) can be solved either explicitly or implicitly as a system of nonlinear scattering equations at each frequency or by using a nonlinear regression model over the entire frequency range. disadvantage of microwave measurements of the complex permittivity in waveguides is that specimens must have very small dimensional tolerances. If the specimen does not fill perfectly the entire cross section of the waveguide, corrections must be made for air gaps. For high permittivity samples in either rectangular or coaxial transmission lines, air gaps can lead to dielectric depolarization, which yields severe underestimates of actual specimen permittivity. The usual procedure t o avoid depolarization effects with these measurement techniques is to metallize the specimen at all surfaces in contact with the waveguide. Generally, there is greater uncertainty in waveguide
546
Dielectric Materials and Devices
transmission-line methods at the lowest measurement frequencies. The real permittivity is very sensitive to measured phase and sample length. Greater uncertainty in permittivity at low frequencies results from the very small phase shift over the length of the sample at low frequencies. Resonance Met hods
Resonance methods employing either closed and open cavities or dielectric resonators provide the highest measurement accuracy for evaluating complex permit tivity and dielectric loss tangent of low-loss materials at microwave frequencies [18-211. We define low-loss materials as those where tan 6 << 0.001. Generally, the (real) permittivity is calculated from the measured resonant frequency of one of the dominant modes of the resonant measurement system and the dimensions of both specimen and the resonant structure. As long as specimen losses are low, they do not affect resonant frequencies. Exact relations between permittivity, sample dimensions, and resonant frequency exist only for simple rectangular, cylindrical, or spherical resonant structure geometries and when any permittivity inhomogeneity in the measurement fixture varies in only one of the principal coordinate directions. Resonant fixtures commonly used in practice for cylindrically shaped disk or rod specimens are shown in Fig. 6. An eigenvalue relation, derived from Maxwell’s equations and application of boundary conditions for t\he particular fixture of interest, always exists for specific electromagnetic field structure. The eigenvalue equation relates permittivity, resonant frequency f T , sample and fixture dimensions in the form of a transcendental equation,
F( f T ,
E:,
dimensions) = 0.
(35)
The permittivity is a numerical root of eq ( 3 5 ) for a given resonant frequency, sample and fixture dimensions. Although there is more than one (mathematical) root to the eigenvalue equation, it is usually possible t o pick the correct root, since many roots are nonphysical or the permittivity is approximately known. It is also possible, in principle, to obtain a unique solution by using two independent measurements with different (mode) field configurations or by using two samples having different dimensions. The resonant fixtures shown in Fig. 6 are practical, when properly dimensioned, for complex permittivity evaluations of low-and medium-loss materials over the frequency range 1-50 GHz. The TEolp mode cavity (Fig. 6a) has high accuracy, since there is no capacitive coupling between sample and the metal walls of the cylindrical cavity. Permittivity with TEol, mode structure is evaluated in the plane of the cylindrical specimen. If the sample is internally elevated into an electric field maximum in the cavity, higher sample partial electric energy filling
Dielectric Materials and Devices
547
Metal Cavity
I
'
Metal Cavity
/
\ I
Cylindrical Dielectric Samples
/
(a)
Metal Plate
\
Cylindrical Dielectric Samples
Figure 6: Typical cylindrical cavities and dielectric resonator configurations used for permittivity and dielectric loss tangent measurements: (a) TEolp mode cavity, (b) TNIolo mode cavity, (c) TEoll mode dielectric resonator in parallel plate waveguide, (d) TEompdielectric sleeve resonator with rod specimen.
