Ceramic and Composite PDF

CERAMICS AND COMPOSITE
MATERIALS: NEW RESEARCH
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
CERAMICS AND COMPOSITE
MATERIALS: NEW RESEARCH
B.M. CARUTA
EDITOR
Nova Science Publishers, Inc.

New York
Copyright 2006 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
For permission to use material from this book please contact us:
Telephone 631-231-7269; Fax 631-231-8175
Web Site: http://www.novapublishers.com
NOTICE TO THE READER
The Publisher has taken reasonable care in the preparation of this book, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or
omissions. No liability is assumed for incidental or consequential damages in connection with
or arising out of information contained in this book. The Publisher shall not be liable for any
special, consequential, or exemplary damages resulting, in whole or in part, from the readers
use of, or reliance upon, this material.
This publication is designed to provide accurate and authoritative information with regard to
the subject matter covered herein. It is sold with the clear understanding that the Publisher is
not engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF
THE AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
Library of Congress Cataloging-in-Publication Data

Available upon request
ISBN 978-1-60876-508-9 (E-Book)
Published by Nova Science Publishers, Inc. New York

CONTENTS
Preface vii
Chapter 1 New Approaches for Estimating and Improving the Grain-
Boundary Conduction in Stabilized Zirconia 1
Jong-Heun Lee
Chapter 2 Effective Elastic Moduli of Alumina, Zirconia and Alumina-
Zirconia Composite Ceramics 31
Willi Pabst and Eva Gregorov
Chapter 3 Progress in Bioceramic Materials 101
P. N. De Aza
Chapter 4 Novel Bioactive Hydroxyapatite-Base Ceramics and Gelatin-
Impregnated Composites 133
V.S. Komlev and S.M. Barinov
Chapter 5 Thermal Stresses in Particle-Matrix System and Related
Phenomena. Application to Sic-Si3N4 Ceramics 147
Ladislav Ceniga
Chapter 6 Novel Bone-Repairing Materials: Bioactive Organic-Inorganic
Hybrids 197
Masanobu Kamitakahara
Chapter 7 Effect of Sample Size and Distribution Parameters in Estimation
of Confidence Lower Bounds for Weibull Percentiles 215
Burak Birgoren
Index 237
PREFACE
In stabilized zirconia, which is one of the representative solid electrolytes, the grain-
boundary resistivity is known to be ~102-104 times higher than the grain-interior one. For the
applications such as solid oxide fuel cells and electrochemical gas sensors operated at the
intermediate (~600oC) and low temperature (~400oC) regime, the improvement in ionic
conduction across the grain boundary becomes significant. For this, a precise estimation of
the grain-boundary conduction via impedance spectroscopy is essential. In Chapter 1, New
Approaches for Estimating and Improving the Grain-Boundary Conduction in Stabilized
Zirconia, Jong-Heun Lee suggests the new methods for improving and estimating the grain-
boundary conduction in stabilized zirconia. In the first part, the various approaches to
improve the grain-boundary conduction are discussed. The addition of Al2O3 is known to
scavenge the siliceous grain-boundary phase. However, this might deteriorate the grain-
interior conduction when the sintering temperature becomes very high (>1600C). Therefore,
new routes for improving the grain-boundary conduction using two-stage sintering process
are suggested. The formation of a Si-containing phase in a discrete configuration and the
dewetting of the intergranular liquid phase were suggested to be the mechanisms for
scavenging via pre- and post-sintering heat treatments, respectively. In the second part, a local
impedance technique using a sub-millimeter-scale electrode array was suggested to estimate
the spatially uneven distribution of the grain-boundary resistivity, which was named as
Millicontact Impedance Spectroscopy. The fundamentals and validity of the technique were
explained, and the analyses of the dynamic rearrangement of an intergranular liquid phase are
given as an example.
In the second chapter, Effective Elastic Moduli of Alumina, Zirconia and Alumina-
Zirconia Composite Ceramics, Willi Pabst and Eva Gregorov investigate from the
theoretical point of view, with a side-glance on experimental results and applications. In the
first section alumina, zirconia and alumina-zirconia composites are introduced as structural
materials, relations of elastic moduli to other properties are recalled and targets of
microstructural design are formulated. In the second section elastic properties are defined
from the viewpoint of rational mechanics for anisotropic and isotropic materials in general.
The difference between adiabatic and isothermal elastic moduli is explained and estimated for
alumina and zirconia. In the third section effective elastic properties are defined and discussed
from the viewpoint of micromechanics and composite theory. General formulae are given for
the calculation of effective elastic moduli of polycrystalline materials from monocrystal data.
Further, the Voigt-Reuss bounds for the effective elastic moduli of multiphase materials are
viii B. M. Caruta
given, as well the Hashin-Shtrikman bounds for the special case of two-phase materials. For
porous materials the dilute approximations are recalled as well as the predictions following
from the self-consistent, Mori-Tanaka, differential, Gibson-Ashby and Coble-Kingery
approach as well as the functional equation approach recently advocated by the authors. A
comprehensive survey of model relations for the porosity dependence of elastic moduli is
given, including exponential and power-law relations and a new relation recently proposed by
the authors, which seems to be the simplest relation allowing for the occurrence of a
percolation threshold (critical porosity). In the fourth section all these micromechanical
bounds and relations are applied to the alumina-zirconia system. Using a theoretically sound
and experimentally confirmed set of elastic moduli (and Poisson ratios) for dense (i.e. fully
sintered) polycrystalline alumina and zirconia the Hashin-Shtrikman bounds of dense
alumina-zirconia composites are calculated and compared to experimentally measured values.
Several predictions for porous alumina, zirconia and alumina-zirconia composites are
compared to the data measured for ceramics with convex interconnected pores prepared by
the starch consolidation casting technique. A master fit curve is given for porous ceramics
with this type of matrix-inclusion microstructure and explicit numerical expressions are given
throughout. The last section gives examples of the mathematical modeling of other effective
properties and their dependence on composition and microstructure and an outlook is given to
future research aims. In particular, the significance of interfaces is emphasized and ideas on
the way from micromechanics to nanoscience towards a general mixture theory are
outlined.
A strong interest in the use of ceramics for biomedical engineering applications
developed in the late 1960s. Used initially as alternatives to metallic materials in order to
increase the biocompatibility of implants, bioceramics have become a diverse class of
biomaterials presently including three basic types: relatively bioinert ceramics maintain their
physical and mechanical properties in the host and form a fibrous tissue of variable thickness;
surface reactive bioceramics which form a direct chemical bonds with the host; and
bioresorbable ceramics that are dissolved with the time and the surrounding tissue replaces it.
In Progress in Bioceramic Materials, P. N. De Aza gives a review of the composition,
physicochemical properties and biological behaviour of the principal types of bioceramics,
based on the literature and some of our own data. The materials include, in addition to
bioceramics, bioglasses and bio-glass-ceramics. Special attention is given to structure as the
main physical parameter determining nor only the properties of the materials, but also their
reaction with the surrounding tissue.
In addition, biomaterials with appropriate three-dimensional geometry (appropriate
porous structure) are highly recommended because are able to become osteoinductive
(capable of osteogeneis), and can be effective carriers of bone cell seeds. A new way of
preparing in situ porous bioactive ceramics with interconnected porosity from a dense
bioactive ceramic is presented. To this purpose the binary system wollastonite- tricalcium
phosphate is selected, taking into account the different bioactivity behavior of both phases.
The present status of research and development of bioceramic is characterized as first
step in the solution of complex problems at the confluence of materials science and
engineering, biology and medicine.
In Novel Bioactive Hydroxyapatite-Base Ceramics and Gelatin-Impregnated Composites,
V.S. Komlev and S.M. Barinov introduce a study aimed at the development of novel ceramic
and composite materials intended for application in bone tissue engineering. A method to
Preface ix
produce porous spherical hydroxyapatite (HA) granules was proposed. The method is based
on liquids immiscibility effect using the HA/gelatin suspension and oil as liquids. The
granules of 50 to 2000 m diameter contain open pores of 10 nm to 10 m size in amount up
to 45 vol.%. A route for the fabrication of porous HA ceramics having two population of
open pores was developed. Ceramics contain intragranular, up to 10 m size pores and
intergranular, hundreds micrometer size interconnected pores were prepared. The
interconnections in the intergranular pores are the pathway to conduct cells and vessels for the
bone ingrowth. The gelatin solution impregnation into the pores was demonstrated to enhance
strength significantly. The strength of HA/gelatin composites was increased by 6 to 10 times
compared to that of initial ceramics. The ceramics can further be filled with a drug resulting
in a drug delivery system of slow drug release rate.
In Chapter 5, Thermal Stresses in Particle-Matrix System and Related Phenomena.
Application to Sic-Si3N4 Ceramics, Ladislav Ceniga deals with elastic thermal stresses in an
isotropic particle-matrix system of homogeneously distributed spherical particles in an
infinite matrix, divided into cubic cells containing a central spherical particle embedded in a
matrix of a dimension equal to an inter-particle distance. Originating during a cooling process
as a consequence of the difference in thermal expansion coefficients between the matrix and
the particle, and investigated within the cubic cell, the thermal stresses of the isotropic multi-
particle-matrix system are thus functions of the spherical particle volume fraction, v, and are
transformed for v = 0 to those of the isotropic one-particle-matrix system. With regard to the
particle-matrix system yield stress, the thermal stresses are derived for such temperature
range within which the particle-matrix system exhibit elastic deformations. Similarly, the
particle-matrix boundary adhesion strength is also considered.
In addition to the thermal-stress-induced elastic energy of the cubic cell, the elastic
energy gradient within the cubic cell, calculated by two equivalent mathematical techniques
and representing a surface integral of the thermal-stress-induced elastic energy density, is
presented to derive the thermal-stress strengthening in the spherical particle and the cubic cell
matrix.
Considering a curve integral of the thermal-stress-induced elastic energy density, the
critical particle radii related to crack formations in ideal-brittle particle and matrix, functions
describing crack shapes in a plane perpendicular to a direction of the crack formation in the
particle and the matrix, and consequently particle and matrix crack dimensions are derived
along with the condition concerning a direction of the particle and matrix crack formation.
The former parameters for v = 0 are derived using a spherical cell model for the spherical cell
radius, Rc .
The derived formulae are applied to the SiC-Si3N4 multi-particle-matrix system, and
calculated values of investigated parameters are in an excellent agreement with those from
published experimental results.
In general, artificial materials implanted into bone defects are encapsulated by a fibrous
tissue. Some ceramics, such as Bioglass, sintered hydroxyapatite and glass-ceramic A-W,
however, form a bone-like apatite layer on their surfaces in the living body and bond to living
bone through this apatite layer, i.e. they show bioactivity. Although these bioactive ceramics
are used clinically as important bone-repairing materials, they are essentially brittle and hence
limited in their applications. It is desirable to develop new types of deformable bioactive
materials. Organicinorganic hybrids prepared by a sol-gel method are expected to exhibit
x B. M. Caruta
bioactivity as well as deformability. Polydimethylsiloxane (PDMS)-CaO-SiO2-based hybrids

were prepared.
To evaluate their potential of bone-bonding property, their ability of apatite formation
was examined in a simulated body fluid (SBF) in Chapter 6, Novel Bone-Repairing
Materials: Bioactive Organic-Inorganic Hybrids by Masanobu Kamitakahara. Since the
apatite formation on bioactive materials in the body can be reproduced even in SBF, the
author can estimate the bioactivity of the material by using SBF. Mechanical properties were
also examined. Some of them showed apatite-forming ability and mechanical properties
analogous to those of human cancellous bone. These CaO-containing hybrids, however,
showed a decrease in mechanical strength in SBF. These hybrids containing no CaO do not
form apatite on their surfaces in SBF. The author then prepared CaO-free PDMS-TiO2 and
poly(tetramethylene oxide) (PTMO)-TiO2 hybrids in which anatase was precipitated by hot-
water treatments. These hybrids also showed apatite-forming ability on their surfaces in SBF
and flexibility, and they showed little decrease in mechanical strength in SBF. In conclusion,
it was revealed that we can design new bioactive materials for bone-repair by controlling the
compositions and structures in organic-inorganic hybrid systems.
The Weibull distribution is widely used for modelling fracture strength of composite
materials. Estimating lower percentiles of the Weibull distribution has been a major concern,
because small sizes of fracture strength experiments leads to unreliable estimates. Therefore,
several recent studies have focused on estimating confidence lower bounds on the lower
percentiles, namely A-basis and B-basis material properties. The general linear regression
methods with different probability indices and weight factors, and the maximum likelihood
method have been used for the estimation; the performance of these methods were compared
in recent studies and different methods were proposed using different criteria for comparison.
Aside from this ambiguity, the criteria and comparison studies do not allow any practical
interpretation. In particular, experimenters would like to know, in practical terms, how well a
method performs with respect to others and how its performance is affected by the sample
size. Burak Birgoren, in Effect of Sample Size and Distribution Parameters in Estimation of
Confidence Lower Bounds for Weibull Percentiles, intends to clarify answers to these
questions. Percent departures of the lower bounds from the true percentiles convey such a
practical meaning. Simulating these departures, probabilistic upper bounds for maximum
percent departures are determined; tables are provided for computing the upper bounds for a
wide range of sample sizes. Using the upper bounds as a criterion for comparison, it is
observed that the maximum likelihood method is the best as compared to several variations of
the general linear regression methods. Its superiority is illustrated by comparing the sample
sizes necessary for different methods to achieve the same level of performance. For the
maximum likelihood method, the upper bound levels are shown to increase at a higher rate
for sample sizes less than 30 as compared to larger sizes. Also, the Weibull modulus is
observed to have a very serious effect on the upper bounds. Finally, a two-stage experimental
setup is proposed for sample size determination; after running a pilot experiment, the setup
enables the experimenter to calculate the number of additional experiments necessary for
lowering the upper bounds to a certain level. Its implementation is illustrated on a practical
example.
In: Ceramics and Composite Materials: New Research ISBN: 1-59454-370-4
Editor: B.M. Caruta, pp. 1-30 2006 Nova Science Publishers, Inc.
Chapter 1
NEW APPROACHES FOR ESTIMATING AND

IMPROVING THE GRAIN-BOUNDARY
CONDUCTION IN STABILIZED ZIRCONIA
Jong-Heun Lee*
Division of Materials Science and Engineering,
Korea University, Seoul 136-701, Korea
Abstract
In stabilized zirconia, which is one of the representative solid electrolytes, the grain-boundary
resistivity is known to be ~102-104 times higher than the grain-interior one. For the
applications such as solid oxide fuel cells and electrochemical gas sensors operated at the
intermediate (~600oC) and low temperature (~400oC) regime, the improvement in ionic
conduction across the grain boundary becomes significant. For this, a precise estimation of the
grain-boundary conduction via impedance spectroscopy is essential. This article suggests the
new methods for improving and estimating the grain-boundary conduction in stabilized
zirconia. In the first part, the various approaches to improve the grain-boundary conduction
are discussed. The addition of Al2O3 is known to scavenge the siliceous grain-boundary phase.
However, this might deteriorate the grain-interior conduction when the sintering temperature
becomes very high (>1600C). Therefore, new routes for improving the grain-boundary
conduction using two-stage sintering process are suggested. The formation of a Si-containing
phase in a discrete configuration and the dewetting of the intergranular liquid phase were
suggested to be the mechanisms for scavenging via pre- and post-sintering heat treatments,
respectively. In the second part, a local impedance technique using a sub-millimeter-scale
electrode array was suggested to estimate the spatially uneven distribution of the grain-
boundary resistivity, which was named as Millicontact Impedance Spectroscopy. The
fundamentals and validity of the technique were explained, and the analyses of the dynamic
rearrangement of an intergranular liquid phase are given as an example.
*
Corresponding author Jong-Heun Lee, Ph. D., Associate Professor, Division of Materials Science and Engineering,
Korea University, 1, 5-ka, Anam-dong, Sungbuk-ku Seoul 136-701, South Korea. Tel: 82-2-3290-3282, Fax:
82-2-928-3584. Email: jongheun@korea.ac.kr
2 Jong-Heun Lee
1 Introduction
When divalent or trivalent oxides such as CaO, MgO, Y2O3, Yb2O3 and Sc2O3 are added to
ZrO2, the high-temperature cubic phase becomes stabilized at room temperature. [1] The
effective negative charge of the dopant cation in the Zr4+ site is compensated by the formation
of a doubly ionized oxygen vacancy (VO), which leads to the stabilized ZrO2 being a pure
oxygen ion (O2-) conductor. [1,2] A high ionic conductivity is essential in order to achieve a
high-energy efficiency and a low operation temperature in a solid oxide fuel cell (SOFC) and
an electrochemical sensor. [3] The improvement in the ionic conductivity in stabilized
zirconia has been studied extensively in the viewpoints of effective dopants and their
optimum concentrations. [1,2,4,5] Among the types of stabilized zirconia, Yb2O3-stabilized or
Sc2O3-stabilized zirconia (YbSZ or ScSC) showed the highest conductivity, which is
explained by the minimum lattice mismatch between the dopant cation (Yb3+ and Sc3+) and
the host Zr4+ ion. [4] However, these efforts to increase the grain-interior conductivity are in
the mature stage.
Siliceous
phase
[VO]
2-
O
2-
O
(a) (b)
Fig.1 The origins of the grain boundary resistance. (a) space charge layer and (b) siliceous intergranular
phase
Another approach is to minimize the grain-boundary resistance. The grain boundary is

known to be specifically ~102-104 times more resistive than the grain interior. [6-8] Two
reasons for the origins for high grain-boundary resistivity have been suggested. One is the
existence of a space charge layer near the grain boundary with the lower oxygen vacancy
concentration, which becomes significant in pure materials. (Fig.1(a)) [9-12] The other is an
oxygen-ion-blocking siliceous phase or film at the grain boundary. (Fig.1(b)) [8,13] Even 100
ppm of SiO2 is known to significantly deteriorate the grain-boundary conduction. [14]
Considering that SiO2 is the most ubiquitous impurity, siliceous contamination is almost
inevitable during ceramic processing. Therefore, the studies on improving the grain-boundary
conduction have been focused on the addition of a scavenger to clean the resistive grain-
boundary phase. Thus far, Al2O3 is known to be the most effective scavenger.[14-20]
However, when the sintering temperature becomes >1600oC, the improvement in the grain-
boundary conduction is known to be nullified by the deterioration of grain-interior conduction.
[21,22] This is related to the incorporation of Al2O3 into the ZrO2 lattice and the subsequent
defect association between AlZr and VO.[23]
Therefore, in order to avoid the adverse effect of Al2O3, two new approaches for
improving the grain boundary conduction without the addition of Al2O3 were suggested.
New Approaches for Estimating and Improving 3
These methods are based on the configuration change in the intergranular phase from
continuous one to a rather discrete one by the additional heat treatments before or after the
sintering process. The mechanism for enhancing the grain-boundary conduction was
investigated. The key idea of this method, which is the sensitive correlation between the
grain-boundary resistance and intergranular phase, can be also applied to a new research field.
In the second part of this contribution, a local impedance method combined with a
submillimeter-scale-electrode array, which is called millicontact impedance spectroscopy,
was suggested to investigate the spatially-uneven grain boundary. The dynamic
rearrangement of the intergranular liquid during liquid-phase sintering, which had not been
observed in scanning electron microscopy, could be successfully observed via the
millicontact impedance technique.
2 The Model of Grain-Boundary Resistivity in Stabilized

Zirconia Ceramics
The complex impedance spectrum of stabilized zirconia generally consists of three
semicircles. (Fig.2) From the low frequency, three semicircles can be attributed to the
contributions by electrode polarization, grain boundary and grain interior, respectively.
Typically, the grain-boundary capacitance (Cgb) is much larger than the grain-interior
capacitance (Cgb) because the grain boundary is very thin compared with grain size. However,
the grain-boundary resistance (Rgb) was reported to be comparable with the grain-interior
resistance (Rgi). Therefore, the relaxation time constant for the grain boundary (gb), the
product between Rgb and Cgb, becomes markedly larger than that that of the grain interior (gi).
This is the reason why the contributions from the grain interior and grain boundary can be
separated effectively in a complex impedance plane.
gi gb ep
Cgi Cgb Cep
Rgi Rgb < Rep
-Z()
freq.
Z()
Fig.2 Schematic complex impedance spectrum of the stabilized zirconia. R, C, and mean the
resistance, capacitance and relaxation time constant, respectively. From the low frequency, subscripts
ep, gb, and gi mean the electrode polarization, grain boundary and grain interior, respectively.
In a stabilized zirconia specimen with a defined thickness (l) and electrode area(A)
(Fig.3(a)), the grain-interior resistivity (gi) can be calculated by normalizing the
deconvoluted Rgi value using a shape factor.
4 Jong-Heun Lee
l
Rgi = gi (1)
A
l
Porous Pt
electrode
A
O2-
(a)
gb
dg
O2-
(b)
Fig.3 (a) Schematic diagram showing the ionic conduction in polycrystalline stabilized zirconia
specimen with a resistive grain boundary and (b) its simplification by the brick layer model. (A:
electrode area, l: specimen thickness, dg: average grain size, gb : grain boundary thickness)
However, because l is not the dimension of the grain boundary, normalization of Rgb by
shape factor does mean not the specific grain-boundary resistivity (gbsp) but the apparent
grain-boundary resistivity (gbapp).
l
Rgb = gb
app
(2)
A
In order to calculate gbsp, the number and thickness of the grain boundary across the
electrode needs to be considered. In order to make the analysis simple, the brick layer model
is generally being used as an effective approximation. [7,13,24-26] This model assumes that
the electrically identical cubic grains (size=dg) are surrounded by the same the grain boundary
thickness (gb). (Fig.3(b)) When the grain boundary is highly conducting compared with the
grain interior, ionic conduction will become a competition between the fast ionic conduction
along the narrow grain boundary and the slow ionic conduction along the wide grain interior.
Therefore, a parallel equivalent circuit is established. (Fig.4(a)) However, in stabilized
zirconia, the grain boundary is generally 102-104 times more resistive than the grain interior.
In this situation, the conduction along the resistive grain boundary can be neglected.
Therefore, a series of two RC lumps can be summarized as an equivalent circuit. (Fig.4(b))
dg gb
Rgb
Cgb
O2-
Rgi
Cgi
(a) parallel equivalent circuit: gi gbsp
Rgi Rgb
O2-
Cgi Cgb
(b) serial equivalent circuit: gi gbsp
Fig.4 (a) Schematic diagrams showing the ionic conduction in specimen with highly conductive grain
boundary and the corresponding parallel equivalent circuit. (b) Schematic diagrams showing the ionic
conduction in the specimen with a highly resistive grain boundary and the corresponding serial
equivalent circuit.
Under the assumption of a brick layer model and serial equivalent circuit, gbsp can be
calculated by the following equation: [8]
dg
gbsp = gbapp (3)
gb
However, in real situations, a precise estimation of gb is very difficult because the grain-
boundary segregation and intergranular liquid phase are too thin to observe and the thickness
of the space charge layer is difficult to determine. Therefore, the normalization of gbapp by
the grain-boundary density (d: number of grain boundary per unit length) can be used to
compare the grain-boundary conduction in a specific manner. [26] The grain-boundary
density can be calculated by the reciprocal of the average grain size (dg).
gbapp
Rgbs = = gb
app
d g = gb
sp
gb (4)
d
The Rgbs value means the resistance for the unit grain-boundary area and its physical
meaning is the product between gbsp and gb. Therefore, without a precise determination of
gb, whether the grain-boundary conduction has improved or deteriorated can be determined.
6 Jong-Heun Lee
As a measure for the improvement in the grain-boundary conduction, the scavenging

factor SF can be defined as follows:
Rgbs ( AS )
SF = (5)
Rgbs ( BS )
where Rgbs(BS) and Rgbs(AS) are the Rgbs values before and after the scavenging reaction,
respectively. SF<1 and SF>1 mean an improvement and deterioration of grain-boundary
conduction, respectively and a SF below 1 means the more effective scavenging of the
resistive intergranular phase.
3 Improvement of Grain-Boundary Conduction by the Addition

of Al2O3
Many additives such as Fe2O3 [6], Bi2O3 [6], TiO2 [27], and Al2O3 [6,13-20,27-29] have been
used to improve the grain-boundary conduction of stabilized zirconia via the scavenging of
resistive siliceous phase. Among the additives used, Al2O3 is known to enhance the ionic
conduction in the grain boundary. [14-20] However, there are also contrasting reports.
[6,26,29] Table 1 summarizes the literature on the positive and negative role of Al2O3. [13-
18,26,30] The improvement in the grain-boundary conduction is usually explained in terms of
the scavenging of the siliceous phase by Al2O3. Butler and Drennan [19] examined the TEM
microstructure of Al2O3-added Y2O3/Yb2O3-stabilized zirconia ceramics. They reported that
most intragranular Al2O3 particles contain Zr-rich or (Si+Zr)-rich inclusions and that
intergranular Al2O3 particle are frequently associated with (Si and Al)-rich amorphous cusp.
From the observation, they suggested that Al2O3 scavenged the siliceous second phase during
the pinning time that was assisted by fast grain-boundary diffusion.
In order to examine the effect of Al2O3 configuration on the scavenging efficiency, Lee et
al. [21] added various Al2O3 particles sized 0.3, 4, and 10 m to 8 mol% of Yttria-Stabilized
Zirconia (8YSZ) via ball milling and ultrasonic dispersion. The pure 8YSZ powder without
Al2O3 was also ball-milled or ultrasonically dispersed to exclude the effect of background
impurities, which might be introduced during the mixing process, with the results shown in
Figure 5. The small and large black spots in Fig.5(b) and (c) represent the fine and coarse
Al2O3 particles added, respectively. The 0.3 m Al2O3 particles dispersed well (Fig.5(b)) at
the inter- and intra-granular points. Although the same mol% of Al2O3 was added, the average
spacing between the 10 m Al2O3 particles was very long (~20 m) compared with that
between the 0.3 m Al2O3 particles because the total number of Al2O3 particles was quite
small. (Fig.5(c)) Despite these large differences in the concentration and size of Al2O3, all the
specimens showed almost the same scavenging factor (Rgbs(AS)/Rgbs(BS) =~0.2). (Fig.5(a))
This strongly supports the suggestion by Butler and Drennans [19] that scavenging occurs
via the fast grain-boundary diffusion. The majority of research results showed an
improvement in grain-boundary conduction as a result of Al2O3 addition.
Table 1. The effect of Al2O3 addition on the grain-boundary conduction of stabilized zirconia
1
or gb
app
Materials Al2O3 Tsint(oC) gi Comments Ref.
Rgbs
10 wt%
2.5YSZ 1600 [15]
(~20 mol%)
Solubility limit of
1 wt%
8YSZ 1500 alumina= 0.6~1.2 mol% [16]
(~2 mol%)
at 1500oC
Solubility limit of
9.9YSZ 2 mol% 1350 alumina= 0.1mol% at [17]
1300oC
9YSZ 1.5,2 mol% 1600 [18]
8YSZ 1 mol% 1600 [14]
Solubility limit of
8YSZ 0.5 mol% 1700 alumina= 0.5mol% at [26]
1700oC
Explained by the
increase of grain-
8YSZ 0.4 mol% 1650 [29]
boundary space-charge
potential
8.5YSZ 0.78 mol% 1267 [6]
0-1wt%
Minimum gb at
app
Al2O3
3YSZ 1500 [13]
0-1wt% SiO2/Al2O3=1
SiO2
1450 Dissolving of alumina at
high Tsint
15CSZ 1 mol% [30]
1600 deterioration of grain-
boundary conduction
However, the deterioration in grain-boundary conduction via the addition of Al2O3 has
also been reported. [6,26,29] This inconsistent tendency might be due to the different
siliceous impurity and/or the different interaction between Al2O3 and siliceous impurity. For
example, Gdikemier et al. [13] varied the SiO2/Al2O3 molar ratio systematically in 3 mol%
Yttria-Stabilized Zirconia (3YSZ) and attained the largest Rgbs value near SiO2/Al2O3=1. Both
the SiO2-rich or Al2O3-rich compositions showed smaller Rgbs values than Rgbs at
SiO2/Al2O3=1. Lee et al. [30] added Al2O3 sized 4 and 10 m to 15 mol% Calcia-Stabilized
Zirconia (15CSZ) and attained two opposite roles of Al2O3 in grain-boundary conduction
(Fig.6). When the sintering temperature is lower than 1525oC, the grain-boundary
conductivity improved up to 15 times by the addition of Al2O3. However, as the sintering
temperature becomes higher than 1550oC, the grain-boundary conductivity deteriorates
significantly. Figure 6(b) and (c) show typical microstructures for the 15CSZ specimens
doped with 1 mol% of coarse Al2O3 particles (size=10 m). The 10 m sized Al2O3 particle
preserved its shape (points A and B in Fig.6(b)) when sintered at 1450oC, while most of the
Al2O3 particles disappeared (see empty pores C in Fig.6(c) sized about ~10 m) and a wetting
8 Jong-Heun Lee
intergranular liquid phase could be observed (see arrows in Fig.6(c)). Therefore, The
deterioration in the grain-boundary conduction at >1525oC can be attributed to the formation
of wetting intergranular liquid phase as a result of the dissolution of Al2O3 into siliceous
intergranular liquid.
BM: Ball Milling UD: Ultrasonic Dispersion
Rgbs(AS)/Rgbs(BS) 1.0
0.8
0.6
0.4
0.2
0.0
2000 gb 2.0
[.cm]
Rgbs[.cm ]
2
Rgbs
1500 1.5
app
1000 1.0
gb
500 0.5
0 0
8Y 02A03 1A03 8Y 02A4 02A10 1A10
Samples
(a)
(b) (c)
10 m 10 m
Fig.5 The grain-boundary resistivity (gb), resistance per unit grain-boundary area (Rgbs), and
scavenging efficiency (SF) of 8YSZ specimens with various concentrations and sizes of Al2O3 particles.
(8Y: 8mol% Yttria-Stabilized Zirconia, 02A03: the addition of 0.2 mol % Al2O3 sized 0.3 m, 02A4: the
addition of 0.2 mol % Al2O3 sized 4 m, 1A10: the addition of 1.0 mol % Al2O3 sized 10 m) All the
specimens were sintered at 1500oC for 4h and the impedance was measured at 400oC in air, according
to [21].
The above results suggest that the role of Al2O3 is not always the same but depends on
the degree of siliceous contamination, the sintering temperature and the temperature for
forming the intergranular liquid phase. When a high sintering temperature is combined with a
large concentration of siliceous impurity, the possibility for the deterioration of grain-
boundary conduction due to Al2O3 dissolution into the intergranular liquid becomes large. In
order to minimize such possibility, a lower level of siliceous impurity is essential. Indeed,
when the siliceous impurity is low, the addition of Al2O3 is known to effectively improve the
grain-boundary conduction of the stabilized zirconia. However, direct evidence for the
scavenging role of Al2O3 in specimens with a segregation-level impurity is difficult to find.
This arises from the difficulties in the observation of impurity segregation or the extremely
thin intergranular phase. For example, even a sub-monolayer segregation of the siliceous
phase, which is very difficult to estimate, is known to affect the grain-boundary conduction.
[8] Even if there were a siliceous film at the grain boundary, the observation of many
nanometer-scale intergranular films with statistical confidence is still difficult.
10
15CSZ-1A4 Deterioration of
G.B. conduction
15CSZ-1A10
Rgbs(AS)/Rgbs(BS)
(c)
0.1
Improvement of G.B. conduction
1450 1500 1550 1600

o
Sintering temperature( C)
(a)
Scavenging resistive g.b. phase Dissolution of Al2O3 into intergranular liquid
by 10 m Al2O3 particles
(b) (c) C
b)
A
A
C
B C
B C
10 m B 10 m
Fig.6 (a) The scavenging efficiency (SF) of the 15CSZ specimens as a function of varying sintering
temperatures: 15CSZ1A4: 1-mol%-Al2O3(size=4m)-doped 15CSZ specimens, 15CSZ1A10: 1-mol%-
Al2O3(size=10m)-doped 15CSZ specimens. The microstructures of 15CSZ1A10 specimens sintered (b)
at 1450oC and (b) 1600oC for 4h, according to [30].
Lee et al. examined the spatial distribution of 100 ppm-level-siliceous impurity in 8YSZ
and 1-mol%-Al2O3-doped 8YSZ specimens using imaging secondary ion mass spectroscopy.
[14,21] From the secondary-ion images of 28Si-, and 27Al16O- for two specimens, it was found
that the siliceous impurity is segregated at the grain boundary in pure 8YSZ specimen and
gathers near the Al2O3 particles in 1-mol%-Al2O3-doped 8YSZ specimen. This provided the
direct and visual evidence for the presence of siliceous segregation and its scavenging as a
result of Al2O3 addition.
10 Jong-Heun Lee
On the other hand, Guo [29] reported that the addition of Al2O3 changes the structure of
the space charge layers as well as the grain-boundary contamination. He formulated the
deterioration of the grain-boundary conduction by the increase in the grain-boundary space-
charge potential within the solubility limit of Al2O3. However, he suggested that the addition
of Al2O3 above the solubility limit improves the grain-boundary conduction via a scavenging
reaction.
4 Improvement in Grain-Boundary Conduction Via Two-Stage

Sintering Approach without Additive
4.1 Improvement in the Grain-Boundary Conduction by Pre-sintering
Heat Treatment
Although the addition of Al2O3 is a convenient approach for improving the grain-boundary
conduction, it is also known to decrease the grain-interior conductivity. [16,21,22,26] In
Table 1, it can be seen that the addition of Al2O3 always deteriorates the grain-interior
conduction. This is explained in terms of the defect association between AlZr and VO [23]
after the incorporation of Al2O3 into ZrO2 lattices. Therefore, it tends to become significant at
higher sintering temperatures (typically > 1600oC). [21,22,26]
In order to prevent the adverse effect of Al2O3, Lee et al. suggested a new scavenging
method without the addition of Al2O3. [22] The schematics for the scavenging mechanisms
by Al2O3 additive and pre-sintering HT are compared in Fig.7. In contrast to additive
scavenging, which uses the interactive reaction between Al2O3 and siliceous intergranular
phase (Fig.7(a)), the new process employs a two-stage sintering process. (Fig.7(c)) This
process was named as precursor scavenging because it can be achieved by the precursor
itself without any additives.
Figure 8 shows the variation in the impedance spectra by additive scavenging and
precursor scavenging in 8YSZ. [22] For additive scavenging, 1 mol% of Al2O3 with various
sizes (A03: 0.3 m, Asol: Al2O3 sol, A10: 10 m) were added. For precursor scavenging, the
pure 8YSZ specimens were heat-treated at 1200oC for 40h prior to sintering at 1500oC for 4h.
(denoted as 8Y-1200(40)-1500 specimen) In the figure, BM and UD denote ball milling
and ultrasonic dispersion, respectively. The semicircle in the intermediate frequency regime
corresponds to the grain-boundary contribution. When various forms of Al2O3 were added,
the grain-boundary resistivity decreased to a large extent. (Fig.8(a),(b),(d), and (e)) This can
be explained by the scavenging reaction of Al2O3. Approximately the same degree of
improvement in the grain-boundary conduction was achieved as a result of pre-sintering heat
treatment (HT) at 1200oC. (Fig.8(c) and (f)) Note that gi increased in the additive (Al2O3)-
scavenged specimen (Fig.8(e)) while it remained constant in the precursor-scavenged
specimen (Fig.8(f)) when sintered at 1600oC. This clearly demonstrates that precursor
scavenging can improve the grain-boundary conduction significantly without affecting the
grain-interior one even at high sintering temperatures.
1600 (a) Al2O3

1200 g.b.
Additive
800 scavenging
400
Temperature ( C)
0
o
(b)
1200 Siliceous
g.b. phase
800
400
0 (c) Discrete
Si-containing
phase
1200
800 Precursor
1200(40)-1500 scavenging
400
HT Temp. HT time g.b.
0
0 20 40 60
Time (h)
Fig.7 The temperature profiles for sintering and a schematic diagram of scavenging mechanism. (a)
Scavenging by the Al2O3 additive (additive scavenging), (b) before scavenging and (c) precursor
scavenging by pre-sintering heat treatment(HT).
(a) (d)

1
8Y-BM-1600
8Y-BM-1500
8Y-UD-1500 8Y-UD-1600
0
-|Z|sin [k.cm]
(b) (e)

1 8Y1A03-BM-1500 8Y1A03-BM-1600

8Y1Asol-BM-1500 8Y1Asol-UD-1600

8Y1A10-UD-1500 8Y1A10-UD-1600 gi
0
(c) (f)
: 543 Hz
1 : 5.12 kHz
: 48.4 kHz
8Y-1200(40)-1500
8Y-1200(40)-1600
: 586 kHz
0
4 5 6 7 8 4 5 6 7 8
|Z|cos [k.cm]
Fig.8 Complex impedance spectra at 400oC of the various 8YSZ specimens sintered at (a), (b), (c) 1500
and (d), (e), (f) 1600oC. (8Y: 8YSZ, BM: Ball Milling, UD-Ultrasonic Dispersion, 1A03: the addition of
1 mol% of Al2O3 sized 0.3 m, 1A10: the addition of 1 mol% of Al2O3 sized 10 m, 1Asol: the addition
of 1 mol% of Al2O3 sol, 1200(40)-1500: heat treatment at 1200oC for 40 h prior to sintering at 1500oC
for 4h) The semicircles indicated by arrows mean the grain-boundary contributions, according to [22].
12 Jong-Heun Lee
2.0 2.0
gbapp (a)
Rgbs
1.5 1.5
[k.cm]
Rgbs[.cm ]
2
1.0 1.0
app
gb 0.5 0.5
0.0 0.0
1050 1100 1150 1200 1250 1300 1350
HT temp. before sintering at 1500oC (oC)
2.0 2.0
(b)
gbapp
1.5 Rgbs 1.5
[k.cm]
Rgbs[.cm2]
1.0 1.0
app
gb
app
and Rgbs regime of
gb
scavenging by Al2O3
0.5 0.5
0.0 0.0
0 10 20 30 40
HT time before sintering (h)
Fig.9 The variation of apparent grain-boundary resistivity (gbapp) and resistance per unit grain-
boundary area (Rgbs) at 400oC with varying (a) HT temperature (HT time=10h) and (b) HT time (HT
temperature=1200oC) prior to sintering at 1500oC for 4h. Pure 8YSZ specimens with the 100 ppm of
SiO2 impurity were used, according to [22].
Figure 9 shows the effect of the heat-treatment temperature (HTT) and holding time of
HT on the grain-boundary conduction. The Rgbs and gbapp values showed the minimum at
HTT=1200oC (Fig.9(a)). At the optimal HTT (1200oC), holding time > 20h was necessary to
achieve the saturated scavenging efficiency. Because the reaction between the precursor and
impurity occurs at the impurity level, the elucidation of the precursor-scavenging mechanism
will require a further detailed study. However, one possible explanation is the gathering of a
siliceous intergranular phase into a discrete Si-containing phase such as ZrSiO4 by additional
HT.
The formation of ZrSiO4 is known to be quite difficult [31,32] due to its small formation
energy [33] and large activation energy for the combination reaction between ZrO2 and SiO2.
[34] For example, only a low concentration of ZrSiO4 is reportedly formed even by the
calcination of the precursor at 1500-1600oC when the heating rate was 300oC/h. [34] ZrSiO4
is known to form at > ~1200oC when prepared by the sol-gel method [35] and grow via an
epitaxial mechanism. [36] Indeed, the seeding of 0.5 2 wt% ZrSiO4 is known to accelerate
ZrSiO4 formation. [35-37] Therefore, the long duration at 1200oC results in sufficient ZrSiO4
nucleation and growth during the subsequent heating at higher temperatures. If the epitaxial
growth mechanism is considered, the increase in the Rgbs value at HTT 1250oC can be
attributed to the low nucleus (ZrSiO4) density due to the fast passing through the region of
nucleation. The conductivity improvement via HT at T=1200oC despite the subsequent
sintering at T=1500-1600oC indicates that the Si-gathered phase formed during HT is stable
up to the sintering temperature. Pure ZrSiO4 is known to dissociate at >1700oC [34] although
this depends on the sample purity. [38,39] Therefore, the thermal stability of ZrSiO4 up to the
sintering temperature (1500-1600oC) coincides well with the above explanation.
Table 2. The 8YSZ specimens with the various siliceous impurity concentrations and their
scavenging efficiency as a result of the additional HT at 1200oC for 40h prior to sintering at
1500oC for 4h, according to [42].
Impurity conc.
Heat treatment Rgbs ( BS ) Rgbs ( AS ) Rgbs ( AS )
Spec. (ppm by weight)
2
and sintering (cm ) 2
(cm ) Rgbs ( BS )
SiO2 Al2O3 Fe2O3
8Y 1500 1.20
As received 0.167
8Y-PS 1200(40)-1500 0.20
8Y100 1500 1.45
100 29 28 0.131
8Y100-PS 1200(40)-1500 0.19
8Y160 1500 3.76
160 28 25 0.231
8Y160-PS 1200(40)-1500 0.87
8Y310 1500 1.58
310 29 24 0.943
8Y310-PS 1200(40)-1500 1.49
8Y1000 1500 1.43
1000 30 26 0.916
8Y1000-PS 1200(40)-1500 1.31
sintered at 1500oC for 4h.
heat-treated at 1200oC for 40h prior to sintering at 1500oC for 4h
The 8 mol% Ytterbia-Stabilzied Zirconia (8YbSZ) specimens with an 80 and 170 ppm-
level background impurity showed a similar precursor-scavenging behavior except for the
slight difference in the optimum HTT (1250oC). [40,41] This indicates that precursor
scavenging can be extended to other stabilized zirconia systems. Other supporting evidence
for the scavenging siliceous phase as a result of pre-sintering HT is the decrease in grain size.
The grain size of 8YbSZ specimen containing 170 ppm of SiO2 was 8.1 m, which is larger
than the 5.2 m for the 8YbSZ specimen containing 80 ppm of SiO2. This means that the
trace siliceous impurity enhances grain growth. As a result of the pre-sintering HT at 1250oC,
the grain-boundary conduction not only improved but the grain size decreased from 8.1 to 5.9
m. This decrease in grain size can be attributed to the scavenging of siliceous impurity.
Indeed, all the YSZ and YbSZ specimens showed a decrease in grain size by the pre-sintering
14 Jong-Heun Lee
HT and a higher scavenging efficiency always accompanied the larger decrease in grain size.
[22, 40-42]
However, precursor scavenging was not always valid for all the stabilized zirconia
specimens. Table 2 shows the 8YSZ specimens containing 0-1000 ppm of siliceous impurities
and their scavenging efficiencies by pre-sintering HT at 1200oC. [42] As shown in the table,
the specimens with a low SiO2 impurity (< 200 ppm) showed an approximately 5-fold
increase in grain-boundary conduction, while those with the high SiO2 impurity (> 300 ppm)
did not show an improvement. This was explained by the inhomogeneous distribution of the
siliceous phase for the heavily-SiO2-doped specimens and its influence on the scavenging
reaction. The slow kinetics for forming ZrSiO4 might be one reason for the difficult precursor
scavenging of a large siliceous impurity concentration.
4.2 The Improvement in the Grain-Boundary Conduction by Post-sintering

Heat Treatment
The precursor scavenging with pre-sintering HT is a convenient and effective approach to

improving the grain-boundary conduction of stabilized zirconia with the segregation-level
siliceous impurity. However, an intergranular liquid is usually formed when the siliceous
impurity is abundant. [43,44] In this condition, the scavenging by forming a discrete Si-
containing phase cannot be applied and a new concept is necessary. This study attempted to
enhance the grain-boundary conduction via the configurational change in the intergranular
liquid.
1600
1200
(a) (b)
40
800
400 105 Hz
103 Hz 101 Hz
0 0
(c) (d)
"[k.cm]
1200
Temp.( C)
40
o
800
400 105 Hz
103 Hz 101 Hz
0 0
1200
(e) (f)
40
800
103 Hz
400 105 Hz
101 Hz
0 0
0 10 20 30 0 40 80 120 160
Time(h) '[k.cm]
Fig.10 The heating schedules of 15CSZ specimens and their corresponding impedance spectra at 400oC
in air: (a),(b) normal sintering at 1550oC for 4h, (c),(d) pre-sintering HT at 1300oC for 10h + sintering
1550oC for 4h. (e),(f) sintering 1550oC for 4h + post-sintering HT at 1300oC for 10h, according to [45].
Figure 10 shows the complex impedance of the 15CSZ specimens with the variation in
the HT schedules. [45] In the specimen sintered at 1550oC, the size of the grain-boundary
semicircle was almost identical to that of the grain-interior semicircle. (Fig.10 (a) and (b)).
Although the pre-sintering HT at 1300oC for 10h decreased the grain-boundary resistivity, the
change was not significant. (Fig.10 (c) and (d)) However, post-sintering HT at 1300oC for
10h greatly decreased the size of the grain-boundary semicircle. (Fig.10 (e) and (f)) From the
systematic investigation on the HTT and HT time (Fig.11), it was found that the optimum
HTT is 1300oC and the saturated scavenging efficiency can be attained by HTT at 1300oC for
10h. At the maximum scavenging efficiency, the grain-boundary conductivity improved 6.5
times by post-sintering HT.
(a) (c)
1600 1600
1200
Temp.(oC)
1200
Temp.( C)
o
800 800
400 400
0 0
0 10 20 30 0 10 20 30 40 50
Time(h) Time(h)
100 100
(b) (d)
gb (no HT)=78.1 k.cm gb (no HT)=78.1 k.cm
app app
80 80
[k.cm]
[k.cm]
60 60
app
app
40 40
gb
gb
20 20
0 0
1000 1100 1200 1300 1400 0 10 20 30 40
HT Temperature(oC) HT time at 1300oC(h)
Fig.11 The apparent grain-boundary resistivity (gbapp) at 400oC for 15CSZ specimens (a), (b) with the
variation in the HT temperatures (HT time=10h) and (c),(d) with the variation in the HT time (HT
temperature=1300oC), according to [45].
Table 3 shows the changes in the intergranular phase as a result of the post-sintering HT.
If the additional grain growth had occurred during post-sintering HT, it can decrease the
gbapp value by decreasing the grain-boundary density. However, the average grain size did
not change as a result of post-sintering HT. This confirms that the significant improvement in
grain-boundary conduction emanated not from the changes in the grain-boundary density but
from the change in the physico-chemical properties of the grain boundary. From TEM
analysis, it was found the dihedral angles in the triple junction and between the two grains
increased significantly as a result of the post-sintering HT. In addition, intergranular phases in
the specimen with post-sintering HT were crystalline, while only the amorphous intergranular
liquid was observed in the specimen without the HT. (see electron diffraction patterns in
Table 3)
16 Jong-Heun Lee
The dewetting phenomenon can be explained in relation to the crystallization of an

intergranular liquid. Fig.12 illustrates a schematic diagram. As shown in the equation in
Fig.12, the dihedral angle of the intergranular liquid was determined by the ratio between the
grain-boundary energy (SS) and solid(matrix)-liquid interface energy (SL). When the liquid
is crystallized, the SL term changed into solid(matrix)-solid(crystallized intergranular phase)
interface energy (SS). Because the SS value is usually larger than the SL value [46], the
dihedral angle can increase via crystallization. When a composition of glass melt is
crystallized during cooling, the degree of crystallization is generally explained by Time-
Temperature-Transformation curve (T-T-T curve). Usually, the optimum crystallization
temperature, which is known as the knee temperature, is present because the crystallite
nucleation is low near the melting temperature and crystallization is slow at the low
temperature. [47] For crystallization, the cooling schedule should be designed to pass the
crystallization regime as much as possible. Therefore, the optimum HTT (1300oC) is believed
to be the knee temperature in the T-T-T diagram and the gbapp increase with increasing HT
temperature from 1300 to 1350oC can be explained in the same way.
Liquid melt
Crystalline
Temp.
Amorphous SL < SS Crystalline

SL < SS
SL SS '
SS = 2 SL cos SS = 2 SS ' cos
2 2
Time
Fig.12 Schematic diagram showing the crystallization and dewetting of the intergranular phase during
post-sintering HT.
Finally, in order to compare the scavenging efficiency as a result of the addition of Al2O3,
the 15CSZ specimens were sintered at 1450-1600oC and then heat-treated at 1300oC for 10h.
Figure 13 shows the scavenging factor. When sintered at 1450oC, a ~2-fold increase in the
grain-boundary conduction occurred, which is quite small compared with the ~15-fold
increase by the addition of 1 mol% Al2O3. (see Fig.6(a)). However, at the high sintering
temperature (1500-1600oC), a 5-fold increase in grain-boundary conduction was achieved,
while the addition of Al2O3 deteriorates the grain-boundary conduction as a result of Al2O3
dissolution into the siliceous intergranular liquid. This suggests that the present post-sintering
HT is a useful approach for improving the grain-boundary conduction of 15CSZ particularly
when the sintering temperature is high.
Table 3. The comparison of the intergranular liquid phase between the 15CSZ specimens with
and without post-sintering HT at 1300oC for 10h, according to [45].
Sintering at 1550oC for 4h + Post-

Sintering at 1550oC for 4h
sintering HT at 1300oC for 10h
Intergranular phase at
the triple junction
0.5 m
0.5 m
Electron diffraction
pattern of
intergranular liquid
Average grain size 15.7 m 15.6 m

(triple junction) 49o 36o 89o 24o
(# of observation =27) (# of observation =52)
(between 8 o 8o 89o 27o
two grains) (# of observation =28) (# of observation =30)
10
Post-sintering HT >
Rgbs(AS)/Rgbs(BS)
Al2O3 scavenging
1
0.1
1450 1500 1550 1600

o
Sintering temperature( C)
Fig.13 The scavenging factor (SF) of 15CSZ specimens by post-sintering HT at 1300oC for 10h. The
gbapp value was measured at 400oC in air, according to [45].
18 Jong-Heun Lee
5 Estimation of the Intergranular-Liquid-Phase Distribution

Using Local Impedance Technique
Impedance spectroscopy has been a useful tool for separating the grain-interior and grain-
boundary contributions in many electroceramics. [48,49] In most analyses, the brick layer
model, which assumes electrically and dimensionally homogeneous grain and grain boundary,
was used. [7,13,24-26] (Fig.3(b)) However, the distribution of the gi and gb values can
become spatially uneven for the specimens under an electric field [50-53] or with a
functionally graded composition. [54]
Therefore, Rodewald, Fleig and Maier proposed the local impedance technique using an
array of the micro-scale electrodes. [50,51,55-57] When the specimen thickness was
sufficiently large compared with the size of the microelectrode, the impedance between the
microelectrode and large counter electrode mainly represents that of the near-microelectrode
hemisphere where most of the potential drops [55,58]. This technique is useful because it not
only spatially resolves the electrical properties within a large grain but also can be used to
directly investigate the individual grain boundary.
(a)
Liquid
Phase
Sintering
(b)
Impedance
Measurement
-Z -Z -Z
Z Z Z
Fig.14 The schematic diagram of (a) dynamic liquid rearrangement during liquid phase sintering and
(b) estimation of the intergranular-liquid distribution by millicontact impedance spectroscopy.
However, the conduction for a single grain boundary can become heterogeneous due to
differences in the wetting of a resistive liquid [13], the liquid distribution [59], the grain-
boundary type (low angle or high angle) [60], boundary microstructure (straight or undulated)
etc. Moreover, during liquid phase sintering, the intergranular liquid phase can flow
dynamically to reduce the total liquid-vapor interface energy. Generally, the liquid phase is
known to agglomerate at the center of the specimen in the initial stage of sintering and the
agglomerated liquid flows outward as a result of further sintering. (Fig.14(a)) [61-63] Indeed,
the uneven distribution of the resistive intergranular liquid has been reported in stabilized
zirconia. [59,64,65] In this case, the average impedance for a small volume rather than the
impedance for a single grain boundary would be more advantageous in examining the spatial
variation in grain-boundary conduction. (Fig.14(b))
5.1 Local Impedance Measurement by Successive Thinning of a Specimen
In order to estimate the spatial distribution of the intergranular liquid in 1-mol%-Al2O3(size =

0.3m)-doped 15CSZ specimen, a series of complex impedance spectra were attained by the
successive thinning of a specimen and subsequent measurement. (Fig.15) [66] The gbapp
decreased to approximately 60% of its original value by removing 1/20 of the original
specimen thickness (t) from both outer regions. This strongly suggests that the outer part of
specimen has a more resistive grain boundary than the grain interior. The gbapp value
decreased further as the specimen was further thinned and then saturated at a specimen
thickness < 0.6t.
10
-|Z|sin [k.cm]
0.7t 0.8t 0.9t 1.0t

5
543.5Hz
48.4kHz 57.6Hz
5.129kHz
0
25 30 35 40 45
|Z|cos [k.cm]
Fig.15 The complex impedance spectra of 1-mol%-Al2O3(size=0.3m)-doped 15CSZ specimens

sintered at 1600oC for 4h. 1/20 of original sample thickness (t) was removed successively from both
outer regions and the corresponding impedance was measured at 450oC in air, according to [66].
The results attained by successive thinning can be analyzed in terms of the successive
attachment of the outer layers from the specimen core. Because the addition of every outer
layer results in a data variation in a cumulative manner, even a small increase in the
cumulative gbapp value means a large increase in the real gbapp value. Indeed, from the
analysis, it was calculated that the grain boundary at the outermost portion is ~15 times more
resistive than that at the inner core. From the uniform grain-size distribution throughout the
specimen, this heterogeneous gbapp profile was attributed to the outward rearrangement of the
intergranular liquid phase.
20 Jong-Heun Lee
5.2 Millicontact Impedance Spectroscopy: The Fundamental Concept
Although the measurement of the local impedance by the successive thinning of a specimen is
a convenient tool for analyzing the spatially-uneven grain-boundary characteristics, there are
some restrictions. Because the data is in a cumulative form, unless the variation in the gbapp
value is very large and monotonous, it is not easy to observe the presence of spatial variation.
Therefore, instead of an indirect thinning method, a straightforward measurement of the local
impedance is desirable when one wishes to characterize the spatial variation of the electrical
properties with precision.
Therefore, a local impedance technique using 1-dimensional or 2-dimensional array of
the sub-millimeter-scale electrodes was suggested to estimate spatially the heterogeneous
distribution of the resistive intergranular liquid. [67] This was referred to as millicontact
impedance spectroscopy. As shown in Figure 14(b), the core concept is the determination of
the spatial distribution in the grain-boundary resistance by measuring the local impedance for
a small volume.
(a) 1
B A
1
(b)
B A
Fig.16 Equi-potential contour maps of (a) single crystal and (b) polycrystalline materials with highly
resistive grain boundary (gb =100gi, gb=0.02dg). The changes of line and color indicate equi-
sp
potential lines. Potential drop between two neighboring lines is 1/20 of the applied voltage, according
to [67].
In order to check the validity of the present technique, the equi-potential contours of
single crystal and polycrystalline materials with a highly resistive grain-boundary phase (gbsp
=100gi, gb=0.02dg) were calculated using a simulation and the results are given in Fig.16.
In a single crystal specimen, the local impedance measured between electrode 1 and 1 are the
approximate contribution from the volume element (A) defined by the electrode area and the
specimen thickness. In particular, a circumventing conduction path between the electrode
edges (see current line B in Fig.16(a)) should be considered and it becomes important when
the specimen is thick and the electrode area is small. However, in a thin specimen with a
sufficient electrode area, the overall gi profile can be attained unless the spatial variation of
gi value is not large near the electrode edge.
When highly insulating grain boundaries are present, the equi-potential line near the edge
becomes more parallel to the electrode (Fig.16(b)), which significantly narrows the
circumventing conduction. The narrowing effect will be further enhanced as the grain size
decreases or as the gbsp value increases. Therefore, millicontact impedance spectroscopy can
be used as an effective and valid tool to estimate the spatial distribution of electrical
properties in stabilized zirconia with a micro-meter-sized grain and a highly resistive grain
boundary.
5.3 Millicontact Impedance Spectroscopy: The Distribution of

Intergranular-Liquid Phase
The presence of a small amount of a liquid phase is known to enhance the sintering rate by
providing a rapid material-transport path. According to the literature [61-63], the liquid can
rearrange dynamically during liquid-enhanced sintering, which results in a spatially uneven
liquid distribution. When the liquid is abundant, such a liquid rearrangement process can be
observed using a conventional scanning electron microscope (SEM). However, at the small
liquid content, SEM is not sufficient for observations because the liquid usually presents as an
intergranular film with a few-nanometer thickness. In this case, the complex impedance,
which is very sensitive to the configuration of intergranular phase, can be used as a measure
for the liquid distribution.
Figure 17(a) shows the electrode array for measuring the local impedance. The specimen
was cut along the diameter of a cylindrical specimen with a thickness of 0.5 mm. A 2-
dimensional array of cube-shape electrode (0.5mm X 0.5mm) was made in both surfaces by
screen printing a Pt paste and subsequent heat treatment. The impedance was measured at
400oC between two opposite electrodes. Fig.17 (b) shows the complex impedance spectra as a
function of the distance from the surface along the axial direction (X=5.25) for the 1-mol%-
Al2O3-doped 15CSZ specimen sintered at 1650oC. [67] The gbapp value attained from the
intermediate-frequency (~102-103 Hz) semicircle changed to a large extent from 56 to 3.6
kcm, while the gi value attained from the high-frequency semicircle remained constant as
82 kcm. Impedances for all the electrodes were measured, and the Rgbs values were
calculated from the gbapp value and the average grain size. (Fig.17(c)) Note that the profile of
Rgbs was the same as that of gbapp because the grain size was uniform throughout the
specimen. The Rgbs value was high at the surface, and decreased abruptly at the second
electrode showing a saturated value (25-15 cm2) at the core region (x > 2mm).
Consequently, the grain boundary at the surface was ~10-15 times more resistive than that
located at the inner part. The profile of Rgbs clearly demonstrates that the spatial distribution
of the grain-boundary conductivity can be estimated by the present technique.
In order to investigate the reason for the large difference in gbapp, the Al2O3
concentration and sintering temperature were varied. For this measurement, 1-dimensional
array of electrodes (size: 2 mm X 0.5 mm) in Fig.18(a) was employed. Figure 18(b), (c), (d),
(e) show the results for the 15CSZ specimens sintered at 1600oC for 4h with the variation in
the Al2O3 concentrations. [68] In pure 15CSZ specimen, the gbapp value increased with
increasing distance from the surface (x) and became saturated at x 2mm. However, the
22 Jong-Heun Lee
tendency for increasing gbapp disappeared at 0.5-mol%-Al2O3 (Fig.18(c)) and completely

reversed at 1.0- and 1.5-mol%-Al2O3. (Fig.18(d), (e))
Surface
(b)
105 104
103
102
10 Hz
d from surface(mm)
50
-|Z|sin [k
0.25
40 1.25
2.25
30 3.25
(a) 20 4.25
. cm]
10 5.25
Pt electrode array Center 6.25
Area: 0.5 x 0.5 mm
2 0 40 80 120 160
Spacing: 0.5 mm |Z|cos [k.cm]
Y
Center 13 mm
X (c)
t=0.5 mm
=11 mm
300
Rgbs [.cm 2]
200
0
Y po
1 100
2
s i t io
3
4
n (m
5 0
6 ce nter 4 2
m)
8 6
10
X position (mm)
Fig.17. (a) Schematic diagram of a 2-dimensional array of electrodes for measuring the local impedance.
(b) The impedance spectra of 1-mol%-Al2O3(size 0.3 m)-doped 15CSZ specimen measured at 400oC
in air as a function of distance from the surface along the cylinder axis. (The specimen was sintered at
1650oC for 4h) (c) The distribution of the resistance per unit grain-boundary area (Rgbs) as a function of
specimen location, according to [67].
Figure 19 shows the gbapp profiles of 1-mol%-Al2O3-doped 15 CSZ specimens as a

function of the sintering temperature. [59,69] When sintered at 1650oC for 4h, the grain
boundary at the surface was ~15 times more resistive than that in the center region. According
to TEM analysis, there was abundant intergranular liquid at the surface, which penetrated
almost all the grain boundary, while that in the center region showed a discontinuous
configuration with a frequent lenticular pocket. The average dihedral angles of the liquid
pockets were ~20 at the surface and ~100 at the center. In addition, the liquid composition
was different from each other, which indicates phase separation of the intergranular liquid.
surface center
(a) 100 =5.33,d =15.

g
Pt electrode Cylinder-shape 80
0.5 x 2 mm2 specimen
Spacing: 0.5mm 60
40 (b)
13 mm
20 =5.49,dg=31.
(c)
[k.cm]
x(mm)
11 mm 0
x=0.25 x=5.25
0.5 mm
40 =5.45,dg=35.
20 (d)
app
0
gb
Impedance
Measurement =5.45,dg=37.
40
(e)
20
0
0 1 2 3 4 5 6
x from surface (mm)
Fig.18 (a) Schematic diagram of a 1-dimensional array of electrodes for measuring the local impedance
and the profile of the apparent grain-boundary resistivity (gbapp) of 15CSZ the specimens at 400oC with
the variation of Al2O3 (size=0.3 m) amounts: (b) pure 15CSZ, (c) 0.5-mol%-Al2O3-doped 15CSZ, (d)
1.0-mol%-Al2O3-doped 15CSZ and (e) 1.5-mol%-Al2O3-doped 15CSZ. All the specimens were sintered
at 1600oC for 4h, according to [68].
However, when 1-mol%-Al2O3-doped 15 CSZ specimen was sintered at 1450oC for 0 h,

the profile of gbapp became opposite. The gbapp value was higher in the center region.
(Fig.19(e)) This is analogous to the pure 15CSZ specimen sintered at 1600oC (Fig.18(a)).
Note that two specimens (Fig.19(e) and Fig.18(b)) show a significantly low density(5.28-
5.33) compared with the other specimens. (5.41-5.45) During liquid phase sintering, it is
generally acknowledged that the liquid agglomerates at the center region in the initial stage of
sintering to decrease the total solid-liquid interface energy. [61] Indeed, TEM analysis
showed that the liquid pocket size in the central region was ~ 3 times larger than that near the
surface in the pure 15CSZ specimen (Fig.18(b)). Therefore, the large gbapp values in
Fig.19(e) and Fig.18(b) can be attributed to liquid coagulation at the initial stage of sintering.
From the different composition of wetting and non-wetting intergranular liquids in the
specimen sintered at 1650oC (Fig.19(a)), the phase separation can be deduced. As the
sintering temperature was increased from 1450 to 1650oC, grain growth and the densification
occurred simultaneously. [Fig.19 (d), (c), (b), (a)] During this process, among the two liquids,
a liquid with a composition that improves the wetting ability will flow preferentially to the
surface along the capillary. The decrease in the pore size as a result of densification will be
advantageous for a capillary rise. [70] Besides, grain growth is known to be necessary for
liquid to flow into the pores. However, densification was always accompanied by grain
growth as a result of the addition of Al2O3 (Fig.18) and the increase in the sintering
temperature (Fig.19), which make the separation of the variables difficult. Therefore, the low-
density specimen (5.27) with almost the same grain size (29.9 m) was fabricated by
24 Jong-Heun Lee
sintering a low-density green body. (Fig.20(c)) A relatively flat profile was attained, which
indicates that not only grain growth but also densification is a prerequisite for the capillary
action.
Wetting 60 =5.44,dg=45.2m
Liquid (a)
40
20
Non-wetting
gbapp[k.cm]
0
Liquid
40 =5.45,dg=35.1m
(b)
20
Outward
0 Capillary rise
40 =5.47,d =17.9m
g
(c)
20
0
=5.41,dg=9.7m
20
(d) Liquid Phase
0 Coagulation Separation
40
(e)
20
=5.28
0
0 1 2 3 4 5 6
x from surface (mm)
Fig.19 The profile of the apparent grain-boundary resistivity (gbapp) of 1.0-mol%-Al2O3(size=0.3 m)-
doped 15CSZ the specimens with the variation of the sintering time and temperatures: (a) sintered at
1450oC for 0h, (b) sintered at 1450oC for 4h, (c) sintered at 1500oC for 4h, (d) sintered at 1600oC for 4h
and (e) sintered at 1650oC for 4h. The impedance was measured at 400oC in air, according to [59,69].
Figure 20(b) shows the gbapp profiles when 1-mol%-Al2O3-doped 15CSZ specimen was
cooled rapidly (800oC/min) after sintering 1600oC for 4h. Compared to the specimen cooled
at a rate of 200oC/min (Fig.20(a)), the profile moved upward. This means a significant
increase in the average gb value for the entire specimen. The configuration of the
intergranular liquid can be changed dynamically during cooling because the
wetting/dewetting characteristics depend on temperature. The increase in the overall profile
by fast cooling qualitatively means that the high-temperature configuration of the
intergranular liquid is more resistive. In addition, the grain-boundary resistance is known to
depend on the thermal history, [71] sintering atmosphere, [72], and superplastic deformation
[73]. These phenomena can be better understood when the local electrical properties are
characterized using a spatially resolved impedance technique.
40 =5.45,dg=35.1m (a)
20
0
=5.46,dg=35.0m (b)
60
gbapp[k.cm]
40
20
0
40 =5.27,dg=29.9m
(c)
20
0
0 1 2 3 4 5 6
x from surface (mm)
Fig.20 The profile of the apparent grain-boundary resistivity (gbapp) of 1.0-mol%-Al2O3-doped 15CSZ
the specimens with the variation of cooling rate and specimen density: (a) cooling rate: 200oC/h,
density: 5.45 g/cm3, dg: 35.1 m, (b) cooling rate: 800oC/h, density: 5.46 g/cm3, dg: 35.0 m, and (c)
cooling rate: 200oC/h, density: 5.27 g/cm3, dg: 29.9 m. All the specimens were sintered at 1600oC for
4h, according to [67].
The redistribution of the liquid phase could also be observed in the 8YSZ specimen
containing 0.5 mol% of SiO2 and 0.5 mol% Al2O3. Figure 21 shows the gbapp profiles for the
specimens sintered at different temperatures. [74] When sintered at 1450oC for O h, liquid
coagulation at the center region could be noted. (Fig.21(c)) As the sintering temperature
increased, the liquid phase was spread outward to make a uniform distribution. (Fig. 21(b)
and (a)) The reason why the concentration of the wetting intergranular liquid was not found
near the surface region can be explained by there being no phase separation of a liquid phase
in 8YSZ system. Therefore, the liquid coagulation at the center and its subsequent
redistribution outward could be observed as a typical liquid phase sintering behavior. Figure
21(d) and (e) show the SEM microstructures of the center and surface regions for the
specimen sintered 1450oC O h. despite the significant difference in the gbapp value, the
microstructure as well as the grain size was the same. This demonstrates that the
rearrangement of a small amount of intergranular liquid, which cannot be observed by SEM,
can be observed using the local impedance technique.
26 Jong-Heun Lee
30
20
10
gbapp[k.cm]
40
30
(d) (e)
20
50
40
5 m 30 5 m
0 1 2 3 4 5
x from surface (mm)
Fig.21 The profile of the apparent grain-boundary resistivity (gbapp) of 8YSZ the specimens containing
0.5 mol% Al2O3 and 0.5 mol% SiO2 with the variation of sintering temperature: The gbapp profiles for
the specimens (a) sintered at 1550oC for 4h, (b) sintered at 1550oC for 0h, and (c) sintered at 1450oC for
0h. The impedance was measured at 400oC in air. The microstructures of a specimen sintered at 1450oC
for 0h (d) at the central region and (e) at surface region, according to [74].
5.4 Millicontact Impedance Spectroscopy: The Investigation of Liquid

Penetration
Millicontact impedance spectroscopy can be also applied to the characterization of the

functionally graded materials or to the determination of material transport near the interface.
For example, the interface reaction between the LaMnO3-based cathode and the silicate
impurity diffused from YSZ can form a deleterious lanthanum silicate phase at the interface,
which affects the long-term stability of the SOFC.[75] In order to understand this
phenomenon, the migration of a siliceous impurity across the YSZ specimen should be
characterized. Figure 22(a) shows the experimental setup for providing an artificial impurity
on the 8YSZ specimen by dropping a dilute solution.[76] The impedance spectra are given in
Fig. 22(b). The gi values in the first and second electrodes (x=0.25, 1.25 mm) from the
surface were ~7-8 kcm, which was slightly larger than the value of the lower part (~6
kcm). (x=2.25-7.25 mm) The gbapp showed more dramatic changes. The uppermost gbapp
value (at x=0.25) was ~ 32 times larger compared with those at x=1.25-7.27 mm. This
indicates that the penetration depth of the impurity is ~ 0.5-1.5 mm. The main advantage of
millicontact impedance spectroscopy is that both the spatial variations of the grain-interior
and grain-boundary conductivities can be examined separately.
(a) CaSi2O5 drop
8YSZ Sintered at
green body 1550oC for 4h
x
3
10 Hz
(b)
)
10
mm
-|Z|sin
8 105 Hz TOP
ace(
0.25
6 2.25
[k
surf
4 4.25
m
2 6.25
x fro
. cm]
0 BASE
10 20 30
|Z|cos [k.cm]
Fig.22 (a) Schematics of the experimental set up for liquid penetration and the electrode configuration.
(b) The Complex impedance spectra of the 8YSZ specimen sintered at 1550oC for 4h after dropping the
ethanol solution containing TEOS and Ca(NO3)4H2O. The total amount of doping was 0.0025g in
terms of CaSi2O5 and the impedance was measured at 400oC in air as a function of the distance from the
surface (x), according to [76].
6 Conclusions
New methods for improving the grain-boundary conduction of stabilized zirconia were
suggested. In contrast to the scavenging of the resistive intergranular phase by the addition of
Al2O3, the configurational change in the intergranular phase via a two-stage sintering process
was used to enhance the grain-boundary conduction, which was called precursor scavenging.
For the 8YSZ and 8YbSZ specimens with a segregation-level siliceous impurity (70-170
ppm), pre-sintering heat treatment (HT) at 1200-1250oC for > 20h resulted in approximately
the same scavenging efficiency as that by Al2O3 addition. The formation of a discrete Si-
containing phase such as ZrSiO4 with the assistance of additional HT was suggested to be the
scavenging mechanism. In the 15CSZ specimen with the intergranular liquid, the post-
sintering HT at 1300oC for > 10h showed ~6.5-fold increase in the grain-boundary
conductivity, which could be explained by the dewetting of the intergranular phase due to
crystallization. In the second part, a new local impedance technique, called Millicontact
Impedance Spectroscopy, was used to estimate the spatially uneven grain-boundary character.
The dynamic rearrangement of a small amount of liquid phase, which could not be observed
28 Jong-Heun Lee
by SEM, was successfully determined using the present local impedance technique combined
with a sub-millimeter-scale electrode array. In conclusion, millicontact impedance
spectroscopy is a useful tool for characterizing functionally graded materials and the spatial
distribution of electrical conductivity in many electroceramics.
References
1. Estell, T. H.; Flengas, S.N., Chem. Rev. 1970, 70, 339-376.
2. Subbarao, E. C.; Maiti, H. S., Solid State Ionics 1984, 11, 317-338.
3. Steele, B.C. H., J. Mater. Sci. 2001, 36, 1053-1068.
4. Kilner, J. A., Solid State Ionics, 1983, 8, 201-207.
5. Kilner, J. A.; Steele B. C. H., In Nostoichiometric Oxides; Srensen, O. T., Academic
Press: New York, NY, 1981, pp.233-269.
6. VerkerkM. J.; Winnubst, A. J. A.; Burggraaf, A. J, J. Mater. Sci. 1982, 17, 3113-3122.
7. Verkerk, M. J.; Middelhuis, B. J.; Burggraaf, A. J., Solid State Ionics 1982, 6, 159-170.
8. Aoki, M.; Chiang, Y. -M.; Kosacki, I.; Lee, J. -R.; Tuller, H.; Liu, Y. J. Am. Ceram. Soc.
1996, 79, 1169-1180.
9. Guo, X., Solid State Ionics 1995, 81, 235-242.
10. Guo, X.; Sigle, W; Fleig, J.; Maier, J., Solid State Ionics 2002, 154-155, 555-561.
11. Guo, X.; Computational Materials Science 2001, 20, 168-176.
12. Lee, J. S.; Kim, D. Y. J. Mater. Res. 2001, 16, 2739-2750.
13. Gdickemier, M.; Michel, B.; Orliukas, A.; Bohac, P.; Sasaki, K.; Gauckler, L.; Henrich,
H.; Schwander, P.; Kostorz, G.; Hofmann, H.; Frei, O. J. Mater. Res. 1994, 9, 1228-1240.
14. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Komatsu, M.; Haneda, H. J. Am. Ceram.
Soc. 2000, 83, 1273-1275.
15. Rajendran, S.; Drennan, J.; Badwal, S. P. S., J. Mater. Sci. Lett. 1987, 6, 1431-1434.
16. Feighery, A. J.; Irvine, J. T. S., Solid State Ionics 1999, 121, 209-216.
17. Filal, M.; Petot, C.; Mokchah, M.; Chateau, C.; Carpentier, J. L., Solid State Ionics 1995,
80, 27-35.
18. Guo, X.; Tang, C. -Q.; Yuan, R. -Z., J. Euro. Ceram. Soc. 1995, 15, 25-32.
19. Butler, E. P.; Drennan, J. J. Am. Ceram. Soc. 1982, 65, 474-478.
20. Drennan, J.; Badwal, S. P. S. In Science and Technology of Zirconia III, Somiya, S.;
Yamamoto, N.; Yanagida, H., American Ceramic Society: Colombus, OH, 1988, pp.807-
817.
21. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Komatsu, M.; Haneda, H. Electrochemistry
2000, 68, 427-432.
22. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Komatsu, M.; Haneda, H., J. Electrochem.
Soc. 2000, 147, 2822-2829.
23. Macdrodt, W. C.; Woodrow, P. M., J. Am. Ceram. Soc. 1986, 69, 277-280.
24. Dijk, T. van; Burggraaf, A. J., Phys. Stat. Solid A (a) 1981, 63, 229-240.
25. Ioffe, M.; Inozemtsev, M. V.; Lipilin, A. S.; Perfilev, M. V.; Karpachov, S. V., Phys. Stat.
Solid A 30, 87 (1975).
26. Miyayama, M.; Yanagida, H.; Asada, A., Am. Ceram. Soc. Bull. 1986, 65, 660-664.
27. Radford, K. C.; Bratton, R. J., J. Mater. Sci. 1979, 14, 66-69.
28. Yuzaki, A.; Kishimoto, A. Solid State Ionics 1999, 116, 47-51.
29. Guo, X., J. Am. Ceram. Soc. 2003, 86, 1867-1873.

30. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Takenouchi, S., Ceramics International
2001, 27, 269-276.
31. Spearing, D. R., J. Am. Ceram. Soc. 1997, 81, 1964-1966.
32. Mori, T.; Yamamura, H.; Kobayashi, K.; Mitamura, T., J. Am. Ceram. Soc. 1992, 75,
2420-2426.
33. Ellison, A. J. G.; Navrotsky, A. J. Am. Ceram. Soc. 1992, 75, 1430-1433.
34. Kanno, Y., J. Mater. Sci. 1989, 24, 2415-2420.
35. Mori, T; Yamamura, H.; Kobayashi, K.; Mitamura, T., J. Mater. Sci. 1993, 28, 4970-
4973.
36. Vilmin, G.; Komarneni, S.; Roy, R., J. Mater. Sci. 1987, 22, 3556-3560.
37. Shi, Y.; Huang, X.; Yan, D., J. Euro. Ceram. Soc. 1994, 13, 113-119.
38. Butterman, W. C.; Foster, W. R., Am. Mineral. 1967, 52, 880-885.
39. Curtis, C. E. Snowman, H. G., J. Am. Ceram. Soc. 1953, 36, 190-198.
40. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Drennan, J.; Kim, D. Y. J. Am. Ceram. Soc
2000, 84, 2734-2736.
41. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Drennan, J.; Kim, D. Y., J. Electrochem.
Soc. 2002, 149, J35-J40.
42. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Drennan, J.; Kim, D. Y., J. Mater. Res.
2001, 16, 2377-2383.
43. Mecartney, M. L., J. Am. Ceram. Soc. 1987, 70, 54-58.
44. Beekmans, N. M.; Heyne, L., Electrochimca Acta. 1976, 21, 303-310.
45. Jung, Y. S.; Lee, J. -H.; Lee, J. H.; Kim, D. Y., J. Electrochem. Soc. 2003, 149, J49-
J53.
46. Chiang, Y. M.; Birnie III, D.; Kingery, W. D., Physical Ceramics, John Wiley, New
York, NY, 1997, p.362 and p.427.
47. Chiang, Y. M.; Birnie III, D.; Kingery, W. D., Physical Ceramics, John Wiley, New
York, NY, 1997, p.432.
48. Bonanos, N.; Steele, B. C. H.; Butler, E. P.; Johnson, W. B.; Worrell, W. L.; Macdonald,
D. D.; McKubre, M. C. H. in Impedance Spectroscopy; Macdonald, J. R., Wiley: New
York, NY, 1987, pp.191-316.
49. Sinclair, D.C.; West, A. R., J. Appl. Phys. 1989, 66, 3850-3856.
50. Rodewald, S.; Fleig, J.; Maier, J. J. Am. Ceram. Soc. 2000, 83, 1969-1976.
51. Rodewald, S.; Fleig, J.; Maier, J. J. Euro. Ceram. Soc. 1999, 19, 797-801.
52. Waser, R.; Baiatu, T.; Hrdtl, K. H. J. Am. Ceram. Soc. 1990, 73, 1645-53.
53. Waser, R.; Baiatu, T.; Hrdtl, K. H. J. Am. Ceram. Soc. 1990, 73, 1654-1662.
54. Snchez-Herencia, A. J.; Moreno, R.; Jurado J. R., J. Euro. Ceram. Soc. 2000, 20, 1611-
1620.
55. Fleig, J.; Rodewald, S.; Maier, J., Solid State Ionics 2000, 136-137, 905-911.
56. Fleig, J.; Rodewald, S.; Maier, J., J. Am. Ceram. Soc. 2001, 84, 521-530.
57. Fleig, J.; Maier, J., Solid State Ionics 1996, 85, 9-15.
58. Skapin, A. S.; Jamnik, J.; Pejovnik, S. Solid State Ionics 2000, 133, 129-138.
59. Lee, J. H.; Lee, J. H.; Kim, D. Y., J. Am. Ceram. Soc. 2002, 85, 1622-1624.
60. Fisher, C. A.; Matsubara, H., J. Euro. Ceram. Soc. 1999, 19, 703-707.
61. Kwon, O. J.; Yoon, D. N. In Sintering and Related Phenomena; Kuczynski, G. C.,
Plenum Publishing Company: New York, NY, 1980, pp.233-269.
30 Jong-Heun Lee
62. Yoo, Y. -S.; Kim, J. J.; Kim, D. Y., J. Am. Ceram. Soc. 1987, 70, C322-C324.
63. Kim, Y. S.; Park, J. K.; Yoon, D. N., J. Powder Metall. 1985, 21, 30-37.
64. Chaim, R.; Brandon, D. G.; Heuer, A., Acta. Metall. 1986, 34, 1933-1939.
65. Souza, D. P. E. De; Souza, M. F. De., J. Mater. Sci. 1999, 34, 4023-4030.
66. Lee, J. -H.; Mori, T.; Li, J. G.; Ikegami, T.; Takenouchi, S. J. Euro. Ceram. Soc. 2001,
21, 13-17.
67. Lee, J. -H.; Jung, Y. S.; Woo, H. S.; Chung, Y. C.; Kim, D. Y., J. Euro. Ceram. Soc.
2004, 24, 1129-1133.
68. Lee, J. H.; Lee, J. H.; Jung, Y. S.; Kim, D. Y., J. Am. Ceram. Soc. 2003, 86, 1518-
1521.
69. Choi, J. H.; Lee, J. H.; Kim, D. -Y.; J. Kor. Ceram. Soc. 2002, 39, 818-821.
70. Leed, J. S. Principles of Ceramic Processing, Wiley, New York, NY, 1995, p.25.
71. Badwal, S. P. S.; Drennan, J., J. Mater. Sci. 1989, 24, 88-96.
72. Badwal, S. P. S.; Hushes, A. E. J. Euro. Ceram. Soc. 1992, 10, 115-12.
73. Drennan, J.; Swain, M. V.; Badwal, S. P. S., J. Am. Ceram. Soc. 1989, 72, 1279-1281.
74. Jung, Y. S.; Lee, J. -H.; D. Y., J. Euro. Ceram. Soc. 2003, 23, 499-503.
75. Kuer, D.; Holc, J.; Hrovat, M.; Bernik, S.; Samardija, Z.; Kolar, D. Solid State Ionics
1995, 78, 79-85.
76. Jung, Y. -S.; Choi, J. H.; Lee, J. H. Mater. Sci. & Eng. B. 2004, 108, 219-222.
Chapter 2
EFFECTIVE ELASTIC MODULI OF ALUMINA,

ZIRCONIA AND ALUMINA-ZIRCONIA
COMPOSITE CERAMICS
Willi Pabst and Eva Gregorov

Department of Glass and Ceramics
Institute of Chemical Technology in Prague (ICT Prague)
Technick 5, 166 28 Prague, Czech Republic
Abstract
The effective elastic moduli of polycrystalline alumina and zirconia as well as alumina-
zirconia composite ceramics are investigated from the theoretical point of view, with a side-
glance on experimental results and applications. In the first section alumina, zirconia and
alumina-zirconia composites are introduced as structural materials, relations of elastic moduli
to other properties are recalled and targets of microstructural design are formulated. In the
second section elastic properties are defined from the viewpoint of rational mechanics for
anisotropic and isotropic materials in general. The difference between adiabatic and
isothermal elastic moduli is explained and estimated for alumina and zirconia. In the third
section effective elastic properties are defined and discussed from the viewpoint of
micromechanics and composite theory. General formulae are given for the calculation of
effective elastic moduli of polycrystalline materials from monocrystal data. Further, the
Voigt-Reuss bounds for the effective elastic moduli of multiphase materials are given, as well
the Hashin-Shtrikman bounds for the special case of two-phase materials. For porous
materials the dilute approximations are recalled as well as the predictions following from the
self-consistent, Mori-Tanaka, differential, Gibson-Ashby and Coble-Kingery approach as well
as the functional equation approach recently advocated by the authors. A comprehensive
survey of model relations for the porosity dependence of elastic moduli is given, including
exponential and power-law relations and a new relation recently proposed by the authors,
which seems to be the simplest relation allowing for the occurrence of a percolation threshold
(critical porosity). In the fourth section all these micromechanical bounds and relations are
applied to the alumina-zirconia system. Using a theoretically sound and experimentally
confirmed set of elastic moduli (and Poisson ratios) for dense (i.e. fully sintered)
polycrystalline alumina and zirconia the Hashin-Shtrikman bounds of dense alumina-zirconia
composites are calculated and compared to experimentally measured values. Several
predictions for porous alumina, zirconia and alumina-zirconia composites are compared to the
32 Willi Pabst and Eva Gregorov
data measured for ceramics with convex interconnected pores prepared by the starch
consolidation casting technique. A master fit curve is given for porous ceramics with this type
of matrix-inclusion microstructure and explicit numerical expressions are given throughout.
The last section gives examples of the mathematical modeling of other effective properties
and their dependence on composition and microstructure and an outlook is given to future
research aims. In particular, the significance of interfaces is emphasized and ideas on the way
from micromechanics to nanoscience towards a general mixture theory are outlined.
1 Introduction
Macroscopically (i.e. from a continuum viewpoint), brittle materials (e.g. ceramics, including
glasses) exhibit purely elastic response upon loading until fracture occurs when a critical
stress level, the strength, is exceeded. Since under usual conditions (i.e. moderate
temperatures and pressures) fracture is commonly encountered before the region of nonlinear
elasticity, plasticity, creep, viscoelasticity or viscosity can be attained, linear elasticity is the
appropriate theoretical framework to treat the stress-strain behavior of most ceramics and
glasses. The elastic properties of ceramics, defined as tensorial or scalar coefficients in linear
constitutive equations [Torquato 2002, Pabst & Gregorov 2003a], determine their (static and
dynamic) mechanical behavior and can have a decisive influence on many other properties, cf.
e.g. [Billington & Tate 1981, Cook & Pharr 1994, Davidge 1979, Green 1998, Irwin 1958,
Komeya & Matsui 1994, Lawn 1993, Menk 1992, Munz & Fett 1999, Salmang & Scholze
1982, Wachtman 1996]. According to the early theories of brittle fracture [Griffith 1920,
Orowan 1934, 1949] the theoretical strength 0 of a ceramic material should be directly
proportional to the square root of the elastic modulus E (tensile modulus), i.e.
E
0 ,
d
with being the specific surface energy (surface tension) and d the size of a typical defect
(Griffith) or interatomic spacing (Orowan). That means that the elastic modulus, together
with the specific surface energy, determines the maximum strength which a material can
principally achieve when it can be prepared in a defect-free state. For example, the smooth,
defect-free surface is the inherent reason for the high strength of glass fibers compared to
bulk glass. At the same time it is evident, however, that pores (which can be viewed as
defects) will reduce the strength by more than what should expected based on the decrease of
E alone (large pores more than small ones). Similarly, the critical value of the stress
intensity factor K IC , the so-called fracture toughness, which arises within the theory of linear
fracture mechanics [Irwin 1958], is directly proportional to the elastic modulus
K IC E .
There have been attempts to relate the elastic constants to other, operationally defined,
mechanical properties, like hardness or wear performance. It is remarkable, e.g. that using
Effective Elastic Moduli of Alumina, Zirconia and Alumina-Zirconia ... 33
Sanditovs simple formula for the (micro-) hardness H of glasses [Sanditov 1977, Volf
1988]
1 2
H = E ,
6 + 6
with being the Poisson ratio, one obtains for alumina and zirconia (inserting for E and
the values recommended below) hardness values of 29 GPa and 10 GPa, respectively, which
are not too far from the typical Vickers, Knoop or Berkovich indentation hardness values
experimentally measured and reported in the literature, which are approx. 24 GPa and 12 GPa,
respectively (orientational values only), cf. also [Conway 1986, Lawn & Howes 1981, Luo &
Stevens 1999, Marshall et al. 1982, Zeng et al. 1992]. Moreover, as a result of Hertzian
contact mechanics [Hertz 1882, Johnson 1985, Lawn 1993] the fracture toughness is related
to the elastic modulus and the indentation hardness via relations of the type
E f
K IC ,
H c3/ 2
where f is the loading force and c the radius of the cracks originating from the indentation.
In contrast to hardness, which is approximately twice as high for alumina than for zirconia,
both strength and fracture toughness are usually significantly higher for zirconia ceramics and
alumina-zirconia composites than for pure alumina ceramics. For zirconia they can attain
peak values of approx. 1500 MPa and 15 MPam1/2, respectively, compared to approx. 400
MPa and 4 MPam1/2 for alumina (orientational values only) [Becher & Rose 1994, Cawley &
Lee 1994, Hannink et al. 2000, Lee & Rainforth 1994, Munro 1997, 2002, NIST 2002,
Wachtman 1996]. Elastic properties, together with the (tensile) strength and the thermal
expansion coefficient (and possibly thermal conductivity or thermal diffusivity) also
determine a materials response to thermal shocks. Accordingly, all thermal shock parameters
R are of the form [Davidge 1979, Menk 1992, Salmang & Scholze 1982, Wachtman 1996]
(1 )
R .
E
That means, the thermal shock resistance is best for high-strength, low-expansion
materials with low elastic constants (low tensile modulus and low Poisson ratio).
From an atomistic point of view (discrete approach) the elastic constants of a material are
dependent on bond strength (force constant) and bond density (number of bonds per unit
volume). That means, all factors controlling bond strength and bond density will have an
influence on the elastic constants, e.g. composition (type of ions, including their size and
deformability), crystal structure (particle packing, coordination number, molar volume) and
bonding type (covalent or ionic) [Born & Huang 1954, Newnham 1975, Pauling 1968, Sahimi
2003a]. Although in principle it may be possible to calculate the elastic constants from the
force-distance relationships between atoms (viewed as harmonic oscillators) [Born & Huang
1954], this has been successful only for the simplest structures and bonding types and is
usually not done in practice. With increasing temperature the bond strength and density
decrease, due to lattice expansion (i.e. due to anharmonic effects), resulting in a decrease in
the elastic moduli [Ashcroft & Mermin 1976, Kittel 1988].
It is common practice to measure elastic constants and their temperature dependence
experimentally, either via static (e.g. three- or four-point bending, XRD or neutron scattering
in connection with strain gauges) or via dynamic tests (e.g. sound velocity methods or
resonant frequency techniques). It is well known from solid state physics, cf. e.g. [Ashcroft &
Mermin 1976, Kittel 1988], that the elastic constants determine the velocity of sound waves
in solids. For example, the velocity of transversal waves (shear waves) vT is given by
G
vT =

and the velocity of longitudinal waves (compressional waves) v L is given by the relation
vL =
3K + 4G
=
E (1 ) ,
3 (1 + ) (1 2 )
where G is the shear modulus, K the bulk modulus, E the tensile modulus, the Poisson
ratio and the density (bulk density of the material). Similar relations hold for surface
waves (Rayleigh waves). Therefore sound wave velocity measurements (e.g. by Brillouin
scattering, pulse-echo or ultrasonic techniques) can vice versa be used to determine the elastic
constants experimentally [Aksel & Riley 2003, Asmani et al. 2001, Hardy & Green 1995,
Lima et al. 2003, Martin et al. 1998a, 1998b, Shen & Hing 1997], e.g. from the shear wave
velocity vT the tensile modulus can be calculated via the relation
E = 2 G (1 + ) = 2 vT2 (1 + ) .
Materials selection and engineering design is a complex task which must take account of
various requirements depending on the intended applications. In many applications proper
materials selection and design implies a compromise between high stiffness (elastic modulus),
strength and fracture toughness and at the same time minimal weight, i.e. density. When this
is the case, certain ratios between the elastic moduli or strength and various powers of density
(e.g. E , E 2 or E 3 , depending on the prevailing loading mode) can be the critical
selection and design parameters, cf. e.g. [Brezny & Green 1994, Gibson & Ashby 1997].
Certain requirements concerning mechanical performance and reliability must be fulfilled
in all applications, e.g. thermally or electrically insulating materials (dense alumina for
electronic circuit supports, thermally insulating closed-pore alumina or zirconia), catalyst
supports, filter membranes and membrane supports (open-pore alumina), ionic conductors,
oxygen sensors (dense zirconia) and oxygen-ion-conducting membranes in solid oxide fuel
cells, oxygen separators and catalytic membrane reactors (open-pore zirconia) [Atkinson &
Seluk 2000, Fleig et al. 2003, Koga 2003, Tsuru 2003, Yamada 2003] and other functional
applications using electrical, magnetic or optical properties. Structural applications proper,
however, are those where mechanical performance is of primary concern. For alumina and
zirconia they range from engine parts, ball bearings, thread guides and cutting tools [Komeya
& Matsui 1994, Schwartz 1992, Takebayashi 2003, Fukuhara 2003] to bioinert implants for
bone and joint replacement [Boutin et al. 1988, Christel et al. 1989, Hulbert 1993, Kohn &
Ducheyne 1992, Kokubo et al. 2003, Park & Lakes 1992, Soltesz & Richter 1984]. The latter
may certainly be considered as one of the most demanding applications. The reasons are
manifold. A typical origin for post-operative or long-term implant failure is aseptic loosening,
caused by the large difference in elastic moduli between bone and implant. The order-of-
magnitude of the tensile moduli is between approx. 0.2 GPa for cancellous (trabecular) bone
and approx. 20 GPa for dense (cortical) bone, while titanium alloy implants (the group of
commonly used metal implants with the lowest elastic moduli) have tensile moduli of approx.
100 GPa, which is clearly considered to be elasto-incompatible [Kohn & Ducheyne 1992,
Ontan et al. 2000, Park & Lakes 1992, Sumner et al. 1998, Weinans et al. 2000]. Naturally,
elastocompatible solutions are generally not sought among ceramic materials but among
polymer-based composites, since for ceramic implant materials like alumina and zirconia the
elastic mismatch between bone and implant is even greater than for metals. Alternative
solutions to circumvent this principal problem, e.g. by allowing for ingrowth of viable bone
tissue into a porous implant and thus forming of a continuous and strong bone-implant
interface, is largely complicated by the severe and very specific microstructural demands: the
porosity must be open and connected, while the pore size (channel size) should be at least 150
m for bony ingrowth to be effective [Park & Lakes 1992]. It is clear that such complex
demands are very difficult to fulfil, especially with brittle materials, without the danger of a
critical deterioration of other mechanical properties (strength) and wear behavior.
Apart from the basic requirements on strength and fracture toughness, less well defined
mechanical properties like hardness, wear (abrasion and erosion), fatigue and ageing (both
intimately connected with corrosion) play a major role in all of these applications. This leads
to the very intricate questions of lifetime and reliability, which mostly have to be solved
empirically by proof testing [Davidge 1979, Menk 1992, Salmang & Scholze 1982,
Wachtman 1996]. In spite of the frequently cited inertness of alumina [Cawley & Lee 1994,
Lee & Rainforth 1994] it has to be kept in mind that, similar to other oxide ceramics, also
alumina is liable to stress-corrosion cracking and subcritical crack growth in the presence of
water vapor, which can seriously limit the lifetime of loaded alumina parts in quite common
environments. Zirconia can additionally undergo hydrothermal ageing (i.e. a transformation
from tetragonal to monoclinic at moderately elevated temperatures in the presence of water
vapor), a phenomenon which led to the general prohibition of steam sterilization for zirconia
implants and seems to affect also the long time-behavior of zirconia parts in physiological
solutions and in vivo at body temperature [Chevalier et al. 1997, 1999, 2004 Gremillard et al.
2004], although in the early 1990es it was believed to be completely inert with respect to
ageing in physiological solutions [Christel et al. 1989, Kohn & Ducheyne 1992].
Although as is evident from these considerations it should not be concealed that some
microstructural targets probably cannot be achieved with ceramic materials and some of the
earlier expectations have turned out to be too optimistic, for many of the above applications
alumina, zirconia and alumina-zirconia composite ceramics are quite suitable. Without doubt,
during the second half of the 20th century alumina has developed into the most frequently
used advanced engineering ceramic, while from the 1970es onwards zirconia has gained its
remarkable reputation as ceramic steel for its unusally high fracture toughness values (still
one order of magnitude lower than typical fracture toughness values of metals, however),
which is a consequence of the two toughening mechanisms occurring in zirconia and
zirconia-containing composite ceramics, viz. transformation toughening and microcrack
toughening, cf. [Cawley & Lee 1994, Cook & Pharr 1994, Fantozzi & Olagnon 1993, Green
et al. 1989, Hannink et al. 2000, Komeya & Matsui 1994, Lee & Rainforth 1994, Wachtman
1996].
In the present chapter we are focusing on the elastic behavior of alumina, zirconia and
alumina-zirconia composite ceramics as expressed by their effective elastic properties, which
are mechanical key properties in the sense exposed above. However, since the viewpoint
adopted in this chapter is primarily a theoretical one, with only a side-glance on practical
issues and experimental experience, the exposition will be general enough for many of the
relations to be applicable to other materials as well (at least in the field of ceramics) and may
serve as a guideline of analogous research in other ceramic systems. The second section
(following this introductory section) defines elastic properties from the viewpoint of rational
mechanics for anisotropic and isotropic materials in general. In the third section effective
elastic properties are discussed from the viewpoint of micromechanics and composite theory.
General formulae are given for the calculation of effective elastic moduli of polycrystalline
materials from monocrystal data. Further, the Voigt-Reuss bounds for the effective elastic
moduli of multiphase materials are given, as well the Hashin-Shtrikman bounds for the
special case of two-phase materials. A rather comprehensive survey of model relations for the
porosity dependence of elastic moduli is given, including exponential and power-law
relations and a new relation recently proposed by the authors. In the fourth section all these
micromechanical bounds and relations are applied to the alumina-zirconia system. The last
section gives examples of the mathematical modeling of other effective properties and their
dependence on composition and microstructure and an outlook is given to future research
aims. In particular, the significance of interfaces is emphasized and some ideas on the way
from micromechanics to nanoscience towards a general mixture theory are outlined.
2 Elastic Moduli from the Viewpoint of Rational Mechanics

The elastic behavior of brittle materials can be described within the framework of linear
elasticity theory. The reason is that, as a response to loadings above a critical value, the
stress-strain curve is cut off and brittle materials react by (brittle) fracture, i.e. the materials
rupture and loose integrity at stress or deformation levels lower than those that would be
necessary for the material to enter the nonlinearly elastic regime. For more details the reader
can refer e.g. to [Atanakovic & Guran 2000, Billington & Tate 1981, Gurtin 1972, Haupt
2000, Landau & Lifshitz 1986, Love 1927, Nye 1957, ilhav 1997, Slaughter 2002,
Sneddon & Berry 1958, Truesdell & Noll 2003, Truesdell & Toupin 1960] and the references
cited therein. For the purposes of the present chapter a rather special theory can be used,
which is linear with respect to the material model (physical linearization of the constitutive
equations) and to the kinematic measures (geometrical linearization of the strain measures for
small deformations). We note in advance that, although this theory must be considered as
rather special from the rational mechanics point of view, it is the usual setting in which elastic
properties of ceramics and most other materials are treated in the literature. With respect to
the applications intended in this chapter we first present the linear elastic properties of
anisotropic materials and then those of isotropic materials, cf. also [Pabst & Gregorov
2003a].
2.1 Elastic Properties of Anisotropic Materials
In the case of small deformations (i.e. invoking geometrical linearization) Hooke's law for
anisotropic elastic solids (i.e. the physically linearized constitutive equation) can be written in
direct tensor notation as
T = C E, (1)
where T is the Cauchy stress tensor (a symmetric second-order tensor), C the stiffness
tensor (a fully symmetric fourth-order tensor, also called elasticity tensor or tensor of elastic
constants) and
E
1
2
(
grad u + grad u T ) (2)
the so-called small strain tensor (a symmetric second-order tensor), with u being the
displacement vector and superscript T denoting tensor transposition. Using the more explicit
tensor index notation, Hooke's law can also be written in the form
Tij = C ijkl E kl . (3)
Alternatively, the constitutive equation of linearly elastic materials can be written in the
form
E = S T = C -1 T (4)
(inverse Hookes law), where S = C is the compliance tensor (again a fully symmetric
-1
fourth-order tensor, also called tensor of elastic coefficients), with the superscript 1 denoting
tensor inversion. In tensorial index notation
Eij = S ijkl Tkl . (5)
Due to the symmetry of T and E , the number of components of the stiffness and
compliance tensors is reduced from a total of 81 to 36 independent ones. Thus, it is possible
to represent these fourth-order tensors alternatively in the form of (6 x 6) matrices (which, of
course, do not have the transformation properties of a tensor), and to express Hooke's law and
inverse Hookes law in direct matrix notation (engineering notation) as
( ) = (C ) ( ) , (6)
( ) = (S ) ( ) (7)
or, alternatively, in matrix index notation as
i = C ij j , (8)
i = S ij j , (9)
respectively. In this notation ( ) and ( ) are 6-dimensional column vectors referring to

stress and strain, respectively, the components of which are defined as follows
1 T11 , 2 T22 , 3 T33 , 4 T23 , 5 T31 , 6 T12 (10 a-f)
1 E11 , 2 E 22 , 3 E33 , 4 2 E 23 , 5 2 E31 , 6 2 E12 (11 a-f)
and (C ) and (S ) are the stiffness and compliance matrices, respectively, the elements of
which are given by the substitution of index pairs as follows:
11 1 , 22 2 , 33 3 , 23 4 , 31 5 , 12 6 .
When it is assumed, as is usually done,1 that the stiffness and compliance tensors are
additionally symmetric with respect to their diagonals, the total number of independent
components is reduced from 36 to 21 (so-called Green elasticity or hyperelasticity, in contrast
to the so-called Cauchy elasticity, where this is not the case). Thus in the most general case of
well-defined anisotropy (triclinic monocrystals) the (6 x 6) stiffness or compliance matrices
or, alternatively, the fourth-order stiffness or compliance tensors, have 36 elastic constants or
coefficients, respectively, 21 of which can be assumed to be independent, cf. [Pabst &
Gregorov 2003a]. For reasons of convenience we confine ourselves to the stiffness matrices
in the sequel. It is understood, however, that completely analogous relations and symmetry
considerations are valid in the case of the compliance matrices.
1
This assumption is based on a more fundamental assumption concerning the elastic energy (stored energy
function). If the elastic energy, which is a potential function for the stress tensor, vanishes in the unstrained
state and can be expressed by a symmetric quadratic form, then the stiffness matrix is symmetric, i.e. the
elasticity tensor is fully symmetric.
The stiffness matrix of triclinic monocrystals can be written as:
C11 C12 C13 C14 C15 C16

C 21 C 22 C 23 C 24 C 25 C 26
C C C 33 C 34 C 35 C 36
(Ctriclinic ) = 31 32 (12)
C 41 C 42 C 43 C 44 C 45 C 46

C 51 C 52 C 53 C 54 C 55 C 56
C C 66
61 C 62 C 63 C 64 C 65
For the off-diagonal elements in this stiffness matrix and the following ones we
automatically assume validity of the symmetry condition for Green elasticity (hyperelasticity),
i.e.
C ij = C ji (13)
For materials of higher symmetry (monocrystals or polycrystalline bodies) the number of

independent elastic constants is further reduced as follows:
Monoclinic monocrystals (13 independent elastic constants):
C11 C12 C13 0 C15 0

C 21 C 22 C 23 0 C 25 0
C C32 C33 0 C35 0
(C monoclinic ) = 31 (14)
0 0 0 C 44 0 C 46

C 51 C 52 C 53 0 C 55 0
0 C 66
0 0 C 64 0
Orthorhomic monocrystals and orthotropic polycrystalline bodies (9 independent

elastic constants):
C11 C12 C13 0 0 0

C 21 C 22 C 23 0 0 0
C C 32 C 33 0 0 0
(C orthorhombic ) = 31 (15)
0 0 0 C 44 0 0

0 0 0 0 C 55 0
0 C 66
0 0 0 0
Trigonal monocrystals (6 independent elastic constants):
C11 C12 C13 C14 0 0

C 21 C 22 C 23 C 24 0 0
0
(Ctrigonal ) = CC31 CC32 C 33 0 0
0
(16)
41 42 0 C 44 0
0
0 0 0 C 55 C 56
0
0 0 0 C 65 C 66
with the additional conditions
C11 = C 22 , C13 = C 23 , C 44 = C 55 (17)
and
C14 = C 24 = C 56 , C 66 =
1
(C11 C12 ) (18)
2
Note that most textbooks claim that in the case of trigonal and tetragonal monocrystals it
is necessary to distinguish two cases of elastic symmetry, depending on the crystal class
4
(point group): one with 7 independent elastic constants (point groups 3 , 3 4 , 4 ,) and
m
one with 6 independent elastic constants (point groups 32 , 3m , 3 m , 422 , 4mm , 4 2m ,
4 2 2
). As shown in a recent paper by Forte and Vianello [Forte & Vianello 1996], this
mmm
claim is wrong.
For tetragonal monocrystals (6 independent elastic constants) the general form of the
orthorhombic (orthotropic) stiffness matrix, Equation (15), together with conditions (17)
applies. The same holds true for materials with higher than tetragonal symmetry. In these
cases, however, the following additional conditions hold for the non-zero elements:
Hexagonal monocrystals and transversely isotropic polycrystalline bodies (5

independent elastic constants):
C 66 =
1
(C11 C12 ) (19)
2
Cubic monocrystals (3 independent elastic constants):
C11 = C 22 = C 33 , C12 = C 23 = C 31 , C 44 = C 55 = C 66 (20)

Isotropic materials (2 independent elastic constants):
C11 = C 22 = C 33 , C12 = C 23 = C 31 , C 44 = C 55 = C 66 =
1
(C11 C12 ) . (21)
2
2.2 Elastic Properties of Isotropic Materials
In the case of isotropic materials the stiffness matrix can be written as follows (using the
definitions C12 and C 44 ):
+ 2 0 0 0

+ 2 0 0 0
0
(Cisotropic ) = 0 + 2 0

0
0
(22)
0 0 0
0
0 0 0 0
0
0 0 0 0
The elastic constants (elastic moduli) and are called Lam constants (or Lam
moduli, units [GPa]). Switching over from matrix notation to tensor notation the elasticity
tensor is
Cijkl = ij kl + ( ik jl + il jk ) (23)
where the 's are Kronecker deltas. Inserting the stiffness tensor given by Equation (23) into
Hooke's law, Equation (3), yields the Cauchy-Hooke law for isotropic materials,
[ ]
Tij = ij kl + ( ik jl + il jk ) E kl = ij kl E kl + 2 Eij = ij E kk + 2 Eij
(24)
Switching over from tensor index notation to direct tensor notation this corresponds to
T = (tr E ) 1 + 2 E , (25)
where tr denotes the trace of a tensor and 1 is the second-order unit tensor. The inverse
Cauchy-Hooke law for isotropic materials is

(tr T ) 1 .
1
E= T (26)
2 2 (3 + 2 )
The two Lam constants occurring in Equations (22) through (26) are one possible choice
of elastic constants which can be used in the case of isotropic materials. Depending on the
application in question, other elastic constants can be more advantageous, e.g. the tensile
modulus (Youngs modulus) E (units [GPa]), the shear modulus G (units [GPa]), the bulk
modulus K (units [GPa]) and the Poisson ratio (dimensionless). Some of these constants
are preferable from the practical point of view, since they can be relatively easily determined
by standard test procedures ( E and G ), while others are preferable from the theoretical
point of view, e.g. for micromechanical calculations ( G and K ). Note, however, that even
in the case of isotropic materials always two of these elastic constants are needed to
determine the elastic behavior completely.
In terms of E and the Cauchy-Hooke law for isotropic materials can be written as
E
T= E+ (tr E ) 1 (27)
(1 + ) (1 2 )
and the inverse Cauchy-Hooke law as
E=
(1 + ) T (tr T) 1 .
E (28)
(1 + )
From Equation (27) it is evident that for the Poisson ratio the values 0.5 and 1 are
not allowed. Actually, as a consequence of the second law of thermodynamics the following
inequality must hold for isotropic materials [Gurtin 1972, Torquato 2002]:
1 < < 0 .5 . (29)
Although it has been known very early that anisotropic materials (e.g. monocrystals) can
exhibit negative Poisson ratios in certain directions [Love 1927], until very recently most
textbooks on the mechanics of materials suggested that according to experience with real
materials the Poisson ratio should be positive for all isotropic materials (i.e. 0 < < 0.5 ), cf.
e.g. [Slaughter 2002]. This misleading suggestion was obviously supported by physical
intuition. Guided by a correct interpretation of the physical meaning of the Poisson ratio (viz.
that in the limit of incompressible materials 0.5 , while the case 0 corresponds to
a compressible material that does not contract in uniaxial extension perpendicular to the
extension direction), it was for a long time considered as evident that isotropic materials are
not allowed to extend perpendicular to the extension direction during uniaxial extension.
Research in materials science for more than two decades has clearly shown, however,
that this apparently obvious conclusion is wrong and that isotropic materials with negative
Poisson ratio, so-called auxetic materials, do exist and can be designed and produced
[Almgren 1985, Bathurst & Rothenburg 1988, Caddock & Evans, Choi & Lakes 1992, Evans
1989, Evans & Caddock 1989, Evans et al. 1992, Friis et al. 1988, Gibson et al. 1982, Lakes
1987, 1991, 1992, Milton 1992, 2002, Rothenburg et al. 1991, Sigmund 1994, 1995, Torquato
2002]. These materials show the unexpected behavior, that when extended in one direction,
they extend in all perpendicular directions. Of course, almost all ceramic materials, including
glasses, exhibit Poisson ratios in the range 0.1 0.4 and in the absence of more precise
information one usually estimates the value of the Poisson ratio for a ceramic materials
between 0.2 (or 0.17) and 0.3 (or 0.33). Nevertheless, newer research in ceramic science does
take the possibility of negative Poisson ratios into account [Roberts & Garboczi 2000].
In a similar way as the Young modulus E and the Poisson ratio are connected to the
uniaxial extension test, the shear modulus G and the bulk modulus K are connected to
simple shear and isotropic deformation (i.e. dilatation or compression). Note that, accidentally,
it turns out that the shear modulus G equals the second Lam constant . Since for
isotropic materials only two of the elastic constants are independent, the knowledge of any
pair of them is sufficient to calculate the other constants and thus to describe the elastic
behavior of isotropic materials completely. For easy reference in this chapter we list the most
important interrelations between the elastic constants:
(3 + 2 ) 9 KG
E= = 2G (1 + ) = 3K (1 2 ) = (30)
+ 3K + G
E 3K (1 2 ) 3EK
G== = = (31)
2(1 + ) 2(1 + ) 9K E
3 + 2 E 2G (1 + ) EG
K= = = = (32)
3 3 (1 2 ) 3 (1 2 ) 3 (3G E )
3K 2G E 2G 3K E
= = = = (33)
2 ( + ) 2 (3K + G ) 2G 6K
In the present chapter we will refer to these relations as the elasticity standard relations.
All elastic constants (and of course also its tensorial counterparts, the stiffness tensors)
are temperature-dependent, e.g.
E = E (T ) , (34)
T being the absolute temperature. From the theoretical point of view such a temperature
dependence arises naturally within the framework of linear thermoelasticity [Carlsson 1972,
ilhav 1997]. Note, however, that an alternative and equally valid theory of linear
thermoelasticity can be built by replacing the temperature with its Legendre-transformed
counterpart, the specific entropy s [ilhav 1997]. This results in entropy-dependent elastic
constants, e.g.
E~ = E~ (s ) . (35)
Concluding this section we would like to add some practical remarks. It seems that no
complete theory is available to predict the temperature (or entropy) dependence of the elastic
constants and thus usually they have to be determined from experimental measurements.
Elastic constants determined by static measurements, e.g. from load-deflection curves in
three-point bending via the formula [Menk 1992]
F L3
E= (36)
48 I Y
are called isothermal, because the measurement is slow enough to permit arbitrary heat
exchange, so that during measurement the specimen is in thermal equilibrium with its
environment. In Equation (36) F is the loading force, L the distance between the supports
(effective specimen length), Y the deflection of the specimen below the load and I the
moment of inertia of the cross-section, e.g. for circular and rectangular cross sections
D4 bh 3
I= and I= , (37 a, b)
64 12
respectively (with D being the specimen diameter, b the specimen width and h the height
of the specimen in the direction of deflection). On the other hand, elastic constants
determined via dynamic measurements, e.g. by resonant frequency techniques via the formula
[Aksay & Riley 2003]
E~ = (2vL )
2
(38)
(with being the bulk density, L the specimen length and v the frequency of the
longitudinal mode of vibration, i.e. the ratio of velocity of the compressional wave and its
wavelength) are called adiabatic or isentropic, because the measurement is so fast that heat
exchange will not occur to any sensible degree during signal registration, so that the specimen
is de facto thermally isolated until measurement is completed.
A rough estimate of the difference between isothermal and adiabatic elastic constants (e.g.
E and E~ , respectively) can be made as follows: Since it is known that the isothermal and
adiabatic shear moduli are equal [ilhav 1997], i.e.
~=G
G (39)
and on the other hand that the isothermal and adiabatic bulk moduli are related by the ratio of
specific heats at constant pressure (stress) and at constant volume (deformation), c P and cV ,
respectively [Ashcroft & Mermin 1976, Callen 1960, ilhav 1997], i.e.
K~ c P
= , (40)
K cV
the following formula can be given for the adiabatic tensile modulus E :
~
cP
9 KG
cV
E~ = . (41)
c
3 P K +G
cV
~
In order to obtain an estimate for the difference between K and K or E and E ,
~
respectively, we invoke the well-known relation for the difference between the specific or
molar heats at constant pressure and volume c P and cV , respectively, for isotropic
condensed phases,
c P cV = TVK (3 ) ,
2
(42)
known from equilibrium thermodynamics [Guggenheim 1957, Kittel 1988, Meyer 1977],
where T is the Kelvin temperature, V the specific or molar volume, respectively and K the
isothermal bulk modulus (i.e. the inverse compressibility) and the coefficient of linear
thermal expansion for the temperature in question. Further, we invoke the approximate
relation
3
cV = VK , (43)

known from solid state physics, where the so-called Grneisen constant is assumed to be 2,
a typical value for crystalline solids [Ashcroft & Mermin 1976, Grneisen 1926, Meyer 1977,
Salmang & Scholze 1982]. For example, inserting for alumina and zirconia the molar
volumes 25.5 cm3/mol and 20.2 cm3/mol, respectively (assuming densities of 4.0 g/cm3 and
6.1 g/cm3, respectively) or alternatively the specific volumes 0.250 cm3/g and 0.164 cm3/g,
6
respectively, the approximate linear thermal expansion coefficients 8 10 K-1 and
10 10 6 K-1, respectively [Menk 1992, Munro 1997, NIST 2002] and the bulk moduli 247
GPa and 184 GPa, respectively (see below), we obtain at room temperature (25 C or 298 K)
differences of the molar or specific heats c P cV of 1.081 Jmol-1K-1 ( = 0.0106 Jg-1K-1)
for alumina and 0.997 Jmol-1K-1 ( = 0.0081 Jg-1K-1) for zirconia. For the absolute values of
cV the rough estimates obtained via Equation (43) using = 2 are 75.6 Jmol-1K-1 ( = 0.741
Jg-1K-1) for alumina and 55.8 Jmol-1K-1 ( = 0.453 Jg-1K-1) for zirconia. These values are in
satisfactory agreement with literature values [Munro 1997, NIST 2002, Salmang & Scholze
1982]. With this input information at hand, Equations (40) and (41) (the latter in connection
with approximate values for the shear and bulk moduli, cf. Table 8 below for the definite
values) can now be used to obtain estimates for the differences that have to be expected at
room temperature between adiabatic elastic constants (measured via dynamic techniques) and
isothermal elastic constants (measured via static techniques). For alumina and zirconia
cP
= 1.015 K~ = 1.015 K and E~ = 1.003 E , (44)
cV
cP
= 1.018 K~ = 1.018 K and E~ = 1.002 E , (45)
cV
respectively. Note that the adiabatic elastic constants are always slightly higher than the
isothermal ones, since always c P > cV . It is evident, however, that for alumina and zirconia
near room temperature the difference is small enough to be negligible for most practical
purposes (1-2 % for the bulk modulus and 0.2-0.3 % for the tensile modulus).
3 Effective Elastic Moduli from the Viewpoint of

Micromechanics and Composite Theory
It is a long-standing aim of mixture theory to predict the mixture properties (and in general all
mixture quantities) from those of the constituents by calculation only, without direct
measurement. There are a few fortunate cases where this can be done with an accuracy
sufficient for practical purposes. Such is the case for (non-reacting) multiphase mixtures or,
more generally, materials with microstructure, particularly in cases where the contribution of
the phase boundaries (interfaces) is negligible. The latter can be said for many properties of
polycrystalline materials with crystal grains (crystallites) of a size in the tenths-of-micrometer
range and above (at least > 0.1 m), i.e. classical microcomposites, including the special case
of monophase polycrystalline materials. In contrast to nanocomposites (and nanocrystalline
materials in general) with typical dimensions < 0.1 m microcomposites (and
microcrystalline materials in general) can usually be modeled by considering volume
fractions and other microstructural characteristics of the constituent bulk phases only, while
the phase boundaries are considered to be sharp, i.e. without assignable volume fraction (in
contrast to nanomaterials, where it can be useful to treat the phase boundaries as diffuse, i.e.
with a finite thickness). Solid materials of this type are the main subject of the present chapter.
With regard to the purpose of this chapter (application to the alumina-zirconia system) we
confine ourselves to linearly elastic materials throughout, with special emphasis on one-phase
polycrystalline materials and two-phase materials (composites and porous materials).
Polycrystalline materials are one-phase materials with microstructure or morphology
(in contrast e.g. to glasses, which do not exhibit microstructure in this sense and are therefore
called amorphous) or multiphase materials (composites). They can be uniform (in the
absence of macroscopic gradients) but are always heterogeneous (at the microscopic scale).
Polycrystalline materials consist of small crystal grains (crystallites), isometric or anisometric
(elongated / prolate or flaky / oblate), held together by strong interaction forces (chemical
bonds) and exhibiting random orientation or (a certain degree of) preferential orientation.
Examples are densely sintered ceramic bodies. Polycrystalline materials with preferential
orientation of anisometric crystallites are in general anisotropic, but an adequate structural
description of partially anisotropic materials probably requires the elaboration of new,

fuzzy symmetry concepts in the future. Short-fiber composites e.g. with only a slight
tendency to fiber orientation, can be almost isotropic, although from the viewpoint of
classical composite theory they are usually treated as transversely isotropic. In the first case
two elastic constants would be needed for a complete description of elastic behavior, in the
second five. With respect to the purpose of this chapter (application to the alumina-zirconia
system) we confine ourselves in the sequel to polycrystalline materials and particulate
composites with isometric particles only. Thus, for evident reasons we consider
macroscopically isotropic materials with microstructures exhibiting random orientation and
disorder (i.e. non-periodicity). For similar reasons the theoretical treatment is focused on
spherical pores, although in principle deviations from sphericity can be accounted for within
the newly introduced concept of intrinsic elastic moduli (see below). The effective stiffness
tensor C e is defined via the linear constitutive equation
T = Ce E (46)
(i.e. averaged Hookes law for the heterogeneous material from the macroscopic point of
1
view), while its inverse, the effective compliance tensor S e = C e is defined via the inverse
equation
E = Se T . (47)
In these relations angular brackets denote volume averages of the Cauchy stress tensor
and the small strain tensor, respectively. For the correct use of volume and ensemble averages
in connection with random and periodic microstructures the reader should consult [Torquato
2002].
In his classical contribution [Voigt 1889] Voigt has shown, based on the assumption of
uniform strain (isostrain assumption), that in the case of isotropic polycrystalline materials
with a uniform and random microstructure the scalar elastic moduli can be calculated from
the components of the elasticity tensor (stiffness matrix), cf. [Voigt 1910]. In another
classical paper [Reuss 1929] Reuss has proposed another result for such materials, based on
the assumption of uniform stress (isostress assumption) and using the components of the
inverse elasticity tensor (compliance matrix), cf. [Nye 1957, Hearmon 1961, Wachtman 1996,
Green 1998]. In both approaches averaging is performed over all possible orientations of the
crystallites. Hill [Hill 1952] was apparently the first to recognize that the Voigt and Reuss
values were in fact the upper and lower bounds, respectively, to the property in question.
Moreover, it turned out that for most practical purposes the Voigt and Reuss bound calculated
for polycrystalline materials from monocrystal data are sufficiently close together for their
arithmetic average (the so-called Voigt-Reuss-Hill average, abbreviated VRH average in the
sequel) to be a satisfactory estimate of the property value [Hill 1952]. Comparison with
experimental data has confirmed the validity of the Voigt and Reuss bounds and the
usefulness of the VRH average for property prediction purposes [Green 1998]. The procedure
to calculate the Voigt and Reuss bounds for elastic moduli of polycrystalline materials from
monocrystal data is recalled in this chapter, since this procedure can generally serve to
establish reliable values of elastic moduli of dense polycrystalline one-phase materials, which
are the necessary input information for the prediction of effective elastic moduli of
composites or can significantly facilitate data fitting of effective elastic moduli of porous
materials (because one fit parameter can be considered as fixed).
The effective properties (sometimes also called gross, overall or macroscopic properties)
of the multiphase materials with microstructure defined above can in principle be predicted
exactly when the properties of the constituent phases and all details of the microstructure are
known. Of course the problem with this statement is the fact, that the microstructural details
must be known quantitatively in order to properly formalize the abstract statement
Effective property = Function (phase properties, microstructure).
The theoretical framework to attack this task is called micromechanics (sometimes also
mechanics or theory of composites or heterogeneous materials). The interested reader can
consult one of the excellent monographs in this field [Beran 1968, Christensen 1979, Markov
2000, Milton 2002, Nemat-Nasser & Hori 1999, Sahimi 2003a, Torquato 2002] for additional
information, but it has been attempted to make the present chapter to a large degree self-
contained. Micromechanics provides theoretical concepts for quantifying microstructural
information to an arbitrary degree of precision and including it into the description of a
material in the form of so-called correlation functions [Beran 1968, Jeulin 2001, 2002,
Markov 2000, Sobcyk & Kirkner 2001, Torquato 2002]. The lowest-order microstructural
information (one-point correlation function) concerns only the volume fractions of the phases.
Thus it is only a measure of composition, replacing the mass fractions or mole fractions from
more general mixture theories. Two-, three- and multi-point correlation functions involve
information on the size, shape, orientation and mutual arrangement of the phases. Such
higher-order correlation functions, however, are extremely difficult to determine. Evidently,
for real materials higher-order microstructural information is accessible only via tomographic
techniques (direct 3D information) or, approximately, by image analysis of planar sections
and the application of stereological approaches (3D information indirectly inferred from 2D
information). Of course, computer simulations may be a useful tool to generate model
materials and analyze their (virtual) microstructure, cf. e.g. [Roberts & Garboczi 2000] and
the extensive work of Torquato and his coworkers cited in [Torquato 2002].
For the remaining part of the present chapter we confine ourselves to isotropic materials,
i.e. the scalar effective elastic moduli M (where M stands for the effective tensile modulus
E , shear modulus G or bulk modulus K , respectively and in contrast to Equations (46)
and (47) above the subscript e denoting effective is omitted for convenience) as
functions of the phase moduli M i and microstructural information of the lowest order, i.e.
the volume fractions i (only compositional information). In this sense we introduce the
following basic assumption:
M = f (M i ,i ) , (48)
where M i ( i = 0, 1, 2 n ) are the phase properties of all n constituent phases and i the
volume fractions of these n phases. Note that the volume fractions of all n phases sum up to
unity, i.e.

i =0
i = 1, (49)
so that only n 1 phase volume fractions are independent. Note that although Equation (48)
is a frequently used starting point for modeling material properties and although at first sight
it might seem sufficiently general to be of almost universal validity, closer inspection makes
clear that this is by no means the case [Torquato 2002]. A simple counter-example
demonstrates this: an alumina suspension with 50 vol.% alumina can easily be prepared, but a
porous alumina ceramic body containing 50 vol.% of (open, water-saturated) pores can also
be prepared. In both cases the alumina volume fraction (or, alternatively, the water volume
fraction or porosity) is 0.5, but the first system is a viscous suspension ( E , G , K all zero /
undefined, viscosity non-zero / finite), the second an elastic porous solid ( E , G , K all non-
zero / finite, viscosity infinite / undefined). Thus, Equation (48) must be considered as a
convenient working hypothesis only. Nevertheless, in the lack of more detailed
microstructural information it is often the only feasible starting point for micromechanical
modeling. It is the purpose of this section to provide, on the basis of Equation (48), combined
with a minimum of additional qualitative information (e.g. concerning pore shape and
connectivity), an easy-to-grasp micromechanical framework to model porous alumina and
zirconia, and (dense and porous) alumina-zirconia composites. With respect to the intended
applications the emphasis is on two-phase materials. Of course, the relations given in this
section can be used for all brittle materials (ceramics) to mathematically describe the
dependence of effective elastic properties on the volume fractions of the constituent phases.
3.1 Effective Elastic Moduli of Polycrystalline Materials
As detailed in the preceding section, the elastic behavior of anisotropic materials

(monocrystals and anisotropic composites) is characterized by a fourth-order elasticity tensor,
which can be represented, as a consequence of symmetry, as a (6 x 6) matrix. This matrix is
again symmetric with respect to its diagonal (due to energetic reasons), so that the maximum
number of independent elements is 21 in the least symmetric case (triclinic monocrystal). For
higher symmetries this number of independent elements is smaller, cf. the discussion above
and [Gurtin 1972, Hearmon 1961, Nye 1957, Pabst & Gregorov 2003a].
When Hookes law is written in engineering matrix notation as
i = Cij j , (8)
where i is the 6-dimensional stress vector and j is the 6-dimensional small strain vector,
C ij is called stiffness matrix. On the other hand, using the so-called compliance matrix
S ij = C ij1 (the inverse of the stiffness matrix), Hookes law can be written in the inverse
form
i = S ij j . (9)
Note that the stiffness matrix and the compliance matrix have the same number of
independent components (called elastic constants and elastic coefficients, respectively) and
zero elements at the same positions, cf. [Nye 1957, Hearmon 1961]. The compliance matrix
can be calculated from the stiffness matrix via matrix inversion as follows
subdet (C ij )
S ij = , (50)
det (C ij )
where subdet (C ij ) is the subdeterminant obtained by omitting the i th row and the j th
column from C ij and det (C ij ) is the determinant of C ij . An analogous relation holds with
the roles of S ij and C ij interchanged. While for triclinic and monoclinic crystals the matrix
inversion is rather lengthy, for materials of higher symmetry it simplifies considerably. E.g.
for orthorhombic monocrystals and orthotropic polycrystalline materials or composites the
number of independent components is 9 and one obtains [Hearmon 1961]
C 22 C 23
C 23 C 33
S11 = etc. (51)
C11 C12 C13
C12 C 22 C 23
C13 C 23 C 33
1
S 44 = etc. (52)
C 44
For materials of even higher symmetry the inversion relations are as follows, cf. e.g.
[Hearmon 1961, Green 1998, Pabst & Gregorov 2004a]:
Trigonal monocrystals (6 independent components):
C 33 C C
S11 + S12 = , S11 S12 = 44 , S13 = 13 , (53 a,b,c)
K K K
C14 C + C12 C C12

S14 = , S 33 = 11 , S 44 = 11 . (53 d,e,f)
K K K
Tetragonal monocrystals (6 independent components):
C 33 1 C
S11 + S12 = , S11 S12 = , S13 = 13 , (54 a,b,c)
K C11 C12 K
C11 + C12 1 1
S 33 = , S 44 = , S 66 = . (54 d,e,f)
K C 44 C 66
Hexagonal monocrystals and transversely isotropic polycrystalline materials or

composites (5 independent components):
C 33 1 C
S11 + S12 = , S11 S12 = , S13 = 13 , (55 a,b,c)
K C11 C12 K
C11 + C12 1
S 33 = , S 44 = . (55 d,e)
K C 44
Cubic monocrystals (3 independent components):
C11 + C12 C 1
` S11 = , S12 = 12 , S 44 = . (56 a,b,c)
K K C 44
In these relations K , K and K have the following meaning:
K = C 33 (C11 + C12 ) 2C132 , (57 a)
K = C 44 (C11 + C12 ) 2C142 . (57 b)
K = (C11 C12 ) (C11 + 2C12 ) . (57 c)
Completely analogous relations are valid with the stiffnesses C ij interchanged by the
compliances S ij . Note, however, that for trigonal monocrystals, hexagonal monocrystals
(including transversely isotropic polycrystalline materials or composites) and isotropic
materials the elastic constant (stiffness) C 66 is given by
C 66 = 12 (C11 C12 ) , (58)
while the elastic coefficient (compliance) S 66 is given by [Nye 1957]
S 66 = 2 ( S11 S12 ) . (59)
This is the only case in which the complete analogy between stiffnesses and compliances
is disturbed. Of course, for isotropic materials C 66 and S 66 are identical to C 44 = C 55 and
S 44 = S 55 , respectively, cf. [Pabst & Gregorov 2003a], and therefore Equation (56 c) is
redundant, since S 44 = 1
2 ( S11 + S12 ) :
Isotropic materials (2 independent components):
C11 + C12 C
S11 = , S12 = 12 . (60 a,b)
K K
Voigt [Voigt 1889, 1910] has shown that, under the assumption that the strain inside the
material is uniform (isostrain assumption), the effective elastic moduli of a dense (i.e. pore-
free) polycrystalline material, e.g. a densely sintered ceramic, composed of crystallites of
arbitrary symmetry can be calculated from the 9 elastic constants (stiffnesses) C11 , C 22 , C 33 ,
C 44 , C 55 , C 66 , C12 , C 23 , C 31 , when the material as a whole is statistically or
macroscopically isotropic (quasi-isotropic). Of course, this requires random orientation of
the anisotropic (and possibly anisometric) crystallites. On the other hand, Reuss [Reuss 1929]
has shown that, under the assumption that the stress inside the material is uniform (isostress
assumption), the effective elastic moduli of a polycrystalline material composed of
crystallites of arbitrary symmetry can be calculated from the 9 elastic coefficients
(compliances) S11 , S 22 , S 33 , S 44 , S 55 , S 66 , S12 , S 23 , S 31 , when the material is
macroscopically isotropic. It should be noted that, according to Voigt and Reuss, even for
polycrystals composed of crystallites (monocrystals) with a symmetry lower than
orthorhombic / orthotropic only nine elastic constants or elastic coefficients completely
determine the elastic response of the polycrystalline aggregate [Green 1998, Hearmon 1961,
Hill 1952, Pabst & Gregorov 2004a, Wachtman 1996].
According to Voigt the effective tensile modulus of a macroscopically isotropic
polycrystalline material is
( A B + 3C )( A + 2 B)
EV = , (61)
2 A + 3B + C
the effective shear modulus

A B + 3C
GV = , (62)
5
and the effective bulk modulus
A + 2B
KV = . (63)
3
In these expressions A , B , C are given by
C11 + C 22 + C 33
A= , (64)
3
C12 + C 23 + C 31
B= , (65)
3
C 44 + C 55 + C 66
C= . (66)
3
According to Reuss the effective tensile modulus of a macroscopically isotropic

polycrystalline material is
5
ER = , (67)
3 X + 2Y + Z
the effective shear modulus
5
GR = , (68)
4 X 4Y + 3Z
and the effective bulk modulus
1
KR = . (69)
3( X + 2Y )
In these expressions X , Y , Z are given by
S11 + S 22 + S 33
X = , (70)
3
S12 + S 23 + S 31
Y= , (71)
3
S 44 + S 55 + S 66
Z= . (72)
3
Hill [Hill 1952] has shown that the Voigt and Reuss values are the upper bound and the
lower bound, respectively, of the effective elastic moduli M of statistically
(macroscopically) isotropic polycrystalline materials (Hills theorem), i.e.
M R M MV . (73)
Experience with the effective elastic properties of polycrystalline materials has shown
that the Voigt bound and the Reuss bound are for practical purposes sufficiently close [Green
1998, Hill 1952]. Therefore, following Hills recommendation [Hill 1952], it is common
practice to use the arithmetic mean of the Voigt and Reuss values,
MV + M R
M VRH = , (74)
2
as an estimate (approximate prediction) of the respective effective elastic modulus of dense

polycrystalline materials (VRH average). There is of course no rational reason to prefer the
arithmetic mean over other averages, e.g. the harmonic or geometric mean. But as long as the
Voigt and Reuss values are sufficiently close, the various means yield practically identical
results.
3.2 Voigt and Reuss Bounds for the Elastic Moduli of Multiphase Materials
When in a multiphase material the properties of the individual phases (constituents) are
known, it can be expected that certain bounds on the effective property of the multiphase
material (mixture) can be given without further microstructural information apart from
compositional information (phase volume fractions). For the elastic moduli (and many other
properties that are well defined via linear constitutive equations) this is indeed the case, cf.
[Beran 1968, Markov 2000, Milton 2002, Nemat-Nasser & Hori 1999, Sahimi 2003a,
Torquato 2002]. It seems plausible e.g. that in a multiphase material the value of an effective
property whose carrier are the bulk phases (and not the phase boundaries) should neither be
lower nor higher than the value of that property for any of the individual phases. Therefore it
can be concluded that such an effective property of the multiphase material, e.g. the effective
elastic modulus M , must be some kind of average value of all the phase (constituent)
properties, e.g. the elastic moduli M i . The most general average value is the general power
mean (weighted by volume fractions) [Pabst & Gregorov 2004a]:
M X = ( i M iN ) N .
1
(75)
In this expression the summation extends over all n phases, the exponent N determines
the type of mean value and the subscript X denotes this respective type. Setting e.g.
N = 1 results in the harmonic mean M H , N 0 yields the geometric mean M G ,
N = 1 the arithmetic mean M A , N = 2 the quadratic mean M Q , and N = 3 the cubic
mean M C etc. For these types of averages the so-called majority relation holds:
M H MG M A MQ MC . (76)
This is a mathematical law (law of logic), i.e. a proposition that can be rigorously proved
without further assumptions. Interestingly, however, it is found that the effective shear and
bulk moduli of multiphase materials always lie between the arithmetic and the harmonic
mean. This is a physical law (law of nature), i.e. a finding which can be (and has been in
micromechanics) rigorously proved for model materials with well defined microstructures. Of
course, since its proof is based on model assumptions, its applicability to real materials is and
remains, strictly speaking, a question of experience.
The arithmetic mean (upper bound) corresponds to the Voigt bound of the shear and bulk
modulus, GV and K V , respectively,
GV = G A = i Gi , (77 a)
KV = K A = i K i , (77 b)
and the harmonic mean (lower bound) corresponds to the Reuss bounds of the elastic moduli
MR
1
1 1
MR = MH = i = = i . (78)
Mi MR MH Mi
In the case of two-phase materials one of the two volume fractions is redundant
( 1 1 and 2 , because of Equation (49)) and the Voigt and Reuss bounds reduce
to
GV = (1 ) G1 + G2 (79 a)
K V = (1 ) K 1 + K 2 (79 b)
and
1 1 M 1M 2
= + MR = , (80)
MR M1 M 2 (1 ) M 2 + M 1
respectively, where is the volume fraction of one of the phases. When, additionally, one of
the phases is the void phase (with zero elastic moduli M 2 = 0 ), the elastic moduli of the
solid matrix phase are denoted as M 1 M 0 as usual, the Voigt bounds reduce to
M V = (1 ) M 0 (81)
and the Reuss bounds, Equation (80), degenerate to zero identically. In this case denotes
the volume fraction of pores and is called porosity. In the micomechanical literature the
Voigt-Reuss bounds for multiphase materials are called one-point bounds [Torquato 2002] as
a consequence of the fact that only the one-point correlation functions (i.e. the volume
fractions) are required as input information. It appears that the first application of the Voigt-
Reuss bounds (one-point bounds) to composites is due to Paul [Paul 1960].
Note that for composites we have written the Voigt bounds only for the shear modulus
G and the bulk modulus K , although in the literature it is sometimes tacitly assumed that
they hold also for the tensile modulus E [Eduljee & McCullough 1993]. In fact, strictly
speaking, the upper bound (Voigt bound) of the tensile modulus EV corresponds to the
arithmetic mean E A only if the Poisson ratios of all phases are equal [Sahimi 2003a,
Torquato 2002], a case barely encountered in practice. For this reason the Voigt bound of the
tensile modulus EV should be calculated from the arithmetic means (Voigt values) of the
shear and bulk moduli, GV and K V , respectively, via the standard relation
9 K V GV
EV = . (82)
3K V + GV
This is possible because the elasticity standard relations must hold for all isotropic
continua, whether heterogeneous on the microscale or not [Torquato 2002]. It will be shown
below, that in the case of the dense alumina-zirconia composite ceramics (Poisson ratios 0.23
and 0.31 for alumina and zirconia, respectively) the deviation of the Voigt bound EV
calculated via Equation (82) from the arithmetic mean turns out to be very small (< 0.6 %), so
that in practice the arithmetic mean will be a sufficiently precise approximation of the Voigt
bound. In the case of porous materials (degenerate case of two-phase composites where one
phase is the void phase) the Voigt bound of the tensile modulus EV corresponds to the
arithmetic mean E A exactly, cf. Equation (81).
We would like to close this subsection with a few words on the effective Poisson ratio.
From elasticity theory it is well known that the Poisson ratio has several exceptional
properties (e.g. in contrast to the elastic moduli E , G and K it can adopt negative values in
principle, as mentioned above). For the effective Poisson ratio of a polycrystalline or
multiphase, possibly porous, material the situation seems to be still rather unclear and would
need further investigation. There is considerable confusion in the literature concerning this
point. However, the following can be said: In general the effective Poisson ratio does not
obey the Voigt and Reuss bounds. It is not even true that the effective Poisson ratio must lie
between that of the constituent phases [Zimmerman 1994]. Nevertheless, in lack of a better
alternative it is often recommended to calculate the effective Poisson ratio of composites via
the mixture rule (i.e. as an arithmetic mean weighted by volume fractions). Unfortunately this
simple remedy fails in the case of porous materials (degenerate case of composites where one
phase is the void phase exhibiting zero elastic moduli). Here the mixture rule for the effective
Poisson ratio evidently breaks down, since the effective Poisson ratio is known to approach
finite values (mostly assumed to be in the range 0.2-0.33) for porosities close to 100 % (e.g.
in foams [Gibson & Ashby 1997, Roberts & Garbozci 2000]), while the close-to-zero value
corresponding to 100 % porosity (strictly speaking undefined for pores), and the value
corresponding to the inclusion phase in general, is attained only due to a singularity
(boundary layer) in the - -diagram [Zimmerman 1991a, 1994]. This is clear evidence of
the fact that the effective Poisson ratio can violate the Voigt bound. Markov [Markov 2000]
and Zimmerman [Zimmerman 1991a, 1994] discuss these points in great detail and derive the
self-consistent, differential and Mori-Tanaka predictions (see below), respectively, of the
asymptotic value towards which the effective Poisson ratio tends when approaching
*
100 % porosity. According to the self-consistent and differential approach (see below) this
asymptotic value is invariably
* = 0.2 , (83)
while according to the Mori-Tanaka approach [Mori & Tanaka 1973] the predicted
asymptotic value lies somewhere between the matrix Poisson ratio 0 and the value 0.2, and
is for porous materials (with spherical voids) given by the formula
1 + 5 0
* = . (84)
9 + 5 0
In the case of dense (pore-free) composites where the contrast between the moduli is not
too large and therefore the VRH average or, better, the arithmetic average of the Hashin-
Shtrikman bounds (HS average, see below) can be expected to yield a satisfactorily sharp
estimate of the effective moduli ( M VRH or M HS ) we recommend calculating the effective
Poisson ratio via one of the the standard relations (cf. the discussion above)
E 2G 3K 2G 3K E
= = = , (85)
2G 2(3K + G ) 6K
where the subscripts VRH or HS have been omitted for convenience. It will be shown below
that in the alumina-zirconia system the difference between the effective Poisson ratios thus
calculated and those calculated by the simple mixture rule is sufficiently small to be neglected
(< 1.4 %, corresponding to a maximum absolute error of 0.004, i.e. in the third decimal).
3.3 Hashin-Shtrikman Bounds for the Elastic Moduli of Two-Phase

Materials
In contrast to the case of polycrystalline monophase materials (see above) the Voigt-Reuss
bounds of two-phase materials are often too far apart to be useful for prediction purposes.
Hashin and Shtrikman [Hashin & Shtrikman 1963] derived the best possible bounds on the
effective elastic moduli of macroscopically isotropic two-phase composites given just
volume-fraction information. Actually the Hashin-Shtrikman bounds are two-point bounds,
but accidentally the key integral involving this two-point information has a form that reduces
to one-point information, i.e. volume-fraction information, in the case of macroscopically
isotropic composites [Torquato 2002]. When K 1 > K 2 and G1 > G 2 (the usual, so-called
well-ordered case, in contrast to the badly-ordered case ( K 1 K 2 ) (G1 G 2 ) 0 ,
considered by Walpole [Walpole 1966a]) the Hashin-Shtrikman upper bounds (in the sequel
HS upper bounds) for the shear modulus G and the bulk modulus K of two-phase
materials take the form
6 (K 1 + 2 G1 )
1
1
G +
= G1 + + 1 2 (86)
G2 G1 5 G1 (3 K 1 + 4 G1 )
HS
and
1
1 3
K +
= K1 + + 1 2 , (87)
K 2 K1 3 K 1 + 4 G1
HS

respectively. The Hashin-Shtrikman lower bounds (HS lower bounds) G HS and K HS are
obtained by interchanging the subscripts 1 and 2, cf. [Beran 1968, Markov 2000, Milton 2002,
Torquato 2002]. In the special case where one of the phases is the void phase (with zero
elastic moduli G 2 = 0 , K 2 = 0 ), where 2 is the porosity and the elastic moduli of the
solid matrix phase are denoted as G1 G0 , K1 K 0 as usual, the HS upper bounds reduce to
6 (K 0 + 2 G0 )
1
1
G +
= G0 + + (1 ) , (88 a)
5 G0 (3 K 0 + 4 G0 )
HS
G0
1
1
+ (1 )
3
.
+
K = K 0 + (89 a)
+
HS
K 0 3 K 0 4 G 0
and the HS lower bounds degenerate to zero identically. An alternative formulation of the HS
upper bounds is
+
G HS 15 K 0 + 20 G0
= 1 , (88 b)
G0 9 K 0 + 8 G0 + (6 K 0 + 12 G0 )
+
K HS 3 K 0 + 4 G0
= 1 . (89 b)
K0 3 K 0 + 4 G0
The HS bounds have been theoretically derived for the shear modulus G and the bulk
modulus K . The corresponding HS bounds for the tensile modulus E can be obtained via
the standard relation [Pabst & Gregorov 2004a]
9 K HS G HS
E HS = . (90)
3K HS + G HS
In the case of porous materials with a matrix or skeleton Poisson ratio of 0.2 ( 0 = 0.2 ,
corresponding to 3K 0 = 4G0 ) it can easily be shown that the HS upper bounds reduce to
+
E HS G+ K+ 1
= HS = HS = . (91)
E0 G0 K0 1+
Note that for this case the HS upper bounds are identical to the Mori-Tanaka predictions
for random materials of matrix-inclusion type with spherical inclusions [Mori & Tanaka 1973,
Weng 1984, Zimmerman 1994] and that Equation (91) corresponds to a relation known in
geophysical context under the name Kuster-Toksz relation [Kuster & Toksz 1974,
Zimmerman 1991b]. It is shown below that in the alumina-zirconia system the Kuster-Toksz
relation, Equation (91), is an excellent approximation to the HS upper bound for the tensile
modulus (error < 0.1 %) and for the shear modulus (error < 2.6 %) but not for the bulk
modulus (error < 14.3 %). For the purpose of later reference we note that Equation (91) can
be approximated by the following second-order polynomial:
+
E HS G+ K+
= HS = HS 1 1.71 + 0.71 2 . (92)
E0 G0 K0
Note that, in contrast to the relatively wide Voigt and Reuss bounds, which are security
bounds (worst case bounds, which certainly cannot be exceeded), the much tighter Hashin-
Shtrikman bounds are optimal or realizable in the sense that microstructures can be devised
for which these bounds can be attained. Without microstructural information of higher order
(for real materials) or more special assumptions (for model materials) they cannot be
improved, cf. [Beran 1968, Markov 2000, Milton 2002, Nemat-Nasser & Hori 1999,
Torquato 2002]. Generally the microstructure corresponding to the HS upper bound is the so-
called Hashin assemblage [Hashin 1962], consisting of polydisperse composite spheres
containing concentric spherical inclusions. In the case of macroscopically isotropic porous
materials the Hashin assemblage can be imagined as a porous material consisting of hollow
spheres with an infinitely wide size distribution that enables space filling with a fractal
microstructure.
3.4 Dilute Approximations for the Effective Elastic Moduli of Porous

Materials
In the case of porous materials it is convenient to define a relative elastic modulus as
M
Mr , (93)
M0
where M is the effective elastic modulus (as above) and M 0 the elastic modulus of the
matrix phase (in the case of porous materials of the matrix-inclusion type, i.e. porous
materials with essentially closed pores) or else the elastic modulus of the solid skeleton phase
(in the case of bicontinuous porous materials, i.e. open-pore cellular solids). For convenience
we will refer to elastic moduli (and the Poisson ratio) with subscribed index 0 as matrix
elastic moduli (and matrix Poisson ratio) in the sequel, irrespective of the underlying concept
of pore topology. The Voigt bounds of the relative elastic moduli M rV of porous materials
are linearly decreasing with a slope of 1 ,
MV
M rV = = 1 (94)
M0
while the HS upper bounds are nonlinearly decreasing (e.g. with an initial tangent slope of
1.71 for a material with a matrix Poisson ratio of 0 = 0.2 ), cf. Equations (88 b), (89 b)
and (92). For very low porosities ( 0 ), where mutual interactions between the pores can
be neglected (so-called dilute approximation) it is justified to assume a linear dependence of
the relative elastic moduli on the porosity,
M r = 1 [M ] , (95)
where [M ] is the intrinsic elastic modulus defined as
M r 1
[ M ] lim . (96)
0
Note the new sign convention in this definition, which is in contrast to our earlier
definition [Pabst & Gregorov 2003c, 2004a, 2004b, 2004c] and to common practice in
suspension rheology [Krieger 1972] but contributes much to clarity and simplicity, cf. also
[Pabst 2005, Pabst et al. 2006].
Based on the exact classical solution of the spherical-cavity problem (single spherical
void embedded in an infinite elastic medium [Goodier 1933, Timoshenko & Goodier 1951]),
dilute-limit approximations to the problem of non-interacting spherical cavities (i.e. single
pores or a dilute system of pores) in an elastic matrix have been obtained by Dewey [Dewey
1947] and Mackenzie [Mackenzie 1950], cf. also [Christensen 1979, Christensen 2000,
Nemat-Nasser & Hori 1999, Torquato 2002]. For the relative shear modulus G r , the relative
bulk modulus K r , the relative tensile modulus E r and the relative Poisson ratio r we have
15 (1 0 )
Gr = 1 , (97)
7 5 0
3 (1 0 )
Kr = 1 , (98)
2 (1 2 0 )
3 (1 0 ) (9 + 5 0 )
Er = 1 , (99)
2 (7 5 0 )
3 (1 02 ) (1 5 0 )
vr = 1 + . (100)
2 0 (7 5 0 )
An analogous dilute-limit approximation is available for the relative Lam constant .