548
Dielectric Materials and Devices
factors result. Higher sample electric energy filling factors increase sensitivity of the resonant system to dielectric loss. The TMolo mode cavity (Fig. 6b), on the other hand, enables permittivity and dielectric loss tangent evaluation along the axis of rod-shaped specimens. If the specimen is uniaxially anisotropic and oriented along the axis of anisotropy, this measurement technique can be used to determine dielectric properties along the anisotropy axis, while a technique using TEolp electric field structure can be used to measure the properties normal to the anisotropy axis. It is important to note that depolarization effects can occur in the TMolo cavity if there are gaps between the top or bottom of the sample and the cavity endplates; hence special flatness requirements for sample preparation are imposed. The TEoll dielectric resonator system (Fig. 6c) is the well-known Hakki-Coleman or Courtney technique [22, 231. It is one of the most accurate for low-loss dielectric specimens. For typical sample dimensional uncertainties of f0.0254 mm, relative uncertainties in real permittivity and dielectric loss tangent are 0.3% and 1x 10-5. The use of dielectric sleeve resonators (Fig. 6d) has several advantages. First, the sleeve resonators may be utilized to control the nominal resonant frequency at which dielectric characterization of an inserted rod specimen is performed. With the use of several low-loss ring resonators, the dielectric properties of a single specimen may be evaluated at several discrete frequencies. Second, if the specimen unknown loss characteristics are relatively high, this approach often permits an accurate determination, since the unloaded Q-factor of the composite system may be measurable while t h a t using the specimen as a single dielectric resonator is not. Third, smaller rod test specimens are required at low frequencies for characterization than for cavity or waveguide transmission methods. The dielectric characteristics of the sleeve resonator(s) must first be accurately measured; then typical uncertainties of the inserted rod specimens are similar to those of the classical Courtney technique for properly identified TEomp modes. Some examples of variable temperature dielectric sleeve resonator measurements of ferroelectric materials are given in Figs. 7 an 8. Several practical resonant fixtures possess geometries for which analytical solutions are not available [21]. These fixtures may be analyzed with numerical RayleighRitz, finite element, or mode-matching methods. These mathematical methods for electromagnetic field analysis permit accuracy improvements in dielectric characterization. In general, measurement uncertainties for (real) permittivity depend on 0
Specimen and resonant system dimensional uncertainties,
0
Uncertainties in resonant frequency measurement,
0
Uncertainties due to any depolarization effects that result from the presence of
Dielectric Materials and Devices
549
900
-
.0 Ba&r,TiO&O
wt. % MgO Ba,S 4sTiOj/20wt. % MgO -8a,Sr',TiOJ40 wt. % MgO Ba,Sr',TiO&CI wt. % MgO icBaSSc:,,TiO3/30 wt % MgO
A
800 700
4
7 A
+ Ba&.,TiOJ/4U
wt % MgO
)r
.- 600 % c1 -bl
*Ea
500
\ \ \
e !j 400 .c,
U
300 200
100 0 -50 -30 -10 10 30 50 70 90 110130150
Temperature (CeIsi us)
Figure 7: Variable temperature dielectric sleeve resonator measurements of ferroelectric materials at frequencies near 1 GHz. Relative permittivity is shown for various barium strontium titanates with added weight percentages of magnesium oxide.
550
Dielectric Materials and Devices
0.01
0.00 1
0.0001
I
-50 -30 -10 10 30 50 70 90 110 130 150 TemDerature (Celsius)
Figure 8: Variable temperature dielectric sleeve resonator measurements of ferroelectric materials at frequencies near 1 GHz. Dielectric loss tangent is shown for various barium strontium titanates with added weight percentages of magnesium oxide.
Dielectric Materials and Devices
55 1
air gaps between sample and fixture conductors (for resonant mode structure leading to a discontinuous electric field component normal to the gap), 0
Uncertainties from electromagnetic numerical analysis.