The generalization to triaxial ellipsoids is due to Eshelby [Eshelby 1957]. Explicit formulae
for the special limiting cases of needle-like and disk-like inclusions have been given by
Torquato [Torquato 2002]. Nemat-Nasser and Hori have shown that these relations have been
derived under the assumption of prescribed macrostrain [Nemat-Nasser & Hori 1999]. When
macrostress is prescribed the results are of the form [Zimmerman 1991b]
1
Mr = , (101)
1 + [M ]
which can be developed into a series expansion in and truncated after the first-order term
(in ) to give again the dilute-limit expressions corresponding to Equations (97) through
(100), cf. Equation (95). It is evident that the first-order coefficients (intrinsic elastic moduli)
are all functions of the matrix Poisson ratio 0 . Note that, according to the dilute
approximation, in the special case 0 = 0.2 the relative Poisson ratio of a porous material
with spherical pores is equal to unity, i.e. the effective Poisson ratio remains unchanged with
increasing porosity. Note also that the so-called self-consistent approach [Bruggeman 1937,
Budiansky 1965, Hill 1965], which in a certain sense takes interactions into account, results
in very similar relations, except for the fact that the intrinsic elastic moduli are functions of
the effective Poisson ratio instead of the matrix Poisson ratio 0 . Of course, in the special
case of porous materials with spherical pores and 0 = 0.2 both the dilute approximation (in
the dilute limit 0 ) and the self-consistent approach (principally intended for finite )
lead to the identical result
M r = 1 2 , (102)
i.e. the intrinsic elastic modulus equals 2 for the case of spherical pores in a 0 = 0.2 matrix
material. Obviously, deviations from this value can be attributed either to deviations of the
pore shape from sphericity (including a topological transition from isolated to connected) or
to deviations of the matrix Poisson ratio from the magic value 0.2.
Values of the intrinsic elastic moduli [G ] , [ K ] and [ E ] are listed in Table 1 in
dependence of the matrix Poisson ratio 0 .
Note that for the normal matrix Poisson ratios 0 between 0 and 0.5 the intrinsic
tensile modulus [ E ] remains very close to 2, increasing from 1.929 (for 0 = 0) to a
maximum value of 2.006 (for 0 = 0.268667), followed by a decrease to 1.917 (for 0 = 0.5).
Due to this anomalous behavior of [E ] the value 2 is attained exactly for two different
values of 0 (0.2 and 1/3). The limiting values of the bulk and tensile moduli [K ] and [E ]
for (extremely auxetic) materials with a (negative) matrix Poisson ratio ( 0 = 1 ) are 1,
corresponding to the Voigt bound (values < 1 cannot occur). Curiously, the limiting value of
the intrinsic shear modulus [G ] for 0 = 1 materials is 2.5, obviously the counterpart of
the value 2.5 [Einstein 1906] for the intrinsic shear viscosity occuring in suspension rheology,
cf. [Pabst 2004]. Note also that for the typical matrix Poisson ratios 0 in the range
between 0.17 and 0.33 the values of [G ] are still relatively close to 2 (range 1.875-2.024),
while for [K ] this is not the case. For a matrix Poisson ratio of 0 = 0.5 (corresponding to a
totally incompressible matrix) the intrinsic bulk modulus diverges, i.e. a very small amount of
pores would be extremely efficient (detrimental) in such a case. In other words, there would
be a singularity in the K - -diagram at = 0 , where the effective bulk modulus steeply
falls down from the value of the matrix bulk modulus K 0 to zero. In this connection we note
that, due to the condition [ M ] 1 , all linear relations of the type (95), whether based directly
on the dilute approximation or derived from the self-consistent approach, predict a critical
porosity C = [ M ] 1 1 , for which the effective elastic moduli become zero, i.e. the
material looses integrity. In the context of percolation theory C can be interpreted as a
percolation threshold [Kovak 1999, Markov 2000, Sahimi 1994, Sahimi 2003a, Stauffer &
Aharony 1985, Zallen 1983].
Table 1. Intrinsic elastic moduli [G ] , [K ] , [E ] for different matrix Poisson ratios 0 .
0 [G ] [K ] [E]
1.0 2.5 1 1
0.5 2.368 1.125 1.539
0.2 2.250 1.286 1.800
0.1 2.200 1.375 1.870
0 2.143 1.5 1.929
0.1 2.077 1.688 1.973
0.17 2.024 1.886 1.994
0.2 2 2 2
0.23 1.974 2.139 2.004
0.25 1.957 2.25 2.005
0.26 1.947 2.313 2.006
0.268667 1.939 2.371 2.006
0.27 1.938 2.380 2.006
0.3 1.909 2.625 2.005
0.31 1.899 2.724 2.004
0.333333 1.875 3 2
0.4 1.800 4.5 1.980
0.5 1.667 1.917
3.5 Nonlinear Relations for the Effective Elastic Moduli of Porous

Materials
Experience shows that usually the porosity dependence of the effective elastic moduli is
not linear. Thus, since it is clear that the linear relation (95) claims validity and practical
significance only in the case of very small porosities (dilute approximation), there have been
numerous attempts to extend it to higher porosities by allowing for a nonlinear dependence.
The simplest way to do so is the Coble-Kingery approach [Coble & Kingery 1956], which is
as follows [Pabst & Gregorov 2003b, 2004a]: Take the linear relation, Equation (95), for the
matrix Poisson ratio in question and add a quadratric term in (second-order polynomial),
M r = 1 [M ] + 2 . (103)
Then determine the value of the coefficient via the condition that M r = 0 at least for
= 1 (which is necessary in order not to violate the Voigt bound). In general one obtains
[Pabst & Gregorov 2004a, 2004c]
M r = 1 [ M ] + ([ M ] 1) 2 . (104)
In the special case of porous materials with spherical pores and 0 = 0.2 this reduces to
M r = (1 )
2
(105)
[Pabst & Gregorov 2003b, 2003c, 2004a, 2004c]. This relation is identical with the
prediction of the differential approach for porous materials with 0 = 0.2 [McLaughlin 1977,
Norris 1985, Norris et al. 1985, Zimmerman 1991a, 1991b]. An alternative derivation of this
relation, using a functional equation approach, has been given recently [Pabst & Gregorov
2003c]. Interestingly, the same result was found for the tensile modulus by fitting a large
amount of experimental data on real materials with the semi-empirical Gibson-Ashby model
for open-pore cellular solids [Gibson & Ashby 1982, 1997]. Also the Gibson-Ashby relation
for the shear modulus has been found to match the general form
Gr = (1 )
2
(106)
where is an adjustable parameter to be determined e.g. by fitting. It has been shown

recently, however, that the value = 3 (1 + 0 ) / 4 proposed by Gibson and Ashby
[Gibson & Ashby 1982, 1997] is probably too low and should be replaced by
= 5 (1 + 0 ) / 6 in order to ensure that Equation (106) reduces to Equation (105) in the
case of porous materials with 0 = 0.2 , cf. [Pabst et al. 2006]. Irrespective of the matrix
Poisson ratio the binomial relation (105) may be expected to be reasonable for the tensile
modulus E and, depending on the matrix Poisson ratio, also for the shear modulus G , but
certainly not for the bulk modulus K . This is connected with a principal problem of Coble-
Kingery relations, cf. [Pabst et al. 2006]: if [ M ] > 2 , the second-order polynomials exhibit a
minimum with negative M r values, which is clearly nonsense from the physical point of
view. Obviously, this problem is harder in the case of the bulk modulus than with the other
two moduli, cf. Table 1. Therefore an entirely different relation has been proposed for the
bulk modulus of low-density cellular solids (i.e. for very high porosities) [Christensen 1986,
Warren & Kraynik 1988]:
1 2 0
Kr = (1 ) . (107)
3
Evidently the linear form of this relation prohibits its use in the range of low porosities.
In particular, in the case of porous materials with 0 = 0.2 it predicts K r = 0.2 (1 ) ,
which is clearly wrong. A possibility to overcome this problem, in lack of a better alternative,
is to combine the Warren-Kraynik-Christensen relation (which can be a realistic prediction in
the high-porosity region) with the Coble-Kingery relation (which can be realistic in the low-
porosity region) [Pabst et al. 2006]. To ensure in any case positivity of the K r values it has
been recommended to use the porosity X at the crossover of the two curves (calculated by
simply equating the Coble-Kingery relation (104) and the Warren-Kraynik-Christensen
prediction (107) for the matrix Poisson ratio in question) as the point of continuation, cf.
[Pabst et al. 2006].
In the remaining part of this subsection let us concentrate on the tensile modulus. For
evident reasons (discussed above) it is in practice unnecessary to emphasize the condition
0 = 0.2 in this case, because for pores of spherical shape the intrinsic tensile modulus [ E ]
is always very close to the benchmark value of 2 (1.97-2.01), even when the matrix Poisson
ratio varies in the wide range between 0 = 0.1 and 0 = 0.4 (which covers practically the
whole range of interest for ceramic materials), cf. Table 1. Notwithstanding the problems
with the bulk modulus the Coble-Kingery relations can yield a prediction of the porosity
dependence of the relative tensile modulus that fulfils all basic criteria of physical
plausibility: it ensures that M r = 0 for = 1 and it does not violate the HS upper bound.
For example in the case of porous materials with 0 = 0.2 the initial tangent slope is 2
and the curvature is 1, whereas the HS upper bound is characterized approximately by an
initial tangent slope of 1.71 and a curvature of 0.71 in this case, cf. Equation (92).
Through the years several other relations have been proposed to predict or describe the
porosity dependence of the relative tensile modulus, and recently the most important ones
have been collected and rationally classified by the authors [Pabst & Gregorov 2003b, 2003c,
2004a, 2004b, 2004c, Pabst et al. 2006]. In the sequel we give a brief overview of these, in
unified notation, using the critical porosity C and the newly introduced concept of intrinsic
elastic moduli [ M ] wherever feasible. It is understood, that in practice these parameters will
usually be treated as fit parameters to be determined from experimentally measured data a
posteriori. This is particularly true for the critical volume fraction C , where reliable and
sufficiently precise a priori estimates will hardly be available [Sahimi 2003a]. Note that for
porous materials there exists no reliable benchmark value for C comparable to the value of
approx. 0.64 (i.e. 64 %) for the packing density of monodisperse rigid spheres in random
close packed (rcp) arrangement [Aste & Weaire 2000, Bernal 1959, Bernal & Mason 1960,
Rintoul & Torquato 1996, Scott 1960]. Nevertheless, in exceptional cases there exist
promising results of computer simulations of porous model materials [Roberts & Garbozci
2000]. When these model materials are considered to reflect the microstructure of any real
material to a sufficient degree, then of course the following relations may be used for
comparing experimental data with model predictions or even for property prediction purposes.
The same holds when a sufficient number of data has been measured for materials having
certain typical microstructural features in common (e.g. a certain pore space topology). Also
in this case (in fact a real experiment in contrast to the virtual computer experiment) the
following relations can be used to predict the behavior of similar materials.
Spriggs [Spriggs 1961] suggested the use of a simple exponential relation of the form
E r = exp ([ E ] ) , (108)
where the intrinsic tensile modulus is principally a parameter to be determined by fitting

experimentally measured data. For small porosities ( 0 ) Equation (108) approximates
Equation (95), because of the truncated series expansion
exp ([ E ] ) 1 [ E ] + ... . (109)
A derivation of the Spriggs relation via the so-called functional equation approach has
been given recently by the authors [Pabst & Gregorov 2003c]. Zimmerman was apparently
the first to use a differential scheme approach to derive a Spriggs-type relation [Zimmerman
1984, 1991b], which for the special case of porous materials with spherical pores reduces to
the simple prediction
E r = exp (2 ) , (110)
Of course, the Spriggs relation (including its special case Equation (110)), suffers from
the serious principal drawback that E r is not zero for = 1 , i.e. the Spriggs relation
necessarily violates the HS upper bound and even the Voigt bound. For this reason
Hasselman [Hasselman 1962], based on previous work by Hashin [Hashin 1962], suggested a
relation which can be written as
1
Er = , (111)
1 + CH
where, again, C H has to be determined principally by fitting experimentally measured data.

This relation is clearly nonlinear and monotonically decreasing, and E = 0 is guaranteed for
= 1 . Unfortunately, however, the inverse of C H cannot be interpreted as a critical
porosity since in the limit 1 C H relation (111) diverges ( E ), i.e. 1 C H
must always lie outside of the interval 0 < < 1 . Thus the interpretation of 1 / C H in
terms of a critical porosity is principally inadmissible and there is no physical meaning left in
the Hasselman relation (111), cf. [Pabst & Gregorov 2004a, 2004c, Pabst et al. 2006]. This
finding is also corroborated by the fact that fitting of experimentally measured values
invariably leads to positive values of C H [Pabst et al. 2004b, 2005], which implies that the
tentative critical porosity 1 / C H would always be a negative quantity (which is nonsense
from the physical point of view). Note, however, that a very special case of the Hasselman
relation can in principle be useful for predictive purposes, viz. the case C H = 1 . In this case
the Hasselman relation reduces to the Kuster-Toksz relation [Kuster-Toksz 1974]
1
Er = , (112)
1+
which is identical to the Mori-Tanaka prediction and the HS upper bound for porous materials
with spherical pores and 0 = 0.2 , as discussed above, cf. Equation (91).
Recently, it has been recognized [Pabst & Gregorov 2004a, 2004b] that the modified
exponential relation
[E]
E r = exp (113)
1
also circumvents the aforementioned compatibility problem at = 1 . Most surprisingly, it

seems that this simple fact has not been realized before in the literature. In [Pabst &
Gregorov 2004b] a derivation of this relation has been given via a functional equation
approach. Trivially, in the case of porous materials with spherical pores the modified
exponential leads to the prediction
2
E r = exp . (114)
1
Note that this exponential model results in a zero relative tensile modulus only in the
limiting case of 100 % porosity. Without doubt, porosities close to 100 % can in principle be
achieved, e.g. in some aerogels [Gibson & Ashby 1997]. However, the usual case
encountered in practice will be a complete structural breakdown (loss of integrity) at
significantly lower porosity levels. In order to allow for the possibility of E r = 0 at
porosities lower than 100 % (i.e. < 1 ), it is necessary to include a critical volume fraction
C in the modulus-porosity relation, which is able to take the possible occurrence of a
percolation threshold into account. This results in a Mooney-type exponential relation
[Mooney 1951, Pabst & Gregorov 2004a, 2004b]
[E]
E r = exp (115)

1 C
Again, for small porosities ( 0 , which implies << C ) accordance with Equation
(95) is guaranteed. In contrast to the Hasselman relation (111) the physical interpretation of
C in relation (115) as a critical porosity is in principle admissible. As before, under the
assumption of spherical pores we can set [ E ] = 2 , which reduces the number of adjustable
fit parameters from two to one.
Power-law relations represent another class of fit models, principally different from the
exponential relations just presented. The simplest relation of this kind is Archies relation
[Archie 1942, Markov 2000]
E r = (1 ) [ E ] , (116)
a derivation of which via the functional equation approach has been given recently [Pabst &
Gregorov 2003c]. Of course, in the case of the tensile modulus the Archie relation can be
considered as a generalization of the Coble-Kingery relation
E r = (1 ) 2 . (117)
Neither of these relations exhibits a compatibility problem at = 1 and, again, in the

dilute limit ( 0 ) they reduce to Equation (95), because of the truncated series expansion
(1 ) [ E ] 1 [ E ] + ... . (118)
Moreover, as before with the exponential relations, in order to allow for the possibility of
the occurence of a percolation threshold, i.e. E r = 0 for porosities < 1 , an additional
parameter C can be introduced. This results in a Krieger-type power-law relation [Krieger
1972, Pabst 2004], in elasticity context often called Phani-Niyogi relation [Phani & Niyogi
1987a, 1987b]:
[ E ] C

E r = 1 (119)
C
As before, under the assumption of spherical pores we can set [ E ] = 2 , which again
reduces the number of adjustable fit parameters from two to one. Evidently, from a principal
point of view, there is nothing to be said against the interpretation of C in relation (119) as a
critical porosity. We would like to emphasize again that, although sometimes considered to be
purely empirical, all the exponential and power-law relations mentioned above can be derived
via a functional equation approach, cf. [Pabst & Gregorov 2003c, 2004b, Pabst 2004]. What
makes them semi-empirical fit models in the end is only the fact that, due to variations in pore
shape, the first-order coefficient (intrinsic tensile modulus) may not be reliably known and
that, due to the difficulty to assess and quantify pore size distribution and connectivity, a
priori estimates of the critical porosity are usually not available.
Recently, a new relation has been proposed by the authors [Pabst & Gregorov 2004a,
2004c]:
( ) (1(1 ) ) .
E r = 1 [ E ] + ([ E ] 1) 2 C
(120)
This relation has been found heuristically in analogy to the Robinson relation in
suspension rheology [Robinson 1949, Mooney 1951, Pabst 2004] and reduces to the linear
approximations (95) or (99) in the dilute limit ( 0 , implying << C ). Furthermore it
ensures that E r = 0 when = C , as required. Setting [ E ] = 2 our relation adopts the
extremely simple form

E r = (1 ) 1 , (121)
C
which seems to be the simplest thinkable relation allowing for a percolation threshold (via the
critical porosity C ). In the absence of a percolation threshold (i.e. C = 1 ) it reduces to the
Coble-Kingery relation (117), as required. We recall that [ E ] = 2 is always a reasonable
approximation for porous materials with spherical or isometric pores, independently of the
precise value of the matrix Poisson ratio 0 , cf. Table 1 and the discussion above. Of course,
unless the critical porosity C is known a priori (a rather exceptional case) and an application
of Equation (121) for predictive purposes is intended, this discussion is of more or less
philosophical character. As long as Equation (121) is used as a fit model its application to
cases where the assumption [ E ] = 2 is (approximately) justified is semi-empirical (because
its form is a special case of Equation (120)), while its application to cases where the
assumption [ E ] = 2 is not justified must be considered as purely empirical. In any case,
contrary to the situation with the parameter 1 / C H occurring in the Hasselman relation
(111), cf. the discussion above, there are no principal objections against the interpretation of
C in terms of a critical volume fraction. In other words, even if taken as a purely empirical
fit equation, our relation (121) remains principally meaningful from the physical point of
view.
In principle, it can be attempted to interpret deviations of the intrinsic tensile modulus
determined by fitting with any of the relations above from the benchmark value 2 in terms of
an influence of pore shape. This is possible because usually the deviations in [E ] caused by a
variation of the matrix Poisson ratio are in practice negligible. Apart from the obvious advice
that such an interpretation should be based on a reliable determination of [E ] (e.g. by

comparing the coincidence of fitting results using several of the relations above) it should be
kept in mind, that significant deviations in [E ] require a considerable degree of pore
anisometry (elongation or flattening). Non-spherical pores which are more or less isometric
(e.g. of polyhedral shape or pores with concave faces) cannot be expected to be responsible
for measurable deviations in the intrinsic elastic moduli. Their intrinsic tensile modulus will
still be close to 2 ( [ E ] 2 ). Moreover, this discussion indicates that, due to the principal
non-uniqueness of the solution, such an interpretation will always remain on the level of a
tentative hypothesis. It can, of course, be useful for the relative comparison of sample series,
e.g. when only one processing parameter is changed at one time.
In concluding this section we would like to emphasize that of course all semi-empirical
relations presented for the relative tensile modulus can be used for fitting experimentally
measured data for any kind of elastic modulus (and many other properties as well). Also it
may be attempted to interpret the values obtained for the intrinsic properties by fitting in
terms of a pore shape influence. However, the intrinsic value of 2 in the case of porous
materials with spherical pores and 0 = 0.2 is specific to the tensile modulus (where
[ E ] = 2 exactly for 0 = 0.2 and 0 = 1 / 3 and [ E ] 2 for all matrix Poisson ratios
0 < 0 < 0.5 ) and can be a good approximation for the shear modulus for the matrix
Poisson ratios commonly encountered in practice, but certainly not for the bulk modulus or
other properties. That means, only for the tensile modulus (and approximately for the shear
modulus) the parameter-free special relations (110), (114) and (117) given above (specialized
by setting [ E ] = 2 in their more general counterpart relations), including the Kuster-Toksz
relation (112), can be expected to provide useful predictions. Relation (110), of course, is
principally disqualified for a completely different reason (viz. because it violates the upper
HS bound and even the Voigt bound, cf. the discussion above), while the Kuster-Toksz
relation represents no improvement over the HS upper bound (because it is identical to it).
Similarly, we recall that our new relation adopts its elegant and simple form (121) only in the
case [ E ] = 2 , otherwise its general form would be (120). This relation, however, is not per
se meant to be a predictive model (since the critical porosity C is usually not known a
priori), but a fit equation. As such, of course, it can be used quite universally.
The overview presented in this section is rather exhaustive, at least for isometric pores.
Of course, apart from the modulus-porosity relations mentioned here, more sophisticated
model relations have been proposed, e.g. for anisometric pores [Ondracek 1987, Boccaccini et
al. 1993] or remarkable models containing a size ratio information [Boccaccini & Fan 1997],
but these are beyond the scope of this chapter. Some well-known relations, e.g. in
[Ramakrishnan & Arunachalam 1993], are formally similar to the Hasselman relation (111)
and suffer from the same drawbacks, while others, e.g. in [Ishai & Cohen 1967], violate even
the HS upper bounds and should therefore be strictly avoided. It seems that all reasonable
non-exponential models intending to describe the porosity dependence of the tensile modulus
should contain the Coble-Kingery relation (117) as a special case (for the tensile modulus at
least this can be considered as a benchmark relation). Models for which this is not the case
should be considered with great scepticism. The interested reader can refer e.g. to [Bert 1985,
Herakovich & Baxter 1999, Kwan et al. 2000, Luo & Stevens 1999, Nielsen 1982, Rice 1998,
Wagh et al. 1991] and the references cited therein for some of the (less rational) modulus-
property relations not mentioned here.
4 Effective Elastic Moduli of Alumina, Zirconia

and Alumina-Zirconia Composite Ceramics
Modern alumina ceramics consist of substantially pure -Al2O3 (corundum), which is the
thermodynamically stable modification at room temperature and atmospheric pressure.
Corundum single crystals (monocrystals) are trigonal (rhombohedric), structurally
characterized by the point group 3 m (space group R3 m ), cf. e.g. [Lee & Rainforth 1994].
Zirconia ceramics are much more complicated. Apart from a less important orthorhombic
phase, there are at least three basic phases which can play a major role in zirconia ceramics:
cubic (c-ZrO2), tetragonal (t-ZrO2) and monoclinic (m-ZrO2). In the absence of stabilizing
agents the monoclinic phase is the stable one at room temperature and atmospheric pressure.
However, as a consequence of the martensitic t-m transformation which occurs during
cooling at approx. 950 C and is connected with a shear strain of approx. 0.16 and with a
volumetric expansion of approx. 4 %, sintered parts of pure zirconia cannot be obtained, since
cracks are formed and the ceramic part looses integrity. In order to obtain zirconia ceramics
that can be used in structural applications other oxides, so-called stabilizing agents, must be
added (usually already during powder synthesis), e.g. CaO, MgO, CeO2, Y2O3 and several
other oxides, mainly of the rare earth elements. Based on the three fundamental phases, the
stabilizing agents and their contents and the typical microstructures one distinguishes at least
three types of pure zirconia ceramics: fully stabilized zirconia (purely cubic phase, used
mainly in electrochemical applications), partially stabilized zirconia (PSZ, consisting of a
matrix of c-ZrO2 with dispersed inclusions or precipitates of t-ZrO2) and tetragonal zirconia
polycrystals (TZP). For detailed information the reader can consult one of the many excellent
review articles in this field, e.g. [Hannink et al. 2000] and the references cited therein. For a
compact overview cf. e.g. [Lee & Rainforth 1994]. If not explicitly stated otherwise, by
zirconia we mean the tetragonal phase of zirconia (t-ZrO2) with point group 4 / m m m
(space group P 4 2 / n m c ), or else tetragonal zirconia polycrystals (TZP), in this chapter.
4.1 Effective Elastic Moduli of Polycrystalline Alumina and Zirconia
The elastic behavior of trigonal and tetragonal monocrystals is fully described by 6

independent elastic constants (stiffnesses) or, equivalently, elastic coefficients (compliances).
The most reliable values currently available for alumina (-Al2O3) and zirconia (t-ZrO2, more
precisely t-ZrO2 with 12 mol% CeO2), which are due to [Wachtman et al. 1960] and [Kisi &
Howard 1998], respectively, are listed in Table 2. The alumina data have been measured by
X-ray diffraction directly for monocrystals, while the zirconia data have been inferred from
neutron diffraction measurements on polycrystalline samples and are in this sense biased, cf.
[Kisi & Howard 1998] and the discussion in [Pabst et al. 2004a]. Note that, due to the
unavailability of t-ZrO2 monocrystals of sufficient size, unbiased monocrystal data are
currently not available. The values listed in Table 2 are mutually consistent in the sense that
they fulfill the matrix inversion relations, Equations (9) and (10), and several other
plausibility criteria, cf. the critical comparison presented in [Pabst et al. 2004a].
Table 2. Components of the stiffness matrix C ij and the compliance matrix S ij for alumina
and tetragonal zirconia (t-ZrO2), respectively [Wachtman et al. 1960, Kisi & Howard 1998].
C ij [GPa] Alumina t-ZrO2 S ij [GPa-1] Alumina t-ZrO2

C11 496.8 1.8 327 S11 0.002353 0.00346
C33 498.1 1.4 264 S33 0.002170 0.00406
C44 147.4 0.2 59 S 44 0.006940 0.0170
C12 163.6 1.8 100 S12 0.000716 0.00096
C13 110.9 2.2 62 S13 0.000364 0.00059
C14 23.5 0.3 S14 0.000489
C66 (calc.) 166.6 64 S 66 (calc.) 0.003274 0.0154
Table 3. Components of the stiffness matrix C ij and the compliance matrix S ij for
monoclinic zirconia (m-ZrO2 ) according to [Nevitt et al. 1988] and [Chan et al. 1991],
respectively.
C ij [GPa] Nevitt et al. Chan et al. S ij [GPa-1] Nevitt et al. Chan et al.
C11 358 361 S11 0.00345 0.00341
C 22 426 408 S 22 0.00335 0.00503
C33 240 258 S 33 0.00537 0.00678
C44 99.1 99.9 S 44 0.0114 0.0104
C 55 78.7 81.2 S 55 0.0153 0.0147
C 66 130 126 S 66 0.00873 0.00828
C12 144 142 S12 0.00128 0.00160

C13 67 55 S13 0.000116 0.000608
C15 25.9 21.3 S15 0.00173 0.00165
C 23 127 196 S 23 0.00178 0.00371

C 25 38.3 31.2 S 25 0.00268 0.00319
C 35 23.3 18.2 S 35 0.00242 0.00310
C 46 38.8 22.7 S 46 0.00342 0.00188
In Tables 3, 4, 5 and 6 we list, for reasons of comparison and reference purposes, the
components of the stiffness and compliance matrix for monoclinic [Chan et al. 1991, Nevitt et
al. 1988] and cubic zirconia [Kandil et al. 1984, Kisi & Howard 1998], including the values
cited by [Kandil et al. 1984] for cubic zirconia (c-ZrO2) monocrystals in dependence of the
yttria content, cf. also [Green 1998].
Table 4. Components of the stiffness matrix C ij and the compliance matrix S ij for c-ZrO2
with 8.1 mol.% Y2O3, according to [Kandil et al. 1984] and [Kisi & Howard 1998].
Kisi & Kisi &

C ij [GPa] Kandil et al. S ij [GPa-1] Kandil et al.
Howard Howard
C11 402 390 S11 0.00274 0.0034
C44 56 60 S 44 0.0179 0.0168
C12 95 162 S12 0.00052 0.001
Table 5. Components of the stiffness matrix C ij for cubic zirconia (c-ZrO2) in dependence of
the yttria content (Y2O3 in mol.%), according to [Kandil et al. 1984, Green 1998].
Yttria content [mol.%]

C ij [GPa] 8.1 11.1 12.1 15.5 17.9
C11 401.8 403.5 405.1 397.6 390.4
C44 55.8 59.9 61.8 65.8 69.1
C12 95.2 102.4 105.3 108.6 110.8
Table 6. Components of the compliance matrix S ij for cubic zirconia (c-ZrO2) in dependence
of the yttria content (Y2O3 in mol.%), calculated from [Kandil et al. 1984, Green 1998].
Yttria content [mol.%]

-1 8.1 11.1 12.1 15.5 17.9
S ij [GPa ]
S11 0.002737 0.002762 0.002765 0.002849 0.002929
S 44 0.017921 0.016694 0.016181 0.015198 0.014472
S12 0.00052 0.00056 0.00057 0.00061 0.00065
Table 7 lists the Voigt bounds (subscript V), Reuss bounds (subscript R) and VRH
averages of the effective elastic moduli of polycrystalline alumina and tetragonal zirconia (t-
ZrO2), calculated from the components of the respective stiffness and compliance matrices
due to [Wachtman et al. 1960, Kisi & Howard 1998], listed in Table 2. For the effective
Poisson ratio, for which the Voigt-Reuss bounds do not hold, only a range of values is given,
cf. the discussion in [Pabst et al. 2004a].
In order to compare these theoretically calculated values with experimentally measured
or extrapolated values for the respective polycrystalline ceramics one has to take account of
the fact that real polycrystalline materials always contain a certain amount of porosity.
Moreover, experimentally determined values depend on the temperature during measurement

and the measurement technique (static or dynamic). While the latter seems to play only an
insignificant role in the alumina-zirconia system (i.e. the difference between isothermal and
adiabatic elastic moduli is negligibly small compared to other, statistical and systematic,
errors of measurement), special care has been taken in [Pabst et al. 2004a] to restrict the
comparison to experimental values determined at room temperature (25 3 C) for samples
with a total porosity < 3%. For alumina (with a theoretical density of 4.0 g/cm3) and
tetragonal zirconia (with a theoretical density of 6.1 g/cm3) the latter requirement corresponds
to bulk densities values of at least 3.88 g/cm3 and 5.92 g/cm3, respectively. A detailed
comparison, essentially based on [Munro 1997] and the NIST (National Institute of Standards
and Technology, Gaithersburg, Maryland, USA) databases [Munro 2002, NIST 2002] has
shown that for both alumina and zirconia (t-ZrO2 with 12 mol% CeO2) the agreement
between theoretically calculated and experimentally determined values is satisfactory [Pabst
et al. 2004a]. Note, however, that in the case of t-ZrO2 with 3 mol% Y2O3 the stiffness and
compliance matrix has not been determined so far and thus the effective elastic moduli for
this zirconia type must rely solely on experimentally measured values and the proximity to
the values of t-ZrO2 with 12 mol% CeO2. Table 8 summarizes the effective elastic moduli and
Poisson ratios that can be recommended as reference values for pore-free (i.e. densely
sintered), macroscopically isotropic polycrystalline alumina and tetragonal zirconia (t-ZrO2
with 3 mol% Y2O3 or 12 mol% CeO2, respectively). These values have been obtained after
careful consideration of theoretical calculated and experimentally determined values available
[Pabst et al. 2004a]. Note that, in contrast to many other sets of values occurring in the
literature, for each material presented here the ultimate value sets listed in Table 8 naturally
obey the standard elasticity relations, Equations (30) through (33).
Table 7. Effective elastic properties (elastic moduli and Poisson ratios) of polycrystalline
alumina and tetragonal zirconia (t-ZrO2), calculated from the components of the respective
stiffness and compliance matrices [Wachtman et al. 1960, Kisi & Howard 1998]; Voigt bound
(subscript V), Reuss bound (subscript R) and Voigt-Reuss-Hill average (subscript VRH).
Effective elastic property Alumina t-ZrO2

EV 408.2 209.9
ER 397.1 192.2
EVRH 402.7 201.1
GV 166.0 82.7
GR 160.7 74.7
GVRH 163.4 78.8
KV 251.4 151.8
KR 250.4 149.3
K VRH 250.9 150.6

(range) 0.2290.236 0.2700.285
Table 8. Recommended reference values for the effective elastic properties (elastic moduli
and Poisson ratios) of pore-free, macroscopically isotropic polycrystalline alumina and
tetragonal zirconia, according to [Pabst et al. 2004a].
t-ZrO2 with t-ZrO2 with

Effective elastic property Alumina
3 mol% Y2O3 12 mol% CeO2
E0 400 210 200
G0 163 80 76
K0 247 184 175
0 0.23 0.31 0.32
Fig. 1 shows the dependence of the effective tensile modulus of polycystalline zirconia
on the yttria content (VRH averages calculated from the monocrystal data for c-ZrO2 listed in
Tables 5 and 6 [Kandil et al. 1984, Green 1998], the value for 3 mol% Y2O3 being the value
recommended for t-ZrO2 in Table 8 [Pabst et al. 2004a]).
300
250
Tensile modulus [GPa]
200
150
100
50
0
0 2 4 6 8 10 12 14 16 18 20
Yttria content [mole %]
Figure 1. Dependence of the effective tensile modulus of zirconia (full squares: c-ZrO2, empty squares
t-ZrO2 ) on the yttria content.
According to these data the dependence of the effective tensile modulus E 0 [GPa] of
pore-free polycrystalline zirconia on the yttria content (mole fraction) cY can be fitted with
the linear expression
E 0 = 205 + 2.05 cY , (122)

(cf. the straight line in Fig. 1). Fig. 2 shows the temperature dependence of the effective
tensile modulus E 0 [GPa] of pore-free polycrystalline alumina and zirconia (m-ZrO2)
according to the measurements of Goto and Anderson [Goto & Anderson 1989] and the
values cited by Munro [Munro 2002], respectively.
450
400
350
Tensile modulus [GPa]
300
250
200
150
100
50
0
0 200 400 600 800 1000 1200 1400 1600 1800
Temperature [C]
Figure 2. Temperature dependence of the effective tensile modulus of alumina (empty squares, upper
curve) and zirconia (full squares, lower curve) [Goto & Anderson 1989, Munro 2002].
The solid curves have been obtained by fitting with the Wachtman-Anderson relation
[Anderson 1966, 1995, Wachtman et al. 1961, Wachtman 1996]
T
E 0 (T ) = E 0 T exp 0 ,

(123)
T

in which E 0 is the effective tensile modulus at absolute zero, T is the Kelvin temperature
and and T0 are empirical constants. As the temperature decreases to absolute zero the
tangent slope of the E 0 (T ) curve approaches zero, as required by the third law of
thermodynamics [Wachtman 1996]. For alumina the Wachtman-Anderson fit is
297.4
E 0 (T ) = 410 0.0583 T exp , (124 a)
T
while for zirconia the Wachtman-Anderson fit is

973.7
E 0 (T ) = 200 0.116 T exp . (124 b)
T
In concluding this subsection we remind the reader that from the viewpoint of solid state
physics the elastic moduli of crystalline solids are a consequence of bond strength and bond
density [Green 1998, Newnham 1975], cf. the introductory section of this chapter. Since the
stabilizing agents in zirconia are built into the crystal lattice they modify naturally the elastic
modulus (this may also be imagined as being a consequence of the different ionic radius of
the stabilizing agent in comparison to the ions making up the host lattice). It has to be
emphasized however, that a simple prediction of the elastic modulus e.g. of t-ZrO2 with 3
mol% Y2O3 by a simple mixture rule or the like is principally not possible, although the room
temperature elastic constants of pure polycrystalline yttria (density 5.0 g/cm3) are well known
( E 0 = 179 GPa, G0 = 69 GPa, K 0 = 149 GPa, 0 = 0.30, cf. [Munro 2002, Phani & Niyogi
1987b]). Note that micromechanical approaches can generally only be applied to multiphase
mixtures or materials with microstructure, but not to one-phase mixtures like solid solutions
(mixed crystals) or glasses, even if the constituent properties are known.
4.2 Effective Elastic Moduli of Dense Alumina-Zirconia Composite

Ceramics
First let us focus on two-phase (i.e. pore-free, densely sintered) mixtures of alumina and
tetragonal zirconia (t-ZrO2 with 3 mol.% Y2O3). We confine ourselves to classical
microcomposites, i.e. the grain size is assumed to be sufficiently large (> 100 nm) for the
phase boundary contribution to the volume fractions to be negligible. Moreover we confine
ourselves to isometric particles in random orientation resulting in a macroscopically isotropic
composite, to which the Voigt-Reuss bounds and the Hashin-Shtrikman bounds apply, and
use the (effective) elastic moduli G0 and K 0 recommended in Table 8 as an input
information for the following calculations. From now on let us consider the elastic properties
of the pure (i.e. single-phase) polycrystalline end members to be given once and for all and
reserve the term effective exclusively for the composite properties. In general (except for
the Reuss bound) the effective tensile moduli E have to be calculated from the effective
shear and bulk moduli G and K via the elasticity standard relation (30). Fig. 3 shows the
Voigt bound of the tensile modulus EV (as a function of the zirconia volume fraction Z )
calculated exactly from the Voigt values of the shear modulus GV = GV ( Z ) and the bulk
modulus K V = K V ( Z ) (crosses slightly above the upper solid line) and the approximate
Voigt bound EV

= EV ( Z ) calculated via the mixture rule, i.e. as the arithmetic average
EV E A = (1 Z ) E1 + Z E 2 , (125)
where E1 = 400 GPa (alumina) and E 2 = 210 GPa (zirconia). In Fig. 3 the latter is drawn as
the upper solid line.
500
Effective tensile modulus [GPa]

400
300
200
100
0
0 0.2 0.4 0.6 0.8 1
Zirconia volume fraction [1]
Figure 3. Effective tensile modulus of dense alumina-zirconia composite ceramics; Voigt bound
(crosses slightly above the upper solid line calculated from the Voigt values of the effective shear and
bulk moduli), approximate Voigt bound (according to the mixture rule, upper solid line), Reuss bound
(results of both calculations identical, crosses and lower solid curve), upper and lower Hashin-
Shtrikman bounds (dashed curves) and values measured by the resonant frequency method for dense
(porosity < 3 %) alumina-zirconia composite ceramics prepared by slip casting.
It is evident that in practice the difference between EV

( Z ) and EV ( Z ) is negligibly
small for the alumina-zirconia system (< 0.6 %), so that the mixture rule can well be used to
calculate the Voigt bound for the effective tensile modulus of alumina-zirconia composite
ceramics. The Reuss bound E R = E R ( Z ) calculated directly as a harmonic average via
Equation (80) is of course identical with that calculated indirectly from the Voigt values of
the shear modulus G R = G R ( Z ) and the bulk modulus K R = K R ( Z ) via the elasticity
standard relations (30). Apart from the Voigt and Reuss bounds Fig. 3 shows the upper and
lower Hashin-Shtrikman bounds (dashed curves in between the Voigt and Reuss bounds) and
values measured by the resonant frequency technique for dense (porosity < 3 %) alumina-
zirconia composite ceramics prepared by slip casting. It is evident that in the case of the
alumina-zirconia system the Hashin-Shtrikman bounds are sufficiently close together for their
arithmetic average to be an excellent prediction of the effective tensile moduli, that will be
precise enough for most practical purposes. The underlying reason is of course that the ratio
between the tensile moduli of the end members (alumina and zirconia) is only 2, a contrast
that must be considered as small from the viewpoint of micromechanics. The excellent
agreement of the HS prediction with the experimentally measured values fully confirms the
adequacy of the approach chosen and gives a feeling for the magnitude of errors (statistical
scatter) that can be expected in practice. Similar conclusions hold in principle for the effective
shear modulus and the effective bulk modulus. In order to fully exploit the usefulness of the
HS prediction in practical applications the HS averages in dependence of the zirconia volume

fraction Z have been fitted by second-order polynomials [Pabst et al. 2005]. This results in
the following handy formulae (with all terms in GPa):
E = 400 245.9 Z + 56.3 Z2 , (126)
G = 163 110.4 Z + 28.3 Z2 , (127)
K = 247 73.6 Z + 10.8 Z2 . (128)
For the calculation of the effective Poisson ratio the mixture rule is usually recommended.
In the case of alumina-zirconia composite ceramics it takes the form (dimensionless)
= 0.23 + 0.08 Z . (129)
The indirect calculation via the elasticity standard relation (33) using the HS averages of
E and G , Equations (126) and (127), results in effective Poisson ratios which can be
expressed as
= 0.23 + 0.0669 Z + 0.0111 Z2 . (130)
The difference between the two formulae is small (< 1.4 %) and affects only the third
decimal of the effective Poisson ratio: at zirconia volume fractions of approx. 50 % the linear
formula yields values which are higher by approx. 0.004. With respect to the difficulty of
experimental measurements of this difference should be negligible.
All the above formulae allow the direct and easy calculation of the effective elastic
moduli and Poisson ratios of dense alumina-zirconia composites for a given zirconia volume
fraction Z . Since usually the composition of alumina-zirconia composites (in the sequel
abbreviated as AZ composites) is given in weight percent (weight fractions wZ ) the
volume fractions Z have usually to be calculated via the (theoretical) density. In the
extreme case of AZ composites with only tetragonal zirconia (density 6.1 g/cm3) the effective
(theoretical) density is (with all terms in g/cm3)
0 (wZ ) = 4.0 + 1.41 wZ + 0.3 wZ2 + 0.4 wZ3 , (131)
while in the other extreme case of AZ composites with only monoclinic zirconia (density 5.6
g/cm3) the effective density is
0 (wZ ) = 4.0 + 1.10 wZ + 0.3 wZ2 + 0.2 wZ3 , (132)

cf. [Pabst et al. 2005]. These handy formulae can be used to calculate the theoretical density
of classical AZ microcomposites when the zirconia weight fraction is given. We emphasize,
however, that they may be inadequate to predict the theoretical density of AZ nanocomposites
(prepared e.g. via sol-gel techniques). Trivially, the corresponding dependences on the
zirconia volume fraction Z are given by simple linear mixture rules,
0 ( Z ) = 4.0 + 2.1 Z , (133)
0 ( Z ) = 4.0 + 1.6 Z (134)
for the extreme cases of AZ composites with only t-ZrO2 and only m-ZrO2, respectively.
With the help of these formulae it can easily be verified that under the assumption of 100 %
prevalence of t-ZrO2 the zirconia volume fraction Z is 0.105 for zirconia-toughened
alumina with a zirconia weight fraction wZ of 0.15 (in the sequel this special composition
with 15 wt.% of zirconia will be abbreviated ZTA) and 0.726 for alumina-containing
tetragonal zirconia polycrystals with a zirconia weight fraction wZ of 0.80 (in the sequel this
special composition with 80 wt.% of zirconia will be abbreviated ATZ). Both ZTA and
ATZ are typical and commercially successful types of AZ composites. When the
aforementioned formulae for the effective elastic moduli are applied to these ZTA and ATZ
composites one obtains E = 375 GPa, G = 151 GPa, K = 239 GPa for the ZTA composite
and E = 251 GPa, G = 97 GPa, K = 199 GPa for the ATZ composite.
4.3 Effective Elastic Moduli of Porous Ceramics in the Alumina-Zirconia

System
In the following let us reserve the term effective for the macroscopic properties of porous
materials, since, based on the findings of the preceding sections, it is justified to assume that
the elastic moduli of the pore-free (i.e. densely sintered) polycrystalline ceramics (alumina,
zirconia and alumina-zirconia composite ceramics) can be reliably predicted. Experimental
evidence confirms that the predictions are reliable and in good agreement with reality. In
particular, the Hashin-Shtrikman average yields a prediction for the elastic moduli of dense
AZ composites which should be sufficiently precise for most practical purposes. This success
of the prediction based on the HS average must be ascribed to the fact that the contrast
between the phase moduli of the end members (alumina and zirconia) is small from the
viewpoint of micromechanics, as mentioned above. The application of the simple HS bounds
in the form given by Equations (88) and (89) is of course restricted to macroscopically
isotropic two-phase composites. Although generalizations of these bounds to anisotropic
multiphase (number of phases > 2) composites exist [Torquato 2002, Walpole 1966a, 1966b,
1969], in many cases it is not necessary to complicate the calculations with such a
generalization.
Doubtlessly, porous AZ composites would strictly have to be modeled by a three-phase
model, particularly when interstitial porosity is considered, i.e. porosity that results from the
packing of particles during the shaping process [German 1989]. Naturally this interstitial
porosity, remnants of which may remain after partial sintering [De Jonghe & Rahaman 2003,
German 1996, Rahaman 1995, Rice 1998, 2003], is characterized by a typical pore size
smaller than (but of the same order of magnitude as) the typical grain size. On the other hand,
the usual way to produce ceramics with controlled porosity is not via partial sintering [Hardy
& Green 1995, Lam et al. 1994, Nanjangud et al. 1995] but via introducing artificial porosity
characterized by a pore size significantly larger than the grain size. This can principally be
done in at least four ways [Rice 1998, 2003]: first by replication of polymeric, especially
polyurethane or carbon-based, foams (which leaves after drying, burnout and firing highly
porous open-cell ceramic foams or networks with hollow struts), second by blowing agents
ensuring gas development during due to heating and / or chemical reactions (e.g. calcium
sulfate or gypsum, leaving CaO, which is a harmless or even benign additive in many zirconia
ceramics), third by ceramic balloons (hollow spheres), which are commercially available for
alumina and zirconia (primarily used for thermal insulation purposes) and fourth by
introducing pore-forming agents which burn out during firing (e.g. carbon / graphite and
various polymers of organic and bio-organic origin, like saw dust, crushed nut shells and
starch). The latter three result usually in porous materials with essentially closed pores (which
of course excludes their application for catalysis, filtering and implants designed for bone
tissue ingrowth), although possibly interconnected by narrow throats. In contrast to open-cell
ceramic foams [Gibson & Ashby 1997] materials with a closed-pore microstructure are more
corrosion-resistant, combine desirable mechanical properties with relatively light weight and
can be useful e.g. for thermal or electrical insulation purposes. However different the
processing details may be, all four techniques of porosity control just mentioned have one
feature in common: the porosity introduced in this artificial way is characterized by a pore
size which is generally at least 12 orders of magnitude larger (typically tens of micrometers)
than the grain size (typically around 1 micrometer). Pores of such a size, embedded in a fine-
grained matrix of host material, will not contribute to overall shrinkage during firing (even at
temperatures and in time schedules where the matrix becomes densely sintered, i.e. attains
almost theoretical density) and will remain as void inclusions in the final ceramic.
For such materials it is possible to consider the matrix (dense host material, e.g. alumina,
zirconia or AZ composites) as a homogenized medium and to invoke the Voigt bounds and
the Hashin-Shtrikman upper bounds in their simple form for two-phase materials. Of course,
in the case of porous materials (which can be considered as materials with an extremely,
substantially infinitely, high contrast in the phase properties from the viewpoint of
micromechanics) both the Reuss bound and the lower HS bound of the effective elastic
moduli degenerate to zero (in mathematical terms almost everywhere, i.e. everwhere except
at the singular point = 0 where they are equal to the matrix values M (0 ) = M 0 ), cf. Fig.
4 for the relative tensile modulus.
Note that the HS upper bound has been drawn as one single curve in Fig. 4,
corresponding to Equation (91), which is identical to the Mori-Tanaka prediction and the
Kuster-Toksz relation (112). Indeed, a comparison of the values obtained by this simple
+ + +
expression with the values obtained when E HS is calculated from G HS and K HS via
Equation (90) reveals that the difference is smaller than 0.1 %, i.e. absolutely negligible. Also
for the shear modulus the difference is small enough (< 2.6 %) to be negligible and thus
Equation (112) may certainly be used to replace the HS upper bound for alumina, zirconia
and AZ composites in practice, cf. Fig. 5. This is not the case for the bulk modulus, however,
cf. Fig. 6 (error up to 14.3 %).
0.9
Relative tensile modulus [1]
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Porosity [1]
Figure 4. Voigt bound (solid straight line) and Hashin-Shtrikman upper bound (solid curve) for the
relative tensile modulus of porous ceramics (e.g. alumina, zirconia or alumina-zirconia composites); the
Reuss bound and the lower Hashin-Shtrikman bound (dashed lines along the axes) degenerate to zero
for all porosities > 0.
0.9
Relative shear modulus [1]
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Porosity [1]
Figure 5. HS upper bounds for the relative shear modulus of alumina (lower curve) and zirconia (upper
curve); ZTA and ATZ composite ceramics are in between (not shown).
0.9
0.8
Relative bulk modulus [1] 0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Porosity [1]
Figure 6. HS upper bounds for the relative bulk modulus of alumina (upper solid curve), ZTA (upper
dashed curve), ATZ (lower dashed curve) and zirconia (lower solid curve).
The HS upper bound for the relative bulk modulus of alumina (with 0 = 0.23) and
zirconia (with 0 = 0.31) can be approximately expressed by the second-order polynomials
+
K HS ( 0 = 0.23) 1 1.78 + 0.78 2 , (135 a)
+
K HS ( 0 = 0.31) 1 2.00 + 1.00 2 , (135 b)
respectively, while for the tensile and shear modulus the corresponding expressions are much
closer to (i.e. almost identical with) Equation (91).
Now the question arises to what degree it is possible to predict the porosity dependence
of the elastic moduli, in particular of those elastic moduli (e.g. K ) which have not been or
cannot be easily measured for the material in question. This question, although of an immense
practical interest, cannot be answered in a unique way, because in principle microstructures
can be imagined that ensure finite moduli up to porosities close to 100 %, e.g. in the open-cell
ceramic foams mentioned earlier [Gibson & Ashby 1997], while on the other hand closed-
pore structures formed with blowing agents can exhibit the possibility of overlapping (i.e.
consist of penetrable pores, which can easily form large connected pore spaces), can loose
integrity (corresponding to zero elastic moduli) at very small porosities. With regard to this
fact the experimental measurement of the porosity dependence of the elastic moduli is in the
first instance an indispensable tool to gain experience with a certain type of material
microstructures. This experience may then, with due caution, be used to make certain
statements about or give rough estimates concerning the porosity dependence of the same or
other elastic moduli for materials with the same or a similar type of microstructure.
0.9
Relative tensile modulus [1]

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Porosity [1]
Figure 7. Relative tensile modulus of porous ceramics (measured values, predictions and master fit); HS
upper bound (thin solid curve), predictions for spherical pores (thin dotted: special case of the Spriggs
relation Eq. (110), thin dashed: Coble-Kingery Eq. (117), thick solid: modified exponential relation Eq.
(114)), experimentally measured values (squares: alumina, diamonds: ZTA, triangles: ATZ, circles:
zirconia, empty: potato starch as a pore-forming agent, full: corn starch as a pore-forming agent) and
master curve (thick dotted curve obtained by fitting with the Pabst-Gregorov relation, Eq. (121),
critical porosity 0.729).
Fig. 7 shows the porosity dependence of the tensile moduli experimentally measured (by
the resonant frequency technique) for alumina, zirconia and alumina-zirconia composites
(ZTA and ATZ) prepared from submicron oxide powders using starch as a pore-forming
agent (corn starch with a median size of approx. 14 micrometers and potato starch with a
median size of approx. 50 micrometers). The processing method is starch consolidation
casting, cf. the details given and references cited in [Pabst et al. 2004b, 2005].
Due to its organic character, the pore-forming agent (starch) is burnt out and vanishes
completely during heating up to the sintering temperature of the ceramics. The microstructure
is essentially of matrix-inclusion type: there are large pore bodies, interconnected by small
and narrow pore throats. The scatter of the measured values is relatively large, obviously
mainly due to errors in the exact determination of porosity (because the resonant frequency
technique used for measuring is rather precise, cf. Fig. 3, and there can be no doubt that also
the reference elastic moduli, used to calculate the relative moduli, are reliably known).
Nevertheless, within experimental scatter, there is a common trend for all six materials, as
expected. All values shown in Fig. 7 obey the HS upper bound. Evidently, the special case of
the Spriggs relation (110) is useless for prediction, because it violates even the HS bounds.
The Coble-Kingery relation is evidently better: it satisfies at least the HS bounds and is
clearly an improvement over the latter when a rough estimate is needed. However, when all
data are fitted with one master curve using the Archie relation (116) the resulting intrinsic
tensile modulus obtained from fitting is [ E ] = 2.51, cf. Table 9. Similarly, when the Phani-
Niyogi relation (119) is used for fitting the intrinsic tensile modulus is [E ] = 2.50 (the
exponent [ E ] C being 2.45 with the critical volume fraction being close to unity, i.e. C =
0.979), cf. Table 9. This indicates, that when the exponent 2 in the Coble-Kingery relation
(117) is allowed to vary freely (as an adjustable fit parameter), the fitted curve corresponds to
a value significantly larger than 2, and is clear evidence of the fact that the Coble-Kingery
prediction is not satisfactory in this case.
Table 9. Fit parameters determined for the master curve of the porosity dependence of porous
ceramics (alumina, zirconia and AZ composites) with matrix-inclusion type microstructure
prepared by starch consolidation casting with starch as a pore-forming agent.
Fit model [E ] C
Modified [E] 2.09
Exponential E r = exp
1
Mooney-type 2 1.006
with [E ] = 2 E r = exp
1 C
Mooney-type [E] 2.48 1.737
E r = exp
1 C
Archie E r = (1 ) 2.51
[E]
E r = (1 C )
Phani-Niyogi [ E ] C 2.50 0.979
Pabst-Gregorov E r = (1 ) (1 C ) 0.729
By far the best prediction is achieved with the modified exponential relation (114). We
emphasize that this is, similar to the HS upper bound and the Coble-Kingery prediction, an
unbiased a priori prediction without the need for fitting or input parameters of any kind. It is
solely based on the assumption that the pores are spherical or isometric. In this case, if the
intrinsic tensile modulus is allowed to vary, i.e. if we consider [E ] as an adjustable
parameter to be determined by fitting according to relation (113), the result is [ E ] = 2.09,
which is a value very close to 2. Similarly, when a critical porosity is introduced as a fit
parameter, i.e. the Mooney-type relation (115) is used with [E ] = 2, this critical porosity
turns out to be C = 1.006, i.e. C 1. Both findings confirm the quality of the prediction
via the modified exponential relation (114). Note, however, that the two-parameter fit via the
general Mooney-type relation (115) can lead to physically unreasonable results (typically C
> 1), cf. Table 9 and [Pabst et al. 2004b, 2005].
In concluding this section we address the most intricate question: assuming the porosity
dependence of one effective elastic modulus (say E ) to be known and analyzed (e.g. by
fitting), can an approximate estimate be given for the porosity dependence of another
effective elastic modulus (say K ) which has not been measured ? To be concrete, our task is
the following: Given the porosity dependence of the tensile modulus as
2
E r = exp (114)
1
or, slightly more precisely, as

E r = (1 ) 1 , (136)
0.729
what is the porosity dependence of the effective bulk modulus ? A first guideline to the
tentative answer is the fact that for porous materials with spherical pores and a matrix Poisson
ratio of 0 = 0.2 the intrinsic elastic modulus should be equal to 2 for all moduli, i.e. also
[ K ] = 2, cf. Equations (102) and (105). Therefore, for such materials also the porosity
dependence of the bulk modulus should be given by the modified exponential relation
2
K r ( 0 = 0.2) = exp . (137)
1
Comparison with the approximate expressions (135 a) and (135 b) shows, that this
relation does not violate the HS upper bound even for zirconia, as can be seen from the initial
tangent slope of this relation (which is equal to 2 at = 0), although of course relation
(137) must be expected to be a better approximation for alumina (with 0 = 0.23) than for
zirconia (with 0 = 0.31). However, as mentioned several times before, the intrinsic bulk
modulus [K ] is a much more sensitive function of the matrix Poisson ratio 0 than the
intrinsic tensile modulus [E ] . For alumina and zirconia the intrinsic bulk modulus [K ]
should be 2.139 and 2.724, respectively, cf. Table 1. Therefore it might be justified to replace
the modified exponential relation (137) by
2.139
K r ( 0 = 0.23) = exp , (138 a)
1
2.724
K r ( 0 = 0.31) = exp (138 b)
1
in the case of alumina and zirconia, respectively. But the problem may be tackled from
another side as well: if a reliable estimate can be given for the porosity dependence of the
effective Poisson ratio , then it should be possible to calculate the effective bulk modulus
K from E and via the elasticity standard relation (32). In order to estimate the porosity
dependence of the effective Poisson ratio of alumina and zirconia we invoke the asymptotic
values for porosities close to 100 %, given by Equations (83) and (84). According to the

differential approach this value is invariably 0.2 (independent of the matrix Poisson ratio),
while according to the Mori-Tanaka approach, Equation (84), the asymptotic value of is

0.21 for alumina and 0.24 for zirconia. This leads to the following estimates for the porosity
dependence of the effective Poisson ratios (with = 0.03 and 0.02 for alumina and = 0.11
and 0.07 for zirconia according to the differential and the Mori-Tanaka approach,
respectively),
=0 . (139)
The various predictions for the porosity dependence of the relative bulk modulus K of
alumina and zirconia are compared in Fig. 8. Alumina-zirconia composite ceramics of any
composition will show intermediate behavior.
0.9
Relative bulk modulus [1]
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Porosity [1]
Figure 8. Relative bulk modulus of porous alumina (red) and zirconia (blue); modified exponential
relation for spherical pores and matrix Poisson ratio 0.2 (green solid curve), predictions for alumina
with intrinsic bulk modulus 2.139 (red solid curve) and for zirconia with intrinsic bulk modulus 2.724
(blue solid curve), dashed and dotted curves are relative bulk moduli calculated via elasticity standard
relations using the differential and Mori-Tanaka prediction, respectively.
5 Outlook
In this chapter we have treated the effective elastic moduli of alumina, zirconia and alumina-
zirconia composite ceramics. We have invoked the theory of linear elasticity [Gurtin 1972],
which is the appropriate framework for the macroscopic mechanical behavior of alumina and
zirconia at room temperature and above (up to a few hundred C, say). Polycrystalline
ceramics in general can be considered as completely brittle at these temperatures, because
dislocation motion is impeded by the absence of appropriate slip systems (in contrast to single
crystals) and the fact that grain boundary sliding, the principal mechanism of plastic
deformation and creep in polycrystalline ceramics, leads to crack nucleation and opening
[Davidge 1979, Green 1998, Munz & Fett 1999, Wachtman 1996]. It is a common experience
that at these temperatures, when a critical stress value (the strength) is exceeded,
polycrystalline ceramics respond by brittle fracture long before inelastic modes of behavior
can operate and even before elasticity can become nonlinear. This is also the deeper reason
why linear fracture mechanics [Irwin 1958] can be readily applied to these materials, why the
mechanical strength shows in dependence of accidental microstructural defects
considerable scatter (more than e.g. the strength values of metals or polymers) and why
Weibull statistics (with its weakest link assumption) [Weibull 1939, 1947, 1951] is ideally
suited to describe this scatter. However, at higher temperatures (still well below the sintering
temperature) both alumina and zirconia exhibit plastic deformation and creep, and especially
zirconia is well known for its so-called superplastic behavior [Nieh et al. 1997].
Based on rational thermomechanics [Truesdell & Toupin 1960, Truesdell 1984, Truesdell
& Noll 2003] it is possible today to extend the theory of linear elasticity presented in this
chapter in a rigorous and well-defined way to linear thermoelasticity [Carlson 1972, ilhav
1997] and to use this theory as the basis for a detailed investigation of thermo-mechanical
properties (i.e. apart from the elastic constants thermal conductivity, specific heats and
coefficient of thermal expansion). For alumina, zirconia and alumina-zirconia composite
ceramics this working programme has been launched by the authors recently. Similarly, based
on [Eringen & Maugin 1989, 1990], an extension of this research programme to electrical,
magnetic and electromagnetic (including optical) properties is thinkable and can be
performed in the future. It goes without saying that plastic deformation and high-temperature
creep of alumina and zirconia are major challenges to mathematical modeling, because the
appropriate constitutive equations are in general nonlinear, but in principle the basic
apparatus for solving this task in a rational way is available in the literature on rational
thermomechanics and theory of materials [Haupt 2000, Krawietz 1986]. Rational theory,
originating from the work of Gibbs [Gibbs 1961], can also tackle in a rigorous way with
phase boundaries and phase transformations [Grinfeld 1991, ilhav 1997], the latter of
which would be of primary importance in the case of zirconia. This is what can be modeled
from the viewpoint of rational theory without invoking mixture theories. Note that chemical
interactions like stress-corrosion cracking and subcritical grain growth of alumina (e.g. in the
presence of water vapor), closely related to fatigue, as well as hydrothermal ageing (low-
temperature surface degradation) of zirconia (also in the presence of water vapor, e.g. at
elevated temperatures in the range 150-250 C), including the accompanying diffusion
phenomena (e.g. yttria diffusion inside and between the zirconia grains) and surface chemical
reactions (e.g. with oxygen bonds linking to OH groups from the ambient atmosphere) can
principally only be attacked by mixture theories. So far, rational mixture theories have not
been applied to these complex phenomena; they are commonly modeled by more traditional
concepts (equilibrium thermodynamics and kinetics).
We would like to emphasize at this point that strength, fracture toughness, hardness and
wear behavior, although important mechanical parameters for materials selection and
engineering design, are not well-defined material properties, because they are intimately
related to (and physically meaningful only for) a certain (and often very special) test
geometry and procedure and (what is worse) specimen and probe geometry. The same holds
true for several other quantities, e.g. thermal shock parameters. All these parameters arise in
the context of very specific model situations and should be used only for the purpose of
relative comparison (i.e. in order to compare the behavior of other materials in the same
situation as the material tested). For a refreshing account of brittle fracture, largely untouched
by engineering tradition but inspired by percolation theory, the reader may consult [Sahimi
2003b]. Of course, during the second half of the 20th century the engineering literature on
these properties for alumina, zirconia and alumina-zirconia composite ceramics has grown
immensely.
Elastic properties, without doubt, are the perfect playground for micromechanics.
Together with viscosity they belong to a class of properties which arises in the context of
class B problems in the sense of Torquato [Torquato 2002] and are in the general case
characterized by fourth-order tensors. Hookes law is the constitutive equation of linearly
elastic solids in the same way as Newtons law is the constitutive equation of linearly viscous
fluids. Note that in general, i.e. in the case of anisotropic fluids, also viscosity is a fourth-
order tensor, similar to the stiffness tensor used in this chapter, cf. e.g. [ilhav 1997,
Truesdell 1984] (in the case of isotropic fluids we obtain again two constants, e.g. the shear
viscosity and the bulk viscosity, in complete analogy with the two elastic constants needed for
isotropic solids, cf. e.g. [Billington & Tate 1981]). Other properties (related to Torquatos
class A problems) are generally described by linear constitutive equations involving
second-order tensors, e.g. thermal conductivity (Fouriers law), electrical conductivity
(Ohms law), dielectric constant and magnetic permeability, which in the case of isotropic
materials reduce to only one constant. The coefficient of thermal expansion is a similar
property (symmetric second-order tensor, i.e. three constants in general and scalar, i.e. one
constant, for isotropic materials).
Of course, the micromechanical relations (bounds, approximations or fit models)
presented in this chapter for the effective elastic properties are by no means restricted to the
alumina-zirconia system but can be applied to many types of ceramics and ceramic
composites. On the other hand they cannot be expected to be automatically applicable to
matrix-inclusion type composites in cases where the matrix consists of nonlinearly elastic
materials (polymers), viscoelastic materials (glasses or porcelain at high temperature) or
elastoplastic materials (metals). In particular, they cannot be a priori expected to be justified
for materials of biological origin, although their application to many of these materials, e.g.
bone, might be seductive and dictated by practical needs. With respect to the inherent
anisotropy and the hierarchical microstructure of these materials [Ontan et al. 2000],
however, any mathematical modeling or description of their composition-structure-property
relationships has to be performed with due caution.
All micromechanical relations presented in this chapter are based on the volume fractions
of the constituent phases (i.e. one-point correlation functions and a special part of the two-
point correlation functions) as the only input information. If complete two-point and three-
point information would be available (e.g. from stereological analysis or 3D tomography), the
Hashin-Shtrikman bounds (i.e. the two-point bounds) could be improved in principle, which
leads to the three-point bounds due to [Beran & Molyneux 1966] and [Milton & Phan-Thien
1982], cf. also [Jeulin 2001, 2002, Markov 2000, Milton 2002, Sobczyk & Kirkner 2001,
Torquato 2000, 2002]. Of course, in the case of dense alumina-zirconia composites it is quite
evident, that (due to the relatively small contrast in the phase properties) already the Hashin-
Shtrikman bounds are precise enough to permit property prediction to a sufficient degree. In
the case of porous materials, however, the three-point bounds could in principle improve the
upper bound.
Throughout this chapter macroscopic isotropy was assumed. In the case of anisometric
constituent phases (inclusions), e.g. alumina platelets (tabular alumina) or fibers (e.g. the so-
called Saffil fibers), this assumption implies of course statistical random orientation. In the
case of preferred orientation of anisometric inclusions (e.g. flow-induced fiber or platelet
alignment) the resulting composites are usually anisotropic (e.g. transversely isotropic). The
generalization of the Hashin-Sthrikman bounds to anisotropic materials is due to Willis
[Willis 1977], and exact series expansions for anisotropic composites have been given by
Torquato [Torquato 1997, 1998]. Anisometric inclusions and voids play a major role in
partially stabilized zirconia (PSZ) ceramics, viz. first as (oriented) lenticular lamellae of t-
ZrO2 in a matrix of c-ZrO2 and second as microcracks around those lamellae which have
transformed to m-ZrO2. Both can be modeled as oblate spheroids or flat disks, the first of
finite thickness, the latter infinitely thin (in this exceptional limiting case the volume fraction
has to be replaced by the so-called crack-density parameter). Explicit expressions for
spheroidal (disk-like and needle-like) inclusions and voids are given by [Markov 2000,
Milton 2002, Torquato 2002].
The authors current work in progress concerns the application of analogous
micromechanical relations (bounds, approximations and fit models) to the thermal
conductivity and the thermal expansion coefficient of alumina, zirconia and alumina-zirconia
composites, both dense and porous and the subsequent description of experimentally
determined microstructure-property relationships, i.e. effective properties and their
dependence on the zirconia volume fraction and porosity, respectively. In complete analogy
to the elasticity context of this chapter, it is possible to derive a Coble-Kingery-type relation
for the effective thermal conductivity [Pabst 2005] of porous ceramics and, based on this, a
Pabst-Gregorov-type relation allowing for a critical porosity (percolation threshold) [Pabst
& Gregorov 2005]. Based on the effective bulk modulus predicted in this chapter it will also
be possible to calculate the effective thermal expansion coefficient of alumina-zirconia
composites via the Levin formula [Levin 1967], a so-called cross-property relation. In a
similar spirit, cross-property bounds have been constructed between bulk or shear modulus
and thermal conductivity [Berryman & Milton 1988, Gibiansky & Torquato 1996, Sahimi
2003a, Torquato 1992, 2002].
Note that the micromechanical approach chosen throughout this chapter can be expected
to be realistic only for classical microcomposites with particle sizes not much smaller than
100 nm, i.e. in cases where the phase boundaries (interfaces) can be considered and modeled
as singular discontinuities [Grinfeld 1991, ilhav 1997]. In other words, the phase
boundaries are considered as sharp interfaces between two bulk phases, without structure and
properties and without contribution to the volume fraction. In the framework of rational
mixture theories this corresponds to the case of so-called multiphase mixture theories, cf. e.g.
[Dobran 1991, Passman et al. 1984]. Such a conception will certainly not be realistic for
nanocrystalline materials and nanocomposites, where the volume fraction of atoms in the
grain boundaries can be as high as 30 vol.% in the case of 10 nm grains, compared to only 3
vol.% for 100 nm grains, cf. e.g. [Ajayan 2003, Gleiter 1989, Niihara 1991, Suryanarayana &
Koch 1999, Winterer 2002]. We emphasize, however, that the apparent grain size dependence
of elastic properties which can develop as a consequence of the growing importance of phase
boundaries has nothing to do with the ubiquitous and well-known influence of grain size on
strength and other mechanical properties. All available experience indicates that the influence
of grain size on the elastic properties is negligible unless the grain size becomes extremely
small (< 10 nm, say) [Krstic et al. 1993, Poole & Owens 2003].
From an atomistic point of view, phase boundaries (e.g. grain boundaries or surfaces in
general) are always characterized by a higher concentration of defects (e.g. due to curvature)
or a lower degree of crystallinity in comparison to the adjacent bulk phase. Obviously, in
contrast to the micromechanical conception of a sharp interface, in reality there must be a
gradient in structure (degree of crystallinity) and properties which is more adequately
modeled by considering the interfaces as diffuse. The effective thickness of such a diffuse
interface can be expected to be of the order of 1 nm up to a few nm. In nanocystalline
materials of two constituents the interface thus plays the role of a third phase (i.e. with a finite
volume fraction) but with diffuse structure and undefined properties, which are usually not
accessible to measurement. Although it might be seductive for such cases to invoke the so-
called mesophase concept [Theocaris 1987] which treats the phase boundary as a separate
phase, this and similar models occurring in the micromechanical literature are complicated
without really solving anything, since the properties of the third (meso-) phase are a priori
unknown. This situation is somewhat similar to the situation in rational mixture theories
[Samohl 1999, 1997, Truesdell 1984, Truesdell & Toupin 1960], where the partial quantities
are introduced as primitive concepts and only the mixture quantities but not the partial
quantities (which are of course different from the constituent quantities before mixing or
outside the mixture) are accessible to measurement [Samohl & ilhav 1990], with one
remarkable exception: the chemical potential (i.e. the specific Gibbs free energy). Note that in
the case of solids the chemical potential is not a scalar quantity but a second-order tensor
(Eshelby tensor), cf. [Samohl & Pabst 1997, 2004].
Glass scientists and engineers are very familiar with this situation as well: usually the
effective (e.g. elastic) properties of glasses cannot sensu strictu be predicted a priori from the
constituent properties but have in principle always to be determined by measurement. The
underlying deeper reason is the absence of microstructure or, more precisely, the fact that
microstructure is lost in the smeared-out phase boundaries, which represent in a certain sense
a new phase and make up the whole volume of the glass body (amorphous material). From
this point of view nanocrystalline materials and nanocomposites can be considered as
intermediate between classical (micro-) composites (which can be called immiscible
mixtures) and glasses (which can be called miscible mixtures or solid solutions). In the
first case (heterogeneous materials), volume fractions of the constituent phases are the
appropriate concentration measures and the effective properties can be predicted when the
properties of the constituent bulk phases (i.e. the components before mixing or outside the
mixture) and the microstructural details are known. In the latter case (homogeneous
materials), weight (mass) or molar fractions are the appropriate concentration measures (since
volume fractions are undefined), microstructural details are irrelevant (homogeneous
materials do not have microstructure in this sense) and the effective properties cannot be
predicted from the properties of the components. This statement holds true for elastic
properties as well as other well-defined properties (e.g. thermal conductivity and thermal
expansion coefficient). In glass science and technology a pragmatic empirical solution is
usually preferred: from a large amount of data, experimentally measured for glasses of
various composition (mixture properties), one can deduce partial properties, which can (a
posteriori) be used for tentatively predicting (interpolating or extrapolating or at least
approximately estimating) the effective properties of yet unmeasured glass compositions with
the help of empirical relations, cf. the rich literature on this subject cited in [Volf 1988].
A first step to rationalize the modeling of nanomaterials could be the elaboration of a
theoretically sound model of phase boundaries and phase transitions based on modern
continuum thermodynamics [Grinfeld 1991, ilhav 1997]. A second step towards the
prediction of effective properties of nanomaterials would then be the elaboration of a rational
mixture theory for solids in analogy to the so-called chemical thermodynamics for fluids,
which is essentially based on Gibbs key paper [Gibbs 1961], cf e.g. [Guggenheim 1957,
Samohl 1982, 1999]. Partial quantities will then be available for the constituents of solid
mixtures in a similar way as the molar partial quantities available for fluid mixtures (i.e. gas
mixtures and liquid solutions). The chemical potential tensor [Samohl & Pabst 1997, 2004]
might play a key role in this step.
Acknowledgement:
This work was part of the frame research programme Preparation and Research of
Functional Materials and Material Technologies using Micro- and Nanoscopic Methods,
supported by the Ministry of Education, Youth and Sports of the Czech Republic (Grant No.
MSM 6046137302). It is based on fundamental research carried out within the project
Mechanics and Thermomechanics of Disperse Systems, Porous Materials and Composites,
supported by the Czech Science Foundation (Grant No. 106/00/D086). The authors would
like to thank Dr. M. ern (Prague) for Youngs modulus measurements and Prof. D. Jeulin
(Fontainebleau), Prof. I. Samohl (Prague), Prof. S. Torquato (Princeton) and Prof. R. W.
Zimmerman (London) for valuable hints and their encouraging interest in our work.
References
Ajayan, P. M. In Nanocomposite Science and Technology; Ajayan, P. M.; Schadler, L. S.;
Braun, P. V.; Eds.; Wiley-VCH: Weinheim, Germany 2003; pp 1-75.
Aksel, C.; Riley, F. L. J. Eur. Ceram. Soc. 2003, 23, 3089-3096.
Almgren, R. J. Elast. 1985, 15, 427-430.
Anderson, O. L. Phys. Rev. 1966, 144, 553-557.
Anderson, O. L. Equations of State of Solids for Geophysics and Ceramic Science; Oxford
University Press: New York, NY, 1995.
Archie, G. E. Trans. AIME 1942, 146, 54-62.
Ashcroft, N. W.; Mermin, N. D. Solid State Physics; Saunders College: Philadelphia, NJ,
1976; pp 421-509.
Asmani, M.; Kermel, C.; Leriche, A.; Ourak, M. J. Eur. Ceram. Soc. 2001, 21, 1081-1086.
Aste, T.; Weaire, D. The Pursuit of Perfect Packing; Institute of Physics Publishing: Bristol,
UK, 2000; pp 20-34.
Atanackovic, T. M.; Guran, A. Theory of Elasticity for Scientists and Engineers; Birkhuser:
Boston, MA, 2000; pp 85-111.
Atkinson, A.; Seluk, A. Solid State Ionics 2000, 134, 59-66.