Evaluation of the dielectric loss tangent for any resonant system containing a dielectrically isotropic specimen is always based on the expression,
where &o is the resonant system unloaded &-factor, P,o,specimen, Pe,sUpport are the partial electric field energy filling factors for sample and any dielectric support, Rs is the surface resistance of any conductor shields, G is the geometrical factor of the resonant system for the particular mode structure under examination, and &;I represents radiated power losses. The partial electric energy filling factor in the specimen under test is defined as the ratio of the electric field energy in the specimen to the total electric field energy in the resonant system. A similar statement can be made for the partial electric field energy filling factor in any dielectric support. The filling factors and geometric factor are derivable as analytic expressions when the electric and magnetic field expressions for specific mode structures of interest are known. For complex dielectric resonator systems, it is often more tenable to evaluate the partial electric energy filling factor from the material perturbation theorem [24],
In general, we choose resonant systems where radiated losses Q;l are negligible, where conductor losses R s / G are minimal, and where the sample partial electric energy filling factor is greatest. This yields maximum sensitivity to sample dielectric loss in low-loss materials and, usually, lowest dielectric measurement uncertainty. Conductor losses, which decrease as the surface resistance becomes small and as the geometric factor increases, must be well-characterized a t the measurement frequency and temperature. A common procedure to minimize conductor losses is to situate the dielectric specimen in a position away from the conductor walls (Fig. 9). The TE016 (0 < S < 1) mode is typically used for dielectric loss tangent and for temperature coefficient measurements. For this mode, geometric factors approach a maximum as L, and B , increase. The optimal value of the geometric factor (optimal positioning of the specimen relative to metal shielding) depends, however, on specimen permittivity. If the distance of the metal shield from the specimen becomes larger than the optimum values, the electric energy filling factor of the sample decreases rapidly, and the field distribution becomes essentially the same as in an empty TEoll cavity. A lower
552
Dielectric Materials and Devices
Figure 9: TEols dielectric measurement system. bound in dielectric loss tangent determination for optimal shielding conditions and for relative permittivities greater than or equal to 30 is approximately 5x 10-7 when conductor losses are accurately known. Another approach to minimize conductor losses is to use superconductors as shielding material. However, this can only be done at temperatures equal to or below the critical temperature of the superconducting material [25, 261. The most effective way to decrease conductor losses (or increase the geometric factor) for accurate dielectric loss tangent measurements is to use a disk-shaped specimen as an isolated dielectric resonator, and to examine higher-order hybrid modes that have high azimuthal mode numbers. Hybrid modes are resonant modes in which all six electric and magnetic field components are present. However, there are hybrid modes which are dominantly TE or T M that can be used for dielectric measurements of low-loss anisotropic materials. As the azimuthal mode number increases relative t o the radial wavenumber in the specimen, more and more of the propagating electromagnetic energy is confined to the specimen perimeter and an inner (caustic) surface. Borrowing from the acoustics literature where they were first observed, these near- surface wave modes are termed whispering gallery modes. A whispering gallery mode resonant system suitable for variable temperature measurements is illustrated in Fig. 10. Because conductor losses decrease very rapidly with increasing azimuthal mode number, whispering gallery modes may be
Dielectric Materials and Devices
553
Figure 10: Whispering-gallery mode resonant system. used for accurate dielectric loss tangent measurements of ultra-low loss materials (tan6 < 1 x 10-'), such as single crystal oxides (e.g., sapphire). With the use of whispering gallery modes, the dielectric loss tangent becomes the reciprocal of the unloaded system Q-factor. Unloaded Q-factors in excess of 10' have been measured at 23 GHz and 18 K with identified quasi-TE whispering gallery modes on oriented sapphire [27], corresponding t o dielectric characterization normal t o the c-axis. The use of whispering gallery modes for dielectric loss tangent evaluation of ultra-low loss, pure crystals enables one to examine the vaildity of various theoretical models on intrinsic and extrinsic sources of loss and, perhaps, quantum mechanical limits on dielectric loss. Use of whispering gallery modes is made difficult, however, by proper mode identification and by the fact that observed resonances are not always Lorentzian in character. When new ceramic phases for oxide systems are being developed, it often becomes difficult to fabricate single phase specimens that are sufficiently large for dielectric evaluation at a nominal frequency of about 1 GHz. In addition, in the development of new ceramic materials for microelectronic applications, the material under test may have higher dielectric losses than acceptable for use as a dielectric resonator. For these types of measurements a coaxial reentrant cavity (Fig. 11) may effectively be employed [28-301. Coaxial reentrant cavities may be used effectively over the nominal frequency range between 100 MHz and 1 GHz. Because
554
Dielectric Materials and Devices
Figure 11: Coaxial reentrant cavity. the electric field in the sample is parallel to the sample’s cylindrical axis and the permittivity is measured in that direction?this method is subject t o depolarization effects in any air gaps between the sample and the center conductor posts. Hence, the top and bottom surfaces of the samples should be metallized. With proper sample preparation, real relative permittivities as high as 600 have been measured with this resonant system. An estimated upper limit on the uncertainty in dielectric loss tangent evaiuation is approximately 5 x 1 0 - ~ . The most commonly used dielectric materials in the electronics industry are those used for printed circuit boards. Typically these dielectric materials exhibit losses in the range, 10-4 < t a n 6 < 10-2. Because printed wiring board substrates are relatively thin dielectric sheets, the use of a sample as a TEoll dielectric resonator is frequently not practical. In addition, for rapid evaluation and checks for dielectric imhomogeneity, a nondestructive measurement technique is desired that is sensitive to dielectric losses in the range indicated. For thin laminar samples, one convenient measurement technique is a dielectrically-loaded waveguide junction (or split, tuned dielectric post resonator system) illustrated in Fig. 12 [31-331. This nondestructive method is practical at frequencies from 1 to 10 GHz. At frequencies above 50 GHz, the dimensions of conventional dielectric resonator system structures become small, leading to high dimensional uncertainties. At these frequencies? a typical method (Fig. 13) for complex permittivity measurements is
Dielectric Materials and Devices
555
Figure 12: Split dielectric resonator system for dielectric evaluation of laminar mat erials . a confocal or semi-confocal Fabry-Perot type resonator {34], although for very low loss materials, the whispering gallery mode technique could also be used. Free space transmission and reflection techniques previously described may also be used t o characterize materials from 50 GHz to visible light frequencies.
SUMMARY Various microwave techniques for evaluating dielectric properties on bulk solid dielectric materials have been reviewed. The preferred technique depends on the frequency and temperature, specimen shape and size, loss and anisotropy, and desired accuracies. The use of waveguide transmission methods, although broadband, have limited accuracy for dielectric loss determination and are constrained by potential depolarization effects. The use of both T E and TM mode structure for dielectric resonator specimens having dielectric loss tangents lower than 10-3 permits dielectric characterization of uniaxially anisotropic materials. When specimens have higher dielectric losses, conventional cylindrical or coaxial reentrant cavities may be employed. The most accurate technique for dielectric loss evaluation of ultra-low loss materials is to employ the specimen as a whispering gallery mode dielectric resonator. The high order whispering gallery modes force resonant system conductor losses to be negligibly small; their use, however, is made difficult by proper mode
556
Dielectric Materials and Devices
c
I I
Upper Semi-sphericai
/IMirror
I I
Mirror
Figure 13: Fabry-Perot semiconfocd resonant system. identification and non-Lorentzian character of hybrid modes.
ACKNOWLEDGMENTS The authors wish to acknowledge Dr. James Baker-Jarvis and Dr. Claude Weil for many useful discussions. We also thank the Army Research Laboratory for providing ferroelectric specimens for testing and far providing insight into the effects of microstructure on dielectric properties of ferroelectric materials. We &o wish to express our appreciation to Dr. Stuart Wolf for partial support of ferroelectric property measurements in the FAME DARPA program.
REFERENCES
J.D. Jackson, Classical Electrodynamics. John- Wiley, New York, 1975. 2A. Nussbaum, Electromagnetic and Quantum Properties of Materials. PrenticeHall, New Jersy, 1966. 3J.C. Maxwell, A Treatise on Electricity and Magnetism. Dover Publ., 1891,
Dielectric Materials and Devices
557
4K .W. Wagner, “Erklarung der Dielectrischen Nachwirkungs Worgange auf Grund Maxwellscher,” Archiv. Electrotechnik, 20 371 (1914).
R.W.P. King and G.S. Smith, Antennas in Matter. MIT Press, Cambridge, MA, 1981. %.S. Dukhin, “Dielectric Properties of Disperse Systems,” in E. Matijevic, ed., Surface and Colloid Science, 3, Wiley Interscience, New York, 1969. 7R.G. Geyer, J . Mantese, J. Baker-Jarvis, “Effective Medium Theory for FerriteLoaded Materials,” Natl. Inst. Stand. Technol. Tech. Note 1371, 1994. 8J.B. Hasted, Aqueous Dielectrics. Chapman and Hall, London, 1973.
’P. Debye, Polar Molecules. Chemical Catalog Co., New York, 1929. l0C.P. Smyth, Dielectric Relaxation and Molecular Correlation in Dielectric and Related Molecular Processes, The Chemical Society, London, 1966. “R.G. Geyer, “Dielectric Characterization and Reference Materials,” Natl. Inst. Stand. Technol. Tech. Note 1338, 1990. 12K.S. Cole and R.H. Cole, “Dispersion and Absorption in Dielectrics, J . Chem. Physics, 9 341-351 (1941). 13G. Williams and D.C. Watts, “Non-Symmetrical Dielectric Relaxation Behavior Arising from a Simple Empirical Decay Function,” Trans. Faraday Soc., 66 80-85 (1970). 14A.K. Jonscher, “The Universal Dielectric Response, a Review of Data and Their New Interpretation,” Chelsea Dielectric Group, Univ. London, 1979. 15A. Von Hippel, Dielectrics and Waves. Wiley, New York, 1954. 16C.J. Bottcher, Theory of Electric Polarization, vol. 1, 2. Elsevier, New York, 1978.