Bathurst, R. J.; Rothenburg,, L. Int. J. Eng. Sci. 1988, 26, 373-383.
Becher, P. F.; Rose, L. R. F. In Structure and Properties of Ceramics; Swain, M. V.; Ed.;
Materials Science and Technology A Comprehensive Treatment; VCH: Weinheim,
Germany, 1994; Vol. 11, pp 409-461.
Beran, M. J. Statistical Continuum Theories; Interscience Publishers / Wiley: New York, NY,
1968; pp 18-49, 181-256.
Beran, M. J.; Molyneux, J. Quart. Appl. Math. 1966, 24, 107-118.
Bernal, J. D. Nature 1959, 183, 141-147.
Bernal, J. D.; Mason, J. Nature 1960, 185, 910-911.
Berryman, J. G.; Milton, G. W. J. Phys. D: Appl. Phys. 1988, 21, 87-94.
Bert, C. W. J. Mater. Sci. 1985, 20, 2220.
Billington, E. W.; Tate, A. The Physics of Deformation and Flow; McGraw-Hill: New York,
NY, 1981; pp 129-193, 547-609.
Boccaccini, A. R.; Ondracek, G.; Mazilu, P.; Windelberg, D. J. Mech. Behaviour Mater. 1993,
4, 119-128.
Boccaccini, A. R.; Fan, Z. Ceramics International 1997, 23, 239-245.
Born, M.; Huang, K. Dynamical Theory of Crystal Lattices; Clarendon Press: Oxford, UK,
1954.
Boutin, P.; Christel, P.; Dorlot, J. M.; Meunier, A.; de Roquancourt, A.; Blanquaert, D.;
Herman, S.; Sedel, L.; Witvoet, J. J. Biomed. Mater. Res. 1988, 22, 1203-1232.
Brezny, R.; Green, D. J. In Structure and Properties of Ceramics; Swain, M. V.; Ed.;
Germany, 1994; Vol. 11, pp 463-515.
Bruggeman, D. A. G. Ann. Physik Leipzig 1937, 29, 160-178.
Budiansky, B. J. Mech. Phys. Solids 1965, 13, 223-227.
Caddock, B. D.; Evans, K. E. J. Phys. D: Appl. Phys. 1989, 22, 1877-1882.
Callen, H. B. Thermodynamics; Wiley: New York, NY, 1960; pp 213-237.
Carlson, D. E. In Encyclopedia of Physics; Flgge, S.; Ed.; Springer: New York, NY, 1972;
Vol. VIa/2, pp 297-346.
Chan, S.-K.; Fang, Y.; Grimsditch, M.; Li, Z.; Nevitt, M. V.; Robertson, W. M.; Zouboulis, E.
S. J. Am. Ceram. Soc. 1991, 74, 1742-1744.
Cawley, J. D.; Lee, W. E. In Structure and Properties of Ceramics; Swain, M. V.; Ed.;
Germany, 1994; Vol. 11, pp 85-117.
Chevalier, J.; Drouin, J. M.; Cales, B. Bioceramics 1997, 10, 135-138.
Chevalier, J.; Cales, B.; Drouin, J. M. J. Am. Ceram. Soc. 1999, 82, 2150-2154.
Chevalier, J.; Deville, S.; Mnch, E.; Jullian, R.; Lair, F. Biomaterials 2004, 25, 5539-5545.
Choi, J. B.; Lakes, R. S. J. Mater. Sci. 1992, 27, 5375-5381.
Choi, K.; Kuhn, J. L.; Ciarelli, M. J.; Goldstein, S. A. J. Biomech. 1990, 23, 1103-1113.
Christel, P.; Meunier, A.; Heller, M.; Torre, J. P.; Peille, C. N. J. Biomed. Mater. Res. 1989,
23, 45-61.
Christensen, R. M. Mechanics of Composite Materials; Wiley: New York, NY, 1979; pp 1-
348.
Christensen, R. M. J. Mech. Phys. Solids 1986, 34, 563-578.
Christensen, R. M. Int. J. Solids Struct. 2000, 37, 93-104.
Coble, R. L.; Kingery, W. D. J. Am. Ceram. Soc. 1956, 39, 377-385.

Conway, J. C. J. Mater. Sci. 1986, 21, 2525-2527.
Cook, R. F.; Pharr, G. M. In Structure and Properties of Ceramics; Swain, M. V.; Ed.;
Germany, 1994; Vol. 11, pp 339-407.
Davidge, R. W. Mechanical Behaviour of Ceramics; Cambridge University Press: Cambridge,
UK, 1979; pp 1-165.
De Jonghe, L. C.; Rahaman, M. N. In Handbook of Advanced Ceramics Vol. I: Materials
Science; Somiya, S.; Aldinger, F.; Claussen, N.; Spriggs, R. M.; Uchino, K.; Koumoto,
K.; Kaneno, M. Elsevier Academic Press: Amsterdam, The Netherlands, 2003; pp 187-
264.
Dewey, J. M. J. Appl. Phys. 1947, 18, 578-586.
Dobran, F. Theory of Structured Multiphase Mixtures; Springer: Berlin, Germany, 1991; pp
1-207.
Eduljee, R. F.; McCullough, R. L. In Structure and Properties of Composites; Chou, T. W.;
Ed.; Materials Science and Technology A Comprehensive Treatment; VCH: Weinheim,
Germany, 1993; Vol. 13, pp 381-474.
Einstein, A. Ann. Phys. 1906, 19, 289-306.
Eringen, A. C.; Maugin, G. Electrodynamics of Continua I; Springer: New York, NY, 1989.
Eringen, A. C.; Maugin, G. Electrodynamics of Continua II; Springer: New York, NY, 1990.
Eshelby, J. D. Proc. Roy. Soc. London A 1957, 241, 376-396.
Evans, K. E. J. Phys. D: Appl. Phys. 1989, 22, 1870-1876.
Evans, K. E.; Caddock, B. D. J. Phys. D: Appl. Phys. 1989, 22, 1883-1887.
Evans, K. E.; Nkansah, M. A.; Hutchinson, I. J. Acta Metall. Mater. 1992, 40, 2463-2469.
Fantozzi, G.; Olagnon, C. In Structure and Properties of Composites; Chou, T. W.; Ed.;
Germany, 1993; Vol. 13, pp 183-238.
Fleig, J.; Kreuer, K. D.; Maier, J. In Handbook of Advanced Ceramics Vol. I: Materials
Science; Somiya, S.; Aldinger, F.; Claussen, N.; Spriggs, R. M.; Uchino, K.; Koumoto,
K.; Kaneno, M. Elsevier Academic Press: Amsterdam, The Netherlands, 2003; pp 59-105.
Forte, S.; Vianello, M. J. Elasticity 1996, 43, 81.
Friis, E. A.; Lakes, R. S.; Park, J. B. J. Mater. Sci. 1988, 23, 4406-4414.
Fukuhara, M. In Handbook of Advanced Ceramics Vol. I: Materials Science; Somiya, S.;
Aldinger, F.; Claussen, N.; Spriggs, R. M.; Uchino, K.; Koumoto, K.; Kaneno, M.
Elsevier Academic Press: Amsterdam, The Netherlands, 2003; pp 333-346.
German, R. M. Particle Packing Characteristics; Metal Powder Industries Federation: New
York, NY, 1989; pp 23-308.
German, R. M. Sintering Theory and Practice; Wiley: New York, NY, 1996; pp 142-207.
Gibbs, J. W. The Scientific Papers of J. W. Gibbs Vol. I: Thermodynamics; Dover: New
York, NY, 1961; pp 55-353.
Gibiansky, L. V.; Torquato, S. Proc. Roy. Soc. London A 1996, 452, 253-283.
Gibson, L. J.; Ashby, M. F. Proc. Roy. Soc. London A 1982, 382, 43-59.
Gibson, L. J.; Ashby, M. F.; Schajer, G. S.; Robertson, C. I. Proc. Roy. Soc. London A 1982,
382, 25.
Gibson, L. J.; Ashby, M. F. Cellular Solids Structure and Properties; Cambridge University
Press: Cambridge, UK, 1997 (second edition); pp 175-234.
Gleiter, H. Prog. Mater. Sci. 1989, 33, 223.

Goodier, J. N. Trans. ASME 1933, 55, 39.
Goto, T.; Anderson, O. L. J. Geophys. Res. 1989, 94, 7588-7602.
Green, D. J. An Introduction to the Mechanical Properties of Ceramics; Cambridge
University Press: Cambridge, UK, 1998; pp 1-336.
Green, D. J.; Hannink, R. H. J.; Swain, M. V. Transformation Toughening in Ceramics; CRC
Press: Boca Raton, FL, 1989.
Gremillard, L.; Chevalier, J.; Epicier, T.; Deville, S.; Fantozzi, G. J. Eur. Ceram. Soc. 2004,
24, in press.
Griffith, A. A. Phil. Trans. Roy. Soc. London A 1920, 221, 163-198.
Grinfeld, M. Thermodynamic Methods in the Theory of Heterogeneous Systems; Longman
Scientific and Technical: New York, NY, 1991; pp 1-399.
Guggenheim, E. A. Thermodynamics; North-Holland: Amsterdam, The Netherlands, 1957;
pp 105-106.
Gurtin, M. E. In Encyclopedia of Physics; Flgge, S.; Ed.; Springer: New York, NY, 1972;
Vol. VIa/2, pp 1-162.
Hannink, R. H. J.; Kelly, P. M.; Muddle, B.C. J. Am. Ceram. Soc. 2000, 83, 461-486.
Hardy, D.; Green, D. J. J. Eur. Ceram. Soc. 1995, 15, 769-775.
Hashin, Z. J. Appl. Mech. 1962, 29, 143-150.
Hashin, Z.; Shtrikman, S. J. Mech. Phys. Solids 1963, 11, 127-140.
Hasselman, D. P. H. J. Am. Ceram. Soc. 1962, 45, 452-453.
Haupt, P. Continuum Mechanics and Theory of Materials; Springer: Berlin, Germany, 2000;
pp 177-583.
Hearmon, R. F. S. Introduction to Applied Anisotropic Elasticity; Clarendon Press: Oxford,
UK, 1961; pp 24-44.
Herakovich, C. T.; Baxter, S. C. J. Mater. Sci. 1999, 34, 1595-1609.
Hertz, H. Miscellaneous Papers; Macmillan: London, UK, 1882; pp 163-183.
Hill, R. Proc. Phys. Soc. Lond. A 1952, 65, 349-355.
Hill, R. J. Mech. Phys. Solids 1965, 13, 213-222.
Hulbert, S. F. In An Introduction to Bioceramics; Hench, L. L.; Wilson, J.; Eds.; World
Scientific: Singapore, 1993; pp 25-40.
Irwin, G. R. In Encyclopedia of Physics; Flgge, S.; Ed.; Springer: Berlin, Germany, 1958;
Vol. VI, pp 551-590.
Ishai, O.; Cohen, L. J. Int. J. Mech. Sci. 1967, 9, 539.
Jeulin, D. In Mechanics of Random and Multiscale Microstructures; Jeulin, D.; Ostoja-
Starzewski, M.; Eds.; Springer-Verlag: Wien, 2001; pp 33-91.
Jeulin, D. In Morphology of Condensed Matter; Mecke, K.; Stoyan, D.; Eds.; Springer: Berlin,
Germany, 2002; pp 3-36.
Johnson, K. L. Contact Mechanics; Cambridge University Press: Cambridge, UK, 1985.
Kandil, H. M.; Geiner, J. D.; Smith, J. F. J. Am. Ceram. Soc. 1984, 67, 341-346.
Kisi, E. H.; Howard, C. J. J. Am. Ceram. Soc. 1998, 81, 1682-1684.
Kittel, C. Einfhrung in die Festkrperphysik; Oldenbourg: Mnchen, Germany, 1988 (eighth
edition); pp 90-158.
Koga, K. In Handbook of Advanced Ceramics Vol. I: Materials Science; Somiya, S.;
Kohn, D. H.; Ducheyne, P. In Medical and Dental Materials; Williams, D. F.; Ed.; Materials
Science and Technology A Comprehensive Treatment; VCH: Weinheim, Germany,
1992; Vol. 14, pp 29-109.
Kokubo, T.; Kim, H.-M.; Kawashita, M. In Handbook of Advanced Ceramics Vol. I:
Materials Science; Somiya, S.; Aldinger, F.; Claussen, N.; Spriggs, R. M.; Uchino, K.;
Koumoto, K.; Kaneno, M. Elsevier Academic Press: Amsterdam, The Netherlands, 2003;
pp 385-416.
Komeya, K.; Matsui, M. In Structure and Properties of Ceramics; Swain, M. V.; Ed.;
Germany, 1994; Vol. 11, pp 517-565.
Kovik, J. J. Mater. Sci. Lett. 1999, 18, 1007-1010.
Krawietz, A. Materialtheorie; Springer: Berlin, Germany, 1986; pp 158-188, 361-382.
Krieger, I. M. Adv. Colloid Interface Sci. 1972, 3, 111-136.
Krstic, V.; Erb, U.; Palumbo, G. Scripta Metall. Mater. 1993, 29, 1501.
Kuster, G. T.; Toksz, M. N. Geophysics 1974, 39, 587-606.
Kwan, Y. B. P.; Stephenson, D. J.; Alcock, J. R. J. Mater. Sci. 2000, 35, 1205-1211.
Lakes, R. S. Science 1987, 235, 1038-1040.
Lakes, R. S. J. Mater. Sci. 1991, 26, 2287-2292.
Lakes, R. S. Nature 1992, 358, 713-714.
Lam, D. C. C.; Lange, F. F.; Evans, A. G. J. Am. Ceram. Soc. 1994, 77, 2113-2117.
Landau, L. D.; Lifshitz, E. M. Theory of Elasticity; Butterworth-Heinemann: Oxford, UK,
1986 (third edition); pp 1-37.
Lawn, B. Fracture of Brittle Solids; Cambridge University Press: Cambridge, UK, 1993
(second edition); pp 1-387.
Lawn, B. R.; Howes, V. R. J. Mater. Sci. 1981, 16, 2745-2752.
Lee, W. E.; Rainforth W. M. Ceramic Microstructures Property Control by Processing;
Chapman and Hall: London, UK, 1994; pp 255-387.
Lima, L. F. C. P.; Godoy, A. L. E.; Muccillo, E. N. S. Mater. Lett. 2004, 58, 172-175.
Love, A. E. H. A Treatise on the Mathematical Theory of Elasticity; Cambridge University
Press: Cambridge, UK, 1927 (fourth edition); pp 32-165.
Luo, J.; Stevens, R. Ceramics International 1999, 25, 281-286.
Mackenzie, J. K. Proc. Phys. Soc. London B 1950, 63, 2-11.
Markov, K. Z. In Heterogeneous Materials Micromechanics Modeling Methods and
Simulations; Markov, K. Z.; Preziosi, L.; Eds.; Birkhuser: Boston, MA, 2000; pp 1-162.
Marshall, D. B.; Noma, T.; Evans, A. G. J. Am. Ceram. Soc. 1982, 65, C175-C176.
Martin, L. P.; Nagle, D.; Rosen, M. Mater. Sci. Eng. A 1998, 246, 151-160.
Martin, L. P.; Poret, J. C.; Danon, A.; Rosen, M. Mater. Sci. Eng. A 1998, 252, 27-35.
McLaughlin, R. Int. J. Eng. Sci. 1977, 15, 237-244.
Menk, J. Strength and Fracture of Glass and Ceramics; Elsevier: Amsterdam, The
Netherlands, 1992; pp 17-180.
Meyer, K. Physikalisch-Chemische Kristallographie; VEB Verlag fr Grundstoffindustrie:
Leipzig, Germany, 1977 (second edition); pp 49-57.
Milton, G. W. The Theory of Composites; Cambridge University Press: Cambridge, UK,
2002; pp 1-719.
Milton, G. W. J. Mech. Phys. Solids 1992, 40, 1105-1137.
Milton, G. W.; Phan-Thien, N. Proc. Roy. Soc. London A 1982, 380, 305-331.
Mooney, M. J. Colloid Sci. 1951, 6, 162-170.

Mori, T.; Tanaka K. Acta Metall. 1973, 21, 571-574.
Munro, R. G. J. Am. Ceram. Soc. 1997, 80, 1919-1928.
Munro, R. G. (2002). Elastic moduli data for polycrystalline ceramics [NIST
Property Data Summaries (SRD Database Number 30, NISTIR 6853)]
www.ceramics.nist.gov/srd/summary
Munz, D.; Fett, T. Ceramics Mechanical Properties, Failure Behaviour, Materials
Selection; Springer: Berlin, Germany, 1999; pp 1-278.
Nanjangud, S. C.; Brezny, R.; Green, D. J. J. Am. Ceram. Soc. 1995, 78, 266-268.
Nemat-Nasser, S.; Hori M. Micromechanics Overall Properties of Heterogeneous
Materials; North-Holland / Elsevier: Amsterdam, The Netherlands, 1999 (second
edition); pp 1-786.
Nevitt, M. V.; Chan, S.-K.; Liu, J. Z.; Grimsditch, M. H.; Fang, Y. Physica B 1988, 150, 230-
233.
Newnham, R. E. Structure-Property Relations; Springer-Verlag: Berlin, 1975; pp 173-221.
Nieh, T. G.; Wadsworth, J.; Sherby O. D. Superplasticity in Metals and Ceramics;
Cambridge University Press: Cambridge, UK, 1997; pp 91-116, 192-202.
Nielsen, L. F. Mater. Sci. Eng. 1982, 52, 39-62.
Niihara, K. J. Ceram. Soc. Jpn. 1991, 99, 945.
NIST (2002). Structural Ceramics Database (SCD). www.ceramics.nist.gov/srd/scd
Norris, A. N. Mech. Mat. 1985, 4, 1-16.
Norris, A. N.; Callegari, A. J.; Sheng, P. J. Mech. Phys. Solids 1985, 33, 525-543.
Nye, J. F. Physical Properties of Crystals; Clarendon Press: Oxford, UK, 1957 (reprint
1985); pp 131-149.
Ondracek, G. Reviews on Powder Metallurgy and Physical Ceramics 1987, 3, 205-322.
Ontan, M.; Aparicio, C.; Ginebra, M. P.; Planell, J. A. In Structural Biological Materials;
Elices, M.; Ed.; Pergamon Press: Amsterdam, The Netherlands, 2000; pp. 33-71.
Orowan, E. Z. Kristallogr. A 1934, 89, 327-343.
Orowan, E. Rep. Prog. Phys. 1949, 12, 185-232.
Pabst, W. Ceramics-Silikty 2004, 48, 6-13.
Pabst, W. J. Mater. Sci. 2005, 40, 2667-2669.
Pabst, W.; Gregorov, E. Ceramics-Silikty 2003, 47, 1-7.
Pabst, W.; Gregorov, E. J. Mater. Sci. Lett. 2003, 22, 959-962.
Pabst, W.; Gregorov, E. J. Mater. Sci. Lett. 2003, 22, 1673-1675.
Pabst, W.; Gregorov, E. Ceramics-Silikty 2004, 48, 14-23.
Pabst, W.; Gregorov, E. J. Mater. Sci. 2004, 39, 3213-3215.
Pabst, W.; Gregorov, E. J. Mater. Sci. 2004, 39, 3501-3503.
Pabst, W.; Gregorov, E. Ceramics International 2006, 32, 89-91.
Pabst, W.; Tich, G.; Gregorov, E. Ceramics-Silikty 2004, 48, 41-48.
Pabst, W.; Gregorov, E.; Tich, G.; Tnov, E. Ceramics-Silikty 2004, 48, 145-153.
Pabst, W.; Tich, G.; Gregorov, E.; Tnov, E. Ceramics-Silikty 2005, 49, 77-85.
Pabst, W.; Gregorov E.; Tich, G. J. Eur. Ceram. Soc. 2006, 26, 1085-1097.
Park, J. B.; Lakes, R. S. Biomaterials: An Introduction; Plenum Press: New York, NY, 1992
(second edition), pp. 169-183, 192-207, 344-354.
Passman, S. L.; Nunziato, J. W.; Walsh, E. K. In Rational Thermodynamics; Truesdell, C.;
Ed.; Springer: New York, NY, 1984 (second edition); pp 287-325.
Paul, B. Trans. Metall. Soc. AIME 1960, 218, 36-41.

Pauling, L. Die Natur der chemischen Bindung; Verlag Chemie: Weinheim, Germany, 1968
(third edition); pp. 61-98.
Phani, K.K; Niyogi, S. K. J. Mater. Sci. 1987, 22, 257-263.
Phani, K.K; Niyogi, S. K. J. Am. Ceram. Soc. 1987, 70, C362-C366.
Poole, C. P.; Owens, F. J. Introduction to Nanotechnology; Wiley: Hoboken, NJ, 2003; pp
133-141.
Rahaman, M. N. Ceramic Processing and Sintering; Marcel Dekker: New York, NY, 1995.
Ramakrishnan, N.; Arunachalam, V. S. J. Am. Ceram. Soc. 1993, 76, 2745-2752.
Reuss, A. Z. Angew. Math. Mech. 1929, 9, 49-58.
Rice, R. W. Porosity of Ceramics; Marcel Dekker: New York, NY, 1998.
Rice, R. W. Ceramic Fabrication Technology; Marcel Dekker: New York, NY, 2003; pp
283-292.
Rintoul, M. D.; Torquato, S. J. Chem. Phys. 1996, 105, 9258-9265.
Roberts, A. P.; Garboczi, E. J. J. Am. Ceram. Soc. 2000, 83, 3041-3048.
Robinson, J. V. J. Phys. Colloid Chem. 1949, 53, 1042.
Rothenburg, L.; Berlin, A.; Bathurst, R. J. Nature 1991, 354, 470.
Sahimi, M. Applications of Percolation Theory; Taylor and Francis: London, UK, 1994; pp 1-
258.
Sahimi, M. Heterogeneous Materials I Linear Transport and Optical Properties; Springer:
New York, NY, 2003; pp 410-642.
Sahimi, M. Heterogeneous Materials II Nonlinear and Breakdown Properties and Atomistic
Modeling; Springer: New York, NY, 2003; pp 247-451.
Salmang, H; Scholze, H. Keramik I Allgemeine Grundlagen und wichtige Eigenschaften;
Springer: Berlin, Germany, 1982 (sixth edition); pp. 227-259.
Samohl, I. Rational Thermodynamics of Chemically Reacting Mixtures (in Czech);
Academia: Prague, Czech Republic 1982; pp. 94-277.
Samohl, I. J. Non-Linear Mech. 1997, 32, 235-240.
Samohl, I. Arch. Rational Mech. Anal. 1999, 147, 1-45.
Samohl, I.; Pabst W. J. Non-Linear Mech. 1997, 32, 227-233.
Samohl, I.; Pabst W. J. Non-Linear Mech. 2004, 39, 247-263.
Samohl, I.; ilhav, M. Arch. Rational Mech. Anal. 1990, 109, 299-321.
Sanditov, D. S. Fiz. I chem. stekla 1977, 3, 14.
Scott, G. D. Nature 1960, 188, 908-909.
Schwartz, M. Handbook of Structural Ceramics; McGraw-Hill: New York, NY, 1992; pp
8.1-8.91.
Shen, T. K.; Hing, P. J. Mater. Sci. 1997, 32, 6633-6638.
Sigmund, O. Mech. Mater. 1995, 20, 351-368.
Sigmund, O. Int. J. Solids Struct. 1994, 31, 2313-2319.
ilhav, M. The Mechanics and Thermodynamics of Continuous Media; Springer: Berlin,
Germany, 1997; pp 1-504.
Slaughter, W. S. The Linearized Theory of Elasticity; Birkhuser: Boston, MA, 2002; pp 193-
254, 331-388.
Sneddon, I. N.; Berry, D. S. In Encyclopedia of Physics; Flgge, S.; Ed.; Springer: Berlin,
Germany, 1958; Vol. VI.
Sobczyk, K.; Kirkner, D. J. Stochastic Modeling of Microstructures; Birkhuser: Basel,

Switzerland, 2001, pp 149-245.
Soltesz, U.; Richter, H. In Metal and Ceramic Biomaterials, Vol. II: Strength and Surface;
Ducheyne, P.; Hastings, G. W.; Eds.; CRC Press: Boca Raton, FL, 1984; pp 23-61.
Spriggs, R. M. J. Am. Ceram. Soc. 1961, 44, 628-629.
Stauffer, D.; Aharony, A. Introduction to Percolation Theory; Routledge / Taylor and
Francis: London, UK, 1985 (revised second edition); pp 1-181.
Sumner, D. R.; Turner, T. M.; Igloria, R.; Urban, R. M.; Galante, J. O. J. Biomech. 1998, 31,
909-917.
Suryanarayana, C.; Koch, C. C. In Non-Equilibrium Processing of Materials; Suryanarayana,
C.; Ed.; Pergamon Press: Amsterdam, The Netherlands, 1999; pp. 311-344.
Takebayashi, H. In Handbook of Advanced Ceramics Vol. I: Materials Science; Somiya, S.;
Theocaris, P. S. The Mesophase Concept in Composites; Akademie-Verlag: Berlin, Germany
1987.
Timoshenko, S.; Goodier J. N. Theory of Elasticity; McGraw Hill: New York, NY, 1951
(second edition); pp 359-362.
Torquato, S. In Macroscopic Behavior of Heterogeneous Materials from the Microstructure;
Torquato, S.; Krajcinovic; Eds.; ASME: New York, NY, 1992; Vol. 147; pp 53-65.
Torquato, S. J. Mech. Phys. Solids 1997, 45, 1421-1448.
Torquato, S. J. Mech. Phys. Solids 1998, 46, 1411-1440.
Torquato, S. Int. J. Solids Struct. 2000, 37, 411-422.
Torquato, S. Random Heterogeneous Materials Microstructure and Macroscopic
Properties; Springer: New York, NY, 2002; pp 1-701.
Truesdell, C. Rational Themodynamics; Springer: New York, NY, 1984 (second edition); pp
219-395.
Truesdell, C.; Noll, W. The Non-Linear Field Theories of Mechanics; Springer: New York,
NY, 2003 (third edition); pp 1-426.
Truesdell, C.; Toupin, R. A. In Encyclopedia of Physics; Flgge, S.; Ed.; Springer-Verlag:
Berlin, 1960; Vol. III/1, pp 226-793.
Tsuru, T. In Handbook of Advanced Ceramics Vol. I: Materials Science; Somiya, S.;
Voigt, W. Wiedemanns Ann. Physik Leipzig 1889, 38, 573-587.
Voigt, W. Lehrbuch der Kristallphysik; Teubner: Leipzig, 1910 (reprint 1928); pp 954-964.
Volf, M. B. Mathematical Approach to Glass; Elsevier: Amsterdam, 1988; pp 87-103, 266-
275.
Wachtman, J. B. Mechanical Properties of Ceramics; Wiley: New York, NY, 1996; pp 23-35.
Wachtman, J. B.; Tefft W. E.; Lam D. G.; Stinchfield, R. P. J. Res. Natl. Bureau of Standards
A. Physics and Chemistry 1960, 64A, 213-227.
Wagh, A. S.; Poeppel, R. B.; Singh, J. P. J. Mater. Sci. 1991, 26, 3862-3868.
Walpole, L. J. J. Mech. Phys. Solids 1966, 14, 151-162.
Warren, W. E.; Kraynik, A. M. J. Appl. Mech. 1988, 55, 341-346.
Wei, G. J. Chem. Phys. 1992, 96, 3226-3233.

Weibull, W. Swed. Inst. Eng. Res. Proc. 1939, 151, 1-45.
Weibull, W. J. Appl. Mech. 1947, 14, 63.
Weibull, W. J. Appl. Mech. 1951, 18, 293.
Weinans, H.; Sumner, D. R.; Igloria, R.; Natarajan, R. N. J. Biomech. 2000, 33, 809-817.
Weng, G. J. Int. J. Eng. Sci. 1984, 22, 845-856.
Willis, J. R. J. Mech. Phys. Solids 1977, 25, 185-202.
Winterer, M. Nanocrystalline Ceramics Synthesis and Structure; Springer: Berlin, 2002; pp
1-260.
Yamada, T. In Handbook of Advanced Ceramics Vol. I: Materials Science; Somiya, S.;
Zallen, R. The Physics of Amorphous Solids; Wiley: New York, NY, 1983; pp 49-59, 135-
204.
Zeng, K.; Breder, K.; Rowcliffe, D. J.; Herrstrom, C. J. Mater. Sci. 1992, 27, 3789-3792.
Zimmerman, R. W. Int. J. Rock Mech. Mining Sci. 1984, 21, 339-343.
Zimmerman, R. W. Mech. Mater. 1991, 12, 17-24.
Zimmerman, R. W. Compressibility of Sandstones; Elsevier: Amsterdam, The Netherlands,
1991; pp 110-127.
Zimmerman, R. W. Appl. Mech. Rev. 1994, 47, S38-S44.
Chapter 3
PROGRESS IN BIOCERAMIC MATERIALS
P. N. De Aza*
Instituto de Bioingeniera, Universidad Miguel Hernndez de Elche,
Edificio Torrevaillo. Avda. de la Universidad s/n. 03202- Elche, Alicante, Spain.
Abstract
A strong interest in the use of ceramics for biomedical engineering applications
developed in the late 1960s. Used initially as alternatives to metallic materials in order to
increase the biocompatibility of implants, bioceramics have become a diverse class of
biomaterials presently including three basic types: relatively bioinert ceramics maintain their
physical and mechanical properties in the host and form a fibrous tissue of variable thickness;
surface reactive bioceramics which form a direct chemical bonds with the host; and
bioresorbable ceramics that are dissolved with the time and the surrounding tissue replaces it.
A review of the composition, physicochemical properties and biological behaviour of the
principal types of bioceramics is given, based on the literature and some of our own data. The
materials include, in addition to bioceramics, bioglasses and bio-glass-ceramics. Special
attention is given to structure as the main physical parameter determining nor only the
properties of the materials, but also their reaction with the surrounding tissue.
In addition, biomaterials with appropriate three-dimensional geometry (appropriate
porous structure) are highly recommended because are able to become osteoinductive
(capable of osteogeneis), and can be effective carriers of bone cell seeds. A new way of
preparing in situ porous bioactive ceramics with interconnected porosity from a dense
bioactive ceramic is presented. To this purpose the binary system wollastonite- tricalcium
phosphate is selected, taking into account the different bioactivity behavior of both phases.
The present status of research and development of bioceramic is characterized as first
step in the solution of complex problems at the confluence of materials science and
engineering, biology and medicine.
Introduction
Ceramics, glasses, and glass-ceramics include a broad range of inorganic/non-metallic
materials whose chemical compositions, chemical bond types and properties vary over a very
*
E-mail address: piedad@umh.es
102 P. N. De Aza
wide range. Bioceramic are those engineered materials that find their applications in the field
of medicine [1]. Traditionally, ceramics materials have been limited as far as their
applications by their brittleness, their low mechanical fracture toughness and their low
resistance to the impact. Ceramics are generally hard; in fact the measurement of hardness is
calibrated against ceramic materials. Diamond is the hardest, with a harness index on Mohs of
10, and talc is the softest ceramic (Mohs hardness 1), while ceramic such as alumina (Mohs
hardness 9), quartz (Mohs hardness 8), and apatite (Mohs hardness 5), are in the middle range.
Other characteristics of ceramic materials are their high melting temperatures and low
conductivity of electricity and heat. Nevertheless, a strong interest in the use of ceramics for
biomedical engineering applications was developed in the late of sixties. New ceramics, with
very improved properties were contributing to increased possibilities for use of ceramic in
biomedicine and their use has extended considerably since then [2,3].
The great chemical inertia of the ceramics, their high compression strength and their
aesthetic appearance, caused that these materials were begun to use in dentistry, mainly in
dental crowns. Later their use extended to orthopaedic applications [4-6].
A basic reason for the high demand for spare body parts is the damage to organic
tissues which occurs with age. Bone tissue is particularly subject to age related changes.
Women are vulnerable to this to a greater degree, from 30 to 65 years of age the bone strength
of women decreases by approximately 40% (men by 20%). Bone density decreases with age
because the osteobleasts (cells which generate bone tissue) became less productive in new
tissue formation. Decreasing density substantially lowers the strength of spongy tissue located
at the end of long bones and in vertebrae. A consequence of this are numerous breaks of the
hip bone stem in old people, or deformation of the vertebrae and pain in the spin column.
A basic problem in the use of bioceramics is replacement of damage bone with a material
which can function successfully throughout the entire lifetime of a patient. Since the average
lifetime in developed countries is now more than 80 years, the need for implants begins at
about 60, the desired service life of a bioceramic is 20 years or more. The requirements of life
long functionality must be satisfied under conditions which are extreme for a bioceramics:
corrosive body fluids at 37C, simultaneous action of ferments and complex cell systems
under the conditions of variable multiaxial and cylindrical loading. The great efficacy of
specially created ceramics for biomedical applications under such conditions is one of the
most important achievements of research and manufacturing in the present century.
Classification of Bioceramics
Used initially as alternatives to metallic materials in order to increase the biocompatibility of
implants, bioceramics can be classified from different points of views [7-10]:
1. According to the type of answer of the living host

2. According to the application to which they are destined
3. According to the characteristics of the material
Progress in Bioceramic Materials 103
1 According to the Type of Answer of the Living Host
Bioinert: Relatively bioinert ceramics maintain their physical and mechanical properties
while in the host. They are those stable bioceramics that do not react appreciably when they
are implanted in the body. The implant does not form a bond with bone. Alumina (-Al2O3)
is a typical example of ceramic bioinert. Other examples, as we see next, are the zirconia
ceramics (ZrO2) and pirolitics carbon ceramics.
Bioactive or surface reactive: The concept of bioactive material is intermediate between a

bioinert material and biodegradable or resorbable material. Upon implantation in the host,
surface reactive ceramics form a strong bond with adjacent tissue. The bioactive materials
(ceramics, glasses and glass-ceramics) bone to living bone through a carbo-hydroxyapatite
layer (CHA) biologically active, which provides the interface union with the host. This phase
is chemical and structurally equivalent to the mineral phase of the bone, and the responsible
of the interface union.
Biodegradable or resorbable: As the name implies, the ceramic dissolves with time and
are gradually replaced by the natural tissues. A very thin or non-existent interfacial thickness
is the final results. This type of bioceramics would be the ideals, since only remain in the
body while their function is necessary and disappear as the tissue regenerates. Their greater
disadvantage is that their mechanical strength diminishes during the reabsorption process.
One of the few bioceramics that fulfil partially these requirements is the tricalcium phosphate
(TCP).
2 According to the Application to which They Are Destined
Structural ceramics: They are bioceramics of high mechanical strength and generally
bioinerts. Typical examples are alumina (-Al2O3) and zirconia (ZrO2).
Non-structural ceramics: They are generally biodegradable or bioactive bioceramics,

dense or porous, with low mechanical strength, since they do not have to support great loads.
Typical examples, the hydroxyapatite (HA) and the tricalcium phosphate (TCP).
3 According to the Characteristics of the Material
Dense and inert bioceramics: They are materials without porosity. The bond to the bone
is morphologic; by growth of the tissue in the superficial irregularities of the implants or bond
through acrylic cement or by fit the implant in a defect by pressure. Typical example of this
group is monocrystalline as much as polycrystalline alumina.
Porous and inert bioceramic: The bond to the bone is mechanical and the fixation is
biological. The growth of the bone takes place through the pores of implants. Porous
polycrystalline alumina is also a typical example of the group.
104 P. N. De Aza
Dense or porous bioactives bioceramics: The bond to the tissue is of chemical type and
the fixation bioactive. Typical example of the group is HA, bioglasses and bioactive glass-
ceramics
Dense and resorbable bioceramics: The implant is replaced slowly by the bone. Belong
to this group the tricalcium phosphate (TCP) and other phosphates as well as the calcium
sulphate (CaSO4H2O = plaster of Paris).
Bioinert Ceramics
Bioinert ceramics undergo little or no chemical changes when they are exposed to
physiological environments. The answer of the host to these bioceramics is the formation of a
very fine fibrous tissue capsule of varying thickness, several micrometers or less, that
surround implant materials. The fixation of implants in the body is made through a strong
mechanical interlocking, by tissue ingrowth into undulating surfaces [11]. When high strength
is required, the bond is made by perforations in the implants using threads, cements, etc.
When so high strength are not required can be used porous inert bioceramics, with sizes of
pore between 100 and 150 m, which guarantee the growth of the tissue towards within
implants assuring its fixation [12-14].
Alumina (-Al2O3)
Alumina of high density and purity (99.5 % in weight of -Al2O3), with an average grain
size < 4 m is probably, the bioinert ceramic material of greater biological interest. This
material was developed, as alternative to used metallic alloys in load-bearing hip prosthesis
and in dental implants, to display an excellent biocompatibility, good resistance to the
corrosion, to form a very fine fibrous capsule, to have a low coefficient of friction and good
mechanical properties as much high strength as wear resistance [15].
The main source of high-purity alumina is bauxite and it is obtained by Bayer process.
After that, the material is putting under a purification process and heat treatment to obtain the
phase -Al2O3 and milling until sizes < 0,3 m. Pieces of alumina are obtained by cold
isostatic press of these high purity powder and sintering at temperatures ranging from 1600 to
1700C depending the properties of the raw material. Nowadays also alumina is obtained by
hot isostatic press (HIP) achieving grain sizes < 4 m. The obtained material polishes with
diamond until superficial roughness of 0.02 m. A very small amount of MgO
(approximately a 0.5 % in weight) is added to the alumina in order to inhibit the growth of the
grain [16,17].
According to the International Standards Organization (ISO), the purity of the alumina
that is used in biomedical applications has to be over 99.5 %, being the rest of the impurities
(SiO2, Na2O, K2O, CaO, etc.) below 0.1 % in weight, in order to avoid the formation of a
liquid phase during the sintering and consequence a large grain size increase. An increase of
average grain size to 7 m can decrease mechanical properties by about 20 %. Table I shows
the requirements for alumina implants according to ISO 6474-81.
Table I. Physical Characteristics of Al2O3 Implants
Property ISO 6474-81

% by weight of Al2O3 99.50
Density (g/cm3) 3.90
Average grain size (m) <7
Hardness (Vickers) > 2000
Compressive strength (MPa) 4000
Bending strength (MPa) 400
Young modulus (GPa) 380
The most common application of alumina bioceramics is in the area of orthopaedics as

component in hip and knee prostheses. The first clinical use of total hip prosthesis with an
alumina head and alumina socket was in 1971. The long term coefficient of the pair
alumina/alumina and the wearing down index decrease with time and approaches the value of
a normal healthy joint [18].
The main problem with total hip system is the loosening of the acetabular component,
which is caused by wear debris. For this reason alumina is used only in the head of femur
since numerous clinical studies indicates that alumina/ultrahigh molecular weigh
polyethylene (UHMWPE) pair reduces wear debris by a factor of 10 or greater [19].
The main limitations of the alumina/UHMWPE pair are their sensitivity to the watery
means, which can cause static fatigue, and their low resistance to the fracture, that can get to
be critical in zones of stress concentrations. These conditions are present in the hip implants
(watery atmosphere, severe cycles of tensions, overloads, impacts, etc.)[20,21]. Nevertheless,
the main problem is the relaxation of the acetabular component with emission of particles.
These particles are accumulated and been able to induce very adverse tissue reactions.
Other clinical applications of alumina are the used for screws, tooth-root implants, orbital
wall, implants for maxillofacial reconstructions, dental implants etc. Between these last ones,
the most well-known application is Tbingen implant [22-24] employee like substitute dental
immediately after the extraction or in edentulous regions. The effectiveness of these implants
is of 92.5 %. The cross-sectional fracture is the most frequent failure of these implants, due to
relatively low resistance to the flexion, being the main cause of failure. Due to it, the use of
monocrystalline alumina began to used as dental implant [25,26] with resistance to the
flexion 3 times over polycrystalline alumina. However, the use of monocrystals does no
eliminate the basic disadvantage of alumina bioceramics (low fracture toughness), which for
monocrystalline alumina is equal to 4-5 MPa m1/2. These monocrystals are obtained generally
by grown and crystallization from the melt or by a process of fusion in flame with germ by
the Verneuil method [27].
Zirconia (ZrO2)
Zirconia, has become a popular alternative to alumina as structural ceramics because,

properly treated, it has a greater resistance to the fracture (greater fracture toughness) of any
monolithic ceramic. On the other hand, zirconia is also exceptionally inert in the
physiological environment and presents very good static fatigue strength. In addition, the
106 P. N. De Aza
zirconia/UHMWPE pair displays lower coefficient of friction than the alumina/UHMWPE

pair, with the consequent diminution of particles to the physiological environment [28,29].
The oxide of pure zirconium (Zirconia = ZrO2) is obtained from sands of zircon (ZrSiO4)
or baddeleyite (ZrO2) by a chemical processes via of chlorination and thermal decomposition,
alkali oxide decomposition or lime fusion or by plasma decomposition [30].
Zirconia presents three different crystalline forms: monoclinic, from room temperature to
1100 C, tetragonal, from this temperature to 2372 C, and cubic, from this last
temperature to the melt at 2680 C. The monoclinic tetragonal transformation is of
martensitic type, and therefore totally reversible when warming up or to cool through the
transition temperature (1100 C), with an increase of volume of the order of 3 to 5 %, which
entails the cracking or fracture of the pieces. In order to avoid this transformation, and even
increase the toughness of the zirconia materials, is recommended the use of total or partial
stabilization of the tetragonal or cubic phase by suitable additives. Zirconia of higher
toughness usually is obtained stabilizing its tetragonal phase of high temperature by doping
with yttria (3 mol %), followed on the suitable heat treatments, considering the binary
diagram ZrO2 - Y2O3 [31]. By this way the zirconia bioceramic is a totally a tetragonal
material with a microstructure formed by very small grains of the order of 0.2 to 0.5 m. The
additives most used in biomedical applications are the yttrium oxide (Y2O3 = yttria) and the
magnesium oxide (MgO = the magnesia).
Between all the ceramic materials, zirconia are those where their mechanical properties
depend more by the sintering process, because is necessary to have a balance between the
density and the average grain size of the bioceramic [32].
Table II shows the physical characteristic of tetragonal zirconia stabilized with yttria
(TZP = tetragonal circonia polycrystals) and magnesium oxide partially stabilized zirconia
(Mg-PSZ = magnesia partially stabilized zirconia [33].
Table II.- Physical Characteristic of Two Types of Zirconia Materials
Characteristic TZP Mg-PSZ

Purity 97 % 96.5
% Y2O3 3 mol %
% MgO 3.4 mol %
Density(g/cm3) 6.05 5.72
Average grain size (m) 0.2 0.4 0.42
Bending strength (MPa) 1000 800
Compressive strength (MPa) 2000 1850
Young Modulus (GPa) 150 208
Hardness(Vickers) 1200 1120
Fracture toughness (KIc) (MPam) 78 8
Nevertheless, recently studies of new materials of TZP, doped with yttria tested in
simulated body fluids and in animals, have shown slight decreases in fracture toughness and
Young Modulus [34-36]. The observed strength, after two years, is sill much higher that the
strength of alumina bioceramic tested under similar conditions. Because the zirconia/zirconia
pair display a wear rate 5000 times greater than the alumina/alumina pair, those do not have
to be used in articulated surfaces.
As far as the potential radioactivity of the zirconia prosthesis, although the detected
activity has been small, the long term effects of the alpha radioactivity still must be evaluated.
French government, very recently, has prohibited the manufacture, distribution, export
and use of these prosthesis in France by decision of the July 22, 2003, because of having
detected frequent breakage of femoral heads made in zirconia [37].
Composite Materials of Al2O3-ZrO2
Recently, researches have been made with materials composed of alumina reinforced
with zirconia (until 15 % in volume of zirconia) with the purpose of improving the reliability
of the single-phase alumina and zirconia ceramic implants. De Aza et. at [38] has put in
evidence that, in general, these new composite materials can display not only a greater
toughness (KIC) that the monolithic materials previously mentioned, otherwise, and what it is
more important, a greater tension threshold for the stress intensity factor (KI0), below which
some of crack does not take place propagation (Table III). Therefore, in the case of the
ceramics prostheses, this tension threshold provides a rank of intensity of tensions of total
security for the use of the composite material under mechanical efforts.
Table III
Material Fracture Tougnhess Hardness

threshold KIc(MPam) H (Vickers)
KI0(MPam)
Alumina (Al2O3) 2.5 0.2 4.2 0.2 1800
Zirconia (3Y-TZP) 3.5 0.2 6.1 0.2 1290
Composite (Al2O3-10% ZrO2) 4.0 0.2 5.9 0.2 1530
On the other hand, since the hardness and the chemical stability are equally important in
the field of the prosthesis, these composite materials, with relatively low contents of zirconia
(10 % in volume), display values of hardness similar to those of alumina and they are not
susceptible to the hydrothermal instability observed in some cases of the stabilised zirconia
bioceramics. Therefore, these materials appear like an alternative, to consider in the future,
for the production of ceramics prostheses. Figure 1 shows a microstructure of one of these
composite alumina-zirconia materials.
Carbons
Carbon presents a great variety of forms: amorphous carbon, graphite, diamond, vitreous
carbon and pyrolitic carbon. Some of them display the most excellent properties of
biocompatibility, chemical inertia and thromboresistance that any other bioceramic. On the
other hand, another advantage of these materials is that their physical characteristics are next
to those of the bone [40]. Thus, their densities, according to the type carbon, change between
1.5 - 2.2 g/cm3, and their elastic modules between 4 - 35 GPa. In spite of all the mentioned
108 P. N. De Aza
varieties, only three types of carbon are used for biomedical devices: the pyrolitic carbon, in
its two varieties of low temperature isotropic (LTI) and ultra-low temperature isotropic
(ULTI), and the vitreous carbon. The three types of carbon have disordered lattice structures
and are collectively referred to as turbostratic carbons. While the microstructure of
turbostratic carbon might seem very complicated due to its disordered nature, it is in fact
quite closely analogous to the structure of graphite, but with at random oriented layers [41].
Figure 1. Microstructure of a material composed of alumina (dark phase) with a 10% in volume of
zirconia (clear phase) [39].
Pyrolitic carbon is widely used for implant fabrication. It is normally used as surface
coating. It is also possible to coat surfaces with diamond. Although the techniques of coating
with diamond have the potential to revolutionize medical device manufacturing, they are not
yet commercially available. Pyrolitic carbon devices are typically made via a chemical
vapour-deposition, from a hydrocarbon gas in a fluidized-bed using at a controlled
temperature and pressure.
Its great cellular biocompatibility with the blood and the soft tissue as well as its
excellent thromboresistance, does carbon material to be used fundamentally in applications of
the circulatory apparatus, blood vessel and mechanical cardiac-valve prosthetic devices, being
this last the most extended application. Nowadays, most of the modern heart valves are made
with a coating of LTI on a polycrystalline graphite substrate or like a monolithic material [42].
In both cases is frequently added up to 10 % in weight of silicon, often in the form of
discrete sub-micron -SiC particles randomly dispersed in a matrix of roughly spherical
micron- size subgrain of pyrolitic carbon. The doping with silicon improve the mechanical
properties of pyrolitic carbons, issue of great importance in the heart valves, where the joint is
subject to degradation by cyclic fatigue or stress corrosion and possible cavitations by erosion
that can happened during the life of the patient.
Whereas the pyrolitic carbons coating have been applied in zones in contact with the
blood, due to their excellent thromboresistance, the vitreous carbons have been studied
mainly to bond to soft and hard tissues, without inflammatory answer in the adjacent tissues.
Similar behaviour has been registered for pyrolitic carbons LTI and ULTI.
Apart from of the mistral and aortic heart valves, there are applications in dental field
[43] and middle ear reconstruction and in devices of LTI coating on titanium to make easy the
circulation of the blood. Recently, success was achieved in coating ULTI onto the surfaces of
blood vessel implants made of polymers. The coating has excellent compatibility with blood
and is thin enough not to interfere with the flexibility of the grafts [44].
Bioactive Ceramics
The appearance of this type of bioceramics born of the need to eliminate the interface
movement that takes place with the implantation of bioinert ceramics. Consequently, L. L.
Hench proposes in 1967 to the U.S.A. Army Medical Research and Development Command,
a research based on the modification of the chemical composition of ceramics and glasses so
that they have chemical reactivity with the physiological system and form chemical bond
between the adjacent tissue and the surface of implant materials.
Hench and collaborators show for the first time in 1971 that a material done by the man
can be bond to the bone. This material was called Bioglass; beginning therefore the field
of the bioactive materials [45].
Materials that are bioactive develop an adherent interface with tissue that resist
substantial mechanical forces. In many cases, the interfacial strength of adhesion is equivalent
to or greater than the cohesive strength of the implant material or the tissue bonded to the
bioactive implant. Generally, the break takes place in the implant or in the bone but almost
never in the interface.
Hench and Wilson [46] show that the surface chemical reactions result in the formation
of a carbo-hydroxyapatite (CHA) layer to which bone can bond. This CHA layer is
biologically activates and it is chemical and structurally equivalent to the constituent mineral
phase of the bone [47-49].
Hydroxyapatite (Ca10(PO4)6(OH)2)
Hydroxyapatite (HA) is the primary mineral content of bone representing 43 % by

weight. HA is a calcium phosphate whose stoichiometric formula corresponds to a:
Ca10(PO4)6(OH)2, with a Ca/P molar ratio = 1.67. HA belong the mineral family of Apatites
whose name derived from Greek means deception or deceit, due to the facility it
was confused with other mineral species like the beryl or the tourmaline [50].
Apatites form an important series of minerals. They occur as a minor constituent of many
igneous rocks, although a few large igneous deposits are known, like the Kola peninsular.
Apatite is also present in most metamorphic rocks, especial crystalline limestone. Less well
crystallised deposits of rather variable composition, usually referred to as rock phosphates or
phosphorite, occur in large deposits, some of which were formed by the reaction between
phosphatic solutions from guano and calcareous rock, or precipitated from sea water.
HA displays ionic character, and its crystalline structure can be describe like a compact
hexagonal packing of oxygen atom with metals occupying the tetrahedral and octahedral
holes of the periodic network. The basic apatite structure is hexagonal with space group
P63/m and approximate lattice parameters a = 9.4 and c = 6.9 , being the fluorapatite
110 P. N. De Aza
(Ca10(PO4)6F2) the representative example of this structure. Nevertheless, HA presents a low

symmetry, monoclinic, due to the distortion of ion OH- with respect to the ideal model that
would represent the position of spherical ion F- in the fluorapatite. Nevertheless, in most of
the works in the field of biomaterials it assumed that HA has a fluorapatite structure but with
lattice parameter a = b = 9.418 and c = 6.884 . = = 90 and = 120, Z = 2. The unit
cell contains 10 Ca+2, 6 PO43- and 2 groups OH- [51].
HA allow the substitutions of many ions in its structure. Change for calcium ions (Ca2+)
or phosphoate groups (PO4)3 or hydroxyls groups (OH). Consequences of these
substitutions are the change in its properties like lattice parameters, morphology, solubility
etc., without significant change in the symmetry. Thus for example, the substitution of F by
OH involve a contraction in the a axis without change in the c axis, associated to an increase
in the crystalllinity and imparts greater stability to the structure. On the other hand, this imply
that the fluorapatite is less soluble than the HA and that all F- free apatites including
biological apatites. Many other ions can enter in the HA network, affecting the properties,
crystallyinity, thermal stability, speed of dissolution etc.
The mechanical properties of the HA are similar to those of the most resistant
components of the bone. HA has an elastic modulus of 40-100 GPa, dental enamel 74 GPa,
the dentine 21 GPa, the compact bone18-12 GPa. Nevertheless, dense bulk compact of HA
has a mechanical resistance of the order of 100 MPa in front of to the 300 MPa of the human
bone, diminishing drastically this resistance in the case of porous bulk compact.
Biological apatites are those components of the enamel and dentine of the teeth as well as
of the human bones and some pathological calcifications (salivary and urinary stones). These
biological apatites differ from pure HA in physical and mechanical properties, stoichiometry,
composition, crystallinity and other properties, but in general terms they are possible to be
assimilated more appropriately to the calcium deficient HA and carbohydroyiapatite, where
group (CO3)2- replace group (PO4)3- and where the Ca2+ is replaced by Na+ to balance the
charges [52,53].
The most current methods of obtaining pure HA are by hydrotermal and solid state
reactions. However, when prepared from aqueous reactions either by precipitation or
hydrolysis methods, the HA obtained is usually calcium deficient (Ca/P molar ratio lower
than the stoichiometric value of 1.69). When the precipitation reaction is carried out under
very basic conditions, the precipitate will contain carbonate, which will make the Ca/P molar
ratio higher than the stoichiometric value.
The interest of the HA as biomaterial comes clearly by its similarity with the mineral
phase of the bone tissue. In principle, HA would be the suitable material as much as for
restoration as for bone substitution. However, the relatively low strength and toughness of
HA, produced little interest among researchers when the focus of attention is on dense
structural samples. Therefore, its use is restricted to all those applications where mechanical
efforts are not required, finding its application concentrated as coating on metallic substrate
with the object to accelerate and to increase the fixation of the prosthesis to the bone [54,55].
Industrial and laboratory techniques, used for HA coating and other ceramics materials
onto metallic substrates, include plasma spraying, laser ablation, electrophoretic deposition,
sputtering and hot isostatic pressing. The most extended is first, mainly in the coating of hip
prosthesis, obtaining what has been denominated biological fixation of the prosthesis [56,57].
Among other applications are: a) coating of dental and maxillofacial prosthesis; b) dental
implants; c) middle ear reconstruction; d) augmentation of alveolar ridge; e) periodontal

defects; f) spinal surgery; g) pulp-capping materials [58].
SiO2-Based Ceramic: Wollastonite (CaSiO3)
Silicates represent the dominant category of the traditional ceramics and glass industries.
These materials are economical due to the abundant availability of raw materials. Also,
silicates provide adequate mechanical, thermal and optical properties for a wide range of
traditional and advanced materials applications.
Ceramics and glasses are distinguished primary by the presence respectively of
crystalline or noncrystalline structure on the atomic scale. For noncrystalline silicate glass,
biomedical applications have been a significant demand because the development of
Bioglass.
Since the development of Bioglass, many bioactive glasses and glass-ceramics in the
complex systems containing CaO-SiO2-P2O5 as the main components have been produced
[45,59-61]. Hench et al. [62,63], Gross et al. [64,65], Karlson et al. [66,67] and Kokubo et al.
[68,69] pointed out that these glasses and glass-ceramics bond to a living bone through
hydroxyapatite (HA) layer which is formed on their surfaces.
Conversely, Ohura et al. [70] have observed that CaO-SiO2 glasses, free of P2O5 as well
as those containing very small amounts of P2O5, form the HA layer on their surfaces
(bioactivity), when they are soaked in simulate body fluid (SBF), whereas CaO-P2O5 glasses
free of SiO2 do not form HA layer in SBF. This seems to indicate that bioactive materials can
be obtained with compositions based on the CaO-SiO2 system rather than in the CaO-P2O5.
Others researchers still claim that to show bioactivity, glasses and glass-ceramics must
contain both CaO and P2O5, which are the main components of the hydroxyapatite [71,72].
On these bases, two polycrystalline chain silicate minerals have been developed synthetically
for use as bioactive ceramic materials. Diopside (CaMgSi2O6) [73,74] and wollastonite
(CaSiO3) [75,85].
Pseudowollastonite (CaSiO3), the high temperature form of wollastonite, displays in its
structure calcium ion chains easily removals by protons, as it was shown by Bailey and
Reesman [86] in the study of the kinetic of dissolution of the wollastonite in H2O-CO2.
system. This fact suggests the possibility of extraction of calcium ions, from wollastonite
structure by protons in an appropriate medium, favouring therefore the precipitation and
formation of a layer of HA on the surface of the material.
De Aza et al. have demonstrated the formation of a HA-like layer on the surface of
pseudowollastonite ceramic both in vitro [75,76,79,83-85] and in vivo [80,81].
Experiments in vitro involved suspension of the material in simulated body fluid with an
ion concentration, pH and temperature virtually identical to human blood plasma, and in
human parotid saliva.
Figure 2 shows the overall microstructure of the polished cross-section of the
pseudowollastonite pellet after one month immersion in simulated body fluid and its relevant
X-ray maps for silicon, calcium and phosphorous elements. This compositional
microcharacterisation of the interface showed that the reaction zone was composed of two
chemically dissimilar layers formed on the pseudowollastonite surface. The outer layer, with
an average thickness of about 10 m, was composed of a CaO/P2O5-rich phase, identified as
112 P. N. De Aza
HA-like phase by thin film X-ray diffraction [77], while the much thinner underlayer (about
1/10 of the thickness of the outer layer) in direct contact with the pseudowollastonite
substrate was rich in silicon but depleted in calcium.
Figure 2. X-ray elemental maps of Si, Ca, P, and a relevant SEM image of the cross-section of the
pseudowollastonite ceramic after 1 month immersion in SBF.
The reaction begins with an ionic exchange of Ca2+ from the pseudowollastonite network
for 2H3O+ from the surrounding fluid, which progressively transforms the pseudowollastonite
crystals into an amorphous silica phase. As the reaction proceeds, the calcium concentration
and the pH of the surrounding fluid increase, creating appropriate conditions for partial
dissolution of the amorphous silica and subsequent HA-like precipitation. This type of HA-
like layer has not been observed for non-bioactive materials, indicating that the HA-like layer
plays an essential role in forming the tight chemical bond between the bioactive material and
the bone [7,69,110].
The authors also have evaluated the cytotoxicity of the pseudowollastonite and the
suitability of the material as a substratum for cell attachment and the ability to effect
osteoblast at a distance from the material surface [83,85]. These experiments demonstrate that
extracts of pseudowollastonite do not show any significant cytotoxic effects and confirm the
biocompatibily of this material. On the other hand, osteoblastic cell can rapidly adhere to the
material surface and that process is aided by the binding of serum proteins to the surface HA
layer. In addition, osteoblastic cells at a distance form the material exhibited a dramatic
alteration on their appearance. This reinforces the suggestion that the release of soluble
factors as well as direct interaction of osteoblactic cell with the surface of the material may
mediate the osteogenic effects of bioactive materials.
In vivo experiments were consisted of implanting small cylinders of
pseudowollastonite into rat tibias [80,81]. Histological observations twelve weeks after
implant show that the bone in contact with the surface of the pseudowollastonite appeared to
be progressively replaced by bone with laminar arrangement. The new bone was growing in
direct contact with the peudowollastonite implants after only 3 weeks [80].
Figure 3 SEM image of the polished cross-sections of the pseudowollastonite ceramic implant after
twelve weeks of implantation and its individual X-ray maps of the silicon, calcium and phosphorous
distribution.
Figure 3 shows SEM micrograph of the polished cross-sections of the pseudowollastonite

ceramic implants after twelve weeks of implantation. The individual X-ray maps of the
silicon, calcium and phosphorous distribution are also included. This compositional
microcharacterisation of the interface showed that there are two chemically separate layers
formed on the implant surface. The outer layer contained Ca and P, while the underlayer was
rich in Si and depleted in Ca.
Overrall, these results suggest that pseudowollastonite is biocompatible and
osteoinductive and it can be used for substitution or repair of bone tissues in places where
mechanical solicitations are not high.
Composite Bioeutectic Material (CaSiO3-Ca3P2O8)
Natural and synthetic materials have been used clinically for many years to reconfigure
anatomic structures for aesthetic and therapeutic reasons in several different surgical
situations, however many outstanding problems still remain unsolved. The most important is
the process of total osteointegration of ceramic implants in the human body. When bioactive
materials are implanted in a living body, the interaction between the bone tissue and these
materials usually take place on their surfaces, with the remaining bulk of the material
unchanged, often causing a harmful shear stress.
To improve the ingrowth of new bone into implants (osteointegration), and therefore the
strength of the bonding with time, the use of materials with an appropriate interconnected
114 P. N. De Aza
porous structure has been recommended [87-89]. The design of a porous ceramic implant
material has the potential of controlling the bone ingrowth, but its initial mechanical
properties are usually inferior to the more compact material. To overcome this problem the
idea of creating artificial bone, making use of coral, has been developed [89,90-92]. At
present, however, there is no evidence in vivo to suggest that the highly organised
macropore structure of the coral-derived HA is better, with regard to bone ingrowth and
bioresorption, than the random pore structures of the more easily prepared synthetic porous
implants.
A new way of making strong porous ceramics with interconnected porosity has been
developed by De Aza et al. [93-98]. This is based on designing dense bioactive ceramic
materials with the ability to develop an in situ porous hydroxyapatite-like (HA) structure
when they are implanted into a living body. The material was composed of two phases,
pseudowollastonite (-CaSiO3 = psW) of bioactive properties, and -tricalcium phosphate
(-TCP) of resorbable properties. Taking into account the bone structure and the eutectic
nature of the wollastonite - tricalcium phosphate system [99], the microstructure of the
material was developed by slow solidification (0.5C/h) of the eutectic composition (60 wt%
psW and 40 wt% -TCP) through the eutectic temperature region (1402 3C).
Due to the volume fraction of the two phases at eutectic point (psW = 0.61 and a-TCP =
0.38) and their high dimensionless entropy of fusion >2, where is defined as = Sf /R ; Sf
= entropy of fusion, R = gas constant; psW = 3.71 and -TCP = 9.66 [100-103] the obtained
microstructure was of irregular eutectic structure. Figure 4 shows a SEM image of the
microstrutrure of the eutectic material (Bioeutectic), which consists of quasispherical
colonies formed by alternating lamellae of pseudowollastonite and -TCP.
Figure 4. SEM image of the microstructure of the eutectic material
The material of this composition was initially exposed to simulated body fluid (SBF) for
one month and the cross section of the specimen is shown in Figure 5. Two HA-like layers
with different morphologies were found to form below and on the surface of the material, by
two different and consecutive mechanisms. At first the material reacts with the surrounding
SBF by dissolving the psW phase and forming a porous structure of HA-like phase that
mimic porous bone, in a pseudomorphic transformation of the -TCP, according to the
reaction:
3[Ca3(PO4)2] + Ca2+ + 2(OH)- Ca10(PO4)6(OH)2
During a second stage a dense HA-like layer is formed by precipitation on the outer
surface of the material according to the reaction:
6[HPO4=] + 10 Ca2+ + 8(OH)- Ca10(PO4)6(OH)2+ 6(H2O)
Figure 5. SEM image of a polished cross-section of the eutectic material after three weeks immersed in
SBF
The first stage of the reaction mechanism, which at the beginning is very fast, is slowed
down due to the combined effect of the difficulty of diffusion of the H3O+ from the SBF to
reach the reaction interface as the distance is increased and due to the closure of the channels
as HA is formed on the surface of the simple.
Conversely, the last stage of the reaction does no occur when the eutectic sample is
immersed in slow stream of SBF (50 cc/h) instead of using a static solution. Then the
complete transformation of the eutectic material takes place given rise to a HA artificial
porous bone. Figure 6A show a SEM image of a fresh fracture surface of the eutectic material
after immersion in a slow stream of SBF for three weeks and for comparison in figure 6B
shows the SEM picture of a human bone.
In parallel to this study, in vivo experiments are being carried out and it is expected
that this material might behave in a similar way as in vitro tests, facilitating the
osteointegration of the implant.
The procedure developed by the authors opens the opportunity to obtain a new family of
bioactive materials for which the general name of Bioeuctectics is proposed by the authors.
The material may present an interesting alternative in biomedical applications. Bioeutectic
material is the only bioactive material, at present, which is totally colonized in simulated
body fluid.
116 P. N. De Aza
A B
Figure 6. (A) SEM image of a fresh fracture surface of the Bioeutectic material after immersion in a
slow stream of SBF for three weeks. (B) SEM image of a human bone.
Bioactive Glasses
Since the discovery by Hench, at the beginning of the Seventies, of which some glasses
belonging to the SiO2-CaO-NaO-P2O5 system do not induce the formation of fibrous tissue
capsule that surround implant materials [45], but the bioactive glasses have chemical
reactivity with the physiological system and form chemical bond between the adjacent tissue
and the surface of implant materials various types of glasses and glass-ceramics have been
developed and used as bone replacement.
Probably the divided point, to select the glass, is that it would not have to contain strange
or poisonous elements to the organism and between the components would have to be
calcium and the phosphorus, because both are the greater components of the mineral phase of
the bone tissue.
The first bioactive glass (Bioglass) developed by Hench and col., was the denominated
Bioglass 45S5, which presents a great bioactivity, and bond to soft and hard tissues. It is also
the composition of bioglass more studied. Its composition in % in weight is the following:
45% SiO2, 24.4% of CaO, 24.5 NaO and 6% P2O5 [45].
The composition of the Bioglass 45S5 was studied in "in vitro" and "in vivo" with three
different microstructures: amorphous, partially crystalline and totally crystalline. All the
implants were bond to the cortical bone, in a model of rat femur after 6 weeks. There was no
difference between amorphous implants and crystalline implants. [46].
Taking the Bioglass 45S5 as the base composition, Hench and col., developed and
studied a great series of glasses within the quaternary system SiO2-CaO-NaO-P2O5,
maintaining in all of them a constant content of 6% in weight of P2O5. Some of these
compositions are in Table IV.
In the initial stages of the development of the bioglasses, one thought that the presence of
P2O5 in the composition was essential so that the glasses were bioactive. Nevertheless, Li and
col [104] have observed that the SiO2-CaO-NaO glasses free of P2O5, obtained by sol-gel
method, were bioactive even until 85 mol % in SiO2. Also, Ohura and col. [70] have observed
that a binary glass of SiO2-CaO named CS, with a composition in weight percent of: 48.3
CaO and 51.7 SiO2, formed with the bone tissue a stable bond. Later, De Aza and col. [75-
77,79] demonstrated, for the first time, that a crystalline material, different from the
hydroxyapatite and free of P2O5, the wollastonite (CaSiO3), in their high and low temperature
phases, display a high bioactivity in simulated body fluid and in parotidea human saliva, and
even bond to bone tissue [80,81].
Table IV- Glass Compositions in Weight %
Biovidrio SiO2 CaO Na2O P2O5 B2O3 F2Ca K2O