17J.Baker-Jarvis, M.D. Janezic, J.H. Grosvenor, and R.G. Geyer, “Transmission/ Reflection. and Short-circuit Line Methods for Measuring Permittivity and Permeability, Natl.
558
Dielectric Materials and Devices
Inst. Stand. Technol. Note 1355-R, 1993. “D. Kajfez and P. Guillon, Dielectric Resonators. Artech House, 1986. ”R.G. Geyer, C. Jones, and J. Krupka, “Microwave Characterization of Dielectric Ceramics for Wireless Communications,” Advances in Dielectric Ceramic Materials, Am. Cer. Soc. Trans., 88 75-91 (1998). 20E.J. Vanzura, R.G. Geyer, and M.D. Janezic, “The NIST 60- Millimeter Cylindrical Cavity Resonator: Performance Evaluation for Permittivity Measurements,” Natl. Inst. Stand. Technol. Tech. Note 1354, 1993.
21J. Krupka and R.G. Geyer, “Loss-Angle Measurements,” in J.G. Webster, ed., Encyclopedia of Elec. and Electron. Eng., John Wiley, New York, 1999. 22B.W. Hakki and P.D. Coleman, “A Dielectric Resonator Method of Measuring Inductive Capacities in the Millimeter Range,” IEEE Trans. Microwave Theory Tech., MTT-8 402-410 (1960). 23W.E. Courtney, “Analysis and Evaluation of a Method of Measuring the Complex Permittivity and Permeability of Microwave Insulators,” IEEE Trans. Microwave Theory Tech., MTT-18 476-485 (1970). 24Y. Kobayashi, T. Aoki, and Y. Kabe, “Influence of Conductor Shields on the Q-factors of a TEo dielectric resonator,” IEEE M TT -S Int. Microwave Symp. Dig., St. Louis 281-284 (1985). 25J. Krupka, R.G. Geyer, M. Kuhn, and J.H. Hinken, “Dielectric Properties of Single Crystals of A1203, LaA103, NdGaO3, SrTiO3, and MgO at Cryogenic Temperatures,’’ IEEE Trans. Microwave Theory Tech., 42 1886-1890 (1994). 26R.G. Geyer and J. Krupka, “Microwave Dielectric Properties of Anisotropic Materials at Cryogenic Temperatures,” IEEE Trans. Instrum. Meas., 44 329-331 (1995). 27R.G. Geyer, J. Krupka, and M. Tobar, ((Microwave Dielectric Properties of LowLoss Materials at Low Temperature,” Trans. Am. Cer. Soc.: Dielectric Materials and Devices (2000). “A. Milewski, “Coaxial Lumped Capacitance Resonators for the Investigations of
Dielectric Materials and Devices
559
Dielectrics,’’ Electron. Technol., 10 71-98 (1977). 29W. Xi, W.R. Tinga, W.A.G. Voss, and B.Q. Tian, “New Results for Coaxial Re-Entrant Cavity with Partially Dielectric Filled Gap,” IEEE Trans. Microwave Theory Tech., 40 747-753 (1992).
30J. Baker-Jarvis and B .F. Riddle, “Dielectric Measurements Using a Reentrant Cavity: Mode-Matching Analysis,” Natl. Inst. Stand. Technol. Tech. Note 1384, 1996.