45S5 45 24.5 24.5 6.0
45S5F 43 12 23 6.0 16
45B15S5 30 24.5 24.5 6.0 15
45B5S5 40 24.5 24.5 6.0 5
KCP1 45 24.5 6.0 24.5
45S5-N 50 24.5 19.5 6.0
45S5-C 50 19.5 24.5 6.0
From the point of view of their composition, it can be say that, most of glasses that are
able to bond to the in vivo tissue contain phosphates, and the totality includes silicates with
the purpose of providing a matrix with low solubility, since the single phosphate glasses are
excessively soluble. All of them can be divided in two great groups: a) glasses rich in alkaline
(glasses of Hench and col.), with alkaline oxide contents greater to 20% in weight and b) poor
in alkaline, containing not more of 5% in weight of these oxides.
Of the all exposed is deduced that exists a generous rank of compositions of bioactive
glasses. This allows, in principle, to adapt the reactivity of glasses for different clinical
applications. This versatility of the glass materials is a remarkable characteristic that
difference them of the other biomaterials.
Bioactive glasses are produces by conventional glass manufacturing methods. Purity of
raw materials must be assured, since its quality affects greatly the properties of the obtained
glass. Thus, usually quartzes or silica sands of great purity are used, sodium or potassium
carbonate reagents, in some cases wollastonite (CaSiO3) of high purity, calcium phosphates
without adsorption of water etc.
The materials weighed and mixed are melts in Pt or Pt/Rh crucibles, homogenizing the
melt and paying attention to the possible loss by volatilisation of some components of high
steam tension such as NaO2 or P2O5. Depending on the composition, the temperatures of
melting can oscillate between 1200 and 1450 C. The melt is cast in graphite or steel molds,
since they do not contaminate and not adhere to the glass. The annealing of the glass is
fundamental, since all the bioglasses usually have high coefficients of thermal expansion,
changing the temperature of annealing, according to the composition of the glass, between
400 and 550 C approximately.
In general, the bioglasses are relatively soft glasses, reason why they can be easy get the
required dimensions by machining, using generally, diamond discs or dental handpieces with
abundant refrigeration. When powdered or granulated bioglasses are required, the glass can
be ground to the wished sizes or the melt can be quenched directly in water, obtaining what it
is denominated glass frit and facility the milling. The glass frit must be dry quickly to avoid
the possible hydrolytic corrosion.
Since it has been exposed previously, the essential requirement for an artificial material
to bond to living bone is believed to be the formation of a biologically active apatite-like
118 P. N. De Aza
layer on their surfaces in the body. The surface chemical reactions results in the formation of
a carbo-hydroxyapatite (CHA) being this, perhaps, the only common characteristic of all the
bioactive materials that are known.
The mechanism by which the CHA layer forms on the exposed bioglass to physiological
surrounding medium was proposed for the first time by Hench [7] and it consists of the
following stages:
1) Leaching by ionic interchange of proton of the physiological medium by labile ions
(Na+, K+, Ca2+, Mg2+, etc.).
Si O Na+ + H+ + OH Si OH+ + Na+ + OH
The rate of reaction, as well as the formation of the silanol groups (Si OH) in the
interface bioglass/solution is controlled by a diffusion process and exhibits a t- dependence.
The cationic interchange increases the concentration of hydroxyl (OH-) groups in the
interface bioglass/solution, giving rise to an increase of pH until levels of the order of 10.5, as
it has been demonstrated by De Aza and col. by measures of interfacial pH with a
microsensor (ISFET) in different bioactive materials [105].
2) The increase of pH facilitates the dissolution of the network and formation of
additional silanol groups, according to the reaction:
- Si O Si - + H O H 2 [Si OH+]
As well as the release of silica into the solution in the form of Si(OH)4.
3) Polymerization of the silica (SiO2) rich layer by condensation the neighbouring groups
of Si-OH, giving rise to a rich amorphous silica layer
O O O O
| | | |
O - Si - OH + HO - Si - O O - Si - O - Si - O + H2O
| | | |
O O O O
4) In this stage, one takes place an ion migration Ca2+ and PO43- to the surface of the rich
silica layer, forming an amorphous layer rich in CaO-P2O5 followed by the growth of this
layer due to the incorporation of Ca2+ and PO43- coming from the solution.
5) Cystallization of the amorphous CaO-P2O5 layer by incorporation of OH, CO32, or F
anions from the solution to form a carbonateapatite or fluorapatite layer.
As opposed to the mechanism described by Hench, nowadays seems to admit the
unnecessary migration of calcium and phosphate ions through the amorphous silica layer
(stage 4) to facilitate the nucleation of calcium ions and phosphate ions coming from
physiological surrounding medium, especially when Ohura and col [70] and De Aza and col.
[75-77,79] have demonstrated the formation of the CHA layer on CaO-SiO2 glasses and
wollastonite (CaSiO3) respectively. Both materials are exempts of phosphate ions, and form a
chemical bond with the bone.
Finally, another hypothesis, perhaps simpler, suggested by De Aza and col. [105] is that
pH interface is the responsible of the precipitation, from the solution, of Ca2+ ions and PO43-
ions. This hypothesis seems to justify itself by the fact that the amorphous silica layer
between the layer of CHA and the vitrocermic of A/W developed by Kokubo and col. [68,69]
neither in the Bioeutectics materials of psW/-TCP obtained by De Aza and col. [93-98] has
never been found.
Bioactive Glass-Ceramics
Glass-ceramics materials are those material obtained from a glass by suitable heat treatments,
with the objective to produce the nucleation and growth of certain crystalline phases that are
immersed in the residual vitreous matrix.
There are two fundamentals reason why there is a great interest by the bioactive glass-
ceramics materials for biomedical applications. First, it can be obtained very complex forms
of products or devices made by glass-ceramics, since the initial stage is to obtain a glass, and
like so, the well-known technologies of the glass production, economic and precise, such as
quenching, blowing, pressed or laminated can be used. Second it is that, as a result of the
crystallization, the glass-ceramics products usually display a very fine microstructure with
very small or null residual porosity. Such microstructures tend to provide better mechanical
properties of the final product or device.
Nevertheless, the crystallization process requires a strict knowledge of the mechanisms of
nucleation and growth of the crystalline phases. Figure 7 represents a scheme of the
nucleation rates and growth of the crystalline nuclei based on the temperature.
The rate of growth of the crystalline nuclei (length by time unit) shows a maximum,
normally about 100 C below the temperature of liquidus (TL). As it can be observed, there is
normally a big margin of temperatures below the temperature of liquidus (TL) where does not
detect.
Vn
Vc
Vn or Vc
Tg Temperature TL
Figure 7.- Scheme of the nucleation rate (Vn) and growth (Vc) respectively, based on the temperature.
TL = temperature of liquidus; Tg = vitreous transition temperature
After obtained the glass to room temperature, it is heated up until the optimal temperature
of nucleation. Figure 8 shows a typical scheme of a process in two stages to produce a glass-
ceramics. The rate of heating is not critical, but gradients of temperatures must be avoided.
The normal rates usually are 2 to 5 C/minute, but it depends of the product of start with. It is
advisable a high speed of nucleation, preferably near the maximum, whose temperature stays
120 P. N. De Aza
constant by a certain time. Alternatively, after the glass is conformed, it cans cold directly
until the temperature of nucleation.
The following stage is critical. The temperature must be increased slowly, normally not
more than 5 C/minute, trying to avoid the generation of tensions caused by the changes of
volume that take place during the crystallization, which could even originate cracks or the
breakage of the material. Slow rates of heating in this stage, give rise that the possible
tensions generated by the volumetric changes, are eliminated by viscous flow of the residual
glass, and as the crystalline fraction increases the material becomes progressively more rigid.
Reached the temperature of "growth", this stays during a time until reaching the wished
degree of crystallinity. Finally the glass-ceramic is cold until room temperature.
Temperature
growth
nucleation
Time
Figure 8. - Simplified scheme of the process of heat treatment in two stages for the preparation of a
glass-ceramic material.
Depending on the composition of the glass, the crystallization sometimes takes place
from the surface and not distributed homogenous in the volume, giving rise to a glass-
ceramics of low resistance. In order to help the nucleation in volume and to obtain an
appropriate microstructure it is frequently the use of additives. They are generally metallic
particles (Cu, Ag, Au, Pt, etc.) and other times metallic oxides (SnO2, Ce2O3, TiO2, etc.). In
the field of the bioactives glass-ceramics for biomedical applications, the P2O5, component of
all products and devices developed commercially, acts like nucleate agent in the same way
that the TiO2.
Most of the glass-ceramics for biomedical applications are based on compositions similar
to the bioglasses of Hench (Bioglass), although in all of them present very low contents in
alkaline oxide. Table V shows the analyses and characteristics of the bioactive glass-ceramics
with clinical applications, compared with the Bioglass 45S5.
Ceravital
The first glass-ceramics material for clinical applications was developed by H. Brmer and E.
Pfeil in 1973 [108,109] called Ceravital. This name includes a number of compositions. In its
origins Ceravital were considered very optimistically, to even replace bones in zones of load
and teeth. Nevertheless, like it is possible to be observed, in mentioned Table V, its
mechanical properties (flexural, tensile strength) are below of the 160 MPa that presents the
cortical human bone.
On the other hand, long in vivo experiments put in doubt the stability of these materials,
although, Gross and Strunz [48,64] developed other compositions with the intention to
diminish the solubility of the glass-ceramics, using metal oxides like nucleates agents.
Ceravital bioactivity index (BI = 5.6) is the half of the Bioglass 45S5 (BI = 12.5).
Nowadays, the only field in which Ceravital are used is in the replacement of the ossicular
chain in the middle ear, where the loads are minimal and therefore the mechanical properties
of material are sufficient.
Table V - Characteristics of the Bioactives Glass-Ceramics with Clinical Applications,

Compared with the Bioglass 45S5.
Compositin Bioglass Ceravital Cerabone Ilmaplant Bioverit
(% weight) 45S5. A/W
Na2O 24.5 5-10 0 4.6 3-8
K2O 0 0.5-3.0 0 0.2 0
MgO 0 2.5-5 4.6 2.8 2-21
CaO 24.5 30-35 44.7 31.9 10-34
Al2O3 0 0 0 0 8-15
SiO2 45.0 40-50 34,0 443 19-54
P2O5 6.0 10-50 6.2 11.2 2-10
CaF2 0 0 0.5 5.0 3-23
Phases Glass Apatite + Apatite + Apatite + Apatite +
Glass - - Flogopite +
wollastonite wollastonite Glass
+ Glass + Glass
Flexural, tensile 42 100 150 220 170 100 160
strength. (MPa)
Compressive n.d 500 1060 n.d 500

strength (MPa)
Young Modulus 35 n.d 117 n.d 70 - 88

(GPa)
Cerabone A/W
One of the glass-ceramics with more clinical success, is probably the denominated A/W on
the basis of their constituent crystalline phases: apatite (oxyfluorapatite = Ca10(PO4)6(O,F2)
and wollastonite (-CaSiO3), developed by Kokubo and col. [60,68,69,110]. This material is
commercially known as Cerabone A/W.
After several insolvent attempts to obtain a glass-ceramics in the pseudoternary system
Ca3(PO4)2 CaSiO3 MgCa(SiO3)2, starting off a monolithic glass and even from powder
glass, the A/W glass-ceramic could be obtained adding small amounts of CaF2 to the
composition of the parent glass (Table V). The glass was ground until 5m in average size,
isostaticaly pressed (200 MPa) in the wished forms and heated up to 1050C; precipitating
successively at 870 and 900 C the phases oxyfluorapatite and -wollastonite respectively.
Obtaining therefore a glass-ceramic with a very fine microstructure, free of cracks and pores.
The special developed microstructure causes that this glass-ceramic presents the best
mechanical properties (f = 220 MPa) of all the materials exposed in Table V. Almost the
122 P. N. De Aza
twice that (115 MPa) of dense hydroxyapatite and even offers a resistance greater than the
human cortical bone (160 MPa). Its fracture toughness is of 2.0 MPa m and its Vickers
hardness is approximately 680 [68,111,112].
Like in other bioactive glass-ceramics, when glass-ceramic A/W is immersed in
simulated body fluid the same type of carbo-hydroxyapatite (CHA) layer is formed on the
surface. The compositional and structural characteristics of this CHA layer are similar to
those of the apatite in the natural bone and its expected that osteoblast would proliferate on
the surface of the CHA layer. This suggests that this CHA layer plays and essential role in
forming the chemical bond of the glass-ceramics to the bone.
Nevertheless, in contrast to the bioglasses, in glass-ceramics A/W it has not been possible
to find the amorphous silica layer, between the layer of CHA and glass-ceramics A/W, even
not at level of high resolution transmission electronic microscopy. Nevertheless Kokubo and
col., consider that the silanol groups formed on the surface of the glass-ceramic are the
responsible for the CHA formation, because provides favourable sites for apatite nucleation
and growth.
Due to its good properties of biocompatibility, bioactivity and mechanical properties
glass-ceramics A/W is used from the Eighties, for the reconstructions of the iliac crests,
artificial vertebrae, intervertebral discs. Glass-ceramics A/W can easily machined into various
shapes using discs and reels of diamond, even into screws and in powder form can be used to
repair and fill bone defects.
Ilmaplant-L1
Berger and col [61] prepared another type of glass-ceramics of apatite/wollastonite called
Ilmaplant-L1. As it is possible to appreciate in Table V, the fundamental difference with the
glass-ceramics A/W is the alkaline contents and the greater proportions of CaF2, SiO2 and
P2O5 and minor content of CaO. Due to its smaller mechanical resistance to flexion (Table V)
its use has been restricted to maxilofacials implants.
Bioverit
In 1983 Holand and col. [113] at the University of Jena, developed a series of bioactives
glass-ceramics that they denominated under the name of Bioverit type I (Table V). These
glass-ceramics are easily machined into various shapes with standard tools and even
retouched in the operating theatre. They have a very complex composition and they are
obtained from of a silico-phosphate glass type within the SiO2-(Al2O3)-MgO-Na2O-K2O-F-
CaO-P2O5 system. Its process of obtaining consists of generating a phase separation glass by
means of the control of the nucleation, crystallization and later growth of the crystalline
phases by heat treatments between 610 and 1050 C. The product thus obtained is formed by
a residual vitreous phase, and a mixture of apatite crystal (1 - 2m) and mica of the
fluoroflogopite type (Na/K Mg3[AlSi3O10F2]) that is the responsible of the easy mechanized
of the glass-ceramic.
Later, such authors developed another family of glass-ceramic, also machineable, called
Bioverit II consisting of a new type of curved mica of fluorophlogopite-type as the main
crystal phase and secondary crystals of cordierite (Mg2[Si5Al4O4]). The basis for the
nucleation and crystallization is also the phase separation of the glasses which has been
caused by Na2O, K2O and F- additions to the glass.
Finally, Vogel and col.[114,115] developed another new family of Bioverit, called type
III from a phosphate glass without any silica content. The processing starts form phosphate
invert glasses of the P2O5-Al2O3-CaO-Na2O system, which structure is formed only by mono
and diphosphate units doping it with Fe2O3 o ZrO2. The glass-ceramic thus obtained is formed
by a residual vitreous phase, and a mixture of apatite crystal as main phase, berlinite (AlPO4)
and a complex phosphate structures of varulite type (phosphate of Na-Ca-Fe).
More that 1000 implants of Bioverit have been successfully used as bone substitution
by the middle of the Nineties. In these case we include orthopaedic surgery (replacement of
vertebrae, reconstruction of the root of the acetabulum in a dislocated hip, osteotomy of the
tibial head, partial replacement of vertebrae, etc) and head and neck surgery (reconstruction
of the posterior wall of the auditory canal, reconstruction of the skull base, rhinoplasty, etc).
Bioactive Composites
As it has been exposed, the fundamental limitation by the application of the ceramic, glass
and glass-ceramics are their low mechanical properties. In order to try to avoid these
limitations was proposed the use of these materials as part of a composite material.
Bioceramic composite synthesis serves the purpose of enhancing the mechanical behaviour of
the parent ceramic, glass or glass-ceramic, while maintaining its excellent biocompativility
and bioactivity.
Bioceramic composites can be divided in bioiner, bioactive and biodegradable and the
ceramic phase can be the reinforcing material, the matrix or both.
There are several bioceramic composite which have been manufactures and used in
clinical applications, the first bioactive composite was stainless steel/bioactive glass.
The preparation of a metal fiber/bioactive glass composite includes preparation of the
fiber, impregnating the perform with glass matrix, and heat treating the composite. For a
effective stress transfer between the matrix and the reinforced metal fibers, there must be a
good bond between the glass and the metal fibers. This is achieved through the oxidation of
the metal fibers before immersion in the molten glass.
The titanium fiber bioactive glass composite represents a fundamental advance in
biomaterials, which derives from consideration of the failure mode of current implants with
bioactive coatings as bonding vehicles. Dental implants or hip prostheses with plasma
sprayed calcium phosphate coatings are known to fail at the ceramic to metal interface. The
advent to the titanium fiber-glass composite represents a new era in that continuity between
metal implant and bioactive coating may be achieved, using fibers that are sintered onto the
implant substrate and the run throughout the coating up the surface [116,117].
Ducheyne and Hench [118] developed a composite made by Bioglass 45S5 and
stainless steel AISI 316L metal fibers. This type of biomaterial is obtained by a process of
immersion of metallic fibers in the melted glass. The composite material thus obtained has a
greater mechanical resistance, better ductility and a similar Young modulus to the human the
cortical bone.
Other composite material composed with a glass-ceramic is Ceravital reinforced with
titanium particles [119] and A/W glass-ceramic reinforced with particles of zirconia partially
stabilized [120].
124 P. N. De Aza
Some of the properties of these composite materials that make them apt for their use as
dental implant or in orthopaedic surgery are: low modulus of elasticity; good resistance to the
stress; good resistance to the impact and easily machineable.
Biodegradable or Resorbable Ceramics

The resorbibles ceramics began to be used in 1969. These types of bioceramics dissolve with
time and are gradually replaced by the natural tissues. A very thin or non-existent interfacial
thickness is the final results. They would be the ideals implants, since only remain in the body
while its function is necessary and disappear as the tissue regenerates. Their greater
disadvantage is that their mechanical strength diminishes during the reabsorcion process.
Consequently, the function of these materials are to participate in the dynamic process of
formation and resorpcion that takes place in bone tissues; so they are used like scaffolding or
filling spaces allowing to the tissues their infiltration and substitution [121].
All the resorbibles ceramics, except plaster (CaSO4H2O), are based on calcium
phosphates, varying their biodegradability in the sense:
-TCP > -TCP >>>> HA
The rate of biodegradation is increased, as it is logical, as: a) the specific surface

increases (the powders are more quickly biodegraded that porous solids and these more than
dense solids); b) when the crystallinity decreases; c) when the grain and crystal size decrease;
d) when there are ionic substitutions of CO32-, Mg2+ and Sr2+ in the HA.
The factors that tend to decrease the rate of biodegradation include: a) substitution of ion
F- in the HA; b) Mg2+ substitution in the - TCP and c) lower - TCP/HA ratios in two-phase
compounds.
The biodegradation or resorption of calcium phosphates is caused by three factors:
1) Physicochemical dissolution, which depends on the solubility product of the material

and pH of its surroundings. New phases can be formed in the surface, such as
amorphous calcium phosphates, dihydrated dicalcium phosphate, octacalcium
phosphate and even HA replaced in an anionic way.
2) Physical disintegration in small particles as result of a preferential attack to the grain
boundaries.
3) Biological factors, such as phagocytosis, which gives rise to a local diminution of pH,
the cellular activity and the site of implantation.
One of the few bioceramics that satisfy partially these requirements is the tricalcium
phosphate (TCP).
Tricalcium Phosphate (Ca3(PO4)2)
The tricalcium phosphate (TCP) is the biodegradable phosphate par excellence. TCP has a
stoichiometric formula Ca3(PO4)2 with a Ca/P molar ratio = 1.15. The TCP is a neutral
compound in which the 6 positive charges of the Ca+2 ions are compensated by 6 negative
charges of the anions PO43-. It belongs to the family of the Whitlockites that respond to
general formulate (Ca,Mg)3 (PO4)2, so that the calcium can be replaced partial or totally by
magnesium.
The TCP is a material that displays polymorphism. Exist three polymorphic stated that
are known like the phases: , , `, from low to high temperature of stability respectively
[122,123].
The phase crystallizes in the hexagonal system, and their characteristic parameters of
network are: a = 10.429 and c = 37.38. == 90 and =120. The phase crystallizes in
the orthorhombic system, and their characteristic parameters of network are: a = 15.22 , b =
20.71 and c = 9.109. == = 90. The phase crystallizes in the monoclinic system,
with lattice parameters: a = 12.887, b = 27.280 and c = 15.219. == 90 and =126.2
[124].
The phase transition ' at 1475 5C is totally reversible in both senses. Also it
happens to the transition at 1150 10C. When cooling below this temperature the
inverse transformation is due to produce , nevertheless if the cooling in very fast the
phase is metastably preserved at room temperature.
Like the synthesis of HA, the synthesis of -TCP, in aqueous solution, is affected by
numerous physical and chemical parameters, which originates great difficulty to obtain a
product 100% pure. However, pure -TCP can be easy obtained by solid state reaction of
calcium and phosphorous compounds with relationships Ca/P = 1.5.
Applications are the temporary fillings and periodontal defects.
Acknowledgment
The author thanks to CICYT the financial support of the Project MAT2003-08331-C02-02
References
1. Williams, D.F. (1987). Definitions of Biomaterials. Amsterdam: Elsevier.
2. Shackelford, J.F. (1999). Bioceramics: Current status and future trends. Mater. Sci.
Forum 293, 99-106.
3. Dubok, V.A. (2000). Bioceramics - Yesterday, today, tomorrow. Powder metall. met
C+ 39 (7-8), 381-394.
4. Vallet-Regi, M. (2001). Ceramics for medical applications. J Chem Soc Dalton (2), 97-
108.
5. Williams, D.F. (1985).The Biocompatibility and Clinical Uses of Calcium Phosphate
Ceramics. In D.F. Williams editors, Biocompatibility of Tissue Analogs (II 43-66). Boca
Raton, FL: CRC Press.
6. Hulbert, S.F., Bokros, J.C., Hench, L.L., Wilson, J. & Heimke, G. (1987). Ceramics in
Clinical Applications: Past, Present and Future. In P. Vincenzini editor, High Tech
Ceramics (189-213). Amsterdam, The Netherlands: Elsevier.
126 P. N. De Aza
7. Hench, L.L. (1991). Bioceramics: From Concept to Clinic. J. Am. Ceram. Soc., 74 (7),
1487-510.
8. Hulbert, S.F., Hench, L.L., Forces, D., & Bowman, L. (1983). History of bioceramics.
In P. Vincenzini editor, Ceramics in Surgery (3-29). Amsterdam, The Netherlands:
Elsevier.
9. Hench, L.L. (1975). Prosthetic implant materials. Ann. Rev. Mater. Sci. 5, 279-300.
10. De Groot, K., (1991). Medical Applications of Calcium Phosphate Bioceramics. The
Centennial Memorial Issue, 99 (10), 943-53.
11. Hayashi, K., Matsuguchi, N., Uenoyama, K, et al. (1992). Re-evaluation of the
biocompatibility of bioinert ceramics in vivo. Biomaterials 13 (4), 195-200.
12. Hulbert, S.F. & Klawitter, J.T. (1970). Potential of ceramic materials as permanently
implantable skeletal prostheses. J. Biomed. Mater. Res. 4, 433-456.
13. De Groot, K. (1988). Effect of Porosity and Physicochemical Properties on the Stability
Resorption and Strength of Calcium Phosphate Bioceramics. Ann. N.Y. Acad. Sci., 523,
227.
14. Cheung, H.S. & Haak, M.H. (1989). Growth of osteoblasts on porous calcium
phosphate ceramic: an in vitro model for biocompatibility study. Biomaterials 10 (1),
63-67.
15. Christel, P., Meunier, A., Dorlot, J.M., Crolet, J.M., Witvolet, J., Sedel, L. & Boutin,
P.M. (1988). Biomechanical compatibility and design of ceramic implants for
orthopedic surgery. In P. Ducheyne & J.E., Lemons editors, Bioceramics: Material
Characteristics Versus In Vivo Behaviour (Vol. 523, 237) New York. Eds. Annals of
New York Academy of Sciences.
16. Howlett, C,R.,, Zreiqat, H., Odell, R., et al. (1994). The effect of magnesium-ion
implantation into alumina upon the adhesion of human bone-derived cells. J Mater Sci-
Mater Med 5 (9-10), 715-722.
17. Ratner, B. D., Hoffman, A. S., Schoen, F.J., & Lemon, J. E. (1996). Biomaterials
Science. An introduction to Materials in Medicine. Publ. Academic Press. ISBN 0 12
582461 0
18. Sedel, L., Bizot, P., Nizard, R., et al. (2000). Long term clinical results of total joint
replacement with alumina/alumina articulation. Key Eng Mat 192-1, 969-973.
19. Lancaster, J. G., Dowson, D., Isaac, G.H., et al. (1997). The wear of ultra-high
molecular weight polyethylene sliding on metallic and ceramic counterfaces
representative of current femoral surfaces in joint replacement. P I Mech Eng H 211 (1),
17-24.
20. Bragdon, C.R., Jasty, M., Kawate K., et al. (1997). Wear of retrieved cemented
polyethylene acetabula with alumina femoral heads. J Arthroplasty 12 (2), 119-125.
21. De Santis, E., Maccauro, G., Proietti, L., et al. (2000). Histological and ultrastructural
analysis of alumina wear debris. Key Eng Mat 192-1, 995-998.
22. Schulce, W. (1984) The intra-osseous Al2O3 (Frialit) Tbingen implant, development
status after eight years. Quintessence International 15, 1-39.
23. Schulte, W., Busing, B. & Heimke, G. (1980). Endosseous implants os aluminium oxide
ceramics. A 5 year study in human. In Kawahara Edit. Proceedings of International
Congress on Implantology and Biomaterials in Stomatology. (157-167). Kyoto, Japan.
24. Clark, A. E. & Anusavice, K. J. (1991). Dental applications. Engineered Materials

Handbook: Ceramics and Glasses (Vol. 4, Sec. 14, 1091-99) Park, Ohio: ASM
International, Materials.
25. Kawahara, H., Hirabayashi, M. & Shikita, T. (1980). Syngle cristal alumina for dental
implants and bone screws. J. Biomed. Mat. Res. 14, 597-605.
26. Yamagami, A. (1988). Single crystal alumina dental implant. 13-year long following-
ups and statistical examination. In H. Oonishi, H. Aoki &K. Sawai Editors. Proceedings
of 1st International Symposium on Ceramics in Medicine. (332-337).Ishiyaku Euro-
America Inc.
27. Verneuil Process. En Encyclopdia Britanica, Micropedia Vol. 12, pg. 324,
28. Garvie, R. C., Urban, D. Kennedy, D. R. & McMeuer, J. C. (1984). Biocompatibility of
Magnesium Partially Stabilized Zirconia (Mg-PSZ Ceramics). J. Mat. Sci, 19, 3224.
29. Taeishi, T. & Yunoki, H. (1987). Research and Development in Advance Biomaterials
and Application to the Artificial Hip Joint. Bull. Mech. Engr. Lab. 45, 1.
30. Stevens, R.(1986). An Introduction to Zirconia. Publ. Magnesium Elektron Ltd.
31. Pascual, C. & Durn, P. (1983).The System ZrO2-Y2O3. J. Am. Ceram. Soc., 66 [1], 23-
27.
32. Butler, E.P. (1985).Transformation Tughened Zirconia Ceramics. Critical Assessment.
Mat. Sci. and Tecnlogy, (Vol. 1, 417-432).
33. Hulbert, S. F. (1993). The use of alumina and zirconia in surgical implants. In L.L.
Hench &J. Wilson editors. Introduction to Bioceramics, (25-40) Publ. World Scientific.
34. Josset, Y., Oum'Hamed, Z., Dupont, C., et al. (2000). Examination of zirconia, alumina
ceramics and titanium interactions on human osteoblasts in culture. Key Eng Mat 192-1,
329-332.
35. Nkamgueu, E. M., Adnet, J. J., Bernard, J., et al. (2000). In vitro effects of zirconia and
alumina particles on human blood monocyte-derived macrophages: X-ray microanalysis
and flow cytometric studies. J Biomed Mater Res 52 (4), 587-594.
36. Serra, E., Tucci, A., Esposito, L., et al. (2002). Volumetric determination of the wear of
ceramics for hip joints. Biomaterials 23 (4), 1131-1137.
37. Ministre de la Sant, de la famille et des Personnes Handicapes. Le Journal Officiel
NOR: SANM0222500S, 22 Juillet 2002.
38. De Aza, A. H., Chevalier, J., Fantozzi, G., Schehl, M. & Torrecillas, R. (2002). Crack
growth resistance of alumina, zirconia and zirconia toughened alumina ceramics for
joint prostheses. Biomaterials 23, 937-945.
39. Chevalier, J., De Aza, A. H., Fantozzi, G., Schehl, M. & Torrecillas, R. (2000).
Extending the lifetime of ceramic orthopaedic implants. Advanced Materials. 12 n21,
1619-1621.
40. Mitura, S., Niedzielski, P., Jachowicz, D., et al. (1996). Influence of carbon coatings
origin on the properties important for biomedical application. Diam Relat Mater 5 (10),
1185-1188 .
41. Cui, F.Z. & Li, D.J. (2000). A review of investigations on biocompatibility of diamond-
like carbon and carbon nitride films. Surf Coat Tech 131 (1-3), 481-487.
42. Glasmacher, B., Nellen, E., Reul, H., et al. (1999). In vitro hemocompatibility testing of
new materials for mechanical heart valves. Material Wiss Werkst 30 (12), 806-808.
43. Lee, S.Y., Chiang, H.C., Lin, C.T., et al. (2000). Finite element analysis of thermo-
debonding mechanism in dental composites. Biomaterials 21 (13), 1315-1326.
128 P. N. De Aza
44. Lewandowska, S., Zumiel, M., Komender, J., Gorecki, A., et al (1997). Fixation of
carbon fibre-reinforced carbon composite implanted into bone. J Mater Sci-Mater Med
8 (8), 485-488.
45. Hench, L.L., Splinter, R.J., Allen, W.C. & Greenlee, T.K. (1971). Bonding mechanisms
at the interface of ceramic prosthetic materials. J. Biomed. Mater. Res. Symp. 2 (1), 117-
41.
46. Hench, L.L., & Wilson, J. (1984). Surface-active biomaterials. Science, 226, 630.
47. Hench, L.L. (1988). Bioactive ceramics. In Ducheyne, P. Lemons, J.E. editors.
Bioceramics: Materials Characteristics Versus In Vivo Behaviour, (Vol 523, 54) New
York: Annals of New York Academy of Science
48. Gross, U., Kinne, R., Schmitz, H.J. & Strunz, V. (1988). The response of bone to
surface active glass/glass-ceramics (4 (2)) CRC. Crit. Rev. Biocompat.
49. Williams, D.F. (1987). Tissue-Biomaterial Interaction. J. Mater. Sci., 22, 3421-3445
50. Deer, Howie & Zussman. (1972). Non-silicates: sulohates, carbonates, phosphates and
Halides. An introduction to Rock-forming minerals. (volume 5B: Second edition). John
Willey & Sons.
51. Young, R.A. (1973). Some aspects of crystal structural modelling of biological apatites.
Coll. Int. CNRS n 230, 21-40.
52. Kerebel, B., Daculsi, G. & Verbaere, A. (1976) High resolution electron microscopy
and crystallographic study of some biological apatites. J. Ultrastruc. Res. 57, 266-275.
53. Jackson, S., Cartwright, A.G. & Lewis, D. (1978). The morphology of bone minerals
crystals. Calcif. Tissue Res. 25, 217-222.
54. Clries, L., Fernndez-Pradas, J. M. & Morenza, J.L. (2000). Bone growth on and
resorption of calcium phosphate coatings obtained by pulsed laser deposition. J. Biomed.
Mater. Res. 49, 43-52.
55. Fernndez-Pradas, J.M., Clries, L., Sardin, G. & Morenza, J.L. (2002).
Characterization of calcium phosphate coatings deposited by Nd:YAG laser ablation at
355 nm : influence of thickness. Biomaterials 23, 1989-1994.
56. Hayek E. & Neweseley H. (1963). Pentacalcium monohydroxyorthophosphate
(hydroxlapatite). Inorganic Syntheses; 7, 63-65.
57. De Groot, K., Klein, C.P.A.T., Wolke, J.G.C. & De Bliek-Hogervorst, J. (1990).
Chemistry of calcium phosphate bioceramics. In Yamamuro, T., Hech, L.L. & Wilson,
J., Editors. Handbook of Bioactive Ceramics, Vol. II, Calcium phosphate and
hydroxylapatite Ceramics (3-15). Boca Ratn, FL.CRC Press.
58. Yubao L., Klein C.P.A.T., de Wijn, J., Van de Meer, S. & de Groot, K. (1994). Shape
change and phase transition of nedddle-lie non-stoichiometric apatite crystals. J. Mater.
Sci. Mater. Me.,: 263-28.
59. Blenke, B.A., Pfeilbromer, E., H. & Kas, H.H. (1973). Langenbechs Arch. Chir. Suppl.
Forum 107.
60. Kokubo. T. (1993).Novel Ceramics for Biomedical Applications. In P. Durn & J.F.
Fernandez editors. Third Euro-Ceramics. (Vol. 3, 1-16).
61. Berger, G., Sauer, R., Steinborn, G., Wihsmann, F., Thieme, V., Kholer, St. & Dressel,
H. (1989). Clinical application of surface reactive apatite/wollastonite containig glass-
ceramics. Mazurin & O.V., Nauka, editors. Proceedings of XV International Congress
on Glass (120-26) Leningrado.
62. Hench L.L., (1981). Fundamental Aspects of Biocompatibility. D.F. Williams, CRC
Press.
63. Hench, L.L. &. Clark, D.E. (1978). Physical chemistry of glass surfaces. J. Non-Cryst.
Solids 28 (1) 83.
64. Gross, U. M., Mller-Mai, C. & C Voigt. (1993). Ceravital bioactive Glass-Ceramic. In
Hench LL, &Wilson J, editors. An introduction to bioceramics. (105-124) Singapore:
World Scientific Publishing Co.
65. Gross, U. &. Strunz, V (1982). Clinical Application of Biomaterials. Ed. A.J.C. Lee, T.
Albrektsson, & P. Branemark., (237) New York: John Wiley & Son.
66. Karlsson, K. H. Frberg, K. & Ringbom,T. (1989). A structural approach to bone
adhering of bioactive glasses. J. Non-Cryst. Solids, 112, 69-72.
67. Anderson, . H. Karlsson K. H. &. Kanganiemi, K (1990.). Calcium phosphate
formation at the surface of bioactive in vivo. J. Non-Cryst. Solids, 119, 290-296.
68. Kokubo, T. (1993) .A-W Glass-Ceramic: Processing and Properties. In L.L.Hench & J.
Wilson editors. Introduction to Bioceramcs (75-88). Singapore.
69. Kokubo, T., Ito,. S., Sakkay, S. & Yamamuro. Y. (1986). Formation of a High Strength
Bioactive Glass-Ceramic in the System MgO-CaO-SiO2-P2O5. J. Mater. Sci., 21, 536-
540.
70. Ohura, K., Nakamura, T., Yamamuro T., Kokubo, T. Ebisawa, T., Kotoura, Y. & Oka,
M. (1992). Bioactivity of CaO-SiO2 glasses added with various ions. J. Mater. Sci. Med.,
3, 95-100.
71. Abe, Y. & Fukui, H. (1975) Shikarikougaku-xasshi 16, 196.
72. Pernot, F., Zarzycki, J., Bonnel, F., Rabischong, P. & Baldet, P. (1979). New glass-
ceramics materials for prosthetic applications. J. Mater. Sci. 14, 1694-706.
73. Miake, Y., Yanagisawa, T., Yajima, Y., Noma, H., Yasui, N. & Nonami, T. (1995).
High-resolution and analytical electron microscopic studies of new crystals induced by
a bioactive ceramic (diopside). Journal of Dental Research. 74 (11), 1756-1763.
74. Nonami, T. (1995). In vivo and in vitro testing of diopside for biomaterials. Journal
Society Materials Enginnering for Resources of Japan 8 (2), 12-18.
75. De Aza, P.N., Guitian, F. & De Aza, S. (1994). Bioactivity of wollastonite ceramics: in
vitro evaluation. Scripta Metall. et Mat., 31 (8), 1001-1005.
76. De Aza, P.N., Guitian, F. & De Aza, S. (1995) Polycrystalline wollastonite ceramics.
Biomaterials free of P2O5. In Vicenzini, P. editor. Advances in Science and Technology,
12. Materials in Clinical Application. (vol. 12, 19-27). Techna Srl.
77. De Aza, P.N., Luklinska, Z.B., Anseau, M.R., Guitian, F. & De Aza, S. (1996)
Morphological studies of pseudowollastonite for biomedical application. Journal of
Microscopy-Oxford, 182, 24-31.
78. De Aza, P.N., Guitian, F., De Aza S. &. Valle, F. J. (1998). Analytical control of
wollastonite for biomedical applications by use of AAS and ICP-AES. The Analyst,
123: 681-685.
79. De Aza, P.N., Luklinska, Z.B., Anseau, M.R., Guitian, F. & De Aza, S. (1999)
Bioactivity of pseudowollastonite in human saliva. Journal of Dentistry. 27, 107-113.
80. De Aza, P.N., Luklinska, Z.B., Martinez, A., Anseau, M., Guitian, F. & De Aza, S.
(2000). Morphological and structural study of pseudowollastonite implants in bone. J.
Microccopy-Oxford, 197 (1), 60-67
130 P. N. De Aza
81. De Aza, P.N., Luklinska, Z.B, Anseau, M., Guitian, F. & De Aza, S. (2001).
Transmission electron microscopy of the interface between pseudowollastonite implants
and bone in vivo. J. Microccopy-Oxford. 201 (1), 33-43.
82. Fernandez-Pradas, J.M., Serra, P., Morenza, J.M. & De Aza P.N. (2002). Pulser laser
deposition os pseudowollastonite coatings. Biomaterials, 23, 2057-2061
83. Dufrane, D., Delloye, C., Mc.Kay, I. De Aza, P.N., De Aza S., Shneider. J. & Anseau,
M. (2003). Indirect cytotoxicity evaluation of pseudowollastonite. J. Mat. Sci. Mater
Med. 14 (1), 33-38.
84. De Aza, P.N., Fernandez-Pradas, J. M. & Serra. P. (2004). In vitro bioactivity of laser
abation pseudowollastonite coating. Biomaterials, 25;1983-1990.
85. Sarmento, C., Luklinska Z. B., Brown, L., Anseau, M., De Aza, P. N., De Aza, S.,
Hughes, F. & McKay I. (2004). The in vitro behaviour of osteoblastic cell cultured in
the presence of pseudowollastonite ceramic. J.Biomed. Mater. Res. 69A (2), 351-358.
86. Bailey. A. & Reesman, A.L. (1971). Study on the Kinetics of the Dissolution of
Wollastonite in H2O-CO2 and Buffered System at 25C. Am. J. Sci. 27, 464-472.
87. De Groot, K. & Le Geros, R.(1988). Significance of Porosity and Physical Chemistry of
Calcium Phosphate Ceramics. In Ducheyne, P. Lemons editors. Bioceramics Material
Characteristics Versus In Vivo Behaviour. (268-277). New York: Publ. J. Am. Acad.
Sci.
88. De Groot, K. (1988). Effect of Porosity and Physicochemical Properties on the Stability
Resorption and Strength of Calcium Phosphate Bioceramics. In Ducheyne, p. Lemons
Bioceramics Material Characteristics Versus Vivo Behaviour. Ed. (227-235) New York,
Publ. J. Am. Acad. Sci.
89. Tencer, A.F., Shors, E.C., Woodard, P.L., & Holmes, R.E. (1990). Mechanical and
biological properties of a porous polymer-coated coralline ceramic. In T. Yamamuro,
L.L. Hench, & J. Wilson editors. Handbook of Bioactive Ceramics (v. II. 209-221).
Boca Raton Florida. CRC. Press.
90. Roy, D.M. & Linnehan, S.K. (1974) Hydroxyapatite formed from Coral Skeletal
Carbonate by Hydrothermal Exchange. Nature 274, 220.
91. Holmes, R., Mooney, V., Bucholz, R. & Tencer, A. (1984). A coralline Hydroxyapatite
graft substitute. Preliminary report. Clin. Orthop. 188. 252-262.
92. Sivakumar, M., Dampath Kumar, T.S., Dhantha, K.L. & Pandurangsa Rao, K. (1996).
Development of Hydroxyapatite derived from Indian Coral. Biomaterials 17, 1709-1714.
93. De Aza, P.N., F.Guitian & De Aza, S. (1997). Bioeutectic: a new ceramic material for
human bone replacement. Biomaterials 18, 1285-1291.
94. De Aza, P.N., Luklinska Z., Anseau M., Guitian, F. & De Aza, S. (1998). Electron
microscopy of interfaces in a wollastonite-tricalcium phosphate bioeutectic material.
Journal of Microscopy-Oxford 189 (2), 145-153
95. De Aza, P.N., Guitian, F. & De Aza, S. (1998). A new bioactive material which
transforms in situ into porous hydroxyapatite. Acta Materialia. 46 (7), 2541-2549.
96. De Aza, P.N., Guitian, F. & De Aza, S. (1998). Eutectics structures that mimic porous
human bone. In A. P. Tomsia, & A. Glaeser Editors. Ceramic Microstructure: Control
al Atomic Level. (761-769). Plenum Publ.
97. De Aza, P.N., Guitian, F. & De Aza, S. (1999). Bioeutectics. a new bioceramic
materials. In P. Vincenzini editor. Advances in Science and Technology 28 Materials in
Clinical Application. (25-32). Techna Srl. ISBN 88-86538-29-4
98. De Aza, P.N., Luklinska Z., Anseau M., Hector, M., Guitian, F. & De Aza, S. (2000).
Reactivity of a wollastonite-tricalcium phosphate bioeutectic. Biomaterials. 21 (17),
1735-1741.
99. De Aza, P.N., Guitian, F. & De Aza, S. (1995). Phase diagram of wollastonite-
tricalcium phosphate. J.Am. Ceram. Soc. 78 (6), 1653-56.
100. Ashbrook, R.L. (1977). Directionally Solidified Ceramic Eutectics. J.Am. Ceram. Soc.,
60, 428-435.
101. Elliot, R. (1977). Eutectic Solidification. Int. Metals Reviews. 161-186.
102. Kurz, W. & Fisher, D.J. (1984). Fundamental of Solidification. Switzerland. Trans.
Tech. Publications.
103. Barin, I. (1989). Thermodynamical Data of Pure Substances, Vol. I D-6940 Weinhein
(Germany), VCH Verlagsgesellschaft mbH, 326-342.
104. Li, R., Clark, A. E. & Hench, L.L. (1991). An investigation of bioactive glass powders
by sol-gel processing. J. Appl. Biomaterials, 2, 231-239.
105. De Aza, P.N., Guitian, F. Merlos, A., Lora-Tamayo, E. & De Aza, S. (1996).
Bioceramics-simulated body fluid interface: ph and its influence on hydroxyapatite
formation. J. Mat. Science: Materials in Medicine. 7 (7), 399-402.
106. Walter, M.M. (1977). An Investigation into the Bonding Mechanisms of Bioglass.
Master Thesis, University of Florida, Gainsville.
107. Andersson, . H., Liu, G., Karlsson, K., Miettinen, H. & Juhanoja J. (1990). In vivo
behaviour of glasses in the SiO2-Na2O-CaO-P2O5-Al2O3-B2O3. J. Maters. Sci., Maters.
In Med. 1, 219-227.
108. Brmer, H., Pfeil, E. & Kos, M. (1973).German Patent. No 2, 326,100.
109. Brmer, H., Deutscher, K., Blenke, B., Pfeil, E. & Strunz, V. (1977). Properties of the
bioactive implant material Ceravital. Sci. Ceram. 9, 219-25.
110. Kokubo, T., Hayashi, T., Sakka, S., Kitsugi, T. & Yamamuro, T. (1987). Bonding
between Bioactive Glass, Glass-Ceramic or Ceramics in Simulated Body Fluid. Yogyo-
Kyokai-hi., 95 (8), 785-791.
111. Kokubo, T., Ito, S., Shigematsu, M., Sakka, S.& Yamamuro, T. (1985). Mechanical
Properties of a New Type of Apatite Containig Glass-Ceramic for Prothectic
Application. J. Mater. Sci., 20, 2001-2004.
112. Kokubo, T., Ito, S., Shigematsu, M., Sakka, S. & Yamamuro, T. (1987). Fatigue and
Life-Time of Bioactive Glass-Ceramic A/W Containing Apatite and Wollastonite. J.
Mater. Sci., 22, 4067-4070.
113. Hland, W., Naumann, J., Vogel, W., Gummel, J., Wiss, Z. & Schieler, F. (1983).Univ.
Jena Math. Naturwiss, Reihe 32, 571-80.
114. Vogel, W. & Hland, W. (1987). The development of bioglass ceramics for medical
applications. Angew. Chem. Int. Ed. Engl. 26 527-44.
115. Vogel, W. & Hland, W. (1988). Development, properties and application of bioglass-
ceramics in medicine. Adv. Fusion Glass, 33.16-33.
116. Rogier, R. & Pernot, F. (1990). Elastic and mechanical properties of glass-ceramic and
metal composites for orthopedic applications. In T. Yamamuro, L.L. Hench & J. Wilson.
132 P. N. De Aza
Handbook of Bioactive Ceramics, Vol I, Bioactive Glasses and Glass Ceramics, (183-
91). Boca Raton, FL. CRC Press.
117. Sakamoto, O. & Ito, S. (1990). Mechanical properties of bioactive glass-ceramic
composite by SiC whiskers. In T. Yamamuro, L.L. Hench & J. Wilson. Handbook of
Bioactive Ceramics, Vol. I, Bioactive Glasses and Glass Ceramics, (155-60). Boca
Raton, FL. CRC Press.
118. Ducheyne, P. & Hench, L.L. (1982). The processing of static mechanical properties of
metal fibre-reinforced bioglass. J. Mater. Sci. 17, 595-606.
119. Mller-Mai, Ch., Schmitz, H.J., Strunz, V., Fuhrmann, G., Fritz, Th. & Gross, U.M.
(1989). Tissues at the surface of the new composite material titanium / glass-ceramic for
replacement of bone and teeth. J. Biomed. Mater. Res. 23, 1149-61.
120. Kasuga, T., Nakajima, K., Uno, T. & Yoshida, M. (1990). Bioactive glass-ceramic
composite toughened by tetragonal zircona. In T. Yamamuro, L.L. Hench & J. Wilson.
Handbook of Bioactive Ceramics, Vol. I, Bioactive Glasses and Glass Ceramics, (137-
42). Boca Raton, FL. CRC Press.
121. Neo, M., Kotani, S., Fujita, Y., et al. (1992). Differences in ceramic bone interface
between surface-active ceramics and resorbable ceramics - a study by scanning and
transmission electron-microscopy. J. Biomed. Mater. Res. 26 (2), 255-267.
122. Welch, J.H., Gutt, W. (1961). High-temperature studies of the system calcium
oxidephosphorous pentoxide. J. Chem. Soc. 4442-4444.
123. Nurse, R.W., Welch, J.H. & Gutt, W. (1969). High-temperature equibria in the system
dicalcium silicate-tricalcium phosphate J. Chem. Soc.1077-1083.
124. Frondel, C. (1941). Whitlockite: a new calcium phosphate Ca3(PO4)2. Am. Miner; 26,
145-152.
Chapter 4
NOVEL BIOACTIVE HYDROXYAPATITE-BASE

CERAMICS AND GELATIN-IMPREGNATED
COMPOSITES
V.S. Komlev1,2 and S.M. Barinov1

1
Institute for Physical Chemistry of Ceramics, Russian Academy of Sciences,
Ozernaya 48, Moscow 119361, Russia;
2
Department of Sciences Applied to Complex Systems,
Polytechnic University of Marche, Via Breccee Bianche, I60131 Ancona, Italy.
Abstract
This study was aimed at the development of novel ceramic and composite materials
intended for application in bone tissue engineering. A method to produce porous spherical
hydroxyapatite (HA) granules was proposed. The method is based on liquids immiscibility
effect using the HA/gelatin suspension and oil as liquids. The granules of 50 to 2000 m
diameter contain open pores of 10 nm to 10 m size in amount up to 45 vol.%. A route for the
fabrication of porous HA ceramics having two population of open pores was developed.
Ceramics contain intragranular, up to 10 m size pores and intergranular, hundreds
micrometer size interconnected pores were prepared. The interconnections in the intergranular
pores are the pathway to conduct cells and vessels for the bone ingrowth. The gelatin solution
impregnation into the pores was demonstrated to enhance strength significantly. The strength
of HA/gelatin composites was increased by 6 to 10 times compared to that of initial ceramics.
The ceramics can further be filled with a drug resulting in a drug delivery system of slow drug
release rate.
1 Introduction
The use of synthetic materials to repair of traumatically or pathologically damaged bone
tissue offers certain advantages compared to the traditional biological methods of bone-defect
management, e.g. autografting and allografting of cancellous bone [1]. Since bone grafts are
avascular and dependent on diffusion, the size of the defect and the viability of the host bed
134 V.S. Komlev and S.M. Barinov
can limit their application. Furthermore, the new bone maintenance can be problematic due to
unpredictable bone resorption. There can be significant donor site morbidity associated with
infection, pain, and hematoma. Allografting introduces the risk of disease and infection
causing lessening of the bone inductive factors, etc. During the last decade, the tissue
engineering approach becomes very important allowing the diminish these risks.
Tissue engineering method is based on combining bone-forming cells and a supportive
scaffold before implantation in bony site [2, 3]. Progenitor cells are seeded onto resorbable
scaffold. The cells grow outside the body and become differentiated and mimic naturally
occurring tissues. These tissue-engineered constructs are then implanted into the patients to
replace diseased or damaged tissues. With time the scaffolds are resorbed and replaced by
host tissues that include a viable blood supply and nerves. The living tissue-engineered
constructs adapt to the physiological environment and should provide long-lasting repair [2].
The particular requirements for the bone tissue engineering determined the scaffold to be
in a porous form, because in this form large number of cells can be delivered. The materials
commonly used for this purpose are polymers, ceramics or composites. Ceramic phosphate
scaffolds guarantee clear advantages. Functional scaffolding devices need to sustain
physiologically applied loads, and ceramics - although fragile - possess an intrinsic strength.
Calcium phosphate phase, hydroxyapatite (HA), is the major mineral constituent of
human bone. HA-based ceramics are of a limited number of materials which form strong
chemical bonds with bone in vivo. Biocompatibility of HA is based on chemical resemblance
to bone mineral. A number of porous HA ceramics were developed for application in tissue
engineering [4-10]. The pores are known to have a direct influence on bone formation. Porous
ceramics body function is to serve as a substrate for proliferation and differentiation of cells
followed by the bone ingrowth into the pores. These processes are affected by pore size,
morphology, volume content and connectivity [9-11]. In particular, it has been reported that
the volume of bone ingrowth increases with an increase of pore size [12]. An optimal pore
size occurs for cells infiltration and host tissue ingrowth: 5-15 m for fibroblasts, 20-125 m
for adult mammalian skin tissue, and 100-135 m for bone tissue cells [13, 14]. No
unanimous opinion exists about optimal pores content in an implanted device. However, it
becomes obvious that the pores degree of interconnections directly influences the biological
fluids, especially in cells and vessels, that favor tissue nutrition and condition of new bone
formation [11]. As a consequence, some contradictions appeared concerning an optimal pore
size if their degree of interconnection is not taken into account [9]. Generally, a minimum
pore size of 100-135 m for sustained healthy bone ingrowth is accepted [9].
However, flexural strength, strain-to-failure, and fracture toughness values of HA-
ceramics are significantly less than those of bone, whereas the elastic modulus is much higher
[8, 9]. These mechanical mismatches influence the reliability of ceramics when implanted
into the bone tissue. To improve the mechanical compatibility, a composite approach may be
promising. The combination of different materials within a composite structure may lead to a
composite material that reveals specific physical, chemical and/or mechanical properties,
particularly resulted from a synergy principal.
However, current applications of composite biomaterials in medicine are still remarkably
less than expected a few years ago. In many biomedical applications, the research and the
testing of composites has been introduced and highly developed, but only in a very few cases
an industrial production and commercial distribution of medical devices partially or entirely
Novel Bioactive Hydroxyapatite-Base Ceramics 135
made of composites has started [15]. The main critical issues is that the available fabrication
methods may limit the possible reinforcement configurations, may be time consuming,
expensive, highly skilled.
Generally, there are two main types of HA-containing composites under development: (i)
ceramics reinforced with small particles, discrete or continuous fibres and (ii) biocompatible
polymers reinforced with small ceramic particles and satisfy to the aim of the tissue
engineering approach, e.g. to be of porous microstructure and of load-bearing capacity [16].
An alternative approach is to introduce a polymer into a porous ceramic matrix to obtain
composites with a continuous ceramic skeleton, in contrast to those reported in [16]. The
biological and mechanical properties of such ceramic/polymer composites can differ
markedly from those of the polymer/ceramic materials.
This paper presents some results of the study aimed at the development of a technique to
fabricate a porous HA matrix followed by the vacuum impregnation of the matrix with a
biopolymer solution to obtain a porous HA/biopolymer composite usable in the bone tissue
engineering applications. The important aspect of the tissue engineering approach is a
combination of a porous scaffold with a controlled-in-time drug release system. The pores of
ceramics scaffold could be impregnated with a drug like antibiotic, growth factor, bone
morphgenetic proteins, etc. to hinder the bacterial colonization of the scaffold or to reach the
osteoinductive behaviour.
2 Materials and Methods

2.1 Materials
As the starting materials, fine HA powder, gelatine and a vegetable oil were used. The
powder of HA was fabricated by precipitation from aqueous solution described in detail
elsewhere [17]. The analytical grade Ca(NO3)2, (NH4)2HPO4 and NH4OH were used as the
starting reagents. The synthesis reaction was
10Ca(NO3)2 + 6(NH4)2HPO4 + 8NH4OH Ca10(PO4)6(OH)2 + 20NH4NO3 + 6H2O (1)
Raw product of the reaction, white powder, was calcined at 9000C for 6 hours in an air
furnace. The characteristics of the calcined product were as follows: Ca/P = 1.67 (wet
chemical analysis); BET specific area as determined by nitrogen gas adsorption (a
Quantachrome Autosorb) is about 5.2 m2/g; the median powder agglomerates size is about 2.2
m (a Coulter Counter); the density is 2.16 g/cm3.
Food-grade bovine gelatin P-11 (Russian Standard GOST 11293-89) was supplied by
Gelatin Factory, Moscow, Russia. Vegetable oil (Rape oil, Russian Standard GOST 8988), as
a dispersion media, was of food-grade refined oil supplied by Cereol Magyarorszag RT
(Hungary). The relative density of the oil is 0.917 g/cm3, its viscosity is 0.585 Pa s at 25 0C.
Other solvents and reactants were of analytical grade.
2.2 Porous Spherical Granules Preparation
Aqueous solution of 10% gelatin content was prepared by dissolution of 10 g gelatin in 100
ml distilled water at 39 0C. Fine HA powder was added to the above solution in amounts of 1
g HA powder per 1.8 3.0 ml of gelatin solution. After that, the HA/gelatin slurry was
dispersed in 500 ml of oil in a reaction flask by stirring with a glass paddle stirrer at 200 500
rpm. The oil temperature was from 15 0C to 30 0C. The stirring results in the beads formation
due to the surface tension forces.
The precipitated beads were washed in acetone followed by ethanol, filtered onto a
Buechner cone filter and dried in air [18].
2.3 Porous Ceramics and Composites Preparation
Porous hydroxyapatite blocks were prepared by sintering of porous HA/gelatin microgranules.

Disc samples of 10 mm diameter and about 4-6 mm thickness were uniaxial pressed at 10-100
MPa pressure at a room temperature. Green bodies were further sintered at 12000C for 1 hour
in air [19].
To fabricate the composites, 1, 4, 7 and 10 vol % food grade bovine gelatin P-11
solutions in distilled water were used. The samples of porous HA ceramics were immersed in
the polymer solution under a vacuum of 1.33 Pa for 10 or 30 min, and without vacuum for 30
min. The temperature of the solution was 400C. After vacuuming at this temperature, the
samples were dried for 24 h in an air furnace [20].
2.4 Testing of Materials
Microstructure was monitored with a Neophot-32 microscope (Karl Zeiss, Jena) equipped
with a Videolabs Desktop Color Video Camera (Videolabs Inc.), as well as by scanning
electron microscopy (SEM) (a JEOL JSM-35-CF microscope). Open pores size distribution
(an Autoscan Qunatachrome Porometer) was measured; their content was estimated by a
common hydrostatic weightening in distilled water.
The disc specimens were tested for diametral compression loading to estimate the rupture
tensile stress (Brazilian test) [21]. To calculate the strength, the following equation was used
= 2P/dh (2)
where P is rupture load; d is diameter, and h is thickness of the disc. The samples were loaded
in an UTS-100 stiff testing machine (UTS Testesysteme GmbH) at a cross-head speed of
loading device of 1.0 mm/min. An error of the rupture load measurement was kept at 1%.
3 Results and Discussion

The method described allows the production of spherical HA granules of wide dimension
range, from <50 m to above 2000 m depending on the route. In Fig. 1 the size distribution
plots for the granules prepared at a stirrer rotation rate of 200 500 rpm at oil temperatures 15
0
C to 30 0C are shown. The granules size distribution obeys the normal distribution function.
The mathematic expectation of granule diameter is equal to 1735 m at a stirring rate of 200
rpm and an oil temperature of 15 0C, while at a stirring rate of 500 rpm and an oil temperature
of 30 0C it falls down to 670 m. Thus, content of the coarser granules increases generally
with a decrease in the stirring rate and the oil temperature. The temperature effect is due to
both the increased viscosity and surface tension of the oil. It is revealed by the direct
measurements of the granule size that the granulation process is practically completing during
the first 12 min from the beginning of the suspension stirring, as it can be seen from Fig. 2
as an example. The granules were fractionated using standard test sieves, and these of 400 m
to 600 m in diameter were used in further experiments.
50 50
45 b
45 a -200 r.p.m.
40
Relative content (vol. %)

40 -300 r.p.m.
Relative content (vol. %)
35 35 -400 r.p.m.
30 30
-500 r.p.m.
25 25
20 20
15 15
10 10
5 5
0 0
100- 500- 1000- 1500- > 2000 100- 500- 1000- 1500- > 2000
500 1000 1500 2000 500 1000 1500 2000
Granules size ( m)
(m) (m)
Granules size ( m)
Fig.1. Size distribution of microgranules produced at oil temperatures 15 0C (a) and 30 0C (b).
800
700
Mean granules size (mm)
600
500
400
300
200
100
0
0 2 4 6 8 10
Stirring time (min)
Fig.2. Mean granule size versus stirring time (stirring rate 500 rpm and oil temperature 25 0C).
HA granule
Intergranular
pore
HA block
Intragranular
pore
Fig. 3. A structure model of the body sintered of porous spherical granules.
45
d
40
Open pores content (vol. %)
35 c
30
b
25
a
20
15
950 1000 1050 1100 1150 1200 1250
Sintering temperrature ( C)
Fig. 4. Open pores content versus heat treatment temperature of the granules: gelatin content in starting
slurry 2.0 ml/g (a); 2.5 ml/g (b); 3.0 ml/g (c), and 3.5 ml/g (d).
Fig. 3 represents a structure model of porous bi-modal pore-size distribution ceramics.

The bi-modality is due to the intergranular and intragranular pores in the body. Intragranular
pores size and content are dependent strongly on the preparation route, e.g. on the gelatin/HA
ratio in the initial suspension and on the heat treatment regime. Fig. 4 shows the mercury
porometry data on open pores content versus the heat treatment temperature for the granules
prepared of varied initial gelatin/HA content in the initial suspension. The porosity in the
range from about 43 to 17 vol.% can be obtained. Dominant pores size (effective diameter) is
in the range 4 to 10 nm, the pores size increases within above range with an increase in the
heat treatment temperature. These pores occupy from 34 to 68 vol.% of the total open pores
within the granules. Other pores are these of from 1 to 10 m in diameter. The existence of
the latest pores population is confirmed by microscopy observations. It can be supposed,
according to [22], that the pores of tens micrometers diameter are connected either directly or
by narrow slit-like channels. In the last case the mercury porometry data are accounted for
these channels size only, because the penetration of the mercury under the pressure is limited
by these channels size.
Intergranular pores result from the granules packing features. Monosize particles can be
free packed into a green body having the relative open pores content 0.3 to 0.4 of the total
body volume [23], if the packing is uniform and no arch effect of packing exists. Such
packing density corresponds to the coordination number, N, of about 7 in the particles
package
N = 11.6 (1 ) (3)
where is the pores content [24]. This N value is close to the coordination number of a body
centered cubic (b.c.c.) structure model. In the frame of this model, minimum pore cross-
section can be evaluated from a simple geometric consideration of the packing density in the
most dense-packed (110) plane of the b.c.c. structure. This minimum size is equal to about
0.2 x 0.4 of the spherical granule diameter. Therefore, the packing of monosize spherical
granules of 500 m in diameter provides the open interconnecting pores of up to 100 x 200
m2 cross-section. This size is in accordance to the requirement of a minimum pores
dimension of 100-135 m essential for sustained healthy bone ingrowth into implanted device
[9]. To increase the pores size, coarser granules must obviously be used.
The situation becomes more complicated in the case of the polydisperse granules packing.
A model of spherical granules packing for bi-size granules mixtures has been devised in [25].
It has been shown that the open intergranular pores percentage varies from 12.5 to about 35
vol.%, the value being dependent on the granules dimension ratio and their relative contents
in the mixture. If the granules of two diameters, d1 and d2, d1 5d2, are packed, so a minimum
of the intragranular pores content can be predicted as
min = kV2/[(1 k)V1 + V2] (4)
where k is a coefficient in the range from 0.3 to 0.4; V1 and V2 are the volumes occupied by
coarse and fine spherical granules, respectively [25].
Fig. 5. Scanning electron microscopy picture of sintered at 12000C ceramics being compacted
at 40 MPa.
Microstructure of ceramics produced by sintering the granules depends on the green body
compacting pressure. The higher is the compaction pressure; the lower is the interpores size
and content. Shown in Fig. 5 is the pores view near an individual granule in the sintered
ceramics, the green body being compacted at a pressure of 40 MPa. The pores size is up to
100 m. Fig. 6 gives microstructure of the surface of the granule in the sintered ceramics. The
pores of from 1 to 10 m dimensions are clearly evidenced.
Fig. 6. Scanning electron microscopy picture of the granule surface in the sintered ceramics.
70
60
50
40
30
20
10
0
0 20 40 60 80 100
Pressure (MPa)
Fig. 7. Total open pores content in sintered bodies versus compaction pressure.
Mercury porometry measurements performed on individual 500 m diameter granules

after the heat-treatment at 12000C for 1 hour revealed that the open pores content equals to
about 18 vol.%. Shown in Fig. 7 is the total open pores content in sintered at 12000C body
versus the compaction pressure plot. The open pores content decreases gradually with an
increase of the compaction pressure until the pressure of about 60 MPa is reached. This is due
to the granules rearrangement and deformation, just because the granules contain gelatin
making the granules plastic. Measured for the samples prepared at very low compaction
pressure, the porosity is as high as 65 vol.%. That value is higher than the sum of the
intragranular porosity (18 vol.%) and the predicted intergranular porosity for the uniform
packing case (30 to 45 vol.%). Therefore, a non-uniformity of the packing exists, probably
resulting from the sticking the granules together due to the gelatin. If the pressure exceeds 60
MPa, so the pores content in sintered body becomes independent on the compaction pressure
being equals to about 18 vol.%. This value corresponds to the open pores content in the
granules themselves. Therefore, open porosity in the sintered bodies which are compacted at a
pressure of 60 MPa and above is the intragranular porosity. Bodies prepared at the
compacting pressure of 10 and 20 MPa are weak and may be crushed by hand easy. So, the
ceramics obtained at 30 to 50 MPa densification pressure can only be considered as being of
interest for practice.
Given in Table 1 are the mercury porometry measurements data on the open pores
content and distribution within the sintered at 12000C blocks in dependence on the green
body unaxial compacting pressure. The ceramics containing intragranular fine pores and
intergranular coarse these, of about 100 m effective diameter, can be produced with the use
of the 500 m size spherical granules batch compacted at 30 MPa pressure.
Table 1. Open pores characteristics of ceramics sintered at 12000C.
Dominant open pores size and content

Total open
Pressure, Intergranular pores Intragranular pores
pores content,
MPa Content, Content,
vol.% Size, nm Size, m
vol.% vol.%
10 67 2-10 and 17-18 >100 48-49
5000-14000
30 51 2-10 and 17-18 about 100 33-34
5000-14000
50 28 2-10 and 17-18 about 30 9-10
5000-14000
5
Tensile strength (MPa)
0
30 35 40 45 50
Fig. 8. Tensile strength estimated by diametral compression test versus open pores content in ceramics.
Fig. 8 gives the dependence of the rupture tensile stress of ceramics on the open pores
content. The strength reduces drastically with an increase of porosity in the range from about
30 to 40 vol.% reaching approximately constant level with the further increase of porosity.
The strength is known influenced strongly by pores. Generally, an exponential dependence is
used,
= 0 exp (-) (5)
where 0 is the strength of a non-porous body, = 7-8 is a constant, to predict the pores
effect [23]. The other common equation is
= 0(1 )m (6)
where exponent m equals to 3-10 for ceramics [26, 27]. Estimations give that the pores
content of about 30 vol.% decreases the strength by factor about 8-12. Therefore, the
measured strength value seems to be no contradictory one, because the tensile strength of
dense hydroxyapatite ceramics varies from 10 to 100 MPa [28].
The relative weight increment of the gelatin-impregnated samples increased with an
increase in the concentration of solution, initial porosity of ceramics, and impregnation time.
Maximum weight increment of about 7.3 % was reached by using 10 % gelatin solution,
respectively, under a vacuum degree of 1.33 Pa for 30 min impregnation duration.
SEM investigation indicates high compatibility between HA and gelatin. The polymer
formed thin layers on the surface of the ceramic phase. Fig. 9 illustrates the microstructure of
HA/gelatin composite, which have been fabricated using 7 % gelatin solution, under a
vacuum of 1.33 Pa for 30 min. Consequently, about 80 % of the pore space remains free of
gelatin that is useful for the osteointegration process.
Fig. 9. Scanning electron microscopy picture of the microstructures of HA/gelatin composites.

14 14
a b
12 12

10 10
8 8
6 6
4 4
2 2
0 0
1 4 10 1 4 10
Concentration of solution (%) Concentration of solution (%)
Fig. 10. Tensile strength estimated by diametral compression test of HA/gelatin composite materials
using ceramics obtained at (a) 50 and (b) 100 MPa pressure, as a function of concentration of solution
( vacuum 1.33 Pa for 30 min, vacuum 1.33 Pa for 10 min, and without vacuum for 30
min).
Fig. 10 shows the tensile strength of the composites. The increase in strength is
proportional to the polymer concentration in solution. The samples infiltrated with 10%
gelatin solutions exhibit the maximum tensile strengths of 9.8 and 13.7 MPa; these values
were reached if the infiltration was performed under vacuum 1.33 Pa for 30 min for ceramics
obtained at 50 and 100 MPa, respectively.
5
Value of
0
0 2 4 6 8 10 12
Concentration of gelatin solution (%)
Fig. 11. The value as a function of the gelatin solution concentration.
The value of in Equation 5 is a measure of the porosity effect on the strength. A wide
range of change in resulted from its dependence on the pore geometry: an increase of the
stress concentration factor created by pore increases the value. Shown in Fig. 11 is the
value as a function of the gelatin solution concentration. The value of was calculated as =
ln(0/)/, according to Equation 5, where is strength of the body of open pores content
being fabricated with the use of certain gelatin solution. The impregnation can be seen to
reduce value, e.g. strength sensitivity to the stress concentration created by pores. The
impregnation is supposed to fill the pore volume at the necks between sintered granules
reducing the stress concentration factor due to both the geometry of pore and the elastic
module mismatch.
Preliminary in vitro and in vivo testing of porous (non-impregnated) scaffolds for
antibiotics release and biocompatibility demonstrated high antibiotic adsorption efficiency
(465 %) and prolonged (up to 42 days) its release [29]. The gelatin impregnation, even being
at maximum reached of 7.3 wt. % level, is not expected to influence the pharmokinetics
significantly. The ceramics were demonstrated to possess excellent osteoconduction and
could be useful for the treatment of chronic osteomyelitis as well as infected joint arthroplasty
[29].
4 Conclusion
Novel HA ceramics of bimodal pore size distribution were developed. The strength of
ceramics was demonstrated to enhance significantly by means of gelatin solution
impregnation. The impregnation degree as low as 7.3 wt. % resulted in strength increasing
from about 2 MPa for initial ceramics to 13.7 MPa for composite, mainly due to the stress
concentration decrease at the necks between the sintered ceramics granules. These porous
HA/gelatin composite materials can further be filled with a drug and are expected to be useful
as drug delivery systems for the treatment of osteomyelitis and as scaffolds in bone tissue
engineering.
Acknowledgments
The supports available from the RFBR grant N 03-03-32230 and the Moscow Government
grant N 1.1.58 are deeply appreciated.
Partially supported by funds from the KMM NoE European project and INTAS Nr 04-
84-289.
References
1. Burg, K.J.L., Porter, S. & Kellam, J.F. (2000). Biomaterial developments for bone tissue
engineering. Biomaterials, 21, 2347-2359.
2. Hench, L.L. & Polak, J.M. (2002). Third-Generation Biomedical Materials. Science, 295,
1014-1017.
3. Griffith, L.G. & Naughton G. (2002). Tissue engineeringcurrent challenges and
expanding opportunities. Science, 295, 1009-1014.
4. Aoki, H. Science and Medical Application of Hydroxyapatite. Tokyo: JAAS; 1991.
5. Jones, J.R. & Hench, L.L. (2003). Regeneration of trabecular bone using porous ceramics.
Current Opinion in Solid State and Materials Science, 7, 301-307.
6. Hing, K.A., Best, S.M. & Bonfield, W. (1999). Characterization of porous hydroxyapatite.
J. Mater. Sci.: Mater. Med., 10, 135-145.
7. Kim, H.W., Knowles, J.C. & Kim, H.E. (2004). Hydroxyapatite/poly( -caprolactone)
composite coatings on hydroxyapatite porous bone scaffold for drug delivery.
Biomaterials, 25, 1279-1287.
8. Uchida, A., Nade, S.M., McCartney, E.R. & Ching, W. (1984). The use of ceramics for
bone replacement. A comparative study of three different porous ceramics. J. Bone Joint.
Surg., 66B, 269-275.
9. Hing, K.A., Best, S.M., Tanner, K.E., Bonfield, W. & Revell, P.A. (1999). Quantification
of bone ingrowth within bone-derived porous hydroxyapatite implants of varying density.
J. Mater. Sci.: Mater. Med., 10, 663-670.
10. Eggli, P.S., Mller, W. & Schenk, R.K. (1987). Porous hydroxyapatite and tricalcium
phosphate cylinders with two different size ranges implanted in the cancellous bone of
rabbits. Clin. Orthoped., 232, 127-138.
11. Liu, D.M. (1996). Fabrication and characterization of porous hydroxyapatite granules.
Biomaterials, 17, 1955-1957.
12. Jarcho, M. (1981). Calcium phosphate ceramics as hard tissue prosthetics. Clin. Orthop.
Rel. Res., 157, 259-278.
13. Yamamoto, M., Tabata, Y., Kawasaki, H. & Ikada, Y. (2000). Promotion of fibrovascular
tissue ingrowth into porous sponges by basic fibroblast growth factor. J. Mater. Sci.:
Mater. Med., 11, 213-218.
14. Hulbert, S.F. & Klawitter J.J. (1972). Application of porous ceramics for the attachment
of load-bearing internal orthopedic applications. Biomed. Mater. Symp., 2, 161-229.
15. Salernitano, E. & Migliaresi C. (2003). Composite materials for biomedical applications:
a review. J. Appl. Biomater. & Biomechanics, 1, 3-18.
16. Cao, W. & Hench, L.L. (1996). Bioactive materials. Ceramics Int., 22, 493-507.
17. Orlovskii, V.P., Komlev, V.S. & Barinov, S.M. (2002). Hydroxyapatite and
hydroxyapatite-based ceramics. Inorg. Mater. (Russia), 38, 973-984.
18. Komlev, V.S., Barinov, S.M. & Koplik, E.V. (2002). A method to fabricate porous
spherical hydroxyapatite granules intended for time-controlled drug release. Biomaterials,
23, 3449-3454.
19. Komlev, V.S. & Barinov, S.M. (2002). Porous hydroxyapatite ceramics of bi-modal pore
size distribution. J. Mater. Sci.: Mater. Med., 13, 295-299.
20. Komlev, V.S., Barinov, S.M. & Rustichelli F. (2003). Strength enhancement of porous
hydroxyapatite ceramics by polymer impregnation. J. Mater. Sci. Lett., 22, 1215-1217.
21. Barinov, S.M., & Shevchenko, V.Ja. Strength of Engineering Ceramics. Moscow:
Science; 1996.
22. Strelov, KK. Structure and Properties of Refractories. Moscow: Metallurgy; 1972.
23. Dulnev, G.N., & Zarichnjak, Yu. P. Heat Conductivity of Mixtures and Composite
Materials. Leningra: Energy; 1974.
24. Kaganer, M.G. Heat Insulation in Low Temperatures Technology. Moscow:
Mashinostrojenije; 1966.
25. Krasulin, Yu. L., Timofeev, V. N., Barinov, S.M., Ivanov, A. B., Asonov, A. N., &
Shnyrev, G. D. Porous Structural Ceramics. Moscow: Metallurgy; 1980.
26. Andrievski, R. A. (1982). Strength of sintered bodies. Powder Metal. (Russia) 1, 37-43.
27. Andrievski, R. A., & Spivak, I. I. Strength of Refractory Compounds and Related
Materials. Tcheljabins: Metallurgy; 1989.
28. Suchanek, W. & Yoshimura, M. (1998). Processing and Properties of Hydroxyapatite-
Based Biomaterials for Use as Hard Tissue Replacement Implants. J. Mater. Res., 13, 94-
117.
29. Hasegawa, M., Sudo, A., Komlev, V.S., Barinov, S.M. & Uchida, A. (2004). High
release of antibiotic from a novel hydroxyapatite with bimodal pore size distribution. J.
Biomed. Mater. Res. Part B: Appl. Biomater., 70B, 332-339.
In: Ceramics and Composite Materials: New Research ISBN 1-59454-370-4
Editor: B.M. Caruta, pp. 147-195 c 2005 Nova Science Publishers, Inc.
Chapter 5
T HERMAL S TRESSES IN PARTICLE-M ATRIX

S YSTEM AND R ELATED P HENOMENA .
A PPLICATION TO S I C-S I3N 4 CERAMICS
Ladislav Ceniga
Institute of Materials Research Slovak Academy of Sciences
Watsonova 47 043 53 Kosice Slovak Republic
Phone: +421 55 63 381 15
Fax: +421 55 63 371 08
Abstract
The paper deals with elastic thermal stresses in an isotropic particle-matrix system
of homogeneously distributed spherical particles in an infinite matrix, divided into cu-
bic cells containing a central spherical particle embedded in a matrix of a dimension
equal to an inter-particle distance. Originating during a cooling process as a conse-
quence of the difference in thermal expansion coefficients between the matrix and the
particle, and investigated within the cubic cell, the thermal stresses of the isotropic
multi-particle-matrix system are thus functions of the spherical particle volume frac-
tion, v, and are transformed for v = 0 to those of the isotropic one-particle-matrix
system. With regard to the particle-matrix system yield stress, the thermal stresses are
derived for such temperature range within which the particle-matrix system exhibit
elastic deformations. Similarly, the particle-matrix boundary adhesion strength is also
considered.
In addition to the thermal-stress-induced elastic energy of the cubic cell, the elastic
energy gradient within the cubic cell, calculated by two equivalent mathematical tech-
niques and representing a surface integral of the thermal-stress-induced elastic energy
density, is presented to derive the thermal-stress strengthening in the spherical particle
and the cubic cell matrix.
Considering a curve integral of the thermal-stress-induced elastic energy density,
the critical particle radii related to crack formations in ideal-brittle particle and matrix,
E-mail address: ceniga@saske.sk, ceniga@imrnov.saske.sk
148 Ladislav Ceniga
functions describing crack shapes in a plane perpendicular to a direction of the crack

formation in the particle and the matrix, and consequently particle and matrix crack
dimensions are derived along with the condition concerning a direction of the particle
and matrix crack formation. The former parameters for v = 0 are derived using a
spherical cell model for the spherical cell radius, Rc .
The derived formulae are applied to the SiC-Si 3N4 multi-particle-matrix system,
and calculated values of investigated parameters are in an excellent agreement with
those from published experimental results.
Keywords: particle-matrix system, thermal stress, strengthening, crack formation, SiC,

Si3 N4 .
1 Introduction
Investigated usually by approximate computational and experimental methods [1]-[3], ther-
mal stresses represent an important phenomenon observed in materials. With regard to
material science, thermal stresses originate as a consequence of the difference in ther-
mal expansion coefficients between individual material components, consequently influ-
encing mechanical properties [4]-[8], superconductivity [9]-[11], and diffusion processes
[12]-[17].
Resulting from an analytical model of elastic thermal stresses in isotropic multi-particle-
matrix system of homogeneously distributed spherical particle embedded in an infinite
matrix [18], the paper is devoted to the calculation of the elastic energy determining the
thermal-stress equilibrium state, to the calculation of the elastic energy gradient influenc-
ing a dislocation motion and being a reason of the thermal-stress strengthening, and along
with an analyse of conditions for particle and matrix crack formations, the results, pre-
sented in the section Theoretical background and consequently applied to the SiC-Si 3 N4
multi-particle-matrix system, accordingly enable to predict and design materials properties,
as additionally confirmed by an excellent agreement of calculated values of investigated
parameters with those from experimental results [19]-[21].
2 Theoretical Background
2.1 Isotropic Multi-Particle-Matrix System
2.1.1 Cell Shape
The thermal radial and tangential stresses, r and , , derived in the system of the spher-
ical coordinates (r, , ) related to the Cartesian system (Ox1 x2 x3 ) (Fig. 1), are investigated

in the point P of a continuum along the axes xr and x , x of the Cartesian system Pxr x x ,
where r is the distance from the spherical particle centre, O.
To derive the thermal stresses acting in the system of the homogeneously distributed
spherical particles embedded in the infinite matrix, the multi-particle-matrix system is di-
Thermal Stresses in Particle-Matrix System and Related Phenomena... 149
Figure 1: The axes xr , x , x , and the point P of the position determined by the coordinates
(r = |OP| , , ), where O is the spherical particle centre. The radial and tangential stresses in the
particle (q = p) and in the matrix (q = m), rq and q , q , act in the point P along the axes xr and
x , x , respectively.
vided into cells of the shape depending on the particle distribution, and the thermal stresses
are thus investigated within the cell [22].
Depending on the distribution of the spherical particles of the radius R (Fig. 2), the infi-
nite matrix is divided into cubic cells of the dimension d, representing an inter-particle dis-
tance, and consequently the spherical particle volume fraction of the multi-particle-matrix
system, v, as the ratio of the spherical particle volume to the cubic cell volume, is derived
as [18]
3
4 R
v= 0, , (1)
3 d 6
where /6 results from d = 2R.