31G. Kent, “An Evanescent-Mode Tester for Ceramic Dielectric Substrates,” IEEE Truns. Microwave Theory Tech., 36 1451-1454 (1988). 32A.G. Yushchenko and V.V. Chizhov, “Precision Microwave Testing of Dielectric Substrates,” IEEE Trans. Instr. Meas., 46 507-510 (1997). 33M.D. Janezic and J. Baker-Jarvis, “Full-Wave Analysis of a Split-Cylinder Resonator for Nondestructive Permittivity Measurements,” IEEE Truns. Microwave Theory Tech., 47 2014-2020 (1999). 34A.L. Cullen and P.K. Yu, “The Accurate Measurement of Permittivity by Means of an Open Resonator,’’ Proc. R. Soc. London, Ser. A , 325 493-509 (1971)
560
Dielectric Materials and Devices
AUTHOR AND KEYWORD INDEX A M , 155 Amorphous SrTiO,, 141 Ang, C., 339 Antiferroelectricity, 5 19 Aqueous synthesis, 25 Araujo, E., 123, 131 Auger spectra, 263 Bai, Y., 205 Ballato, A., 369 Balmori-Ramirez, H., 13 Barium dissolution, 247 Barium titanate, 13, 25, 35, 247, 269, 279,289,301 -based Ni-MLCC, 95 powders, characterization of, 35 Bhalla, A., 113, 339 Bismuth titanate, 57 Bless, P.W., 457 Bloating, 1 Blum, F.D., 279 Bonnell, D.A., 503 Bowen, P., 25 Braski, D.N., 263 Bridger, K., 179 BST, 113 Buffer layer, 47 Burns, A.E., 421 Buscaglia, M.T., 25 Buscaglia, V., 25 Cao, G.Z., 221, 349 Ceramic-polymer 0-3 composites, 205 Ceramic tapes, 457 Chazono, H., 95,411 Chen, H., 257 Chen, I., 519 Cheng, 2.-Y., 205 Chiang, C., 247
Dielectric Materials and Devices
Cho, S., 213 Ciftci, E., 35, 279 Cofiring, 493 Composition dependence, 43 1 Conduction properties, 349 Conductivity, 42 1 Core-shell microstructure, 95,411 Crystallization, 123, 131, 483 Cygan, S.P., 311 Dai, S.X., 483 Degradation, 239 Delamination, 1 Diamond, 263 Dielectric loss, 339, 431 Dielectrical applications, 323 Dielectrics, 77, 123, 349,421,483 properties of, 169, 179,227,257,289 Dilatometry, 13 Dogan, F., 69 Domain, 123 structures, 503 switching, 355 Doping, La3+and Ca2+,349 Doubly reentrant cavities, 187 Eiras, J., 105, 123, 131 Elastic modulus, 3 11 Electrical conductivity, 43 1 Electrical properties, 41 1 Electromechanical properties, 179 Electrooptical properties, 131 Electrostrictive properties, 179 Endo, A., 47 Feingold, A.H., 457 Feng, J., 69 Ferrite, Mg-Cu-Zn, 467 Ferroelectricity, 369, 5 19
561
Ferroelectrics, 13,77, 113, 123, 131, 187,221,349,503 properties of, 105 Fluxes, ZNO-based, 289 Forbess, M.J., 221, 349 Force-distance analysis, 503 Fracture toughness, 3 11 Friddle, P.A., 257 Fujikawa, Y., 473 Gadow, R., 323 Garcia, D., 105, 113, 123, 131 Gennanate glass, 431 Geyer, R.G., 187,533 Gharti, V., 205 Glazounov, A.E., 355 Gomez-Yanez, C., 13 Gopalan, P., 289 Gordon, A., 227 Grain surface fractal, 169 Guo, R., 113, 339
562
Jain, H., 431 Jean, J., 247 Jones, L.E., 263 Jordovic, B., 269 Kalinin, S.V., 503 Kanert, O., 43 1 Killinger, A., 323 Kim, D., 155,213 Kim, K., 155 Kishi, H., 95,411 Kocic, L., 169 Kolleck, A., 355 Krishnaswami, S., 431 Krueger, D.S., 227 Krupka, J., 533
Hagiwara, T., 95 Hennings, D., 1 High dielectric constant, 205 Highly accelerated life test, 239, 443 Hill, J.W., 311 Hoffmann, M.J., 355 Hong, K., 213 Hot forging, 57 Howe, J.Y., 263 Huang, R., 483 Hydrolysis, 301 Hydrothermal synthesis, 35, 279 Hypersonic flame spraying, 323 Hysteresis, 123