Due to the isotropy of the multi-particle-matrix system, the shape symmetry of the
spherical particle and the cubic cell, the symmetrical distribution of the cubic cells resulting
from the matrix infinity, the thermal stresses are thus sufficient to be investigated within one
twenty-fourth of the cubic cell (Fig. 3), then for r h0, rci, h0, /4i, h34 , /2i,
where rc = |OC|, 34 = (x2 , OC34), C is a point on the cubic cell surface C1C2C3C4 , and
then [18]
d
rc = , (2)
2 cos sin
150 Ladislav Ceniga
Figure 2: The multi-particle-matrix system of the homogeneously distributed spherical particles

of the radius, R, embedded in the infinite matrix divided into the cubic cells of the dimension, d,
representing an inter-particle distance, and the Cartesian system (Ox1 x2 x3 ) in the point O in the
spherical particle centre.
Figure 3: The surface C1C2C3C4 of one eighth of the cubic cell and the point C on the surface
C1C2C3C4 of the position determined by the coordinates (rc , , ), where rc = |OC|, O is the spherical
particle centre, d is the cubic cell dimension (Fig. 2). The thermal stresses are thus sufficient to
be investigated within one twenty-fourth of the cubic cell, then for r hR, rc i, h0, /4i,
h34 , /2i.

|OC12 | 1
34 = arctan = arctan . (3)
|C12C34 | cos
2.1.2 Thermal Stresses

The radial and tangential stresses of the cubic cell, acting in the spherical particle ( q = p)
for r h0, Ri and in the cubic matrix (q = m) for r hR, rci (see Eq. (2)), rq and q , q ,
respectively, have the forms [18]
rp = p = p = pb , (4)
( "
3 #)
R
rm = pb 1 + c6 1 , (5)
r
( " #)
1 R 3
m = m = pb 1 + c6 1 + , (6)
2 r
where > 0 and < 0 represent the tensile and compressive stresses, respectively.
The compressive or tensile particle-matrix boundary radial stress, pb > 0 or pb < 0,
respectively, is derived as [18]
ZTi
pb = c7 (m p )dT, (7)
T
where the coefficient ci (i = 1 19) is presented in the section Appendix (see Eqs. (112)-
(127), (130)-(133)); Ti and T are the initial and final temperatures of a cooling process,
respectively; q is the thermal expansion coefficient of the particle ( q = p) and the matrix
RTi
(q = m). Considering q to be temperature-independent, the integral q dT is replaced by
T
q (Ti T ).
2.1.3 Temperature Range of Cooling Process

With regard to a real particle-matrix system exhibiting plastic deformation, the initial tem-
perature, Ti 0.5 Tmq , represents the homologous temperature below which the stress
relaxation does not occur as a consequence of thermal-activated processes [20, 21], where
Tmq is the melting temperature of the particle (q = p) or the matrix (q = m), and if either
Tmp < Tmm or Tmp > Tmm then either Tmp or Tmm is considered, respectively. Slowly cooling
the real particle-matrix system at temperature T > Ti , the thermal stresses are completely
released by plastic deformations [20, 21].
If at least one of individual components of the particle-matrix system represents a ce-
ramic material, the initial temperature, Ti Tmq , where Tmq is the melting temperature of
the ceramic component [21].
152 Ladislav Ceniga
The thermal stresses exhibit an maximal value on the particle-matrix boundary, acted
by the radial stress, pb (see Eq. (7)). Accordingly, the presented calculation is considerable
in the temperature range T hTc , Tii, where Tc is the critical final temperature of a cooling
process at which the stress, |pb | = yq , where yq = ycq or yq = ytq is a yield stress
in compression or tension related to the compressive or tensile stress, pb > 0 or pb < 0,
respectively, and if yp < ym or yp > ym then yq = yp or yq = ym , respectively. With
regard to Eq. (7), the critical temperature, Tc, can be derived from

ZTi

c7 (m p ) dT yq = 0, (8)

T
c
for a concrete isotropic multi-particle-matrix system by a numerical method, and consider-

ing q (q = p, m) to be temperature-independent, Tc has the form
yq
Tc = Ti . (9)
|c7 (m p )|
If the particle-matrix boundary bond is characterized by the adhesion radial stress, pa ,
and loaded by the tensile radial stress, pb < 0, the presented calculation is considerable for
|pb | pa , the critical temperature, Ta , related to the condition, |pb | = pa , can be derived
from Eq. (8), or has the form given by Eq. (9), both for yq pa . Finally, if Ta > Tc, the
presented calculation is considerable in the temperature range T hTa , Ti i.
2.1.4 Elastic Energy

Applying the Hooke laws for an isotropic continuum [23]
11 = s11 11 + s12 (22 + 33 ) , (10)
22 = s11 22 + s12 (11 + 33 ) , (11)
33 = s11 33 + s12 (11 + 22 ) , (12)

into the formula for the elastic energy density [23]
1 3
w= i j i j ,
2 i, j=13
(13)
and with regard to i j , i j = 0 (i 6= j) and the subscript transformations, 11 r, 22 ,

33 , the elastic energy density of the thermal stresses in the spherical particle and in the
cubic cell matrix, w p and wm , respectively, are derived as
3p2b
wp = (s11p + 2s12p ) , (14)
2
" 6 #
3p2b R
wm = 2c28 (s11m + 2s12m) + c26 (s11m s12m) . (15)
4 r
where the elastic moduli, s11q, s12q, for the particle (q = p) and the matrix (q = m) are
presented in the section Appendix (see Eqs. (110), (111)) [23].
Consequently, the elastic energy of the thermal stresses in the spherical particle and in
the cubic cell matrix, Wp and Wm, respectively, have the forms
Z Z ZR
Wp = w p dVp = w p r2 drd = 2R3 p2b (s11p + 2s12p ) , (16)
Vp 0
Z Z Zrc
1
Wm = wm dVm = 24 2
wm r drd = p2b R3 2c28 (s11m + 2s12m) 1
v
Vm R

+ c6 (s11m s12m) 2 36vc3 ,
2
(17)
where dVq = r2 drd represent an infinitesimal volume of the spherical particle ( q = p)

or the cubic cell matrix (q = m) for r h0, Ri, h0, 4i or r hR, rci, h0, /6i,
respectively. With regard to the term r6 in Eq. (15), d dd for h0, /4i,
h34 , /2i (Fig. 3), and consequently rc and 34 are given by Eqs. (2) and (3), respectively.
The elastic energy of the cubic cell is then W = Wp +Wm .
2.1.5 Surface Integral of Elastic Energy Density

The surface integral, Wsp and Wsm1 , of w p and wm (see Egs. (14), (15)) over the sur-
faces P2 P3 P4 k x2 x3 and P1 P2 P4 P5 k x2 x3 (Fig. 4), representing the elastic energy gradient
within the spherical particle and the cubic cell matrix along the axis x1 h0, Ri (Figs. 2-4),
Wsp = Wp /x1 and Wsm1 = Wm /x1 , respectively, as the elastic energy surface density,
equivalent to those along the axes x2 , x3 due to the isotropy of the multi-particle-matrix
system, has the form
Z Z/2ZR2
3p2b
Wsp = w p dS p = 4 w p r23 dr23 d = (s11p + 2s12p ) R2 x21 ,
2
Sp 0 0
x1 h0, Ri, (18)
Z Z/4ZR23
Wsm1 = wm dSm = 8 wm r23 dr23 d
Sm 0 R2
154 Ladislav Ceniga
Figure 4: The planes P1 P2 P3 P4 P5 k x2 x3 and P6 P7 P8 k x2 x3 in the positions x1 h0, Ri and x1

hR, d/2i, respectively, in the cubic cell of the dimension d, containing the spherical particle of the
radius R (Fig. 2).
( " 2/3 #
3p2b 4
= 4c28 (s11m + 2s12m) R2 R2 x21
8 3v

+ R2 c26 (s11m s12m) 4c9 R4 , x1 h0, Ri, (19)
and then, for x1 h0, Ri, dSq = r23 dr23 d is an infinitesimal part in the point P of the spher-
ical particle (q = p) or the cubic cell matrix (q = m) on the surfaces S p P2 P31 P32 k x2 x3
or Sm P11 P2 P32 P33 P12 k x2 x3 for r23 h0, R2 i or r23 hR2 , R23 i (Figs. 4, 5), respec-
tively, where P2 P31 P32 , P11 P2 P32 P33 P12 P1 P2 P3 P
q4 P5 (Fig. 4); r23 = |P31 P|; P31 P k x2 x3 ;
R = |OP2 | = |OP32 |; R2 = |P31 P2 | = |P31 P32 | = R2 x21 ; R23 = |P31 P23 | = d/(2 cos );
q
P31 P2 , P31 P23 k x2 x3 ; and the coordinate r in Eq. (15) is derived as r = |OP| = r23 2 + x2 .
1
Due to the isotropy of the multi-particle-matrix system, the angle is sufficient to be varied
in the interval h0, /4i.
The surface integral, Wsm2 , of wm (see Eq. (15)) over the surface P6 P7 P8 k x2 x3 (Fig. 4),
representing the elastic energy gradient within the cubic cell along the axis x1 hR, d/2i
(Figs. 2-4), Wsm2 = Wm /x1 , equivalent to those along the axes x2 , x3 , has the form
Z Z/4ZR68
Sm 0 0
Figure 5: The planes P2 P31 P32 k x2 x3 and P11 P2 P32 P33 P23 P12 k x2 x3 of the spherical particle of the
radius R and of the cubic cell matrix of the dimension d (Fig. 2), included, along with the point P,
in the plane P1 P2 P3 P4 P5 k x2 x3 in the position x1 h0, Ri (Fig. 4).
( 2/3
3p2b R2 4
= 4c28 (s11m + 2s12m)
8 3v
4 )
R d
+ c26 (s11m s12m ) 4c9 x41 , x1 R, , (20)
x1 2
and then, for x1 hR, d/2i, dSm = r68 dr68 d is an infinitesimal part in the point P of the
cubic cell matrix on the surface Sm P62 P72 P73 P63 k x2 x3 (Figs. 4, 6) for r68 = |P72 P|
h0, R68i, respectively, where P62 P72 P73 P63 P6 P7 P8 (Fig. 4); P72 P k x2 x3 ; R68 = |P
q72 P68 | =
d/(2 cos ); P72 P68 k x2 x3 ; and the coordinate r in Eq. (15) is derived as r = |OP| = r68
2 + x2 .
1
Due to the isotropy of the multi-particle-matrix system, the angle is sufficient to be varied
in the interval h0, /4i
Alternatively, the surface integral, Wsm2 , of wm (see Eq. (15)) over the surface P6 P7 P8 k
x2 x3 (Fig. 4), is also derived as
Z Z73 Z/2
"
3p 2 2
R 4 2/3
Wsm2 = wm dSm = 4 wm r2 dd = b 4c28 (s11m + 2s12m )
8 3v
Sm 0 6
4 #
2 R d
+ c6 c15 (s11m s12m ) , x1 R, , (21)
x1 2
156 Ladislav Ceniga
Figure 6: The plane P62 P72 P73 P68 P63 k x2 x3 of the cubic cell matrix of the dimension d (Fig. 2),
included, along with the point P, in the plane P6 P7 P8 k x2 x3 in the position x1 hR, d/2i (Fig. 4).
and then, for x1 hR, d/2i, dSm = r2 dd represents an infinitesimal part in the point P of
the cubic cell matrix on the surface Sm P62 P72 P73 P63 k x2 x3 (Figs. 4, 7) for r = |OP| =
x1 / (cos sin ), h0, 73 i, h6 , /2i, where P62 P72 P73 P63 P6 P7 P8 (Fig. 4); and the
angles 73 = (OP72 , OP73 ), 6 = (x3 , OP6 ) are presented in the section Appendix (see
Eqs. (128), (129)).
The x1 -dependence of Wsp + Wsm1 and Wsm2 exhibits concave and convex courses in
the intervals x1 h0, Ri and x1 hR, d/2i, respectively. As a reason of material properties
(e.g. dislocation motion, coercivity) [20, 21], the maximum, Wsmax = (Wsp +Wsm1 )x1 =0 (see
Eq. (18), (19)), has the form
( " 2/3 #
3p2b R2 3p2 R2 4
Wsmax = (s11p + 2s12p ) + b 4c28 (s11m + 2s12m )
2 8 3v

+ c26 (s11m s12m ) 4c9 R4 . (22)
2.1.6 Particle and Matrix Thermal-Stress Strengthening
With regard to (Eq. (13)) and 11 = , ii = 0 (i = 2, 3), considering the planes P1 P2 P3 P4 P5

and P6 P7 P8 to be loaded by the stress, , constant over the individual plane in the positions
x1 h0, Ri and x1 hR, d/2i, respectively, the elastic energy density, wq , induced by the stress,
Figure 7: The plane P62 P72 P73 P63 k x2 x3 of the cubic cell matrix of the dimension d (Fig. 2),
included, along with the point P, in the plane P6 P7 P8 k x2 x3 in the position x1 hR, d/2i (Fig. 4).
q , acting along the axis x1 in the particle (q=p) and the cubic cell matrix (q = m), is derived
as
s11q 2q
wq = . (23)
2
The elastic energy gradient within the spherical particle and the cubic cell along the axis
R
x1 (Figs. 2-4) is Wsq = wq Sq , where S p P2 P3 P4 , Sm P1 P2 P4 P5 and Sm P6 P7 P8 for
Sq
x1 h0, Ri and x1 hR, d/2i, respectively. The stresses, p and mi , derived from the conditions,
Ws p = Wsp and Wsmi = Wsmi (i = 1, 2) (see Eqs. (18)-(21)), and accordingly inducing the
same influence as induced by the thermal stresses and thus representing the thermal-stress
strengthening within the spherical particle and the cubic cell matrix, as a resistance against
compressive, tensile and tensile, compressive mechanical loading for p > m , p < m ,
respectively, have the forms
s
2s12p
p = pb 3 1 + , (24)
s11p
r
Wsm1
m1 = , (25)
c16
r
Wsm2
m2 = , (26)
c17
158 Ladislav Ceniga
where Eqs. (25) and (26) are functions of x1 h0, Ri (i = 1) and x1 hR, d/2i (i = 2),
respectively, and (m1 )x1 =R = (m2 )x1 =R (see Eqs. (19)-(21), (130), (131)).
With regard to Eq. (1), the average of thermal-stress strengthening in the cubic cell
matrix, m , is defined as [24]

ZR Zd/2
2
m = m1 dx1 + m2 dx1
d
0 R

1/3 ZR r Zd/2r
2 3v Wsm1 Wsm2
= dx1 + dx1 , (27)
R 4 c16 c17
0 R
where the integrals can be derived for a concrete isotropic particle-matrix system by a nu-
merical method.
2.1.7 Curve Integral of Elastic Energy Density
The curve integrals within the cubic cell, Wcp , Wcm1 and Wcm2 , as integrals of w p , wm and
wm along the abscissae P1 P2 , P2 P3 and P4 P5 in the plane x1 x3 in the positions x1 h0, Ri and
x1 hR, d/ (2 cos )i (Fig. 8), respectively, as the elastic energy curve density, equivalent
to those along the abscissae in the planes x1 x2 , x2 x3 due to the isotropy of the multi-particle-
matrix system, are derived as
Figure 8: The abscissae P1 P2 P3 and P4 P5 in the plane x12 x3 (Fig. 3) in the positions x1 h0, Ri
and x1 h0, d/ (2 cos)i, respectively, perpendicular to the axis x12 , where x12 x1 x2 and =
C (x12 , x1 ) h0, /4i (Fig. 3).
Z q
3p2b
Wcp = w p dx3 = (s11p + 2s12p ) R2 x21 , x1 h0, Ri, (28)
2
P1 P2
" #
Z 1/3 q
3p2 4
Wcm1 = wm dx3 = b5c28 (s11m + 2s12m) R 2 R2 x21
20 3v
P2 P3
5

2 2R (3v)1/3
+ c6 R (s11m s12m ) 1 q ,

R2 (4)2/3 + 4x21 (3v)2/3
x1 h0, Ri, (29)
Z 1/3
3p2 R 4
Wcm2 = wm dx3 = b 5c28 (s11m + 2s12m)
20 3v
P4 P5
5
5

1/3
R 2x1 (3v)
+ c26 (s11m s12m) 1 q ,
x1

R2 (4)2/3 + 4x2 (3v)2/3 1

d
x1 R, , (30)
2
q
where = (x1 x12 ) h0, 2i and the axis x12 x1 x2 (Fig. 3); |P1 P2 | = R2 x21 ; |P2 P3 | =
q
d/2 R2 x21 and |P4 P5 | = d/2 for x1 h0, Ri and x1 hR, d/2i, respectively. With
regard to the term r6 in Eq. (15), and due to the isotropy of the multi-particle-matrix
system, the elastic energy accumulated in the cubic cell matrix between the points P2 and
P3 , and between the points P4 and P5 , q
is equal to that accumulated between the points at
radii r = |OP2 | = R and r = |OP3 | = (d/2)2 + x21 , and at the radii r = |OP4 | = x1 and
q
r = |OP5 | = (d/2)2 + x21 , for x1 h0, Ri and x1 hR, d/2i, respectively. Accordingly,
the term r6 dx3 (see Eqs. (15), (28), (29)) is replaced by r6 dr, where r hR, |OP3 |i and
r hx1 , |OP5 |i for x1 h0, Ri and x1 hR, d/2i, respectively.
2.1.8 Particle and Matrix Crack Formation

Provided that pb < 0 and then for m < p , or pb > 0 and then for m > p (see Eqs. (4)-
(7)), the spherical particle or the cubic cell matrix are acted by the tensile thermal stresses,
or the tensile radial thermal stress and the compressive tangential thermal stresses, respec-
tively, and consequently, as a consequence of releasing of the elastic energy of the thermal
stresses, equal circular cracks are assumed to be equivalently formed in the planes x1 x2 ,
160 Ladislav Ceniga
x1 x3 , x2 x3 of the spherical particle or the cubic cell matrix (Figs. 9-14), due to the isotropy
of the multi-particle-matrix system. Resulting from the isotropy of the multi-particle-matrix
system, the formulae (28)-(30) are also related to the plane x01 x3 , and accordingly the particle
or matrix cracks, symmetrical with respect to the formation plane x1 x2 (Figs. 9-14), exhibit
the same shape in all planes x12 x3 , where the axis x12 x1 x2 and = (x1 x12 ) h0, 2i
(Fig. 3).
The curve integrals, Wcp and Wcm1 (see Eqs. (28), (29)), represent decreasing and in-
creasing functions of x1 h0, Ri, respectively, both depending on the parameter R, where
(Wcp )x1 =R = 0, and accordingly the sum, Wcp + Wcm1 , is a decreasing, increasing or de-
creasing-increasing dependence on x1 h0, Ri, exhibiting a minimum as a function of
the parameter R. With regard to the decreasing, increasing or decreasing-increasing x1 -
dependence of the sum, the circular particle crack is thus formed from the particle centre,
O, to the particle surface (Fig. 9), vice versa (Fig. 10), or simultaneously from the particle
centre and the particle surface (Fig. 11), resulting from the conditions, (Wcp +Wcm1 )x1 =0 >
(Wcm1)x1 =R , (Wcp +Wcm1 )x1 =0 < (Wcm1 )x1 =R or (Wcp +Wcm1 )x1 =0 = (Wcm1 )x1 =R , from which
the coefficient, c18 (see Eq. (129)), is determined as c18 > 0, c18 < 0 or c18 = 0, and con-
sequently, the function, f p , represents a decreasing, increasing or decreasing-increasing
x1 -dependence, respectively. Similarly, with regard to c18 > 0, c18 < 0 or c18 = 0, the ma-
trix crack is formed from the position x1 = 0 to the position x1 = R (Fig. 12), vice versa
(Fig. 13), or simultaneously from the positions x1 = 0 and x1 = R (Fig. 14), respectively.
Considering c18 > 0, c18 < 0 and c18 = 0, critical particle radii, as reasons of the formation
of the particle and matrix cracks of the infinitesimal length, are determined with respect
to positions of the maximum value of Wcp + Wcm1 , then for x1 = 0, x1 = R and x1 = 0 or
x1 = R, respectively.
The coefficient c18 depends on the spherical particle volume fraction, v, and on the
material parameters of the particle (q = p) and the matrix (q = m), s11q , s12q , and considering
c18 as a function of v, the critical spherical particle volume fraction, v0 (0, /6), related
to determination of a direction of the particle cracking, can be derived from the condition,
c18 = 0, for a concrete isotropic particle-matrix system by a numerical method.
Particle crack formation I

After the particle crack formation, the elastic energy, dWp = (Wcp +Wcm1 ) x1 ddx1 /3, accu-
mulated in the cubic cell infinitesimal volume of the dimensions x1 d dx1 d/2, and be-
ing a consequence of the particle cracking of the infinitesimal surface area, dS p = x1 dds p ,
is in the equilibrium state with the energy of an infinitesimal crack surface of the parti-
cle, dWcsp = p ds p dx1 d, and then dWp = dWcsp [19]-[21], where the multiplication factor
1/3 results from the formation of equal circular particle cracks in the planes x1 x2 , x1 x3 ,
x2 x3 (Fig. 1); p = s11p KICp
2
[19]-[21] is the particle crack surface energy per unit surface
q
area; KICp is the particle fracture toughness; ds p = dx1 1 + ( f p /x1 )2 [24] is infinitesi-
mal length of the curve, f p , describing the particle crack shape in the interval x1 h0, Ri
(Figs. 9-11), representing a function of x1 and the parameter R, corresponding to the x1 - and
R-dependence of Wcp +Wcm1 . With regard to dWp = dWcsp , we get
r 2
fp 1 2 2
= (Wcp +Wcm1 ) 3s11p KICp , (31)
x1 2
3s11p KICp
and accordingly, the energy condition for the particle crack formation, is derived as
2
Wcp +Wcm1 3s11p KICp > 0, (32)
fulfilled for the particle radius greater than critical, as a reason of the particle cracking.
Figure 9: The particle crack formed in the plane x12 x3 for R > R p1 from the particle centre, O, to
the particle surface; the particle crack radius, r p , for R hR p1 , R p2 i.
Considering such v when c18 > 0 (see Eq. (129)), the critical particle radii, R p1 and
R p2 , as a reason of the particle crack formation from the particle centre, O, to the particle
surface, and as a reason of the particle crack tip on the particle surface (Fig. 9), represent
roots of Eqs. (33) and (34) in the forms
2
(Wcp +Wcm1 )x1 =0 3s11p KICp = 0, (33)
2
(Wcp +Wcm1 )x1 =R 3s11p KICp = 0, (34)
as functions of the variable R, resulting from the condition for the particle crack tip in the
particle centre in the position x1 = 0 and on the particle surface in the position x1 = R,
( f p /x1 )x1 =0 = 0 and ( f p /x1 )x1 =R = 0, respectively, related to an ideal-brittle particle
[19]-[21], where R p1 < R p2 for c18 > 0. With regard to Eqs. (28), (29), (33), (34), R p1 and
R p2 are derived as
162 Ladislav Ceniga
( " #
1/3
1 p2b 1 4
= 2
s11p + 2s12p + 5c28 (s11m + 2s12m) 2
R p1 2s11pKICp 10 3v
" 5/3 #)!
3v
+ c26 (s11m s12m ) 1 32 , (35)
4
1/3
1 p2b 4
= 2
5c28 (s11m + 2s12m)
R p2 20s11p KICp 3v
5

2 2 (3v)1/3
+ c6 (s11m s12m ) 1 q . (36)

(4)2/3 + 4 (3v)2/3
The particle crack radius (Fig. 9), r p , determining the particle crack tip position, repre-
senting a root of the equation
2
Wcp +Wcm1 3s11p KICp = 0, (37)
as a function of the variable x1 and the parameter R hR p1 , R p2 i, resulting from the con-
dition for the particle crack tip in the position x1 = r p , ( f p /x1 )x1 =r p = 0 (see Eq. (31)),
can be derived from Eq. (37) for a concrete isotropic multi-particle-matrix system by a
numerical method, where r p = 0 and r p = R p2 for R = R p1 and R = R p2 , respectively.
With regard to Eq. (31), the function, f p , describing the particle crack shape in the
intervals x1 h0, r p i and x1 h0, Ri for R hR p1 , R p2 i and R R p2 (Fig. 9), respectively,
has the form
" Z r
#
1 2
fp = 2
p 2 2
(Wcp +Wcm1 ) 3s11p KICp dx1 , (38)
3s11p KICp
representing a decreasing function of x1 , and accordingly the sign - in Eq. (31) is considered.
The integral can be derived for a concrete isotropic particle-matrix system by a numerical
method, and with respect to the boundary conditions given by Eqs. (39) and (40) in the
forms
x1 = r p , f p = 0, (39)
R = R p2 , x1 = R p2 , f p = 0, (40)
related to the particle radius R hR p1 , R p2 i and R R p2 , respectively, the integration con-

stant, p , is derived for R hR p1 , R p2 i as
"Z r #
2
p = (Wcp +Wcm1 )2 3s11p KICp
2 dx1 , (41)
x1 =r p
and for R R p2 as
"Z r #
2
p = (Wcp +Wcm1 )2 3s11p KICp
2 dx1 . (42)
x1 =R p2 ; R=R p2
Particle crack formation II

Considering such v when c18 < 0 (see Eq. (129)), the critical particle radii, R p2 and R p1 (see
Eqs. (35), (36)), are a reason of the particle crack formation from the particle surface in the
position x1 = R to the particle centre, O, and a reason of the particle crack tip in the particle
centre in the position x1 = 0 (Fig. 10), respectively, where R p2 < R p1 for c18 < 0.
Figure 10: The particle crack formed in the plane x12 x3 (Fig. 8) for R > R p2 from the particle
surface to the particle centre, O; the particle crack radius, r p , for R hR p2 , R p1 i.
With regard to Eq. (31), the function, f p , describing the particle crack shape in the
intervals x1 hr p , Ri and x1 h0, Ri for R hR p2 , R p1 i and R R p1 (Fig. 10), respectively,
has the form
" Z r
#
1 2
fp = 2
p + 2
(Wcp +Wcm1 ) 3s11p KICp 2 dx1 , (43)
3s11p KICp
representing an increasing function of x1 , and accordingly the sign + in Eq. (31) is consid-
ered. With respect to the boundary conditions given by Eqs. (44) and (45) in the forms
164 Ladislav Ceniga
x1 = r p , f p = 0, (44)
R = R p1 , x1 = 0, f p = 0, (45)
related to the particle radius R hR p2 , R p1 i and R R p1 , respectively, the integration con-
stant, p , is derived for R hR p2 , R p1 i as
"Z r #
2
p = (Wcp +Wcm1 )2 3s11p KICp 2 dx1 , (46)
x1 =r p
and for R R p1 as
"Z r #
2
p = 2 2
(Wcp +Wcm1 ) 3s11pKICp dx1 . (47)
x1 =0; R=R p1
The particle crack radius (Fig. 11), r p (see Eq. (46)), represents a root of Eq. (37), as
a function of the variable x1 and the parameter R hR p2 , R p1 i, and can be derived from
Eq. (37), along with the integral in Eq. (43), for a concrete isotropic multi-particle-matrix
system by a numerical method, where r p = 0 and r p = R p1 for R = R p2 and R = R p1 ,
respectively.
Particle crack formation III

Considering such v when c18 = 0 (see Eq. (129)), the critical particle radii, R p1 , R p2 and
R p1/2 , are a reason of the particle crack formation simultaneously from the particle centre
in the position x1 = 0 and from the particle surface in the position x1 = R, and a reason of
the particle crack tip in the position x1 = R/2 (Fig. 11), where R p1 = R p2 for c18 = 0, and
R p1/2 > R p1 represents a root of Eqs (33) as a function of the variable R on the condition
x1 = R/2 in the form
" 1/3 #
1 p2b 4
= 2
5c28 (s11m + 2s12m) 3
R p1/2 20s11pKICp 3v
5

1/3

2 2 (3v)
+ c6 (s11m s12m) 1 q . (48)

(4)2/3 + (3v)2/3
The particle crack is described by the functions, f p1 and f p2 , decreasing and increasing
in the intervals x1 h0, r p1 i, x1 h0, r p i and x1 hr p2 , Ri, x1 hr p , Ri (Fig. 11), respectively,
related to the disconnected, interconnected particle cracks, as depended on the parameter
R. The particle crack radii, r p1 , r p2 , and the position in which the function, f p = f p1 + f p2 ,
describing the interconnected particle cracks exhibits a minimum, r p , represent roots of

Eq. (37), as a function of the variable x1 and the parameter R, for R R p1 , R p1/2 and
R R p1/2 , respectively, and can be derived from Eq. (37) for a concrete isotropic multi-
particle-matrix system by a numerical method, where r p1 = r p2 = 0 and r p1 = r p2 = R p1/2 /2
for R = R p1 and R = R p1/2 , respectively.
Figure 11: The particle crack simultaneously formed in the plane x12 x3 (Fig. 8) for R > R p1 = R p2
from the particle centre, O, to the particle surface, and vice versa; the particle crack radii, r p1

and r p2 , of the disconnected cracks for R R p1 , R p1/2 ; the position, r p , of the minimum of the
interconnected cracks for R R p1/2 .
With respect to the boundary conditions given by Eqs. (49), (50) and (51) in the forms
x1 = r p1 , f p1 = 0, (49)
x1 = r p2 , f p2 = 0, (50)
R p1/2
R = R p1/2 , x1 = , f p1 = f p2 = 0, (51)
2

related to the particle radius R R p1 , R p1/2 and R R p1/2 , respectively, the functions,
f p1 and f p2 , are given by Eqs. (38), (41), (42), and Eqs.

(43), (46),
(47), on the conditions
x1 = r p1 and x1 = r p2 in Eqs. (41) and (46) for R R p1 , R p1/2 , respectively, and on the
condition x1 = R p1/2 , R = R p1/2 in Eqs. (42), (47) for R R p1/2 .
166 Ladislav Ceniga
Matrix crack formation I

Considering the particle-matrix boundary adhesion radial stress, pa , provided that pb > 0
and then for m < p , the ca -multiple of rm , m , m , and c2a -multiple of w p , wm , the
latter resulted from the dependence, w 2 (see Eq. (13)), are concerned in the matrix
crack formation, where the multiplier, ca , is derived as
pa
ca = 1 . (52)
pb
After the matrix crack formation, the elastic energy related to the positions x1 h0, Ri
and x1 h0, d/2i, dWm = c2a (Wcp +Wcm1 ) x1 ddx1 /3 and dWm = c2aWcm2 x1 ddx1 /3, re-
spectively, accumulated in the cubic cell infinitesimal volume of the dimensions x1 d
dx1 d/2, and being a consequence of the matrix cracking of the infinitesimal surface area,
dSm = x1 ddsm , is in the equilibrium state with the energy of an infinitesimal crack surface
of the matrix, dWcsmi = m dsmi dx1 d (i = 1, 2), and then dWm = dWcsmi [19]-[21], where the
multiplication factor 1/3 results from the formation of equal circular particle cracks in the
planes x1 x2 , x1 x3 , x2 x3 (Fig. 1); m = s11mKICm
2
[19]-[21] is the matrix crack
q surface energy
per unit surface area; KICm is the matrix fracture toughness; dsmi = dx1 1 + ( fmi /x1 )2
[24] is infinitesimal length of the curves, fm1 and fm2 , describing the matrix crack shape in
the intervals x1 h0, Ri and x1 hR, d/ (2 cos )i (Fig. 12), respectively, representing func-
tions of x1 and the parameter R, corresponding to the x1 - and R-dependence of Wcm2 , where
= (x1 x12 ) h0, 2i and the axis x12 x1 x2 (Fig. 3). With regard to dWm = dWcsmi, we
get
q
f m1 1 2
= c2a (Wcp +Wcm1 )2 3s11mKICm
2 , (53)
x1 2
3s11mKICm
q
f m2 1 2 3s 2
2
= c2aWcm2 11m KICm , (54)
x1 2
3s11mKICm
and accordingly, the energy condition for the particle crack formation, is derived in the
interval x1 h0, Ri as
2
ca (Wcp +Wcm1 ) 3s11mKICm > 0, (55)
and in the interval x1 hR, d/ (2 cos )i as
2
caWcm2 3s11m KICm > 0, (56)
both fulfilled for the particle radius greater than critical, as a reason of the matrix cracking.
Considering such v when c18 > 0 (see Eq. (129)), the critical particle radii, Rm1 and Rm2 ,
as a reason of the matrix crack formation from the position x1 = 0 to the position x1 > 0,
and as a reason of the matrix crack tip in the position x1 = R (Fig. 12), represent roots of
Eqs. (57) and (58) in the forms
2
ca (Wcp +Wcm1 )x1 =0 3s11mKICm = 0, (57)
2
ca (Wcm2 )x1 =R 3s11mKICm = 0, (58)
as functions of the variable R, resulting from the conditions for the matrix crack tip,
( f m1 /x1 )x1 =0 = 0 and ( fm1 /x1 )x1 =R = 0, respectively, related to an ideal-brittle par-
ticle [19]-[21], where Rm1 < Rm2 for c18 > 0. With regard to Eqs. (28), (29), (57), (58), Rm1
and Rm2 are derived as
Figure 12: The matrix cracks formed in the plane x12 x3 (Fig. 8) from the position x1 = 0 to the
position x1 h0, rm1 i, x1 h0, Ri and from the position x1 = R to the position x1 hR, rm2 i, x1
hR, d/ (2 cos)i for R hRm1 , Rm2 i, R Rm2 and R hRm2 , Rm3 i, R Rm3 , respectively; the matrix
crack radii, rm1 , rm2 .
( " #
1 ca p2b 1 2 4 1/3
= 2
s11p + 2s12p + 5c8 (s11m + 2s12m ) 2
Rm1 2s11mKICm 10 3v
" 5/3 #)!
2 3v
+ c6 (s11m s12m ) 1 32 . (59)
4
1/3
1 ca p2b 4
= 2
5c28 (s11m + 2s12m)
Rm2 20s11mKICm 3v
5

1/3
2 (3v)
+ c6 (s11m s12m) 1 q
2
, (60)

2/3 2/3

(4) + 4 (3v)
168 Ladislav Ceniga
The matrix crack radius (Fig. 12), rm1 , determining the matrix crack tip position, repre-
senting a root of the equation
2
ca (Wcp +Wcm1 ) 3s11mKICm = 0, (61)
as a function of the variable x1 and the parameter R hRm1 , Rm2 i, resulting from the condi-
tion for the particle crack tip in the position x1 = rm1 , ( fm1 /x1 )x1 =rm1 = 0 (see Eq. (53)),
numerical method, where rm1 = 0 and rm1 = Rm2 for R = Rm1 and R = Rm2 , respectively.
With regard to Eq. (53), the function, f m , describing the matrix crack shape in the
intervals x1 h0, rm1 i and x1 h0, Ri for R hRm1 , Rm2 i and R Rm2 (Fig. 12), respectively,
has the form
Z q
1 2
f m1 = 2
m1 2 2 2
ca (Wcp +Wcm1 ) 3s11mKICm dx1 , (62)
3s11mKICm
representing a decreasing function of x1 , and accordingly the sign - in Eq. (53) is considered.
The integral can be derived for a concrete isotropic particle-matrix system by a numerical
method, and with respect to the boundary conditions given by Eqs. (63) and (64) in the
forms
q
x1 = rm1 , fm1 = R2 rm1 2 , (63)
x1 = R, ( fm1 )x1 =R = ( fm2 )x1 =R , (64)

related to the particle radius R hRm1 , Rm2 i and R Rm2 , respectively, the integration con-
stant, m1 , is derived for R hRm1 , Rm2 i as
q
m1 = 2
3s11mKICm 2
R2 rm1
Z q
2
+ c2a (Wcp +Wcm1 )2 2
3s11mKICm dx1 , (65)
x1 =rm1
and for R Rm2 as
Z q
2
m1 = m2 2 3s
c2aWcm2 2
11m KICm
q
2 2
2
2
x1 =R
where the function, f m2 , and the integration constant, m2 , are given by Eqs. (81) and (84),
(85), respectively.
Matrix crack formation II

Considering such v when c18 < 0 (see Eq. (129)), the critical particle radii, Rm2 and Rm1 ,
are a reason of the matrix crack formation from the position x1 = R to the position x1 < R,
and a reason of the matrix crack tip in the position x1 = 0 (Fig. 13), respectively, where
Rm2 < Rm1 for c18 < 0.
With regard to Eq. (53), the function, fm1 , describing the matrix crack shape in the
intervals x1 h0, rm1 i and x1 h0, Ri for R (Rm2 , Rm1 ) and R Rm1 (Fig. 13), respectively,
has the form
Z q
1 2
f m1 = 2
m1 + c2 (W +W
a cp cm1 )2
3s K 2
11m ICm dx 1 , (67)
3s11mKICm
representing an increasing function of x1 , and accordingly the sign + in Eq. (53) is consid-
ered. With respect to the boundary conditions given by Eqs. (63) and (64) related to the
particle radius R hRm2 , Rm1 i and R Rm1 , respectively, the integration constant, m1 , is
derived for R hRm2 , Rm1 i as
q
m1 = 2
3s11mKICm 2
R2 rm1
Z q
2
c2a (Wcp +Wcm1 )2 2
3s11mKICm dx1 , (68)
x1 =rm1
and for R Rm1 as
Z q
2
m1 = m2 2 3s
c2aWcm2 2
11m KICm
q
2 2
2
+ 2
x1 =R
(85), respectively. The matrix crack radius (Fig. 13), rm1 (see Eq. (68)), represents a root
of Eq. (61), as a function of the variable x1 and the parameter R hRm2 , Rm1 i, and can be
derived from Eq. (61), along with the integral in Eq. (67), for a concrete isotropic multi-
particle-matrix system by a numerical method, where rm1 = 0 and rm1 = Rm1 for R = Rm2
and R = Rm1 , respectively.
Matrix crack formation III

Considering such v when c18 = 0 (see Eq. (129)), the critical particle radii, Rm1 , Rm2 and
Rm1/2 , are a reason of the matrix crack formation simultaneously from the positions x1 = 0,
x1 = R to the positions x1 > 0, x1 < R, respectively, and a reason of the matrix crack tip in
the position x1 = R/2 (Fig. 14), respectively, where Rm1 = Rm2 for c18 = 0, and Rm1/2>Rm1
170 Ladislav Ceniga
Figure 13: The matrix cracks formed in the plane x12 x3 (Fig. 8) from the position x1 = R to the
positions x1 hrm1 , Ri, x1 h0, Ri and x1 hR, rm2 i, x1 hR, d/ (2 cos)i for R hRm2 , Rm1 i, R
Rm1 and R hRm2 , Rm3 i, R Rm3 , respectively; the matrix crack radii, rm1 , rm2 .
represents a root of Eq. (57) as a function of the variable R on the condition x1 = R/2 in the
form
" 1/3 #
1 ca p2b 4
= 2
3(s11p + 2s12p ) + c28 (s11m + 2s12m) 3
Rm1/2 4s11mKICm 3v
5

1/3
c26 2 (3v)
+ (s11m s12m) 1 q . (70)
5
2/3 2/3

(4) + (3v)
The matrix crack is described by the functions, f m11 and fm12 , decreasing and increas-
ing in the intervals x1 h0, rm11 i, x1 h0, rm1 i and x1 hrm12 , Ri, x1 hrm1 , Ri (Fig. 14),
respectively, related to the disconnected, interconnected matrix cracks, as depended on
the parameter R. The matrix crack radii, rm11 , rm12 , and the position in which the func-
tion, f m1 = fm11 + fm12 , describing the interconnected particle cracks exhibits a minimum,
rm1 ,
represent roots
of Eq. (61), as a function of the variable x1 and the parameter R, for
R Rm1 , Rm1/2 and R Rm1/2 , respectively, and can be derived from Eq. (61) for a con-
crete isotropic multi-particle-matrix system by a numerical method, where rm11 = rm12 = 0
and rm11 = rm12 = Rm1/2 /2 for R = Rm1 and R = Rm1/2 , respectively.
With regard to Eq. (53), the functions, f m11 and fm12 , have the forms
Z q
1 2
f m11 = 2
m11 2 2 2
3s11mKICm
Figure 14: The matrix cracks simultaneously formed in the plane x12 x3 (Fig. 8) from the positions
x1 = 0 and x1 = R to the positions x1 h0, rm11 i, x1 h0, rm1 i and x1 hrm12 , Ri, x1 hrm1 , Ri for

R Rm1 , Rm1/2 , R Rm1/2 , respectively, and from the position x1 = R to the position x1 hR, rm2 i,
x1 hR, d/ (2 cos )i for R hRm2 , Rm3 i, R Rm3 , respectively; the matrix crack radii, rm11 , rm12 ,
of the disconnected cracks; the position, rm1 , of the minimum of the interconnected cracks.
Z q
1 2
fm12 = 2
m12 + 2 2 2
3s11mKICm
representing increasing and decreasing function of x1 , and accordingly the sign - and + in
Eqs. (71) and (75) are considered, respectively. The integrals can be derived for a concrete
isotropic particle-matrix system by a numerical method, and with respect to the boundary
conditions given by Eqs. (73), (74) and (75), (79) in the forms
q
x1 = rm11 , fm11 = R2 rm11 2 , (73)
q
x1 = rm12 , fm12 = 2 ,
R2 rm12 (74)
x1 = rm1 , fm11 = fm12 , (75)
x1 = rm1 , fm11 = fm12 , (76)

related to the particle radius R Rm1 , Rm1/2 and R Rm1/2 , respectively, the integration
constants, m11 and m12 , are given by Eqs.

(38) and (47), replacing the term rm1 by the
terms rm11 and rm12 , respectively, for R Rm1 , Rm1/2 , and derived for R Rm1/2 as
172 Ladislav Ceniga
Z q
2
11m = m2 + 2 c2a (Wcp +Wcm1 )2 2
3s11mKICm dx1
x1 =rm1
Z q
2 2
2
2
ca (Wcp +Wcm1 ) 3s11mKICm dx1
x1 =R
Z q
2 3s 2
2
c2aWcm2 11m KICm dx1 , (77)
x1 =R
Z q
2
m12 = m2 c2a (Wcp +Wcm1 )2 2
3s11mKICm dx1
x1 =R
Z q
2 2
2
2
caWcm2 3s11mKICm dx1 , (78)
x1 =R
(85), respectively.
Matrix crack formation IV

With regard to the function, f m2 , describing the matrix crack shape in the interval x1
hR, d/ (2 cos )i (Figs. 3, 13, 14), the critical particle radius, Rm2 , given by Eq. (60), is also
a reason of the matrix crack formation from the position x1 = R to the position x1 > R.
Consequently, the critical particle radius, Rm3 , resulting from the condition for the matrix
crack tip on the cell surface in the position x1 = d/ (2 cos ), ( fm2 /x1 )x1 =d/(2 cos) = 0,
related to an ideal-brittle matrix [19]-[21], corresponding to a connection of the matrix
cracks in neighbouring cells between the points C1 and C12 (Fig. 3), where Rm2 < Rm3 ,
represents a root of Eq. (58) as a function of the variable R on the condition x1 = d/ (2 cos ),
and regarding Eq. (1) is derived as an increasing function of the angle in the form
( 1/3
1 ca p2b 4
= 2
5c28 (s11m + 2s12m)
Rm3 20s11mKICm 3v
5/3 " #)
3v 1
+ 32c26 cos5 (s11m s12m ) 1 . (79)
4 (1 + cos2 )5/2
The matrix crack radius (Figs. 13, 14), rm2 , determining the matrix crack tip position,
representing a root of the equation
2
caWcm2 3s11m KICm = 0, (80)
as a function of the variable x1 and the parameter R hRm2 , Rm3 i, resulting from the condi-
tion for the particle crack tip in the position x1 = rm2 , ( fm2 /x1 )x1 =rm2 = 0 (see Eq. (54)),
numerical method.
With regard to Eq. (54), the function, fm2 , describing the matrix crack shape in the
intervals x1 hR, rm2i and x1 hR, d/ (2 cos )i for R hRm2 , Rm3 i and R Rm3 (Figs. 13,
14), respectively, has the form
Z q
1 2
f m2 = 2
m2 c2W 2 3s
a cm2 K
11m ICm
2 dx 1 . (81)
3s11mKICm
The integral can be derived for a concrete isotropic particle-matrix system by a numer-
ical method, and with respect to the boundary conditions given by Eqs. (82) and (83) in the
forms
x1 = rm2 , fm2 = 0, (82)
d
R = Rm3 , x1 = , fm2 = 0, (83)
2 cos
related to the particle radius R hRm2 , Rm3 i and R Rm3 , respectively, the integration con-
stant, m1 , is derived for R hRm2 , Rm3 i as
Z q
2
m2 = 2 2 2
caWcm2 3s11mKICm dx1 , (84)
x1 =rm2
and for R Rm3 as

Z q
2
m2 = 2
c2aWcm2 2
3s11mKICm dx1 . (85)
x1 =d/(2 cos ); R=Rm3
2.2 Isotropic One-Particle-Matrix System

As presented in the section 3, concerning the v-dependences of thermal-stress related pa-
rameters of the SiC-Si3 N4 particle-matrix system in the interval v h0, /6), the formulae
related to the isotropic one-particle-matrix system represented by one spherical particle
of the radius R embedded in the infinite matrix, as transformation of those related to the
isotropic multi-particle-matrix system are for v = 0, are required to be derived.
2.2.1 Thermal Stresses

The radial and tangential stresses, acting in the spherical particle ( q = p) for r h0, Ri and
in the infinite matrix (q = m) for r hR, ), rq and q , q , respectively, have the forms
[25]
rp = p = p = pb , (86)
174 Ladislav Ceniga
3
R
rm = 2m = 2m = pb , (87)
r
where > 0 and < 0 represent the tensile and compressive stresses, respectively.
The compressive or tensile particle-matrix boundary radial stress, pb > 0 or pb < 0,
respectively, is derived as [25]
ZTi
1
pb = (m p ) dT, (88)
c19
T
and Eqs. (4)-(7), (112)-(115) are transformed for v = 0 to Eqs. (86)-(88), (130), for
RTi
(pc )v=0 = 0. Considering q to be temperature-independent, the integral q dT is replaced
T
by q (Ti T ).
2.2.2 Temperature Range of Cooling Process

With regard to the section 2.1.3, the presented calculation is considerable in the temperature
range T hTc , Ti i, and the critical final temperature of a cooling process, Tc , can be derived
from

Ti
1 Z
( )dT yq = 0, (89)
c19 m p
T
c
for a concrete isotropic multi-particle-matrix system by a numerical method, and consider-

ing q (q = p, m) to be temperature-independent, Tc has the form

c19

Tc = Ti yq . (90)
(m p )
2.2.3 Elastic Energy

With regard to Eq. (13), and consequently to i j = 0 (i 6= j) and the subscript transforma-
tions, 11 r, 22 , 33 , the elastic energy density and the elastic energy of the
thermal stresses in the infinite matrix, wm and Wm , respectively, are derived as
6
3p2b R
wm = (s11m s12m) , (91)
4 r
Z Z2Z
Wm = wm dVm = wm r2 drd = p2b R3 (s11m s12m) , (92)
Vm 0 R
where dVm = r2 drd represent an infinitesimal volume of the matrix, and h0, 4i.
The elastic energy density and the elastic energy of the thermal stresses in the spherical
particle, w p and Wp , respectively, are given by Eqs. (14), (16), considering the particle-
matrix boundary radial stress, pb , derived by Eqs. (88), (130). The elastic energy of the
cubic cell is then W = Wp +Wm .
2.2.4 Surface Integral of Elastic Energy Density

The surface integral, Wsp , of w p (see Eg. (14)) over the surface P2 P3 P4 k x2 x3 (Fig. 15)
is given by Eq. (18), considering the particle-matrix boundary radial stress, pb , derived by
Eqs. (88), (130). Determining the surface integral of the elastic energy density, the spherical
particle is considered to be embedded in the spherical cell of the radius Rc (Fig. 15), and
derived formulae then represent transformations of the integrals for Rc .
Figure 15: The planes P1 P2 P3 P4 P5 k x2 x3 and P6 P7 P8 k x2 x3 in the positions x1 h0, Ri and x1

hR, Rc i, respectively, in the spherical cell of the radius Rc , containing the spherical particle of
the radius R.
The surface integral, Wsm1 , of wm (see Egs. (88), (91), (130)) over the surface P1 P2 P4 P5 k
x2 x3 (Fig. 15), representing the elastic energy gradient within the spherical cell matrix of the
radius Rc along the axis x1 h0, Ri, Wsm1 = Wm/x1 , as the elastic energy surface
density, equivalent to that along the axes x2 , x3 due to the isotropy of the one-particle-matrix
system, has the form
Z Z/2ZR1
Sm 0 R2
176 Ladislav Ceniga
" 4 #
3p2b R2 R 2 2
Rc 3pb R
= (s11m s12m ) 1 = (s11m s12m) ,
8 Rc 8
x1 h0, Ri, (93)
and then, for x1 h0, Ri, dSq = r23 dr23 d is an infinitesimal part in the point P of the
spherical particle (q = p) or the spherical cell matrix (q = m) on the surfaces S p P2 P31 P32 k
x2 x3 or Sm P1 P2 P32 P33 k x2 x3 , P2 P31 P32 , P1 P2 P32 P33 P1 P2 P3 P4 P5 (Fig. 15), for r23
h0, R2i or r23 hR2 , R1 i (Fig. 16), respectively, where r23 = |P31 P|,qP31 P k x2 x3 , R = |OP2 | =
|OP32 |, Rc = |OP1 | = |OP33 | (Fig. 5), R1 = |P31 P1 | = |P31 P33 | = R2c x21 , R2 = |P31 P2 | =
q
|P31 P32 | = R2p x21 , P31 P1 , P31 P2 k x2 x3 ; and the coordinate r in Eq. (91) is derived as
q
r = |OP| = r23 2 + x2 . Due to the isotropy of the particle-matrix system, the angle is
1
sufficient to be varied in the interval h0, /2i.
Figure 16: The planes P2 P31 P32 k x2 x3 and P1 P2 P32 P33 P23 k x2 x3 of the spherical particle of the
radius R and of the spherical cell matrix of the radius Rc , included, along with the point P, in
the plane P1 P2 P3 P4 P5 k x2 x3 in the position x1 h0, Ri (Fig. 15).
The surface integral, Wsm2 , of wm (see Eqs. (88), (91)) over the surface P6 P7 P8 k x2 x3
(Fig. 15), representing the elastic energy gradient within the spherical cell of the radius
Rc along the axis x1 hR, Rci (Fig. 1), Wsm2 = Wm /x1 , equivalent to those along the
axes x2 , x3 , has the form
Z Z/2ZR68
Wsm2 = wsm dSm = 4 wm r68 dr68 d
Sm 0 0
4 " 4 #
3R2 p2b R R
= (s11m s12m ) 1
8 x1 Rc

Rc 3pb R
2 2
R 4
= (s11m s12m ) , x1 hR, Rci , Rc , (94)
8 x1
and then, for x1 hR, Rci, dSm = r68 dr68 d is an infinitesimal part in the point P of the
spherical cell matrix on the surface Sm P62 P72 P73 P63 k x2 x3 (Fig. 17) for r68 = |P72 P|
q R68i, respectively, where P62 P72 P73 P63 P6 P7 P8 (Fig. 15); P72 P k x2 x3 ; R68 =q
h0, |P72 P68 | =
R2c x21 ; P72 P68 k x2 x3 ; and the coordinate r in Eq. (91) is derived as r = |OP| = r68
2 + x2 .
1
Due to the isotropy of the particle-matrix system, the angle is sufficient to be varied in the
interval h0, /2i.
Figure 17: The plane P6 P72 P73 k x2 x3 of the spherical cell matrix of the radius Rc , included,
along with the point P, in the plane P6 P7 P8 k x2 x3 in the position x1 hR, Rc i (Fig. 15).
Consequently, the maximum, Wsmax = (Wsp +Wsm1 )x1 =0 , derived from Eqs. (18), (93)
for x1 = 0, has the form
3R2 p2b
Wsmax = [4 (s11p + 2s12p ) + s11m s12m ]. (95)
8
2.2.5 Particle and Matrix Thermal-Stress Strengthening

The elastic energy gradients within the spherical cell matrix along the axis x1 , Wsm1 and
Wsm2 , equivalent to those along the axes x2 , x3 due to the isotropy of the one-particle-matrix
system, induced by the stresses, m1 and m2 , constant on the plane perpendicular to the
178 Ladislav Ceniga
axis x1 , representing integrals of wm over the surfaces P2 P3 P4 and P1 P2 P4 P5 for x1 h0, Ri

and x1 hR, Rci, respectively, have the forms
Z
s11m2 2
Wsm1 = wm dSm = Rc R2 , x1 h0, Ri, Rc , (96)
2
Sm
Z
s11m2 2
Wsm2 = wm dSm = Rc x21 , x1 hR, Rci , Rc . (97)
2
Sm
With regard to Rc , the thermal-stress strengthening in the infinite matrix of the

isotropic one-particle-matrix system, mi (i = 1, 2), derived from the condition, Wsmi =
Wsmi , is derived as
mi = 0, (98)
and the thermal-stress strengthening in the spherical particle of the isotropic one-particle-
matrix system, p , is given by Eq. (24), considering the particle-matrix boundary radial
stress, pb , derived by Eqs. (88), (130).
2.2.6 Curve Integral of Elastic Energy Density

The curve integrals within the spherical cell of the radius Rc , Wcm1 and Wcm2 , as inte-
grals of wm along the abscissae P4 P5 and P7 P8 in the plane x1 x3 in the positions x1 h0, Ri
and x1 hR, Rci (Fig. 15), respectively, as the elastic energy curve density, equivalent to
those along the abscissae in the planes x1 x2 , x2 x3 due to the isotropy of the multi-particle-
matrix system, are derived as
Z " 5 #
3Rp2b R
Wcm1 = wm dx3 = (s11m s12m) 1
20 Rc
P4 P5
Rc 3Rp2b
= (s11m s12m) , (99)
20
Z 5 " 5 #
3Rp2b R x1
Wcm2 = wm dx3 = (s11m s12m) 1
20 x1 Rc
P7 P8
5
Rc 3Rp2b R
= (s11m s12m) . (100)
20 x1
With regard to the term r6 in Eq. (91), and due to the isotropy of the one-particle-matrix
system, the elastic energy accumulated in the cubic cell matrix between the points P2 and
P3 , and between the points P4 and P5 , is equal to that accumulated between the points at
radii r = |OP2 | = R and r = |OP3 | = Rc, and at the radii r = |OP4 | = x1 and r = |OP5 | = Rc ,
for x1 h0, Ri and x1 hR, Rci, respectively. Accordingly, the term r6 dx3 (see Eqs. (91),
(99), (100)) is replaced by r6 dr, where r hR, |OP3 |i and r hx1 , |OP4 |i for x1 h0, Ri
and x1 hR, Rci, respectively.
The curve integral within the spherical particle, Wcp , as an integral of w p along the
abscissae P3 P4 in the plane x1 x3 in the position x1 h0, Ri (Fig. 15), is given by Eq. (28),
considering the particle-matrix boundary radial stress, pb , derived by Eqs. (88), (130).
2.2.7 Particle and Matrix Crack Formation

With regard to Eqs. (28), (99), Wcp + Wcm1 and Wcm1 represent decreasing function of x1 ,
depending on the parameter R, and consequently the particle and matrix cracks exhibit
shapes as shown in Figs. 9, 12.
Particle crack formation

The critical particle radii, R p1 and R p2 , where R p2 > R p1 , as a reason of the particle crack
formation from the particle centre, O, in the position x1 = 0, to the particle surface (Fig. 9),
and as a reason of the particle crack tip on the particle surface in the position x1 = R,
represent roots of Eqs. (33) and (34) in the forms
2
20s11p KICp
R p1 = , (101)
p2b [10 (s11p + 2s12p ) + s11m s12m]
2
20s11p KICp
R p2 = . (102)
p2b (s11m s12m)
The particle crack radius (Fig. 9), r p , determining the particle crack tip position, repre-
senting a root of Eq. (37), as a function of the variable x1 and the parameter R hR p1 , R p2 i,
is derived as
" #2 1/2
1
2
2s11p KICp R
rp = R2 (s11m s12m ) , (103)
(s11p + 2s12p )2 p2b 10
and the function, f p , describing the particle crack shape in the intervals x1 h0, r p i and
x1 h0, Ri for R hR p1 , R p2 i and R R p2 (Fig. 9), is given by Eqs. (38), (41) and (42),
respectively.
Matrix crack formation
The critical particle radii, Rm1 and Rm2 , as a reason of the matrix crack formation from the
position x1 = 0 to the position x1 = R, and as a reason of the matrix crack tip in the position
x1 = R to the position x1 > R (Fig. 12), represent roots of Eqs. (57) and (58) in the forms
20s11mKICm2
Rm1 = , (104)
ca p2b (s11m s12m )
180 Ladislav Ceniga
2
20s11mKICm
Rm2 = . (105)
ca p2b [10 (s11p + 2s12p ) + s11m s12m]
The matrix crack radii (Fig. 12), rm1 h0, Ri and rm2 > R, both for R > Rm , representing
roots of Eqs. (61) and (80), as functions of the variable x1 and the parameter R, for R
hRm1 , Rm2 i and R Rm2 , respectively, are derived as
( 2 )1/2
2
2s11mKICm
1 R
rm1 = R2 (s11m s12m) , (106)
(s11p + 2s12p )2 2
ca pb 10
1/5
ca R6 p2b (s11m s12m )
rm2 = 2
, (107)
20s11mKICm
and the matrix crack (Fig. 12) is thus described by Eqs. (62), (65), (66) and (81), (84) for
x1 h0, rm1 i and x1 hR, rm2i, related to R Rm1 and R Rm2 , respectively.
3 Thermal Stresses and Related Phenomena in SiC-Si3 N4

Particle-Matrix System
Compared to high strength and wear resistance, ceramic materials are characterized by low
fracture toughness, usually strengthened by addition of particles of higher thermal expan-
sion coefficient than that of a matrix. Increasing fracture toughness of the Si 3 N4 matrix by
addition of the SiC particles of higher thermal expansion coefficient than that of the Si 3 N4
matrix, p > m (Tab. 1), the SiC particle and the Si3 N4 matrix are thus acted by the tensile
thermal stresses, rp = p = p = pb > 0 (see Eqs. (4), (7)) (Fig. 18), and by tensile
and compressive radial and tangential thermal stresses, rm > 0 and m = m < 0 (see
Eqs. (5)-(7)) (Fig. 19), respectively.
Representing a resistance against tensile mechanical loading, the tensile radial ther-
mal stress in the Si 3 N4 cubic cell matrix, as a reason of the Si 3 N4 fracture toughness
increase, and the compressive tangential radial stresses, both decrease within the cubic
cell matrix (Fig. 19), the former from the maximal value on the particle-matrix boundary,
(rm )r=R = pb , where the particle-matrix boundary radial stress, pb , as a function of
the spherical particle volume fraction, v h0, /6), exhibits the maximum for the critical
particle volume fraction, vc = 0.132 (Fig. 18). Similarly, the tensile thermal stresses in
the SiC particle represents a resistance against compressive mechanical loading. Resulting
from experimental results [19]-[21], the SiC-Si 3 N4 multi-particle-matrix system exhibits a
maximal fracture toughness for the SiC particle volume fraction, v 0.15, corresponding
to the calculated value, vc = 0.132.
With regard to the yield stress in tension of the SiC particle, yt p = 1000 MPa, and to
the initial temperature of a cooling process, Ti = 1700C, the v-dependence of the critical
final temperature of a cooling process, Tc (see Eq. (9)), exhibits values < -273.15C in the
1.2 -200
1-4: T = 20, 400, 800, 1200C
Ti = 1700C
1.0 = = = -pb
rp p p
0.8 1: -pb -225
-pb [GPa]
Tc [C]
0.6 2: -pb
0.4 3: -pb -250
0.2 4: -pb
Tc
0.0 -275
0.0 0.1 0.2 0.3 0.4 0.5
v
Figure 18: The tensile radial and tangential thermal stresses in the SiC spherical particle, rp > 0
and p = p > 0, respectively, equal to the particle-matrix boundary radial stress, pb , and the
critical final temperature of a cooling process, Tc , as functions of the spherical particle volume
fraction, v h0, /6).
intervals v h0, 0.07) h0.193, /6) and the maximum for v = vc = 0.132 (Fig. 18), and
the presented calculation is thus considerable in the temperature range T hTc, Ti i.
Table 1: The material constants of the SiC particle and the Si 3 N4 matrix [19]-[21]: the Youngs
modulus, E; the Poissons number, ; the thermal expansion coefficient, ; the yield stress in tension
and compression, yt and yc , respectively; the fracture toughness, KIC ; the SiC particle radius and
volume fraction, R and v, respectively.
E yt yc KIC R v
[GPa] [106 K1 ] [GPa] [GPa] [MPam1/2 ] [nm]
SiC 360 0.19 4.14/5.05 1 5 3.25 10-500 0.05-0.3
Si3 N4 310 0.235 2.35/3.75 1 5 5.25 - -

20 C/1100C
With regard to the material constants listed in Tab. 1 [19]-[21], the dependences in
Figs. 18, 19 and the following ones are generated on the condition of the linear T -dependence
of the thermal expansion coefficient of the particle ( q = p) and the matrix (q = m), q , de-
rived the form
q2 (T T1 ) q1 (T T2 )
q = , (108)
T2 T1
where q1 and q2 are related to the temperature T1 = 20C and T2 = 1100C, respectively,
182 Ladislav Ceniga
1.0
1-4: T = 20, 400, 800, 1200C
1 Ti = 1700C
0.8
v = 0.132
R = 250 nm
rm [GPa]
0.6 2 d = 794 nm
0.4 3
0.2 4
0.0
250 275 300 325 350 375 400
a x1[nm]
0.0
4
-0.1
m= m [GPa]
-0.2 2
1-4: T = 20, 400, 800, 1200C
Ti = 1700C
-0.3 1
v = 0.132
R = 250 nm
d = 794 nm
-0.4
250 275 300 325 350 375 400
b x1[nm]
Figure 19: The tensile and compressive radial (a) and tangential (b) thermal stresses in the Si 3 N4
cubic cell matrix, rm < 0 and m = m > 0, respectively, as functions of the position r = x1
hR, d/2i (Figs. 1-3).
for the spherical particle volume fraction equal to a critical, v = vc = 0.132 (Fig. 18), for
the average SiC particle radius, R = 250 nm, consequently for the inter-particle distance,
d = 794 nm, for the initial and different temperature of a cooling process, Ti = 1700C and
T = 20, 400, 800, 1200C, respectively, and regarding the interval v h0, /6), the formulae
presented in the section 2.2 are considered for v = 0. Furthermore, the v-dependences of
the investigated parameters also result from the fact that a real multi-particle-matrix system
is characterized by different local v, as a consequence of non-homogeneously distributed
particles of different dimensions.
10
1-4: T = 20, 400, 800, 1200C 1 1-4: T = 20, 400, 800, 1200C
Ti = 1700C 8 Ti = 1700C
8 1
6
Wm/ R3 [MJ]
Wp/ R3 [MJ]
6
2
4
4 2
2 2 3
3
4 4
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.0 0.1 0.2 0.3 0.4 0.5
a v b v
20
1-4: T = 20, 400, 800, 1200C
Ti = 1700C
1
15
W/ R3 [MJ]
10
2
5
3
4
0
0.0 0.1 0.2 0.3 0.4 0.5
c v
Figure 20: The elastic energy of the thermal stresses in the SiC spherical particle ( a), Wp /R3 , in
the Si3 N4 cubic cell matrix (b), Wm /R3 , and in the SiC-Si 3 N4 cubic cell (c), W = (Wp +Wm ) /R3 , as
functions of the spherical particle volume fraction, v h0, /6).
The v-dependences of the elastic energy of the thermal stresses acting in the SiC spheri-
cal particle, Wp /R3 (see Eq. (16)), in the Si3 N4 cubic cell matrix, Wm/R3 (see Eq. (17)), and
184 Ladislav Ceniga
in the SiC-Si3 N4 cubic cell, W = (Wp +Wm )/R3 , exhibit the maximum for v = vc = 0.132,
the minimum for v = 0.451 and the maximum for v = 0.057, respectively (Fig. 20). Al-
though the function, W = W (v), does not exhibit the minimum, v = 0.451 may be assumed
to represent a value corresponding to a thermal-stress equilibrium state of the isotropic
SiC-Si3 N4 multi-particle-matrix system. With regard to the parameters, Wp ,Wm,W R3 ,
the v-dependences of Wp , Wm , W are related to the spherical particle radius R = 1.
250
1-4: T = 20, 400, 800, 1200C 1-4: T = 20, 400, 800, 1200C
400 1 Ti = 1700C Ti = 1700C
200
v = 0.132 1 v = 0.132
Wsm [10-9Jm-1]
Wsp [10-9Jm-1]
300 R = 250 nm R = 250 nm

d = 794 nm 150 d = 794 nm
200 2
100 2
100 3 50
3
4 4
0 0
0 50 100 150 200 250 0 100 200 300 400
x1[nm] x1[nm]
600
1 1-4: T = 20, 400, 800, 1200C
500 Ti = 1700C
v = 0.132
R = 250 nm
Ws [10-9Jm-1]
400
d = 794 nm
300 2
200
3
100
4
0
0 100 200 300 400
x1[nm]
Figure 21: The elastic energy gradient of the thermal stresses in the SiC spherical particle and
the Si3 N4 cubic cell matrix, Wsp (a) and Wsm1 , Wsm2 (b), as functions of the position x1 h0, Ri
and x1 h0, Ri, x1 hR, d/2i (Figs. 1-3), respectively, along with the elastic energy gradient of the
thermal stresses in the SiC-Si 3N4 cubic cell, Ws = Wsp +Wsm1 +Wsm2 (c).
If a dislocation moves from the cubic cell surface to the particle-matrix boundary,
the elastic energy gradient within the SiC-Si 3 N4 cubic cell, Ws = Wsp + Wsm1 + Wsm2 (see
Eqs. (18)-(21)), as an increasing function of x1 h0, d/2i (Fig. 21), and the maximum for
x1 = 0, Wsmax (see Eq. (22), (95)), as a function of the spherical particle volume fraction
v h0, /6), exhibiting the maximum for v = 0.096 (Fig. 22), represent an energy barrier
and its height, respectively, influencing a dislocation motion [20, 21]. The elastic energy
gradient, Wsm2 , derived in the interval x1 hR, d/2i by Eq. (20) and (21) exhibits identical
courses. With regard to Wsmax R2 , the v-dependence of Wsmax is related to the spherical
particle radius R = 1.
10
1 1-4: T = 20, 400, 800, 1200C
Ti = 1700C
8
Wsmax / R2 [MJm-1]
6
2
4
2 3
4
0
0.0 0.1 0.2 0.3 0.4 0.5
v
Figure 22: The maximum of the elastic energy gradient of the thermal stresses in the SiC-Si 3 N4
cubic cell, Wsmax /R2 , as a function of the spherical particle volume fraction, v h0, /6).
The elastic energy gradients, Wsm1 and Wsm2 , correspond to thermal-stress strengthening
within the Si 3 N4 cubic cell matrix, m1 and m2 , in the intervals x1 h0, Ri and x1 hR, d/2i
(see Eqs. (24)-(26)) (Fig. 23), respectively. The thermal-stress strengthening in the SiC
spherical particle, p (see Eq. (24)), and the average thermal-stress strengthening in the
Si3 N4 cubic cell matrix, m (see Eq. (27)), as functions of the spherical particle volume
fraction, v h0, /6), are presented in Fig. 24, where | p | exhibits the maximum for v =
vc = 0.132, and m is independent on the parameter R, as resulted from the integration in
Eq. (27) by a numerical method. With regard to p > m , the parameter, m , represents the
strengthening related to tensile mechanical loading. The m v dependence for the SiC-
Si3 N4 multi-particle-matrix system is in an excellent agreement with experimental results
adverting to the tensile strength increase in the range Rm = 0.4 0.45 GPa for v 0.15
and T = 20C [19]-[21].
Resulting from the tensile thermal stresses in the SiC particle and consequently with
regard to the particle crack formation, the SiC particle radius can be varied within the in-
tervals, R (0, R p1 ) and R (0, R p2 ), for v h0, v0 i and v hv0 , /6i (Fig. 25a), where
R p1 and R p2 , are reasons of the particle crack formation from the particle centre, O, to the
particle surface, and vice versa (Figs. 9, 10), respectively, and v0 = 0.487. In addition,
Fig. 25b shows the T -dependence of R p0 = (R p1 )v=v0 = (R p2 )v=v0 for v = v0 , as a reason
186 Ladislav Ceniga
0.6
1-4: T = 20, 400, 800, 1200C
0.5 Ti = 1700C
1 v = 0.132
0.4 R = 250 nm
d = 794 nm
m[GPa]
2
0.3
0.2 3
0.1 4
0.0
0 100 200 300 400
x1[nm]
Figure 23: The thermal-stress strengthening in the Si 3 N4 cubic cell matrix, m1 and m2 , as func-
tions of the position x1 h0, Ri and x1 hR, d/2i (Figs. 1-3), respectively.
of the particle crack formation simultaneously from the particle centre, O, and the particle
surface (Fig. 11).
Although, the critical particle radii, R p1 , R p2 , are extremely high compared to the SiC
particle radius of the interval, R h10, 500i [nm ], as an illustration, Fig. 26 shows the
function, f p (see Eq. (38)), describing the particle crack shape in the interval x1 h0, r p i, in
the form
f p = 1.183 105 0.598x1 + 4.256 103 x21 2.591 109 x31 + 8.383 1014 x41
1.418 1020x51 + 1.452 1025x61 8.815 1029 x71 + 3.246 1034 x81
6.569 1038x91 + 5.624 1042x10
1 ... (109)
and accordingly the contribution of the term x21 is considerable for x1 r p , and the contri-
bution of the terms xn1 for n 3 is neglecting, where r p = 24.966 m for v = vc = 0.132,
R = 45 m. The quasi-linear shape corresponds to that from experimental results [9]-[11],
[19]-[21].
4 Conclusions
The paper is the continuation of the calculation published in [18] and [25], presenting the
thermal stresses in the isotropic multi- and one-particle-matrix systems, respectively, origi-
nating during a cooling process as a consequence of the difference of thermal expansion co-
efficients between the particle and the matrix. The isotropic multi- and one-particle-matrix
systems are represented by the homogeneously distributed spherical particles embedded in
0.0
1-4: T = 20, 400, 800, 1200C 4
Ti= 1700C 3
-0.2
2
-0.4
p [GPa] 1
-0.6
-0.8
-1.0
-1.2
0.0 0.1 0.2 0.3 0.4 0.5
a v
1.0
1-4: T = 20, 400, 800, 1200C 1
Ti= 1700C
0.8
2
0.6
m [GPa]
3
0.4

4
0.2
0.0
0.0 0.1 0.2 0.3 0.4 0.5
b v
Figure 24: The thermal-stress strengthening in the SiC spherical particle, p , and the average
thermal-stress strengthening in the Si 3 N4 cubic cell matrix, m , as functions of the spherical particle
volume fraction, v h0, /6).
188 Ladislav Ceniga
10.00
1-4: T = 20, 400, 800, 1200C
Ti= 1700C 4
1.00 3
Rp [mm] 2
1
0.10
0.01
0.0 0.1 0.2 0.3 0.4 0.5
a v
1E+3
Ti= 1700C
1E+2
Rp0 [mm]
1E+1
1E+0
1E-1
0 500 1000 1500
b T [C]
Figure 25: The critical SiC particle radii, R p1 and R p2 (a), as functions of the spherical particle
volume fraction, v h0, v0 i and v hv0 , /6), respectively, and the critical SiC particle radius, R p0 =
(R p1 )v=v0 = (R p2 )v=v0 (b), as a function of the final temperature of a cooling process, T h20, 1700i
[ C ] for v0 = 0.487.
12
T = 20C
10 Ti = 1700C
v = 0.132
8 R = 45 m
fp [ m]
6
0
0 5 10 15 20 25
x1[ m]
Figure 26: The function, f p , describing the SiC particle crack for the spherical particle volume
fraction, v = vc = 0.132, and for the SiC particle radius, R > (R p1 )v=vc = 38.6 m (Fig. 25a).
the infinite matrix divided into cubic cells, containing one central particle [18], and by one
spherical particle embedded in the infinite matrix [25], respectively.
Contributed to the results published in [18] and [25] (see sections 2.1.1, 2.1.2, 2.2.1),
the results of the presented calculation related to isotropic multi- and one-particle-matrix
systems are as follows:
The initial and critical final temperature of a cooling process between which the
isotropic particle-matrix systems are acted by elastic thermal stresses are derived (see
sections 2.1.3, 2.2.2).
The thermal-stress induced elastic energy density in a point, consequently on a sur-

face, and along a curve in the spherical particle, in the cubic cell matrix and in the
infinite matrix of the isotropic particle-matrix systems are derived (see sections 2.1.4,
2.1.5, 2.1.7, 2.2.3, 2.2.4, 2.2.6).
Resulting from the surface elastic energy density, the thermal-stress strengthening
in the spherical particle, the cubic cell matrix and the infinite matrix of the isotropic
particle-matrix systems are derived (see sections 2.1.6, 2.2.5).
Resulting from the curve elastic energy density, the sections 2.1.8, 2.2.7 devoted to
the particle and matrix crack formation include
the condition concerning a direction of the crack formation in the spherical

particle, the cubic cell matrix and the infinite matrix of the isotropic particle-
matrix systems is derived,
190 Ladislav Ceniga
the critical particle radii, as a reason of the crack formation in the spherical
particle, the cubic cell matrix and the infinite matrix of the isotropic particle-
matrix system are derived,
the functions describing cracks in the spherical particle, the cubic cell matrix
and the infinite matrix of the isotropic particle-matrix systems are derived.
Applying the derived formulae to the SiC-Si 3 N4 multi-and one particle-matrix systems,
the main results included in the section 3 are as follows:
The SiC spherical particle thermal stresses, the particle-matrix boundary radial stress
and the critical final temperature of a cooling process, as functions of the spherical
particle volume fraction, v, are presented (Fig. 18).
The critical SiC particle volume fraction, vc = 0.132 (Fig. 18), corresponding to max-
imal calculated thermal stresses, is in an excellent agreement with the SiC particle
volume fraction, v 0.15, corresponding to a maximal fracture toughness as ob-
tained from published experimental results [19]-[21].
The thermal stresses, as functions of the position in the Si 3 N4 cubic cell matrix, are
presented for the v = vc = 0.132 and for the average SiC particle radius, R = 250 nm
(Fig. 19).
The thermal-stress induced elastic energy in the SiC spherical particle, the Si 3 N4
cubic cell matrix and SiC-Si3 N4 cubic cell, as functions v, are presented (Fig. 20).
The thermal-stress induced elastic energy gradient, as a function of the position in

the SiC-Si3 N4 cubic cell, is presented (Fig. 21) along with the v-dependence of the
gradient maximum (Fig. 22).
The thermal-stress strengthening, as a function of the position in the Si 3 N4 cubic

cell matrix, is presented (Fig. 23) along with the v-dependences of the thermal-stress
strengthening in the SiC spherical particle and the average thermal-stress strength-
ening in the Si 3 N4 cubic cell matrix (Fig. 24), being in an excellent agreement with
published experimental results [19]-[21].
The critical SiC particle radii, as reasons of the SiC particle crack formation and
as functions of v, are presented along with the temperature dependence of the SiC
critical particle radius (Fig. 25).
The function describing the particle crack for the critical particle volume fraction and
for the SiC particle radius greater than critical is presented (Fig. 26).
Acknowledgement
This work was supported by the Slovak Grant Agency VEGA (2/4173/04, 2/4175/04,
2/4062/04, 1/1111/04), by NANOSMART, Centrum of Excellence, Slovak Academy of
Sciences, by Science and Technology Assistance Agency under the contract No. APVT-51-
049702, by EU 5th FP project GRD1-2000-25352 SmartWeld, and by COST Action 536
and COST Action 538.
The author is thankful to his dearest parents for their support.
5 Appendix
The elastic modulus of the particle (q = p) and the matrix (q = m), sq , has the forms
1
s11q = , (110)
Eq
q
s12q = , (111)
Eq
where Eq and q are the Youngs modulus and Poissons number, respectively.
The coefficient ci (i = 1 19) is as follows
3 (s11m + s12m)
c1 = , (112)
2 ( 6v)
c2 = c1 + s11p + 2s12p (s11m + 2s12m) , (113)
!3
Z/4
p cos4 1 cos2 d,
c3 = p (114)
1 + cos2 3 1 + cos 2
0
where the coefficient c3 can be derived by the Taylor series for the integrated function.
Consequently, after integration of the Taylor series in terms of ( 0 )n about the point
0 = 0, the coefficient c3 = 0.337497 for n = 100;
2
c1
c4 = 2v (s11p + 2s12p )
c2
"
1 2 6vc1 2
+ 2 (1 v) (s11m + 2s12m) 1
( 6v)2 c2
2 #
c1
+ v (s11m s12m ) 2 36vc3 1 , (115)
c2
192 Ladislav Ceniga

2c1 (s11p + 2s12p ) 1 6vc1
c5 = + 12 (1 v) (s11m + 2s12m ) 1
c2 ( 6v)2 c2

c1
+ (s11m s12m ) 2 36vc3 1 , (116)
c2
[c2 c4 + vc5 (c1 c2 )]

c6 = , (117)
( 6v) (c2 c4 + vc1 c5 )
c2 c4 + vc1 c5
c7 = , (118)
c22 c4
v [c2 c5 6 (c2 c4 + vc1 c5 )]
c8 = , (119)
( 6v) (c2 c4 + vc1 c5 )
4/3
3 + 8 3v
c9 = , x1 = 0, (120)
2R4 4
Z/4
" #2
4 (3v)2/3 cos2 d
c9 = d, x1 0, , (121)
R2 (4)2/3 + 4x1 (3v)2/3 cos2 2
0
where the coefficient c9 for x1 (0, d/2i can be derived for a concrete isotropic particle-
matrix system by the Taylor series for the integrated function in terms of ( 0 )n about
the point 0 = 0;
h i
2x1 R (12v)1/3 3R2 (4)2/3 + 20x21 (3v)2/3
c10 = 373 + h i2 , (122)
R2 (4)2/3 + 4x21 (3v)2/3
Z73
c11 = 6 cos4 d, (123)
0
Z22
R4 (4)4/3 cos5 R (4)1/3
c12 = h i2 d = q
0 R2 (4)2/3 cos2 + 4x21 (3v)2/3 R2 (4)2/3 + 4x21 (3v)2/3
(
R2 (4)2/3 + 3x21 (3v)2/3

R2 (4)2/3 + 2x21 (3v)2/3
" #)
2x1 2 3v 2/3 R2 (4)2/3 + 3x21 (3v)2/3 1 R 2 4 2/3
ln 1 + ,
R 4 R2 (4)2/3 + 4x2 (3v)2/3 1
2 x1 3v
(124)
Z73
R2 (4)2/3 cos5 (3v)1/3
c13 = d = q
R2 (4)2/3 cos2 + 4x21 (3v)2/3 R2 (4)2/3 + 4x21 (3v)2/3
0
(
R2 (4)2/3 4x21 (3v)2/3 4R3
h i
R (12v)1/3 3 (3v)1/3 R2 (4)2/3 + 4x21 (3v)2/3
" #)
6vx41 1 R 2 4 2/3
+ ln 1 + , (125)
R3 2 x1 3v
Z73
(4)4/3 R4 cos7 1
c14 = h i2 d = q
0 R2 (4)2/3 cos2 + 4x21 (3v)2/3 R2 (4)2/3 + 4x21 (3v)2/3

R2 (4)2/3 8x2 (3v)1/3 3R4 (4)4/3 16x41 (3v)4/3
1 h i
+
R (4) 1/3
4R (4)1/3 R2 (4)2/3 + 2x21 (3v)2/3
4R3
h i
3 R2 (4)2/3 + 4x21 (3v)2/3
h i "
5/3 4/3 4/3 #
1 R 2 4 2/3
5
2R (4) 4 4
R (4) 6x1 (3v)
+ h i4 ln 1 + ,(126)
2 x1 3v

R2 (4)2/3 + 4x21 (3v)2/3
1/3 " 2 2/3 #

3c10 2x1 3v 2x1 3v
c15 = 3c11 + c12 + 4c13 c14 , (127)
16 R 4 R 4
where the coefficient c11 can be derived for a concrete isotropic particle-matrix system by
the Taylor series for the integrated function in terms of ( 0 )n about the point 0 = 0;
" 1/3 #
R 4
73 = arctan , (128)
2x1 3v
" 1/3 #
2x1 3v
6 = arctan , (129)
R cos 4
" #
s11m 2 4 2/3
c16 = R R2 x21 , x1 h0, Ri, (130)
2 3v
2/3
s11mR2 4
c17 = , (131)
2 3v
194 Ladislav Ceniga
n
c18 = 5v2 (s11m + 2s12m )[c5 (6vc1 c2 ) + 6c2 c4 ]2
o
(s11p + 2s12p ) ( 6v)2 (c2 c4 + vc1 c5 )2
5/3
2 2 3v
+ 16 (s11m s12m) [vc5 (c1 c2 ) + c2 c4 ]
4
5

1/3

(4)
1q , (132)

(4)2/3 + 4 (3v)2/3
1
c19 = s11p + 2s12p + (s11m s12m ) . (133)
2
References
[1] Sivak, P.; Ostertag, O.; Ivanco, V.; Kostolny, K. Acta Mech. Slovaca 1998, vol. 2,
201-204.
[2] Pesek, L.; Novotny, L.; Ivanco, V. Acta Mech. Slovaca 1999, vol. 3, 171-174.
[3] Ivanco, V.; Kubn, K.; Kostolny, K. Finite element method I; ISBN 80-96731-4-0;
Elfa: Kosice, SK, 1994; Vol. 1, pp 225-243 (in Slovak).
[4] Hvizdos, P.; Lofaj, F.; Dusza, J. Metall. Mater. 1995, vol. 6, 473-483.
[5] Lofaj. F.; Dusza J.; Richarz, B. Metall. Mater. 1997, vol. 35, 247-256.
[6] Lofaj, F.; Dorcakova, F. Metalurgija 2003, vol. 42, 229-233.
[7] Lofaj, F.; Hvizdos, P.; Dorcakova, F.; Satet, R.; Hoffmann, M.J.; Arellano
Lopez, A.R. Mater. Sci. Eng. 2003, vol. A357, 181-187.
[8] Lofaj, F.; Dorcakova, F.; Kovalck, J.; Hoffmann, M.J.; Arellano Lopez, A.R. Metall.
Mater. 2003, vol. 41, 145-157.
[9] Diko, P. Supercond. Sci. Technol. 1998, vol. 11, 68-72.
[10] Sefckova, M.; Zmorayova, K.; Diko, P.; Babu, H.; Cardwell, D. Acta Metallur. Slo-
vaca 2004, vol. 10, 889-891.
[11] Zmorayova, K.; Sefckova, M.; Diko, P.; Babu, H.; Cardwell, D. Acta Metallur. Slo-
vaca 2004, vol. 10, 892-894.
[12] Janovec, J.; Guth, A.; Vyrostkova, A.; Stefan, B. Metall. Mater. 1988, vol. 26, 714-
724.
[13] Janovec, J.; Vyrostkova, A.; Holy, A. J. Mater. Sci. 1992, vol. 27, 6564-6572.
[14] Janovec, J.; Vyrostkova, A.; Svoboda, M. Metall. Mater. Trans. 1994, vol. 25A, 267-
275.
[15] Janovec, J.; Vyrostkova, A.; Holy, A. Canadian. Metall. Quart. 1994, vol. 33, 227-
232.
[16] Kovac, F.; Dzubinsky, M.; Sidor, Y. J. Mag. Mag. Mater. 2004, vol 269, 333-340.
[17] Sidor, Y.; Dzubinsky, M.; Kovac F. Mater. Character. 2003, vol. 51, 109-116.
[18] Ceniga, L. J. Therm. Stress. 2004, vol. 27, 425-432.
[19] Skocovsky, P.; Bokuvka, O.; Konecna, R.; Tillova, E. Materials Science for Mechan-
ical Engineers; ISBN 80-7100-831-1; EDIS Technical University in Zilina: Zilina,
SK, 2001; pp 36-63 (in Slovak).
[20] Skocovsky, P.; Palcek, P.; Konecna, R.; Varkoly, L. Structural materials; ISBN 80-
7100-608-4; EDIS Technical University in Zilina: Zilina, SK, 2000; pp 41-45 (in
Slovak).
[21] Hidveghy. J.; Dusza, J. Nonmetallic Structural Materials. Plastic and Structural Ce-
ramics; ISBN 80-7090-363-4; Technical University in Kosice: Kosice, SK, 1998;
pp 126-144 (in Slovak).
[22] Mizutani, T. J. Mater. Res. 1996, vol. 11, 483-494.
[23] Trebuna, F.; Jurica, V.; Simcak, F. Theory of elasticity; ISBN 80-7099-478-9; Vien-
ala: Kosice, SK, 2000; Vol. 2, pp 72-81 (in Slovak).
[24] Kluvanek, I.; Misk, L.; Svec, M. Mathematics; Alfa: Bratislava, SK, 1959; Vol. 1,
pp 685-687 (in Slovak).
[25] Ceniga, L.; Kovac, F. Mater. Sci. Eng. 2001, vol. B86, 178-181.
Chapter 6
NOVEL BONE-REPAIRING MATERIALS: BIOACTIVE

ORGANIC-INORGANIC HYBRIDS
Masanobu Kamitakahara*
Graduate School of Materials Science, Nara Institute of Science and Technology,
8916-5, Takayama-cho, Ikoma-shi, Nara 630-0192, Japan
Abstract
In general, artificial materials implanted into bone defects are encapsulated by a fibrous tissue.
Some ceramics, such as Bioglass, sintered hydroxyapatite and glass-ceramic A-W, however,
form a bone-like apatite layer on their surfaces in the living body and bond to living bone
through this apatite layer, i.e. they show bioactivity. Although these bioactive ceramics are
used clinically as important bone-repairing materials, they are essentially brittle and hence
limited in their applications. It is desirable to develop new types of deformable bioactive
materials. Organicinorganic hybrids prepared by a sol-gel method are expected to exhibit
bioactivity as well as deformability. Polydimethylsiloxane (PDMS)-CaO-SiO2-based hybrids
were prepared. To evaluate their potential of bone-bonding property, their ability of apatite
formation was examined in a simulated body fluid (SBF). Since the apatite formation on
bioactive materials in the body can be reproduced even in SBF, we can estimate the
bioactivity of the material by using SBF. Mechanical properties were also examined. Some of
them showed apatite-forming ability and mechanical properties analogous to those of human
cancellous bone. These CaO-containing hybrids, however, showed a decrease in mechanical
strength in SBF. These hybrids containing no CaO do not form apatite on their surfaces in
SBF. We then prepared CaO-free PDMS-TiO2 and poly(tetramethylene oxide) (PTMO)-TiO2
hybrids in which anatase was precipitated by hot-water treatments. These hybrids also showed
apatite-forming ability on their surfaces in SBF and flexibility, and they showed little decrease
in mechanical strength in SBF. In conclusion, it was revealed that we can design new
bioactive materials for bone-repair by controlling the compositions and structures in organic-
inorganic hybrid systems.
*
E-mail address: kamitaka@ms.naist.jp. Tel: +81-743-72-6122, Fax: +81-743-72-6129
198 Masanobu Kamitakahara
1 Introduction
Bone supports our body, protects our organs, such as brain and heart against external force,
and enables us to do various kinds of motion. Once even one of the bones is damaged by a
disease or an accident, quality of our lives is much decreased. Defects of bones have been
repaired by autogenous or allogenic bones. In the case of autogenous bones, however, a
healthy part of the patient is damaged and available amount of bone is limited. In the case of
allogenic bones, infections are sometimes caused through viruses, bacteria, etc. Therefore,
artificial materials, which are free from virus and bacteria and not limited in its available
amount, are needed to repair the defects of bones. Artificial materials implanted into bone
defects, however, are generally encapsulated by a non-calcified fibrous tissue to be isolated
from the surrounding bone [1].
In early 1970s, Hench et al. showed that some glasses in the system Na2O-CaO-SiO2-
P2O5 spontaneously bond to living bone without forming the fibrous tissue around them and
named them Bioglass [2-4]. Since the discovery of Bioglass, various kinds of ceramics,
such as glass-ceramic Ceravital containing crystalline apatite [5, 6], sintered hydroxyapatite
(Ca10(PO4)6(OH)2) [7, 8], glass-ceramic A-W containing crystalline apatite and wollastonite
[9, 10], glass-ceramic Bioverit containing crystalline apatite and fluorophlogopite ((Na,
K)Mg3(AlSi3O10)F2) [11, 12], MgO-CaO-SiO2-P2O5 glasses [13] and CaO-SiO2 glasses [14],
have been found to bond to living bone. These materials are called bioactive materials, and
some of them are already clinically used as important bone substitutes, such as artificial iliac
crests, artificial vertebrae, artificial intervertebral discs, bone fillers, etc [15-24].
They, however, occupy only 10 to 30 % of whole the bone grafts needed, because they
are brittle, poor in fracture toughness and deformability, and too high in elastic modulus.
Among the problems described above, the low fracture toughness has been solved by
development of bioactive metals with high fracture toughness. For example, Kim et al.
showed that titanium metal and its alloys bond to living bone, if they are subjected to prior
NaOH and heat treatments to form a sodium titanate layer on their surfaces [25-34]. Miyazaki
et al. showed that tantalum metal also bonds to living bone, if it is subjected to prior NaOH
and heat treatments to form a sodium tantalate layer on its surface [35-37]. These metallic
materials, however, have higher elastic moduli than that of human cortical bones, and hence
are liable to induce resorption of the surrounding bone due to stress shielding by them.
Bioactive materials with low elastic moduli and high deformability are desired to be
developed.
Living bone is a composite of 70 wt% inorganic component, apatite, and 30 wt% organic
component, collagen, as shown in Fig. 1 [38, 39]. Due to its unique composition and structure,
natural bone can show not only high strength and high fracture toughness, but also
deformability and low elastic modulus. Therefore, it is expected that a bioactive material with
lower elastic modulus and deformability could be obtained if a bioactive inorganic substance
is combined with some organic substance. Bonfield et al. early developed a composite of
hydroxyapatite granules with polyethylene [40, 41]. These composites, however, lose
deformability when the hydroxyapatite content exceeds 40 vol%. When the hydroxyapatite
content is limited to less than 40 vol%, the composite can not show high bioactivity [3].
Novel Bone-Repairing Materials: Bioactive Organic-Inorganic Hybrids 199
2 Bone-Bonding Mechanism
It has been shown that the bioactive materials bond to living bone through an apatite layer
which is formed on their surfaces in the body [3, 31, 42-49]. The schematic representation is
shown in Fig. 2. Therefore, it is believed that the essential requirement for an artificial
material to bond to living bone is formation of an apatite layer on their surfaces in the living
body. This apatite layer consists of calcium-deficient and carbonate-containing nano-sized
hydroxyapatite with low crystallinity and defect structure [50, 51]. This apatite is similar to
the bone mineral in composition and structure, and hence bone-producing cell called
osteoblast can proliferate preferentially on this surface apatite layer, and differentiate to
produce bone matrix composed of apatite and collagen [4, 52, 53]. As a result, the
surrounding bone can come into direct contact with the surface apatite layer. When this
occurs, a tight chemical integration is formed between the bone apatite and the surface apatite
on bioactive materials in order to decrease interfacial energy between them.
Kokubo et al. [54] previously showed that the apatite formation on the surfaces of
bioactive materials in the living body can be reproduced even in an acellular protein-free
simulated body fluid (SBF) with ion concentrations nearly equal to those of human blood
plasma. Ion concentrations of human blood plasma and SBF are shown in Fig. 3. This
indicates that bioactivity of a material can be evaluated even in vitro by examining the
apatite-formation on its surface in SBF.
3 Design of Bioactive Hybrids

Currently, organically modified sol-gel-derived metal oxides are paid attention because of
their unique properties, such as low elastic modulus and deformability [55-59]. In these
hybrid materials, organic components are chemically incorporated into inorganic networks at
nanometer level. It has been revealed that some sol-gel-derived metal oxide gels, such as
silicon oxide [60-63], titanium oxide [63-65], tantalum oxide [66], zirconium oxide [67, 68]
and niobium oxide [69], form a bone-like apatite layer on their surfaces in SBF. Therefore, it
is expected that if some organic component is incorporated into a network of these metal
oxide gels by a sol-gel method, the obtained hybrids could show not only apatite-forming
ability but also deformability. Moreover, if calcium ions are incorporated into these hybrids,
the obtained hybrid could show higher apatite-forming ability, because the incorporated
calcium ions are released into the surrounding body fluid to accelerate apatite nucleation by
increasing the ionic activity product of apatite. [70].
4 Polydimethylsiloxane (PDMS)-CaO-SiO2-Based Hybrids

Modification of CaO-SiO2 system with organic polymer may provide bioactive organic-
inorganic hybrids. Tsuru et al. [71] prepared polydimethylsiloxane (PDMS)-CaO-SiO2
hybrids by a sol-gel method. Tetraethoxysilane (TEOS) was hydrolyzed and reacted with
PDMS, and calcium nitrate (Ca(NO3)2) was incorporated. The model structure of the PDMS-
CaO-SiO2 hybrid is shown in Fig. 4. Some of the resultant hybrids formed an apatite layer on
their surfaces in SBF. This indicates that these hybrids are expected to form a bone-like
apatite layer on their surfaces in the living body and bond to living bone through this apatite
layer [72]. However, they did not show their mechanical properties because they could not
prepare large enough specimens for mechanical test. Recently, Chen et al. [73-77] and we [78,
79] successfully synthesized large samples enough for mechanical test in PDMS-CaO-SiO2-
TiO2 and PDMS-CaO-SiO2 hybrids and revealed that some of the hybrids show both apatite-
forming ability on their surfaces in SBF and deformability.
We prepared the PDMS-CaO-SiO2 hybrids whose starting compositions are given in

Table 1 [78, 79]. The surfaces of the hybrids before and after soaking in SBF were observed
by scanning electron microscopy (SEM). Figure 5 shows SEM photographs of the surfaces of
the PDMS-CaO-SiO2 hybrids before and after soaking in SBF for 7 days. Apatite-forming
ability of the PDMS-CaO-SiO2 hybrids increased with decreasing PDMS content. When
PDMS content increases, the amount of silanol groups which induce apatite nucleation
decreases, accordingly the apatite-forming ability decreases. The CaO-free hybrid (Ca0) did
not form apatite on its surface. The calcium-containing hybrids release calcium ions and the
released calcium ions accelerated apatite nucleation of apatite [70]. Once the apatite nuclei
are formed, they can grow spontaneously by consuming calcium and phosphate ions from the
surrounding SBF, since SBF is highly supersaturated with respect to apatite. Figure 6 shows
stress-strain curve of one of the PDMS-CaO-SiO2 hybrids, in comparison with that reported
for human cancellous bone [80]. Unlike usual brittle ceramics, the hybrid was deformable and
showed mechanical properties analogous to those of human cancellous bone.
Modification of inorganic component CaO-SiO2 with organic component PDMS gives a
hybrid material with apatite-forming ability and mechanical properties analogous to those of
human cancellous bone.
5 Degradation of PDMS-CaO-SiO2-Based Hybrids in SBF

For the clinical application of these hybrids, to know their changes in mechanical properties
in the body environment is important. The mechanical properties of PDMS-CaO-SiO2-TiO2
hybrids were examined before and after soaking in SBF [81]. The starting compositions of
PDMS-CaO-SiO2-TiO2 hybrids are given in Table 2. The changes in mechanical strength and
strain at failure of the PDMS-CaO-SiO2-TiO2 hybrids before and after soaking in SBF are
shown in Fig. 7. After soaking in SBF, the bending strength and strain to failure of hybrids
Ti0Ca0 and Ti10Ca0 only slightly decreased, whereas those of hybrids Ti0Ca15, Ti10Ca15,
Ti30Ca0, and Ti30Ca15 significantly decreased after soaking in SBF. The decreases in both
the bending strength and the strain to failure of the calcium-containing hybrids due to soaking
in SBF were larger than those of the corresponding calcium-free hybrids. It has been reported
that calcium ions inhibit the polycondensation of TEOS in the preparation of a CaO-SiO2
glass by a sol-gel method [82]. Therefore, it is considered that calcium ions inhibit the
formation of a continuous three-dimensional network of the hybrids in the calcium-containing
hybrids. In such cases, calcium ions are easily released via exchange with hydronium ions in
SBF and water molecules also easily enter into the network of calcium-containing hybrids to
break the Si-O-Si bonds of the hybrids. These may be the reasons why the deterioration of the
calcium-containing hybrids was larger than that of the calcium-free hybrids in SBF.
It can be concluded that incorporation of a large amount of CaO into the PDMS-CaO-
SiO2-based hybrid system results in deterioration of the hybrids in the body environment.
6 PDMS-TiO2 Hybrids
Preparation of bioactive CaO-free PDMS-modified hybrids that can form apatite on their
surfaces was attempted. Recently, it has been shown that apatite formation on titania gel
significantly depends on its structure. The Ti-OH groups in the anatase structure are most
effectively induce apatite nucleation, whereas Ti-OH groups in the amorphous and rutile
structure are not so effective [65]. It is expected that even a CaO-free hybrid can show high
apatite-forming ability if the structure of a TiO2-based hybrid is controlled.
A PDMS-TiO2 hybrid was synthesized and its structure was controlled by hot-water
treatment [83]. A polydimethylsiloxane (PDMS)-TiO2 hybrid was prepared by a sol-gel
method from PDMS (M=550) and tetraethylorthotitanate with PDMS/tetraethylorthotitanate
molar ratios at 1.35. The as-prepared hybrid was amorphous and did not form apatite on its
surface after soaking in SBF for 7 days. The PDMS-TiO2 hybrid was then soaked in hot water
at 60C or 80C for various periods. The structural change was examined by thin film X-ray
diffraction (TF-XRD). Figure 8 shows TF-XRD patterns of the surfaces of the PDMS-TiO2
hybrid subjected to the hot-water treatment at 60C or 80C for various periods. Anatase
peaks were observed for the PDMS-TiO2 hybrid treated with hot water at 60C within 7 d and
at 80 C within 1 d. The peak intensities of anatase increased with increasing period and
temperature of the hot-water treatment. Transmission electron microscopic (TEM)
observation showed that the hybrid took a homogeneous amorphous structure before the hot-
water treatment, and precipitated anatase particles 10-20 nm in size after the hot-water
treatment at 80C for 7 days (Fig. 9). The hot-water-treated hybrids were soaked in SBF. The
SEM photographs of surfaces of the hot-water-treated PDMS-TiO2 hybrids after soaking in

SBF are shown in Fig. 10. The apatite was formed on the hybrid treated with hot water at
60C for the periods longer than 3 days and at 80C for the periods longer than 1 day. The
amount of apatite formed on them increased with increasing period and temperature of the
hot-water treatment. This indicates that the anatase precipitated in the hybrid induces apatite
formation on the surface of the hybrid. It has been shown that apatite formation on titania
depends mainly on its crystalline phase, and anatase is more effective for apatite formation
than amorphous and rutile-type titania [65].
The obtained hybrid showed rubber elasticity even after the hot water treatment, as
shown in Fig. 11. The representative stress-strain curves of the PDMS-TiO2 hybrid before and
after the hot-water treatment, and after both hot-water treatment and SBF treatment are shown
in Fig. 12. Both tensile strength and Youngs modulus of the hybrid decreased, and its strain
to failure increased after the hot-water treatment. It is considered that the matrix of the hybrid
was enriched with PDMS due to the segregation of TiO2 by the hot-water treatment and hence
the mechanical properties of the hybrid subjected to the hot-water treatment became close to
the matrix rich in PDMS, since the continuous phase predominantly governs the overall
mechanical behavior of composite materials. The observation that the strain energies stored
until failure were not significantly changed by the hot-water treatment might indicate that
rearrangement of the hybrid by the hot-water treatment does not cause deterioration of the
hybrid. After soaking in SBF, its strain to failure decreased, but its tensile strength did not
change and its Youngs modulus a little increased. It is considered that an increase in its
Youngs modulus after soaking in SBF is due to apatite formation on its surface, because
apatite shows much higher Youngs modulus than the hybrid. In this case, its strain to failure
decreases when its tensile strength does not change.
Since the environment in hot water is more severe than in the body environment, the
unreacted reagents or solvents remaining in the hybrid might be released into hot water, and
therefore the resultant hybrid is expected to be non-toxic and stable in the living body at
around 36.5 C.
A highly deformable PDMS-TiO2 hybrid with apatite-forming ability and durability in
SBF was obtained by the present method. Besides deformability, rubbery elasticity of the
hybrid would also be an advantage, because the hybrid can return to the original shape, even
after a large deformation.
7 Poly(Tetramethylene Oxide) (PTMO)-TiO2 Hybrids

The PDMS-TiO2 hybrid treated with hot water showed apatite-forming ability and high
deformability. The mechanical strength (0.4 MPa) and Youngs modulus (0.2 MPa) of the
obtained hybrid, however, were much lower than those of human bones. It is expected that if
PDMS is replaced with poly(tetramethylene oxide) PTMO, a deformable bioactive material
with higher mechanical strength might be obtained [84].
Poly(tetramethylene oxide) (PTMO)-TiO2 hybrids were prepared by a sol-gel method
from triethoxysilane functionalized PTMO (Si-PTMO) and tetraisopropyltitanate with weight
ratios of 30/70, 40/60 and 50/50 (hybrids PT30, PT40 and PT50, respectively), and
subsequently subjected to a hot-water treatment at 95 C for 2 d [85]. The structural model of
the PTMO-TiO2 hybrid is shown in Fig. 13. All of the obtained hybrids were amorphous
before the hot-water treatment, and precipitated nanosized anatase after the hot-water
treatment. The amount of precipitated anatase increased with decreasing PTMO content.
Apatite was not formed on the surfaces of the hybrids in SBF before the hot-water treatment,
but was formed after the hot-water treatment, and its amount increased with decreasing
PTMO content, as shown in Fig. 14. This indicates that apatite-formation was induced by the
anatase precipitated by hot-water treatment, as the case of the PDMS-TiO2 hybrid. Figure 15
shows representative stress-strain curves of hybrid PT40 before and the hot-water treatment,
and that after both the hot-water treatment and SBF treatment for 28 days, in comparison with
that reported for human cancellous bone. Hybrid PT40 showed strength and Youngs
modulus analogous to those of human cancellous bones, and high deformability after the hot-
water treatment. The PDMS-TiO2 hybrid treated with hot water showed strength of only 0.4
MPa and Youngs modulus of 0.2 MPa. The mechanical properties of the hybrid were
improved by changing the organic component from PDMS to PTMO, from the point of view
of strength and Youngs modulus. Although the PDMS-TiO2 hybrid treated with hot water
can deform up to 200%, such extremely high deformability is not needed and 10%
deformability of the PTMO-TiO2 hybrid treated with hot water is considered to be enough for
a bone-repairing material. After soaking in SBF, its Youngs modulus increased, and its
bending strength and strain to failure decreased. The increase in its Youngs modulus after
soaking in SBF might be due to the apatite-formation on its surface because apatite shows
much higher Youngs modulus than the hybrid. The decreases in its bending strength and
strain to failure after soaking in SBF might be due to the degradation of the hybrid. These
decreases were, however, much lower than those of a CaO-containing PTMO-CaO-TiO2
hybrid [81].
8 Summary
Development of new types of deformable bioactive materials is desired for bone repair. A
bioactive material which shows mechanical properties analogous to those of human bones
and high durability in the body environment can be obtained by controlling not only the
compositions but also the structures of sol-gel-derived organic-inorganic hybrids. The hybrid
material can be a candidate for a novel bone-repairing material.
References
[1] Hulbert, SF. The use of alumina and zirconia in surgical implants. In: Hench LL, Wilson
J. editors. An Introduction to Bioceramics. Singapore: World Scientific; 1993; 25-40.
[2] Hench, LL; Splinter, RJ; Allen, WC; Greenlee, TK. Bonding mechanism at the interface
of ceramics prosthetic materials. J. Biomed. Mater. Res. Symp., 1971, 2, 117-141.
[3] Hench, LL. Bioceramics: From concept to clinic. J. Am. Ceram. Soc., 1991, 74, 1487-
1510.
[4] Hench, LL; Andersson, . Bioactive glasses, In: Hench LL, Wilson J. editors. An
Introduction to Bioceramics. Singapore: World Scientific; 1993; 41-62.
[5] Brmer, H; Pfeil, E; Ks, HH. Glass-ceramic material. German Patent No. 2, 1973, 326,
100.
[6] Gross, UM; Mller-Mai, C; Voigt, C. Ceravital bioactive ceramics. In: Hench LL,
Wilson J. editors. An Introduction to Bioceramics. Singapore: World Scientific; 1993;
105-124.
[7] Jarcho, M; Kay, JL; Gumaer, RH; Drobeck, HP. Tissue, cellular and subcellular events at
bone-ceramic hydroxyapatite interface. J. Bioeng., 1997, 1, 79-92.
[8] LeGeros, RZ; LeGeros, JP. Dense hydroxyapatite. In: Hench LL, Wilson J. editors. An
[9] Kokubo, T; Shigematsu, M; Nagashima, Y; Tashiro, M; Nakamura, T; Yamamuro, T;
Higashi, S. Apatite- and wollastonite-containing glass-ceramics for prosthetic application.
Bull. Inst. Chem. Res., Kyoto Univ., 1982, 60, 260-268.
[10] Kokubo, T. A/W glass-ceramic: Processing and properties. In: Hench LL, Wilson J.
editors. An Introduction to Bioceramics. Singapore: World Scientific; 1993; 75-88.
[11] Hland, W; Nawmann, K; Vogel, W; Gummel, J. Machinable bioactive glass ceramic.
Wiss. Ztschr. Friedrich-Schiller-Univ. Jena, Math.-Naturwiss. Reihe., 1983, 32, 571-580.
[12] Hland, W; Vogel, W. Machinable and phosphate glass-ceramics. In: Hench LL, Wilson
J. editors. An Introduction to Bioceramics. Singapore: World Scientific; 1993; 125-137.
[13] Kitsugi, T; Yamamuro, T; Nakamura, T; Kokubo, T. Bone bonding behavior of MgO-
CaO-SiO2-P2O5-CaF2 glass (mother glass of AW glass-ceramics). J. Biomed. Mater. Res.,
1989, 23, 631-648.
[14] Ohura, K; Nakamura, T; Yamamuro, T; Kokubo, T; Ebisawa, T; Kotoura, Y; Oka, M.
Bone-bonding ability of P2O5-free CaOSiO2 glasses. J. Biomed. Mater. Res., 1991, 25,
357-365.
[15] Reck, R; Strkel, S; Meyer, A. Bioactive glass-ceramics in middle ear surgery: An 8-year
review. In: Ducheyne P, Lemons JE editors. Bioceramics: Material characteristics versus
in vivo behavior Vol. 523. New York: The New York Academy of Science; 1988; 100-
106.
[16] Yamamuro, T. Replacement of the spine with bioactive glass-ceramic prostheses. In:
Yamamuro T, Hench LL, Wilson J. editors. Handbook of bioactive ceramics Vol. 1:
Bioactive glasses and glass-ceramics. Boca Raton; CRC Press Inc.; 1990; 343-351.
[17] Ono, K; Yamamuro, T; Nakamura T; Kokubo, T. Apatite-wollastonite containing glass
ceramic granule-fibrin mixture as a bone graft filler: Use with low granular density. J.
Biomed. Mater. Res., 1990, 24, 11-20.
[18] Gummel, J; Zippel, H; Hahnel, H. Replacement of the lumbar vertebral bodies with
machinable bioactive glass ceramic. Z. Klin. Med., 1988, 43, 1791-1793.
[19] Merwin, GE. Review of bioactive materials for otologic and maxillofacial applications.
In: Yamamuro T, Hench LL, Wilson J. editors. Handbook of bioactive ceramics Vol. 1:
Bioactive glasses and glass-ceramics. Boca Raton; CRC Press Inc.; 1990; 323-328.
[20] Yamamuro, T; Shikata, J; Okumura, H; Kitsugi, T; Kakutani, Y; Matsui, T; Kokubo, T.
Replacement of the lumber vertebrae of sheep with ceramic prostheses. J. Bone and Joint
Surg., 1990, 72-B, 889-893.
[21] Hench, LL; Stanley, HR; Clark, AE; Hall, M; Wilson, J. Dental applications of bioglass
implant. In Bonfield W, Hastings GW, Tanner KE editors. Bioceramics Vol. 4. Oxford:
Butterworth-Heinemann Ltd; 1991; 231-238.
[22] Wilson, J; Yli-Urpo A; Happonen, RP. Bioactive glasses: Clinical applications. In: Hench
LL, Wilson J. editors. An Introduction to Bioceramics. Singapore: World Scientific;
1993; 63-73
[23] Yamamuro, T. A/W glass-ceramic: Clinical applications. In: Hench LL, Wilson J. editors.
An Introduction to Bioceramics. Singapore: World Scientific; 1993; 89-103.
[24] Shors, EC; Holmes, RE. Porous hydroxyapatite. In: Hench LL, Wilson J. editors. An
[25] Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T. Preparation of bioactive Ti and its alloys
via simple chemical treatment. J. Biomed. Mater. Res., 1996, 32, 409-417.
[26] Kokubo, T; Miyaji, F; Kim HM; Nakamura, T. Spontaneous formation of bonelike
apatite layer on chemically treated titanium metals. J. Am. Ceram. Soc., 1996, 79, 1127-
1129.
[27] Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T. Apatite-forming ability of alkali-treated
Ti metal in body environment. J. Ceram. Soc. Japan, 1997, 105, 111-116.
[28] Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T. Effect of heat treatment on apatite-
forming ability of Ti metal induced by alkali treatment. J. Mater. Sci.: Mater. Med., 1997,
8, 341-347.
[29] Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T. Bonding strength of bonelike apatite
layer to Ti metal substrate. J. Biomed. Mater. Res.: Appl. Biomater., 1997, 38, 121-127.
[30] Kim, HM; Miyaji, F; Kokubo, T; Nishiguchi, S; Nakamura, T. Graded surface structure
of bioactive titanium prepared by chemical treatment. J. Biomed. Mater. Res., 1999, 45,
100-107.
[31] Yan, WQ; Nakamura, T; Kobayashi, M; Kim, HM; Miyaji, F; Kokubo, T. Bonding of
chemically treated titanium implants to bone. J. Biomed. Mater. Res., 1997, 37, 267-275.
[32] Yan, WQ; Nakamura, T; Kawanabe, K; Nishiguchi, S; Oka, M; Kokubo, T. Apatite
layer-coated titanium for use as bone bonding implants. Biomaterials, 1997, 18, 1185-
1190.
[33] Nishiguchi, S; Nakamura, T; Kobayashi, M; Kim, HM; Miyaji, F; Kokubo, T. The effect
of heat treatment on bone-bonding ability of alkali-treated titanium. Biomaterials, 1999,
20, 491-500.
[34] Nishiguchi, S; Kato, H; Fujita, H; Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T.
Enhancement of bone-bonding strengths of titanium alloy implants by alkali and heat
treatments. J. Biomed. Mater. Res.: Appl. Biomater., 1999, 48, 689-696.
[35] Miyazaki, T; Kim, HM; Miyaji, F; Kokubo, T; Kato, H; Nakamura, T. Bioactive tantalum
metal prepared by NaOH treatment. J. Biomed. Mater. Res., 2000, 50, 35-42.
[36] Miyazaki, T; Kim, HM; Kokubo, T; Miyaji, F; Kato, H; Nakamura, T. Effect of thermal
treatment on apatite-forming ability of NaOH-treated tantalum metal. J. Mater. Sci.:
Mater. Med., 2001, 12, 683-687.
[37] Kato, H; Nakamura, T; Nishiguchi, S; Matsusue, Y; Kobayashi, M; Miyazaki, T; Miyaji,
F; Kim, HM; Kokubo, T. Bonding of alkali- and heat-treated tantalum implants to bone. J.
Biomed. Mater. Res.: Appl. Biomater., 2000, 53, 28-35.
[38] Park, JB; Lakes, RS. Biomaterials 2nd Ed. New York: Plenum Press; 1992; 185-222.
[39] Park, JB. Biomaterials. New York: Plenum Press; 1979; 105.
[40] Bonfield, W; Grynpas, MD; Tully, AE; Bowman, J; Abram, J. Hydroxyapatite reinforced
polyethylene a mechanically compatible implant. Biomaterials, 1981, 2, 185-186.
[41] Bonfield, W. Design of bioactive ceramic-polymer composites. In: Hench LL, Wilson J.
editor. An Introduction to Bioceramics. Singapore: World Scientific; 1993; 299-303.
[42] Hench, LL. Stability of ceramics in the physiological environment. In Williams DF editor.
Fundamental Aspects of Biocompatibility Vol. 1. Boca Raton: CRC Press Inc., 1981; 67-
85.
[43] Hench, LL; Clark, AE. Adhesion to bone. In Williams DF editor. Biocompatibility of
Orthopedic Implants Vol. 2. Boca Raton: CRC Press Inc., 1982; 129-170.
[44] Hland, W; Vogel, W; Nawmann, K; Gummel, J. Interface reaction between machinable
bioactive glass-ceramics and bone. J. Biomed. Mater. Res., 1985, 19, 303-312.
[45] Kitsugi, T; Nakamura, T; Yamamuro, T; Kokubo, T; Shibuya, T; Takagi, M. SEM-
EPMA observation of three types of apatite-containing glass-ceramics implanted in bone:
The variance of a Ca-P-rich layer. J. Biomed. Mater. Res., 1987, 21, 1255-1271.
[46] Kokubo, T; Ohtsuki, C; Kotani, S; Kitsugi T; Yamamuro, T. Surface structure of
bioactive glass-ceramics A-W implanted into sheep and human vertebra. In Heimke G
editor. Bioceramics Vol. 2. Cologne: German Ceramic Society; 1990; 113-120.
[47] Kokubo, T. Surface chemistry of bioactive glass-ceramics. J. Non-Cryst. Solids, 1990,
120, 138-151.
[48] Ohtsuki, C; Kushitani, H; Kokubo, T; Kotani, S; Yamamuro, T. Apatite formation on the
surface of Ceravital-type glass-ceramic in the body. J. Biomed. Mater. Res., 1991, 25,
1363-1370.
[49] Neo, M; Kotani, S; Yamamuro, T; Ohtsuki, C; Kokubo T; Bando, Y. A comparasive
study of ultrastructures of the interfaces between four kinds of surface-active ceramic and
bone. J. Biomed. Mater. Res., 1992, 26, 1419-1432.
[50] Kokubo, T; Hayashi, T; Sakka, S; Kitsugi, T; Yamamuro, T; Takagi, M; Shibuya, T.
Surface structure of a load-bearable bioactive glass-ceramic A-W. In Vincenzini P editor.
Ceramics in Clinical Applications. Amsterdam, Elsevier Science Publishing; 1987; 175-
184.
[51] Kokubo, T; Ito, S; Huang, ZT; Hayashi, T; Sakka, S; Kitsugi, T; Yamamuro, T. Ca, P-
rich layer formed on high-strength bioactive glass-ceramic A-W. J. Biomed. Mater. Res.,
1990, 24, 331-343.
[52] Neo, M; Nakamura, T; Yamamuro, T; Ohtsuki, C; Kokubo, T. Apatite formation on three
kinds of bioactive material at an early stage in vivo: A comparative study by transmission
electron microscopy. J. Biomed. Mater. Res., 1993, 27, 999-1006.
[53] Loty, C; Sautier, JM; Boulekbache, H; Kokubo, T; Kim, HM; Forest, N. In vitro bone
formation on a bone-like apatite layer prepared by a biomimetic process on a bioactive
glass-ceramic. J. Biomed. Mater. Res., 2000, 49, 423-434.
[54] Kokubo, T; Kushitani, H; Sakka, S; Kitsugi, T; Yamamuro, T. Solutions able to

reproduce in vivo surface-structure changes in bioactive glass-ceramic A-W. J. Biomed.
Mater. Res., 1990, 24, 721-734.
[55] Schmidt, H. New type of non-crystalline solids between inorganic and organic materials.
J. Non-Cryst. Solids, 1985, 73, 681-91.
[56] Wilkes, GL; Orler, B; Huang, HH. Ceramers: Hybrid materials incorporating
polymeric/oligomeric species into inorganic glasses unitizing a sol-gel approach. Polym.
Prepr., 1985, 26, 300-302.
[57] Huang, HH; Orler, B; Wilkes, GL. Ceramers: Hybrid materials incorporating
polymeric/oligomeric species with inorganic glasses by a sol-gel process. 2. Effect of
acid content on the final properties. Polymer Bulletin, 1985, 14, 557-564.
[58] Mackenzie, JD; Chung, YJ; Hu, Y. Rubbery ormosils and their applications. J. Non-Cryst.
Solids, 1992, 147&148, 271-279.
[59] Mackenzie, JD. Structures and properties of ormosils. J. Sol-Gel Sci. and Tech., 1994, 2,
81-86.
[60] Li, P; Ohtsuki, C; Kokubo, T; Nakanishi, K; Soga, N; Nakamura, T; Yamamuro, T.
Apatite formation induced by silica gel in a simulated body fluid. J. Am. Ceram. Soc.,
1992, 75, 2094-2097.
[61] Li, P; Ohtsuki, C; Kokubo, T; Nakanishi, K; Soga, N; Nakamura T; Yamamuro, T. A role
of hydrated silica, titania and alumina in forming biologically active apatite on implant.
Trans. 4th World Biomater. Congress. Berlin: Society for Biomaterials; 1992; 4.
[62] Li, P; Ohtsuki, C; Kokubo, T; Nakanishi, K; Soga, N; Nakamura, T; Yamamuro, T.
Effects of ion in aqueous media on hydroxyapatite induction by silica gel and its
relevance to bioactivity of bioactive glasses and glass-ceramics. J. Appl. Biomater., 1993,
4, 221-229.
[63] Li, P; Ohtsuki, C; Kokubo, T; Nakanishi, K; Soga, N; de Groot, K. The role of hydrated
silica, titania and alumina in inducing apatite on implants. J. Biomed. Mater. Res., 1994,
28, 7-15.
[64] Li, P; Kangasniemi, I; de Groot, K; Kokubo, T. Bonelike hydroxyapatite induction by a
gel-derived titania on a titanium substrate. J. Am. Ceram. Soc., 1994, 77, 1307-1312.
[65] Uchida, M; Kim, HM; Kokubo, T; Fujibayashi, S; Nakamura, T. Structural dependence
of apatite formation on titania gels in a simulated body fluid. J. Biomed. Mater. Res.,
2003, 64A, 164-170.
[66] Miyazaki, T; Kim, HM; Kokubo, T; Kato, H; Nakamura, T. Induction and acceleration of
bonelike apatite formation on tantalum oxide gel in simulated body fluid. J. Sol-Gel. Sci.
Tech., 2001, 21, 83-88.
[67] Uchida, M; Kim, HM; Miyaji, F; Kokubo, T; Nakamura, T. Bonelike apatite formation
induced on zirconia gel in a simulated body fluid and its modified solutions. J. Am.
Ceram. Soc., 2001, 84, 2041-44.
[68] Uchida, M; Kim, HM; Kokubo, T; Tanaka, K; Nakamura, T. Structural dependence of
apatite formation on zirconia gels in a simulated body fluid. J. Ceram. Soc. Japan, 2002,
110, 710-715.
[69] Miyazaki, T; Kim, HM; Kokubo, T; Ohtsuki, C; Nakamura, T. Apatite-forming ability of
niobium oxide gels in a simulated body fluid. J. Ceram. Soc. Japan, 2001, 109, 929-933.
[70] Ohtsuki, C; Kokubo T; Yamamuro, T. Mechanism of apatite formation on CaO-SiO2-
P2O5 glasses in a simulated body fluid. J. Non-Cryst. Solids, 1992, 143, 84-92.
[71] Tsuru, K; Ohtsuki, C; Osaka, A; Iwamoto T; Mackenzie, JD. Bioactivity of sol-gel

derived organically modified silicates, part I: in vitro examination. J. Mater. Sci.: Mater.
Med., 1997, 8, 157-161.
[72] Kokubo, T. Bioactive glass ceramics; Properties and applications. Biomaterials, 1991, 12,
155-163.
[73] Chen, Q; Miyaji, F; Kokubo, T; Nakamura, T. Apatite formation on PDMS-modified
CaO-SiO2-TiO2 hybrids prepared by sol-gel method. Biomaterials, 1999, 20, 1127-1132.
[74] Chen, Q; Miyata, N; Kokubo, T; Nakamura, T. Bioactivity and mechanical properties of
PDMS-modified CaO-SiO2-TiO2 hybrids prepared by sol-gel process. J. Biomed. Mater.
Res., 2000, 51, 605-611.
[75] Chen, Q; Miyata, N; Kokubo, T; Nakamura, T. Bioactivity and mechanical properties of
PDMS-modified calcia-silica hybrids with added titania. J. Am. Ceram. Soc., 2003, 86,
806-810.
[76] Chen, Q; Kamitakahara, M; Miyata, N; Kokubo, T; Nakamura, T. Preparation of
bioactive PDMS-modified CaO-SiO2-TiO2 hybrids by the sol-gel method. J. Sol-Gel Sci.
Tech., 2000, 19, 101-105.
[77] Chen, Q; Miyata, N; Kokubo, T; Nakamura, T. Effect of heat treatment on bioactivity
and mechanical properties of PDMS-modified CaO-SiO2-TiO2 hybrids via sol-gel
process. J. Mater. Sci.: Mater. Med., 2001, 12, 515-522.
[78] Kamitakahara, M; Kawashita, M; Miyata, N; Kokubo, T; Nakamura, T. Bioactivity and
mechanical properties of polydimethylsiloxane(PDMS)-CaO-SiO2 hybrids with different
PDMS contents. J. Sol-Gel Sci. Tech., 2001, 21, 75-81.
[79] Kamitakahara, M; Kawashita, M; Miyata, N; Kokubo, T; Nakamura, T. Bioactivity and
mechanical properties of polydimethylsiloxane(PDMS)-CaO-SiO2 hybrids with different
calcium contents. J. Mater. Sci.: Mater. Med., 2002, 13, 1015-1020.
[80] Hench, LL; Wilson, J. In: Hench LL, Wilson J. editors. An Introduction to Bioceramics.
Singapore: World Scientific; 1993; 12.
[81] Kamitakahara, M; Kawashita, M; Miyata, N; Kokubo, T; Nakamura, T. Degradation of
bioactive polydimethylxiloxane-CaO-SiO2-TiO2 and poly(tetramethylene oxide)-CaO-
TiO2 hybrids in a simulated body fluid", J. Am. Ceram. Soc., 2004, 87, 235-239.
[82] Hayashi, T; Saito, H. Preparation of CaO-SiO2 glasses by the gel method. J. Mater. Sci.,
1980, 15, 1971-1977.
[83] Kamitakahara, M; Kawashita, M; Miyata, N; Kokubo, T; Nakamura, T. Apatite
formation on CaO-free polydimethylsiloxane(PDMS)-TiO2 hybrids", J. Mater. Sci.:
Mater. Med., 2003, 14, 1067-1072.
[84] Wen, J; Wilkes, GL. Organic/inorganic hybrid network materials by the sol-gel approach.
Chem. Mater., 1996, 8, 1667-1681.
[85] Kamitakahara, M; Kawashita, M; Miyata, N; Kokubo, T; Nakamura, T. Apatite-forming
ability and mechanical properties of CaO-free poly(tetramethylene oxide) (PTMO)-TiO2
hybrids treated with hot water. Biomaterials, 2003, 24, 1357-1363.
Chapter 7
EFFECT OF SAMPLE SIZE AND DISTRIBUTION

PARAMETERS IN ESTIMATION OF CONFIDENCE
LOWER BOUNDS FOR WEIBULL PERCENTILES
Burak Birgoren*
Associate Professor
Krkkale University, Faculty of Engineering,
Department of Industrial Engineering, 71451-Krkkale, Turkey
Abstract
The Weibull distribution is widely used for modelling fracture strength of composite
materials. Estimating lower percentiles of the Weibull distribution has been a major concern,
because small sizes of fracture strength experiments leads to unreliable estimates. Therefore,
several recent studies have focused on estimating confidence lower bounds on the lower
percentiles, namely A-basis and B-basis material properties. The general linear regression
methods with different probability indices and weight factors, and the maximum likelihood
method have been used for the estimation; the performance of these methods were compared
in recent studies and different methods were proposed using different criteria for comparison.
Aside from this ambiguity, the criteria and comparison studies do not allow any practical
interpretation. In particular, experimenters would like to know, in practical terms, how well a
method performs with respect to others and how its performance is affected by the sample size.
This chapter intends to clarify answers to these questions. Percent departures of the lower
bounds from the true percentiles convey such a practical meaning. Simulating these departures,
probabilistic upper bounds for maximum percent departures are determined; tables are
provided for computing the upper bounds for a wide range of sample sizes. Using the upper
bounds as a criterion for comparison, it is observed that the maximum likelihood method is
the best as compared to several variations of the general linear regression methods. Its
superiority is illustrated by comparing the sample sizes necessary for different methods to
achieve the same level of performance. For the maximum likelihood method, the upper bound
levels are shown to increase at a higher rate for sample sizes less than 30 as compared to
larger sizes. Also, the Weibull modulus is observed to have a very serious effect on the upper
bounds. Finally, a two-stage experimental setup is proposed for sample size determination;
*
E-mail address: birgoren@kku.edu.tr. Fax: ++90 318 357 2459
216 Burak Birgoren
after running a pilot experiment, the setup enables the experimenter to calculate the number of
additional experiments necessary for lowering the upper bounds to a certain level. Its
implementation is illustrated on a practical example.
Introduction
The Weibull distribution has been widely used in recent years to model the statistical
variation in the fracture strength of many materials such as advanced ceramics, metallic
matrix composites and ceramic matrix composites [1]; it is also used to describe the fracture
toughness behavior of steels in ductile-brittle transition region [2]. In particular, it is in the
ASTM standards for modeling fracture strength of ceramic materials [3]. The scatter is caused
by various factors such as anisotropy, internal structure, and working environment of
composites, which prevent designers from having a specific strength value to characterize
their mechanical behavior.
The two-parameter Weibull distribution function is given by
F ( ) = 1 exp ( / 0 )
m
(1)
where m is the shape parameter or Weibull modulus, 0 is the scale parameter or

characteristic strength of the distribution, and F is the fracture probability of the material at or
below uniaxial tensile stress . Weibull modulus, m, is related to the scatter of the data: the
higher the m the lower the dispersion of fracture stress. The scale parameter, on the other
hand, is closely related to the mean fracture stress.
The Weibull parameters m and 0 are estimated from a sample of strength measurements
1 , 2 ,..., n . The Maximum Likelihood (ML) and the General Linear Regression (GLR)
methods are commonly employed for estimating these parameters [1, 4, 5]. In practice, it is
only possible to perform a limited number of experiments which result in a small sample of
measured strength data, hence low accuracy of estimates. Therefore, statistical properties of
the estimators and confidence intervals for m and 0 , particularly for m, have been studied
extensively [1, 2, 6-18].
The methods for estimating Weibull parameters are also used in the estimation of
statistically-based design values, namely A-basis and B-basis material properties (or simply
A-basis and B-basis values), which are the 95% confidence lower bounds on the first and
tenth percentiles of a Weibull distribution, respectively. They are of great interest to the
engineer in the design of structural and mechanical components; however, recent research on
their estimation has remained limited [1, 4, 5, 19, 20].
The basis values are often regarded as constants which can be used to help characterize
the material and processing. Since they will always change from one sample (data set) to the
next, treating them as material constants is always an approximation [21]. Therefore, enough
data should be used to reduce the error in the approximation within a tolerable level.
Investigating the effect of sample size on the basis values through the use of simulated data
from the normal distribution, Ref. [21] demonstrates that as the sample size becomes larger,
the mean of the basis values increases while the standard deviation decreases. These
outcomes are normally expected according to the theory of confidence intervals.
Effect of Sample Size and Distribution Parameters 217
This chapter proposes a general approach to quantifying the effect of sample size on the
error of the approximation. The proposed approach measures the error as the percent
departures of the lower bounds from the true percentiles, and it is applied to the ML and GLR
methods. A general formula is derived in expressing the largest amount of error as a function
of both the sample size and the Weibull modulus. The formula allows studying the general
behavior of the error for different parameters and sample sizes for a given method. It is also
used to compare the performance of the methods. Based on this approach, an experimental
setup is proposed for determining the size of an experiment that lowers the error to a level
specified by the experimenter.
The parameter estimation methods, the confidence lower bounds, and results of recent
studies for comparing the methods are described in the following three sections. The
proposed approach and its implementation are discussed later.
Parameter Estimation Methods

Several methods are available for the estimation of Weibull parameters. Among them, the
most commonly used are the ML and the GLR methods [1, 5, 19].
The ML parameter estimates m and 0 should satisfy the following set of equations [22]:
ln i ( i )m n
n
n
ln i n i =1 n

+ =0
m
(2)
i =1 ( i )
m
i =1

and

1/ m
0 = ( i ) / n
n
m
(3)
i =1
where i represents strength measurements.

Although Equation 2 is non-linear, it can be solved by the Newton-Raphson algorithm
using Menons estimate as a starting point [23]: The Fortran source code for this algorithm
can be found in Ref. [24]. Also, Wasserman [25] explains the use of Microsoft Excel
spreadsheets for computing the ML estimates.
Equation 1 becomes a straight line by a double logarithmic transformation:
ln [ ln (1 F ( ) )] = m ln m ln 0 (4)
In order to apply the GLR method, the measurements are ranked from the smallest to the
largest. F-values are assigned according to the rank i of a measurement, ( i ) denoting the ith
smallest: ( 1 ) ( 2 ) ... ( n ) . The most commonly used estimators of F are
F ( ( i ) ) = ( i 0.3) / ( n + 0.4 ) (5a)

218 Burak Birgoren
and
F ( ( i ) ) = i / ( n + 1) (5b)
Considering the familiar form of a regression equation, Y = aX + b , the left side of

Equation 4 corresponds to Y, ln corresponds to X, m corresponds to a, and mln 0
corresponds to b. Using ( i ) and F ( ( i ) ) pairs in Equation 4, a and b are obtained by the
least squares procedure. Then the Weibull parameter estimates are calculated as m = a and
0 = exp ( b / m
).
The regression model in Equation 4 has non-constant error variance [6], then use of the
weighted least squares procedure gives estimates with better statistical properties: Bergman
[6] and Faucher and Tyson [7] proposed using the following weight factors in linear
regression respectively:
wi = [(1 F ( ( i ) ) ) ln (1 F ( ( i ) ) )]
2
(6a)
wi = 3.3F ( ( i ) ) 27.5 1 (1 F ( ( i ) ) )
0.025
(6b)
Also, Hung [15] proposed the following factors where F ( ( i ) ) values are estimated by
Equation 5a.
[(1 F ( ( i ) ) ) ln (1 F ( ( i ) ) )]2
wi = (6c)
[(1 F ( ( i ) ) ) ln (1 F ( ( i ) ) )]
2
Least squares and weighted least squares computations are performed by various
statistics packages such as MINITAB and SPSS. Microsoft Excel spreadsheets can also be
used for this purpose.
While there is no closed form solution to Equation 2 for the ML method, explicit
formulae exist for the GLR estimates as follows [1]:
m =
( w )( w z y ) ( w z )( w y )
i i i i i i i i
(7a)
( w )(w z ) ( w z )
i i
2
i i i
2
and

0 = exp
w y i i wi zi
m
(7b)
m wi
where
zi = ln (i ) (7c)
yi = ln [ ln (1 F ( ( i ) ) )] (7d)
When there is no weighting, the estimates are obtained by setting the weight factors to 1
in Equations 7a and 7b.
Confidence Lower Bounds

An important reason for the estimation of Weibull parameters is the need to determine the
(100p)th percentile, p , for a predefined failure probability p:
Pr [ p ] = p (8)
By setting F ( ) = p in Equation 1, p is obtained as
p = 0 [ln (1 / (1 p ) )]
1/ m
(9)
The estimate of p , p , can be obtained through the use of sample estimates of m and
0 (i.e. m and 0 ) by the following equation:
p = 0 [ln (1 / (1 p ) )]

1/ m
(10)
However, the p values can be quite unreliable, especially when they are estimated from
small samples. Therefore, instead of p values, confidence lower bounds for p have been
used to characterize a material.
Combining Equations 9 and 10, the following relation is obtained [4]:
ln ( p / p ) = mln
m / m ) ln ( ln (1 p ) )
( 0 / 0 ) + (1 m (11)
When m and 0 are estimated by the ML or the GLR method, the statistics
ln ( 0 / 0 ) and m / m are distributed independently of m and 0 , and hence pivotal
m
ln ( p / p ) is also a pivotal statistic according to Equation 11.
statistics [4, 22]. Therefore, m
This property was first indicated by Thoman et al. [26] for the ML estimates, and then used
by Bain [27] for the derivation of exact confidence lower bounds for p .
By setting Pr [ m
ln ( p / p ) c p ] = 1 , the following is obtained:
Pr [ p exp ( c p / m
) p ] =1 (12)
Therefore,
l p = p exp ( c p / m
) (13)
220 Burak Birgoren
is the lower bound for (1 ) level one-sided confidence intervals for p . The l p values for
= 0.05 and p = 0.01 and p = 0.10 are the A-basis and B-basis material properties,
respectively.
Computation of confidence lower bounds requires c p values, which can be obtained from
ln ( p / p ) . Results of a simulation study based
Monte-Carlo simulations of the statistic m
on the ML method were provided in tabular form [21]. In an effort to eliminate the use of
tables, another study proposed closed-form functions, approximating simulated data, for
computing the lower bounds [1]; the functions were provided for the ML method and the
GLR method with two weighting schemes (F-values and weight factors are assigned
according to Equations 5b, and 6a and 6b, respectively). Recently, it was shown that
simulation-run time for the ML computation of the lower bounds does not exceed 30 seconds
even for large simulations: The C++ source code for the computation can be obtained from
the authors of Ref. [20].
Comparison of Lower Bound Estimation Methods

In two recent studies [5, 19], the ML and GLR methods with various combinations of wi and
Fi were chosen for comparison as shown in Table 1. In their simulation study, Barbero et al.
[5] compared the first seven of the methods with respect to the variance of m ln ( p / p ) .
The authors found out that Method 6 has the smallest variance for sample sizes n 7, and
Method 7 has the smallest variance for n > 7 .
Table 1. Estimation methods investigated in the literature
Method Type Equation for Fi Equation for wi

1 Linear regr. 5a wi = 1
2 Linear regr. (weighted) 5a 6a
3 Linear regr. (weighted) 5a 6b
4 Maximum Likelihood
5 Linear regr. 5b wi = 1
6 Linear regr. (weighted) 5b 6a
7 Linear regr. (weighted) 5b 6b
8 Linear regr. (weighted) 5a 6c
Variance is a general measure for comparing estimators, but it may be misleading when
used for confidence lower bounds. There is an established theory for determining the best
confidence intervals. A well-known measure for comparing one-sided confidence intervals is
derived from statistical hypothesis testing: The best method is defined as the one with the
smallest false coverage probabilities Pr [l p p ] for all p < p [28], which are analogous
to Type-II errors in hypothesis testing. Based on this measure, Birgoren [19] chose three
levels of p and compared all the methods in Table 1 using Monte-Carlo simulation for
various sample sizes. He showed that Method 4 (ML) is the best method, while Method 7 is
the second best. Also, the simulation results indicate that all the GLR methods with weighted
least squares (Methods 2, 3, 6-8) performed better than those with no weights (Methods 1 and
5).
On the other hand, false coverage probabilities do not allow any practical interpretation.
In particular, an experimenter would like to know, in practical terms, how well a method
performs with respect to others and how its performance is affected by the sample size. Lets
take a closer look at the behavior of l p . A simulation of l p values, which will be explained
later, yielded the probability density functions for the two methods illustrated in Fig. 1. For
this simulation the parameters are set as m = 5, 0 = 100, p = 0.01, and the sample size is
chosen as n = 20. The density function of l p estimated by the ML method is slightly to the
right of the function estimated by the GLR method 1. Therefore, for any p < p , the ML
method has smaller false coverage probabilities. However, it is not clear whether there is a
significant difference from the viewpoint of an experimenter. Further, Method 1 is one of the
worst methods in Table 1, so the difference between the ML method and the GLR methods
with weighted least squares is expected to be smaller. One practical way for quantifying the
difference is to compute the sample size necessary for the ML method to have approximately
the same density function as the GLR method 1.
Figure 1. Effect of different estimation methods on the probability density function of l p
For this purpose, lets examine the behavior of l p with respect to the sample size. In this
case, the density functions of l p were estimated by the ML method for three sample sizes as
depicted in Fig. 2. The function for n = 16 in Fig. 2 seems to be very close to the function for
the GLR method in Fig. 1. This indicates that confidence lower bounds estimated by the ML
method with a sample size of 16 have almost the same precision of those estimated by the
GLR method 1 with a sample size of 20. The difference between the two sample sizes means
saving four test specimens by preferring the former method.
222 Burak Birgoren
Figure 2. Effect of sample size on the probability density function of l p
Fig. 2 also shows that as n grows larger, the expected value of l p moves toward p and
the variance decreases. This is a typical behavior for any m, 0 and p levels. Note that
Pr [l p p ] = 0.95 for all n.
A Practical Measure for Comparison and Sample Size Selection

The previous case study illustrated in Figs. 1 and 2 helps us understand the impact of method
and sample size selection. The rest of this chapter will discuss the development and
implementation of an approach that will serve as a practical tool for assessing this impact: the
methods are compared, the effects of sample size and Weibull parameters are quantified, and
the minimum sample size satisfying certain precision requirements is calculated.
First note the behavior in Fig. 2: larger sample sizes increase the probability of observing
l p values closer to the true percentile p . Conversely, for small samples, the distance
between p and l p can become undesirably high while Pr [l p p ] = 0.95 holds true.
Therefore, an experimenter would like to know the minimum number of measurements to
obtain a desired closeness between p and l p : the ratio of ( p l p ) to p can be used as a
measure of precision indicating the percent departure of l p from the true percentile p . The
following discussion derives probabilistic upper bounds for the maximum value of this
departure.
After the necessary manipulations, l p in Equation 13 can be reformulated as
( p / p ) c p
l p = p exp m
mln
(14)
m / m
The following term

ln ( p / p ) c p
m
Sp = (15)
m / m
in Equation 14 is a pivotal statistic, that is, its distribution does not depend on m and 0 ,
ln ( p / p ) and m / m , and the value of c p do not depend on
because the distributions of m
m and 0 [5]. Combining Equations 14 and 15 yields
l p = p exp ( S p / m ) (16)
Lets denote the 5th percentile of l p by l p (0.05) , which satisfies Pr [l p l p( 0.05 ) ] = 0.05 . It is
a lower bound for l p as illustrated in Fig. 2. In Equation 16, l p is an increasing function of
S p , then, l p (0.05) can be computed from the 5th percentile of S p , S p( 0.05 ) :
l p( 0.05 ) = p exp ( S p( 0.05 ) / m ) (17)
According to the above relationship, once S p( 0.05 ) for a sample size n is known, l p (0.05) can
be easily computed. Using the facts that Pr [l p p ] = 0.95 and Pr [l p l p( 0.05 ) ] = 0.05 , the
following is obtained:
Pr [l p( 0.05 ) l p p ] = 0.90 (18)
Finally, it follows from Equations 17 and 18 that
Pr [ 0 ( p l p ) p 1 exp ( S p( 0.05 ) / m )] = 0.90 (19)
Here, 1 exp ( S p( 0.05 ) / m ) is a probabilistic upper bound for ( p l p ) p , which is the

precision measure for l p in this study. The remaining task is the generation of the S p( 0.05 )
values for various sample sizes.
Monte-Carlo Simulation
Methods 1, 4 and 7 in Table 1 are selected for generating S p( 0.05 ) values: the motivation
behind this selection is that some experimenters prefer using a GLR method with weighted
least squares; Method 7 is the best in this category as discussed previously. If ordinary least
squares (no weights) is preferred, then the best in this category is Method 1 [19]. Finally,
Method 4 (ML) is the best among all the methods in Table 1.
A two-stage Monte-Carlo simulation study was performed for generating S p( 0.05 ) values.
In the first part, the c p values were estimated following the procedure of Birgoren [19]: for
specified values of and p, a sample of n values are generated from a Weibull distribution
with parameters m = 1 and 0 = 1; then m ln ( p / p ) is calculated for the three methods and
224 Burak Birgoren
this process is repeated N times to find the estimates of c p . Throughout this study = 0.05
and N = 200.000. This simulation was performed for each combination of p = 0.01 and 0.10,
and n = 6 ,7 ,...,70 ,72 ,...,120 .
The second part estimates S p( 0.05 ) using the simulated c p values: N Weibull samples of
size n were generated with m = 1 and 0 = 1; for each sample, the S p values were estimated
for the three methods by Equation 7. Using N simulated values of S p , an estimate of S p( 0.05 )
was obtained. This procedure was repeated for all combinations of p and n. The whole
simulation procedure is illustrated in Fig. 3.
Figure 3. Flow diagram of the Monte-Carlo simulation procedure

The procedure is coded in the C++ language. Weibull variables with the parameters m = 1
and 0 = 1 are generated by the formula i = ln (1 / u ) , where u follows a uniform distribution
on the interval [0,1]. These uniform variables are generated by LEcuyers multiple recursive
random number generator [29].
The estimated S p( 0.05 ) values are presented in Tables A1, A2 and A3 in the appendix.
Also, the l p values in Figs. 1 and 2 were generated using the S p values of the second part.
A Practical Comparison Based on Equivalent Sample Sizes

The methods with the same S p( 0.05 ) values have the same upper bounds for ( p l p ) p
according to Equation 19; this provides a basis for a more formal treatment of the discussion
of Figs. 1 and 2.
The S p( 0.05 ) values for Method 4 (ML) were taken from Table A2 for certain sample sizes,
and the sample sizes giving the closest S p( 0.05 ) values for the other two methods were
determined from Tables A1 and A3. The corresponding sample sizes are provided in Table 2.
Table 2. Sample sizes with equivalent levels of precision for compared methods
Equivalent Sample Sizes (n)

p = 0.01 p = 0.10
Method 4 Method 7 Method 1 Method 4 Method 7 Method 1
10 10 13 10 10 13
20 21 29 20 21 27
30 32 45 30 31 41
40 43 61 40 42 56
50 54 78 50 53 69
60 65 92 60 64 84
70 76 110 70 74 96
For sample sizes of 30 or less in Table 2, Method 7 achieves the level of precision of
Method 4 with one or two additional test specimens, which can be interpreted as the cost of
not using Method 4 (ML). As the sample size grows larger, however, this cost increases
reaching to an additional 4-6 specimens for n = 70. The performance of Method 1 is much
poorer requiring nearly an additional half the sample size of Method 4 (for p = 0.01). In
general, the cost is less severe in the estimation of B-basis values (p = 0.10) than that of A-
basis values (p = 0.01).
These results strongly discourage the use of the GLR method with ordinary least squares.
Use of weighted least squares significantly improves the performance making Method 7 an
alternative to the ML method for small samples (n 30). Nevertheless, the ML method should
be preferred for larger samples.
226 Burak Birgoren
Effects of Weibull Modulus and Sample Size

The previous comparison based on sample sizes suggests that an experimenter should use the
ML method for estimating the confidence lower bounds. Therefore, only this method will be
considered for a detailed study of the effects of Weibull parameters and sample size.
First, it would be very convenient to provide the S p( 0.05 ) values in Table A2 in a closed-
form expression. For each p, the S p( 0.05 ) values were fitted as a function of n by
S p( 0.05 ) = b1 b2 * exp ( ( n b3 ) )
b4
(20)
with the use of non-linear regression analysis. The empirical parameters b1 b4 for p = 0.01
and 0.10 are presented separately in Table 3. When compared to the estimated S p( 0.05 ) values
in Table A2, the maximum error of the empirical function in Equation 20 is less than 1% for
both p = 0.01 and p = 0.10.
Table 3. Parameters for the empirical function for S p( 0.05 ) in Equation 20
b1 b2 b3 b4
p = 0.01 23.57 23.8582 1.9259 0.6502
p = 0.10 12.0175 12.1435 1.8909 0.6171
Combining Equations 19 and 20, the upper bound for the precision measure
( p l p ) p can be modelled as
1 exp ( S p( 0.05 ) / m ) = 1 exp ({b1 b2 * exp ( ( n b3 ) )} m )

b4
(21)
where the parameters b1 b4 are as given in Table 3.