Lanthanum, 77 Layered perovskite, 349 Lead-containing perovskite, 5 19 Lee, M., 155 Lee, S., 155 Lee, W., 239 Leite, E.R., 141 Lejeune, M., 77 Lente, M., 105 LHPG, 113 Limmer, S., 221, 349 Lin, C.H., 257 Lithium borate glass, 421 Lombardo, S.J., 227 Longo, E., 141 Low-fired X7R dielectric, 443 Low-temperature cofired ceramics, 493 Low-temperature sintering, 467 Lu, X., 257
Impedance measurements, 369 Integrated components, 457 Intergranular capacitors, 269 Ionic transport, 421 Iwasaki, A., 47
Maher, G.H., 443 Mandai, H., 493 Mastellaro, V., 141 McKinzie, W.E., 187 Mechanical properties, 311
Dielectric Materials and Devices
Mendes, A.G., 131 Metallorganic CVD, 257 Metzmacher, C., 1 Microstructural development, 69 Microstructure-propertyrelationship, 269 Microwave conductivity, 43 1 Microwave dielectric properties, 213 measurements of, 533 Mirtovic, I., 169 Mitic, V., 169, 269 Mitrovic, I., 269 Mizuno, Y., 95 Mizutani, N., 47 Modulus spectra, 42 1 Morphotropic phase boundary, 5 19 Multilayer ceramic capacitors, 1, 301, 311,473 nickel-based, 41 1 Multilayer ferrite chip, 467 Multilayer module, 493 Murase, T., 467 Nakahata, I., 467 Nakajima, N.,493 Nakano, A., 467 Nanni, P., 25 Nguyen, C.P., 221, 349 Nomura, T., 467,473 Oxidation, 263 Oxides, 323 Oxygen vacancies, 239 Ozomizer, 323 Particulate coating, 179 Passives, 457 Patwardhan, I.S., 57 PbO loss, 69 Phase constitution, 2 13 Phase diagram, 5 19 Phase transformation, 503
Dielectric Materials and Devices
Piaggio, P., 25 Piezoceramics, 369 Piezoelectricity, 369 Piezoelectrics, 77 Pilgrim, S.M, 77, 179 Plasma spraying, 323 PLZT, 69 PMN, 77, 179 Point defects, 1 Polarization screening, 503 Polarization switching, 105 Pontes, EM., 141 Porosity, 1 Porter, W.D., 263 Powder processing, 227 Prakash, D., 289 Precipitation mechanisms, 25 Protons, 1 Pulsed MO-source CVD, 47 Pyroelectric response, 113 PZT, 47, 105, 123,257,355 Rahaman, M., 35,57,279 Rama Mohan, T.R., 289 Rare-earth addition, 4 11 Rare-earth doping, 473 R-curve behavior, 355 Reaction mechanisms, 25 Relaxors, 77, 179, 519 Residual stres, 155 Resistance, 239 Resonators, 369 Riester, L., 3 11 Rossetti, G.A., Jr., 27 RTA, 123, 131 Saiki, A., 47 Sakabe, Y., 301 Sato, S., 473 SBN, 131,221 Scanning probe microscopy, 503 Schneider, G.A., 355
563
Schreinemacher, S., 1 Sehirlioglu, A., 179 Self-heating, 105, 123 Seraji, S., 221, 349 Sharma, B.P., 289 Shende, R.V., 227 Shinozaki, K., 47 Shumsky, M., 35,279 Single crystals, 113, 339 Size effect, 301 Sodium trisilicate glass, 421 Sol-gel, 141 Solid solutions, 227 Solid-phase reaction, 301 Stein, M.A., 457 Stein, S.J., 457 Stress relaxation, 47 Strontium titanate, 227, 339 Strontium zirconate, 227 Suspensions, 247 Switching, 123 Synthesis, 227
Wada, N., 301 Wahlers, R.L., 457 Wakino, K., 493 Wakiya, A., 47 Wereszczak, A.A., 3 11 Wilcox, D., Sr., 483 Winter, M., 77 Wireless applications, 457, 483 wu, Y., 221, 349
XRD, 155,355 Xu, H.S., 205 Yoo, M., 155 Yoon, C.H., 179 Youn, H., 213 Yu, z., 339 Zanetti, S.M., 141 Zhang, Q.M., 205 ZnNb,O,-TiO,, 2 13
Temperature-capacitance characteristic, 473 Temperature-stable LTCC, 483 Tetragonality, 30 1 Texture, 57 Thermal conductivity, 3 11 Thermal expansion, 311 Thermistors, 13 Thickness effect, 257 Thin films, 47, 123, 131, 141, 257, 279, 339 Tunability measurements, 187 Vanadates, 22 1 Varela, J.A., 141 Viswanath, D.S., 227 Viviani, M., 25
564
Dielectric Materials and Devices