A detailed examination of Equation 21 helps to give a comprehensive perspective about
the effect of parameters and sample size on the precision of the confidence lower bounds. The
percent departure of the lower bounds from the true percentile varies from sample to sample.
Equation 21 forms an upper bound for this random quantity with a high probability. Note that
it does not involve the scale parameter 0 , which is nice because the measure of precision is
in the form of percent departure.
Figs. 4 and 5 displays the plots of the function 1 exp ( S p( 0.05 ) / m ) with respect to n for
p = 0.01 and p = 0.10, respectively; m is varied in a wide range from 2 to 40, which are
common Weibull modulus values for composite materials. It is clear from the plots that the
function is decreasing in n and m.
Marginal change in the function values are much more significant when n 30. Moreover,
the higher the value of m is the larger the marginal change occurs. This is illustrated in Table
4 where the percent decrease in the function value is computed for 20 additional
measurements. For example, for m = 20 and p = 0.01, an increase in the sample size from 10
to 30 decreases the value of 1 exp ( S p( 0.05 ) / m ) by about 50 percent. However, an increase
from 30 to 50 will cause only a further 26 percent decrease. In other words, the gain is halved
although the amount of experimental cost remained the same.
Figure 4. Plot of n versus 1 exp ( S p( 0.05 ) / m ) for p = 0.01 (A-basis case)
Figure 5. Plot of n versus 1 exp ( S p( 0.05 ) / m ) for p = 0.10 (B-basis case)
Figs. 4 and 5 also show that the function values for p = 0.10 are lower than the value for
p = 0.01 for all n and m. Therefore, for the same level of precision, estimation of A-basis
material property always requires a larger sample than that of B-basis material property.
When 1 exp ( S p( 0.05 ) / m ) and m were kept constant, the ratio of the sample size for p = 0.10
to the sample size for p = 0.01 was computed, provided that the sample sizes ranged between
10 and 100. This was repeated for various levels of 1 exp ( S p( 0.05 ) / m ) between 0 and 1, and
for m between 2 and 40. The ratios varied between 2.2 and 2.63. Therefore, it is reasonable to
228 Burak Birgoren
decide on the number of measurements as for B-basis and then multiply the resulting n by 2.4
to obtain an approximate A-basis sample size.
Table 4. Percent change in 1 exp ( S p( 0.05 ) / m ) with respect to n
1 exp ( S p( 0.05 ) / m ) Percent change from

m n = 10 n = 30 n = 50 n = 70 10 to 30 30 to 50 50 to 70
2 0.9738 0.7960 0.6820 0.6079 18.3 14.3 10.9
p = 0.01 20 0.3053 0.1470 0.1082 0.0894 51.9 26.3 17.4
40 0.1665 0.0764 0.0557 0.0457 54.1 27.1 17.8
2 0.8624 0.5891 0.4756 0.4107 31.7 19.3 13.6
p = 0.10 20 0.1799 0.0851 0.0625 0.0515 52.7 26.6 17.6
40 0.0944 0.0435 0.0318 0.0261 53.9 27.0 17.8
The same plots also point out achievable levels of precision: for a wide range of m values,
30 measurements will suffice to reduce the upper bound for the percent error of A-basis and
B-basis values below 80% and 60%, respectively. For 50 measurements, the upper bound
values decrease below 70% and 50%. These limit values are obtained for m = 2; larger m
results in much lower upper bounds.
Finally, it should be noted that Figs. 4 and 5 can be easily reproduced for any m and n via
a spreadsheet program.
A Two-Stage Experimental Setup for Sample Size Determination

So far, one application of Equations 19 and 21 is demonstrated; they can also be used to
determine the sample size of an experiment. In this case, the parameters should be assumed
unknown, but Equation 21 involves the unknown Weibull modulus m. As a solution, a pilot
experiment can be conducted to gain some information about m; this is the first stage of the
experiment. Using this information, a decision can be reached about the number of additional
measurements required in order to satisfy certain precision requirements. In other words, the
information from the first stage is used to determine the size of the experiment to be run in
the second stage. The two stages constitute the full experiment.
In the pilot experiment, a confidence interval for m will be helpful because of a nice
property of the function 1 exp ( S p( 0.05 ) / m ) in Equation 21: it is decreasing in n and m as
mentioned before. Let ml be the lower bound for a (1 ) level one-sided confidence
interval for m, that is, Pr [ m ml ] = 1 . Then, if m is greater than ml , the value of the
function at m will be less than the function value at m = ml . The procedure for computing ml
at = 0.05 level is briefly given along with the associated coefficients in Table A4 in the
appendix.
The lower bound value is used as follows: suppose the engineer requires that the
precision measure ( p l p ) p be less than the level u = 0.50 for the full experiment, that is,
the maximum allowable departure of l p from p is set to 50%. Then, he places the lower
bound value ml in Equation 21 and computes the smallest n that reduces the function value
below u = 0.50. The computed value for n gives the size of the full experiment.
This has a simple probabilistic interpretation: for the computed sample size,
Pr [1 exp ( S p( 0.05 ) / m ) u ] = 1 , because Pr [ m ml ] = 1 and the function
1 exp ( S p( 0.05 ) / m ) is decreasing in m. Using this fact and Equation 19 in Bonferroni
Inequality [23], a well-known inequality of probability, it can be shown that
Pr [ 0 ( p l p ) p u ] 0.90 (22)
Therefore, when the full experiment is run, the value of ( p l p ) p will be between 0
and u with at least 0.90 probability. The following section will illustrate the proposed
approach on a practical example.
The remaining important question is the size of the pilot experiment. Behavior of the
confidence intervals for m provides an answer. Barbero et al. [1] plotted the standard
deviation of the statistic m / m , which is used to construct the confidence intervals, with
respect to n. The plot shows a sharp decrease in the standard deviation for small n,
particularly for n < 10. This indicates that with each additional measurement, much larger
confidence lower bounds for m - much closer to p - are obtained for n < 10, as compared to
larger values of n. Larger bounds in Equation 21 result in smaller sample sizes for the full
experiment, and, thus, sample sizes of at least 10 are recommended for the pilot experiment.
The choice of a proper level for u will be discussed after the following example.
Practical Example
We will consider the results of the following experiment [30]: 19 identical composite
specimens were prepared from quasi-isotropic carbon-epoxy sheets with (0)3 configuration,
0.89 mm thickness, and 295 gr/m2 weight. The tension experiments were carried out on an
Instron 8516+ universal testing machine according to the ASTM D3039 standard. A
crosshead speed of 1.33 mm/min was used and room temperature conditions existed during
the tests. The fracture strength values measured are presented in Table 5.
Table 5. Fracture strength values from tension experiments
Test No. 1 2 3 4 5 6 7 8 9 10
Fracture strength
532.7 502.5 442 473 519 502.7 477 510 522 552
[MPa]
Test No. 11 12 13 14 15 16 17 18 19
Fracture strength
522 439 513.6 497.5 521.6 450.9 476.5 507.3 463.5
[MPa]
The parameter estimates and confidence lower bounds computed by Methods 1, 4 and 7
are presented in Table 6. The values of p were calculated from Equation 10, and the values
230 Burak Birgoren
of l p (A-Basis and B-basis values) were calculated from Equation 13, where the c p values
were taken from Tables A1-A3.
Table 6. Weibull estimates from the experimental data
Method 1 Method 4 Method 7

m 17.57 18.86 16.03
0 510.60 510.18 510.83
p 393.00 399.77 383.42

p = 0.01
lp 354.43 353.23 355.04
p 449.22 452.80 443.94

p = 0.10
lp 424.34 423.38 424.53
While Method 4 is better than the others in Table 6, it does not have the largest basis
values, because, as illustrated in Fig. 1, Method 4 produces larger basis values on the average,
which does not guarantee larger values from a single sample. It is also interesting to note that
the difference between the basis values are about 1-2 MPa., which are not practically
significant. Again, the difference may get much larger for another sample.
Now, we will treat these observations as our pilot experiment and determine the size of a
full experiment based on the ML method (Method 4). Using m = 18.86 from Table 6, and
using Table A4, the value of the confidence lower bound for m is computed as ml = 12.82. As
an alternative, the formulae provided by Barbero et al. [1] can be used for this computation.
We will compute the size of a full experiment for each of the basis values for p = 0.01
and p = 0.10. Suppose that we want to reduce the precision measure ( p l p ) p below
u = 0 .20. First, ml = 12.82, is placed in Equation 21 with the empirical parameters for p = 0.01
in Table 3. Then, the values of the function in Equation 21 are computed for different n in
order to find the smallest sample size that lowers the value of the function 1 exp ( S p( 0.05 ) / m )
below 0.20. This sample size can be found with a spreadsheet program, for instance, by
enumerating the function values for all n between 19 and 120.
As an alternative, a scientific calculator can be used after examining Fig. 4: for m = 10,
the function value for n = 20 is about 0.20. It is obvious from the plot that higher n values are
needed for m = 12.82. For n = 20 and 30 the function values are greater than 0.20, but for
n = 40 the value is lower. Next, n is increased successively starting from 31, and finally n = 36
is found to be the smallest sample size lowering 1 exp ( S p( 0.05 ) / m ) below 0.20. Therefore,
this computation can be performed using a calculator with a reasonable amount of trial and
error.
Lets interpret this result according to Equation 22: when additional 17 tests are carried
out in the second stage of the experiment, hence a total of n = 36 observations, the value of
( p l p ) p will be below 0.20 with at least 1 = 0.85 probability. In other words, we are
at least 85% confident that, the percent deviation of the B-basis value from the true 10th
percentile will be less than 20% after the second stage is run.
When this computation is repeated for p = 0.10, it is found that the function value for the
pilot experiment (n = 19) is 0.17, which is less than u = 0 .20. Therefore, no second stage will
be required. According to Equation 22, the value of ( p l p ) p is already below 0.17 with
at least 1 = 0.85 probability.
The choice of u = 0 .20 in this example is subjective. Further, as can be seen from Fig. 4,
if the confidence lower bound for m were ml = 2, it would be impossible to achieve this level
of precision with even 120 measurements. Experimenters need pre-specified values with
respect to which they judge a basis value, for instance, as precise and not precise. Such
values may be specified based on the observed ml values. Revisiting Figs. 4 and 5, for
instance, for ml 10, the levels of u = 0 .20 and u = 0 .10 may be recommended for precise A-
basis and B-basis computations.
Conclusions
This chapter has focused on the estimation of A-basis and B-basis material properties for the
two-parameter Weibull distribution. It has proposed percent departures of the basis values
from true percentiles as a measure of precision for the estimation methods; the ML method
and several variations of the GLR method proposed in the literature have been taken into
consideration. Using Monte-Carlo simulations, probabilistic upper bounds for the maximum
percent departures were determined. The formula and the associated tables proposed for this
purpose allow an experimenter to calculate the upper bounds for a wide range of sample sizes
and Weibull modulus values. Using the measure as a criterion for comparison, the ML
method has been shown to be the best method, that is, it has the smallest upper bounds. Its
superiority has been illustrated by comparing the sample sizes necessary for different
methods to produce the same upper bounds. The ML method produces sample sizes that are
significantly lower than those for other methods except for a certain variation of the GLR
method; it may be considered as an alternative to the ML method for small samples. The
effects of the sample size and the Weibull modulus have been investigated for the ML method.
The upper bound levels have been shown to increase at higher rates for sample sizes less than
30 as compared to larger sizes. Also, the Weibull modulus has been observed to have a very
serious effect on the upper bounds; the higher the Weibull modulus the lower upper bounds.
Finally, a two-stage experimental setup has been proposed for sample size determination;
after running a pilot experiment, the setup enables the experimenter to calculate the number
of additional experiments necessary for lowering the upper bounds to a certain level. As
illustrated on a practical example, the setup is an alternative to classic experiment plans with
pre-determined sample sizes; it may help experimenters reduce the total size of an experiment
as compared to classic experiment plans, conducting experiments only necessary for
achieving the target upper bound levels, hence, cutting down experimental costs.
232 Burak Birgoren
Appendix
Table A1. The values of cp and Sp(0.05) for Method 1 (GLR with no weights)
p = 0.01 p = 0.10 p = 0.01 p = 0.10

n cp Sp(0.05) cp Sp(0.05) n cp Sp(0.05) cp Sp(0.05)
6 4.2189 -13.8452 2.2632 -7.2668 51 1.0451 -2.9519 0.5818 -1.5644
7 3.6209 -11.9068 1.9738 -6.3011 52 1.0316 -2.9003 0.5740 -1.5374
8 3.2718 -10.5799 1.7698 -5.5730 53 1.0326 -2.8807 0.5684 -1.5177
9 2.9766 -9.6085 1.6162 -5.0705 54 1.0145 -2.8429 0.5616 -1.4994
10 2.7589 -8.8341 1.5012 -4.6529 55 1.0067 -2.8224 0.5584 -1.4871
11 2.5644 -8.2253 1.4064 -4.3484 56 0.9993 -2.7784 0.5528 -1.4695
12 2.4064 -7.6707 1.3337 -4.0745 57 0.9862 -2.7476 0.5464 -1.4518
13 2.2781 -7.2460 1.2576 -3.8385 58 0.9811 -2.7340 0.5425 -1.4436
14 2.1822 -6.8574 1.2098 -3.6469 59 0.9699 -2.6950 0.5409 -1.4297
15 2.0759 -6.5046 1.1507 -3.4459 60 0.9697 -2.6760 0.5333 -1.4113
16 2.0023 -6.2461 1.1026 -3.2986 61 0.9587 -2.6475 0.5279 -1.3938
17 1.9336 -5.9865 1.0693 -3.1685 62 0.9489 -2.6144 0.5205 -1.3749
18 1.8680 -5.7409 1.0368 -3.0451 63 0.9432 -2.5861 0.5200 -1.3638
19 1.8150 -5.5408 1.0015 -2.9268 64 0.9308 -2.5669 0.5158 -1.3546
20 1.7570 -5.3511 0.9652 -2.8227 65 0.9255 -2.5395 0.5107 -1.3412
21 1.6972 -5.1799 0.9450 -2.7469 66 0.9082 -2.5076 0.5076 -1.3294
22 1.6581 -5.0389 0.9138 -2.6582 67 0.9092 -2.5050 0.5023 -1.3211
23 1.6123 -4.8710 0.8930 -2.5764 68 0.9054 -2.4878 0.5000 -1.3118
24 1.5684 -4.7155 0.8619 -2.4845 69 0.8975 -2.4582 0.4927 -1.2946
25 1.5333 -4.5933 0.8498 -2.4322 70 0.8878 -2.4299 0.4895 -1.2817
26 1.4960 -4.4883 0.8307 -2.3729 72 0.8723 -2.3996 0.4858 -1.2701
27 1.4750 -4.3902 0.8197 -2.3240 74 0.8655 -2.3605 0.4795 -1.2483
28 1.4469 -4.2796 0.8006 -2.2633 76 0.8518 -2.3263 0.4720 -1.2277
29 1.4164 -4.1917 0.7865 -2.2181 78 0.8411 -2.2820 0.4630 -1.2006
30 1.3991 -4.0997 0.7719 -2.1632 80 0.8269 -2.2498 0.4594 -1.1906
31 1.3669 -4.0160 0.7575 -2.1245 82 0.8211 -2.2184 0.4521 -1.1699
32 1.3377 -3.9313 0.7427 -2.0769 84 0.8121 -2.1886 0.4483 -1.1541
33 1.3191 -3.8522 0.7350 -2.0427 86 0.8041 -2.1599 0.4433 -1.1380
34 1.2908 -3.7762 0.7190 -1.9988 88 0.7965 -2.1277 0.4399 -1.1239
35 1.2767 -3.7097 0.7113 -1.9651 90 0.7814 -2.0956 0.4314 -1.1064
36 1.2581 -3.6481 0.6989 -1.9284 92 0.7742 -2.0656 0.4257 -1.0899
37 1.2473 -3.6035 0.6876 -1.9003 94 0.7619 -2.0437 0.4194 -1.0756
38 1.2226 -3.5178 0.6771 -1.8618 96 0.7633 -2.0274 0.4173 -1.0652
39 1.2055 -3.4646 0.6687 -1.8351 98 0.7536 -1.9995 0.4144 -1.0540
40 1.1946 -3.4219 0.6620 -1.8084 100 0.7476 -1.9831 0.4107 -1.0442
41 1.1781 -3.3730 0.6522 -1.7827 102 0.7409 -1.9621 0.4072 -1.0334
42 1.1607 -3.3287 0.6430 -1.7592 104 0.7263 -1.9320 0.4014 -1.0213
43 1.1444 -3.2774 0.6374 -1.7344 106 0.7227 -1.9149 0.3995 -1.0108
44 1.1344 -3.2263 0.6288 -1.7050 108 0.7140 -1.8873 0.3951 -0.9992
p = 0.01 p = 0.10 p = 0.01 p = 0.10

45 1.1146 -3.1676 0.6205 -1.6817 110 0.7157 -1.8791 0.3922 -0.9900
46 1.1059 -3.1431 0.6128 -1.6620 112 0.7029 -1.8471 0.3882 -0.9747
47 1.0938 -3.0943 0.6025 -1.6295 114 0.7000 -1.8357 0.3843 -0.9668
48 1.0767 -3.0543 0.5977 -1.6157 116 0.6910 -1.8186 0.3793 -0.9583
49 1.0651 -3.0127 0.5948 -1.5973 118 0.6840 -1.7925 0.3778 -0.9473
50 1.0653 -2.9918 0.5895 -1.5777 120 0.6842 -1.7824 0.3743 -0.9372
Table A2. The values of cp and Sp(0.05) for Method 4 (ML)
p = 0.01 p = 0.10 p = 0.01 p = 0.10

6 6.7836 -12.0682 3.5389 -6.4241 51 1.1756 -2.2628 0.6544 -1.2751
7 5.6423 -10.1547 2.9759 -5.4701 52 1.1604 -2.2433 0.6480 -1.2665
8 4.9408 -8.9221 2.6024 -4.8051 53 1.1489 -2.2173 0.6386 -1.2494
9 4.3946 -8.0170 2.3182 -4.3220 54 1.1360 -2.1885 0.6313 -1.2324
10 3.9751 -7.2589 2.1097 -3.9395 55 1.1212 -2.1650 0.6258 -1.2216
11 3.6480 -6.7111 1.9487 -3.6607 56 1.1090 -2.1428 0.6180 -1.2089
12 3.3799 -6.2506 1.8208 -3.4318 57 1.0976 -2.1203 0.6099 -1.1935
13 3.1326 -5.8236 1.6944 -3.2079 58 1.0868 -2.1007 0.6043 -1.1822
14 2.9529 -5.5038 1.6032 -3.0404 59 1.0723 -2.0775 0.5983 -1.1724
15 2.7966 -5.2031 1.5119 -2.8679 60 1.0673 -2.0600 0.5923 -1.1599
16 2.6457 -4.9498 1.4312 -2.7343 61 1.0549 -2.0411 0.5868 -1.1488
17 2.5476 -4.7535 1.3763 -2.6193 62 1.0438 -2.0184 0.5816 -1.1381
18 2.4253 -4.5443 1.3168 -2.5170 63 1.0360 -2.0020 0.5746 -1.1259
19 2.3344 -4.3717 1.2673 -2.4172 64 1.0251 -1.9778 0.5696 -1.1143
20 2.2267 -4.1922 1.2183 -2.3323 65 1.0162 -1.9600 0.5637 -1.1029
21 2.1483 -4.0569 1.1761 -2.2602 66 0.9990 -1.9406 0.5585 -1.0945
22 2.0845 -3.9462 1.1327 -2.1867 67 0.9946 -1.9258 0.5548 -1.0870
23 2.0341 -3.8400 1.1057 -2.1287 68 0.9834 -1.9085 0.5500 -1.0785
24 1.9444 -3.6933 1.0631 -2.0561 69 0.9765 -1.8937 0.5446 -1.0692
25 1.8924 -3.5921 1.0408 -2.0071 70 0.9674 -1.8803 0.5429 -1.0642
26 1.8383 -3.4812 1.0140 -1.9535 72 0.9502 -1.8463 0.5302 -1.0422
27 1.7913 -3.4027 0.9832 -1.9008 74 0.9363 -1.8170 0.5219 -1.0245
28 1.7498 -3.3297 0.9587 -1.8556 76 0.9178 -1.7843 0.5141 -1.0106
29 1.7118 -3.2561 0.9429 -1.8236 78 0.9049 -1.7554 0.5062 -0.9934
30 1.6772 -3.1880 0.9199 -1.7805 80 0.8890 -1.7260 0.4979 -0.9766
31 1.6359 -3.1251 0.8992 -1.7456 82 0.8829 -1.7144 0.4877 -0.9627
32 1.5990 -3.0579 0.8783 -1.7071 84 0.8691 -1.6857 0.4851 -0.9521
33 1.5652 -2.9914 0.8690 -1.6851 86 0.8509 -1.6606 0.4803 -0.9430
34 1.5359 -2.9383 0.8487 -1.6462 88 0.8436 -1.6410 0.4736 -0.9297
35 1.4993 -2.8664 0.8330 -1.6131 90 0.8281 -1.6134 0.4654 -0.9160
36 1.4832 -2.8281 0.8167 -1.5834 92 0.8190 -1.5957 0.4586 -0.9047
37 1.4523 -2.7773 0.8005 -1.5539 94 0.8103 -1.5775 0.4553 -0.8945
38 1.4251 -2.7321 0.7878 -1.5316 96 0.8067 -1.5655 0.4487 -0.8822
234 Burak Birgoren
p = 0.01 p = 0.10 p = 0.01 p = 0.10

39 1.3956 -2.6834 0.7730 -1.5089 98 0.7948 -1.5419 0.4423 -0.8696
40 1.3816 -2.6445 0.7634 -1.4826 100 0.7838 -1.5247 0.4385 -0.8634
41 1.3544 -2.6011 0.7476 -1.4587 102 0.7748 -1.5091 0.4340 -0.8535
42 1.3365 -2.5681 0.7402 -1.4423 104 0.7653 -1.4943 0.4295 -0.8473
43 1.3123 -2.5195 0.7247 -1.4132 106 0.7526 -1.4716 0.4251 -0.8371
44 1.2948 -2.4895 0.7201 -1.4034 108 0.7485 -1.4596 0.4159 -0.8216
45 1.2727 -2.4542 0.7075 -1.3828 110 0.7402 -1.4478 0.4129 -0.8172
46 1.2571 -2.4189 0.6962 -1.3599 112 0.7343 -1.4278 0.4104 -0.8081
47 1.2387 -2.3837 0.6904 -1.3439 114 0.7248 -1.4137 0.4059 -0.8001
48 1.2261 -2.3660 0.6813 -1.3317 116 0.7170 -1.3998 0.4030 -0.7930
49 1.2108 -2.3269 0.6717 -1.3092 118 0.7087 -1.3841 0.4006 -0.7864
50 1.1948 -2.2978 0.6635 -1.2928 120 0.7049 -1.3752 0.3965 -0.7802
Table A3. The values of cp and Sp(0.05) for Method 7 (GLR with weights)
p = 0.01 p = 0.10 p = 0.01 p = 0.10

6 2.8149 -12.5985 1.5494 -6.6763 51 0.8049 -2.3743 0.4678 -1.3192
7 2.3891 -10.6259 1.3385 -5.6819 52 0.7920 -2.3459 0.4637 -1.3065
8 2.1382 -9.3048 1.1995 -4.9821 53 0.7934 -2.3221 0.4621 -1.2930
9 1.9445 -8.3729 1.0981 -4.4894 54 0.7849 -2.2904 0.4551 -1.2722
10 1.7938 -7.5679 1.0208 -4.0793 55 0.7782 -2.2720 0.4551 -1.2663
11 1.6844 -6.9989 0.9688 -3.7953 56 0.7724 -2.2474 0.4529 -1.2544
12 1.5822 -6.4786 0.9214 -3.5389 57 0.7684 -2.2225 0.4456 -1.2355
13 1.4976 -6.0636 0.8684 -3.3050 58 0.7610 -2.2033 0.4431 -1.2254
14 1.4354 -5.7152 0.8408 -3.1320 59 0.7559 -2.1765 0.4411 -1.2151
15 1.3872 -5.3990 0.8049 -2.9466 60 0.7566 -2.1637 0.4354 -1.1995
16 1.3302 -5.1357 0.7749 -2.8127 61 0.7509 -2.1392 0.4348 -1.1900
17 1.3019 -4.9163 0.7587 -2.6959 62 0.7431 -2.1180 0.4293 -1.1762
18 1.2601 -4.7087 0.7364 -2.5883 63 0.7434 -2.1033 0.4302 -1.1693
19 1.2333 -4.5322 0.7168 -2.4859 64 0.7391 -2.0750 0.4263 -1.1528
20 1.1917 -4.3498 0.6979 -2.3914 65 0.7321 -2.0599 0.4228 -1.1446
21 1.1586 -4.2008 0.6801 -2.3163 66 0.7196 -2.0344 0.4211 -1.1362
22 1.1444 -4.0928 0.6652 -2.2488 67 0.7205 -2.0189 0.4174 -1.1241
23 1.1282 -3.9732 0.6550 -2.1868 68 0.7200 -2.0077 0.4156 -1.1170
24 1.0833 -3.8175 0.6336 -2.1058 69 0.7125 -1.9887 0.4143 -1.1080
25 1.0746 -3.7353 0.6278 -2.0565 70 0.7041 -1.9728 0.4098 -1.0999
26 1.0532 -3.6180 0.6174 -2.0016 72 0.7038 -1.9461 0.4058 -1.0809
27 1.0334 -3.5270 0.6046 -1.9513 74 0.6943 -1.9139 0.4009 -1.0628
28 1.0219 -3.4509 0.5966 -1.9078 76 0.6840 -1.8840 0.3974 -1.0500
29 1.0095 -3.3831 0.5928 -1.8715 78 0.6765 -1.8488 0.3905 -1.0297
30 1.0012 -3.3067 0.5845 -1.8311 80 0.6708 -1.8268 0.3887 -1.0184
31 0.9846 -3.2496 0.5732 -1.7920 82 0.6720 -1.8151 0.3825 -1.0027
32 0.9651 -3.1691 0.5638 -1.7547 84 0.6602 -1.7793 0.3813 -0.9911
p = 0.01 p = 0.10 p = 0.01 p = 0.10

33 0.9598 -3.1206 0.5598 -1.7261 86 0.6502 -1.7515 0.3775 -0.9781
34 0.9421 -3.0512 0.5536 -1.6924 88 0.6505 -1.7349 0.3743 -0.9656
35 0.9298 -2.9883 0.5473 -1.6620 90 0.6392 -1.7133 0.3681 -0.9513
36 0.9291 -2.9389 0.5386 -1.6254 92 0.6326 -1.6883 0.3649 -0.9425
37 0.9135 -2.8907 0.5321 -1.5967 94 0.6293 -1.6665 0.3606 -0.9269
38 0.9027 -2.8455 0.5249 -1.5749 96 0.6291 -1.6544 0.3577 -0.9165
39 0.8915 -2.7971 0.5211 -1.5528 98 0.6234 -1.6312 0.3576 -0.9077
40 0.8814 -2.7468 0.5185 -1.5303 100 0.6166 -1.6150 0.3549 -0.8988
41 0.8763 -2.7139 0.5078 -1.5019 102 0.6126 -1.5985 0.3517 -0.8890
42 0.8677 -2.6717 0.5080 -1.4851 104 0.6025 -1.5758 0.3507 -0.8820
43 0.8594 -2.6356 0.4984 -1.4593 106 0.6010 -1.5649 0.3454 -0.8702
44 0.8539 -2.6014 0.4999 -1.4481 108 0.5965 -1.5443 0.3421 -0.8590
45 0.8363 -2.5532 0.4915 -1.4235 110 0.5916 -1.5319 0.3377 -0.8516
46 0.8344 -2.5254 0.4865 -1.4041 112 0.5904 -1.5190 0.3369 -0.8438
47 0.8298 -2.4951 0.4846 -1.3873 114 0.5835 -1.5001 0.3351 -0.8359
48 0.8204 -2.4635 0.4803 -1.3727 116 0.5791 -1.4889 0.3304 -0.8261
49 0.8193 -2.4351 0.4768 -1.3522 118 0.5740 -1.4716 0.3279 -0.8177
50 0.8133 -2.4066 0.4739 -1.3378 120 0.5738 -1.4624 0.3281 -0.8137
Table A4. Coefficients for computing confidence lower bounds for m*
n d n d n d n d n d n d
6 2.4300 21 1.4348 36 1.2961 51 1.2329 66 1.1973 92 1.1620
7 2.1782 22 1.4206 37 1.2903 52 1.2299 67 1.1963 94 1.1597
8 2.0272 23 1.4091 38 1.2840 53 1.2288 68 1.1948 96 1.1589
9 1.9121 24 1.3908 39 1.2778 54 1.2251 69 1.1925 98 1.1562
10 1.8198 25 1.3807 40 1.2750 55 1.2220 70 1.1916 100 1.1544
11 1.7496 26 1.3699 41 1.2690 56 1.2195 72 1.1877 102 1.1523
12 1.6917 27 1.3596 42 1.2662 57 1.2172 74 1.1849 104 1.1499
13 1.6412 28 1.3496 43 1.2612 58 1.2160 76 1.1812 106 1.1479
14 1.6020 29 1.3429 44 1.2576 59 1.2126 78 1.1783 108 1.1471
15 1.5676 30 1.3358 45 1.2532 60 1.2109 80 1.1760 110 1.1457
16 1.5394 31 1.3279 46 1.2501 61 1.2092 82 1.1746 112 1.1441
17 1.5152 32 1.3199 47 1.2454 62 1.2067 84 1.1714 114 1.1427
18 1.4908 33 1.3135 48 1.2436 63 1.2050 86 1.1679 116 1.1412
19 1.4711 34 1.3063 49 1.2411 64 1.2034 88 1.1664 118 1.1394
20 1.4506 35 1.3001 50 1.2374 65 1.2003 90 1.1631 120 1.1379
*
d , ie Pr [ m ml ] = 0.95 .
95% confidence lower bound for m is ml = m
d values were produced by using 200.000 simulated samples for each n.
See Ref. [22] for details.
236 Burak Birgoren
References
[1] Barbero, E.; Fernandez-Saez, J.; Navarro, C. Compos. Part B-Eng. 2000, 31, 375-381.
[2] Barbero, E.; Fernandez-Saez, J.; Navarro, C. J. Mater. Sci. Lett. 2001, 20, 847-849.
[3] Standard Practice for Reporting Uniaxial Strength Data and Estimating Weibull
Parameters for Advanced Ceramics, ASTM C1239-94a (American Society for Testing
and Materials, Philadelphia, PA, 1994).
[4] Fernandez-Saez, J.; Chao, J.; Duran, J.; Amo, J. M. J. Mater. Sci. Lett. 1993, 12, 1493-
1496.
[5] Barbero, E.; Fernandez-Saez., J.; Navarro, C. J. Mater. Sci. Lett. 1999, 18, 1441-1443.
[6] Bergman, B. J. Mater. Sci. Lett. 1986, 5, 611-614.
[7] Faucher, B.; Tyson, W. R. J. Mater. Sci. Lett. 1988, 7, 1199-1203.
[8] Khalili, A.; Kromp, K. J. Mater. Sci. 1991, 26, 6741-6752.
[9] Langlois, R. J. Mater. Sci. Lett. 1991, 10, 1049-1051.
[10] Gurvich, M. R.; Dibenedetto, A. T.; Pegoretti, A. J. Mater. Sci. 1997, 32, 3711-3716.
[11] Gong, J. J. Mater. Sci. Lett. 1999, 18, 1405-1407.
[12] Absi, J.; Fournier, P.; Glandus, J. C. J. Mater. Sci. 1999, 34, 1219-1227.
[13] Gong, J. J. Mater. Sci. Lett., 2000, 19, 827-829.
[14] Davies, I. J. J. Mater. Sci. Lett., 2001, 20, 997-999.
[15] Hung, W. L. Qual. Reliab. Eng. Int. 2001, 17, 467-469.
[16] Griggs, J. A.; Zhang, Y. J. Mater. Sci. Lett. 2003, 22, 1771-1773.
[17] Wu, D.; Jiang, H. J. Mater. Sci. Lett. 2003, 22, 1745-1746.
[18] Davies, I. J. J. Mater. Sci. 2004, 39, 1441-1444.
[19] Birgoren, B. J. Mater. Sci. Lett. 2003, 21, 1121-1124.
[20] Birgoren, B.; Dirikolu, M. H. Compos. Part B-Eng. 2004, 35, 263-266.
[21] U.S. Department of Defence, Military Handbook- MIL-HDBK-17-1F: Composite
Materials Handbook, Vol. 1, 2002. Chp 8: pp 10-14. (available at:
www.knovel.com/knovel2/Toc.jsp?BookID=721 )
[22] Thoman, D. R.; Bain, L. J.; Antle, C. E. Technometrics 1969, 11, 445-460.
[23] Law, A. M.; Kelton, W. D. Simulation Modeling and Analysis; McGraw-Hill: New York,
NY, 2000; pp 303.
[24] Keats, J. B.; Lawrence, F. P.; Wand, F. K. J. Qual. Technol. 1997, 29, 105-110.
[25] Wasserman, G. S. Qual. Engng. 2000, 12, 569-581.
[26] Thoman, D. R.; Bain, L. J.; Antle, C. E. Technometrics 1970, 12, 363-371.
[27] Bain, L. J. Statistical Analysis of Reliability and Life-Testing Models; Marcel-Dekker:
New York, NY, 1978; pp 227-246.
[28] Casella, G.; Berger, R. L. Statistical Inference; Wadsworth: Belmont, CA, 1990; pp 433.
[29] LEcuyer, P.; Oper. Res. 1999, 47, 159-164.
[30] Dirikolu, M. H.; Aktas, A; Birgoren, B. Turkish J. Engng. Environ. Sci. 2002, 26, 45-48.
INDEX
A attention, viii, 101, 110, 117, 200

availability, 111
accuracy, 46, 216 averaging, 47
acetabulum, 123
acetone, 136
acid, 213
B
activation, 12 bacteria, 198
activation energy, 12 behavior, viii, 13, 25, 32, 35, 36, 42, 43, 47, 49,
additives, 6, 10, 106, 120 62, 66, 71, 87, 88, 101, 207, 210, 216, 217,
adhesion, ix, 109, 126, 147, 152, 166 221, 222
adhesion strength, ix, 147 bending, 34, 44, 203, 208
adsorption, 117, 135, 144 benign, 81
aerogels, 67 binding, 112
affect, 8, 35 biocompatibility, viii, 101, 102, 104, 107, 108,
age, 102 122, 126, 127, 144
ageing, 35, 88 biodegradability, 124
agent, 77, 84, 85, 120 biodegradation, 124
Al2O3 particles, 6, 7, 8, 9 biomaterials, viii, 101, 110, 117, 123, 128, 129,
algorithm, 217 134
alloys, 104, 198, 211 biomedical applications, 102, 104, 106, 111, 115,
alternative, 43, 57, 59, 64, 65, 104, 105, 107, 115, 119, 120, 129, 134, 145
135, 225, 230, 231 biopolymer, 135
alternatives, viii, 101, 102 blocks, 136, 141
aluminium, 126 blood, 108, 109, 111, 127, 134, 200
ambiguity, x, 215 blood plasma, 111, 200
anatase, x, 197, 204, 207 blood supply, 134
animals, 106 blowing agent, 81, 83
anisometry, 70 body, ix, x, 24, 35, 49, 91, 102, 103, 104, 106,
anisotropy, 38, 89, 216 111, 113, 114, 115, 117, 118, 122, 124, 131,
annealing, 117 134, 138, 139, 140, 141, 142, 144, 197, 198,
antibiotic, 135, 144, 146 199, 200, 201, 203, 204, 207, 209, 211, 212,
appendix, 225, 228 213, 214
arithmetic, 47, 54, 55, 56, 57, 77, 78 body fluid, x, 102, 106, 111, 114, 115, 117, 122,
arthroplasty, 144 131, 197, 200, 213, 214
aseptic, 35 bonding, x, 33, 113, 123, 197, 210, 211
association, 2, 10 bonds, 33, 88, 198, 203
assumptions, 55, 60 brain, 198
atmospheric pressure, 71 breakdown, 67
atoms, 33, 90 burnout, 81
attachment, 19, 112, 145
238 Index
C concentration, 2, 6, 8, 13, 14, 21, 25, 91, 111, 112,

118, 142, 143, 144
C++, 220, 225 conception, 90, 91
calcium, 81, 104, 109, 110, 111, 112, 113, 116, concrete, 86, 152, 158, 160, 162, 164, 165, 168,
117, 118, 123, 124, 125, 126, 128, 132, 199, 169, 170, 171, 173, 174, 192, 193
200, 201, 203, 214 condensation, 118
capillary, 23 conduct, ix, 133
capsule, 104, 116 conduction, vii, 1, 2, 4, 5, 6, 7, 8, 9, 10, 12, 13, 14,
carbon, 81, 103, 107, 108, 127, 229 15, 16, 18, 20, 21, 27
carrier, 54 conductivity, 2, 7, 10, 13, 15, 21, 27, 33, 88, 89,
case study, 222 90, 91, 102
cast, 117 conductor, 2
catalyst, 34 confidence, x, 9, 215, 216, 217, 219, 220, 221,
cation, 2 226, 228, 229, 230, 231, 235
cell, viii, ix, 2, 81, 83, 101, 102, 110, 112, 130, confidence interval, 216, 220, 228, 229
147, 148, 149, 150, 151, 152, 153, 154, 155, configuration, vii, 1, 3, 6, 21, 22, 24, 27, 229
156, 157, 158, 159, 160, 166, 172, 175, 176, confusion, 57
177, 178, 180, 182, 183, 184, 185, 186, 187, Congress, 4
189, 190, 199 connectivity, 49, 69, 134
cell culture, 130 consolidation, viii, 32, 84, 85
cell surface, 149, 172, 184 contamination, 2, 8, 9
channels, 115, 138 context, 59, 63, 68, 89, 90
chemical bonds, viii, 46, 101, 134 continuity, 123
chemical interaction, 88 control, 81, 122, 129
chemical properties, 15 cooling, ix, 16, 24, 25, 71, 125, 147, 151, 152,
chemical reactions, 81, 88, 109, 118 174, 180, 181, 183, 186, 188, 189, 190
chemical reactivity, 109, 116 cooling process, ix, 147, 151, 174, 180, 181, 183,
chemical stability, 107 186, 188, 189, 190
chlorination, 106 Copyright, 4
circulation, 109 corn, 84
closure, 115 correlation, 3, 48, 56, 89
CO2, 111 correlation function, 48, 56, 89
coagulation, 23, 25 corrosion, 35, 81, 88, 104, 108, 117
coatings, 123, 127, 128, 130, 145 costs, 231
collagen, 198, 199 coverage, 220, 221
colonization, 135 creep, 32, 88
combined effect, 115 critical value, 32, 36
compatibility, 67, 68, 109, 126, 134, 142 crystallinity, 91, 110, 124, 199
competition, 4 crystallites, 46, 47, 52
compliance, 37, 38, 47, 50, 52, 72, 73, 74 crystallization, 16, 27, 105, 119, 120, 122
components, 37, 38, 47, 50, 51, 52, 72, 73, 74, 91, crystals, 50, 77, 112, 122, 128, 129
110, 111, 116, 117, 148, 151, 200, 216 culture, 127
composites, vii, ix, 31, 33, 46, 47, 48, 50, 51, 56, cycles, 105
57, 58, 79, 80, 82, 84, 89, 90, 91, 123, 131, cytotoxicity, 112, 130
133, 134, 135, 216 Czech Republic, 31, 92, 98
composition, viii, 16, 18, 22, 23, 32, 33, 36, 48,
79, 80, 87, 89, 92, 101, 109, 110, 114, 116,
D
117, 120, 121, 122, 198, 199
compounds, 124, 125 damage, 102
compressibility, 45 danger, 35
computation, 220, 230, 231 data set, 216
computer simulations, 48, 66 decomposition, 106
computing, x, 215, 217, 220, 228, 235 defects, ix, 32, 88, 91, 111, 122, 125, 197, 198
Index 239
definition, 61 electron microscopy, 128, 129, 139, 140, 142,

deformability, x, 33, 197, 198, 200, 201, 207, 208 212
deformation, 25, 36, 43, 44, 88, 102, 140, 207 emission, 105
degenerate, 56, 57, 59, 81, 82 energy density, ix, 147, 152, 156, 174, 175, 189
degradation, 88, 108, 208 entropy, 43, 114
degree of crystallinity, 91, 120 environment, 44, 105, 134, 203, 204, 207, 209,
density, 5, 13, 15, 23, 24, 25, 33, 34, 44, 65, 74, 211, 212, 216
77, 79, 80, 81, 90, 102, 104, 106, 135, 139, epitaxial growth, 13
145, 147, 153, 158, 175, 178, 210, 221 equating, 65
deposition, 108, 110, 128, 130 equilibrium, 44, 45, 88, 148, 160, 166, 184
designers, 216 erosion, 35, 108
developed countries, 102 estimating, vii, x, 1, 92, 215, 216, 226
deviation, 56, 229, 230 ethanol, 27
diamonds, 84 EU, 191
dielectric constant, 89 Euro, 28, 29, 30, 127, 128
differential approach, 57, 64, 87 evidence, 8, 9, 13, 57, 80, 85, 107, 114
differentiation, 134 expectation, 137
diffraction, 17, 71 expression, 55, 81
diffusion, 6, 88, 115, 118, 133, 148 extraction, 105, 111
diffusion process, 118, 148
diffusivity, 33 F
direct measure, 46, 137
discs, 117, 122 fabrication, ix, 108, 133, 135
dislocation, 88, 148, 156, 184, 185 failure, 35, 105, 123, 134, 203, 207, 208, 219
disorder, 47 family, 109, 115, 122, 123, 125
dispersion, 6, 10, 135, 216 fatigue, 35, 88, 105, 108
displacement, 37 femur, 105, 116
distilled water, 136 fibers, 32, 90, 123
distribution, vii, x, 1, 9, 14, 18, 19, 20, 21, 22, 25, fibrin, 210
28, 60, 69, 107, 113, 134, 136, 137, 138, 141, fibroblast growth factor, 145
144, 145, 146, 149, 215, 216, 223, 225, 231 fibroblasts, 134
distribution function, 216 fibrous cap, 104
doping, 27, 106, 108, 123 fibrous tissue, viii, ix, 101, 104, 116, 197, 198
drug delivery, ix, 133, 144, 145 fillers, 198
drug release, ix, 133, 135, 145 films, 9, 127
drying, 81 financial support, 125
ductility, 123 fixation, 103, 104, 110
durability, 207, 209 flame, 105
duration, 13, 142 flexibility, x, 109, 197
fluid, 92, 112
E foams, 57, 81, 83
focusing, 36
elaboration, 47, 92 food, 135, 136
elastic deformation, ix, 147 fracture stress, 216
elasticity, 32, 36, 37, 38, 39, 41, 43, 47, 49, 56, France, 107
57, 68, 74, 77, 78, 79, 86, 87, 88, 90, 124, 195, free energy, 91
207 friction, 104, 106
electric field, 18 fuel, 2, 34
electrical conductivity, 28, 89 function values, 226, 227, 230
electrical properties, 18, 20, 21, 25
electricity, 102
G
electrodes, 18, 20, 21, 22, 23, 26
electron diffraction, 15 gel, 200, 204, 213, 214
240 Index
gene, 80 inclusion, viii, 32, 57, 59, 60, 84, 85, 89

generalization, 61, 68, 80, 90 indices, x, 215
generation, 120, 223 induction, 213
Germany, 92, 93, 94, 95, 96, 97, 98, 99, 131 inelastic, 88
glasses, 32, 33, 43, 46, 77, 89, 91, 101, 103, 109, inequality, 42, 229
111, 116, 117, 118, 122, 123, 129, 131, 198, inertia, 44, 102, 107
210, 211, 213, 214 infinite, ix, 49, 61, 147, 148, 150, 173, 174, 178,
grain boundaries, 21, 90, 91, 124 189, 190
grains, 4, 15, 17, 46, 88, 90, 106 influence, 14, 32, 33, 69, 70, 91, 128, 131, 134,
granules, ix, 133, 136, 138, 139, 140, 141, 144, 144, 157
145, 198 input, 45, 48, 56, 77, 85, 89
graphite, 81, 107, 108, 117 instability, 107
groups, 40, 88, 110, 117, 118, 204 Instron, 229
growth, 13, 15, 23, 35, 88, 103, 104, 118, 119, insulation, 81
120, 122, 127, 128, 135 integration, 162, 164, 168, 169, 171, 172, 173,
growth factor, 135 185, 191, 199
integrity, 36, 63, 67, 71, 83
H intensity, 107
interaction, 6, 112, 113
hardness, 32, 33, 35, 88, 102, 107 interactions, 60, 62, 127
heart valves, 108 interest, viii, 65, 83, 92, 101, 102, 104, 110, 119,
heat, vii, 1, 3, 10, 11, 12, 13, 16, 21, 27, 44, 102, 141, 216
104, 106, 119, 120, 122, 123, 138, 140, 198, interface, 16, 18, 23, 26, 35, 91, 103, 109, 111,
211, 212, 214 113, 115, 118, 123, 128, 129, 131, 132, 210
heating, 13, 14, 81, 84, 119, 120 interface energy, 16, 18, 23
heating rate, 13 interpretation, x, 42, 66, 68, 69, 70, 215, 221, 229
height, 44, 185 interrelations, 43
hemisphere, 18 interval, 66, 154, 155, 160, 166, 172, 173, 176,
hemocompatibility, 127 177, 183, 185, 186, 225, 228
hip, 102, 104, 105, 110, 123, 127 intrinsic value, 70
hip joint, 127 intuition, 42
host, viii, 2, 77, 81, 101, 102, 103, 104, 133, 134 inversion, 37, 50, 72
hybrid, x, 197, 200, 201, 204, 207, 209, 214 ionic conduction, vii, 1, 4, 5, 6
hydrolysis, 110 ions, 33, 77, 110, 111, 118, 125, 129, 200, 201,
hydroxyapatite, ix, 103, 109, 111, 114, 117, 118, 203
122, 130, 131, 133, 134, 136, 142, 145, 146, Italy, 133
197, 198, 199, 210, 211, 213
hydroxyl, 118 J
hydroxylapatite, 128
hypothesis, 49, 70, 118, 220 Japan, 126, 129, 197, 211, 213
hypothesis test, 220
K
I
kinetics, 14, 88
ideas, viii, 32, 36 knowledge, 43, 119
image analysis, 48 Korea, 1
immersion, 111, 112, 115, 116, 123
implementation, x, 216, 217, 222 L
impregnation, ix, 133, 135, 142, 144, 145
impurities, 6, 14, 104 laminar, 113
in vitro, 111, 115, 116, 126, 129, 130, 144, 200, language, 225
214 lanthanum, 26
inadmissible, 67 laser ablation, 110, 128
Index 241
lattice parameters, 109, 110, 125 microstructure, viii, 6, 18, 25, 32, 36, 46, 47, 48,
laws, 152 60, 66, 77, 81, 83, 84, 85, 89, 90, 91, 106, 107,
lead, 62, 85, 134 108, 111, 114, 119, 120, 121, 135, 140, 142
Least squares, 218 migration, 26, 118
legal, 4 mixing, 6, 91
liability, 4 mode, 34, 44, 123
lifetime, 35, 102, 127 modeling, viii, 32, 36, 49, 88, 89, 92, 216
likelihood, x, 215 models, 68, 69, 70, 89, 90, 91
limitation, 123 modules, 107
linear dependence, 60 modulus, x, 32, 33, 34, 42, 43, 45, 46, 48, 52, 53,
liquid phase, vii, 1, 5, 7, 8, 17, 18, 19, 21, 23, 25, 54, 55, 56, 58, 59, 60, 61, 62, 64, 65, 66, 67,
28, 104 68, 69, 70, 75, 76, 77, 78, 81, 82, 83, 84, 85,
liquids, ix, 23, 133 86, 87, 90, 92, 105, 110, 123, 124, 134, 181,
location, 22 191, 198, 200, 207, 208, 215, 216, 217, 226,
228, 231
M molar ratios, 204
molar volume, 33, 45
macrophages, 127 mole, 48, 75
magnesium, 106, 125, 126 molecular weight, 126
management, 133 molecules, 203
manufacturing, 102, 108, 117 monolayer, 8
mass, 9, 48, 91 morbidity, 134
material surface, 112 morphology, 46, 110, 128, 134
materials science, viii, 42, 101 Moscow, 133, 135, 144, 145
matrix, viii, ix, 16, 32, 37, 38, 39, 40, 41, 47, 49, motion, 88, 148, 156, 185, 198
50, 57, 59, 60, 61, 62, 63, 64, 65, 69, 70, 71, motivation, 223
72, 73, 74, 81, 84, 85, 86, 87, 89, 90, 108, 117, movement, 109
119, 123, 135, 147, 148, 149, 150, 151, 152, multiphase materials, vii, 31, 36, 46, 48, 55, 56
153, 154, 155, 156, 157, 158, 159, 160, 162, multiplication, 160, 166
164, 165, 166, 167, 168, 169, 170, 171, 172, multiplier, 166
173, 174, 175, 176, 177, 178, 179, 180, 181,
182, 183, 184, 185, 186, 187, 189, 190, 191, N
193, 199, 207, 216
measurement, 19, 20, 21, 44, 74, 83, 91, 102, 136, nanocomposites, 46, 80, 90, 91
217, 229 nanomaterials, 46, 92
measures, 36, 91, 118, 217 Nd, 128
mechanical properties, x, 35, 81, 88, 91, 104, 106, needs, 4, 89
108, 110, 114, 119, 120, 121, 122, 123, 131, Netherlands, 94, 95, 96, 97, 99, 100, 125, 126
132, 134, 135, 148, 197, 201, 203, 207, 208, network, 109, 110, 112, 118, 125, 200, 203, 214
209, 214 niobium, 200, 213
media, 135, 213 nitrogen, 135
median, 84, 135 nitrogen gas, 135
melt, 16, 105, 106, 117 normal distribution, 137, 216
melting, 16, 102, 117, 151 nucleation, 13, 16, 88, 118, 119, 120, 122, 200,
melting temperature, 16, 102, 151 201, 204
melts, 117 nuclei, 119, 201
membranes, 34 nucleus, 13
men, 102 nutrition, 134
mercury, 138, 141
metals, 35, 36, 88, 89, 109, 198, 211 O
micrometer, ix, 46, 81, 133
microscope, 21, 136 observations, 21, 112, 138, 230
oil, ix, 133, 135, 136, 137
242 Index
optical properties, 35, 111 polycondensation, 203

organism, 116 polydimethylsiloxane, 200, 204, 214
orientation, 46, 48, 52, 77, 90 polymer composites, 135, 212
osteomyelitis, 144 polymer-based composites, 35
osteotomy, 123 polymers, 81, 88, 89, 109, 134, 135
oxidation, 123 polymorphism, 125
oxides, 2, 71, 117, 120, 121, 200 polyurethane, 81
oxygen, 2, 34, 88, 109 poor, 117, 198
oxygen sensors, 34 population, ix, 133, 138
porosity, viii, 31, 35, 36, 49, 56, 57, 58, 60, 62,
P 63, 65, 66, 67, 68, 69, 70, 73, 78, 80, 83, 84,
85, 86, 87, 90, 101, 103, 114, 119, 138, 140,
pain, 102, 134 142, 143
parameter, viii, 48, 64, 66, 68, 69, 70, 85, 90, 101, potassium, 117
110, 160, 162, 164, 165, 166, 168, 169, 170, power, viii, 31, 36, 54, 68
172, 179, 180, 185, 216, 217, 218, 226, 229, precipitation, 110, 111, 112, 115, 118, 135
231 prediction, 47, 54, 58, 64, 65, 66, 67, 77, 78, 80,
parameter estimates, 217, 218, 229 81, 84, 85, 87, 90, 92
parameter estimation, 217 preparation, 4, 120, 123, 138, 203
parents, 191 pressure, 44, 45, 71, 103, 108, 136, 139, 140, 141,
parotid, 111 143
particles, ix, 6, 7, 47, 77, 81, 105, 106, 108, 120, principle, 33, 47, 48, 57, 67, 68, 69, 78, 83, 88,
123, 124, 127, 135, 139, 147, 148, 149, 150, 89, 91, 110, 117
180, 183, 186, 204 probability, x, 215, 216, 219, 221, 222, 226, 229,
percentile, 219, 222, 223, 226, 230 230, 231
percolation, viii, 31, 63, 67, 68, 69, 89, 90 probability density function, 221, 222
percolation theory, 63, 89 probe, 89
periodicity, 47 production, 107, 119, 134, 136
periodontal, 111, 125 program, 228, 230
permeability, 89 proliferation, 134
permit, 44, 90 propagation, 107
perspective, 226 proposition, 55
pH, 111, 112, 118, 124 prostheses, 123
phagocytosis, 124 prosthesis, 104, 105, 107, 110
phase boundaries, 46, 54, 88, 90, 91, 92 proteins, 112
phase transformation, 88 protons, 111
phase transitions, 92 purification, 104
phosphates, 104, 109, 117, 124, 128
phosphorus, 116 Q
photographs, 201, 205
physical and mechanical properties, viii, 101, 103, quartz, 102
110
physicochemical properties, viii, 101 R
physics, 34, 45, 77
plasma, 106, 110, 123, 200 radius, 33, 77, 149, 154, 155, 161, 162, 164, 165,
plastic deformation, 88, 151 166, 168, 169, 171, 172, 173, 175, 176, 177,
plasticity, 32 178, 179, 181, 184, 185, 186, 190
platelets, 90 range, ix, x, 35, 43, 46, 57, 62, 65, 73, 74, 88, 101,
plausibility, 65, 72 111, 136, 138, 139, 142, 143, 147, 152, 174,
Poisson ratio, viii, 31, 33, 34, 42, 43, 56, 57, 58, 181, 185, 215, 226, 228, 231
59, 60, 61, 62, 63, 64, 65, 69, 70, 73, 74, 75, rare earth elements, 71
79, 86, 87 raw materials, 111, 117
polarization, 3 reaction mechanism, 115
Index 243
reaction zone, 111 series, 4, 19, 62, 66, 68, 70, 90, 109, 116, 122,
reagents, 117, 135, 207 191, 192, 193
reality, 80, 91 serum, 112
recall, 69, 70 services, 4
reconstruction, 109, 123 shape, 3, 4, 7, 21, 48, 49, 62, 65, 69, 70, 149, 160,
redistribution, 25 162, 163, 166, 168, 169, 172, 173, 179, 186,
regression, x, 215, 218, 226 207, 216
regression analysis, 226 shaping, 81
regression equation, 218 shear, 34, 42, 43, 44, 45, 48, 52, 53, 55, 56, 58,
reinforcement, 135 59, 61, 62, 64, 70, 71, 77, 78, 81, 82, 83, 89,
relationship, 223 90, 113
relationships, 33, 89, 90, 125 sheep, 211, 212
relaxation, 3, 105, 151 shock, 33, 89
relevance, 213 Si3N4, v, ix, 148, 173, 180, 181, 182, 183, 184,
reliability, 34, 35, 107, 134 185, 186, 190
repair, x, 113, 122, 133, 134, 197, 198 SIC, 147
replacement, 35, 102, 116, 121, 123, 126, 130, sign, 61, 162, 163, 168, 169, 171
132, 145 silanol groups, 118, 122, 201
replication, 81 silica, 112, 117, 118, 119, 122, 123, 213, 214
reputation, 36 silicon, 108, 111, 113, 200
resistance, 2, 3, 5, 8, 12, 20, 22, 24, 33, 102, 104, similarity, 110
105, 110, 120, 122, 123, 124, 127, 157, 180 simulation, 20, 220, 221, 223, 224
resolution, 122, 128, 129 Singapore, 95, 129, 210, 211, 212, 214
responsibility, 4 single crystals, 71, 88
retrieval, 4 SiO2, x, 2, 6, 7, 12, 13, 14, 25, 26, 104, 111, 116,
rheology, 61, 62, 69 117, 118, 121, 122, 131, 197, 198, 200, 201,
rights, 4 203, 204, 210, 213, 214
risk, 134 sites, 122
room temperature, 2, 45, 46, 71, 74, 77, 88, 106, skeleton, 59, 60, 135
119, 120, 125, 136, 229 skin, 134
roughness, 104 sodium, 117, 198
rubber, 207 sol-gel, ix, 13, 80, 116, 131, 197, 200, 203, 204,
Russia, 133, 135, 145 207, 209, 213, 214
rutile, 204, 205 solid matrix, 56, 58
solid oxide fuel cells, vii, 1
S solid solutions, 77, 91
solid state, 34, 45, 77, 110, 125
saliva, 111, 117, 129 solubility, 9, 110, 117, 121, 124
sample, x, 13, 19, 70, 115, 215, 216, 217, 219, solvents, 135, 207
220, 221, 222, 223, 224, 225, 226, 227, 228, South Korea, 1
229, 230, 231 Spain, 101
scanning electron microscopy, 3, 136, 201 species, 109, 213
scatter, 78, 84, 88, 216 specific surface, 32, 124
scattering, 34 spectroscopy, vii, 1, 9, 18, 21, 26, 28
schema, 10 spectrum, 3
security, 60, 107 speed, 110, 119, 136, 229
seeding, 13 spin, 102
segregation, 5, 8, 9, 14, 27, 207 spine, 210
self, viii, 31, 48, 57, 62, 63 spongy tissue, 102
semicircle, 10, 15, 21 spreadsheets, 217, 218
sensitivity, 105, 144 stability, 26, 110, 120, 125
sensors, vii, 1 stabilization, 106
separation, 22, 23, 25, 122 stages, 116, 118, 119, 120, 228
244 Index
standard deviation, 216, 229 textbooks, 40, 42

standards, 216 theory, viii, 32, 36, 43, 44, 46, 47, 48, 57, 87, 88,
starch, viii, 32, 81, 84, 85 92, 216, 220
statistics, 88, 218, 219 thermal decomposition, 106
steel, 36, 117, 123 thermal stability, 13, 110
stoichiometry, 110 thermal treatment, 212
strain, 32, 34, 36, 37, 47, 50, 52, 71, 201, 203, thermodynamics, 42, 45, 76, 88, 92
207, 208 threshold, viii, 31, 63, 67, 68, 69, 90, 107
strength, ix, x, 32, 33, 34, 35, 77, 88, 91, 102, 103, thromboresistance, 107, 108
104, 105, 106, 109, 110, 113, 121, 124, 133, time, viii, 3, 6, 12, 15, 24, 32, 34, 35, 42, 70, 81,
134, 136, 141, 142, 143, 144, 180, 197, 198, 101, 103, 105, 109, 113, 116, 118, 119, 120,
203, 207, 208, 211, 212, 215, 216, 217, 229 124, 134, 135, 137, 142, 145, 220
stress, ix, 32, 35, 36, 37, 38, 44, 47, 50, 52, 88, tissue, viii, 35, 81, 101, 102, 103, 104, 105, 108,
105, 107, 108, 113, 123, 124, 136, 142, 143, 109, 110, 113, 116, 117, 124, 133, 134, 135,
144, 147, 148, 151, 152, 156, 157, 158, 159, 144, 145
166, 173, 174, 175, 178, 179, 180, 181, 184, titanium, 35, 109, 123, 127, 132, 198, 200, 211,
185, 186, 187, 189, 190, 198, 201, 207, 216 213
stress intensity factor, 32, 107 topology, 60, 66
stress-strain curves, 207 tradition, 89
strong interaction, 46 transformation, 35, 36, 37, 71, 106, 114, 115, 125,
structural characteristics, 122 173, 217
substitutes, 198 Transformation curve, 16
substitution, 38, 110, 113, 123, 124 transformations, 152, 174, 175
substrates, 110 transition, 62, 106, 119, 125, 128, 216
superconductivity, 148 transition temperature, 106, 119
superiority, x, 215, 231 transmission, 122, 132, 212
surface area, 160, 166 transport, 21, 26
surface energy, 32, 160, 166 trend, 84
surface region, 25, 26 trial, 230
surface structure, 211 Turkey, 215
surface tension, 32, 136, 137
Switzerland, 99, 131 U
symmetry, 37, 38, 39, 47, 50, 52, 149
synthesis, 71, 123, 125, 135 UK, 92, 93, 94, 95, 96, 97, 98, 99
systems, x, 13, 36, 88, 102, 111, 144, 186, 189, uniform, 19, 21, 25, 46, 47, 52, 139, 141, 225
190, 197
V
T
vacuum, 135, 136, 142, 143
talc, 102 validity, vii, 1, 20, 39, 47, 49, 63
tantalum, 198, 200, 211, 212, 213 values, viii, ix, 6, 7, 12, 18, 21, 23, 26, 31, 33, 36,
targets, vii, 31, 35 42, 45, 47, 54, 56, 57, 62, 64, 65, 67, 70, 71,
technology, 91 73, 75, 76, 77, 78, 79, 81, 84, 87, 88, 107, 134,
teeth, 110, 120, 132 143, 148, 180, 216, 217, 218, 219, 220, 221,
temperature, vii, ix, 1, 2, 7, 8, 11, 12, 13, 15, 16, 222, 223, 224, 225, 226, 228, 229, 230, 231,
21, 22, 23, 24, 25, 26, 34, 35, 43, 44, 45, 71, 232, 233, 234, 235
74, 76, 84, 88, 89, 106, 108, 111, 114, 117, vapor, 18, 35, 88
119, 120, 125, 132, 136, 137, 138, 147, 151, variable, viii, 101, 102, 109, 161, 162, 164, 165,
152, 174, 180, 181, 183, 188, 189, 190, 204 167, 168, 169, 170, 172, 179, 180
temperature dependence, 34, 43, 76, 190 variables, 23, 225
tensile strength, 120, 121, 142, 143, 185, 207 variance, 212, 218, 220, 222
tension, 107, 117, 152, 180, 181, 229 variation, 10, 12, 15, 19, 20, 21, 23, 24, 25, 26,
TEOS, 27, 200, 203 69, 216, 231
Index 245
vector, 37, 50 X
vehicles, 123
velocity, 34, 44 X-ray diffraction, 71, 112, 204
versatility, 117 XRD, 34, 204
vertebrae, 102, 211
vessels, ix, 133 Y
Vickers hardness, 122
viruses, 198 yield, ix, 54, 57, 65, 147, 152, 180, 181
viscosity, 32, 49, 62, 89, 135, 137 yttrium, 106
volumetric changes, 120
Z
W
zirconia, vii, 1, 2, 3, 4, 6, 7, 8, 13, 14, 19, 21, 27,
water, x, 35, 49, 88, 109, 117, 197, 203, 204, 207, 31, 33, 34, 35, 36, 45, 46, 47, 49, 56, 58, 59,
214 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82,
water vapor, 35, 88 83, 84, 85, 86, 87, 88, 89, 90, 103, 105, 106,
wear, 32, 35, 88, 104, 105, 107, 126, 127, 180 107, 108, 123, 127, 210, 213
weight ratio, 207 zirconium, 106, 200
wetting, 7, 18, 23, 24, 25
women, 102
words, 57, 62, 69, 90, 227, 228, 230
work, 48, 66, 88, 90, 92, 191

Ceramic and Composite PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Ceramic and Composite PDF

Transféré par

Droits d'auteur :

Formats disponibles

CERAMICS AND COMPOSITE

MATERIALS: NEW RESEARCH

Nova Science Publishers, Inc.

NOTICE TO THE READER

Library of Congress Cataloging-in-Publication Data

Published by Nova Science Publishers, Inc. New York

bioactivity as well as deformability. Polydimethylsiloxane (PDMS)-CaO-SiO2-based hybrids

NEW APPROACHES FOR ESTIMATING AND

Another approach is to minimize the grain-boundary resistance. The grain boundary is

2 The Model of Grain-Boundary Resistivity in Stabilized

(a) parallel equivalent circuit: gi gbsp

(b) serial equivalent circuit: gi gbsp

As a measure for the improvement in the grain-boundary conduction, the scavenging

3 Improvement of Grain-Boundary Conduction by the Addition

BM: Ball Milling UD: Ultrasonic Dispersion

1450 1500 1550 1600

4 Improvement in Grain-Boundary Conduction Via Two-Stage

1600 (a) Al2O3

HT temp. before sintering at 1500oC (oC)

sintered at 1500oC for 4h.

heat-treated at 1200oC for 40h prior to sintering at 1500oC for 4h

4.2 The Improvement in the Grain-Boundary Conduction by Post-sintering

The precursor scavenging with pre-sintering HT is a convenient and effective approach to

The dewetting phenomenon can be explained in relation to the crystallization of an

Amorphous SL < SS Crystalline

Sintering at 1550oC for 4h + Post-

Average grain size 15.7 m 15.6 m

1450 1500 1550 1600

5 Estimation of the Intergranular-Liquid-Phase Distribution

5.1 Local Impedance Measurement by Successive Thinning of a Specimen

In order to estimate the spatial distribution of the intergranular liquid in 1-mol%-Al2O3(size =

0.7t 0.8t 0.9t 1.0t

Fig.15 The complex impedance spectra of 1-mol%-Al2O3(size=0.3m)-doped 15CSZ specimens

5.2 Millicontact Impedance Spectroscopy: The Fundamental Concept

5.3 Millicontact Impedance Spectroscopy: The Distribution of

tendency for increasing gbapp disappeared at 0.5-mol%-Al2O3 (Fig.18(c)) and completely

Figure 19 shows the gbapp profiles of 1-mol%-Al2O3-doped 15 CSZ specimens as a

(a) 100 =5.33,d =15.

However, when 1-mol%-Al2O3-doped 15 CSZ specimen was sintered at 1450oC for 0 h,

5.4 Millicontact Impedance Spectroscopy: The Investigation of Liquid

Millicontact impedance spectroscopy can be also applied to the characterization of the

(a) CaSi2O5 drop

29. Guo, X., J. Am. Ceram. Soc. 2003, 86, 1867-1873.

EFFECTIVE ELASTIC MODULI OF ALUMINA,

Willi Pabst and Eva Gregorov

2 Elastic Moduli from the Viewpoint of Rational Mechanics

2.1 Elastic Properties of Anisotropic Materials

Tij = C ijkl E kl . (3)

Eij = S ijkl Tkl . (5)

or, alternatively, in matrix index notation as

respectively. In this notation ( ) and ( ) are 6-dimensional column vectors referring to

1 T11 , 2 T22 , 3 T33 , 4 T23 , 5 T31 , 6 T12 (10 a-f)

1 E11 , 2 E 22 , 3 E33 , 4 2 E 23 , 5 2 E31 , 6 2 E12 (11 a-f)

The stiffness matrix of triclinic monocrystals can be written as:

C11 C12 C13 C14 C15 C16

For materials of higher symmetry (monocrystals or polycrystalline bodies) the number of

Monoclinic monocrystals (13 independent elastic constants):

C11 C12 C13 0 C15 0

Orthorhomic monocrystals and orthotropic polycrystalline bodies (9 independent

C11 C12 C13 0 0 0

Trigonal monocrystals (6 independent elastic constants):

C11 C12 C13 C14 0 0

with the additional conditions